. b.2231... .quvn tl./A n. . f... . 1‘30uyeu 1r! .3 .13.»... T333211... , .1. (53.....th 1.5:... if}??? 7!. 3.0x. .3. 33))... 5:. v. 17kt. l.9!.. 2"! r.‘£‘a~.. I'llki. ‘1‘1..fvll't.v.tl.‘ 3.3.1.!i3 2...: . {Izin 7 ‘Alfl.‘!|:. \tl .. t.14.!9tv$!av !I.D:.6Fn.... 1.1%.. n .33.“? A q¢;.:aa‘3.tl£. uh... 13......) I .2. flint? v..\: s!" . . . . . ‘ y I? l . V ‘I 57 IE lli‘Pbtv". 5"! ‘ t“ . :yxul... EASY-a... I .. ‘ . ‘ 1 11.1... .. I. :7 7.99.! .11.)!»h125? 1v$l§3$2 \3? ti I. a: In 15.1.. 1.!~§t licr‘ n 1112!!!! I - .l !..(cfirr...anwrrVLH . .7: . ., . ifs.- . 1!. .005“- nlalunvlly'lulivb'. ‘dy'l .Ollvba‘l» t..|.v1l.| 3.7!, -.A:IOXIIII!!.E.I . , )f1.!. svch‘lA SI\ 1, \f-i. ‘ t ‘5‘”:- a:xS t ‘vU II MICHIGAN STATE V RSI LIBRARIES infirm,mm”viiimunm 1 9300st 0113 F LIBRARY Hickman State University A J fl This is to certify that the dissertation entitled FOREIGN LANGUAGE VOCABULARY ATTRITION: A STUDY OF COLLEGE-LEVEL SPANISH STUDENTS OVER AN 18-MONTH PERIOD presented by Marilyn Ruth Bierling has been accepted towards fulfillment of the requirements for Ph. D. Spanish degree in Zea/7002 3. Paw Major professor Date 3’ ”la/(.7 I?0 MS U is an Affirmative Action/Equal Opportunity Institution ‘ 0-12771 PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE ll ' "more ll; mm mm ‘6. It“. flF—fi! MSU lo An Affirmative Action/Equal Opportunity Institution cha-DJ FOREIGN LANGUAGE VOCABULARY ATTRITION: A STUDY OF COLLEGE-LEVEL SPANISH STUDENTS OVER AN lB-MONTH PERIOD BY Marilyn Ruth Bierling A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR or PHILOSOPHY. Department of Romance and Classical Languages 1990 ABSTRACT FOREIGN LANGUAGE VOCABULARY ATTRITION: A STUDY OF COLLEGE-LEVEL SPANISH STUDENTS OVER AN l8-MONTH PERIOD ‘ BY Marilyn Ruth Bierling This dissertation investigates the attrition of foreign language vocabulary in fifteen students of Spanish over an 18-month period. The study examines characteristics of the students and their learning situation, characteristics of the words and the way they were taught and tested, and the relationship between recognition and production vocabulary. The ‘Natural Approach' was used to teach the first semester of college—level Spanish to.these English-speaking students. One group met in a classroom in the United States, and the other group met in Mexico. The first 20 hours of instruction were studied intensively in order to track the students' acquisition of 100 specially-targeted words in Spanish. During the following 18 months the students were retested on these 100 words for both recognition and production in order to study the pattern of attrition. The data were analyzed statistically, and individual cases were studied in depth. The major findings were the following: 1) there were no differences in retention scores based on motivation, effort, sex, frequency of the words in teacher classroom input, or location of the course in the 0.8. or Mexico (when the effects of post-course contact with the 100 target words were eliminated), 2) there were significant differences in I retention scores between words taught and tested for both recognition and production during the course, and words taught and tested for recognition only, 3) there were significant differences in retention scores for words with higher and lower saliency ratings, 4) adjectives tended to be forgotten more quickly than verbs or nouns, 5) retention scores dropped quickly at first and then gradually leveled off, and 6) the best description of recognition and production vocabulary loss was that equal quantities of words were lost over time, rather than that the ratio of production and recognition vocabulary remained constant. The study concludes that requiring students to study for production of vocabulary is an important factor in long— term retention. Other important factors are the saliency of the vocabulary in teacher input, and the mnemonic devices created and structural analyses performed consciously by the student. Such results support a cognitive theory for foreign language acquisition. DEDICATED TO MY HUSBAND NEAL, who through many long hours provided encouragement and support iv ACKNOWLEDGMENTS This dissertation would not have been possible without the encouragement and assistance of many different people who gave of their time to listen to my ideas and to give their advice. I would especially like to thank Dr. Pat Lunn, the director of my dissertation committee. Pat helped me to shape the idea for my research and spent many long hours correcting the drafts and offering her suggestions. I would also like to thank the other members of my committee, Dr. Natalie Lefkowitz, Dr. Ann White, Dr. Grover Hudson, Dr. Susan Gass, and Dr. George Mansour for the help they gave. Dr. Jim Stapleton and William Noble, also of Michigan State University, gave many helpful suggestions regarding the statistical analysis. Several colleagues at Calvin College provided valuable assistance. I especially want to thank Dr. Edna Greenway for her wise advice, Dr. A1 Reynolds for his statistical expertise, and Diane Lucar for the encouragement she gave me in the office we shared through much of the year. Nick Hengeveld and Jeremy Frens of the Calvin Computer Center helped in the preparation of the graphs for the manuscript. Dr. Dick Van Halsema, director of the STS program in Mexico, graciously allowed me to conduct a portion of the research while‘serving as an instructor in this program during the summer of 1988. I also must thank the fifteen students who participated in this study, willingly filling out questionnaires and submitting to interviews over the eighteen months. Robert Knoor is the student who specifically wanted to see his name in print, but all of the students were wonderful individuals who were a joy to teach. Hopefully some of the data can be used to help future students of Spanish. Finally, my husband and three children are to be commended for the encouragement they have given me and the willingness they have shown to take on many of the responsiblities at home so that I would be free to write. Their support through the year has made this dissertation possible. vi TABLE OF CONTENTS LIST OF TABLES O O O O O O O O O O O O O O O O O O O I 0 ix LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . xii Chapter 1. INTRODUCTION. . . . . . . . . . . . . 1 Overview of language acquisition theory and methodology in the 1980s. . . . . . . . 3 Purpose and importance of study . . . . . 10 Brief description of study. . . . . . . . . 11 Research hypotheses . . . . . . . . . . . 11 Subjects. . . . . . . . . . . . . . . 15 Data collection during class sessions . 16 Data collection during retention period . l8 Variables . . . . . . . . . . . . . . . 20 Analysis of data. . .~. . . . . . . . . . 21 Limitations of research . . . . . . . . . 22 Definitions in this study . . . . . . . . . . 24 2. REVIEW OF LITERATURE. . . . . . . . . . . . . . 27 Literature on language attrition. . . . . 27 Different types of language attrition . . 27 Early studies on foreign language attrition . . . . . . ; . . . 30 Research during the 19808 . . . . . . . . 34 Summary of literature on language attrition . . . . . . . . . . . . 47 Literature on foreign language vocabulary . . 50 Importance and organization of the lexicon 50 Vocabulary teaching and learning. . . 52 Word characteristics: word difficulty and frequency in input. . . . . . . . S7 Recognition and production vocabulary . . 59 Summary of research on foreign language vocabulary. . . . . . . . . . . . . . 61 Literature on human memory. . . . . . . . . . 64 Short— term and long-term memory . . . . . 65 Depth of processing . . . . . . . . . . . 73 Cognitive theory. . . . . . . . . . 76 Summary of literature on human memory . . 78 vii The hypotheses. . . . . . . . . . Hypotheses about students . . . . . . . . Hypotheses about words. . Hypotheses about the relationship between recognition and production vocabulary . 3. METHODOLOGY . . . . . . . . . . . . . . . . . . Subjects. . . . . . . . . . . . . . . . . . . Materials and procedures. . . . . Analysis. . . . . . Hypotheses about students . . . . Hypotheses about words. . . . . . . . Hypotheses about the relationship between recognition and production vocabulary . 4. ANALYSIS AND FINDINGS . . . . . . . . . . Analysis by student . . . . . . . . . . . . Statistical study . . . . . . . . . . . Case study. . . ‘ Results for the hypotheses about students Analysis by word. . . . . . . . . . . Statistical study . . . . . . . . . . . . Descriptive word study. . . . Results for the hypotheses about words. The relationship between recognition and production vocabulary . . . . . . . . Statistical study . . . . . . . . . Case study. . . Results for the hypotheses about recognition and production. . . . 5. SUMMARY AND CONCLUSION. . . . . . . . Summary of research . . . . . . . . . . . . . Findings and discussion . . . . . . . . . . . Student characteristics . . . Word characteristics. . . . . . . . Recognition and production. . . . . Implications of research. . . . . . . . For second language acquisition theory. . For foreign language teaching . . Recommendations for further research. . APPENDIX A: STUDENT MATERIALS. . . . . . . . . . . APPENDIX 3: ADDITIONAL DATA AND STATISTICAL RESULTS. LIST OF REFERENCES. . . viii LI ST OF TABLES Table 1. 2. 3. 12. 13. 14. 15. 16. 17. Testing schedule . . . . . . . . . . . . . . . Subject characteristics. . . . . . . . . . Subject characteristics: previous and later contact. 0 O O O O O O O O O O O O O O O O 0 Word characteristics . . . . . . . T-tests for groups in Mexico and 0.8.. . . Correlations of test scores with combined motivation and effort. . . . . . . . . . . . . T—tests for groups based on course grade . . . Correlations of test scores for 15 students over time . . . . . . . . . . . . . . . . . . . . . Correlations with scores at Time 4 . . . . . . T-tests for frequency in input . . . . . . . Correlations of frequency in input with test scores . . . . . . . . . . . . . . . . . . . . One-way analyses of variance for part of speech. One-way analyses of variance for part of speech (Pasos A and B only) . . . . . . . . . . . . . Means and standard deviations for part of speech . Anovas for part of speech, frequency in input, recognition/production: F ratio . . . . . . . Anovas for part of speech, frequency in input, recognition/production: Sig. of F . . . . . . Anovas for part of speech, frequency in input, paso (A-C): P ratio . . . . . . . . . . . . . ix and and and 20 93 94 99 125 127 128 130 131 156 157 158 159 160 163 164 165 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. Anovas for part of speech, frequency in input, and paso (A-C): Sig. of F . . . . . . . . . . . . . . One-way analyses of variance by paso (A-E) . . . . Correlations of test scores with saliency of forms of Spanish words . . . . . . . . . . . . . . . . . Saliency of Spanish words by paso. . . . . . . . . T-tests for groups based on saliency of Spanish word . . . . . . . . . . . . . . . . . . . . . Words best retained at Time 4. . . . . . . . . . . Words with a score of 0 at Time 4. . . . . . . . Other poorly retained words. . . . . . . . . . . . T-tests for saliency of best—retained and worst- retained words . . . . . . . . . . . . . . . . . . Data organized by student. . . . . . . . . . . . Data organized by word . . . . . . . . . . . . . . Ratio of prod. scores to rec. scores: P+R. . . . Difference between prod. and rec. scores: P-R . . Percentage of original vocabulary (Time 0) retained at later times. . . . . . . . . . . T-tests for words taught and tested for both recognition and production and words taught and tested for recognition only. . . . . . . . . . . . Anovas for saliency of Spanish word and teaching. and testing for recognition and production: 3 ratio. . . . . . . . . . . . . . . . . . . . . . Anovas for saliency of Spanish word and teaching and testing for recognition and production: Sig. of P. . . . . . . . . . . . ... . . . . T-tests for words at high levels of recognition and production and words at high levels of recognition but lower production . . . . . . Correlations of test scores for 100 words over time 0 O O O O O C O I O O O O O O O I O 0 Correlations with scores at Time 4 . . . . . . . . 165 166 168 169 170 171 171 172 172 179 179 181 181 183 184 186 186 188 189 190 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. Best predictors at Time 1 for scores at Time 4 The attrition process. . . . . . . . . . . . Data by Data by student 0 O O O O O O O O O O O 0 word 0 O O O O C O I O O O O O O O O 0 Predictions for R4 and P4 (based on R1 and P1) . T-tests One-way contact T-tests T-tests One-way Tf-tBStS for pre-course contact with Spanish. analyses of variance for post-course with Spanish . . . . . . . . . . for differences based on sex for differences based on effort. analyses of variance for motivation. for differences based on motivation. Means for scores by paso . . . . . . . . . . . Correlations of test scores with emotionality of English translations . . . . . . . . . . . . . Correlations of test scores with saliency of English translations . . . . . . . . . . . . . xi 191 194 242 243 248 251 251 252 253 254 254 255 255 255 LIST OF FIGURES Recognition scores for higher and lower grades. Production scores for higher and lower grades SCOIGS SCOIES SCOIES SCOIOS SCOEES SCOEES SCOIES SCOIOS scores SCOIOS SCOIBS SCOIOS SCOIOS SCOIOS SCOIOS of retention scores (recognition and for for for for for for _for for for for for for for for for Sara (student 1) David (student 2). Judy (student 3) Laura (student 4). Rachel (student 5) Gary (student 6) Ed (student 7) . Todd (student 8) Jeff (student 9) . Neal (student 10). Kimberly (student 11). Diane (student 12) Joel (student 13). Karen (student 14) Glen (studentlS). Recognition scores for part of speech . Figure l. 2. 3. Retention 4. Retention 5. Retention 6. Retention 7. Retention 8. Retention 9. Retention 10. Retention 11. Retention 12. Retention 13. Retention 14. Retention 15. Retention 16. Retention 17. Retention 18. Composite production) . . 19. 20. Production scores for part of speech. . xii 129 129 132 132 133 133 134 134 135 135 136 136 137 137 138 138 139 140 161 162 21. 22. 23. 24. 25. 26. Composite of retention scores (recognition and production) Pictures for final Scattergram Scattergram Scattergram Scattergram for R1 for P1 for R1 for P1 oral interview . and R4 . and P4 . and P4 and R4 xiii 132 240 246 245 247 247 CHAPTER 1 INTRODUCTION The decade of the 19805 witnessed an explosion in .research on the processes by which adults acquire a second language. Fueled by new theories and re-worked older ones and a bewildering array of methodologies and movements, this research has attempted to sort out the probable from the improbable and the workable from the unworkable. Closely related to the field of language acquisition research is the field of language attrition research, an area of study that was still declared to be ‘antenatal' in the early 19805 (Berko-Gleason, 1982b:22). A conference on the attrition of language skills, held at the University of Pennsylvania in 1980, called for extensive research into this new field, a call which was heeded during the decade. A survey of the literature in Chapter 2 will demonstrate that most research in the area of second language attrition is.very recent. And, since any kind of longitudinal study in the area of language attrition must obviously stretch over a significant period of time, it is only quite recently that results of such studies are being published. Language attrition can refer to at least four distinct types of loss: 1) loss due to accident, disease, or aging of an individual, 2) loss of a dying dialect or language by a whole group of people, 3) loss of a first language in a second-language environment, and 4) loss of a foreign language in a first-language environment. These different types of attrition have far-reaching ramifications for areas as diverse as psychology, gerontology, dialectology, foreign language education, and 0.3. government policy. Yet until recently, categories 2-4 above were relatively unexplored. In the area of foreign language, specifically, attention has been focused on the processes involved in acquiring the foreign language, rather than on how quickly (or slowly) it is forgotten. Yet research in attrition is inevitably tied to research in acquisition and has a good deal to say about acquisition theory and about the ultimate goal of all our well-intentioned efforts as foreign language teachers. The goal of this study is to focus on one small area of foreign language attrition--the loss of vocabulary that had at one time been either recognized or produced by the foreign language student. In the present chapter I will give a brief overview of second language acquisition theory and methodology in the 1980s and will show the relevance of the present study to this field. I will also give a brief description of the methodology used in this study, state the hypotheses to be investigated, and define terms. Chapter 2 contains an in-depth review of literature in the field of language attrition specifically, as well as recent literature on vocabulary acquisition and human memory. Chapter 3 will explain in detail the methodology used in this research, and Chapter 4 will give an analysis and evaluation of the data. Finally, in Chapter 5, I will discuss the findings and point out some implications for second language acquisition theory and methodology in the foreign language classroom. OVERVIEW OF LANGUAGE ACQUISITION THEORY AND METHODOLOGY IN THE 19808 One of the most influential models for second language acquisition during the 19805 was Stephen Krashen's Monitor Model. This model has provoked research and produced teaching methodology such as the ‘Natural Approach' (Krashen 1982 and 1985, Rrashen and Terrell 1983), as well as much criticism. Briefly stated, the Model consists of five hypotheses: l) the acquisition/learning hypothesis, 2) the natural order hypothesis, 3) the monitor hypothesis, 4) the input hypothesis, and 5) the affective filter hypothesis. Terrell has used these five hypotheses as the theoretical basis for his popular first-year Spanish textbook Dos mundos (1986c), the textbook used in the classroom during the present study. In Dos mundos, vocabulary is taught through a great deal of comprehensible input in the form of teacher talk, utilization of pictures and real objects, and, in the beginning lessons, through Total Physical Response. Because of the emphasis on the importance of comprehensible input, comprehension receives priority over production, at least in the beginning stages of acquisition. Students are encouraged to bind the form of the Spanish word to the meaning, with the goal being automatic retrieval of the meaning of a word. Later on, students begin to use the new words in their emerging speech, hopefully without reliance on English as an intermediary. Krashen's Monitor Model has not gone unchallenged. Although all five hypotheses have been criticized, the most controversial seem to be the acquisition/learning distinction and the input hypothesis. The controversy over the acquisition/learning distinction centers around the emphasis on meaning versus form and subconscious versus conscious learning, and whether learning can turn into acquisition. Sharwood Smith (1981, 1986) has pointed out the value of consciousness raising in the acquisition process. Both VanPatten (1984) and Terrell (l986a,b) minimize the subconscious/conscious dichotomy by saying that binding form to meaning can occur on either a conscious or subconscious level. Some conscious learning activities may be helpful to make certain features more salient and to aid in their comprehension and access. Terrell (1989) states that grammar instruction may be useful as an ‘advanced organizer' (giving the learners information that will help them make sense of the input) and will also provide multiple occurrences of a form for acquisition and for practice in monitoring. Byrnes (1984) argues that there is ample evidence that formally learned knowledge can become part of the subconscious automatic processing capability. Gregg (1984) attacks Krashen's terminology. saying that the acquisition/learning and subconscious/conscious distinctions are ill defined and trivial. Higgs (1985b) asserts that as meaning in the target language becomes obscure, more conscious information becomes important for the student. Long (1983) calls for Krashen to reevaluate the importance of ‘learning', and, indeed, in his 1985 book, Krashen admits that acquisition can be aided by learning. McLaughlin (1979, 1987) also criticizes Rrashen's acquisition/learning hypothesis and asserts that controlled processing can become automatic. McLaughlin's cognitive theory will be discussed further in Chapter 2 in the section on memory. The term ‘consciousness' itself is ambiguous. In a perceptive paper to be published in Applied Linguistics, Richard Schmidt points out that ‘consciousness' can be defined as awareness, as intention, or as knowledge. Schmidt concludes that ‘subliminal language learning is impossible, and that noticing is the necessary and sufficient condition for converting input to intake' (page 1). He believes that incidental learning is possible, although adults are helped by paying attention and by intending to learn. The issue of implicit learning is difficult to resolve, because there is evidence for implicit learning, as well as for the facilitative effect of conscious knowledge and understanding. The input hypothesis states that "humans acquire language in only one way -- by understanding messages, or by receiving ‘comprehensible input'"(Kra5hen 1985:2). Speech is the result of acquisition and not its cause. However, others emphasize the need for output and negotiation of meaning as well. Swain (1985, 1989) points out that after several years of comprehensible input in immersion programs in Canada, children still produce speech with many errors. Input is essential, but not sufficient. When the student is pushed to produce output, he is also forced to produce a more sophisticated analysis of the language, testing the hypotheses that he has formed. What is needed in the immersion programs is greater opportunity for the students to interact with each other in the foreign language, thus producing ‘comprehensible output'. VanPatten (1987, 1989) also believes that negotiation and interaction may be crucial. He calls for more sophisticated accounts of the role that consciousness plays in the processing of input. Other writers have pointed out the difference between input and intake. The term intake was introduced by Corder (1967), who used it to describe ‘what goes in and not what is available to go in' (165). According to Hatch (1983), input is what the learner hears and tries to process, and intake is what is successfully processed. Chaudron (1985a,b) has given ideas for testing the way that a learner processes input and turns it into intake--which may be encoded into short-term or long—term memory. Gass (1988) describes a five-step process: a) apperceived input: input which is noticed by the learner, b) comprehended input: that which is understood by the learner, ranging from semantic comprehension to full structural analysis, c) intake: mental activity mediating between comprehended input and the learner's grammar; process of assimilating linguistic information, d) integration: storage of the information possibly leading to a reanalysis of the learner's grammatical system, and e) output: that which a learner produces. Also related to this general area of input is the question of the nature of input that learners receive. Lightbown (1983, 1987) has extensively explored the relationship between input in the classroom and the output of students. Gass and Madden have edited a book entitled Input in Second Language Acquisition (1985) containing several articles which ‘have a direct link to empirical research' (Scarcella and Perkins 1988:348). Another large area of research and controversy related to the input issue is the relationship of comprehension (both listening and reading comprehension) to production (both speaking and writing). The skills involving comprehension are also referred to as recognition or the receptive skills, and production is also known as recall. Although there is evidence that production in some cases may precede comprehension (Sheldon and Strange 1982), many writers such as Postovsky (1974), Nord (1980), Winitz (1981), Asher (1981), and Krashen and Terrell (1983) insist on the priority of comprehension in foreign language instruction. They say that students will produce when they have received considerable input and are ready to do so. Ostyn and Godim (1985) believe that a minimum vocabulary in a foreign language is 5000 words, a goal which can be met in two years of language study at the college level if more emphasis is placed on recognition and less on production. Davies (1976) calls for well-developed receptive skills--a logical goal for two years of language study in Swedish classrooms; production skills should be reserved for the more motivated students. The tests that accompany Terrell's textbook Dos mundos focus on listening and reading comprehension and vocabulary and grammar recognition through a multiple choice format, though suggestions are also given for speaking and writing tests. Gass (1988:213) points out that ‘there is a lot of mileage that needs to be travelled between the input to the learner...and what the learner produces. We cannot assume that with mere presentation of language information, whether implicitly or explicitly, students will convert it to output'. Hatch (1983) describes how vocabulary may be ‘penciled in', (for example, a certain word x exists, has two syllables, and begins with 22), but there is no evidence to support that ‘penciled in' forms later become part of the learner's system. Rivers (1986) points out that comprehension and speaking are very different skills--in comprehension small details may be ignored, but not in Speaking. In other words, ability to recognize words and forms will not necessarily lead to the ability to produce them. Krashen's Monitor Model is thus related to a great deal of the current research and controversy in the field of second language acquisition. When this model is criticized, as in Gregg (1984, 1986), Spolsky (1985), and McLaughlin (1979,1987), it is most frequently mentioned that Krashen lacks operational definitions and that his hypotheses are tautological or cannot be falsified. This brief overview of second language acquisition theory, controversy, and methodology in the 19805 cannot be complete without mentioning the Oral Proficiency Movement. While an approach such as the ‘Natural Approach' emphasizes the acquisition of a large recognition vocabulary, textbooks based on the Oral Proficiency Movement, such as Spinelli and Rosso-O'Laughlin's Encuentros (1988), emphasize production. In the Oral Proficiency Movement, vocabulary is still essential for the beginning speaker, as pointed out by Higgs (1982, 1984, 1985a,c). In the relative contribution model presented by Higgs, vocabulary and pronunciation are the most critical factors at the lowest level of proficiency, whereas control of grammar is essential to go from the advanced to the superior level. Since many students seem to ‘fossilize' at the advanced (2 or 2+) level, however, it is important to teach grammatical accuracy from the beginning of instruction. Omaggio (1984) stresses concern for development of linguistic accuracy from the beginning. The 10 proficiency movement has been criticized by several writers, including Lantolf and Frawley (1985), Savignon and Bachman (1985), Schulz (1986), VanPatten (1987), and Kramsch (1986,1987), the chief criticism being that grammatical accuracy has been given too much weight. However, the ACTFL proficiency guidelines are becoming an influential base for organizing foreign language courses. PURPOSE AND IMPORTANCE OF STUDY While teaching during the 1987-88 school year with the Natural Approach, an approach that emphasizes large recognition vocabularies at the beginning of second language acquisition as noted above, I became interested in several questions that had direct impact on my teaching. What were the characteristics of the teacher input, of the learning environment, and of the learner that might lead to long-term vocabulary retention? What were the characteristics of the words themselves that led to intake by the student and long- term retention? Through which stages did the students pass in the acquisition and the loss of vocabulary? What is the relationship between amount of recognition and production vocabulary? Does this relationship change over time? Should large vocabularies be taught for recognition only-- and be tested for recognition only--during the beginning stages of second language acquisition? 11 These questions are timely, since they ask about fundamental issues of methodology in current comprehension- based approaches, as well as issues in teaching and testing for recognition and/or production. A basic assumption that was made before beginning this study was that without continued contact with specific vocabulary items, foreign language students over time would lose some of their ability to access these items and that certain patterns in recognition and production would emerge. It is hoped that by studying the attrition of foreign language vocabulary, we can learn something more about human language--how we acquire it and how we lose it. BRIEF DESCRIPTION OF STUDY Research hypotheses The research hypotheses were divided into three groups: 1) hypotheses about students, 2) hypotheses about words, and 3) hypotheses about the relationship of recognition vocabulary to production vocabulary. \ Hypotheses about students: 1. Students who study in a target-language (Spanish- speaking) environment retain more vocabulary than those who study in a native-language (English-speaking) environment, 12 even when the specific vocabulary learned is not rehearsed or used outside of class. 2. Students who have pre-course contact with Spanish (either informally in the culture or formally in the classroom) will retain more vocabulary than those who have no previous contact. 3. Students who have post-course contact with Spanish (either informally in the culture or formally in the classroom) will retain more vocabulary than those who have no further contact, even when the specific vocabulary learned is not rehearsed or used. 4. Vocabulary is retained at different rates according to sex. 5. Students who report more effort while taking a language course retain more vocabulary than those who report less effort. 6. Students who report more intrinsic motivation for the course retain more vocabulary than those who report more extrinsic motivation. 13 7. Students who receive higher grades for the course retain more vocabulary than students who receive lower grades. 8. Students who receive higher scores on the first retention tests continue to receive higher scores on later retention tests. Hypotheses about words: 9. Vocabulary that is repeated more frequently in teacher classroom input is retained at higher levels than vocabulary that is repeated less frequently. 10. Verbs, adjectives, and nouns are retained at different rates. 11. Vocabulary learned earlier in the course will be retained at higher rates than vocabulary learned later in the course. 12. Vocabulary with high emotionality or saliency ratings is retained longer than vocabulary with low ratings. 14 Hypotheses about the relationship of recognition to production: 13. Recognition scores for vocabulary are higher than production scores. 14. Both recognition and production scores for vocabulary drop quickly at first and then gradually level off. 15. The best description of recognition and production vocabulary loss is that equal quantities of words are lost over time rather than that the ratio of production to recognition vocabulary remains constant over time. 16. Smaller percentages of recognition vocabulary are lost than of production vocabulary. l7. Vocabulary taught and tested for both recognition and production is retained at higher rates than vocabulary taught and tested for recognition only. 0 18. Ability to produce a word initially, rather than ability to recognize a word, is the best predictor of both recognition and production later on. 15 Subjects The subjects of this study were college-level students from two beginning Spanish classes that the researcher taught during the summer of 1988. The first class was an intensive 3 1/2-week course taught at Calvin College in Grand Rapids, Michigan, from May 31 to June 22. The nine English-speaking students were in class for 3 hours each day and used audio tapes and completed homework assignments outside of class. This Spanish class was taught, of course, in an English—speaking environment. The second class was an intensive 3 l/2-week course taught in Merida, Mexico, from June 26 to July 18. This class was part of a program for English-speaking college- level students who were interested in working with the Presbyterian Church in Mexico and who followed their 3 1/2- week course with field assignments in Central America and Mexico before returning to the 0.8. and Canada. Six students participated in this class. Both classes were taught entirely in Spanish with the ‘Natural Approach', using the textbook Dos mundos by Tracy Terrell et al. Only the first 20 hours of instruction were included in the acquisition period for the present project, corresponding to the first five 23525 ‘steps' in the textbook. Students were followed for eighteen months after the end of their class. 16 Data collection during class sessions Students at the beginning of each course were given a questionnaire asking about any former contact with Spanish and motivation and attitudes toward learning Spanish. Students were also aware that vocabulary research was being done (though not the details) because of the consent form they were given the option of signing. The teacher did not see these consent forms, nor the final evaluation of the course, until after the final grades were given. Vocabulary was introduced in ‘teacher talk' with the use of pictures and objects and through Total Physical Response. All new words were written on the board and copied in student notebooks. After the first three hours of instruction, students were encouraged to begin responding with one-word answers in Spanish, and by the end of the first 20 hours students were participating in widely-varying communicative activities. Short grammatical explanations had also begun, and students had begun writing short compositions in their workbooks. Data collection during the class sessions occurred through normal classroom activities. The following behaviors in class and for tests were assessed: -motor response to commands -oral response to pictures -picture naming and description (both oral and written) -translation 17 —guided composition -matching of words (oral and written) to objects and pictures -written multiple choice items By using these multiple measures the researcher was able to record which vocabulary items had entered a) the recognition stage and b) the production stage by the end of the first 20 hours. The only obtrusive instrument used was a tape recorder for recording the input of the instructor during class, as well as oral responses during individual oral testing. The tape recorder was also used at the end of the course, when students were asked to do some introspection about their vocabulary learning strategies. Students were asked why they could recall certain words and not others, what methods they used for studying vocabulary, and how often they mentally translated into English. In the final interview eighteen months later, the tape recorder was also used for oral interviews. In addition to the regular vocabulary of over 600 words introduced in Pasos A-E in the text, the researcher added 100 words which were not taught in Dos mundos nor in the second-year book at Calvin. Some of the words came from Spanish dialects other than those spoken in Mexico. These 100 words formed the core of the research in determining the attrition rate, since the students were likely to hear these words as input only during the first 20 hours of 18 instruction. When a student indicated that he or she had further contact with one of these target words, the word was dropped from the data for that particular student. These 100 words belonged to categories of words already introduced in the lesson, such as family members, clothing, and parts of the body, and were introduced along with the other vocabulary. Examples of these target words are madrastra ‘stepmother', cerilla ‘match', and barbilla ‘chin'. Since students studied all words from their vocabulary lists copied off the board, it was not readily apparent that these words were not included in the vocabulary lists at the end of each paso in the text. By the end of the first 20 hours, these words had been checked for both recognition and production in regular tests in order to form a baseline for later retention testing. Data collection during retention period At the end of each 3 1/2 week session, the students were given their first test to measure retention of these 100 words, which were mixed in with other words that had been taught. This test was in written format (see Appendix A) and used pictures that the students had seen before in class. Students were also shown scenes containing several of these 100 vocabulary items during individual oral interviews and were asked to describe these scenes into the tape recorder. 19 The students in the summer 101 class at Calvin continued immediately with Spanish 102 and then had a break between July 21 and September 12, at which time they continued with second-year Spanish. These students retook the written picture test at the endof 102 and again in April of 1989. The students in Merida left immediately for field assignments after finishing lOl--and went to locations in Central America and Mexico where they needed to use Spanish. They met again in Mexico City on August 8 and then returned to the 0.8. and Canada on August 13. During their stay in Mexico City they retook the written picture test, and they repeated the test in May of 1989. These students were asked to indicate which words they had come into contact with in their experiences. Of the 100 target words, only a few were mentioned, and these words were dropped from the data for the student that mentioned them. All fifteen students were asked to retake the written picture test in November and December of 1989. Oral interviews were also performed with the twelve students who were still in the Michigan area. The entire testing schedule for the two groups is illustrated by Table 1 below: 20 TABLE 1 TESTING SCHEDULE CLASS TIME 0 TIME 1 TIME 2 TIME 3 TIME 4 l | I T l #1 |during | June 22,| July 1988 [April | Nov.- Grand Ra-lfirst 20| 1988 | (at end |1989 | Dec. pids, MI |hours of| | of Spanishl | 1989 |class | | 102) | | I I I I | #2 |during | July 18,| August IMay | Dec. Merida, |first 20| 1988 | 1988 (in |1989 | 1989 Mexico lhours of| | Mexico | | ' [class I I City) 1 1 TIME 0: Baseline testing TIME 3: 9 months after TIME 1: at end of Spanish 101 previous testing TIME 2: 3-1/2 weeks after TIME 4: 7 months after previous testing previous testing Variables The dependent variables in this research are on vocabulary recognition and production tests at different points in time: the baseline tests and the scores five retention tests at four different times. Thus there was a set of ten scores per student and also per word, depending on how the data were organized. Several independent variables were chosen for study because of their possible effects on recognition and production scores. One important variable was the learning environment, whether an English-speaking environment or the target-language environment. Other independent variables related to the learners were their contact with Spanish both before and after the course, their sex, their effort and 21 motivation, and the grades they received in the course. Variables relating to the vocabulary words themselves were their frequency in classroom input, their part of speech, the order in which they were taught, their emotionality and saliency ratings, and whether they were originally taught for recognition and production or for recognition only. Potential variables that were controlled in the research were the method (‘Natural Approach"used), the level of the instruction (first semester of college—level Spanish), the intensity and length of the courses (3-1/2 weeks), and the instructor (same instructor for both courses). Analysis of data The data were analyzed using the SPSS (Statistical Package for the Social Sciences) program at Calvin College. Correlations were performed, as well as several tests to test for significant differences between means (T-test, One- way analysis of variance, and Anova). The level of significance chosen was .05. A multiple analysis of variance was performed at Michigan State University in order to study the relationship between initial recognition and production as predictors of later scores. Information not analyzed statistically but important for the final discussion included written work produced by the students during the first 20 hours of instruction, final 22 oral interviews at the end of the course and after the final vocabulary test, and introspective comments by students on how they learned and remembered vocabulary items. To illustrate relationships between recognition and production vocabulary over time, several graphs have been included in the text. Limitations of research .I Because of the desire to conduct the research under natural conditions in actual classrooms, this study has four major limitations: 1. small sample size--The research was conducted during intensive summer courses in a small college and a comparatively small program in Mexico, and thus only fifteen students participated. This small sample size is partially offset by the large amount of data resulting from ten separate tests per student with 100 items per test. However, care must be taken in applying the results of this study to the general student population. 2. non-random groups--In real-world educational settings, classes are not normally formed by random assignment, but rather by student choice or scheduling necessities. By choosing a number of independent variables, the researcher could group the students in different ways, 23 but again care must be taken in applying the results to other educational situations. 3. post-course contact with vocabulary items--Students were asked to identify which of the vocabulary items on the tests they had read or heard in later contact with Spanish, and these items were eliminated from the data for that individual. However, it is possible that some students missed identifying some of these items. 4. repeated tests--Although the later tests were separated by several months and target vocabulary items were interspersed with different non-target items for each test, scores for some students may have been influenced by seeing the target items on an earlier test. An additional limitation was caused by the requirement by Michigan State University that students sign consent forms for the research. Thus students knew that they were subjects of research on foreign language vocabulary, although they did not know the exact nature of the research. Such knowledge may have resulted in the ‘Hawthorne Effect', in which students perform differently because they are aware that they are participants in an experiment. 24 DEFINITIONS IN THIS STUDY Since these terms will be used frequently throughout the study, working definitions are given here in order to clarify their meaning. Second language acquisition: Internalization of the structure and vocabulary of a second language in either a first language or second language environment, by either formal or informal means. In this definition ‘acquisition' is not meant to contrast with ‘learning', and ‘second language' is not meant to contrast with ‘foreign language'. Second language vocabulary: Spanish words taught with pictures, objects, or actions in context in the classroom. Vocabulary retention: The ability to access a word through either recognition or production. Vocabulary attrition: The failure to access a word through either recognition or production. Vocabulary recognition: The ability to match a vocabulary word to a picture, object, or action or to pick the correct word in a multiple choice item. 25 Vocabulary production: The ability to write or say a vocabulary word when shown a picture, object, or action (also known as recall). Input: Teacher talk in the classroom and worksheets containing practice items in Spanish. The Natural Approach calls for methods (such as use of context or pictures) that insure that foreign language input is comprehensible to the students. Intake: The process during which learners assimilate linguistic information noticed and comprehended in input into an incipient linguistic system. Frequency in classroom input: Number of times that instructor repeats vocabulary word in input teacher talk. Saliency of vocabulary word: The ability of a word to attract attention due to its conspicuous, striking, or unusual nature, as indicated by classroom learners on rating scale. Extrinsic motivation: Motivation resulting from pressure outside the student. Intrinsic motivation: Motivation resulting from a desire within the student with no external pressure. 26 Natural Approach: A teaching methodology based on the following principles: 1) comprehension precedes production, 2) speech emerges in stages, 3) speech emergence is characterized by grammatical errors, 4) group work encourages speech, 5) students acquire language only in a low-anxiety environment, 6) the goal is proficiency in communication skills (Terrell et al. 1986:7-9 in Dos mundos Instructor's Manual). Total Physical Response: A teaching methodology in which students respond physically to commands in the foreign language (Asher 1981). Regression Hypothesis: A hypothesis stating that language attrition is the mirror image of language acquisition (that which is last learned is first forgotten). These definitions provide a useful working vocabulary for this and following chapters. CHAPTER 2 REVIEW OF LITERATURE This chapter will review literature in three areas that are relevant to this research: 1) language attrition, 2) foreign language vocabulary, and 3) human memory. At the end of each section the literature will be summarized, and at the end of Chapter 2 each research hypothesis for the present project will be scrutinized in light of what the literature says about it and what gaps still exist in our knowledge. LITERATURE ON LANGUAGE ATTRITION Different types of language attrition Although the purpose of the present investigation is to study the attrition of a foreign language, research in other types of language attrition should also be mentioned. Such research can alert us to patterns that may be significant in foreign language attrition as well. So, before going on to review literature in the area of foreign language attrition specifically, I will briefly review recent research in the following areas: 1) attrition of a first language through accident, disease, or aging, 2) the attrition of dying dialects or languages by entire groups of people, and 3) 27 28 attrition of a first language in a second language environment. Aphasia, the loss of language due to some sort of trauma, has been thoroughly researched. "Studies have indicated...that the dissolution of language in aphasia does not mirror acquisition by children. Aphasic subjects may be reduced to a very limited lexicon, but it rarely consists of baby's first words, such as ‘mommy' and ‘juice', and it is more likely to consist of politeness routines and/or unprintable words" (Berko-Gleason 1982a). Obler (1982) has shown that naming is the most universally impaired skill, both in aphasia and in normal aging, with content words. being harder to recall than function words and nouns harder than verbs. Obler and Albert (1984) found that naming ability increases through age 60, then gradually begins to decrease after 60, and steeply declines in the 705. With disease such as Alzheimer's, naming difficulties occur even in the mild stages (Pan and Berko-Gleason 1986). The regression hypothesis (that which is last learned is first forgotten) has not been substantiated in research on loss of a first language through accident or disease (Caramazza and Zurif 1978:x). Dorian (1973, 1978, 1981, 1982) is often cited in connection with research in the area of languages in danger of extinction, such as East Sutherland Gaelic. In reference to comments made by the less fluent speakers themselves, she notes that "explicit comment on the decline in the quality 29 of their Gaelic focuses almost entirely on the lexicon...: the younger speakers feel sure their elders had more ‘words for things' than they have themselves" (1978:590-91). Her studies, as well as others on languages as diverse as Navajo, Frisian, and Breton, emphasize change in the language or dialect as a whole across generations, with close attention to sociolinguistic factors such as language prestige and official support for bilingualism. Recently, research has increased on the attrition of a first language (L1) in a second language (L2) environment. Much of the data has come from migrant workers in Europe and from second and third-generation immigrants in the United States, for example Brewer-Bomar (1981) on the Spanish of two children in the U.S. and Gonzo and Saltarelli (1979, 1983) on emigrant languages. Gonzo and Saltarelli (1979) found that as Spanish-speaking immmigrants gradually lost their first language, there was increased lexical borrowing from English from one generation to the next. Leyen (1986) studied native language vocabulary decline and concluded that 1) extreme L1 decline occurred only among subjects who had left their L1 environment during early childhood, 2) language attrition was mostly in the area of lexicon, and 3) the production of lower frequency L1 lexicon became impaired over time. Lexicon among bilinguals appeared to be stored in a common master lexicon and impairment occurred in the access rather than in the master lexicon itself. In connection with first-language loss research, Jaspaert, 30 Kroon, and Van Hout (1986:41) make the interesting point that the relation between language loss and time is probably non-linear. Many growth processes show an S-like curve, and this is probably also true in the case of language decline. In summary, studies in these other areas of language attrition point to the critical nature of the lexicon. In the loss of a dialect by a whole group of people or in the loss of a first language in a second language environment, the lexicon seems to be the most sensitive to attrition. In aphasia, diseases such as Alzheimer's, and in normal aging, naming is also the most impaired skill, with nouns being harder to recall than verbs. Thus the investigation of the loss of vocabulary seems to be a potentially important area of research in foreign language attrition as well. Early studies on foreign language attrition Before 1980 there were very few studies on the retention of a foreign language learned in a first-language environment. One of the earliest studies is that of Anderson and Jordan (1928) on the retention of Latin words and phrases. After a period of eight weeks, only one-half of the meanings of Latin words was retained (students were asked to write the English equivalent of the Latin word). It was also noted that those who learned more, retained more, although this result showed up only when comparing the most extreme cases. A few early reports attempted to 31 measure the effect of summer vacation on the loss of language skills. Kennedy (1932) found that his students showed a 15-35% loss in their Latin syntax after the summer vacation, and he also suggested that there was a positive correlation between language retention and original attainment and desire to continue study of Latin. Geoghegan (1950) attempted to replicate and to extend this research. His study included Latin, French, and Spanish high school students, and he tested vocabulary as well as grammar. The results were mixed, with Latin students showing significant losses in both grammar and vocabulary and Spanish students showing significant losses in vocabulary only. French students even showed gains in vocabulary. Scherer (1957) reported on a series of experiments that he had performed over a period of six years. He had tested college students at the end of the first year of German and again at the beginning of the second year in grammar, reading, and vocabulary. Scores in vocabulary and grammar decreased insignificantly, while reading scores increased slightly. Scherer also concluded that a week of review at the beginning of the new school year was enough to bring students back to their previous level. Pratella (1969) studied the effect of summer vacation in all four skill areas--listening comprehension, speaking, reading, and writing. Significant losses were sustained only in the area of listening comprehension. Pratella also concluded that sex and verbal ability were not important factors in loss 32 over the summer vacation. In summary, these short-term studies of loss over three months indicate that vocabulary may be one of the areas most likely to shOw attrition. Smythe, Jutras, Bramwell, and Gardner (1973) conducted two Canadian studies on the retention of French reading and listening skills by high school students in London, Ontario. Over the summer, students showed a small but significant loss in reading comprehension and a small but significant gain in listening comprehension. The second study compared the high school students returning to French class immediately after the summer semester with those who waited to begin study of French again until the second semester. As expected, the group that waited for eight months to resume study of French scored lower than the group that waited only three months; yet the total loss of the lower— scoring group was only 4%. Smythe (1973:405) concluded that ‘it has been demonstrated that only relatively small changes in second language performance were recorded over reasonably long periods during which students were not receiving instruction'. The three—month vacation could even have a facilitating effect on language retention. Cohen (1974, 1975) studied elementary school children who had been acquiring Spanish in an immersion setting in California. In his first project he tested fourteen children on their speaking abilities before and after summer vacation (between the first and second grades). After the summer break the children made more errors overall, and 33 there were more problems with article and adjective agreement, as well as errors in person inflection of verbs. Utterances in general were shorter. The following summer Cohen studied three children intensively to see if the last things learned were the first to be forgotten. Two of the three children provided data that at least partially supported this hypothesis. For example, one child had finally shown general command of the ser/estar distinction just before the end of first grade, but at the end of the summer recess she was overusing ggppg. Cohen also suggested that the forgetting process could produce new incorrect patterns, but also in some cases ‘a pause in the learning process may actually cause a reduction in certain problem areas' due to the elimination of incorrect hypotheses about. the language (1975: 137). Edwards (1976, 1977) studied French-dominant and English-dominant bilinguals in the Canadian Public Service who had completed courses of instruction six months, twelve months, and eighteen months earlier. This study differed from the ones above in that the subjects were adults and continued to have contact with the second language after finishing their course of study. After six months there were no significant differences between language-dominant groups, but after twelve months the English—dominant group showed significant losses in speaking, though they gained in reading skills. More decline in speaking occurred after eighteen months. The French-dominant group showed no 34 decline in their English skills after six, twelve, and eighteen months. In questionnaires this latter group indictated more frequent use of the second language. Edwards concluded that retention was more related to use and a supportive environment than to motivation or attitude. These studies before 1980 show that much more needed to be done in the area of foreign language attrition research. However, some interesting patterns did emerge: vocabulary usually showed some attrition even over short periods of time, while listening and reading comprehension results varied, and in some cases scores in these skills were even higher after a summer vacation. Cohen (1975) concludes that there is some evidence to support the regression hypothesis (at least in speaking), and Edwards (1976, 1977) believes that a supportive environment which encourages use of the second language after the termination of a course is more important to retention than motivation or attitude. Research during the 19805 In May of 1980 a conference on language attrition was held at the University of Pennsylvania, the proceedings of which were published in The loss of language skills (1982), edited by Lambert and Freed. This conference and book provided a major impetus for research in the area of language attrition. The subjects in the book range from the relation of psycholinguistics and sociolinguistics to 35 language loss (articles by Berko-Gleason, Gardner, Dorian, Obler) to guidelines for conducting research (Oxford). In an article entitled ‘The U.S. Government's Foreign Language Attrition and Maintenance Experience', Lowe states that second language atttrition is a serious problem for the U.S. government. The C.I.A. Language School, on the basis of anecdotal information, feels that subjects start losing their language skills as soon as they stop practicing them. Candidates with low-level skills lose them more rapidly than those with higher-level skills. Also Lowe believes that ‘Speaking is least stable, perhaps so unstable as to never be constant; Understanding occupies a middle ground, with some languages being harder to retain understanding in than others; and Reading is the most stable of the skill modalities' (1982:181). In the conclusion to his article on ‘Determining the Linguistic Attributes of Language Attrition' in the same book, R. Andersen states his views on vocabulary attrition: one important area of study should be the ‘quick retrieval of appropriate vocabulary and idiomatic phrasing in on-going speech production. My own experience as an LA [language acquirer] and especially my observations of my children (as well as other individuals) convinces me that this area far outweighs morphosyntactic and phonological aspects of attrition' (1982:113). Gardner in his article in this book gives two hypotheses related to second language retention: ‘1) Since 36 attitudinal/motivational characteristics are related to the level of second language proficiency, they will relate to second language retention, 2) Since attitudinal/motivational characteristics are related to indices of participation in language-related situations, they will relate to attempts to maintain second language skills once training has terminated' (1982:41). Gardner refers to this post-training period as the ‘incubation period'. In more recent research on social factors (Gardner, Lalonde, and MacPherson 1985), attitudinal/motivational factors seem to be more influential during the acquisition period (since they influence the level of competence acquired) than during the incubation period. Attitudes toward the second language were not directly related to use of the language after the completion of training. These conclusions are similar to those of Edwards (1976, 1977). After the 1980 conference and the 1982 publication of The loss of language skills, research in the area of second language attrition drastically increased. 'In October of 1982 Theo Van Els convened a second conference on language attrition at the Catholic University of Nijmegen in the Netherlands, and in March of 1986 a Language Loss Symposium was held in Kerkrade in the Netherlands. The papers from this symposium have been published in a volume entitled Language attrition in progress, edited by Weltens, De Bot and Van Els. Also in 1986 Applied Psycholinguistics devoted an entire issue to the theme of language attrition. One of 37 the most recent testimonies to the importance of this field is that in July of 1989 an entire issue of Studies in Second Language Acquisition was dedicated to discussion of language attrition. Several of the articles coming out of these publications are included as part of the discussion in this chapter. Out of the 1980 conference grew the Language Skills Attrition Project (LSAP), which is a long-term effort to gather parallel data about language attrition in Arabic, Chinese, and Japanese as second languages. In their article in Language attrition in progress, Lambert and Moore (1986) describe progress that has been made in specifying criterion variables by developing diagnostic tests in listening, speaking, and reading (each with a subtest in lexicon and structure), and in specifying predictor variables. In the same book, Ginsberg (1986:20) writes that the major goal of LSAP is ‘to determine the nature and extent of language skill attrition on the part of selected U.S. populations in Arabic, Chinese, and Japanese, both on a global basis and as a function of student-specific, language-specific, initial instruction-specific, and interim exposure/use variables'. It is assumed that ‘language competence is a complex, multidimensional phenomenon whose components are not necessarily acquired or lost at the same rates or for the same reasons' (1986:21). A number of interesting studies have appeared on the loss of a second language by children after returning to 38 their first language environment. Hansen (1980) wrote a dissertation on the acquisition and attrition of Hindi-Urdu by two English-speaking children. The oldest child (eight years old at the end of the study) spent three periods of time in India, but between periods of exposure apparently totally lost the second language and had to relearn it. The younger child (six years old) spent two periods of time in India, but showed a similar pattern of acquisition and attrition. Even though the children had virtually no conscious memory of Hindi-Urdu, previous experience seemed to facilitate the relearning. Hansen also states that the process of forgetting was the reverse of the acquisitional sequence. Olshtain has studied the attrition of English in Hebrew—speaking children who had acquired a native-like mastery of English in the U.S. before returning to Israel (Berman and Olshtain 1983, Olshtain 1986, 1989). She shows that the result of attrition is greater variability in the application of marked rules, resulting from both a reversal of the acquisition process and from language transfer from Hebrew, and lower accessibility of specific lexical items. She says that we can assume ‘reduced accessibility in vocabulary retrieval in all situations of attrition where there is a reduction of language use over longer periods of time. In cases of lexical retrieval difficulty, we can expect all attriters to develop some kind of compensation strategies' (1989:163). The greatest loss was shown by the 39 youngest children (ages five through eight), most of whom had no further contact with English once they returned to Israel. The older children (through age 14) received further exposure to English in Israel, and ‘although they lose command of rapid retrieval of certain vocabulary items and of certain syntactic porperties of English, these children do retain a fair amount of native-like command of English' (1983:233). . Cohen (1989) reports on lexical attrition in a third language, specifically the attrition of Portuguese in his own two children who were already fluent in English and Hebrew. Cohen had Daniel (age 9) and Judy (age 13) retell the same story after one, three, and nine months of discontinued contact with Portuguese. He found a significant decrease in lexical production, especially in the younger child and especially with respect to nouns, although Daniel could still identify the words in an oral comprehension task. The children ‘used at least six lexical production strategies in order to compensate for forgotten words--two of them Ll-based (borrowing and foreignizing) and four of them intralingual (the use of a*general word, approximation, circumlocution, and word abandonment)’ (1989:147). (Foreignizing involves using an L1 root word with an L2 inflection.) The interest in the question of whether language attrition is the reverse of language acquisition is reflected in the studies of Hansen (1980) and Olshtain 40 (1986, 1989) above. The most recent article dealing with the regression hypothesis (Jordens, De Bot, and Trapman, 1989) examines attrition in German case markings. For German-speaking immigrants who had spent at least ten years in the Netherlands (Ll-German), grammatical case preserves its semantic function. However, for Dutch students who had studied German and then had minimal contact with it later (L2sGerman), the nominative case was used as the default case, reflecting the reverse of the acquisition process, in which the nominative case was the first learned. These results tend to support the regression hypothesis for L2 speakers of German, but not for L1 speakers of German who are losing their language in a second language environment. Bahrick (1984a,b,c) carried out some remarkable cross- sectional research at Ohio Wesleyan University by testing 733 alumni, some of whom had never studied Spanish (the control group), some of whom were studying Spanish at the time, and others who had finished studying Spanish one to fifty years before. Independent variables included level of training reached, grades obtained, and opportunities to use Spanish since leaving formal study. Those with more than casual contact with Spanish since finishing their courses were eliminated from the study. Reading comprehension, vocabulary recognition and recall, grammar recognition and recall, idiom recognition and recall, and Spanish word order were all tested. 41 Bahrick's study produced a wealth of interesting data. Up until six years after taking a language course, retention rates dropped, but then remained level for several decades, finally to fall once more. Thus, after an initial loss, ‘the remainder is immune to further losses for at least a quarter of a century, and much of that content survives for fifty years or longer' (1984c:111) in what Bahrick calls ‘permastore'. Bahrick also found that l) rehearsals during the retention interval such as listening to television programs in Spanish or conversing in Spanish were limited and had little if any influence on the retention scores, 2) the original level of training influenced the proportion of knowledge with permastore longevity, 3) production vocabulary stabilized later than recognition vocabulary, 4) the production of grammar declined most rapidly, and 5) reading comprehension was maintained at a level determined by recognition of vocabulary and grammar (1984c:111). Bahrick also notes: 1) much content survives more than fifty years; 2) grades received in courses continue to be valid predictors of performance for several decades following training; 3) training of a single course in the language is likely to leave little if any permanent content; 4) the larger the number of courses taken the greater the portion of content with permastore longevity. It appears that the total amount of content to be forgotten during the five years following training is relatively constant for individuals at different levels of training, but this amount becomes a progressively smaller portion of total knowledge with higher levels of training...; 5) attrition affects smaller portions of recognition vocabulary than of recall vocabulary. This difference may reflect the fact that the recognition vocabulary is larger, and that 42 the absolute amount of attrition is approximately the same (1984c:116). Neisser (1984) questions Bahrick's use of the word ‘permastore'. Instead ‘Bahrick could have made essentially the same argument simply by postulating that some response strengths reach a critical threshold during learning; beyond that threshold, they become immune to interference or decay' (1984:33). Thus, information that is tied into an extensive cognitive structure is resistant to forgetting (1984:34). As an interesting comparison to Bahrick's research, De Bot and Lintsen (1986) and De Bot and Clyne (1989) have gathered data on elderly Dutch-English bilinguals living in Australia. They studied the reversion of these elderly immigrants to their first language (Dutch) to see if there was a final decline in L2 ability as observed by Bahrick. They found language reversion in the immigrants who had never reached the ‘critical threshold' as mentioned by Neisser above, but other immigrants showed ‘surprisingly little loss of proficiency in both Dutch and English over the years' (1989:167). These elderly people, in contrast to the subjects studied by Bahrick, had ample practice in their second language, and thus were able to break through the ‘critical threshold'. Meara (1989) presents an interesting theory about vocabulary retention in which students eventually reach an equilibrium between words forgotten and words spontaneously regenerated. ‘Spontaneous regeneration' refers to 43 remembering words that the student was previously unable to access. The point of his article is that vocabulary testing techniques should look at the underlying pattern of vocabulary attrition rather than superficial surface phenomena. He gives three suggestions for testing programs: 1) the week should be the basic unit of time, 2) the attrition pattern between week 2 and 3 is the best predictor of the final equilibrium state, and 3) use of a simple, neutral assessment instrument, such as having students cross out foreign vocabulary words that they do not know well enough to say what they mean. Although he includes no data with the article, his equilibrium model is interesting because it predicts that learners will eventually reach a state in which scores will not drop further, and because he suggests a very simple instrument for testing vocabulary attrition. He does not deal with the effects of testing students each week or clarify whether the same words would be tested each time. Two fairly recent studies have investigated the effect of summer vacation on second language retention. A dissertation by Robison (1985) reported the results of testing 180 junior and senior high Spanish students before and after the summer vacation. Robison concluded that 1) listening and reading comprehension were not affected, 2) less-advanced students showed significant loss in writing while more advanced students did not, 3) vocabulary recognition and recall were significantly affected, though 44 from a practical point of view loss was minimal, 4) FLES students retained more than non-FLES students, 5) no loss was noted in ability to produce linguistically correct forms and no evidence showed that forgetting is the chronological reverse of learning, 6) a sequence of forgetting was observed, varying according to one's level of achievement, and 7) exposure to Spanish during the summer raised performance. Robison's conclusions that significant loss was noted in vocabulary, but not in listening and reading comprehension, are similar to the conclusions of studies done before 1980 about the effect of summer vacation. The results of a study by Moorcroft and Gardner (1987) seem to contradict Robison's conclusions about the forgetting sequence. Moorcroft and Gardner tested 114 students of French twice: upon their completion of 9th grade and again as they started the 10th grade. Their conclusion is that ‘while grammatical structures in general are more likely to be affected by language loss than vocabulary, it seems to be the most recently acquired structures that are most vulnerable to language loss'. (1987:337) The loss in grammatical accuracy was related to the fact that ‘most grammatical structures are incompletely and recently learned' (1987:338). A number of other recent studies should be mentioned before concluding this section on language attrition research. Godsall-Myers (1981) in a dissertation on the attrition of German in six college students concluded that 45 1) the attrition proceeds at different rates for different skills, 2) the attrition is inversely related to the cumulative average the student had in the low-intermediate German class, and 3) the attrition of L2 patterns will reflect the learning order and the frequency of occurrence of L2 patterns. She also suggests that vocabulary be taught using semantically paired items and synonyms and antonyms and that students be taught principles of word formation in German. In 1984, the same year that Bahrick reported on long- term retention of Spanish taught in the U.S., Clark and Jorden reported on retention of Japanese taught in U.S. schools. As in Bahrick's study, participants were asked to fill out retrospective questionnaires. In contrast to Bahrick's study in which participants had little contact with the foreign language after finishing course work, Clark and Jorden found that many of their participants had extensive contact with Japanese after finishing their courses. Clark and Jorden did not include the many college students whose only contact with Japanese had been a single- year part-time course, since ‘the language acquisition of such students is so minimal that, except for commonly used ritual expressions and the most basic grammatical patterns, attrition is almost total after a comparatively short period of time away from the classroom' (1984:16-17). No correlation was found between attrition of speaking and listening comprehension skills on the one hand and attrition 46 of reading skills on the other hand, showing the independence of these skills. Some other interesting conclusions were that "no differences were found between ‘attriting' and ‘non-attriting' students in the reported ease with which they were able to learn foreign languages in general or Japanese in particular" (1984:52-53) and that ‘decreases in language proficiency since the time of maximum performance are directly related to decreases in the extent to which these skills are in fact currently being used' (1984:53). For those students who originally had a good functional control of Japanese and had seemed to reach a critical point in overall proficiency, renewed exposure to‘ real-life settings was a powerful means for re-acquisition (1984:55,58). Fakhri (1985) examined morphological and syntactic simplification in the interlanguage of American college students of French after periods of three, eight, and ten months. Morphological simplification in the interlanguage occurred in written narratives, and also in written narratives the students increased their use of full noun phrases in order to avoid clitic pronouns. Fakhri called for further study in other areas of language attrition, including the loss of vocabulary in comprehension and production (1985:157). Weltens, Van E15, and Schils (1989) reported on a four- year study on the attrition of French receptive skills in Dutch students who had finished their study of French four 47 and six years earlier, but who had had very little contact with French since then. These students had originally been at a very high proficiency level, since they had received 400-600 hours of training in French, and therefore had probably mastered French above Neisser's ‘critical threshold'. The researchers report gains in listening and reading proficiency and say that these gains might be attributed to ‘general cognitive maturation, further academic training, and continued learning of other foreign languages' (1989:214). Lexical tests and grammar tests showed some attrition. Confirmation of Bahrick's findings were shown in that 1) most loss occurred in the first two years and then leveled off and 2) a fixed amount of material was lost rather than a fixed proportion. However, unlike Bahrick, they found that attrition was independent of training level. Summary of literature on language attrition The research during the past few years in second language attrition has been overwhelming. No longer can the field be called ‘antenatal'. A summary of the research at this point is difficult, but some general observations can be made: 1. Attrition of the lexicon is an extremely important area for study in all types of language loss. For example, 48 R. Andersen believes that it ‘far outweighs morphosyntactic and phonological aspects of attrition' (1982:113). 2. The level that a student reaches affects subsequent attrition (Kennedy 1932, Godsall-Myers 1981, Lowe 1982, Bahrick 1984, Robison 1985). A single foreign language course is unlikely to leave any content (Bahrick 1984, Clark and Jorden 1984). There may be a ‘critical threshold' of language skill that must be reached in order for much to survive (Neisser 1984, Weltens et al 1989, De Bot 1986,1989). 3. Higher grades received in course work are related to higher retention rates (Godsall-Myers 1981, Bahrick 1984). In contrast, Clark and Jorden (1984) report no difference in initial ease in learning the language for ‘attriting' and ‘non-attriting' students. 4. Motivation and attitude may have little to do with attrition rates (Edwards 1976, 1977, Gardner 1985). Kennedy (1932) did report higher retention rates for students who had a desire to continue their study of Latin. 5. The regression hypothesis (first to be learned is last to be forgotten) does not appear to be true for first language loss (Caramazza and Zurif 1978, Berko-Gleason 1982a), but may apply in cases of second language loss (Cohen 1974, 1975, Hansen 1980, Godsall-Myers 1981, Moorcroft and Gardener 1987, Olshtain 1986, 1989, Jordens et a1. 1989). However, more work at this point has been done in morphology and syntax than in vocabulary. 49 6. Language attrition is most rapid at first, but appears to reach a state where it levels off (Bahrick 1984, Weltens et al. 1989). Anderson and Jordan (1928) found rapid initial attrition of Latin vocabulary. Meara's matrix theory (1989) predicts rapid initial attrition of vocabulary with a leveling off and a final state of equilibrium. 7. A fixed amount of material appears to be lost rather than a fixed proportion (Bahrick 1984, Weltens et al. 1989). 8. Production skills (speaking, writing) usually suffer more attrition than recognition skills (listening, reading) (Scherer 1957, Edwards 1976 and 1977, Lowe 1982, Bahrick 1984, Robison 1985). These generalizations will be discussed specifically in relation to the research hypotheses for this project at the end of Chapter 2. SO LITERATURE ON FOREIGN LANGUAGE VOCABULARY Importance and organization of the lexicon With the advent of comprehension-based approaches to foreign language acquisition and the call for communicative competence, there has been renewed interest in the importance of vocabulary. Extensive bibliographies compiled by Meara (1983, 1987) contain entries calling for a massive input of vocabulary from the earliest stages of language learning (such as Judd 1978), an entry insisting on the importance of vocabulary acquisition at the middle stages of language learning (Twaddel 1973), and another stating that vocabulary learning is problem number one at the advanced level (Marton 1977). As noted in Chapter 1, vocabulary and pronunciation are considered the most important components for the novice level in the Oral Proficiency Interview. Meara (1984) cites an article by Levenston (1979) concerning issues and problems in second language vocabulary acquisition and points out that studies of the lexicon in the interlanguage tradition are not very numerous in comparison with the importance accorded to the lexicon. It has been shown that for native speakers lexical errors are more disruptive than grammatical errors, and lexical errors consistently outnumber grammatical errors by three or four to one (Meara 1984:229). One of the reasons for the neglect of the lexicon by linguists is the belief that it, ‘unlike 51 our knowledge of syntax and phonology, is an inherently messy part of our linguistic competence, and that it cannot be handled conveniently using the sorts of rules which have been used to describe the more manageable parts of a language' (1984:230). In response to Meara's assertion that little work has been done on the lexicon, Sharwood Smith (1984) points out that such fields as word morphology, semantics, phonology, and syntactic and pragmatic coding touch on lexical matters. Even generative grammar has become much more lexically oriented, and the area of psycholinguistics has much to offer. Gass (1987) agrees that many areas of research "impinge on questions of the lexicon', but still ‘the I lexicon has been dealt with somewhat tangentially' (1987:130). Fundamental questions remain as to how and what is learned, the organization of the lexicon, and access and retrieval of the lexicon. Meara (1984) also considers what a typical lexical entry might look like in a second language learner's memory. The entry would consist of two parts, a phonological or orthographical representation (perhaps only partially spelled out), and a semantic representation specifying meaning (perhaps as part of a semantic tree or cluster). Meara mentions various methods that one might use for ordering, storing, searching for, and accessing these lexical entries, but there is no principled way of choosing among them. Meara also compares the lexicons of second 52 language learners with the lexicons of native speakers and finds several differences: 1) semantic networks are much more loosely organized and unpredictable for learners, especially for material recently learned, and ‘semantic factors are frequently overridden by extraneous phonological factors' (1984:234); for example, some learners associated prétre ‘priest' with argent ‘money', confusing it with préter ‘to lend'; 2) phonological representations of words may be only partially sketched in, and although initial sounds and consonant clusters tend to be preserved, the placing and nature of vowels is not so reliable; 3) different languages may use different word-handling strategies, for example, syllables may play ‘a much more important role in the representation of words for Spanish speakers than is the case for English' (1984:235). Henning (1973) has demonstrated that beginning language learners tend to encode vocabulary into acoustic clusters (according to similarities in sound), while higher-proficiency learners tend to cluster words semantically, relying more on similarities in meaning than on sound. Vocabulary teaching and learning The realization that words come in semantic networks or sets is extremely important for teachers and learners. According to Ludwig (1984:559), ‘the foreign language teacher cannot look at words in isolation. In any language, 53 each word suggests other formally or functionally similar words...’ D. Brown (1980) encourages the teaching of collocations (words that often occur together, such as £3335 and ppgpp), clines (gradations of words such as freezing, 221$, £22191 warm, hot, and boiling water), and clusters (a group of similar things, such as heat, Fahrenheit, conduct, melt, hot, temperature, sun, and calorie). In addition to semantic associations, structural and mnemonic associations can be formed (Cohen 1987). Structural associations are made when students analyze the root and affixes of a word to understand the meaning. There are several kinds of mnemonic associations; for example, in a chain-type mnemonic the learner links words through their use in a story or rhyme. An example of a verbal mnemonic is a Spanish speaker's use of the word gpppg (strikes) to remind him of the English ghglk, by forming the sentence La tiza se choca con la pizarra (1987:45). An example of an imagery mnemonic is the keyword technique, which has been well researched by Cohen and others. In the keyword mnemonic, the learner thinks of a word in his native language that sounds like the foreign language word. For example, p359 (duck) in Spanish sounds like the English pp_. The next step is to visualize a duck and a pot together, for example, a duck with a pot on his head. The next time that the learner hears the word papp, he will remember the image and the meaning gppk. The keyword mnemonic has been tested for effectiveness in many 54 different research projects. When tested under laboratory conditions, the results are always positive when compared to other methods of learning vocabulary, but when tested in the classroom, results have been mixed. Not all students use the keyword method when instructed to do so, and some students conjure up mental images even when not taught to do so. In the cases of younger children, better results are obtained when the teacher creates the image for the child, but adults seem to do just as well whether they create their own image or use one created for them. Cohen (1987) also answers criticisms of the keyword technique. Though creating a keyword image ties the foreign word to the native language, there is a strong tendency for students to translate to their native language anyway, even in a method such as the Natural Approach. The keyword mnemonic also focuses on just one meaning of the word to be learned, but in cases of words with multiple meanings, perhaps this one meaning can trigger the other meanings. Nouns are definitely the easiest to learn with the keyword approach, and verbs and adjectives are harder. Even though it appears that using the keyword mnemonic would take more time than learning a word through other methods, groups using keywords do better even when under time constraints. Some research has shown that associational devices have lasted up to eight weeks, but there is a need for research over longer periods of time to see if and when the keyword drops out. 55 Although some writers believe that vocabulary can be learned more quickly from word lists than within context (Nation, 1982, cites a number of experiments and concludes that for the initial learning, words in isolation are just as or more effective as words in context), other writers (D. Brown 1980) encourage learners to use the context to find meanings of words. Part of the problem is determining just what ‘context' means. Is context the presence of a nondefining sentence, a defining sentence, or the presence of a story? And what can be done to ensure that the learner is really learning from the context in the foreign language, rather than simply forming a covert bond between the foreign language word and a native language word as if he were learning from a word list (Nation 1982:23)? Roster (1987) emphasizes context in the recognition and comprehension of speech; for example, ‘He sent me a Christmas guard' was often accepted by Dutch speakers as a correct English sentence, since they interpreted gpppg as 223d (1987:25). Schouten-Van Parreren (1989) contends that words should not be taught in isolation since such words present neither a linguistic nor a psychological reality. Isolated words cannot ‘evoke emotions or involvement in the learner, a factor which plays an ...important part in long-term acquisition' (1989:77). Haastrup (1989) provides a cautionary note about context. She believes that if the context is too predictable and the word is too easily guessed, 56 comprehension is aided, but not long-term learning. However, if the context does not make the meaning of the word obvious, and if the learner must do some analysis of the surface structure of the word, the word will have a more distinctive representation in memory and will be better retained. A word judgment of this type may require more effort, but it may pay off later in better recognition and recall because of the ggpph of analysis (see page 73 of this dissertation for a description of Craik and Lockhart's ‘depth of processing' framework). The keyword mnemonic, as well as Asher's Total Physical Response, are techniques that may also increase depth of analysis, owing ‘their success to the dual psychological realities of imagery and concreteness of individual words' (Ludwig 1984:555). ‘Visual information appears to interact with verbal stimuli, demanding that the learner make a linguistically sophisticated analysis'. There is also a good deal of research on the use of pictures in learning foreign language vocabulary, substantiating the value of imagery. Deno (1968) observed that there were problems with using pictures to learn foreign words, since the meaning provoked by the picture may not be the one intended. However, Kellogg and Howe in 1971 demonstrated that, at least for children in grades 4-6, the use of a picture instead of an English word was superior in a paired-associate learning situation. Some of the children, though, learned faster with words than with 57 pictures, showing a variety in learning style. A study by Champagnol in 1972 also showed the superiority of learning with pictures, especially in retention of the words twelve weeks later. The use of pictures with words produced better performance than words without pictures, but the best performance of all was produced by the use of real objects. D. Brown (1980) has compared vocabulary acquisition to a hill that grows taller and taller, but which is constantly threatened by the erosion of forgetting at the bottom. She summarizes the methods for developing vocabulary and preventing forgetting with the formula eight Cs and a G: use of collocations, clines, clusters, cloze procedures, context, consultation (with a book or dictionary), cards (flashcards with native language on one side and target language on the other), creativity, and guessing (with and without context). Word characteristics: word difficulty and frequency in input Another interesting area of research is determining the relative difficulty of words. Nation (1982:17) states that words are difficult when they are hard to pronounce or when ‘their English translations are adjectives, adverbs, or verbs, rather than nouns'. Cohen (1986:148) makes an interesting point when he states that nouns may be easier to forget because they are easier to learn. Foreign words are easy when they are similar in form and meaning to their 58 English translations or when they are ‘morphologically transparent' and their meanings can be understood from the meanings of their parts (Laufer 1989:11). Murray (1986) studied the characteristics of words that would make them easier to translate into a foreign language. He obtained ratings for 145 words according to characteristics such as frequency, memorability, intensity, emotionality, goodness, pleasantness, and concreteness, and then asked the students to translate the English words to French. He noted that variables such as emotionality and imagery, which influence long-term episodic memory, had little effect on translation time, but ‘translation efficiency was most strongly influenced by the frequency of the word in the language, its familiarity, and the similarity of the French equivalent to the English word' (1986:353). Ludwig's article on word characteristics (1984) points out that nouns and concrete words are easier to learn because they can evoke a visual image in the mind. Also words that are positively loaded are recalled better than — words with neutral or negative connotations. A few writers have reported on the number of times a new word must be repeated before students normally remember it. Nation (1982:17) cites evidence that seven repetitions are sufficient for most students. Carpay (1975, cited in Palmberg 1987:208-209) proposed a 4+1+1+1 formula: ‘New words should be used in at least four different contexts in the introductory lesson, followed by (at least) one 59 occurrence in each of three subsequent lessons'. On the other hand, for indirect (incidental) vocabulary learning in context, Nation (1982:17) says that the average number of encounters needed to recognize a word is approximately sixteen. Nation makes other comments on indirect vocabulary learning, saying that even though the greatest share of vocabulary learning is indirect, direct learning of vocabulary is still an important component that speeds vocabulary development (1982:15-16). Recognition and production vocabulary A very important area of study, also mentioned in Chapter 1, is the relationship of recognition vocabulary to production vocabulary. At this point I will review the literature specifically on foreign language vocabulary, but the general concept of recognition and production will again be discussed in the third section of this chapter on studies in human memory. Recognition vocabulary is sometimes known as receptive or passive vocabulary, but the term passive seems to be ill chosen, since it implies that the learner has nothing to do. Melka Teichroew (1982) has written a perceptive article on recognition (receptive) vocabulary versus productive vocabulary. She first reviews various tests for receptive vocabulary, ranging from the rather arbitrary method of 60 checking off words in a dictionary to some very commonly- used methods such as multiple choice and translation tests. The most common production test is to ask the subject to complete a sentence in L2 with the aid of a translation in L1 or a related word in L2. Melka Teichroew goes on to question the point where receptive knowledge ends and productive knowledge begins. She notes that in reception we need to hear or see only enough information to distinguish one word from another, and she refers to Brown and McNeill's (1966) tip-of-the-tongue experiment in which subjects had enough information to recognize a word (and often tell the number of syllables and the first or final letter) but were unable to produce it. She points out that there are many different features indicating familiarity with a word: its spelling, phonology, morphology, and semantic and syntactic properties. A student who can produce a word may be familiar with it to a greater or lesser degree. Melka Teichroew summarizes various reports that indicate that receptive vocabulary in a second language is about twice the amount of productive vocabulary, especially for beginning students, but at later stages the gap seems to close. She concludes that recognition and production are points along a continuum beginning with the first stage of recognition and ending with various stages of production depending on familiarity with the word. The form of a test has strong influence on the results that are produced. She believes that there is only one lexical store which we use 61 receptively or productively according to need, and she presents two questions for further research: 1) is there a meaningful gap between comprehension and production, or is this gap more seeming than real? and 2) does the same comprehension/production problem exist in L2 as in L1? Cohen (1986) also emphasizes that vocabulary acquisition is a continuum. Palmberg (1987:203) asks if all words necessarily pass automatically from recognition to production knowledge, or if some words never reach the production stage and others bypass the recognition stage. He also believes that the relative sizes of the recognition and production vocabulary are ‘independent until the productive vocabulary gets exceedingly large'. Summary of research on foreign language vocabulary Certain ideas recur in the literature on foreign language vocabulary. These generalizations will be discussed specifically in relation to the research hypotheses for this project at the end of Chapter 2. 1. The lexicon is an extremely important component of foreign language acquisition, although it has often been neglected. Lexical errors are more serious and outnumber grammatical errors (Meara 1984). 62 2. Words may be only partially ‘sketched in', with initial sounds and clusters the most important. Recognition does not require as much information as production (Melka Teichroew 1982, Meara 1984). 3. Beginning learners tend to cluster words according to phonological rather than semantic features (Henning 1973, Meara 1984). It is important to encourage semantic clustering of words and to form semantic networks (D. Brown 1980, Ludwig 1984). 4. Mnemonic devices such as the keyword technique show positive results under laboratory conditions, but have mixed results in the classroom (Cohen 1987). 5. Opinions on the effects of context differ. Some writers say that words can be learned as well in isolation (Nation 1982) or caution that too much context may discourage structural analysis of a word (Haastrup 1989). Others stress the importance of using context to find and retain the meaning of a word (D. Brown 1980, Schouten Van-Parreren 1989). 6. Depth of analysis results in better retention of a word (Craik and Lockhart 1972). 63 7. Learning words through pictures and real objects generally results in better performance (Kellogg and Howe 1971, Champagnol 1972, Ludwig 1984). 8. Features that make words easier to learn are part of speech (nouns), similarity to L1, morphological transparency, and frequency in the language (Carpay 1975, Nation 1982, Ludwig 1984, Cohen 1986, Murray 1986, Laufer 1989). Emotionality and imagery had little effect on translation time in Murray's research (1986), but other writers (such as Ludwig 1984) assert the importance of emotionality and imagery in vocabulary learning and retention. 9. Recognition and production are points along a continuum (Melka Teichroew 1982, Cohen 1986). 10. Opinion is divided about the relationship of size of recognition and production vocabulary. Melka Teichroew (1982) cites reports indicating that for beginners, recognition vocabulary is about twice the size of production vocabulary, although the gap seems to close for more advanced learners. Palmberg (1987) believes that the sizes of recognition and production vocabularies are relatively independent, except when productive vocabulary is very large. 64 LITERATURE ON HUMAN MEMORY As Weltens (1987:23) has noted, research on memory and forgetting has a long tradition, going back to Ebbinghaus in the late nineteenth century. However, much of this research lacks relevance for studies in language attrition, since experiments done under laboratory conditions will not necessarily show what actually happens in real language learning and forgetting situations. Bahrick (1979:298) states: Methodologically, most research has been limited to information acquired in the laboratory and tested later for retention. This longitudinal approach offers high control over the conditions of original acquisition, but it limits research to simple material that can be acquired within a few laboratory sessions and to short retention intervals...The point is not that all previous memory research is irrelevant to the acquisition and maintenance of knowledge systems. Past findings dealing with interference, level of processing, distribution of practice, and so on, almost certainly have a bearing on how well knowledge is maintained, but it is impossible to determine the applicability of any findings without a data base relevant to the acquisition and maintenance of knowledge under more ecologically realistic circumstances. As Bahrick points out, language learning and maintenance are problems that must be studied in a real— world context. However, language attrition is a problem intrinsically connected to human memory, and we cannot fail to note the implications of verbal learning experiments in psychology for language attrition studies. This section will first deal with problems related to short-term and 65 long-term memory, including problems of recognition versus recall and intentional versus nonintentional learning. Then this section will turn to recent discussions of depth of processing and declarative and procedural memory and will relate these discussions to the field of language attrition research. Short-term and long-term memory Traditionally, human memory has been divided into three different stores: 1) sensory memory, which holds visual and acoustic information for up to two seconds, 2) short-term memory (also known as primary or working memory), which is limited and holds information up to thirty seconds, and which requires rehearsal if the information is to be retained, and 3) long-term memory (also known as secondary memory), with no known limit and which may last from minutes to years. Loss in this memory store is due to interference from new information. Some psychologists also include tertiary memory, which is long-term memory immune to forgetting through interference. Bahrick (1984a,b,c) refers to this type of memory as ‘permastore'. Short-term memory is thought to have a capacity of seven ‘chunks' of information, plus or minus two. When subjects are asked to repeat words in a list, they show both a primacy effect and a recency effect; thus the first few and last few items are the best remembered, whereas the 66 words in the middle are more likely to be forgotten. The resulting memory curve is known as the serial position curve. Both Meara (1980a) and Spitze and Fischer (1981) have performed interesting experiments in short-term memory with language learners. Whereas native speakers are able to chunk words in a sentence into larger syntactic and semantic groups, thus being able to hold more words accurately in short-term memory, non-native speakers show the typical serial position curve, with the middle items particularly fuzzy, as if the items in the sentence were unrelated. This inability to form semantic and syntactic chunks simultaneously becomes more pronounced the lower the level of the learner. Data from Henning (1973) support the view that low-proficiency students attempt to register vocabulary in memory through sound (acoustic) similarities, whereas high-proficiency students rely more on semantic associations. Pimsleur (1967) provides a ‘memory schedule' for learning individual words which is frequently cited: When a word reaches a probability of 60% of being remembered (perhaps after only 5 seconds in short-term memory, especially if there is intervening material), it should be repeated. Since forgetting is exponential in form, the next repetition of this word should come 52 (or 25) seconds later, the next repetition 5’ (125) seconds (or two minutes) later, and so on. Pimsleur calls this schedule ‘graduated interval recall'. 67 Many experimenters have found that repeating the presentation of items enhances both recognition and recall of these items. However, giving the learner more time to‘ rehearse items on his own improves recognition memory but has little effect on ability in free recall (Eysenck 1982:221). Ebbinghaus performed the first experiments on short- term and long-term memory in the second half of the nineteenth century. His attempts to recall lists of nonsense syllables led to the traditional ‘forgetting curve', and his work set the pattern for further research in serial learning and paired-associate learning in the twentieth century. Serial learning refers to learning a series of items in order, and paired-associate learning refers to a specific stimulus memorized with a specific response. Retention can be measured by 1) recall (in serial recall the items of a list must be recalled in order, in free recall the items can be recalled in any order, and in cued recall the subject is given a clue and must respond appropriately), 2) savings (difference in time taken to learn information a second time), or 3) recognition (either asking the subject to identify an item as ‘old' or ‘new' or giving a multiple-choice test). A great deal has been written in the psychological literature about the difference between recognition and recall. Testing for recognition memory generally produces the higher scores, though scores may also be dependent upon 68 the subjects' knowing which mode they will be tested in (Tversky 1973). Wingfield and Byrnes (1981:31) give these reasons for the superiority of recognition: 1) recognition tests ‘restrict the size of the set of alternatives from which the correct response must be drawn', 2) correct judgments can be made on the basis of partial recall, and 3) if alternatives on the multiple choice test are very different from each other, the test can be very easy (however, if the alternatives require discrimination between fine nuances of meaning, the recognition test may be even more difficult than the recall test). J. Brown (1976) has edited a book entitled Recall and Recognition. In it he states that recognition can be mediated by recall, especially when the subject must search through many items (1976:2). Low levels of learning suffice for recognition, but not for recall, since the amount of information required for a correct response is lower in recognition. Brown also speaks about research design: In assessing differential effects, an extremely tempting within-subject design is available. Retention can be tested first by recall and then by recognition. This greatly increases the precision of the experiment since recall and recognition are then tested on the same subjects, on the same material and on the same occasions. One practical difficulty is that the statistical analysis of the results can be tricky. A simple solution if recognition and recall are assessed .using the same measure may be to analyse in terms of the difference score (recognition minus recall) for each subject: the statistical analysis is then conducted on a table of independent difference scores (1976:34). 69 In the second chapter in the same book, Tulving (1976) questions whether recognition and recall are fundamentally two different processes, or if they are basically the same process of using stored information. He comes to the conclusion, as do Lockhart, Craik, and Jacoby in Chapter 3, that the two do not differ in a crucial way: ‘they are different only in the sense that in recognition re— presentation of the stimulus provides better information from which the initial encoding can be reconstructed' (1976:85). In Chapter 5 Cooper and Monk deal with details of testing. Testing may cause consolidation of the learned material and stereotyping in recall. The same items tend to be recalled whether they are correct or incorrect (in reference to free recall from a list). On repeated recognition tests, false alarms become consistent. A point made by Murdock in Puff (1982:5) is that recognition memory can be tested for latency (time taken for subject to respond) and subject's confidence level, as well as for accuracy. In the same book Bahrick and Karis (1982:433) make the following point about recall and recognition measures: ...with high degrees of original learning and short retention intervals, recognition performance is likely to remain near the ceiling and fail to reveal forgetting, whereas recall measures are more sensitive. With low levels of learning, or very long retention intervals, recall performance may decline to a point where it is no longer sensitive to further retention losses, whereas recognition performance may continue to reflect changes over time....ecological memory investigations that extend over long time periods are likely to require 70 both recall and recognition measures to reflect adequately the changes occurring during the entire retention interval. Research done on the ‘tip of the tongue' phenomenon is related to the concept of recognition and recall. Brown and McNeill ( but other al. 1987) 1966) wrote the first article on this phenomenon, articles have appeared since (for example, Kohn et . In the study by Brown and McNeill (1966), definitions were given, and even when subjects were not able to recall features, number of Sometimes ' and other word. It ‘penciled reached t the words, they were often able to give some such as the beginnings and ends of the words, the syllables, and the location of the primary stress. this ‘tip of the tongue' state ended in recall, times only in recognition when shown the correct seemed that some features of the word were in' more faintly than others. Brown and McNeill he following conclusion: It is consequently possible to recognize words when one has not stored t e complete letter sequence. The evidence is that we do not store the complete sequence if we do not have to. We begin by attending chiefly to initial and final letters and storing these. The order of attention and of storage favors the ends of words because the ends carry more information than the middles. An incomplete entry will serve for recognition, but if words are to be produced (or recalled), they must be stored in full (1966:335). Brown and McNeill add that the need to produce encourage Also the s the registration of the middle letters of a word. growth of vocabulary should force attention to the 71 middle of a word, because more detail is needed to distinguish one word from another as vocabulary increases (1966:337). Retrieval of a word may also be heavily influenced by -context or the way that an event was encoded (Craik 1979:88). According to Weltens (1987:32) ‘recall should be facilitated by a test situation which more or less reflects the learning situation'. By using think-aloud procedures, Cohen (1986) discovered that in addition to using structural, semantic, and mnemonic associations as retrieval strategies, learners attempted to recall the learning situation, such as the textbook or the situation in which the word came up. Wingfield and Byrnes (1981:44) also assert that recall can improve when a test occurs in the same environment where learning occurred. From his experiments in free recall, Jenkins (1974:788) concludes that ‘the most important determiner of recall was the nature of the gygpp the subjects experienced when their task brought them into contact with stimulus words. This finding suggests that subjects recall the quality of the events they have experienced, not stimuli to which they have been exposed'. Questions have also been asked about the effect of intent to learn. Is incidental learning as effective as intentional learning? According to Eysenck (1982:197, 201- 202), most human learning is incidental, since no test of learning is expected. From recent experiments it seems as though the orienting task given before learning and the type 72 of processing performed by the learner are more influential than intent to learn. ‘The most appropriate generalization is probably that intention to learn has little or no effect on recognition memory, whereas its effects on recall are more variable'(1982:206). Another question frequently asked about long-term memory is what causes forgetting. Do memories spontaneously decay over time, or is access blocked because of interference from new material? Most of the research points to the validity of the interference hypothesis, that what we learn before and after any piece of information will influence how well we will remember that information. Time is a factor in forgetting, but not its pgpgg. Bahrick's study on long-term retention of Spanish has already been summarized in the first section of this chapter. He has done other cross-sectional studies of long- term ecological memory, that is, memory in real-life social and cultural settings rather than laboratory settings. In a 1979 study, Bahrick used alumni from.Ohio Wesleyan University for testing maintenance of knowledge of the city of Delaware, Ohio, and compared this retention to the subjects' opportunities for ‘rehearsal' of this knowledge (the frequency, duration, recency, and distribution of their trips back to their alma mater). Bahrick (1982:428) pointed out that we must study variables that are ‘ecologically important rather than just those that are easily manageable...It is acceptable to sacrifice some control over 73 critical variables in order to investigate ecologically important phenomena that would otherwise be neglected'. In connection with longitudinal testing designs, Bahrick (1982:432) states: Testing the same individuals repeatedly has the advantage of control over individual differences; however, there is the danger that later test performance will be affected by the practice or interference resulting from the earlier tests and will therefore fail to yield an unbiased estimate of the amount of forgetting during the interval. To avoid this problem, laboratory research in memory has primarily used between-subject comparisons. Semantic memory content is generally retained over much longer time intervals than the intervals involved in typical laboratory research, and the problems created by successive testing are likely to be trivial because of the long intervals between successive tests. For that same reason, however, the longitudinal approach becomes cumbersome. In this article Bahrick goes on to describe the longitudinal studies done over the summer vacation by Cohen (1975) and Smyth et al. (1973) and gives his own methods for conducting cross-sectional research. Depth of processing Craik and Lockhart (1972) proposed an alternative view to the ‘duplex' model of short-term and longeterm‘memory. They offered a model based on ‘depth of processing', ranging from a very shallow sensory analysis of a stimulus to a deeper analysis based on meaning. They proposed that deeper memory traces or codes are more enduring than shallow ones, 74 and forgetting is a function of the depth of processing (Wingfield and Byrnes, 1981:282). Memory is a result of the cognitive operations that have been performed on the stimulus. Processing for semantic features yields much better word retention than processing for phonological features or for physical structure (for example, if the word is written in capital letters or not). The depth of processing concept has resulted in more calls for meaning and communicative activities in the classroom (for example, Stevick 1976:43). Since the appearance of this model, there have been some amendments to it. Rehearsal does improve recognition, even if it is on the same and not on a deeper level. Also some sensory memories, such as the sound of a voice, can be very enduring. It is difficult to specify the mechanisms that lead to deep processing or to define it, and it is impossible to prove that deep processing has pp; occurred. Depth of processing can easily be confounded with factors such as processing time, difficulty of processing, or distinctiveness of processing. (Depth of processing may also make a memory more distinctive and thus lessen interference from other memories.) However, depth of processing remains a viable model and has been the impulse for a great deal of research. The advantages for learning with pictures can perhaps be explained through depth of processing, because imagery can lead to a more meaningful (or perhaps a more 75 distinctive?) analysis. An alternative explanation is that pictures induce a dual code--an imaginal code and a verbal one. In any event, ‘pictorial presentation and imagery coding substantially facilitate both recall and recognition memory' (Paivio 1976:120). Another interesting item to note is the effect of personal significance on memory. Stevick (1976: 38-39) reports on several experiments showing that memory is better when a subject's ego is involved (for example, when the subject reads a passage supposedly derived from his own personal inventory) or when a word is emotionally loaded (such as money, slut). Schouten-Van Parreren (1989:77) concurs that words are better learned when they evoke emotions or involvement in the learner. There are other terms for talking about memory and dividing it up into different types. Common terms are semantic and episodic memory. Semantic memory refers to general rules that we know, but that are removed from the context in which we learned them-~such as 2 X 2 - 4 or that an apple is fruit (Bahrick 1979:297). Episodic memory refers to memory of particular events related to time and place, such as yesterday's menu or an accident we had. Variables such as word frequency have a greater effect on semantic memory and variables such as emotionality and imagery affect retrieval from episodic memory to a greater extent (Murray 1986). 76 Cognitive theory In their article applying cognitive theory to second language acquisition, O'Malley, Chamot, and Walker (1987) discuss the role of memory. According to them, language is not learned separately from other cognitive skills but is stored and retrieved from memory like other complex cognitive skills. They believe that the short-term, long- term memory model is inadequate for explaining language acquisition, and they prefer the theoretical model developed by John Anderson (1983, 1985). Things we know 32223 (static information) constitute declarative knowledge, and things we know how to do (dynamic information) constitute procedural knowledge. Declarative knowledge is represented in memory by ‘ppggp that are associated with other nodes through connecting associations or lipkg' (1987:290). Schema are interconnected networks of nodes depicting complex concepts. These schema allow us to organize and understand new information. Procedural knowledge is used over and over again to solve problems and to understand and generate language. Unlike declarative knowledge which may be acquired rapidly, procedural knowledge is acquired gradually and with extensive opportunities for practice. Instructional methods for second language acquisition need to focus on communicative activities which give practice for language as a skill rather than as an object of study. The language 77 learner must procede from rule-bound declarative behavior to an automatic procedural stage. During the first or cognitive stage, the knowledge is declarative and the learner tends to be consciously aware of learning. During the second or associative stage the learner eliminates errors in the information and strengthens the skills, resulting in an autonomous stage in which the skill becomes virtually automatic and unconscious (see also McLaughlin 1987). O'Malley et a1. propose that vocabulary is declarative knowledge and predict fairly rapid attrition: ...aspects of the language that are at the first or cognitive stage of acquisition and are therefore represented by declarative knowledge would be forgotten first, whereas those aspects of the language that have become automatic or proceduralized would be retained...Vocabulary is identified with concepts, and thus can be represented as declarative knowledge (1987:303). Jenkins (1974) cautions that complex human behavior is qualitatively different than the simple behaviors often measured by experiments in the verbal-learning tradition. What memory is depends on context. The contextualist takes the uncomfortable position that a ”‘final' analysis is a myth, that analyses mean something only in terms of their utilities for some purposes" (1974:787). ‘What is remembered in a given situation depends on the physical and psychological context in which the event was experienced, the knowledge and skills that the subject brings to the context, the situation in which we ask for evidence for 78 remembering, and the relation of what the subject remembers to what the experimenter demands' (1974:793). The contextualist must work from the top down and remember that human memory is much more than simply a sum of its parts. Summary of literature on human memory The literature on human memory is vast, and a brief but accurate summary is difficult. However, the following generalizations can be made that are relevant to second language learning. 1. It is important to conduct memory research in ecologically realistic circumstances (real-life settings) as well as in the laboratory (Bahrick 1979, 1982, 1984). 2. Short-term memory holds a capacity of about seven chunks for up to thirty seconds. When repeating sentences, native speakers can form larger semantic and syntactic chunks, but non-native speakers show a typical serial position curve, as if the items in the sentence were unrelated (Meara 1980, Spitze and Fischer 1981). 3. Forgetting is exponential in form (Pimsleur 1967). 79 4. Long-term memory can last for years, but is subject to interference from new material. Bahrick (1984) writes about another stage of memory (‘permastore') which is immune to forgetting through interference. 5. Recognition and production vocabulary differ in the following ways: a) rehearsal improves recognition but not production; repeated presentation improves both scores, b) recognition usually produces higher scores, since the amount of information required for a correct response is lower, c) lower levels of learning suffice for recognition, d) with high levels of learning, production is more sensitive, and with low levels of learning, recognition is more sensitive, e) intention to learn has more effect on production than on recognition (Tversky 1973, J. Brown 1976, Wingfield and. Byrnes 1981, Eysenck 1982). 6. Beginnings and ends of words are easier to recall than middles (Brown and McNeill 1966, Kohn et al. 1987). 7. The learning situation in which the word was encoded affects the retrieval of the word (Jenkins 1974, Craik 1979, Cohen 1986, Weltens 1987). Memory is better when ego and emotions are involved (Stevick 1976). 8. Depth of analysis seems to be an important factor. (Craik and Lockhart 1972, Wingfield and Byrnes 1981). Imagery and 80 pictures can lead to a more meaningful analysis of vocabulary (Paivio 1976). 9. Word frequency has a greater effect on semantic memory, and emotionality and imagery have a greater effect on episodic memory (Murray 1986). 10. Cognitive theory predicts fairly rapid attrition of foreign language vocabulary, since it is declarative knowledge (O'Malley et al. 1987). 81 THE HYPOTHESES In this section, the hypotheses for the present research are examined in the light of the literature reviewed in Chapter 2. Several of the hypotheses are supported by the literature, but other hypotheses are more controversial. Gaps in the literature are noted--gaps which the current research seeks to begin to fill. Hypotheses about students 1. Students who study in a target-language environment retain more vocabulary than those who study in a native- language environment, even when the specific vocabulary learned is not rehearsed or used outside of class. None of the literature reviewed in this chapter deals with this question. 2. Students who have pre-course contact with Spanish will retain more vocabulary than those who have no previous contact. None of the literature reviewed in this chapter deals specifically with this question. 3. Students who have post-course contact with Spanish will retain more vocabulary than those who have no further 82 contact, even when the specific vocabulary learned is not rehearsed or used. Several studies cited in this chapter concluded that \ post-course contact with the foreign language was important for retention (Edwards 1976, 1977, Lowe 1982, Gardner 1982, Robison 1985, Clark and Jorden 1984). However, all of these studies assumed that students had the chance to hear and practice many of the items for which they were being tested, amd thus the items were reinforced through time. The present research seeks to eliminate target vocabulary with which the student has contact after the termination of the course . 4. Vocabulary is retained at different rates according to sex. Pratella (1969) concluded that sex was not an important factor in loss over the summer vacation. 5. Students who report more effort while taking a language course retain more vocabulary than those who repprt less effort. None of the literature cited in this chapter deals with this question. 6. Students who report more intrinsic motivation for the course retain more vocabulary than students who report more extrinsic motivation. 83 Kennedy (1932) reported higher retention rates for students who had a desire to continue their study of Latin. Edwards (1976, 1977) and Gardner (1985) indicate that attitudinal/motivational factors were not related to retention. This question has not yet been resolved. 7. Students who receive higher grades-for the course retain more vocabulary than students who receive lower grades. Bahrick (1984) and Godsall-Myers (1981) specifically note that higher grades were related to higher retention rates. However, Clark and Jorden (1984) did not find differences between ‘attriting' and ‘non-attriting' students according to the ease with which they originally learned Japanese. 8. Students who receive higher scores on the first retention tests continue to receive higher scores on later retention seats- The only researcher who retested the same subjects was Meara (1989), but he gave no data on the relationship of the first retention scores to the later retention scores. He suggests that the words retained from Week 2 to Week 3 are a better predictor of future scores and the final equilibrium state than words retained from Week 1 to Week 2. 84 Hypotheses about words 9. Vocabulary that is repeated more frequently in teacher classroom input is retained at higher levels than vocabulary that is repeated less frequently. Godsall-Myers (1981) concluded that attrition reflects the frequency of occurrence of L2 patterns. Murray (1986) noted that translation efficiency from English to French was influenced by word frequency. Nation (1982) believes that words should be encountered seven times in order to be remembered (and sixteen times for incidental learning). Word frequency seems to have a greater effect on semantic memory than on episodic memory, in which a single vivid encounter with a word may fix it in memory (Schouten—Van Parreren 1989:82). 10. Verbs, adjectives, and nouns are retained at different .2223. Obler (1982) has shown that in aphasia and in normal aging, nouns are harder to recall than verbs. Nation (1982) states that nouns are easier to learn than adjectives, adverbs or verbs, and Cohen (1986) makes the point that nouns may be easier to forget because they are easier to learn, since they have not passed through as much processing. The relationship of part of speech to retention merits further investigation. 85 11. Vocabulary learned earlier in the course will be retained at higher rates than vocabulary learned later in the course. Although the regression hypothesis (first-learned, last forgotten) does not appear to be true in the loss of a first language (Caramazza and Zurif 1978, Berko-Gleason 1982), several researchers report finding cases of regression in the loss of a second language (Cohen 1974 and 1975, Hansen 1980, Godsall-Myers 1981, Moorcroft and Gardner 1987, Olshtain 1983, 1986, 1989, Jordens et a1. 1989). However, more work has been done in morphology and syntax than in vocabulary. 12. Vocabulary with high emotionality or saliency ratings is retained longer than vocabulary with low ratings. Emotionality is most important for retention in episodic memory (Murray 1986). Several writers believe that emotionally loaded words are easier to remember than neutral words (Stevick 1976, Ludwig 1984, Schouten-Van Parreren 1989). None of the literature reviewed tested for effects of saliency. 86 Hypotheses about the relationship between recognition and production vocabulary 13. Recognition scores for vocabulary are higher than production scores. In general, production skills suffer more attrition than recognition skills (Scherer 1957, Edwards 1976 and 1977, Lowe 1983, Bahrick 1984, Robison 1985). Bahrick (1984) states that attrition affects smaller portions of recognition vocabulary than recall vocabulary. Several writers on human memory mention that recognition produces higher scores (Tversky 1973, J. Brown 1976, Wingfield and Byrnes 1981). 14. Both recognition and production scores for vocabulary drop quickly at first and then gradually level off. Pimsleur (1967) states that forgetting is exponential in form. Bahrick (1984a,b,c) concludes that retention rates drop for up to six years and then level off. For Weltens et a1. (1989), most loss occurred during the first two years and then leveled off. Meara's matrix model of acquisition (1989) predicts that remembering and forgetting will eventually reach an equilibrium. None of the vocabulary studies presents data from measuring the same learners several times. 87 15. The best description of recognition and production vocabulary loss is that equal quantities of words are lost over time rather than that the ratio of production to recognition vocabulary remains constant over time. Bahrick (1984) believes that the absolute amount of attrition in production and recognition vocabulary is approximately the same. Weltens et al. (1989) also believe that a fixed amount of material is lost rather than a fixed proportion, although they do not refer specifically to vocabulary. Other writers mention size of these recognition and production vocabularies in terms of proportions; for example, Melka Teicroew (1982) cites reports indicating that for beginning students, recognition vocabulary is about twice the size of production vocabulary. This question has not yet been resolved. 16. Smaller percentages of recognition vocabulary are lost than of production vocabulary. If recognition vocabulary is larger than production vocabulary and equal amounts are lost from both, it follows that the percentage loss from recognition will be smaller (see Bahrick 1984c). 17. Vocabulary taught and tested for both recognition and production is retained at higher rates than vocabulary \ taught and tested for recognition only. 88 Brown and Mc Neill (1966) believe that the need to produce a word promotes storage of the word in full. Tversky (1973) states that scores may be dependent upon the subjects' knowing which mode they will be tested in. None 'of the literature reviewed in this chapter specifically tested these two different modes for teaching and learning vocabulary. 18. Ability to produce a word initially, rather than ability to recognize a word, is the best predictor of both recognition and production later on. None of the literature reviewed in Chapter 2 specifically tested for this hypothesis. In conclusion, most literature on foreign language attrition is very recent, and the study of foreign language vocabulary is only beginning to attract the attention it deserves. Both of these areas are related to the study of human memory, including ideas on levels of processing and cognitive theory. There are still many questions about the role of vocabulary in language attrition, questions which the present research can play a part in answering. CHAPTER 3 METHODOLOGY The purpose of this study is to investigate the retention of foreign language vocabulary in college-level students over varying intervals of time. Several characteristics of the students and the learning environment are studied, as well as characteristics of the words themselves. Special attention is paid to the relationship of recognition vocabulary to production vocabulary over the retention period. A brief overview of the methodology has already been given in Chapter 1, and the present chapter will describe in greater detail the 1) subjects, 2) materials and procedures, and 3) methods of analysis. In this study a balance was sought between degree of control (in order to increase internal validity) and degree of approximation to the real-world classroom (in order to increase external validity and generalizability to other situations). If similar patterns can be found in differing learners and differing situations, then the results may be applicable to other learners in other real-world situations. In the previous chapter it was noted that writers such as Bahrick (1979) have called for research in ecologically realistic circumstances. 89 90 SUBJECTS The fifteen subjects were students enrolled in two intensive introductory Spanish courses. The first course was taught at Calvin College in Grand Rapids, Michigan, from May 31 to June 22, 1988, and the second course, composed of different students, was taught in Merida in the Yucatan Peninsula of Mexico from June 26 to July 18 of the same summer. The nine students who enrolled in the Calvin course did so for very different reasons, as indicated by a questionnaire that they were given the first day of class. ' Six students were in the class because of graduation requirements at Calvin, although two of these six also indicated a special interest in the Spanish language. A seventh student was hoping to go to Spain (though his plans changed midway through the course), an eighth student believed that Spanish would help her in her nursing career, and the ninth student (a 16-year old high school student) mentioned his interest in Latin America. Except for this l6-year-old student, the subjects in this class ranged in age from 19 to 22. These nine students also varied widely in their backgrounds of foreign language study. Latin and French were two of the languages that some of these students had studied in high school. Three of the students had already attempted Spanish at Calvin College, had dropped out because 91 of low grades, and were repeating the course. Final grades in the summer course at Calvin ranged from A to D. The six students who enrolled in the course in Merida were involved in the STS (Summer Training Session in MisSions) program in Mexico, conducted for twenty years by the Reformed Bible College in Grand Rapids and presently associated with IDEA ministries in Grand Rapids. In terms of motivation, this was a much more homogeneous group. All of the students mentioned an interest in missions, as well as an interest in Latin America and its people. However, their foreign language background was widely varied: two had never studied a foreign language while others had studied German, French, or Dutch. Only one subject had studied Spanish previously, and for only one semester. One subject had already graduated from college and was teaching in a school in south Texas with many Spanish-speaking children. The ages of these subjects ranged from 20 to 25. Final grades in this class were high, ranging from A to 8-. Since this was a retention study, follow-up questionnaires asked the students to evaluate the course and their effort in it and to detail further contact with Spanish, either formally in the classroom or informally in the culture. Two of the students in the Calvin class had no further contact with Spanish, although most went on to take the second semester of first year (another intensive summer course) and then on to second-year Spanish during the regular school year. The students in the Merida class lived 92 with Mexican families in pairs while taking the class and then moved on in pairs to 3-week field assignments in Mexican or Central American villages and heard and used varying degrees of Spanish, depending on the principal language spoken in the village, whether Spanish or an indigenous language (such as Maya or Tzeltal). Five of these six students returned to the U.S. or Canada at the end of August and took no further classes in Spanish, but one remained in Mexico City to continue her study of Spanish and eventually went to work in Honduras. Although it would have been possible to eliminate this student from the study, in the end she was included, in view of the fact that the attrition of the 100 target words showed the same patterns in her case as in the other students. . Of the total number, eight subjects were male and seven were female. Tables 2 and 3 give a summary of the characteristics of these subjects. The data file in Table 40 in Appendix B indicates how these characteristics were coded for the statistical analysis and also includes the scores for the subjects on each of the recognition and production tests. Although not all of the subjects responded to each of the four retestings, all participated in the final retest; thus, there were no drop-outs. 93 TABLE 2 SUBJECT CHARACTERISTICS LOCATION COURSE SUBJECT SEX AGE OF CLASS MOTIVATION EFFORT GRADE 01 F 20 Mexico interest in missions much 8+ 02 M 21 Mexico interest in missions, some A Latin America 03 F 24 Mexico interest in missions, much A Latin America 04 F 21 Mexico interest in missions, much 8- Latin America 05 F 25 Mexico interest in missions, much A Latin America 06 M 24 Mexico interest in missions, some 8+ Latin America 07 M 22 U.S. graduation requirement, some D career in communica- tions 08 M 21 U.S. graduation requirement much 8- 09 M 20 U.S. graduation requirement much 8- 10 M 19 U.S. graduation requirement, some C interest in Latin America 11 F 20 U.S. graduation requirement much C 12 F 22 U.S. graduation requirement some C 13 M 22 U.S. plans for travel to some 8+ Spain, though later the plans fell through 14 F 20 U.S. career in nursing much A- 15 M 16 U.S. interest in Latin much A America 94 TABLE 3 ' PREVIOUS AND LATER CONTACT SUBJECT CHARACTERISTICS: Subject Previous langppge study Later contact with Spanish 01 02 03 O4 05 06 07 08 O9 10 11 12 13 14 15 2 years junior high French 2 years high school German, 1 semester college German, 2 years college Greek 1 semester college Spanish, 2 years college Dutch, 10-week visit to Peru 2 years high school Latin short trips to Mexico, teacher in bilingual school none 1 semester high school Spanish 1 year college Spanish semester college Spanish years high school Latin years high school Latin 1 2 4 2 years high school Latin, 1 year college Spanish 1 3 4 P year college Latin years high school French hours Spanish (Total hysical Response) some in Mexico, though stayed in village where Tzeltal was spoken; no further formal study 4 weeks with Spanish-speaking family; no further formal study 3 weeks in Spanish-speaking village; no further formal study further language study and stay in Honduras until present 3 weeks in Spanish-speaking village; no further formal study; teacher in bilingual school 3 weeks in Mexico, though stayed in village where Maya was spoken 3 semesters college Spanish 3 semesters college Spanish 3 semesters college Spanish 2 semesters college Spanish none 2 semesters college Spanish none 1 semester college Spanish 3rd year high school Spanish; talked with migrant workers 95 As a control to see what kind of scores would result from a class of Spanish students who had not studied this specific material, the same recognition and production test was given to a class of 24 students at Calvin College. These 102-level students had been taught Spanish by an entirely different approach with a different text (Communicating in Spanish by Lamadrid el al. 1984) and had encountered only two of the target words (pidp and cartel) in their course work. Their mean score on recognition was 1.6 (out of 100 words) and their mean score on production was .2 (also out of 100), scores far below those of the students in the study, even 18 months after the course had ended. MATERIALS AND PROCEDURES Both the course taught in Grand Rapids and the course taught in Merida used exactly the same methods, text, and additional materials. The courses were taught using the ‘Natural Approach' with the text Dos mundos (1986) by Terrell et al. Classes met for three hours each morning for a period of seventeen days in Grand Rapids and for sixteen days in Mexico. Both classes took the same period of time (twenty hours) to cover Pasos A-E (the period intensively studied in this investigation), although the Grand Rapids class was able to go through Capitulo 5 in the text because 96 of the extra day, while the Merida class completed only the material through Capitulo 4. The researcher was the instructor for both classes and used the same lesson plans for both classes, thus insuring uniformity in classroom procedure. The language of instruction for both classes was Spanish. On the first day of class all students were given a questionnaire asking for name, age, class rank, address, previous contact with Spanish or other foreign languages, motivation for studying Spanish, and what the student hoped to learn in the class. This questionnaire was handed in to the instructor and the form is included in Appendix A. Students also filled out a consent form which asked for permission to use classroom data for a study on vocabulary retention, although the exact nature of the study was not specified. This consent form is also included in Appendix A. Students gave these consent forms to a third party, and the instructor did not see them until after she had given the final course grades. All of the students gave their consent as it turned out, and thus the researcher was able to include the data from all fifteen students. New material was first presented in class orally by the teacher through teacher talk with pictures, objects, and commands. Students were asked to indicate comprehension with actions, by answering with s; or pp, or by answering with single words or short sentences. After a new group of vocabulary words was introduced, the instructor wrote the 97 new words on the board and students copied them in their notebooks, often including their own English translation. Students were asked to study the text outside of class after the presentation of new material and to fill out their workbooks and worksheets. Homework assignments (including grammatical explanations) were discussed briefly in class the following day and sometimes collected. By the end of the first twenty hours, students were participating in widely-varying communicative activities and writing short compositions. The material in Dos mundos is organized around language functions and communicative situations. Paso A contains classroom commands, names, adjectives to describe people, colors, clothing, numbers, parts of the body, and expressions for greeting and leave-taking. Paso B teaches informal and formal distinctions, classroom objects, more commands, and more adjectives for describing people. Paso C includes family, expressions for possession, telling time and telling age, and more numbers. Paso D teaches some weather expressions, expressions for origin and nationality, class subjects, and expressions for location. Paso E contains more numbers, addresses and phone numbers, dates and birthdays, and expressions for prices. All together these five pasos teach a total of about 650 new words or expressions. After finishing these preliminary pasos, the students begin the regular capitulos, which are set up in the same basic format as the preliminary pasos, but longer. 98 In addition to these 650 new words, the instructor chose another 100 words which formed the core of the investigation on retention. These 100 words were chosen because of their concreteness and their ability to be represented by pictures (all were nouns, verbs, or adjectives), their possibility of inclusion in the categories already being taught (in order to form networks of meanings and to make them less obvious), and their non-appearance in later chapters in Dos mundos or the second-year book used at Calvin, Charlemos by Jarest and Robinson (1986). Thus these 100 words would not be reviewed after the first twenty hours of class and could be used to measure the attrition rate. On subsequent retention tests students were asked to mark words with which they had had contact after the course, and when any of these 100 words were mentioned, they were eliminated from that student's data. Table 4 gives a list of these 100 words, the paso in which they were taught, their part of speech, their frequency in teacher talk, and whether they were originally taught and tested for production and recognition or for recognition only. 99 TABLE 4 WORD CHARACTERISTICS RTE- QUENCY WORD ENGLISH PART OF IN TAUGHT AND TESTED NO. WORD TRANSLATION PASO SPEECH INPUT FOR REC. AND PROD.? 001 zpellizque! pinch!‘ A verb yes 002 :ande de tiptoe! A verb yes puntillas! 003 :rece! pray! A verb yes 004 :pliegue! fold! A verb low yes 005 3tache! cross out! A verb low yes 006 (amarre! tie! A verb low yes 007 cojo lame A adj. low yes 008 tuerto one-eyed A adj. medium yes 009 manso gentle A adj. high yes 010 hurafio timid A adj. high yes 011 ocioso lazy A adj. high yes 012 agudo sharp,smart A adj. high yes 013 torpe stupid A adj. low yes 014 delantal apron A noun medium yes 015 faja belt A noun medium yes 016 cremallera zipper A noun medium yes 017 aretes earrings A noun medium yes 018 tacén heel A noun low yes 019 chaleco vest A noun medium yes 020 alhajas jewels A noun medium yes 021 mefiique little A noun yes finger 022 tez face A noun yes 023 pufio fist A noun yes 024 pulgar thumb A noun yes 025 panza stomach A noun yes 026 barbilla chin A noun yes 027 quijada jaw A noun yes 028 sien temple A noun yes 029 hocico snout A noun medium yes 030 espinazo back (of A noun low yes animal) 031 garra claw A noun low yes 032 pescuezo neck (of A noun medium yes animal) ‘ 033 mariposa butterfly A noun low yes 034 ciervo deer A noun low yes 035 venado deer A noun low yes 036 ballena whale A noun low yes 037 tiburdn shark A noun medium yes 038 jexprima! squeeze, 8 verb very low yes wring out! 039 jguifie! wink! 8 verb medium yes 040 1ronque! snore! 8 verb low yes 100 TABLE 4 (cont'd) O41 géilbe! whistle! 8 verb medium yes 042 gbostece! yawn! 8 verb low yes 043 descalzo barefoot 8 adj. low yes 044 calvo bald 8 adj. very low yes 045 cano white-haired 8 adj. very low yes 046 mohoso moldy 8 adj. low yes 047 harapiento ragged 8 adj. very low yes 048 arrugado wrinkled 8 adj. very low yes 049 empapado soaked 8 adj. low yes 050 maceta flower pot 8 noun low yes 051 tabla board 8 noun low yes 052 colcha blanket 8 noun low yes 053 cuenco bowl 8 noun medium yes 054 rétulo sign 8 noun low yes 055 cartel poster 8 noun medium yes 056 pasillo hallway 8 noun low yes 057 huella footprint 8 noun medium yes 058 fango mud 8 noun low yes 059 ala wing 8 noun low yes 060 rams branch 8 noun low yes 061 mirlo blackbird 8 noun. high yes 062 nido nest 8 noun low yes 063 antepasado ancestor C noun low yes 064 vastago descendant C noun low yes 065 bisabuelo great-grand- C noun medium yes father 066 tatarabuelo great—great- C noun very low yes grandfather 067 padrastro stepfather C noun very low yes 068 madrastra stepmother C noun very low yes 069 padrino godfather C noun very low yes 070 madrina godmother C noun very low yes 071 mellizo twin C noun low yes 072 légrima tear C noun yes 073 beca scholarship C noun yes 074 cesto basket C noun very low yes 075 incendio fire C noun yes 076 llama flame C noun yes 077 cerilla match C noun yes 078 aguanieve sleet D noun no 079 estanque pool D noun medium no 080 dehesa meadow D noun low no 081 loma hill D noun low no 082 cerro hill D noun low no 083 vereda path D noun very low no 084 sendero path D noun -low no 085 agujero hole D noun low no 086 carpa tent D noun low no 087 aldea village D noun very low no 088 verja window D noun very low no grating 101 TABLE 4 (cont'd) 089 pefia rock D noun very low no 090 bolivar monetary unit E noun very low no (Venezuela) 091 rastrillo rake E noun very low no 092 enano dwarf E noun very low no 093 azulejo tile 8 noun low no 094 navaja blade 8 noun low no 095 cascote garbage E noun medium no 096 hilo thread 8 noun low no 097 semilla seed 8 noun very low no 098 frasco bottle 8 noun medium no 099 globo balloon E noun medium no 100 aguja needle 8 noun low no KEY FOR FREQUENCY IN CLASSROOM INPUT very low: 1-4 repetitions 5-9 repetitions 10-19 repetitions low: medium: high: 20+ repetitions 102 These 100 words were taught just like the other vocabulary words in that they were introduced through teacher talk and the use of commands, pictures, and realia. Like the other vocabulary words, they were written on the board and copied by students into vocabulary notebooks, and since students were asked to study vocabulary from their notebooks, it was not readily apparent that these words were not included in the text. Since students received additional practice with the regular vocabulary words in their workbooks, extra worksheets were assigned to give students written practice with these 100 words, which were mixed in with the regular vocabulary words. Tests were given after Paso A, Pasos B-C, and Pasos D-E. The greater part of each test came directly from the testing manual provided with Dos mundos, but items testing the extra 100 vocabulary words were mixed in with the other items, and students were called in individually for oral interviews after each of these tests. In addition, two quizzes (not in the Dos mundos testing program) were written to test production of vocabulary during the first 20 hours of the course. The two production quizzes and sample items from the tests are included in Appendix A. The tests provided with Dos mundos are mainly multiple choice tests of vocabulary and grammar, thus asking the student to recognize but not to produce the words and forms. Because of the emphasis on recognition, the authors found it possible to include the large number of 650 vocabulary words 103 in the first five pasos of the text. Suggestions are also given in the testing manual for compositions and oral interviews in which the students are asked to produce words and forms, though the students always have a choice about what to say. When the researcher added additional parts to the tests and wrote the two extra quizzes, she insured that each of the 100 additional words in Pasos A-E would be tested at least once for recognition during the first twenty hours. After students had been tested for recognition on a test or quiz, the next quiz or test would ask them for the production of this same vocabulary, and students were asked to study with production in mind. Because production tests lagged behind the recognition tests, students were not tested for production of the extra vocabulary in Pasos D and E and thus did not study for the production of this vocabulary either. This difference in testing during the first twenty hours shows up in Table 4 in the column headed ‘Taught and tested for recognition 222 production', where some words were learned for both recognition and production and other words for recognition only. The data collected during the first twenty hours of class during these tests and quizzes provided the information for setting a baseline for each student: whether the student at one point in time could recognize and produce a word. After the first twenty hours of class, the students did not see or hear these 100 words again in class, and they were told to study only the vocabulary lists in 104 their books (which did not contain the 100 target words) for the final exam. However, at the end of the course, the students unexpectedly received their first retention test. The first part of the retention test consisted of pictures of the 100 words followed by blanks for writing the corresponding words. The pictures had been used many times in the past, but the teacher was also available to identify in English any picture that a student could not interpret. Pictures of other familiar words that the students had learned were also included so that it was impossible for the subjects to determine exactly which words the researcher was interested in. When students had completed as much of this part as possible, they turned it in to the instructor and took the second part, which was identical to the first part, except that a long list of Spanish words was provided for the students to choose from. Students needed to match the letter of the word to the correct picture, thus performing a recognition test. All 100 target words were listed and pictured, but several other pictures and words were also included as distractors. Although the researcher recorded the amount of time each student took to complete each part, no time limit was imposed, since the goal was to learn if the student had retained each word and not to force a quick guess (for a precedent, see Bahrick 1984b:3). On the matching test students were also asked to indicate when they were guessing by starring the word, thus indicating a lack of confidence in their decision. Although this information 105 about confidence level was not analyzed statistically, it will be mentioned again in Chapter 4. Pictures, instead of English translations, were used on the retention test because of the nature of the classroom experience in which pictures, realia, and commands had been used to teach the words. Students were also familiar with these pictures because of the worksheets and tests they had completed earlier. Also, the literature reviewed in Chapter 2 of this work indicates that words are more easily learned and retained when accompanied by pictures (Kellogg and Howe 1971, Champagnol 1972, Paivio 1976:120). The research design in which each subject is tested first for recall and then immediately for recognition of the same vocabulary is mentioned in Chapter 2 of this dissertation. J. Brown (1976:34) has stated that ‘This greatly increases the precision of the experiment since recall and recognition are then tested on the same subjects, on the same material and on the same occasions'. The second retention test was given to the subjects approximately three-and-a-half weeks later. The form for the test was the same: first a production test and then a recognition test over the same words. The test in Grand Rapids was administered by a different instructor at the end of Spanish 102 (second semester of first year), and the test in Mexico was administered by the researcher in Mexico City when the students met again after their field assignments in Mexican and Central American villages. 106 The third retention test was taken by the subjects in April and May of 1989, approximately nine months after the previous test. The format was the same, except that several different words and pictures had been mixed in with the 100 target words. The students were widely dispersed geographically, and for this reason the tests were mailed to the students with explicit instructions about how to complete them. By this time the subjects were quite familiar with the protocol for taking the tests, but in case a student did not recognize a picture, a separate sheet of English words for the pictures was included. The fourth and final retention test was given in November and December of 1989, approximately eight months after the previous test. The.format of the test was the same as in the spring, but this time the researcher was able to personally test twelve of the fifteen subjects. She still needed to rely on mailed forms for the three subjects living in Texas, Minnesota, and Illinois. A copy of the retention test (both production and recognition parts) is included in Appendix A. Although testing all 100 words (along with other words) in each of the four retention tests leaves open the possibility that students retain more because of the tests, Berko Gleason (1982:22) sets a precedent by suggesting that longitudinal studies be conducted using the same measure. Because of the small number of subjects in this study, the researcher felt that more accurate data could be obtained by 107 using approximately the same test each time. However, the possibility remains that some subjects obtained higher scores because of the effect of previous tests, in spite of the relatively long periods of time between tests. All retention tests were scored using the same criteria. On the recognition test an item was either correct or not correct, and one point was awarded for a correct answer. On the production test it was felt that partial answers were important in tracing the attrition pattern, so partial credit was given for some answers. The precedent for giving partial credit is found in Bahrick (l984b:5), and Ginsberg (1986:23,183). On the production test, the researcher gave one point for a totally correct word or for a word that was correct except for an accent mark or for a spelling that would not change the pronunciation of the word (at least in Latin American pronunciation). One-half point was given for a word that could not meet these criteria, but that had at least half of the correct.letters. For example, if the correct word was azulejo (tile), asulejo received one point, azuljes received one-half point, and 3525 received no credit. Raw scores were converted to percentages, thus allowing the researcher to obtain a score even after having deleted those words for which the subject indicated subsequent contact. If a subject indicated that he had had contact with a specific word but still got it wrong on the test, the word was not deleted from the data. 108 In addition to the regular tests and quizzes during the first twenty hours of the course and the later retention tests, the researcher obtained data using other methods. During class a tape recorder was left on so that the researcher could record her input to the students. These recordings were used later to determine the number of times certain words were repeated in the input. The tape recorder was left on even after completion of the first twenty hours of the course to minimize clues about the research design. Written compositions collected by the teacher were scrutinized for any of the 100 target words that the subjects had used spontaneously. Also oral interviews were conducted at the end of 101 and after the final retention 1 tests. During both interviews subjects were asked to describe a picture that contained several of the target words (see Figure 22, Appendix A), and they were also asked to do some introspection about their vocabulary learning and retention. The precedent for think-aloud procedures is found in Hosenfeld (1977), Cohen and Hosenfeld (1981), and in Cohen (1984). Students were asked how they studied vocabulary, if they mentally translated to English, and how they remembered words. After the final retention test they were asked which devices they used to remember specific words. The 100 target words were also rated for emotionality and saliency to see if there was any relationship between these qualities and retention. Brown and Dre (1969) 109 published a list of 650 English words with ratings on a seven-point scale for goodness, pleasantness, emotionality, concreteness, and associative difficulty. This same idea was used to create a five—point scale in the present study for rating words on emotionality and saliency. Classes of Spanish students who were not involved in the present study were asked to rate the words. The rating for emotionality was done by a class of 30 students with the target words in English, since the target words in Spanish were unknown to the students. The following instructions come from Brown and Ure (1969:234): ‘Decide just how strongly emotional the meaning is. Notice that the kin of emotion is not relevant; only the intensity of emotion matters.’ A rating of 1 indicated high intensity and a rating of 5 indicated low intensity. A different class of 26 students rated these same English words on saliency, with a definition of saliency coming from Webster's dictionary. The specific instructions read: -‘Decide just how salient (noticeable, conspicuous) the meanings of these words are. How easy would they be to remember if you had to learn them in Spanish?’ Students again rated these words from 1 to 5. The third class of 25 students received a list of the Spanish words with the English translations in parentheses. They were again asked to rate the words on saliency, but this time the instructions were slightly different: ‘Decide just how salient (noticeable, conspicuous) the forms of the 110 Spanish words are. How easy would they be to remember if you had to learn them?’ Ratings again were from 1 to 5. The results of the one rating for emotionality and the two ratings for saliency are found in Table 41 in Appendix 8, along with the numbers for recoding words according to frequency in input, paso in which they were learned, part of speech, and whether they were taught and tested for both recognition and production. Table 41 also contains the percentage scores for each of the words on each of the recognition and production tests over time. ANALYSI S As can be seen from Appendix 8 (Tables 40 and 41), the variables in this study can be organized in two ways, by student and by word. When the data are organized by student, 3 equals 15, and when the data are organized by word, 3 is equal to 100. Both data files deal with recognition and production, and this distinction can be set apart as a third area of analysis. The research hypotheses are also organized around these three areas (dealing with students, with words, and with recognition/production), and these three areas thus form the basis for the organization of results in Chapter 4. The research hypotheses are listed below and the method of investigating each one is described. 111 The recognition and production tests were scored by hand, and data and control files were created using SPSS (Statistical Package for the Social Sciences), available at Calvin College. Where statistical analysis (such as a T- test, One-way analysis of variance, or Anova) seems appropriate, the hypotheses are restated in null form and a probability level of .05 or lower is cause for rejecting a null hypothesis. T-tests are independent rather than dependent, since the subjects are in only one group at a time. One-tailed T-tests are used when the hypotheses are directional, and two-tailed T-tests are used when the hypotheses are nondirectional. For some hypotheses, a correlation or a graph seems to be the most effective way of studying the problem. In addition, individual case studies make important points that are missed when all the data are lumped together for statistical purposes. Thus, in Chapter 4 on the findings, the questions are looked at from varying perspectives to get a more rounded and more complete picture of the data. One must also remember that patterns are important. If one test is significant at the .05 level, but other similar ones are not, the overall pattern clearly overrides the individual test. This is especially important in the cases where a series of statistical tests are used to analyze recognition and production tests over time. A conservative and careful approach to results should be taken, and results should not be overinterpreted. 112 Hypotheses about students independent variables: classroom location (Mexico or U.S.), pre- and postécourse contact with Spanish, sex, effort, motivation, final grade in course dependent variables: ten scores per student on a series of recognition and production tests over time I other data: initial and final questionnaires (written), interviews at end of 101 and at end of 18-month retention period Hypothesis 1: Students who study in a target-language environment retain more vocabulary than those who study in a native-language environment, even when the specific vocabulary is not rehearsed or used. The corresponding null hypothesis is that there is no difference between the two groups. In order to test this null hypothesis, an independent one-tailed T-test was used to test for significant differences between the Mexico group and the U.S. group on their recognition and production scores at the different points in time. Hypothesis 2: Students who have pre—course contact with Spanish (either informally in the culture or formally in the classroom) will retain more vocabulary than those who have 113 no previous contact. The corresponding null hypothesis is that there is no difference between the students who have pre-course contact and the students who have no pre-course contact. In order to test this null hypothesis, an independent one-tailed T-test was used to test for significant differences between these two groups based on their test scores. Hypothesis 3: Students who have post-course contact with Spanish (either informally in the culture or formally in the classroom) will retain more vocabulary than those who have no further contact, even when the specific vocabulary learned is not rehearsed or used. The corresponding null hypothesis is that there is no difference between the students who do and do not have post-course contact. For this analysis the students were divided into four small groups depending on type of post-course contact and a One- way analysis of variance was run on the test scores. Hypothesis 4: Vocabulary is retained at different rates according to sex. The corresponding null hypothesis is that there is no difference between the sexes in retention of vocabulary. Since this hypothesis is nondirectional, an independent two-tailed T-test was used to test for significant differences in the test scores between groups based on sex. 114 Hypothesis 5: Students who report more effort while taking a language course retain more vocabulary than those who report less effort. The corresponding null hypothesis is that there is no difference based on effort. In order to test the null hypothesis, the students were divided into two groups based on the amount of effort they reported and an independent one-tailed T—test was run on scores of these two groups. Hypothesis 6: Students who report more intrinsic motivation for the course retain more vocabulary than those who report more extrinsic motivation. The corresponding null hypothesis is that there is no difference based on motivation. In order to test the null hypothesis, the students were divided into three groups based on the type of motivation they reported (1 for intrinsic motivation, 2 for mixed motivation, and 3 for extrinsic motivation), and a One-way analysis of variance was performed on test scores to look for significant differences between the groups. An independent one-tailed T-test was also performed to look for significant differences between the most extreme groups (groups 1 and 3). In order to examine the role of effort and motivation from a slightly different perspective, the rating for effort and the rating for motivation were added for each student to obtain ordinal data, and a nonparametric correlation was performed with test scores to determine Spearman's rank correlation coefficient. 115 Hypothesis 7: Students who receive higher grades for the course retain more vocabulary than students who receive lower grades. The corresponding null hypothesis is that there is no difference in scores based on final course grade. In order to test the null hypothesis, the students were divided into two groups (those with grades ranging from A to 8+ and those with grades ranging from 8- to D) and an independent one-tailed T-test was performed on the scores in order to test for differences between the groups. The relationship between the two groups was also graphed. Hypothesis 8: Students who receive higher scores on the first retention tests continue to receive higher scores on later retention tests. The corresponding null hypothesis is that there is no systematic relationship between retention scores on earlier tests and scores on later tests. This hypothesis was examined by performing a Pearson product- moment correlation with test scores and looking for strong correlations among earlier and later recognition and production scores. In addition to performing these statistical tests, graphs of the scores for each of the fifteen students were constructed, as well as a composite graph. Individual cases were discussed in relation to the composite graph. 116 Hypotheses about words independent variables: frequency of words in classroom input, part of speech, words taught earlier or later, emotionality and saliency of words dependent variables: ten scores for each of 100 words other data: interviews by students at end of 101 and at end of 18-month retention period Hypothesis 9: Vocabulary that is repeated more frequently in teacher classroom input is retained at higher levels than vocabulary that is repeated less frequently. The corresponding null hypothesis is that there is no difference in scores between vocabulary that is repeated more frequently and vocabulary that is repeated less frequently in classroom input. In order to test the null hypothesis, the words were divided into two groups according to frequency in input, and an independent one-tailed T-test was run on the scores to test for significant differences. In addition, since the frequencies could also be represented by ordinal data, a nonparametric correlation was run between frequencies recoded on a scale of 2 to 8 and test scores. Hypothesis 10: Verbs, adjectives, and nouns are retained at different rates. The corresponding null hypothesis is that 117 there is no difference in the retention rates of verbs, adjectives, and nouns. In order to test the null hypothesis, a One-way analysis of variance was conducted on all the words in Pasos A-E by part of speech (with the Tukey option to tell where any significant differences occurred). Next, the One-way analysis of variance was repeated on just the words in Pasos A and 8, since these were the only ppppp in which all three parts of speech were taught. A graph was also constructed to illustrate these relationships. In order to study the combined effects of frequency in input, part of speech, and whether a word was taught and tested for both recognition and production or for recognition only, an Anova was run on the scores by these three factors. The size of the F ratio (the ratio of ‘between group variance' to ‘within group variance') is given, as well as the probability of obtaining these results only by chance. The Anova also gives an idea of the relative contribution of each of these factors to the SCOIGS . Hypothesis 11: Vocabulary learned earlier will be retained at higher rates. The corresponding null hypothesis is that there is no difference in scores between vocabulary learned earlier and vocabulary learned later. This null hypothesis was tested by doing a One-way analysis of variance by paso and including the Tukey option in order to discover where significant differences occurred. 118 Hypothesis 12: Vocabulary with high emotionality or saliency ratings is retained longer than vocabulary with low ratings. The corresponding null hypothesis is that there is no difference in scores between vocabulary with high emotionality or saliency ratings and vocabulary with low ratings. For a preliminary analysis, Pearson product-moment correlations were determined between the scores and the one rating for emotionality and each of the two ratings for saliency. The Pearson correlation was chosen because all of these data were interval scaled. The rating for emotionality and the rating for the saliency of the words in English showed practically no correlation with the scores, but the rating for saliency of the forms of the Spanish words showed a weak correlation with test scores. Based on this preliminary analysis, the words were divided into groups of high and low saliency according to the third scale, and an independent one-tailed T-test was run on the scores to test for significant differences between the two groups. The 21 best-retained words and 24 worst-retained words also formed two separate groups, and an independent one-tailed T—test was performed on their saliency ratings in order to look for significant differences between these two groups. As mentioned in connection with Hypothesis 12 above, the best-retained words and the worst-retained words were 119 identified (according to their scores on the final retention tests). These words were further examined by summarizing comments made about them by the students after their final retention tests. Also a brief discussion of networks of meanings is included. Hypotheses about the relationship between recognition and production vocabulary independent variable: whether vocabulary was taught and tested for both recognition and production or for recognition only dependent variables: ten scores for each of 15 students and for each of 100 words other data: written answers on test forms for each of 15 students Hypothesis 13: Recognition scores for vocabulary are higher than production scores. Hypothesis 14: Both recognition and production scores for vocabulary drop quickly at first and then gradually level off. 120 Hypothesis 15: The best description of recognition and production vocabulary loss is that equal quantities of words are lost over time rather than that the ratio of production to recognition vocabulary remains constant over time. Hypothesis l6: Smaller percentages of recognition vocabulary are lost than of production vocabulary. Hypotheses 13-16 are best examined by constructing and studying a graph of the means for recognition and production vocabulary over time. Hypothesis l7: Vocabulary taught and tested for both recognition and production is retained at a higher rate than vocabulary taught and tested for recognition only. The corresponding null hypothesis is that there is no difference in the scores of vocabulary taught and tested in the two different ways. The null hypothesis was tested by running an independent one-tailed T-test for a significant difference in the scores between words from Pasos A-C (words taught and tested for both recognition and production) and words from Pasos D-E (words taught and tested for recognition only). The Anova test included under the hypotheses about words is also reexamined, and another Anova is run to test the combined effects of saliency and whether a word is taught and tested for both recognition and production. 121 Hypothesis 18: Ability to produce a word initially, rather than ability to recognize a word, is the best predictor of both recognition and production later on. This hypothesis was examined from three different perspectives. In the first test a null hypothesis was formulated stating that words that are recognized and produced at high levels initially will not have higher scores on later retention tests than words that are initially recognized at high levels but produced at lower levels. Words were identified that would fit into these two initial groups, and an independent one-tailed T-test was run on their scores for later retention tests. The second perspective included a Pearson product-moment correlation for test scores to look for strong correlations between recognition and production scores. The third perspective included a multivariate analysis run at Michigan State University to determine the best predictors at Time 1 for retention at Time 4. Multiple correlation coefficients are also given. Finally the data from one student are examined closely to follow the attrition process in detail. Interesting features such as the process of forgetting, confidence level, repetition of errors, and the ‘tip-of-the-tongue' phenomenon are briefly discussed. CHAPTER 4 ANALYSIS AND FINDINGS INTRODUCTION This chapter presents the findings of the research described in the previous chapter. Results are shown for statistical tests, but individual cases are also described in order to give a more complete perspective on the data. The chapter is divided into three sections based on the division of the research hypotheses into three groups: analysis of the students, analysis of the words, and analysis of recognition and production of vocabulary words. The conclusions for the hypotheses are given within each section, along with some discussion. An over-all perspective on the meaning of the results is saved for the summary and conclusion in Chapter 5. ANALYSIS BY STUDENT Although the students were grouped in several different ways in order to study the effect of various independent variables on retention rates over time, few important overall patterns emerged. The small sample size of only 122 123 fifteen students may have impeded the emergence of more overall patterns, and indeed differences between groups would need to have been quite large in order to reveal statistical significance at the .05 level. Among the fifteen students involved in this study, there was great variation in total amount of material retained, though some similarities in the pattern of forgetting were observed. The statistical results for the analysis by student are given in this section, but it seems that the most interesting results come from studying individual cases. Statistical study First the students were separated into the two groups representing the two different classroom locations, one in Mexico and the other in the United States. Since the researcher eliminated words with which students had contact outside the classroom, she did not expect to see large differences between the groups. However, the fact that there was no difference between the two groups over time merits further discussion. It seems that in the group in the United States there were more students who had difficulties with the class, but there were also some very good students who offset the lower scores (notice in Table 5 that the standard deviation of the U.S. group is higher than that of the Mexico group). Also it seems that the students in the United States had more time to concentrate on 124 specific course content, while the students in Merida living with Mexican families had a broader communicative goal in mind than simply studying specific course vocabulary. Students in Merida also reported that it was hard to find a time and place for concentrated study. In Table 5 note that 5 stands for a recognition test and P stands for a production test. The number 0 represents the original baseline tests during the first twenty hours of class, and the number 1 represents the first retention tests given at the end of the three-and-a-half week course. Time 2 is approximately a month after Time 1, Time 3 is approximately 9 months after the previous test, and the final test at Time 4 is another eight months later (see testing schedule in Table 1 on page 20). Thus the ten different tests for recognition and production are labeled R0, P0, R1, P1, R2, P2, R3, P3, R4, and P4. 125 TABLE 5 T-TESTS FOR saoups IN MEXICO AND U.S. TEST N MEAN 55 TLVALUE* 1-TAfE“P§08“'“‘ RE‘EEEIEO 6 80.3 III;5 .11 .46 US 9 79.7 14.1 PO MEXICO 6’ 56.0 15.4 -.19 .43 US 9 58.2 25.3 RI MEXICO 6 69.5 ’18.3 .40 .35 US 8 65.0 22.4 PI MEXICO 6 32.0 21.3 -.81 .22 US 8 42.3 24.8 ' RT'EEEon *6 56.3 25.5* -.30* .38 US 7 60.3 21.9 P2 MEXICO 6 26.7 17.0 -.46 .33 us 7 30.9 22.6 R3 MEXICO 5 34.6 25.2 -.37 .18 US 7 40.4 27.5 P3 MEXICO 5 7.4 7.3 -.94 .37 us. 7 12.4 10.2 RZ‘EEEICO 6 29.5 21.3 -.02 .49 US 9 29.8 24.2 PI‘EEETEO* 6 7.0* 6.8 -.59 .28 US l 9 9.8 10.0 Judging from the mean scores on the later tests and their proportionately high standard deviations, one can assume that these scores are positively skewed. Such a distribution results from several of the students scoring near 0 on these tests; they have ‘bottomed out' and show little measurable retention on these tests. Similar T-tests were conducted to see if pre-course contact with Spanish was significant (see Appendix B, Table 43). The first group consisted of seven students who had had some previous contact with Spanish, and the second group consisted of eight students who had not. Although the means for the first group were slightly higher on each test, there were no significant differences on any of the tests. To see if post-course contact was significant, students were 126 divided into four small groups: those who had later contact in both a Spanish course and in the culture (2 students), those who had later contact only through the culture (5 students), those who had later contact only in the classroom (6 students), and those who had no further contact with Spanish (2 students). A One—way analysis of variance showed no differences between any of the groups on vocabulary retention scores (Appendix B, Table 44). This result was not surprising since the researcher eliminated the vocabulary items from tests if students reported having further contact with them. Students were also divided into groups of females (7 students) and males (8 students). T-tests showed that there were no differences between females and males (Appendix 8, Table 45). In their final evaluation of the course, students indicated how much effort they had put into the course, and accordingly were given a 1 (much effort) or a 2 (some effort). Nine students were in the first group and six in the second group. T-tests showed no differences between the groups (Appendix B, Table 46). On the initial questionnaire students were asked to give their reasons for taking the course and were grouped the following three ways: 1) took the course because of interest (8 students), 2) took the course because it was a requirement, but also because of interest (3 students), and 3) took the course because it was 'a requirement for graduation (4 students). On a One-way 127 analysis of variance no differences showed. When T-tests compared the most extreme groups (group 1 and group 3), there was a significant difference only at R1 (Appendix 8, Tables 47 and 48). A slightly different way of looking at this problem was to combine the scores for effort and motivation to arrive at a four-level scale (2 for a combination of high effort and motivation through 5 for low motivation and effort). Since this scale is ordinal in nature rather than interval, a nonparametric correlation was performed to calculate Spearman's Rho. We would expect a negative correlation with test scores since a 2 represents the most combined motivation and effort and a 5 represents the least. The results are shown in Table 6. TABLE 6 CORRELATIONS OF TEST SCORES WITH COMBINED MOTIVATION AND EFFORT TEET 720 90 21 h 91 E2 92 E3 *23 24 FT— EBT.+ -.42 -.45 -.44 -.1s -.18 .09 .11 .29 -.07 .09 EFF. race. .06 .05 .06 .31 .28 .39 .37 .18 :40 .37 It appears that if there is any connection at all between motivation and effort and test scores, it is only at the very beginning. As time goes on, any kind of a relationship disappears. The next question was whether students who did better in class (earned the higher grades) also did better in retaining their knowledge. Again the students were divided into two groups, those who earned an A, A-, or 8+ in the 128 course (8 students) and those who earned a B— or lower (7 students). Results of the T-tests are shown in Table 7. TABLE 7 T-TESTS FOR GROUPS BASED ON COURSE GRADE TE T _; N MEAN 780 T VALUE’ l—TAIL PROB. R0 HIGHER 8 88.4 8T7 4.00 .00 LOWER 7 70.1 8.9 PO HIGHER 8 6978 521.1 3.05 .00 LOWER 7 43.1 9.8 RI HIGHER 8' 79.0 14.1 3.58 .00 LOWE§_ 6 50.8 15.3 91 HIGHER 8 45.5 26.7 1.50 .08 LOWER 6 27.7 13.0 R HIGHER 8* 65.5 20.6 1.48 .08 LOWER 5 47.2 23.3 P2 HIGHER 8 32.3 22.9 .87 .20 LOWER? 5 22.4 12.6 R3 HIGHER ’7 ’42.0 27.5 .62 .27 LOWERp 5 32.4 24.4 PB HIGHER 77* 11.1 9.9 .35 .37 _ LOWER 5 9.2 8.8 R4 HIGHER ’86 35.4 24.3 1.06 .15 LOWER 7 23.1 19.5 94 HIGHER 8 ”’11.3 10.1 1.25 .12 LOWER | 7 5.7 6.2 It is interesting to note that the only significant differences found between the two groups are on the beginning tests, but this is to be expected since course grades were based in part on knowledge of vocabulary words at Time 0. Even though later times do not show significant differences, Group 1 maintains a higher mean each time than Group 2. Note that at the end of the retention period, the mean for production vocabulary for the higher group stays about the same (has possibly leveled off), whereas the mean score for the lower group continues to decline. The relationship between these two groups is illustrated by the graphs in Figures 1 and 2. 129 100 five 30- 20—4 10- r I F I I t I O 20 40 GO .. Till- (In '00:. a mo" 2 Lou-'- FIGURE 1 RECOGNITION SCORES FOR HIGHER AND LOWER GRADES 400 BO - IO - 70 IO 30 §.. 10 20 an A O T fi r . v r u fl; 0 an 40 so no a Mali—"‘mmi ...... FIGURE 2 PRODUCTION SCORES FOR HIGHER AND LOWER GRADES 130 Another way to look at the data is to correlate the test scores for each time. Do students who receive high recognition and production scores at Times 0 and 1 continue to receive high scores as time goes on? Table 8 gives the Pearson correlation coefficients for the fifteen students on their ten tests over time. TABLE 8 CORRELATIONS OF TEST SCORES FOR 15 STUDENTS OVER TIME R1 R2 6R3 R46 P0 0P1 P2 53’ P4 R0 | .76 .61 .35 .47 .87 .67 .43 .33 .55 I R1 | .89 .65 .70 .73 .81 .70 .65 .59 R2 .89 .90 .83 .55 .74 .83 .82 .65 R3 I .65 .90 .92 .28 .65 .88 .89 .76 R4 .70 .83 .92 .50 .81 .82 .97 .88 P0 I .73 .55 .28 .50 .77 .47 .40 .58 P1 .81 .74 .65 .81 .77 .77 .77 .72 P2 .70 .83 .88 .82 .47 .77 .83 .67 P3 .65 .82 .89 .97 .40 .77 .83 .93 I 94 l .59 .65 .76 .88 .58 .72 .67 .93 The strong positive correlations show that students who score high on the original and first tests continue to score high on later tests, although the correlations generally grow weaker over time. Is early production or recognition a better predictor of final scores when we look at the data organized by student? Consider Table 9: 131 TABLE 9 CORRELATIONS WITH SCORES AT TIME 4 RUTand’R4 .47 |RlandR4 .70 IR2 and"R4 .831R3 and R4 .92 R0 and P4 .55 |R1 and P4 .59 |R2 and P4 .65 |R3 and P4 .76 | I I P0 and R4 .50 |P1 and R4 .81 |P2 and R4 .82 |P3 and R4 .97 P0 and P4 .58 [PI and P4 .72 [P2 and P4 .67 |P3 and P4 .93 Although production scores seem to have slightly higher correlations than recognition scores with both R4 and P4, the numbers are too close to draw any conclusions, especially since only fifteen students were involved. This same question will be raised again in Section 3 of this chapter when we specifically discuss the relationship of recognition and production vocabulary. Case study Since variations among individual students were so great, it seems best to study some of the individual cases in depth. Figures 3-17 show the graphs for each student's recognition and production scores over time. All student names have been changed to protect their anonymity. Note that Neal was available for testing only at the beginning and end of the l8-month period, and therefore his graph appears as a straight line. The data on which these graphs are based are found in Table 40 in Appendix B. t“. lure 132 1N .4 10 d o T I I I I I T 0 an 40 O T)- (M u) a ”men 4' MIC! FIGURE 3 RETENTION SCORES FOR SARA (STUDENT 1) 1m ”- 704 ”cl ”-1 404 :4 2" —.~ 104 FIGURE 4 RETENTION SCORES FOR DAVID (STUDENT 2) k." k“. 133 TO - ' u 70 so \ 5° ‘0 a a \ 1° \ o ' l U I I'— r f I 4: U a “ U II 1']. (In *3 D "In" 9 m." FIGURE 5 RETENTION SCORES FOR JUDY (STUDENT 3) m I -( U -( 1' I T r I I I 0 so 40 m N TI- (10.) o Huey-men 9 mm FIGURE 6 RETENTION SCORES FOR LAURA (STUDENT 4) :rL' :7 V‘ I :3 RN. 134 «4 n .1 20 4 1n -( a ‘ I I I t r I a 20 40 U n- (In I...) a WING! :- reaction FIGURE 7 RETENTION SCORES FOR RACHEL (STUDENT 5) FIGURE 8 RETENTION SCORES FOR GARY (STUDENT 6) Score Score 135 «m U. .4 VII-J In 30.1 ‘0" ——O 30.: all-4 10+ . 0 . . . . . . .fi 6 an no as o tunnnwn ”-cm'—-2 "‘NHN" FIGURE 9 RETENTION SCORES FOR ED (STUDENT 7) «on U- are an. or. ”-1 4H 3.4 flu 162 . . . . r . . .4‘ a 20 46 an a WING! "- CM": R’mun FIGURE 10 RETENTION SCORES FOR TODD (STUDENT 8) &Of. Ito's 136 1G I U 70 J In .1 SO .. 40 4 a4 fin an . 10 d o I I I I v— I I I 0 2! 4. U D Yunnan-n U mun” f MIC! FIGURE 11 RETENTION SCORES FOR JEFF (STUDENT 9) «a a .4 n d 70 Q d as J 4. d a; a 10 J a I ‘ ' 0 an 4'“ I 8:! ' I]- (In *) O WON" 4- Wm.” FIGURE 12 RETENTION SCORES FOR NEAL (STUDENT 10) kw. RC. 137 0‘ ' r ' ‘ ‘ flu T 0 20 40 U U Till (In eel-I) a was» ,. mun ‘ . FIGURE 13 RETENTION SCORES FOR KIMBERLY (STUDENT ll) ”-1 04 Q4 :04 «4N/’ a.» and ”-1 I I I I I r a 20 40 as U T). (In ml) a mum 1» Mia! FIGURE 14 RETENTION SCORES FOR DIANE (STUDENT 12) Store “I 138 1a .4 NE E; 504 34 ad ' 4// 104 T r I r f TmeUnI-I) o tunnnwn ‘r humane '4 FIGURE 15 RETENTION SCORES FOR JOEL (STUDENT 13) FIGURE,16 RETENTION SCORES FOR KAREN (STUDENT l4) 139 1U n; ma n1 n2 ”-1 more «a and 10d T r I I I 1' I O a: 40 U U nn 1..“ o qunnmn (n 3 fi‘uamn FIGURE 17 RETENTION SCORES FOR GLEN (STUDENT 15) Figure 18 is a composite of all fifteen graphs. At this point we can make two important observations about the mean curve: 1) retention drops rapidly at first and then levels off and 2) recognition scores are consistently higher than production scores. Later in Section 3 of this chapter we will specifically discuss the relationship of recognition and production vocabulary, while at this point we will focus on individual students and why they might have deviated from the mean curve. 140 ‘IOD 90 IO 70 .- HO 1 SO-( Score 4aq 30- ZO-J ‘10—: I 1 f r I D an 40 60 III rm- ln m C) Iii-cognition Scar. C :- F'Pocuctlon Scar. FIGURE 18 COMPOSITE OF RETENTION SCORES (RECOGNITION AND PRODUCTION) Jeff and Sara are the first students that we will discuss since their scores are quite close to the composite. Jeff was a student in the class in Grand Rapids. He had dropped out of a first semester Spanish class a year and a half earlier at Calvin (with a different text), and was now beginning over with 101. He was taking the course as a college requirement and said that he put a great deal of effort into the class. His final grade for this course was a B-, and he went on to finish three more semesters of Spanish at Calvin. Jeff liked to study vocabulary by associating a Spanish word directly with its picture and trying not to think in 141 English. He used a mnemonic device from time to time (for example, Mastercard to remember masticar ‘to bite'), but he preferred to access the words directly in Spanish. His grade for the course was about average for Calvin students, and his scores on the retention tests were also near the mean. His vocabulary dropped off rapidly at first; in fact his production score was halved by the end of the three-and- a-half-week course. However, his scores on tests at Times 3 and 4 remained nearly the same. Sara was another student whose scores were close to the mean. Sara participated in the course in Mexico because of her interest in missions. She had no background in Spanish, but she came from Canada and had studied two years of French in junior high. Sara’s grade in the Spanish course was a 8+, and she also said that she put a great deal of effort into the course. After finishing the course in Merida, she worked for three weeks in a Tzeltal-speaking village and used Spanish only with her host family. Her production score (but not her recognition score) jumped somewhat after this three-week experience. At the end of the total 18- month period, though, both of her scores were slightly below the mean. After her experience in Mexico, she did not have any further experience with Spanish. Sara preferred to study vocabulary from English-Spanish word lists, first covering up one side and then the other. She used few devices to aid memory, but one example was English tarp to remember Spanish carpa ‘tent'. When she was 142 asked to describe the picture of the classroom (see Appendix A, Figure 22) at the end of 101, she was able to say the following: ‘Un profesora --how do you say teacher?--ensefia un estudiante. Ocho y mediodia...puerta, libro, cuaderno, tiza, borrador.’ After eighteen months all she could say about the same picture was ‘Well, son las ocho. I can't remember anything.’ In contrast, let us consider three students who consistently maintained scores above the mean. David participated in the course in Merida and was interested in missions and in Latin America. He had already studied German and Greek, but had no previous contact with Spanish. He received an A in the course, even though he admitted that he could have spent more time studying. During his field experience he did not have many opportunities to speak Spanish (he stayed in a Mayafspeaking area and used English with his friends); however, he returned to Mexico the following summer and had a few more opportunities to speak Spanish. All together his Spanish-speaking experience totaled about four weeks, and he received no further formal study in Spanish. David studied vocabulary from flashcards he had made. He said that he liked to look for relationships between words, either cognates in English and Spanish or relationships within Spanish. For example, on his final production test he remembered both 33213 ‘needle' and agudo ‘smart, sharp' because a needle is sharp. Barbilla ‘chin' 143 reminded him of barber, and words like padrino and bisabuelo were closely related to other words he knew (padre and abuelo). ‘ Another student who performed consistently well was Joel, a participant in the Grand Rapids class. Joel had studied Latin, but had no previous contact with Spanish. He originally joined the class because he was planning a trip to Spain, but his plans fell through midway in the course, and he admitted that he studied far less toward the end. Joel received a 8+ in the course and had no further contact with Spanish. Joel's scores showed a precipitous decline at first, but they leveled off sooner than the mean and even showed a slight rise at the end. Joel explained that he worked harder on the final test at Time 4, spending more time trying to access the vocabulary. When he studied vocabulary, he reported using as many devices as he could think of to help memorize it, and this showed up on his final test in which he said he remembered the word £2523 ‘stupid' because another student used it to joke with him, the word lagrima sounded like tear in Latin, and £3339 (mud) reminded him of fangs. In his original picture description task at the end of 101 he described the village (see Appendix A, Figure 22) with the following words: ‘La pictura tiene mucho cosas. Hay mariposa y calle, hace sol, hay montafia. Hay perro, un gato. Hay hombre, la méquina, hay loma, hay tree.’ His picture description eighteen months 144 later was not too different: ‘Es la sol. Es la mariposa. Es la iglesias. Es la tractora. Es la ganada. Es la loma. Es la casa. What are dog and cat? I can't remember.’ Note that he was still able to produce the two target words mariposa and 1293. A third student who did very well throughout was Glen, the sixteen—year-old student in the Grand Rapids class. A few years earlier he had received 4 hours of contact with Spanish through TPR (Total Physical Response), but he had no other foreign language background. He joined the class because of an interest in Latin America and said that he studied a great deal. His final grade was an A, and he went on to take more Spanish in high school and also had contact with Spanish—speaking migrants during the summer. Glen said that he tried to think the words in Spanish without using reference to English as a crutch. He learned various pairs of words together, like gang ‘grey-haired' and galgg ‘bald', or huella ‘footprint' and £2222 ‘mud'. Another thing that helped him was that he went back immediately to study any word that he had missed on a test during 101. If he missed a word, he said that it was a sure sign that he would not miss it a second time. Looking at these three students with consistently high scores, certain patterns emerge. All three of them had some intrinsic motivation for taking the course, and all three had shown previous interest in foreign language. Even though amount of effort was not uniform, they all received 145 high grades in the class. All three learned vocabulary through relationships and associations with other words. 0n the other hand, let us look at two students whose scores were less than the mean. Gary started out with a fairly high recognition score, but his production score was less than half of the recognition. By the end of the eighteen months Gary retained very little, either in recognition or in production vocabulary. Gary had no previous foreign language experience before he joined the class in Merida, but he did have an interest in missions and in Latin America. His efforts to study outside of class were sporadic because of the conditions under which he was living, but in class he paid attention and ended up with a course grade of 8+. His field assignment was in a Mayan village where little Spanish was spoken, and after leaving Mexico he had no further contact with Spanish. His time in Mexico was difficult because of problems his wife and two small children had in adjusting to a foreign culture, and by the end of the summer he had made a decision that his family would be happiest staying in the United States. Gary admitted that he had no special method for studying vocabulary--he just tried to remember the word and that was all. He also said that after the course he did not have any special reason for trying to remember the words. A few came to mind because of vivid experiences--seeing the globos ‘balloons' in Mexico or remembering the teacher's 146 story of fishermen catching tiburones ‘sharks' where she had just taken a swim. Kimberly was another student whose scores were below the mean, both in recognition and production. Kimberly had had four years of high school Latin but no previous contact with Spanish. She took the summer language course in Grand Rapids in order to fill a single course requirement in foreign language at another college, and thus had no further contact with Spanish after finishing 101. She reported putting a great deal of effort into the course at Calvin. Her course grade was a C. Kimberly said that she was more acoustically than visually oriented and that she would read words to herself to learn them by sound. She remembered a word like pupitre ‘desk' by knowing that she was ‘pooped' when she sat in it. Although her Latin helped her to remember lagrima, some French she had learned in France the summer after 101 got in her way when she took the final oral interview (fleurs instead of flores, garcon instead of muchacho.) Kimberly was also extremely nervous about tests, and this most likely impeded her ability to do well. Looking at both cases, we see that neither student ended up with a desire to go on in Spanish. Gary did not have an effective way of studying vocabulary. Kimberly tried to learn through association with sounds, but her hold on Spanish was too nebulous after just one course to last for much time, especially after some contact with French. 147 Probably the most instructive cases are those in which the student performed entirely differently than what could be expected from class performance and original test scores. One of these students was Karen, a student who took the class in Grand Rapids because she was interested in nursing and thought she might use Spanish in her career. Karen came to the course with a background of three years of high school French, though she had no previous contact with Spanish. She worked hard in the course and received an A-. She went on to take Spanish 102 at Calvin during the summer but had no further contact with Spanish after that. Karen's original test scores were originally very high, 88% for recognition and 81% for production. Note that the production score is nearly as high as the recognition score. On her original picture description test of the village, she was able to say the following: ‘Es una mafiana y es un ciudad avec--con una iglesia y las casas. Tengo un montafia y los plantos y arbols. Hay una mariposa.’ Even though Karen's recognition score remained high on the first retention test, her production score was only 39%, which was right at the mean. 0n the final tests her recognition score was only 8% and her production score 2%, far below the mean. A look at her final production test (written) shows more French than Spanish: piscine, nez, and oiseaux, for example. Her final oral picture description showed the same problem; all she could say was ‘un mariposa, un eglise, montagne, les casas', very little for an 148 investment of seven intensive weeks and several hundred dollars. Karen seemed to use effective methods for studying vocabulary during the class. She read and pronounced the words and tried to form pictures in her mind. She related the words to French that she knew, and just before a test she studied the word pairs in English and Spanish, covering up the Spanish and giving the words by memory. One interesting comment that she made was that she knew the beginnings of the words much better than the endings. It is difficult to explain why Karen retained so few words, when it seemed like her French should have helped her Spanish. Instead the three years spent studying high school French seemed to remain much stronger in her memory than the seven intensive, but quick weeks of Spanish. It seems that she never reached a ‘critical threshhold' in Spanish over the seven weeks in spite of her excellent grades, and that what was quickly learned was quickly forgotten. A student in whom the opposite effect was observed was Diane. Although Diane came to Calvin with two years of high school Latin, she had difficulty with Spanish and hadv already dropped out of one Spanish course. She took the summer 101 course because of the language requirement and received a C, but might have done better if she had not worked at another job for so many hours that at times she could barely stay awake in class. During the 102 course that followed, Diane again dropped out. Diane finally 149 finished 102 during the following school year, but dropped out once more during second-year Spanish. Diane's original scores were below the mean, 72% for recognition and 42% for production. However, on the first retention test her recognition score was slightly higher than her first score (74%), and her production score had hardly dropped at all (39%). At the final test, her scores were far above the average: 59% for recognition and 19% for production. These are not the results we would have expected when looking at her original scores. After the final test, Diane explained how she remembered so many words--the letters of the words themselves visually represented their meaning. For example, for Diane the word aaiya actually looked like a scalp and the E in pafia looked like a rock. To remember ala (wing) she visualized a bird covered with a's instead of feathers. The two 5's in gaaaa looked like claws, while the two 3's in arrugado looked like wrinkles. The two 3's in ElEEQ looked like two balloons, and the two ll's in llaaa ‘flame' reminded her of two matches. Cuenco ‘bowl' was'a bowl with the number EEEEE ‘five' written on it. Other Spanish words had links to English; for example, aaaaa ‘gentle' reminded Diane of gentleaaa. Pasillo ‘hall' reminded Diane of a pass, which she needed to go down the hall in high school. Lagrima ‘tear' had the word gala in it and dehesa ‘meadow' had the word gay in it. The word mariposa reminded her of the Virgen Mary, someone quite beautiful. Diane's rich 150 visual and acoustic imagery undoubtedly helped her to retain words that other students soon forgot. The final case that we will look at is that of Laura. She is the only one of the Mexico group that stayed in Latin America after the termination of the program. Laura went to Mexico City for further language training and then on to Honduras where she worked in a mission program. She took her final retention tests while home for Christmas in December 1989. At first appearance it seemed that it might be necessary to drop the data from Laura because her exposure to Spanish was obviously far greater than with any of the other students, and by December of 1989 her Spanish was by far the most fluent. However, the attrition of target words closely paralleled that of the other students. For example, in her first rather painful oral description of the picture of the classroom at the end of 101, she mentioned several objects, including maceta ‘flowerpot', one of the target words. In her final oral description of the same scene in December of 1989 her Spanish was very fluent and she told an interesting story about the picture, but all of the original target words (including maceta) were absent from her speech. Laura's original scores were lower than the mean, 61% for recognition and 47% for production, and once the words with which she had contact were eliminated from the final data, her final scores were also low (17% and 3%). Laura's written tests were far from blank, though. She had 151 substituted words from the Spanish dialect where she was living for the target words she had been taught; for example, mojado for empapado ‘wet', fésforo for cerilla ‘match', basura for cascote ‘trash', cubierta for colcha ‘blanket', and canasta for EEEEQ ‘basket'. Laura remarked that,even a North American woman living and speaking Spanish in Honduras for seven years did not know the names for all the pictures, and Laura felt that it was quite impossible to expect her to know them all. The case of Laura seems to be one of forgetting through interference, in which the original vocabulary words are supplanted by words that the subject hears continuously. Although her Spanish had improved a great deal, there was little measurable retention left of the original words. These case studies point out the variation that exists among individual students. A certain pattern seems to be true for all students: rapid attrition at first with a gradual leveling off and recognition scores that are consistently higher than production scores. However the ratio of recognition vocabulary to production vocabulary varies with each student. Although most students who perform well at the beginning also perform well at the end, and vice versa, there are a couple of students who do not follow this general pattern, and they either retain or forget an unexpectedly high amount. Visual and acoustic mnemonic devices relating words to an image or to the sound of an English word seem to be important in the retention of 152 the words in this study, as well as the chance to study Spanish long enough to build up a ‘critical threshold' from which knowledge is not so readily forgotten. Results for the hypotheses about students Hypothesis 1: Students who study in a target-language environment retain more vocabulary than those who study in a native-language environment, even when the specific vocabulary is not rehearsed or used outside of class. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Table 5) Hypothesis 2: Students who have pre-course contact with Spanish (either informally in the culture or formally in the classroom) retain more vocabulary than those who have no previous contact. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Page 125) Hypothesis 3: Students who have post-course contact with Spanish (either informally in the culture or formally in the classroom) retain more vocabulary than those who have no further contact, even when the specific vocabulary learned is not rehearsed or used. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Page 126) 153 Hypothesis 4: Vocabulary is retained at different rates according to sex. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Page 126) Hypothesis 5: Students who report more effort while taking a language course retain more vocabulary than those who report less effort. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Pages 126-127, Table 6) Hypothesis 6: Students who report more intrinsic motivation for the course retain more vocabulary than those who report more extrinsic motivation. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Pages 126-127, Table 6) Hypothesis 7: Students who receive higher grades for the course retain more vocabulary than students who receive lower grades. The corresponding null hypothesis cannot be rejected at the .05 probability level (Table 7). However, Figures 1 and 2 show a pattern of consistently higher scores for students who received higher grades and consistently lower scores for students who received lower grades. 154 Hypothesis 8: Students who receive higher scores on the first retention tests continue to receive higher scores on later retention tests. Strong correlations were noted between scores on early tests with scores on later tests. (Tables 8 and 9) Retention scores for individual students were examined by means of graphs and discussion. Although most students who perform well at the beginning also perform well at the end, and vice versa, there are a few students who do not conform to the overall pattern. Further discussion of the results above and of individual variation will be found in Chapter 5. 155 ANALYSIS BY WORD A second way to look at the data is to analyze it by word. What percentage of the students could recognize or produce a word at each point in time? What changes occur in a word over.time in its recognition and production strength, and what characteristics of the word might affect the likelihood of its long-term retention? Statistical study The independent variables for each word are given in Table 3 in Chapter 3, where each word is listed in Spanish with its English translation, the paso in which it was taught (A,B,C,D,E), its part of speech (noun, verb, or adjective), its frequency in teacher input (very low to high), and whether it was taught and originally tested for both recognition and production or for recognition only. Table 41 in Appendix B shows the recoding of these variables for statistical analysis, as well as each word's rating on one scale of emotionality and two scales of saliency. Also in Table 41 the percentage scores for each word are given for each recognition and production test. Notice that there are no scores for words in Pasos D and E for baseline production, since part of the design was to originally teach and test these particular words for recognition only. Other missing data occur because of problems in carrying out the 156 project--for example, for parts of two lessons in Merida the tape recorder did not function, thereby making it impossible to collect data on input for several words. Whenever data were left out of the tables, they were also left out of the calculations. In order to analyze the role of frequency in input on the retention scores of words, the categories very low and low (fewer than 10 times) were recoded as ‘low,’ and the categories medium and high (10 times or more) were recoded as ‘high'. As a result, 49 words were included in the low category and 25 words in the high. Independent one-tailed T-tests were then run on the two groups of words, with the results shown in Table 10: TABLE 10 T-TESTS FOR FREQUENCY IN INPUT TE___ST no i=0 1'21 P1 in J12? 91:3 1‘33 R4 134 T VALUE -.72 -.15 -.25 -.59 .38 -.36 1.22 -.40 .51 1.00 PROB. .24 .44 .40 .28 .35 .36 .11 .35 .31 .16 The results of this test for significant difference show that retention of words in this project was 223 dependent on the number of times the instructor repeated the words in the input. These results are reinforced by a nonparametric correlation test in which the frequency was recoded on a scale of 2 to 8 and correlated with test SCOIOS . 157 TABLE 11 CORRELATIONS OF FREQUENCY IN INPUT WITH TEST SCORES -'r'_ss' '1'" 3‘0 P_O i1 PT R2 132—‘33 P3 R'Z""'_PT""' Spearman rho .07 .07 .03 .14 -.06 .10 -.11 .07 -.14 -.09 PROB .26 .28 .41 .11 .30 .18 .16 .28 .11 .22 Again, there is no pattern of correlation between the number of times words were repeated in class and the test scores. The difference between repeating a word 5 or 20 times may not have been great enough to make a difference in subsequent test scores, especially since none of these words was repeated again in input after the first twenty hours of class. Another research question was whether part of speech influenced the retention of vocabulary. It must be noted that because of the nature of the vocabulary taught in the beginning pasos of Dos mundos, verbs and adjectives could be inserted only into Pasos A and 8. Thus the number of verbs (11) and adjectives (14) is much lower than the number of nouns (75) taught. Moreover, all words, whether verb, adjective, or noun, were concrete and were taught through actions, pictures, or real objects. Since verbs were taught through Total Physical Response and tested in command form only, they showed no morphological complexity during the first twenty hours of class. Adjectives showed some complexity in that students heard them inflected for gender 158 and number. However, they were tested only in the masculine singular form. The first test performed to find a significant difference between parts of speech was a One-way analysis of variance, with verbs, adjectives, and nouns each forming a group. The results of these tests are shown in Table 12. TABLE 12: ONE-WAY ANALYSES OF VARIANCE FOR PART OF SPEECH TEST R0 5R0 R1 Pl R2 R2 R3 3R3 R4 R? F 3.45 .12 .16 1.88 .51 .62 1.93 .90 .62 1.39 F PROB .04 .89 .86 .16 .60 .54 .15 .41 .54 .28 The test at R0 is the only one that shows a significant difference, and the accompanying Tukey test showed that the difference was between adjectives (with a mean of 89.57) and nouns (with a mean of 78.17). Thus it appears that only on the baseline recognition test was there a significant difference related to part of speech, a difference that does not exist in any of the further testings. Another way to look at the data is to limit the analysis to the words taught in Pasos A and B only, since these were the only pasos to include all three parts of speech. If this limitation is made we are dealing with only 37 nouns rather than with 75. The number of verbs remains at 11 and the number of adjectives at 14. Another One-way analysis of variance provides us with the results in Table 159 13. The set-up of Table 13 is similar to that of Table 12, but the data include only the words from Pasos A and 8. TABLE 13 ONE-WAY ANALYSES OF VARIANCE FOR PART OF SPEECH (PASOS A AND B ONLY) TEST R5 P0 RI P1 ’R2 ’52 R3 P37 R4 P4 F 2.26 .20 1.89 .28 2.75 .80 3.09 .50 2.47 .99 F PROB .11 .82 .16 .76 .07 .46 .05 .61 .09 .38 Although only test R3 reaches significance at the .05 level, an interesting pattern is established in that probabilities on the recognition tests are lower than on ‘ production tests. This seems to result from the adjective scores--which were high on the original tests but are lower on later tests than the scores for verbs and nouns, especially on the recognition tests. Table 14 gives the means and standard deviations for verbs, adjectives, and nouns at the different test times. Note that there are two columns of scores for nouns: the first column includes nouns from all pasos and the second includes nouns from Pasos A and B only. 160 TABLE 14 MEANS AND STANDARD DEVIATIONS FOR PART OF SPEECH PART OF SPEECH VERB ADJECTIVE NOUN (A-E) NOUN (ARET‘ l l R0 MEAN 78.6 89.6 78.2 81.7 SD 13.6 11.3 15.7 14.8 P MEAN 55.5 58.6 57.6 59.0 SD 10.9 15.6 17.1 17.6 RI MEAN 68.6 68.7 66.0 76.4 _f spa 18.3 11.2 21.7 15.6 P1 MEAN 42.0 46.2 35.9 46.6 .1 SD: 23.4 15.6 19.7 17.1 R2 MEAN 6413 55.7 58.4 68.3 SD 20.5 8.8 23.1 18.4 RI‘ERAN 30.5 *33.2 27.7 37.2 SD 15.9 12.7 18.9 18.3 R3 MEAN 48.3 31.2 38.2 45.2 Sp 20.0 15.5 22.7 21.1 R3 MEAN 14.4 10.13 9.8 13.4 _. 89 14.5 8.9 10.3 11.6 R4 MEAN 31.0 24.4 30.7 36.2 SD 10.8 11.1 21.8 20.1 RI‘REAN 12.8 715 8.4 10.4 SD 1 11.6 7.9 8.8 9.5 Figures 19 and 20 on the following pages Show the graphs for part of speech for recognition and for production tests. Notice the pattern that adjectives follow, particularly on the recognition tests. This pattern will be discussed further in Chapter 5. 161 Amu(u mcaoz om mummmm m0 BMdm mom mfldOUm ZOHBHZGOUBm ma mxwam a 3.3 8:02 o 33333 + nmxooa Eu as: am or . ON a p _ C _ . _ . . o 00.. $01.0) aJoog 162 nmn(u mcaoz mommmm m0 kmdm mom mmmoom ON NMDUHm nwu.uuoqu< ow Oh om om our mote) aJoog 163 Anova tests were also performed to study the interactions between the different independent variables, and we find the same pattern repeated in that part of speech seems to make some difference on recognition, but not on production. The following tables Show the Anova tests for the independent variables of 1) part of Speech, 2) frequency in input (rated on a scale of 2-8), and 3) whether a word was taught and tested originally for both recognition and production or only for recognition. Table 15 shows the size of the F ratio (the ratio of ‘between group variance' to ‘within group variance'), and Table 16 shows the probability of obtaining an F value at least that large if all means were equal. The symbols used are § (part of speech), l (frequency in input), and B (taught and tested for both recognition and production or for recognition only). TABLE 15 ANOVAS POR PART OF SPEECH, FREQUENCY IN INPUT, AND RECOGNITION/PRODUCTION: F RATIO "“1flf‘1‘1flf"‘187“'1fi7”"“R2"'”‘R2' R3 R33 'R4' RI S47 .97 - 2.47 .05 *1.49 .43 4.41 .71 4.52 1.31 I 1.21 - .79 1.25 1.17 .47 1.77 1.13] .94] .66 R 9.81 - 39.30 22.24 27.10 17.73 25.02 10.72|25.37 10.50 I SOURCE OF VARIATION EXPLAINED | | 2.744] - I 5.22] 3.97] 3.78] 2.71 4.27] 2.10] 3.78] 2.19 164 TABLE 16 ANOVAS FOR PART OF SPEECH, FREQUENCY IN INPUT, AND RECOGNITION/PRODUCTION: SIG. OF F " """RO """'P"0"' —1_R " ' 'PLI RT 52 R3 R3 R4 P7 3 .39 - .09 .95 .23 .65 .02 .49 .01 .28 I .32 - .56 ‘.29 .33 .80 .13 .35 .46 .65 R .00 — .00 .00 .00 .00 .00 .00 .00 .00 SOURCE OF VARIATION EXPLAINED .01 - .00 .00 .00 .01 .00 .05 .00 .04 As can easily be seen, whether a word has been taught and tested originally for both recognition and production explains by far the greatest share of variation in test scores. This factor will be examined in the third section of this chapter on recognition vs. production. Frequency in input again appears to play no role. However, the pattern of part of speech affecting recognition tests, but not production tests, can again be seen. Tables 17 and 18 illustrate a similar Anova for the following variables: 1) part of speech, 2) frequency in input, and 3) paso (A, B, and C only). Notice that by eliminating Pasos D and E, we have also eliminated the effect of whether a word was taught for both recognition and production or for recognition only, since all words in Pasos A, B , and C were taught and tested originally for both recognition and production. The symbols used are a (part of speech), l (frequency in input), and g (paso). 165 TABLE 17 ANOVAS FOR PART OF SPEECH, FREQUENCY IN INPUT, AND PASO (A-C): F RATIO R04 R0 R1 *R1 R2 R2 R3 R3 R4 R4 S .03 .43 4150 1.79 4.04 1.56 3.90 1.15 3.82 .55 I .83 .90 |1.25 |2.85 |2.01 | .62 1.67 1.40 .74 1.20 P 3.22 2.21 1.13 5.78 1.34 2.50 .19 .78 .06 .10 SOURCE OF VARIATION EXPLAINED | | 1.62l1.34 ]1.40 |2.87 |1.69 I .88 |1.62 |1.00 1.23 [1.00 TABLE 18 ANOVAS FOR PART OF SPEECH, FREQUENCY IN INPUT, AND PASO (A-C): SIG. OF F 4R0 R0 R1 R14 R2 RRRZ R3 R3 R4 R4 .97 .66 .02 .18 .02 .22 .03 .33 .03 .58 .54 .49 .30 .02 .09 .68 .16 .24 .60 .32 .05 .12 .33 .01 .27 .09 .83 .47 .94 .91 mrqm] SOURCE OF VARIATION EXPLAINED .14 .24 .22 .01 .12 .55 .14 .45 .30 .45 Although the probability that the source of variation has been explained is much less in this test, the pattern of the effect of part of speech still holds: significant for the recognition tests but not for the production tests. Why should part of speech have somewhat more effect on recognition than production, and why should adjectives be the least-well remembered? These questions will be brought up again in Chapter 5. The previous test brings up the question of whether a word taught earlier or later in the 20 hours under study has a better probability of being retained. This question is clouded by the research design in which words in the first 166 three pasos were taught and tested for both recognition and production and words taught in the last two pasos were taught and tested only for recognition. As mentioned above and as discussed in the third section of this chapter, the goals for teaching and testing are definitely important. The One-way analysis of variance in Table 19 also clearly shows the difference in scores between Pasos A-C and D-E. TABLE 19 ONE-WAY ANALYSES OF VARIANCE BY PASO (A-E) TEST 4R0 7R0 FRI 77R1 R2 R2 7R3 R3 R4 R4 F RATIO 4.44|4.63|9.45|9.68|6.12|7.08|3.47|3.23|2.53|2.22 F PROB. .00 .01 .00 .00 .00 .00 .01] .02 .05 .07 GROUPS SIG.A/B A/Bl A/Dl A/Dl A/DI A/Dl -- A/E -- I I I DIFFERENT B/D | A/E| A/EI B/D A/El I | AT .05 | l B/Dl B/CI B/Dl | | LEVEL , | | B/El B/Dl B/El | I (TUKEY) 1 | C/Dl B/El 1 l I | l It can be noted that most of the significant differences between groups occur between Pasos A, B, and C (which were taught and tested for both recognition and production) and Pasos D and E (taught and tested for recognition only). In the three places where there are significant differences within A, B, and C, the data show that B has the larger mean, and thus no conclusion can be drawn about words learned earlier or later (Appendix B, Table 49). There are no differences either between Pasos D and E. Recall that data for P0 are missing in this test for 167 Pasos D and E, since these words were not tested for baseline production. . Thus there are no data here that would suggest that first learned is first forgotten or Vice versa. What these data do suggest is that as time goes on, the difference in how the words were taught becomes less important. Notice that the one-way test at P4 no longer shows significance at the .05 level. Another factor that was studied was the role of emotionality or saliency of a word in its retention. The scales were Obtained from second-year college Spanish students as explained in Chapter 3. Two of the scales, the ones for emotionality and for saliency of meaning, were given in English only, and the third scale for the saliency of the Spanish word was given in Spanish followed by an English translation in parentheses. It should be remembered that the students who performed these ratings were from different classes and had not studied these particular words in Spanish. A rating of 1 represents highest emotionality and a rating of 5 represents lowest emotionality. Among the highest rated words were paay! (1.83) and paaa (1.33-- associated with crying), and among the lowest were appaa (4.38), ppaaa and p231 (both 4.17). On saliency (defined as noticeability, conspicuousness), aaaap and stupid both rated high (1.81) and balloon rated a 1.70, while apppa (4.08), animal back (4.10), and ppapg (4.04) rated low. As might be 168 suspected, these two ratings correlated fairly strongly with each other (.65); however, the negative correlation that we might expect with test scores (negative because the lower the number of the rating, the higher the emotion or saliency) is extremely low or nonexistent (Appendix 3, Tables 50 and 51). The students who rated the Spanish words as to saliency of form gave ratings such as 1.75 to papla ‘board', 1.71 to bisabuelo ‘great-grandfather', and 1.68 to globo ‘balloon'. No words were rated a 4, but ratings such as 3.92 were given to hocico ‘animal snout' and 3.80 to dehesa ‘meadow'. When the Pearson correlation coefficients were determined for these saliency ratings and test scores, the results in Table 20 were noted. TABLE 20 CORRELATIONS OF TEST SCORES WITH SALIENCY OF FORMS OF SPANISH WORDS T—__0'—_EST R PO R1 ‘f—‘P R2 Rz R3 P R P r -.04 -;28 -.29 -.13 -.29 -.20 -.24 -.09 —.34 -.21 PROB. .33 .01 .00 .10 .00 .02 .01 .20 .00 .02 The correlation coefficient of this last rating with emotionality is only .11 and with saliency of the English word is only .02. It seems that the form of the Spanish word is somewhat more important than the emotional connotation in determining retention of the word. 169 It is interesting to compare the means for the saliency of Spanish words by paso to see if any one paso contained words that were more salient than another. Table 21 shows .that the means for saliency were quite close in all five Pasos. The most salient words were in Pasos C and E. TABLE 21 SALIENCY OF SPANISH WORDS BY PASO PASO A R C D E MEAN 3.06 3.00 2.70 3.08 . 2.72 The Spanish words were divided into two groups according to whether they were more salient than the mean on this scale (2.96) or less salient, and independent one- tailed T-tests were performed on the scores to see if there was a significant difference between the two groups. 170 TABLE 22 T-TESTS FOR GROUPS BASED ON SALIENCY OF SPANISH WORD TEST N MEAN SD T'VAEUE"‘T=TRIL‘RROR" RU‘RORE 38 81.29 15.92 .84 .20 LESS 57 78.56 15.10 P0 MORE 30 61.33 17.23 1.84 .04 LESS 45 54.49 14.77 RI MORE 39 ’72.67 18.47 2.48 .01 LESS 56 62.46 20.54 PI MORE 41 40.07 19.09 .94 .18 LESS 57 36.25 20.56 R2 MORE 41 64.20 20.463 2.23 .02 LESS 57 54.54 21.64 P2 MORE 414 31.39 18.56 1.25 .11 LESS 57 26.79 17.56 R3 MORE 41 41.59 22.36 1.35 .09 LESS 57 35.53 21.48 P3 MORE 41 9.76 9.67 -.41 .34 LESS 57 10.67 11.44 RI‘HORE 41 36.12 18.72 2.67 .0013 LESS 57 25.60 19.59 RT‘RORER 41 9.49 8.72 .62 .17 LESS 57 8.33 9.44 ' The degree of saliency reaches significance on three of the recognition tests but on only one of the production tests. It appears that saliency is more important for recognition than for production. The list in Table 23 contains the words that were best retained at the last testing. They had recognition scores of at least 50% (at least half the students recognized the word) or production scores of over 15% (at least 2 students could produce the word). 171 TABLE 23 WORDS BEST RETAINED AT TIME 4 RECOGNITION mefiique (50%) pulgar (50%) mariposa (89%) ballena (75%) tiburén (85%) :exprima! (50%) mohoso (50%) cuenco (60%) rama (62%) mirlo (67%) antepasado (64%) padrastro (53%) madrastra (53%) padrino (67%) aguanieve (87%) PRODUCTION :rece! (21%) itache!(29%) mariposa (44%) :ronque! (25%) tibur6n (32%) :exprima! (32 mohoso (25%) cuenco (30%) rama (19%) mirlo (15%) antepasado (17%) bisabuelo (17%) madrina (19%) padrino (19%) aguanieve (23%) llama (30%) %) In contrast, the words in Table 24 are remarkable in that ap students retained them at Time 4 in either their recognition or production vocabulary: TABLE 24 WORDS WITH A SCORE OF 0 AT TIME 4 vistago ‘descendant' cerro ‘hill' verja ‘window grating'v estanque ‘pool' agujero ‘hole' navaja ‘knife blade' The additional words found in Table 25 had scores of under 20% (remembered by two students or fewer) in both recognition and production at Time 4: 172 TABLE 25 OTHER POORLY RETAINED WORDS Iaja ‘belt' quijada ‘jaw' cerilla ‘match' cascote ‘garbage' hilo ‘thread' aguja ‘needle' hurafio ‘timid' delantal ‘apron' sien ‘temple' (of head) manso ‘gentle' agudo ‘sharp, smart' légrima ‘tear' (crying) beca ‘scholarship' vereda ‘path' enano ‘dwarf’ sendero ‘path' aldea ‘village' frasco ‘bottle' When an independent one-tailed T-test is performed on the best-retained words (Table 23) and the worst-retained words (Tables 24 and 25) to learn if their saliency ratings are significantly different, we Obtain the results shown in Table 26. TABLE 26 T-TESTS FOR SALIENCY OF BEST-RETAINED AND WORST-RETAINED WORDS ORDERS . N MEAN 85 T VALUE 1-TRIE‘RR68““‘ RRSTZRRTAINRD 21 2.80 .53 -2.88 .00 WORST-RETAINED 24 3.17 .32 Again the retention rate of the Spanish word seems to be related to.the saliency of its form. Descriptive word study When students were interviewed after taking the test at Time 4 about why they were able to remember certain words, these were the responses given for the words in the best-retained list (Table 23): 173 mefiique ‘little finger' -- it reminds me of mini; it reminds me of miniature pulgar ‘thumb' -- I remember from class and class tests; reminds me of puny (a thumb is short and stubby); reminds me of pulling (on a thumb) mariposa ‘butterfly' -- it's a pretty word (2 students); reminds me of the Virgin Mary (someone beautiful); beautiful 'word; ppaa reminds me of purple (color of butterfly); the a sounds like a flower; I remember the joke you told about the the word butterfly in German and in Spanish (2 students) ballena ‘whale' -- opposite of a small ballerina; it's like a balloon that someone blew up; a balloon as big as a whale; a mellow word; like baleen in English tibur6n ‘shark' -— I learned by rote memorization; sounds mean (2 students); reminds me of jaws; reminds me of tiger shark; a different word, not at all typical :exprima! ‘wring out!’ -- I remember doing this in class (4 students); I think of strangling the cloth mohoso ‘moldy' -- I remember picture; reminds me of mold (3 students); simply sticks in my head 174 cuenco ‘bowl' -- I remember bowl in class; this word was drilled in; I see a 5 (piaap) written over the bowl; I had a hard time with this word so I worked especially hard at it; I missed this word once so studied extra hard; I like the word and associate it with EEEEE ‘basket'; I like the sound of the word rama ‘branch' -- reminds me of bracchia; I see a tree branch; it popped into my head mirlo ‘blackbird' -- I remember the picture; you look at a blackbird like you look in a mirror; interesting word; rote memory antepasado ‘ancestor' -- popped into head; made up of two words, ante and pasado; after + passed; ante means before padrastro ‘stepfather' -- reminds me of padre (3 students); must be distinguished from padrino (godfather) madrastra ‘stepmother' -- like madre (2 students); must be distinguished from madrina (2 students) padrino ‘godfather' -- like padre (5 students) 175 aguanieve ‘sleet' —- gua + nieve; I know two parts of the word; agua is water (2 students); like water-rain; water + (snow; contains the word snow :rece! ‘pray!’ -- the a reminds me of the pa in pray; I can visualize the word :tache! ‘cross out!’ -- I remember doing in class (3 students) 3522323! ‘snore!’ —- sounds like rude; you caught me doing this during class; I remember from class; I like the word bisabuelo ‘great-grandfather' -- contains abuelo; pi means two madrina ‘godmother' -- like madre (3 students) llama ‘flame' -- like the word ‘call' in Spanish (llamar); the 11 looks like two matches; looks like flame; struggled with this word Reasons that students gave for remembering words can be summarized by the following: 1) mnemonic devices relating the Spanish word to the sound of the English word (mefiique and £121) or to another Spanish word (aiaap written on bowl), 2) linking the form of the word to a physical object 176 it represents (11 looks like two matches), 3) structural analysis (2; means two, madrina looks like EEQ£2)v 4) recognizing cognates in English (ballena and baleen), 5) remembering vivid experiences related to learning the word (actions, jokes), 6) having struggled extra hard with the word in the past. It is interesting to note that all but one of these words (aguanieve) comes from the pasos taught for both recognition and production. There seems to be no relationship between the frequency that these words were repeated in classroom input, but some (such as mariposa, :ex rimaz, :ronque!) are definitely linked to vivid classroom experiences as recalled by the students. Only one of these well-remembered words is an adjective. When we look at the lists of the worst-retained words (Tables 24 and 25), we note that none of these words is,a verb and that slightly more than half (13 out of 24) come from Pasos D and E, the lessons that were taught for recognition only (only 23% of total words come from Pasos D and E). Another question can be asked of the words that were well retained. Did words tend to form networks of meanings so that if one word was retained, another closely—related word was usually retained? One closely-related set of words was 21513 ‘blackbird', aaaa ‘branch', ala ‘wing', and alga ‘nest'. None of the students in the final interview could remember how they had 177 been taught these words, but in fact they had been taught together in a story told with pictures in class. Using the pictures from Are you my mother?, a children's storybook in English, the instructor retold the story in Spanish, repeating each of the target words several times. In the final recognition test, there were four students who knew none of these four words, and the data from another two students were eliminated because they had further contact with these words. Of the remaining students, two recognized only £1519. But seven of the students recognized at least two of these words, and four students recognized all four words. In fact, aiplp and papa ended up in the list of best-retained words, both in production and recognition. Huella ‘footprint' and faagp ‘mud' are two more words that were closely related. They were taught with a picture of footprints tracking through mud. There were four students who recognized both huella and faagp, but no students who recognized only one of the words. One word seemed to bring the other one to mind in close association. Results for the hypotheses about words Hypothesis 9: Vocabulary that is repeated more frequently in teacher classroom input is retained at higher levels than vocabulary that is repeated less frequently. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Tables 10 and 11) 178 Hypothesis 10: Verbs, adjectives, and nouns are retained at different rates. The corresponding null hypothesis cannot be rejected at the .05 probability level. (Tables 12 and 13) There is a tendency for adjectives to be forgotten more quickly, a tendency which shows up more on the recognition tests. (Tables 14-18, Figures 19 and 20) Hypothesis ll: Vocabulary learned earlier will be retained at higher rates. The corresponding null hypothesis cannot be rejected at the .05 probability level when applied to data from Pasos A, B, and C. (Table 19) Hypothesis 12: Vocabulary with high emotionality or saliency ratings is retained longer than vocabulary with low ratings. High emotionality or high saliency of English meanings is not correlated with high retention scores. (Page 168) High saliency of the Spanish form is weakly correlated with high retention scores. (Table 20) Spanish words that have higher saliency ratings also have significantly higher scores on recognition tests. (Table 22) The best-retained Spanish words have significantly higher saliency ratings than the worst-retained Spanish words. (Table 26) 179 THE RELATIONSHIP BETWEEN RECOGNITION AND PRODUCTION VOCABULARY Statistical study In order to analyze the relationship of recognition and production vocabulary, the data files organized by both student and word were utilized. Mean scores and standard deviations for times 0, 1, 2, 3, and 4 were calculated for both recognition and production, and the results are displayed in Tables 27 and 28 below: TABLE 27 DATA ORGANIZED BY STUDENT TIME 0 1 2 3 4 - _ i ‘ MEAN SD MEAN SD M’EAN ST: fiAN SD MjE-AN SD REC 80 13 | 67 20 | 59 23 | 38 26 | 30 22 PROD 57 21 | 38 23 I 29 20 | 10 9 I 9 9 TABLE 28 DATA ORGANIZED BY WORD TIME_0_ _1 f2__3 14_ MEAN SD MEAN SD MEAN SD MEAN SD MEAN SD REC 80 15 | 67 20 | 59 21 | 38 22 | 30 20 PROD 57 16 | 38 20 L 29 18 | 10 11 [ 9 9 As can be seen from these tables, the mean scores are the same, whether organized by student over time or by a particular word over time. The standard deviations are somewhat different and can be explained by the two different ways of grouping the data. 180 It is readily apparent from these tables that the size of both recognition and production vocabulary decreases over time, and that at any one time, recognition vocabulary is greater in size than production vocabulary. What is the relationship in size between recognition and production vocabulary? Recall that Melka Teichroew (1982) cites reports indicating that for beginners, recognition vocabulary is about twice the size of production vocabulary, but for more advanced learners the gap seems to close. Her statement permits the inference that production vocabulary will be a certain percentage of recognition vocabulary. Bahrick (1982a:116) states that ‘attrition affects smaller portions of recognition vocabulary than of recall vocabulary. This difference may reflect the fact that the recognition vocabulary is larger, and that the absolute amount of attrition is approximately the same'. Thus Bahrick presents the view that the absolute numbers of words lost from recognition and production vocabulary are about equal. In order to study the numerical relationship in loss of recognition and production vocabulary over time, the quotients and differences in Tables 29 and 30 were calculated. The same results were obtained from using either Table 27 or Table 28. 181 . TABLE 29 RATIO OF PROD. SCORES TO REC. SCORES: P+R TIME 0 41 2 3 4 RATIO 71% 57% 49% 26% 30%, TABLE 30 DIFFERENCE BETWEEN PROD. AND REC. SCORES: P-R TIME 0 1 2 3 4 DIFFERENCE 23 29 30 28 21 From Table 29 it appears that the ratio of production to recognition vocabulary does app remain constant during the attrition process. Rather, the ratio of production vocabulary to recognition vocabulary becomes progressively smaller until production vocabulary begins to level off between Times 3 and 4. Instead of a constant ratio, Table 30 gives a different way of looking at the relationship. The difference between the amount of recognition and production vocabulary remains between 21 and 30 words, with the lowest difference being at Time 4 when production vocabulary has nearly leveled off. In other words, the students lost an average of 13 words (out of 100) between times 0 and l in their recognition vocabulary, but they also lost 19 words (out of 100) in their production vocabulary. Between times 1 and 2, they lost an average of 8 words in recognition and 9 words in production. The averages for between times 2 and 3 are 20 and 18 words, and for between times 3 and 4 the averages are 8 and 2 words (when production has leveled off). Thus the data support the hypothesis that the same amount of recognition and 182 production vocabulary is lost over time, rather than that the ratio of production to recognition vocabulary remains constant. This relationship is graphed below in Figure 21 (the same as Figure 18 in the first section Of this chapter). 100 40 Score Qa- T j I I O 20 40 IO "no (In on“) D WHIRIOE w. «0- *MIM score FIGURE 21 COMPOSITE OF RETENTION SCORES (RECOGNITION AND PRODUCTION) These results confirm Bahrick's view that the absolute amount of attrition is approximately the same for recognition and production vocabulary. Attrition affects smaller portions of recognition vocabulary than of production vocabulary in the present study, illustrated by the fact that after 18 months, 37% of original recognition 183 vocabulary (from Time 0) remains, whereas only 16% of production vocabulary remains. This relationship is illustrated in Table 31. TABLE 31 PERCENTAGE OF ORIGINAL VOCABULARY (TIME 0) RETAINED AT LATER TIMES TIME 0 1 2 3 4 RECOGNITION 100% 84% 74% 48% 37% PRODUCTION 100% 67% 51% 18% 16% The graph in Figure 21 also confirms Hypothesis l4 concerning the rate of attrition. The drop in both recognition and production vocabulary is steepest between Times 0 and l. The Slope is not so steep between Times 1 and 2 and begins to level off between Times 2 and 3 and 3 and 4. Production vocabulary seems to be bottoming out by Time 4, although it would be unwise to make predictions about what would happen if students were tested again several months later. Another important question is one of teaching methodology, stated in Hypothesis 17. IS vocabulary taught and tested for both recognition and production retained at higher rates than vocabulary taught and tested for recognition only? This hypothesis was tested by using independent one-tailed T-tests to find differences in scores for words taught and tested for both recognition and production (Group 1 - Pasos A, B, and C) and for words taught and tested for recognition only (Group 2 - Pasos D 184 and E). The results from these T-tests are summarized in Table 32 below. Note that Time 0 is not included in the table since Pasos D and E were not originally tested for production. TABLE 32 T-TESTS FOR WORDS TAUGHT AND TESTED FOR RECOGNITION AND PRODUCTION AND WORDS TAUGHT AND TESTED FOR RECOGNITION ONLY TEST GROUP MEAN SD T VALUE DF I-TAIL PROB. R1 1 72.12 15.44 5.70 95 .00 2 48.09 23.02 P1 1 42.83 18.08 4.94 98 .00 2 21.91 16.87 R2 1 63.25 17.50 4.23 98 .00 2 43.39 26.15 P2 1 32.73 17.25 4.40 98 .00 2 15.57 13.03 ' R3 1 42.52 20.47 3.77 98 .00 2 24.17 20.42 P3 1 12.13 11.07 3.25 98 .00 2 4.30 5.83 R4 1 33.10 17.84 3.14 98 .00 2 19.04 21.90 P4 1 10.19 9.39 2.99 98 .00 2 4.00 5.92 As shown by Table 32, all of the differences in the recognition and production tests are Significant at the .01 level, showing that words taught and tested for both recognition and production are retained at higher rates than words taught and tested for recognition only. The results of the Anova shown in Tables 15 and 16 in the second section 185 of this chapter also support this finding, showing that teaching and testing for both recognition and production carries far more weight in retention than part of Speech or frequency in input. An alternate explanation for the lower retention of words from Pasos D and E is that these words were taught later than the words in Pasos A, B, and C, and that the last learned is the first to be forgotten. Although it is not possible to reject this explanation entirely, it Should be noted that for the One-way analysis of variance in Section 2 of this chapter (Table 19), there was no evidence that words taught earlier in Paso A were remembered better than words from Pasos B or C; in fact, the retention scores of words from B had a higher mean than the scores of words from A. Because saliency of the form of the Spanish word was also a significant factor in retention (see section 2 of this chapter), Anova tests were run on the test scores to see if teaching and testing the word for both recognition and production, in combination with saliency, could account for the variation. In this test, words were coded a 1 if they were above the mean score on this scale, orla 2 if they were below it. The symbols used are § (saliency) and B (taught and tested for both recognition and production or for recognition only). 186 TABLE 33 ANOVAS FOR SALIENCY OF SPANISH WORD AND TEACHING AND TESTING FOR RECOGNITION AND PRODUCTION: F RATIO _ R0 R0 RI 9 R2 P2 113 R3 114 F; S 2.12 - 9.95 1.88 6.23 2.76 1.39 .05 5.89 .71 R 8.66 - 35.86 24.48 18.33 20.01 11.81 10.00 9.88 9.31 SOURCE OF VARIATION EXPLAINED | | | | | 4.69 I - [14.55] 8.61] 7.83] 7.32] 4.23] 3.52] 4.92] 3.35 TABLE 34 ANOVAS FOR SALIENCY OF SPANISH WORD AND TEACHING AND TESTING FOR RECOGNITION AND PRODUCTION: SIG. OF F R0 1P0 R1 R1 R R2 R3 R3 14 P4 S .15 — .00 .17 .01 .10 .24 .83 .02 .40 R .00 - .00 .00 .00 .00 .00 .00 .00 .00 SOURCE OF VARIATION EXPLAINED | | | | | .00 I - I .00] .00] .00] .00] .01] .02] .00] .02 Together these two factors seem to explain quite well the source of variation in the scores. Notice that the F ratios in the bottom row in Table 31 are higher than those in Table 13, which studied a different combination of factors. Notice also that saliency again seems to be more important for recognition that for production. A different, but related, question, is whether ability to produce a word initially, rather than ability to recognize a word, is the best predictor of both recognition and production later on. Is a word at Time 0 or at Time 1 more likely to be retained if the student can recognize aaa produce it at the beginning than if he can only recognize it but not produce it? Remember that it is necessary to consider Time 1 as well as Time 0, Since not all words were tested for production at Time 0. 187 Three different tests were devised for examining Hypothesis 18 statistically from slightly different perspectives. The first test divided words into two groups so that a T-test could be run. In the first group were words that were recognized at a high level (mean of 80% or higher) and also produced at a high level (mean of over 50%) at either Time 0 or Time 1 or both. In the second group were words that were recognized at a high level (mean of 80% or higher), but produced at a lower level (mean of under 50%) at either Time 0 or Time 1 or both. Some words fell into both groups because of widely varying scores at Times 0 and l, and these words were eliminated from this test. Words that had low production aaa low recognition scores were also eliminated. Group 1 contained many more words (46 words) than the Group 2 (12 words), Since high recognition and high production are so closely correlated. The results of the T-tests are shown below in Table 35. 188 TABLE 35 T-TESTS FOR WORDS AT HIGH LEVELS OF RECOGNITION AND PRODUCTION AND WORDS AT HIGH LEVELS OF RECOGNITION BUT LOWER PRODUCTION TIRE "“' GROUP MEAN SD T VALUE ”OF l-TAIL PROB R2 1 66.80 17.70 .74 56 .23 2 62.33 21.91 P2 1 39.07 18.50 2.78 56 .00 2 23.17 13.58 R3 1 44.59 21.68 .47 56 .32 2 41.33 19.35 P3 1 14.37 12.25 1.57 56 .06 2 8.58 6.96 R4 1 35.65 19.16 1.20 56 .12 2 28.33 17.65 P4 1 12.52 9.89 2.92 56 .00 2 3.83 5.44 The results of these T-tests Show that there are significant differences (at the .01 level) between these groups of words on the later production tests at P2 and P4, which is not unexpected, since there were large differences between these groups at the beginning in production. But words that had high recognition rates at the beginning, even though they were low in production, continue to hold their own in recognition scores. Although their means are slightly below those of the high-recognition, high- production group, there are no significant differences between groups on recognition tests, even at Time 4. It seems that high initial recognition is the best predictor for high recognition at Time 4, and high initial production is the best predictor for high production at Time 4. 189 Another way to look at initial production and recognition as predictors of final production and recognition is to study the Pearson correlation coefficients for the scores over time: TABLE 36 CORRELATIONS OF SCORES FOR 100 WORDS OVER TIME R1 R2 R3 R4 R0 R1 R2 R3 R4 R0 .41 .27 .32 .36 .49 .54 .40 .21 .27 R1 .88 .67 .69 .52 .76 .73 .52 .53 R2 .88 .68 .72 .35 .67 .69 .51 .52 R3 .67 .68 .84 .27 .56 .57 .57 .64 R4 .69 .72 .84 .26 .56 .64 .53 .64 P0 .52 .35 .27 .26 .65 .56 .29 .36 P1 .76 .67 .56 .56 .65 .85 .62 .62 P2 .73 .69 .57 .64 .56 .85 .66 .69 P3 .52 .51 .57 .53 .29 .62 .66 .76 P4 .53 .52 .64 .64 .36 .61 .69 .76 All of these correlations have a probability of .01 of occurring by chance. The strongest correlations are between the same type of test (recognition and recognition, or production and production) on two consecutive testings; for example, R1 and R2 (.88), R3 and R4 (.84), or P1 and P2 (.85) and P3 and P4 (.76). Correlations between recognition tests and production tests given at the same time are also fairly strong: R1 and P1 (.76), R2 and P2 (.69), R3 and P3 (.57), and R4 and P4 (.64). 190 As time passes, correlations between the same tests weaken; for example, the correlation between R1 and R2 is .88, between R1 and R3 is .67, and between R1 and R4 is .69. The correlation between P1 and P2 is .85, between P1 and P3 is .62, and between P1 and P4 is .61. Although correlations exist with tests at R0 and P0, these relationships are the weakest. The greatest attrition occurred between Time 0 and Time 1, and a word correct at Time 1 had a greater chance of being retained over the 18-month period than a word correct at Time 0. When we look at the predictability of a later score, we again notice that recognition seems to be a slightly better predictor of recognition, and production seems to be a slightly better predictor of production. For example, the correlation between R1 and R4 is .69, whereas the correlation between P1 and R4 is lower--.56. The correlation between R1 and P4 is .53, and the correlation between P1 and P4 is higher--.61. Note the following pairs of correlations with tests at Time 4: TABLE 37 CORRELATIONS WITH SCORES AT TIME 4 ’R3 and R4 .84 R3 and P4 .64 45RI and'R4* .69 T R2 and5R4 .72 R1 and P4 .53 R2 and P4 .52 P3 and R4 .53 P3 and P4 .76 h_-—a-—q | I P1 and R4 .56 | P2 and R4 .64 P1 and P4 .61 l P2 and P4 .69 Thus, ability to recognize a word seems to be a slightly better predictor of later recognition than ability 191 to produce that word initially. Likewise, ability to produce a word is a slightly better predictor of later production. The correlations between tests at Time 1 and Time 4 are illustrated by the scattergrams in Appendix B (Figures 23-26). The third perspective on the relationship of recognition and production is given by a multivariate analysis of scores at Time 1 and Time 4, showing how both recognition and production scores at Time 1 can be used to predict recognition and production scores at Time 4. Table 38 gives the multiple regression equations that best predict these final scores. TABLE 38 BEST PREDICTORS AT TIME 1 FOR SCORES AT TIME 4 A FOR FINAL REC.: R - R4 + .612 (RI-RI) + .0918 (Pl-RI) A FOR FINAL PROD.: P - E4 + .0674 (RI-RI) + .229 (Pl-RI) The symbols R4 and R4 stand for the predicted values of a certain word on the final recognition and production tests. The symbols R1, R4, RI, and R4 stand for the means of all words on each of those tests. For example, if we wish to find the predicted recognition score for delantal at Time 4, we need the following information: 192 mean of recognition scores for R4: 30 scores for delantal at R1: 36 (recognition) 25 (production) mean scores at R1: 67 (recognition) 38 (production) /\ Substituting these values in the formula for R4, we get R4 - 30 + .612 (36-67) + .0918 (25-38) - 30 + .612 (-31) + .0918 (-13) - 30 - 19 - 1 - 10 Our projected recognition score for delantal at Time 4 is 10, and the actual recognition score for delantal at Time 4 is 13. The full table for predicted scores and actual scores is found in Table 42 in Appendix B. The formulas Show that in order to predict final recognition, the recognition scores carry more weight than the production scores at Time 1 (.612 vs .0918) and to predict final production, the production scores carry more weight than the recognition scores at Time 1 (.229 vs. .0674). This result is consistent with the earlier tests showing that recognition is the best predictor of recognition and production is the best predictor of production. The multiple correlation coefficient of R4 with R1 and P1 is .693, and the multiple correlation coefficient of P4 with R1 and P1 is .624. When these quantities are squared, we see that scores of R1 and P1 explain about 48% of the variance of R4 and about 39% of the variance at P4. 193 The results Showing the relative importance of recognition and production as predictors when we analyze by word in Tables 36 and 37 are slightly different than the results shown in Tables 8 and 9 when we analyze by student. These results will be taken up again in the conclusions in the final chapter. Case study The final table in this chapter is a case study that shows what happens to the vocabulary of one student over time. The student chosen is Jeff, Since his attrition curve is so close to the mean curve. An X in the appropriate space shows that Jeff could recognize the word on a certain test (a star indicates that he guessed at the answer). A word that caused confusion on the recognition test is enclosed in parentheses. The Spanish word in the Space below the X shows what Jeff actually produced on a specific production test. A horizontal dash indicates that data were not available for that specific slot. 194 TABLE 39 THE ATTRITION PROCESS TIME 0 1 2 3 4 peIlizque I ande de 4R X puntillas ___ _y .1 v_ rece X X X X rece rece rece rece respen pliegue piejan tache amarre X X aparre amarre cojo X tuerto manso ’XR hurafio X ocioso R’ *X *X agudo R agudo torpe X delantal X delanta _, Iaja x x x faja faja faja cremallera X X X cremallera cremallero cremalera aretes X (alhajas) (alhajas) _, arge‘as tac6n X X X 4_ ]_ toquen chaleco X X X alhajas R R5 if 5R alhajas _, arpeja _ mefiique X X X X mennique mefinique meanique mennique tez I X X X #592 tez teza pufio X pulgar 3R 7? (partido) 257 *X . pplgar pulgar ppulgar panza X l 195 TABLE 39 (cont'd.) ‘TIME _0 J g 3 4 Bafbillal X X X bparillo _ .1 quijada X X X sien R; X4 R4 hocico X (horario) (horario) hocio hopico A; espinazo X X eapinazo espinaro espinaza a_ _ garra X | X X *X *X garra gagga gagga gantes pescuezo X X X mariposa X X7 RR; R X maripasa mariposa mariposa mariposa mariposa ciervo X X cervo _f _f _ venado X X X yierbo ]_ _ _ _ ballena X X X X X ballena ballena ballena _ ballena tiburén X X X X X taburén tipuron tipuron tarzO exprima X X X exprima espiren guiHe '-X gpifie guinen ¥_ ronque X X X rpnque ronque ronque silbe X silbe _ _ bostece X X X bostece descalzo X . dascal o _, _ :_ calvo X X X X calvo calvo colva cano X cano if _f mohoso X (manso) (manso) X *X mphoso moaoso mohoso go monado harapiento X X X *X *X harapiento_ _ arrugado X | X X aprugado arragudo empapado X | eppapado _A_ 41 _ maceta X X X X X mesata meceta meseta I mesa 196 TABLE 39 (cont'd.) TIME _0 A_l _2 3 4 tabla X X X X X tabla tapla taplo _7 tabla coIcha X X X X X cplcha colcha colcha _ 4_ cuenco X X X X X openco cuanco _ _ r6tulo X X X X cartel R pasiIlo 5X X *X pasillo _ 4_ huella X X X huella huela _ fingo X (falda) X *X falgo falgo fpejo ala X X X aIa ala ala 4_ rama X X X X R ' .rama ]_ rama rama m1 rfo x x x x E mirlo _ _ nido X X X nido nido _ A_ antepa- x x x E sado aptepasado vastago X véstado _ _ bisabuelo X X Ibisabuelo _ biapuelos tatara— | X X X buelo tatarabuelo tetrabuelo tetrabuelos padrastro -- I I lpadrastro padrestro madrastra -- RI madrastro madrestro _yf _. _ padrino X X X X *X _ ¥_, padpismo madrina X X X *R madrina _ ¥_ madrismo mellizo X X X mellizo mazenas _ ligfima X lagrima Beca beca _ cesto X X X peta ._, _ incendio X X X 7R X incendio incendio I 197 TABLE 39 (cont'd.) TIME" _0 41 2 3 4 llama X X llama ceriIla X aguanie- R ‘R' X X’ X ve :_ .1 _ aguanieve estanque X X X dehesa 7X5 loma X cerro X3 vereda X sendero R agujero carpa X X X X 44X :— carpa carpa aldea X verja 4X pefia X bolivar R 5X5 i i rastrillo X5 enano ’XRR azulejo R’ X if 4;: azulijas azulita navaja X cascote fix 7 hilo RR semilla X X; a:f siIla j: Irasco X X X gIobo X57 754 ii if if :: globo globa aguja X 198 Note some interesting features of the data in Table 39: the general progression of attrition is from 1) recognition and full production, to 2) recognition and partial production, to 3) recognition only, to 4) no observable retention. Jeff shows that his confidence level is not very high on the recognition of some of the final words by indicating guesses with a star. Words that are good examples of this typical attrition pattern are gappa, tibur6n, calvo, maceta, and fango. When Jeff produces part of a word incorrectly, he always gets the first part right. Notice that even when words are grossly misspelled, the first letters, and to some extent the final letters, are preserved, while the middle of the word suffers the most transformations: gapaa > gantes, tibur6n > 22529: mohoso > monado, maceta > aaaa, 33353 > falgp > £2312. Mistakes are sometimes repeated, as in the case of Eaigp, which was repeated on the next test before becoming fpajp on a later test. When mistakes occur in recognition, a word is often mistaken for another that has the same beginning letters; for example, pulgar was replaced by partido on the matching teSt, hocico was replaced by horario, and mohoso was replaced by aaaap. Examples such as these support the belief that words may be ‘penciled in' incompletely in memory, with the first part of the word the ‘darkest', and the middle part the ‘lightest'. This 199 phenomenon is similar to the ‘tip-of-the-tongue' phenomenon discussed in Chapter 2. Jeff gives some examples of missing a word on one test and then remembering the word on a subsequent test (for example, see chaleco, venado, ciervo, pasillo, and others). This phenomenon, though not the norm, should indicate that we need to use caution when we say that a student has ‘forgotten' a word. Although the student may not be able to access a word for a particular test at a particular time, we cannot prove that the word has disappeared entirely from memory and that it cannot be accessed at a future time. Results for the hypotheses about recognition and production Hypothesis 13: Recognition scores for vocabulary are higher than production scores. The graph in Figure 21 shows this hypothesis to be true for the current research. Hypothesis 14: Both recognition and production scores for vocabulary drop quickly at first and then gradually level 23;; The graph in Figure 21 shows this hypothesis to be true for the current research. Hypothesis 15: The best description of recognition and production vocabulary loss is that equal quantities of words 200 are lost over time rather than that the ratio of production to recognition vocabulary remains constant over time. The graph in Figure 21 shows this hypothesis to be true for the current research. This relationship is also illustrated by Tables 29 and 30. Hypothesis 16: Smaller percentages of recognition vocabulary are lost than of production vocabulary. Table 31 Shows this hypothesis to be true for the current research. Hypothesis 17: Vocabulary taught and tested for both recognition and production is retained at a higher rate than vocabulary taught and tested for recognition only. The corresponding null hypothesis can be rejected at the .01 probability level (Table 32). Hypothesis 18: Ability to produce a word initially, rather than ability to recognize a word, is the best predictor of both recognition and production later on. A corresponding null hypothesis states that words that are recognized and produced at high levels initially will not have higher scores on later retention tests than words that are initially recognized at high levels but produced at lower levels. This null hypothesis cannot be rejected, since there were no differences at the .05 probability level 201 between the two groups of words on later recognition tests (Table 35). Correlations between scores and multivariate analysis also showed that early recognition seems to be the best predictor of later recognition, and early production of later production (Tables 36-38). CHAPTER 5 SUMMARY AND CONCLUSION The main focus of this study has been to investigate the attrition of foreign language vocabulary by examining factors such as student characteristics, lexical characteristics, and the relationship between recognition and production vocabulary over time. This final chapter will provide a summary Of the research, discuss the findings and their implications for second language acquisition theory and foreign language teaching, and make recommendations for further research in the area of foreign language attrition. SUMMARY OF RESEARCH Fifteen college-level students of first-semester Spanish were studied intensively during their first 20 hours of the course in order to track their acquisition of 100 specially-targeted words in Spanish. During the following 18 months these students were retested on these 100 words for both recognition and production in order to study the pattern of foreign language vocabulary attrition. Research hypotheses were grouped into three categories: 1) hypotheses about students, 2) hypotheses about words, and 202 203 3) hypotheses about the relationship of recognition and production vocabulary. Most of these hypotheses were tested through statistical procedures such as T-tests, One-way analyses of variance, Anovas, and correlations. Some of the hypotheses, however, were studied by means of descriptive data, and individual cases were also considered. An important component of the research was to gather data from students taught in actual Classrooms under ecologically realistic conditions rather than in a laboratory situation. Although conducting a longitudinal study of this séope involved some problems of control, it was felt that this type of study could provide important data about what happens to real foreign language learners in their very real battle with vocabulary attrition. FINDINGS AND DISCUSSION Before discussing the findings, the limitations of the research should be reviewed. As mentioned in Chapter 1, the sample size of only 15 students is small, which must be taken into account especially when discussing the first eight hypotheses which are based on the data file organized by student. The results for Hypotheses 9-18 are based on the data file organized by word containing 100 items, and since a is larger, statistical significance is easier to achieve. 204 Three factors may have influenced the scores on the retention tests: 1) possible post-course contact with vocabulary items, even though the researcher attempted to eliminate the influence of all such contact, 2) repeated tests over the same vocabulary items, even though these tests were spread far apart in time, and 3) the knowledge of the students that they were involved in some kind of vocabulary research. Any of these three factors could have raised retention scores and made the attrition appear less severe . Student Characteristics Hypotheses 1-8 dealt with student characteristics. Students were grouped as to whether they took the Class in the United States or Mexico, how much pre-course and post- course contact they had with Spanish, sex, type of motivation and effort, and grades they received for the course. No differences at the .05 level were found for these different groups with regard to their retention scores. Graphs Showed, however, that students who finished the course with higher grades did retain slightly more vocabulary than students who finished the course with lower grades. Also, strong correlations were found among retention tests over time, showing that students who scored. higher on the first retention tests also scored higher on the final retention tests. Production scores correlated Q 205 slighty higher than recognition scores with both final production and recognition. Individual case studies of students showed that the students who retained the most vocabulary were ones who had rich capabilities for imagery, who analyzed words structurally, and who looked for relationships with other words. Those who did poorly on retention tests did not make as much use of these memory aids. Also, interference from lexical items from other languages or from another dialect of Spanish seemed to play a role in reducing the retention rate of the target words. When evaluating these results, it was somewhat surprising to find that there were no differences between the students who were taught in the United States or Mexico, or between those who had pre—course or post-course contact with Spanish and those who had none. However, the set-up of the research required that the 100 target words be heard and learned only during the first 20 hours of the course, and that students have no further contact with these words. Thus, the research conditions prevented reinforcement of vocabulary caused by previous or later contact with Spanish either in Mexico or in subsequent Spanish classes. It would be very hard to deny that learning Spanish in a native-language environment and having subsequent contact with Spanish language and culture are very valuable influences on language retention. But the present research 206 seems to indicate that rather than influencing vocabulary retention in a general way, these environmental factors have a more specific result--to reinforce specific vocabulary items. When these vocabulary items are missing from input outside of the course, there can.be no reinforcement from environment, and the vocabulary items will attrite quite rapidly. In fact, a target—language environment may even increase the attrition rate of non-rehearsed vocabulary by adding interference from other words, as in the case of Laura. It was also surprising to note that there was no difference in retention scores based on grouping by type of motivation (extrinsic or intrinsic) or amount of effort reported by the students. This finding on motivation, though, is consistent with data from Edwards (1976, 1977) and Gardner (1985). Gardner states that attitudinal/ motivational factors seem to be more influential during the acquisition period than during the retention period. The case of Karen in the present study is representative of the good student who works hard in class and hopes to use Spanish in a future career, yet who is able to access only a very few words after 18 months. The data in the present study confirm the report by Pratella (1969) that sex seems to have no effect on retention rates. The question of whether course grades can predict retention rates remains open. Course grades are based on 207 many different factors, one of which is initial acquisition of vocabulary. Thus, the T-tests in this research at Time 0 Show a significant difference between students who received higher grades (A, A-, 8+) and students with lower grades (B— or lower). However, this significant difference disappears by Times 2, 3, and 4, even though the graphs continue to show a small difference between groups, especially on the recognition tests. Godsall-Myers (1981) and Bahrick (1984) found that higher grades received in course work were related to higher retention rates, but Clark and Jorden (1984) reported that there was no difference in initial ease in learning Japanese for ‘attriting' and ‘non-attriting' students. In the present research, course grades gave some indication of future retention rates, but there was also a great deal of individual variation. Retention scores at Time 1 were a better indicator of future retention rates than were course grades. It is interesting that Meara (1989) suggests that researchers use retention rates between Weeks 2 and 3 of the retention period as the best predictors of future retention. The strong correlations among scores in the present research Show that once students have shown themselves to be good vocabulary ‘retainers' at Time 1 (about 2 weeks after their last contact with the target vocabulary), they continue to get high scores on later retention tests. The highest correlations occur between subsequent tests in the same mode toward the end of the retention period (for example, the 208 correlation coefficient between R3 and R4 is .92, and between P3 and P4 is .93), but the correlation coefficients between R1 and R4 (.70) and between P1 and P4 (.72) show that the scores at Time 1 can be good early predictors of future retention. Whether production or recognition scores are the best predictors for future retention is really too close to call from the data organized according to student. Although the correlation coefficients slightly favor production, it must be remembered that this data file is organized around only 15 students, and the correlation coefficients are extremely close. This question will be discussed later in this chapter under the section on recognition and production. Word characteristics Hypotheses 9-12 dealt with characteristics Of the 100 target words. The data showed no differences in retention rates for words the instructor repeated with higher or lower frequency during class time or for words that were presented earlier or later. Differences in retention rate for part of speech were noted, caused by the more rapid loss of adjectives than of nouns or verbs. I The 100 target words had been given ratings in the following three categories: 1) emotionality of the English translation, 2) saliency of the meaning of the English translation, and 3) saliency of the Spanish form. Saliency 209 is defined as the ability of a word to attract attention due to its conspicuous, striking, or unusual nature. The first two ratings failed to correlate with retention scores; however, the third rating showed weak correlations with these scores. Moreover, when lists were compiled of the best-retained and the worst-retained Spanish words, the words on the best-retained list had significantly higher ratings in the third category above than the words on the worst-retained list. Recognition tests rather than production tests were more sensitive in detecting differences based on part of speech or saliency rating. It is also interesting to note that on the graphs illustrating the retention rates for students who received higher grades and those who received lower grades, the recognition tests also showed greater differences between groups.‘ Students were asked individually to describe the devices they had used to remember the Spanish words on the best-retained list. Mnemonic devices had frequently been used, such as relating the Spanish word to the sound of an English word or linking the form of the word to a physical object. Students also mentioned using structural analysis, struggling extra hard with a word they had missed in the past, and remembering vivid experiences related to the learning of the word in the classroom. Also, students tended to remember certain words (such as mirlo, nido, ala and rama) as part of a network. 210 Several interesting observations can be made about these findings on word Characteristics and the way in which the words were taught. In the present study, the most influential factor on retention was whether the word was taught and tested for both recognition and production (and this factor will be discussed further in the next section), although the saliency of the Spanish word also appeared to play a role. Judgments about greater saliency were made by Spanish students outside of the main study and seemed to be based on shortness of the word, similarity of the form to English (even when not a cognate), lack of diphthongs and accent marks, and ease of structural analysis. Within the main study, subjects made comments related to saliency during their individual interviews: the Spanish word reminded them of the sound of an English word or phrase or of the actual physical object that the word represented, or the structure of the word was easy to analyze as in aguanieve and madrina. And, the instructor played an important role in making words more salient by associating the words with vivid images or actions during class; the students commented that they remembered some words because of an action, story, or joke they had experienced in class. Such experiences are related to episodic rather than semantic memory. According to Bahrick (1979), episodic memory refers to memory of particular events related to time and place, whereas semantic memory refers to general rules. Schouten-Van Parreren (1989:82) reports that even weak 211 pupils could remember a difficult word that they had encountered only once because they had enjoyed the story in which it had occurred.' Murray (1986) states that variables such as imagery and emotionality affect retrieval from episodic memory. Therefore, it was a surprise to find that the ratings for the emotionality of the English translations of the target words had extremely low or nonexistent correlations with retention scores (Appendix B, Table 50). For the present research, emotionality seemed to be based on what the instructor and the student did with the word when it was presented in class, rather than on characteristics of the English translation. Murray (1986) also states that variables such as word frequency have the greatest effect on semantic memory. The frequency of repetition of words during the first 20 hours of the course may simply have been too limited to Show any correlation with retention scores, or to influence semantic memory to a great degree. Words were remembered because of their saliency in input and because students had to study them for both recognition and production, not because of the numDRr of times students heard them in input during the first 20 hours of class. When we consider the information from Nation (1982) about the number of times a word should be repeated in input (seven for intentional learning and sixteen for Incidental learning), we must keep in mind that Nation is not referring to retention over long periods of 212 time. It appears that for frequency of repetition to be an important factor in retention of vocabulary, words must be repeated far more than sixteen times, a feat that becomes increasingly difficult for an instructor to perform as total amount of vocabulary increases. Another factor which was not significant in the present research was whether a word was learned earlier or later in Paso A, B, or C. Since the words in these three pasos were taught in a matter of a few days, there was not much Spread in time between words taught earlier or words taught later. Thus, there is insufficient evidence to make a judgment on the regression hypothesis for vocabulary retention. A factor which did seem to play a small role was the part of speech of the word, a factor which showed up on recognition tests but not on the less sensitive production tests. Although adjectives scored better than nouns and verbs on the original tests, their scores trailed on later tests. This result was unexpected because research in language loss in aphasia (Obler 1982) and in loss of a second language (Cohen 1986) indicates that nouns are more easily lost than other parts of Speech. Cohen comments that because nouns are easily learned, they are also easily forgotten. Possible explanations for the results in the present research are that 1) adjectives were the only words during the first 20 hoUrs of the course to Show any morphological variation and 2) nouns and verbs are the major constituents of the Spanish sentence and therefore are 213 essential to comprehension and expression; adjectives are somewhat peripheral because they are modifiers. Sapir (1921:117) pointed out the primacy of noun and verb: ‘There must be something to talk about and something must be said about this subject of discourse once it is selected.’ More recently, categorial grammar has presented the concept of syntactic rank; in this analysis of part of speech, nouns are of a higher rank than both verbs and adjectives, Since both verbs and adjectives say something about nouns (Lyons 1977:439). It would be interesting to apply this concept to further research on the role of part of Speech in second language retention. Recognition and production Hypotheses 13-18 dealt with the relationship of recognition and production in vocabulary retention. Graphs showed that recognition scores were higher than production scores at all points in time. Both recognition and production scores dropped the most quickly at the beginning and then gradually leveled off over time. By the end of the l8-month period, production scores had nearly leVeled off and were close to 0 for many of the weaker students, while recognition scores were still continuing to decline. The best description of the relationship of recognition and production vocabulary loss is that equal numbers of words are lost from each type of vocabulary over time. 214 Because production vocabulary represents a smaller absolute number of words, the percentage loss from production vocabulary is greater than the percentage loss from recognition vocabulary. Thus the ratio of production vocabulary to recognition vocabulary is constantly decreasing, until production vocabulary ‘bottoms out' while recognition vocabulary continues to decline. There was a significant difference in retention scores for vocabulary that was taught and tested for both recognition and production, and vocabulary that was taught and tested for recognition only. However, for predicting later recognition and production scores, early recognition scores were slightly better for predicting recognition, and early production scores were slightly better for predicting production. Data from individual students Showed that a typical attrition order was a) recognition and full production, b) recognition and partial production--with the first letters of the word being the easieSt to produce, c) recognition only, and d) no observable retention. Occasionally, words that a student failed to access on a retention test reappeared on a later test, making it impossible to judge at any point whether a certain word was completely erased from memory, or whether the difficulty was in accessing the word. There are many observations about the relationship of recognition and production that can be made from the data. 215 Recognition was the easier Skill, and scores on recognition were always higher than scores on production. In order to score well on production, a student needed more information than simply the first and last sounds in a word. The descriptive data from the production tests showed that students may ‘pencil in' words in memory, with the first letters being the ‘darkest' or easiest to remember. The middle parts Of words were the first to blur over time. When words were confused on the recognition (matching) tests, the two words mixed up were often words that had the same initial sounds, but different middle parts (such as hocico and horario or mohoso and aaaap). By far the most influential factor in determining retention scores was whether a word had been taught and tested for both recognition and production or for recognition only. Requiring students to study for a production test (after a recognition test) during the first 20 hours of the course most likely forced students to pay attention to the middles of words, as well as to review the words again from a different perspective. This different perspective can be related to Craik and Lockhart's depth of processing theory (1972), in which learners process vocabulary on different levels, thereby increasing their retention. ReCOgnition and production skills seem to be different enough that they require processing a word at different levels, and when students are required to go the 216 extra step to produce a word, they add another level to their analysis of a word. Recognition testing was more sensitive for judging lower levels of learning. When some of the students scored close to O on their production tests, their recognition tests still Showed evidence of some retention. In the present research, recognition tests were also more sensitive than production tests for showing differences for students with higher and lower course grades, for part of Speech, and for saliency rating of the Spanish word. In terms of absolute numbers, roughly as many words were lost in recognition vocabulary as in production vocabulary. This result is difficult to explain, but Bahrick (1984c) also reports the same phenomenon. Thus, it does not appear that production vocabulary remains a constant ratio of recognition vocabulary, but that attrition affects larger percentages of production vocabulary than of recognition vocabulary. Production vocabulary that is ‘forgotten' is usually not lost entirely, but rather becomes recognition vocabulary. Near the end of the attrition process, production vocabulary appears to ‘bottom out' first while recognition vocabulary continues to decrease. The data from this research Show that at the end of the 18—month period the ratio of production to recognition vocabulary increased slightly. Retention rates for both recognition and production vocabulary fell rapidly at first and then gradually leveled 217 off. These results are consistent with both Bahrick (l984c) and Weltens et al. (1989), though vocabulary decline in the present study was more drastic. Bahrick's data Show that attrition continues for up to 6 years and then what is retained at that point continues to be remembered in ‘permastore' for at least 25 years. In contrast to the present research, Bahrick's data base (and Weltens' as well) included groups of individuals who had reached a high level of proficiency. Bahrick remarks that content from a single course is not likely to endure at all, a point also made by Clark and Jordan (1984). In his ‘critical threshold' theory, Neisser (1984) states that information must be tied into an extensive cognitive structure if it is to be resistant to forgetting. This kind of critical threshold cannot be supplied by a single language course or by a few repetitions of vocabulary. The data from the present study Show that after 18 months, eleven of the fifteen students were producing less than ten percent of the vocabulary that they had been taught, and five students were recognizing ten percent or less. Only four students could produce more than twenty percent of the words and recognize more than half. It is difficult to predict what would have happened if the testing had continued longer than eighteen months, but for most students it is probable that little vocabulary would remain over a longer period of time or ever reach Bahrick's ‘permastore' state. 218 In the earlier section on student characteristics, the correlation between earlier and later retention scores was discussed and the question was raised of whether early recognition or production scores are the best predictor for future retention. This same question can be examined from a different perspective by studying the data file organized by word. Again the correlations are quite close, but this time the data file is organized around 100 items instead of just 15. The correlations are not as high overall as in the data file organized by student (for example, the correlation coefficient between R3 and R4 is .84 and P3 and P4 is .76, rather than .92 and .93), showing that a strong word is not quite as likely to continue with a high score as a strong student is to continue with a high score. The correlations from this data file slightly favor recognition scores as a predictor of later recognition, and production scores as a predictor of later production. Thus there is no clear-cut indication of the superiority of production over recognition for predicting future retention scores. That production is not necessarily superior to recognition for predicting future scores is borne out by the other tests (see Tables 34 and 37). It seems that production and recognition Skills, although very related, still tap different learner capabilities. 219 IMPLICATIONS OF RESEARCH For second language acquisition theory An overview of current second language acquisition theory was presented in Chapter 1, with discussion centering around Krashen's acquisition/learning distinction and the relationship between comprehension and production and the value of output. In Chapter 2, cognitive theory was discussed with its emphasis on how the language learner proceeds from rule-bound declarative behavior to an automatic procedural stage. According to Krashen's Monitor Model, acquisition is a subconscious process that results from meaningful interaction in a communicative setting. Learning, on the other hand, is a conscious process that comes about when the focus is on the form that the language takes. Cognitive theory, on the other hand, sees the acquisition/learning distinction as unnecessary. Even automatic processes can be brought to consciousness. O'Malley et al. (1987:302) state that ‘a cognitive model of language learning sees active conscious processes involved in all language settings, at least in the initial stages of learning. These processes can be described and used to assist learning instead of relegated to inaccessible unconscious mechanisms'. 220 The data in the present study provide evidence that adult second language learners are highly aware of the process of vocabulary acquisition, even when the Natural Approach is used in the classroom. Although all words were presented in Spanish during the research through a comprehension approach, students wrote down accompanying English translations in their notebooks without prompting from the instructor. Several students, on their own initiative, made flashcards with English translations on the back, even when the instructor urged them to avoid English or to use pictures. Also, the formation of mnemonic devices by the students was a highly intentional activity. The descriptive data from the research Show that some students were still using these same devices eighteen months later. In terms of cognitive theory, the fairly rapid attrition of vocabulary in this study Showed that students had not yet turned their declarative knowledge of words into procedural form. This failure can be attributed to the short duration of the learning period, in which the students did not have adequate time to strengthen their vocabulary skills. Although both the Monitor Model and cognitive theory strongly encourage classroom practice with communicative activities, only the cognitive theory includes an explanation for the rapid attrition of foreign language vocabulary. The data collected in this research have a good deal to say about what constitutes high quality teacher input and 221 consequent learner intake. Whether or not a student retains a word from a beginning language class does not seem to depend on if the teacher repeats the word 5 or 20 times, but rather on the saliency of the foreign language word. This saliency can depend on the form of the Spanish word itself (its concreteness, its length, its morphological transparency, or its similarity to an English sound), or saliency can depend on what the teacher does to make the word conspicuous by presenting a picture or Object, by telling a story or joke, by having a student perform an action, or by tying the word into a network of meaning. The students can also make the word salient for themselves by constructing memory devices such as keywords, by going back to study a word that was especially difficult before, or by analyzing the structure of a word. The data contain evidence that recognition and production, although closely related, are somewhat different skills in vocabulary acquisition and retention. High recognition predicts high recognition later on and high production predicts high production later on. There is no evidence from the data that recognition turns into production over time (without further input), whereas production can disappear and recognition remain as the only proof of retention. The data also Show that whether or not a word has been taught and tested for recognition aaa production is an influential factor in determining the retention of the word 222 over time. The mean scores for such words were significantly different at the .01 level from the mean scores for words taught and tested for recognition only. Forcing students to study for production on tests seems to push them to analyze words at a deeper level, thus aiding in the general retention of these words. Thus, while the importance of comprehensible input is not challenged, the data point to the additional importance of output in the long-term retention of vocabulary. The data also point to the role of consciousness in initial vocabulary learning--even when the words are presented through the ‘Natural Approach'--and emphasize the need to convert declarative knowledge to procedural knowledge if long-term retention is to be achieved.. For foreign language teaching At the conclusion to a dissertation in the field of second language acquisition, readers have the right to ask ‘so what'? Although there are implications for theory, there should also be Implications for pedagogy. As cautioned by Hatch (1978), results of research must be applied with care to the Classroom. Nevertheless, the following recommendations for the classroom seem to be supported by the data in this research: 223 l. The saliency of a word in input is an important factor for later retention. The instructor can increase saliency of vocabulary by using imagery, vivid experiences, Structural analysis, and by encouraging students to use their own mnemonic devices to remember words. This implies teacher training in methodology, but also in language itself, so that the instructor will have the language skills adequate to the use of imagery, narration, and structural analysis in the classroom. 2. Vocabulary from a single course is rapidly forgotten. Students should be encouraged to take a series of courses in the same foreign language, rather than to ‘Sample' a foreign language for a short period of time. Teachers and curriculum planners should remember that students need to form networks of meanings tied into an extensive cognitive structure and that isolated words studied for a short period of time are excellent candidates for attrition. 3. Although this research in no way discredits a comprehension-first approach to language learning, it indicates that vocabulary production is also important if students are to retain words over a long period of time. Production forces the students to pay attention to all parts of a word and to analyze the vocabulary at a different I level, thus increasing the probability of long-term retention. 224 4. If the lexicon plays a crucial role in foreign language development, then attention must be focused on vocabulary acquisition in the classroom. A large vocabulary for comprehension is important, but if long-term retention is desired, some vocabulary size may have to be sacrificed in order to allow time for making vocabulary sufficiently salient, for practicing the vocabulary that has already been introduced, and for encouraging production of vocabulary that students can already recognize. 5. Vocabulary testing should reflect course goals for both recognition and production. Recognition testing requires lower levels of learning, but seems to be the most sensitive measure of retention, especially for beginning students. Production testing encourages students to process words at a deeper level and aids long-term retention. RECOMMENDATIONS FOR FURTHER RESEARCH Although research in second language attrition exploded during the decade of the 1980s, much more needs to be done. The following areas of research are suggested specifically by questions raised in this dissertation. 1. What is the best predictor for future language attrition-~early recognition or early production scores? Although the data from the present research suggest that 225 recognition is the best predictor for recognition and production for production, the research should be repeated with a larger data base. 2. When is the optimal time for predicting future attrition? The correlations in this research suggest that scores taken 2 weeks after the end of the contact period are good early predictors, and Meara (1989) suggests using the retention rate between Week 2 and Week 3 of the retention period. Are there better points in time for making predictions? 3. Can the ratio of production vocabulary to recognition vocabulary predict the state of acquisition or attrition for an individual? 4. Why are recognition and production vocabulary lost in roughly equal amounts over time? 5. Are certain parts of Speech more likely candidates for attrition than other parts of speech? Why in the present study were adjectives more quickly forgotten than nouns or verbs? Can this finding be related to categorial grammar? 6. Is the regression hypothesis true for vocabulary acquisition and attrition? 226 7. How can the acquisition process be tied to post- classroom maintenance? 8. Is forgetting equivalent to an inability to access information because of interference, or to the complete extinction of a memory trace for a word? 9. How does vocabulary attrition in a foreign language fit into a general theory of language attrition? In the answers to questions like these lies the hope of understanding more completely the process of language attrition, which is a small step toward understanding human language in general and a smaller step toward understanding ourselves. The challenge for the foreign language I profession today is not only to understand how best our students can acquire a foreign language, but also how best they can retain it. APPENDIX A STUDENT MATERIALS 10. 11. APPENDIX A STUDENT MATERIALS STUDENT QUESTIONNAIRE Name Age 3. Class rank: F Soph J Sen Grad Current address Phone Permanent address Phone Have you spoken Spanish at home with relatives? Explain. Have you Studied Spanish before? How long and where? Have you Studied other foreign languages? Which ones, how long, and where? .' What contact haVe you had with Spanish outside of the classroom? Explain please. -travel to Spanish-Speaking countries (when, where, how long?) -contact with Spanish-speaking people in U.S. (describe) -listening to radio, TV in Spanish or reading books, newspapers in Spanish (how Often?) Why do you want to study Spanish? Graduation requirement from college I like the way it sounds Interest in Latin America May be useful in my future career in Write more about why you are in this class: What specific things do you want to do and to learn in this Class? 227 228 CONSENT FORM The research being done by Professor Marilyn Bierling is part of a project for a doctoral dissertation on vocabulary acquisition and retention. There are important questions in the area of second language acquisition that affect all language students, questions such as: -Vhat is the effect of an intensive course on vocabulary development? -Vhat is the difference between a Spanish class taught in an English-speaking environment and in a Spanish-Speaking environment? —What is the relationship between the teacher's output of Spanish in class and what the student actually acquires? What stages do Students pass through in acquiring vocabulary? - -Vhat is the relationship between being able to recognize a word in a foreign language and being able to produce it? -At what rate is foreign language vocabulary forgotten? Data for this research will come from your normal written assignments in the workbook, your written tests, and your oral interviews. All of this material is required in the regular 101 courses that the professor teaches. By giving your consent, you allow the professor to use this data for research after the 101 course has been finished and your final grade has been given. If you have given your consent, she may later ask if you would like to participate in a retention Study which would involve Short testing of vocabulary retention through the fall of 1989. ' 'Data for the research will also come from tape recordings of class meetings. These tape recordings will not be available to the professor until after the final grades have been given, and only the recorded utterances of students who have given their consent will be used in the study. The research has been explained to me and I understand the basic purposes and procedures. I freely consent to participate and understand that at any time I may withdraw my consent without penalty. I understand that the Spanish 101 professor will not know if I have consented or not until after she has given the final grades for 101, and only after that time will she analyze any of the data. I understand that all results will be treated with strict confidence and that all subjects will remain anonymous; upon request, results will be made available to me as long as the anonymity of other subjects can be maintained. Name Date 229 SAMPLE ITEMS FROM RECOGNITION TESTS (FIRST 20 HOURS) Multiple choice 10. 11. 12. 13. 14. 15. Escoja 1a respuesta mas lOgica. Un hombre con un ojo es a. rubio. b. grande. c. tuerto. d. ocioso. Un insecto muy bonito es a. la faja. b. e1 gato. c. el chico. d. la mariposa. es parte de la mano. a. BI pulgar b. La sien c. La oreja d. E1 chaleco es parte del zapato. a. El traje b. La tez c. La cremallera d. E1 tacOn Un hombre con una pierna es a. cojo. b. torpe. c. verde. d. joven. 0n hombre calvo no tiene a. cuaderno. b. pelo. c. secretaria. d. huella. gDénde eSta el cartel? a. en la pared b. en la tiza c. en el ejemplo d. en el piso La planta esta en a. la pizarra. b. el nido. c. 1a maceta. d. 1a rama. Gemelo es un Sinénimo de a. meIlizo. b. casado. c. apellido. d. primo. Cuando hay un incendio, hay muchas a. solteras. b. llamas. c. becas. d. primes. ICamine en a. el sendero! b. la piscina! c. e1 estanque! d. la quimica! Hay liquido en a. enero. b. el frasco. c. la cédula. d. el hilo. 0n es un hombre corto. a. mundo b. viudo c. cascote d. enano La planta tiene varias a. camisetas. b. fechas. c. semillas. d. ciudadanias. El padre de mi abuelo es mi a. padrino. b. bisabuelo. c. tatarabuelo. d. padrastro. 230 SAMPLE ITEMS FROM RECOGNITION TESTS (FIRST 20 HOURS) (cont'd.) 13. la carpa “““ 14. 1a aguja LE] 15. la vereda (”’0’ 16. e1 estacionamiento 17. e1 hilo 18. e1 edificio Matching __ 1. el estanque ® ’._ / ' :\ ___ 2. e1 cascote _ 3. la calle NW 4. 1a aldea : 5. el globo fi® _ 6. e1 rastrillo In 6. _ 7. la piscina *fli a r ___ 8. el agujero -- -- - - :‘ - ___ 9. la pefia -] . I 41:53 ___ 10. los piojos ' ' ' _ 11. la aguanieve I a I! I I ___ 12. la dehesa ' 6?) 19. la navaja w 20. la verja 3‘ fifl :Jmf H O 0 '1 0) E v-i 'v w v Hoocvasuusw Tequese.. la boca. . la quijada. HO @3/ u & e1 pufio. 1a nariz. GD . la sien. . el pie. . la espalda. . el mefiique. tron mm on a'n) Apunte las partes del animal. a. la garra b. el espinazo c. e1 pescuezo Silbe. Guifie el ojo derecho. Levante la mano izquierda. Ronque. Ande de puntillas. Pellizque a1 chico. Levantese y dé una vuelta a la derecha. . Bostece. 231 PRODUCTION QUIZZES (FIRST 20 HOURS) £Cd~o se llama astedl exA'g.¢IT0:#1 I. gag: write in Spanish, including the pronoun. (8) I am ° 3° we are you are (informal) you are you are (formal) (plural) II. Tener: Write in Spanish, including the pronoun. (10) I have I we have you have (informal) he has they have III. Escriba lo contrario (the opposite). (16) alto a la derecha guapo escuchen gordo levihtense joven miren abajo _gngln grands corto agudo lacio el hombre lg, negro Buenos dfas. no IV. ascriba las partes del animal. (4) a. b.. c. d. e. f. g. 232 -gl- (fist) (estémago) Sscriba las partes del cuerpo humano. (l h. i. j. k. 1. la la la la la PRODUCTION QUIZZES (FIRST 20 HOURS) (cont'd.) VI. LII-huh)!“ sees 10. 11. 12. I 13. 15. Escriba en espaHol. (15) the page la lame shy lazy sharp, clever one-eyed gentle also this 41th very walk! I Good morning. How are you? ‘77 (skin, complexion) 233 PRODUCTION QUIZZES (FIRST 20 HOURS) (cont’d.) 6C6mo se llama usted? (EXAMENCITO #2 > G Cua’l es la fecha?. I. El verbo estar (include the pronouns) (8) I am we are nosotros estamos you are (informal) she is . they are II. Escriba los mandatos. (8) Squeeze); I Answerzz : Winsz I Whistlel; : Lower your hand! Yawn); 1 I la mano ! Describe); : Snore); : 1 III. Write the adjective to describe the picture. (12) 6M) ‘10— — IEHVLIQI‘)34IE r7»:— 1. (C1911 ‘. §-\ 1 “—0 ('2 IV. Ssgriba los sindhimos. (7) Escriba 1o contrario (o o ite). hurifio adelantado el gemelo e1 e1 mediodfa la pap‘ el mayor meme la v. La familia (£111 in the blanks) (8) otra vez - again 1. Anita es mi madre. :11a es divorciada, pero se casa (marries) otra vez con Victor. Victor es mi . Y YO SOY e1 (la) de Victor. 2. Carlos es mi padre. El es viudo (widower), pero se casa otra vez con Maria. Maria es mi . 3. El padre de mi abuelo es mi , y el abuelo de mi abuelo es mi . . Todos (All) estos hombres son mis . y yo soy el de ellos. 4. Los hijos de mis padres son mis 234 PRODUCTION QUIZZES (FIRST 20 HOURS) (COnt’d.) -2- VI. Escriba en espaHOl. (23) l. the godfather e1 2. the godmother la 1 3. the question a 4. the year el 5. the boyfriend el 6. late 7. You're welcome. 8. our 9. his/her 10. 40 ll. 60 12. 8O VII. Mire a la profesora. (15) 1. la 9. el 2. el 10. la 3. el 11. la 4. e1 12. la 5. la 13. e1 6. la 14. e1 7. el 15. el 8e 13 235 RETENTION TEST (FOR PRODUCTION) {m VOCABULARY Rm DATE 11% AOJECTIVES Zolazy 31. 3.3MY‘G 32. f;.:.¥ 4.5” («made») 33. :w Pd“ 34. 5.6:!”1 Q: (1.003,..th 35. pg: .rlag ma .37.?e! as. E '3 39. 4...! la 40. '3 410 k e. 42. gig 43. (W“)‘. _ 44. @, ® el 51.=gp5 la ”.ma 53. -———->/ a}? 55. gla be! 4”" 57. W |; 53. Russmu el S9. GERMAN e0 236 RETENTION TEST (FOR PRODUCTION) (cont'd.) 89. godmowher ’3 so. 4% Jam) 9‘ 9o. _g/‘3' 91. ' 52. ::.[®E(\ flotsam) las 63. 4:Q./f\ a: g: 9‘ _ as. 95. EGG «~60 "e! * 66. fi 96. .mmr .. __ * ‘ 67. 97.desc¢ndon1‘ e! 68. 98.5,. g, 69. v '15 990-: e. _- 70. ‘0“. WpIC-IOIOH [a 100 ...-D”... ‘3 0‘.” v " 102. fl“ in 1.03. k”. 104.@ e: IOSOW * gonmmoi 106. %: ~' . 107. 39 _z to 108. $5 '. 109. - ! nth/Ea i q 111. a, 3 112-~24 i I. 1.13. ...} 5 - ! 114.\U i. ! as. _ '3 - 115. (F? i ‘ ‘ a 87. Fret-cl _ 117. ' 7K3; 1 33, godfuher el 118. Q5 g ' - 237 WORDS IN ENGLISH LOOK HERE if you don’t know what the pictures are. soooxn O‘U‘5UNH . the . the . the . the . the I! . the . the . the . the . the . the . the . the . the . the . the . the . the . the fat lazy gentle shy curly (hair) smart, sharp, clever stupid one-eyed ragged . barefoot . moldy . wet, soaked . lame . white-haired . wrinkled . bald . the . the . the . the . the blackbird wing nest branch (animal) snout (animal) neck (animal) back (animal) claw rock path (2 ways) hole swimming pool window grating hallway flame match balloon seed poster rake board tear needle thread 42. 43. 44. 45. 46. 47. . the . the . the . the . the . the . the . the . the . the . Russian . German . the fist . the temple . the chin . the jaw . the stomach . the little finger . the thumb . the nose . the moustache . the contact lens . face, complexion . the scholarship the flower pot the Venezuelan monetary unit the clock the footprint the mud the rubbish sleet jewel apron belt vest dress earring zipper heel (high) shirt the blanket . the bottle . the bowl . the dwarf . the fire . the basket . the sign . the deer (2 ways) 1' . the whale . the shark . the butterfly . the cat 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. . the . the . the . the . the . the . the . the . the . the . the . the . the . the . the . the window blade (razor) tile godfather godmother son grandson twins stepmother stepfather stepson ancestor descendant great-great— grandfather . the great-grand- father . the tent . the meadow . the hill (2 ways) N pond village Tie! Squeeze/Wring! Wink! Yawn! Pinch! Fold! Snore! Walk! Open! Whistle! Walk on tiptoe! Cross out! Pray! 238 RETENTION TEST (FOR RECOGNITION) NAME _1 fi __20 lily 3. gCRT‘C 4. Shy (h’mido) __.5 @ pole _ —-6 ’GWb': (Iotcfi’oetc) —" CW _.N08 WW! RECOGNITION W - II tel-id: . :0 salad . fa’u’l - «Sue . ratio In piscina . (009 c! all . h".'o¢ntfl - empapado . laeio calvo . dd,ld¢ - to «an. In [uteri - c! mirlo - cojo - C! nédo . have?“ . mohoso - fl ‘3')". 831.40 . h «an . gudo major 1.. la .pcfi’a 3:. cl wind. by. ed pun-eta y‘n‘ftfiflifltyaarrL.-<7udxntn.~ ct. deacon-pact: dd. descalzo n. H San: 6‘. cl bored. ,3. la «:3: uh. “were H. care 43. arrugado kk. la‘ rant ll. “(tomato fla- l‘l (0:41.18 no. la nude. on. c! otu‘u‘. PP' fl ho‘l‘o gg- Dante rr-tuorfo 3‘. II ‘fffl‘ t!- for” an. Ciel VV' Mensa —3 "I— l. 32. 33. 34. 35. 36. DATE 13¢, :‘EEF sa. RussnAN 59. GERMAN a, la 4’83: b, 0! «adorn 6. la urbat J. .4 am... e. d rastrillo ‘. a H“. g. a! row". h. u like 3. cl la’po'a. J. to arm. x. I. but“ I. I; '0'“ n. I. «am. a. cl ’3“ o. 1! nude 9. 3! ud- ‘. ll pium r. 3! del...“ 3. la aguanieve !. cl ”I'm a. It «new. 9. at page! H. cl «nido :. c! ,1.“ y. cl sue 1. a! (log. a. d {can it. a! beh’vlr cc. c! done; 44. to fit. a. la Hm w. to In. 3,. 3! 89“.. M- 0: fl“; 3;. ll “out 5.3- n tabla n. In plane u. I. an}: on. a! pasillo no. la sill! 00. I: I’m}. pp 4! launder gg. ll under. H. a! lbfl,‘ sh ll “grin. th e! hilo uu- a! bolo'qrn'e w. u cane! RETENTION TEST (FOR RECOGNITION) (cont'd.) @6359) _70. Fae. , complexion -71. ”gun ) _83. 82% _;:;]E u.<7" _87.‘Hm 5.3713 _88. god €41I‘Cr -2- a. ll ‘0‘ . 0. cl IIuIcJO t. c! bigot; 4. cl «one c, Ia cord“) 6. II colcha 3- 13 data h. I: beca ;. e! ciervo J‘ cl incendio K. I: nan-3m I. at brlu an. I: navaja fl. '3 pierna 00 '3 5.753 P' ’3 mujdl‘ 3,- II 5300 r. c! hes-bro s. t! v.39“. ‘0 '3 p.330; 0. la bus I. la urea w. c! «not. 1. cl git. 7. l3 iuijua 1.. la nevi: 33. cl pcnd 55. I! cute 6‘.- la «an; dd. cl r69.“ 0!. II pro’uata K. d pd,er 33. c! Viento M. cl "no. N- n.4adad we,"m ' 9‘ would Cl padrino ° e! ,1“ ' M lentélla ‘0' cl tiburg'a 9" I! pi.- SS'II panza rr. la wooing u. I: ballena '3' '3 mesa uu. la ventaua ' h "Clips; 2 39 godmowhcr' :92 m... _93. “(7% @ gm _::NQ$V 95. ancestor 97. deoceno'anf 98. .. I,I __"99. . —-1°° .52.. max-- .. -‘°" ml ...- 122' N '90 3- cl prune h. .1 mafia“ Go Hot! . d.“ hSJ'o . ’3 main . 53607:! . 36000.“! C i: 5 h. e! tatarabuelo 3- “ puns". j. e! “cum k. e! «no... I. e! vista’o “h iAu-ure! n. It undue": 9- 3‘00601! P° I. In; $' " ”4.3.. r S t a V w I 7 . 0| nieve . G! Ill-um. . mo“! . l! dehesa . e! uh 0 55.53.! - 38496“! - Welling-e! ‘0 iEI'rini! 33. ll VCCQGI’. bk. iAbrd! “o 503,. ! “o C. bilfihtl. a. I: and... ff- Mat-e! ”- e! ”amd. {‘P‘ 0"“. do Pun-CNN no la "n‘. irmue 3° iTe’uef II. It carp. mn- fl ”55.4. on. I: an... 00. h libren’. PP° ""8303! 1,1 iE3¢u¢h¢'. I'V- 353.155,, 38. 36334 ‘. “o :0 (one 80- in'-urn! W. ”out! 240 -fi.‘.x 2 . i /// . #44121 ":“’::,\i '\ m fit ’ ., A, 7‘1 - "“:’—’/ ~m‘ ‘wfg' ' gnmt... l / \ “‘ w E!!! y \‘\\ \/ l l. A f’ > JFQ I FIGURE 22 PICTURES FOR FINAL ORAL INTERVIEW 241 EVALUATION OF SPANISH 101 You have been taught with what can best be described as a modified 'Natural Approach.’ Emphasis during your first semester of Spanish has been on understanding the spoken language, though we have also done some speaking, reading, writing, and grammar study. 1. Have you studied a foreign language before? If so, for how long? How does this class compare to other foreign language study you have done? (more positive experience, about the same, more negative experience, etc.) 2. How do you feel about the amount of Spanish you have learned this semester? (more than you expected, less than you expected, etc.>) 3. What are your feelings during the class period? (bored, scared, challenged, interested, etc.) 4. How does the workload in this course compare to other introductory courses? the difficulty? the grading? 5. How would you describe your own effort in this class? (both in class and outside of class) 6. Please comment on the following components of the course as learning tools: -text (vocab lists, readings, blue pages, actividades) -lab and workbook -quizzes and tests -other classroom activities (singing, bringing in native speakers, cultural points, etc.) - 7. The one thing I appreciate most about Spanish 101 is... The one thing I would change in Spanish 101 is... 8. Any other comments would be very much appreciated! APPENDIX B ADDITIONAL DATA AND STATISTICAL RESULTS APPENDIX B ADDITIONAL DATA AND STATISTICAL RESULTS TABLE 40 DATA BY STUDENT "K""§U'Ffi'ii'fii'fi2‘92‘fi3'93‘fi2‘94 *8 c D 2"? G a I J 01 81 56 7o 35 56 43 42 10 23 05 | 1 03 1 1 1 2 2 2 1 | 02 84 61 80 4o 79 42 71 19 6o 20 1 01 1 2 1 3 2 2 2 | 03 84 54 65 23 47 12 38 02 18 02 1 01 1 1 1 2 1 2 1 | 04 61 47 44 20 22 13 07 04 17 03 1 05 2 1 1 2 2 1 1 | 05 96 82 98 68 92 38 52 09 1 01 1 1 1 2 1 2 1 | 06 76 36 6o 06 42 06 15 02 07 03 1 03 1 2 1 3 2 2 2 | 07 6o 31 58 46 42 21 33 10 36 06 2 10 2 2 2 4 1 3 2 | 08 80 39 41 19 36 14 20 02 10 04 2 05 2 1 3 4 1 3 2 | 09 83 61 57 3o 52 20 3o 06 28 06 2 05 2 1 3 4 1 3 2 | 10 7o 35 06 01 2 07 2 2 2 4 2 3 2 | 11 65 47 31 12 06 01 2 07 2 1 3 4 2 4 1 | 12 72 42 74 39 84 44 72 24 59 19 2 07 2 2 3 5 1 3 1 | 13 99 95 77 64 55 23 36 19 46 26 2 03 1 2 2 4 2 4 2 | 14 88 81 84 39 57 17 08 01 08 02 2 02 1 1 1 2 2 3 1 | 15 99 93 98 89 96 77 84 25 69 23 2 01 1 1 1 2 1 1 2 | KEY A student number R0 baseline recognition score PO baseline production score R1 first recognition test score P1 first production test score R2 second recognition test score P2 second production test score R3 third recognition test score P3 third production test score R4 fourth recognition test score P4 fourth production test score B location of class: 1) Mexico, 2) U.S. C course grade: 1) A, 2) A-, 3) 3+, 4) B, 5) B-, 6) C+, 7) C, 8) C-, 9) D+, 10) D D grade (grouped): 1) A, A-, and 3+, 2) B- and lower E effort: 1) much, 2) some F motivation: 1) enrolled because of interest, 2) enrolled because of -interest and requirement, 3) enrolled to meet requirement G effort + motivation H previous contact with Spanish: 1) yes, 2) no I later contact with Spanish: 1) yes, in class and culture, 2) yes, in culture, 3) yes, in class, 4) no J sex: 1) female, 2) male 242 R 0 $6 $3 $7 $6 $7 R 0: $2 $1 00 $7 0% $0 OM $5 0% $4 on $4 $3 05 $2 $0 $7 M6 $4 $7 $7 m0 0$ 00 $5 0% 0$ on $2 $5 $8 m0 $4 $4 $2 $2 $2 M6 $4 M6 $4 $2 $7 0$ P 0 OH $8 0U $4 ON 0“ OH 0% O$ OH 0% 0% $3 on M6 0% $0 0$ on 0U 0U M6 0% 0% M6 0$ 0” 0H 0N $4 0D $3 on 0$ $7 0” M6 0U 0$ Ofl 05 Ofl $8 M2 M3 mnm41 WWHHU$D R 0: OH $0 0% on 05 0% 0% ON 0% 0M 0$ M5 on ON 0% 0% $0 0% 0% $4 0% M0 on $0 on $5 0% 03 0% 0% 03 m0 0“ on $0 $0 on $0 0“ Ofi 0% ON ON 0% R o 0% 0% on $9 on $0 $0 OH OH M3 0“ 03 0B 0W 0% OH 0% $0 on $0 0” 0% $0 OH $7 ON OH OH Ofl M0 0” $9 on Ofl $5 $5 $0 on 0” M0 $0 $0 0% 0% P N m0 m1 m3 om m7 0m m0 on ma $4 m0 0m m0 m7 m3 on on m7 m7 m3 on on on 05 cm w HHHHHHHHHHHHHHHI—‘HHHHHHHHHHHHHHHHHHHHHHHHHHHH Q NNNNNNNNHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH NNNHHt-u-awawwwwwwwwwwwwwwwwwwwwwwNNNNNNNHHHHH U NNbUG§OWG§§§§Gb5O mabaoo¢>mmmm¢bb§b NNHNNNNHNNNNNHHHH NHNHNHHHHNHHN wNNNNNNunuwwuwNbwwwwwwwnwwbkwwb#NNNNNNNqu-‘Hw O O O I O O O O O O O C O O O O C O O O O HHH e o o o o n O O O O O O O HWNVO‘VO‘NNNNNHHU'HONk#CWOQ§OHOO\\IOUJONO‘UIOONVUIHQN HOO‘O‘OVWH“#0000\JWNOWOWUCUVCNO5UUNCWCWONUONOVCUO NHNNNMNWNNNNNUU§NWUWNNNWUHUwU-JNW#HHNNNNWNUWNW NUNMVLflGm‘t‘WOFQUOUOWmO\lU‘UOwO\O\ 00. HU'IOU1UMQOOCVN$Q8W$WONC©UNUQ§\OUNCQHHkNmU‘#‘OUm8°C QOOOOHOUUGOOUC 0 H 2.96 3.12 3.56 3.20 3.24 3.32 3.58 3.04 3.48 3.20 3.12 2.28 3.32 2.72 3.48 2.96 3.56 3.20 3.12 2.88 2.12 2.32 3.17 2.80 2.71 3.20 3.68 3.92 3.08 2.60 3.60 2.32 3.04 2.80 2.56 3.56 3.52 3.40 3.12 2.76 3.26 2.96 2.58 3.26 NHNNHNNNNHHNHNHNNNNHHNHHHNNNNNHNHNNNNNNNNNNN L4 A o . 047 048 049 050 051 052 053 054 055 056 057 058 059 060 061 062 063 064 065 066 067 068 069 070 071 072 073 074 075 076 077 078 079 080 081 082 083 084 085 086 087 088 089 090 091 092 093 094 095 on 087 080 087 100 087 093 093 073 087 060 067 087 100 o: 030 057 057 057 090 070 063 070 060 053 050 060 093 0.. 050 071 071 086 079 086 086 064 071 079 064 071 093 086 100 043 o 025 054 036 046 050 050 064 036 054 039 039 043 061 064 082 039 ... 046 062 050 069 064 075 069 042 058 058 042 067 075 064 091 069 TABLE 41 (cont'd.) R0 P0 RI PI R2 P2 R3 P3 R4 P4 F - u o . 008 027 023 027 029 042 054 021 033 033 023 046 050 064 055 031 019 000 055 046 015 019 031 042 000 018 027 015 019 031 004 027 000 008 025 008 012 008 008 031 012 004 013 008 015 012 023 004 012 027 027 025 055 064 033 067 042 244 . . 009 005 000 000 000 000 025 000 008 013 000 021 017 018 036 017 008 000 014 013 008 008 010 015 004 027 000 008 000 025 0 0 021 021 020 036 038 033 060 020 020 036 029 029 042 062 067 027 o 000 007 010 004 013 007 030 003 W UIUIUUIUUJ-‘b§####k5§§#wawwwwwwwwwwwwNNNNNNNNNNNNNNNN NNNNNNNNNNNNNNNNNNHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHI—I O 620002000000020000wwwwwwwwwwwwwwwuwwwwwwuwwwwuwwwwwwuwwnNM U auuwwmwm~6~6~uw5~bbo~ #NNNNNO\#&‘UQUU5¢§G©O§U5§NN N HHHHHHNHH HNHNHNH HHHH N HHHHN nnwwwwwwwwwwuwwwwnnnmwnwwaanNwNwwwwnwwwwbwhwwwn O O O O O O O O O O O O O C O O O O O C O . HHHHHHHH HCWVOWWHO‘bh-‘DVWVOOOUWNNVOVON OHO‘QU‘VWHOOO‘O‘OOOVVO‘WO NNNHNNWWNWUWU’NNNUWWMNUUNNWUU’WU’WWWWWWWNWUWUWN#U-INNN O O O O O O O O O C O O O O C O O O O O O O O O O O O O O O WNWNNNNwwNWNUNWWUUWU’NWU’UWNNNNNHWN WWNUWN WWNHUOUJWN O O O O O O O O O O O I O O O O O O O C O O O O O O O O O O O O O O O NNHHHHNH <_. NNHNNH NHNHHHHNNHNHNHNNNNNNHNNNNHHHHHHNH A 0°. 097 098 099 100 245 TABLE 41 (cont'd.) R0 ioééii *91 82 22 +83 P3 +34 94 B 6‘5‘2‘?"§* H I J I 0 ' I .0 O 00' H: 000 H . .° . 093 086 050 067 038 050 017 036 009 5 2 3 3 1 3.36 2.88 2.56 1 073 043 021 039 000 008 000 013 000 5 2 3 6 2 2.92 2.00 2.68 1 087 093 057 100 055 044 011 045 013 5 2 3 6 2 2.57 1.70 1.68 1 053 029 018 031 012 009 004 000 004 5 2 3 5 2.44 2.44 3.27 2 (uh-HI! Q'fl NU nmwwmamgmwmwn» bbw N HHOO KEY word number (see Table 4 for Spanish words with English translations) baseline recognition score baseline production score first recognition test score first production test score second recognition test score second production test score third recognition test score third production test score fourth recognition test score fourth production test score paso: 1) Paso A, 2) Paso B, 3) Paso C, 4) Paso D, 5) Paso E recognition and production: 1) taught and tested for both recognition production, 2) taught and tested for recognition only part of speech: 1) verb, 2) adjective, 3) noun frequency in input: 2) very low (1-4 times), 4) low (5-9 times), 6) medium (10-19 times), 8) high (20+ times) frequency in input (grouped): 1) low (1-9 times), 2) high (10+ times) emotionality of English word (5--1owest emotionality, 1--highest emotionality) saliency of English word (5--lowest saliency, 1--highest saliency) saliency of Spanish form (5--lowest saliency, l--highest saliency) saliency of Spanish form (grouped): 1) high saliency, 2) low saliency 246 CORRELATIONS 0F SCORES AT TIMES 1 AND 4 R4 | 1| 2| 3| 4| 5| 6| 7| 8| 9| i|| R1 FIGURE 23 ,\ SCATTERGRAH FOR R1 AND R4 R4 = ~15.631 + .683(R1) / r a .689 5| ' P4 FIGURE 24 ,\ SCATTERGRAM FOR 91 AND P4 P4 . -1.853 + .280(P1) / r . .617 P4 R4 247 CORRELATIONS 0F SCORES AT TIMES 1 AND 4 (cont’d.) u r «.26!- 50.531- l 31.19}. 27.361. 21.121- 1 15.5:1- I 9d“?— I 4.112:- 9| N 6| 5| 4| 2| fl ..., a 4.62?- to z: 3: 40 s: so 7: v: 9: 1:: R1 ’ FIGURE 25 ,\ SCATTERGRAH FOR R1 AND P4 P4 a -7.36 + .242(R1) / r a .532 P1 FIGURE 26 ,\ SCATTERGRAM FOR P1 AND R4 R4 8 8.73 + .556(P1) / r a .564 248 TABLE 42 PREDICTIONS FOR R4 AND P4 (BASED ON R1 AND P1) A A A B R1 P1 R4 P4 R4 C P4 001 pellizque 071 037 027 007 32.4 -05.4 08.9 -01.9 002 ande de 086 018 043 000 39.9 03.1 05.5 -05.5 puntillas 003 rece 093 054 033 021 47.5 -14.5 14.2 06.8 004 pliegue 057 032 020 003 23.4 -03.4 06.8 -03.8 005 tache 086 071 029 029 44.7 -15.7 17.6 11.4 007 cojo 057 039 020 010 24.1 -04.1 08.4 01.6 008 tuerto 064 025 020 000 27.1 -07.1 05.6 -05.6 009 manso 064 057 015 014 30.0 -15.0 13.0 01.0 010 hurafio 064 043 013 000 28.7 -15.7 09.8 -09.8 011 ocioso 064 032 043 004 27.7 15.3 07.2 -03.2 012 agudo 071 054 014 000 34.0 -20.0 12.7 -12.7 013 torpe 086 036 029 018 41.5 -12.5 09.6 08.4 014 delantal 036 025 013 000 09.9 03.1 03.7 -03.7 015 faja 086 046 007 007 42.4 -35.4 11.9 -04.9 016 cremellera 093 043 033 003 46.5 -13.5 11.7 -08.7 017 aretes 071 039 031 011 32.6 -01.6 09.3 01.7 018 tacén 086 054 046 019 43.2 02.8 13.8 05.2 019 chaleco 057 021 020 007 22.4 -02.4 04.2 02.8 020 alhajas 071 032 020 007 32.0 -12. 07.7 -00.7 021 mefiique 079 043 050 003 37.9 12.1 10.8 -07.8 022 tez 079 057 029 014 39.2 -10.2 14.0 00.0 024 pulgar 071 057 050 013 34.3 . 15.7 13.4 -00. 025 panza 079 039 015 015 37.5 -22.5 09.9 05.1 026 barbilla 079 050 027 010 38.5 -11.5 12.4 -02.4 027 quijada 057 014 007 000 21.8 -14.8 02.6 -02.6 028 sien 086 046 013 003 42.4 -29.4 11.9 -08.9 029 hocico 050 039 027 000 19.8 07.2 07.9 -07.9 030 espinazo 093 043 033 007 46.5 -13.5 11.7 -04.7 031 garra 086 061 040 007 43.8 —03.8 15.4 -08.4 032 pescuezo 050 014 033 003 17.5 15.5 02.2 00.8 033 mariposa 093 079 089 044 49.8 39.2 20.0 24.0 034 ciervo 057 021 021 007 22.4 -01.4 04.2 02.8 035 venado 064 036 031 008 28.1 02.9 08.2 -00.2 036 ballena 100 079 075 013 54.0 21.0 20.4 -07.4 037 tiburén 100 079 085 032 54.0 31.0 20.4 11.6 038 exprima 086 096 050 032 47.0 03.0 23.4 08.6 039 guifie 050 029 020 007 18.9 01.2 05.6 01.4 040 ronque 050 032 039 025 19.1 19.9' 06.3 18.7 041 silbe 043 039 040 007 15.5 24.5 07.4 -00.4 042 bostece 064 018 020 003 26.4 -06.4 04.0 -01.0 043 descalzo 057 043 020 000 24.4 -04.4 09.3 -09.3 044 calvo 086 075 036 004 45.1 -09.1 18.6 -14.6 045 cano 071 057 020 013 34.3 -14.3 13.4 -00.4 046 mohoso 086 071 050 025 44.7 05.3 17.6 07.4 047 harapiento 050 025 021 000 18.5 02.5 04.7 -04.7 249 TABLE 42 (cont'd.) /\ A B R1 P1 R4 P4 R4 c PE D 048 arrugado 071 054 021 007 34.0 -13.0 12.7 -05.7 049 empapado 071 036 020 010 32.3 -12.3 08.6 01.4 050 maceta 086 046 036 004 42.4 —06.4 11.9 —07.9 051 tabla 079 050 038 013 38.5 -00.5 12.4 00.6 052 colcha 086 050 033 007 42.8 -09.8 12.8 -05.8 053 cuenco 086 064 060 030 44.1 15.9 16.0 14.0 054 rétulo 064 036 020 003 28.1 -08.1 08.2 -05.2 055 cartel 071 054 020 000 34.0 -14.0 12.7 -12.7 056 pasillo 079 039 036 014 37.5 -01.5 09.9 04.1 057 huella 064 039 029 000 28.3 00.7 08.8 -08.8 058 fango 071 043 029 010 33.0 ~04.0 10.2 -00.2 059 ala 093 061 042 008 48.1 -06.1 15.8 -07.8 060 rama 086 064 062 019 44.1 17.9 16.0 03.0 061 mirlo 100 082 067 015 54.3 12.7 21.1 -06.1 062 nido 064 039 027 013 28.3 -01.3 08.8 04.2 063 antepasado 079 021 064 017 35.9 28.1 05.7 11.3 064 vastago 043 014 000 000 13.2 -13.2 01.7 —01.7 065 bisabuelo 086 050 036 017 42.8 -06.8 12.8 04.2 066 tatarabuelo 079 039 043 013 37.5 05.5 09.9 03.1 067 padrastro 079 029 053 000 36.6 16.4 07.6 -07.6 068 madrastra 064 025 053 007 27.1 25.9 05.6 01.4 069. padrino 086 032 067 019 41.2 25.8 08.7 10.3 070 madrina 079 043 038 019 37.9 00.1 10.8 08.2 071 mellizo 057 021 033 003 22.4 10.6 04.2 -01.2 072 lagrima 079 050 018 014 38.5 -20.5 12.4 01.6 073 beca 071 057 014 007 34.3 -20.3 13.4 -06.4 074 cesto 043 014 027 007 13.2 13.8 01.7 05.3 075 incendio 057 021 040 000 22.4 17.6 04.2 -04.2 076 llama 057 036 033 030 23.8 09.2 07.7 22.3 077 cerilla 050 014 007 000 17.5 -10.5 02.2 -02.2 078 aguanieve 093 036 087 023 45.8 41.2 10.1 12.9 079 estanque 043 014 000 000 13.2 -13.2 01.7 -01.7 080 dehesa 043 014 020 003 13.2 06.8 01.7 01.3 081 loma 036 025 021 007 09.9 11.1 03.7 03.3 082 cerro 021 011 000 000 -00.5 00.5 -00.5 00.5 083 vereda 014 007 000 003 -05.2 05.2 -01.9 04.9 084 sendero 036 007 013 000 08.3 04.7 -00.4 00.4 085 agujero 050 007 000 000 16.8 -16.8 00.6 -00.6 086 carpa 071 054 047 007 34.0 13.0 12.7 -05.7 087 aldea 021 007 013 000 -00.9 13.9 -01.4 01.4 088 verja 029 004 000 000 03.7 -03.7 -01.5 01.5 090 bolivar 057 007 033 013 21.1 11.9 01.0 12.0 091 rastrillo 071 025 047 000 31.3 15.7 06.1. -06.1 092 enano 050 018 013 000 17.8 -04.8 03.1 -03.1 093 azulejo 064 036 020 003 28.1 -08.1 08.2 -05.2 094 navaja 036 000 000 000 07.6 -07.6 -02.0 02.0 095 cascote 043 011 007 000 12.9 -05.9 01.0 -01.0 250 TABLE 42 (cont'd.) A xx A 8 R1 P1 R4 P4 R4 C P4 D 096 hilo 029 036 000 007 06.6 -06.6 05.8 01:2 097 semilla 086 050 036 009 42.8 -06.8 12.8 -03.8 098 frasco 043 021 013 000 13.8 -00.8 03.3 -03.3 099 globo 093 057 045 013 47.7 -02.7 14.9 -01.9 100 aguja 029 018 000 004 05.0 -05.0 01.7 02.3 KEY word number Spanish word first recognition test score first production test score fourth recognition test score fourth production test score predicted score for fourth recognition test error in prediction for R4 predicted score for fourth production test error in prediction for P4 251 TABLE 43 T-TESTS FOR PRE-COURSE CONTACT WITH SPANISH TEST N MEAN SD T VALUE 1—TAIL PROB RO'SOEE‘EONTACT* 7 82.0 13.4 :60 .28 N0 CONTACT 8 78.0 12.6 P0 SOME CONTACT 7 57.4 23.0 .02 .49 No CONTACT 8 57.3 21.3 R SOME CONTACT 7 70.1 21.5 .58 .29 NO CONTACT 7 63.7 20.9 PI SOME CONTACT 7 44.9 25.5 1.15 .14 NO CONTACT 7 30.9 19.8 RT‘SOEE‘CORTAET* 7 64.1 25.5 .97 .185 NO CONTACT 6 51.8 18.9 T2“SOEE'CONTACTT 75 32.3 23.1 .75 .24 No CONTACT 6 24.0 15.4 RS‘SOEE‘EONTACT 6 46.2 25.6 1.12 .14 N0 CONTACT 6 29.8 24.9 PE‘SOEE‘EORTAET* 67* 11.5 10.5 .43 .34 N0 CONTACT 6 9.2 8.2 RZ"§OEE‘CONTACT* 7 38.9 21.9 1.57 .07 No CONTACT 8 21.6 20.6 PZ‘SOEE'EONTAET 7 9.9. 8.0 .48* .32 NO CONTACT 8 7.6 9.7 TABLE 44 ONE-WAY ANALYSES 0F VARIANCE FOR POST-COURSE CONTACT WITH SPANISH TEST R0 P0 R1 P1 R2 . P2 R3 P3 R4 P4 F .39 .86 .55 .36 .11 .55 .12 .50 .33 .47 P PROD .76 .49 .66 .79 .95 .66 .95 .70 .81 .71 252 TABLE 45 T-TESTS FOR DIFFERENCES BASED ON SEX TEST N"’ MEAN SD TTVALUE 24TAIL PROB R FEMALE 7 78.1 12.7 -.48 .64 MALE:_ 8 81.4 13.4 TU'TEMALE *7 58.4 16T4 .18 .86 MALE 8 56.4 25.9 R1 FEMALE 7 66.6 22.9 —.06 .95 MALE 7 67.3 18.9 PT'TEMALE *7 33.7 18.3 -.66 .52 MALE 7 42.0 27.9 RT'TEEKTE: 6 59.7 25.4 .17 .87 MALE 7 57.4 22.1 PT‘TEEATE 6* 27.8 15.4 -.10 .92 MALE 7 29.0 23.8 RE‘TEEKTE 5 33.4 27.1 -.51 .62 MALE 7 41.3 26.1 P FEMALE 5 8.2 9.5 -.67 .52 MALE 7 11.9 9.2 RZ‘TEMALE 7 26.1 20.9 —.56 .59 EALE4_7 8 32.8 24.4 P4 FEMALE 7 5.9 6.4 -1.19 .26 MALE 8 11.1 10.1 253 TABLE 46 T-TESTS FOR DIFFERENCES BASED ON EFFORT TEST N *EEAN SD TTVALUE 1-TAIL PROB RU'EEEE‘EETOET: 9 81.9 12.6* .74 .24 SOME EFFORT 6 76.8 13.4 P MUCH EFFORT 9 62.2 18.7 1.10 .15 SOME EFFORT 6 50.0 24.5 ET'EUCE'ETEOET 9 6513 24.5 —.38’ .35 SOME EFFORT 5 69.8 10.1 ET‘EEEE‘ETTOET 9 37.2 25.4 -.13 .45 SOME EFFORT 5 39.0 21.0 EE"MU6E‘ETF0RT 8 57.3 25.5 -.23 .41 SOME EFFORT 5 60.4 20.1 EE‘EEEE‘EETOET 8 29.3 22.6 .18 .43 SOME EFFORT 5 27.2 15.9 R3 MUCH EFFORT 7 32.7 726.4 -.84 .21 SOME EFFORT 5 45.4 25.1 P3 MUCH EFFORT 7 7.1 8.5 -1.52 .08 SOME EFFORT 5 14.8 8.8 EZ'Eficfi"ETTOET 9 25.7 21.4 -.84 .21 SOME EFFORT 6 35.7 24.3 EZ'EEEH EFFORT 9 6.1 628 -1.45 .09 ' SOME EFFORT | 6 12.5 10.4 TABLE 47 ONE—WAY ANALYSES 0F VARIANCE FOR MOTIVATION TEST R0 P0 R1 P1 P2 R3 P3 R4 P4 F .73 .84 2.30 1.25 .23 .17 .03 .25 .08 .13 F PROB .50 .46 .15 .32 .80 .84 .97 .78 .92 .88 TABLE 48 T-TESTS FOR DIFFERENCES BASED ON MOTIVATION TEST __gj N MEAN SD T VALUE 1-TAIL PROB R0 MOST 8 83.6 11.9 1.29 .11 LEAST 4 75.0 8.1 POTMOST 8 63.8 19.6 1.56 .08 LEA_ST 4 47.3 9.7 R1 MOST 8 *74.9 18.8 2.09 .03 ‘f LEAST 4 50.8 18.8 P1 MOST 8 40.0 26.9 1.05 .16 LEA§T 4 25.0 11.9 R2 MOST 8 61.4 725.7 .23 .41 LEAST 3 57.3 24.4 EE‘EOST 8 31.0 23.7 .33 .37 LEAST 3 26.0 15.9 EE‘EOST 7 37.9 30.6 —.14 .45 LEAST 3 40.7 27.6 ES'EOST 7 9.0 9.5 —.24 .41 LEAST 3 10.7 11.7 R4 MOST 8 31.8 24.7 .40 .35 LEAST 4 25.8 24.1 P4 MOST 8 8.4 8.5 .17 .43 LEAST 4 7.5 7.9 MEANS FOR SCORES BY PASO 255 TABLE 49 1 PASO B C D R0 *78.6 89.3 77.8 70.6 74.9 P0 53.8 65.1 53.9 —- _- R1 73.9 72.6* 67:3 41.5 54.6 P1 43.0 49.7 31.1 18.8 25.4 R2 66.4 6214 57.0 36.1 51.4 22 35.0 35.4 22.7 13.0 18.4 R3 42.5 42.8 42.3 25.2 23.1 P3 13.8 11.3 9.3 4.9 3.6 R4 31.5 34.2 35.1 18.7 19.5 P4 .8 10.8 10.2 3.6 4.5 TABLE 50 CORRELATIONS 0P TEST SCORES VITH EMOTIONALITY OF ENGLISH TRANSLATIONS TEST R0 P0 R1 P1 R2 P2 R3 F3 R4 P4 r -.06 -.03 -.03 -.03 —.04 -.05 .03 -.20 —.02 -.11 PROB .27 .39 .40 .37 .36 .32 .37 .02 .44 .13 TABLE 51 CORRELATIONS 0F TEST SCORES WITH SALIENCY OF ENGLISH TRANSLATIONS TEST’ R0 P0 R1 P1 R2 92 R3 P3 R4 P4 r -.15 -.03 .12 .05 .10 -.01 .17 .00 .08 .05 PROB .08 .39 .11 .31 .15 .46 .05 .48 .21 .33 LI ST OF REFERENCES LIST OF REFERENCES Andersen, Roger W. 1982. Determining the linguistic attributes of language attrition. The loss of language skills, ed. by Richard D. Lambert and Barbara P. Freed, 83-118. Rowley, MA: Newbury House. ----- (ed.) 1983. Pidginization and creolization as language acquisition. Rowley, MA: Newbury House. Anderson, J.P., and A.M. Jordan. 1928. Learning and retention of Latin words and phrases. Journal of Educational Psychology 19.485-96. Anderson, John R. 1983. The architecture of cognition. Cambridge: Harvard University Press. 4—-—-. 1985. Cognitive psychology and its implications. 2nd edn. New York: Freeman. Asher, James J. 1981. Learning another language through actions: The complete teacher's guidebook. Los Gatos, CA: Sky Oak Productions. Bahrick, Harry P. 1979. Maintenance of knowledge: Questions about memory we forgot to ask. Journal of Experimental Psychology: General 108.296-308. ------ 1984a. Semantic memory content in permastore: fifty years of memory for Spanish learned in school. Journal of Experimental Psychology: General 113.1-29. ------ 1984b. Associations and organization in cognitive psychology: A reply to Neisser. Journal of Experimental Psychology: General. 113.36-7. ----- . 1984c. Fifty years of second language attrition: Implications for programatic research. Modern Language Journal 68.105-18. ----- , and Demetrios Karis. 1982. Long-term ecological memory. Handbook of research methods in human memory and cognition, ed. by C. Richard Puff, 427-465. New York: Academic Press; Berko-Gleason, Jean. 1982a. Converging evidence for linuistic theory from the study of aphasia and child language. Exceptional language and linguistics, ed. by L.K. Obler and L. Menn, 347-56. New York: Academic Press. 256 257 ----- . 1982b. Insights from child language acquisition for second- language loss. The loss of language skills, ed. by Richard D. Lambert and Barbara F. Freed, 13-23. Rowley, MA: Newbury House. Herman, Ruth A., and Elite Olshtain. 1983. Features of first language transfer in second language attrition. Applied linguistics 4. 222-34. Brewer-Bomar, K. 1981. Second language lexical and syntactical inference in the first language of two four-year-old Spanish speakers. DAI 42.5105A. Brown, Dorothy F. 1980. Eight Cs and a G. Guidelines for vocabulary teaching, ed. by Rosemary Khoo, 1-17. (RELC Journal Supplement 3.) Brown, John (ed.) 1976. Recall and recognition. New York: John Wiley. Brown, Roger, and David McNeill. 1966. The ‘Tip of the Tongue' phenomenon. Journal of Verbal Leaning and Verbal Behavior 5. 325-37. Brown, W. P., and Dorothy M. J. Ure. 1969. Five rated Characterstics of 650 word association stimuli. British Journal of Psychology 60. 233-49. Byrnes, Heidi. 1984. The role of listening comprehension: A theoretical base. Foreign Language Annals 17.317-29. Caramazza, Alfonso, and Edgar B. Zurif (eds.) 1978. Language acquisition and language breakdown: parallels and divergencies. Baltimore: John Hopkins University Press. Carpay, J.A.M. 1975. 0nderwijsleerpsychologie en leerangontwikkeling in het moderne vreemdetalenoundervijs. Groningen: Wolters-Noordhoff. Champagnol, R. 1972. Conditions d'apprentissage du vocabulaire. Les languages modernes 66. 413- 26. Chaudron, Craig. 1985a. Intake: on models and methods for discovering learner's processing of input. Studies in Second Language Acquisition 7.1-l4. ----- . 1985b. A method for examining the input/intake distinction. Input in second language acquisition, ed. by Susan M. Gass and Carolyn G. Madden, 285-300. (Issues in second language research.) Rowley, MA: Newbury House. Clark, J.L.D., and E.H. Jorden. 1984. A study of language attrition in former U.S. students of Japanese and implications for design of curriculum and teaching materials. Washington, DC: Center for Applied Linguistics. ED 243 317. 258 Cohen, Andrew D. 1974. The Culver City Spanish Immersion Program: How does summer recess affect Spanish speaking ability? Language Learning 24.55-68. ----- . 1975. Forgetting a second language. Language Learning 25. 127-38. ----- . 1984. Studying second-language learning strategies: How do we get the information? Applied Linguistics 5.101-12. ----- . 1986. Forgetting foreign-language vocabulary. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els. Dordrecht: Foris.143-58. ------ 1987. The use of verbal and imagery mnemonics in second- language vocabulary learning. Studies in Second Language Acquisition 9.43-61. ------ 1989. Attrition in the productive lexicon of two Portuguese third language speakers. Studies in Second Language Acquisition 11.135-49. ----- , and Carol Hosenfeld. 1981. Some uses of mentalistic data in second language research. Language Learning 31.285-313. Cooper, Aldwyn, J.R., and Andrew Monk. 1976. Learning for recall and learning for recognition. Recall and recognition, ed. by John Brown, 131-56. New York: John Wiley. Carder, S.P. 1967. The significance of learners' errors. IRAL 5.161-70. Craik, Fergus I.M. 1979. Human memory. Annual Review of Psychology 30. 63-102. Craik, Fergus I.M., and Robert S. Lockhart. 1972. Levels of processing: A Framework for memory research. Journal of Verbal Learning and Verbal Behavior 11.671-84. Davies, Alan; C. Criper; and A.P.R. Howatt (eds.) 1984. Interlanguage. Edinburgh: University of Edinburgh Press. Davies, Norman F. 1976. Receptive versus productive skills in foreign language learning. Modern Language Journal 60.440-3. De Bot, Kees, and Toke Lintsen. 1986. Foreign-language proficiency in the elderly. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els, 131-41. Dordrecht: Foris. -----, and Michael Clyne. 1989. Language reversion revisited. Studies in Second Language Acquisition 11.167-77. Deno, Stanley L. 1968. Effects of words and pictures as stiumuli in learning language equivalents. Journal of Educational Psychology 59 0 202-6 0 259 Dorian, Nancy. 1973. Grammatical change in a dying dialect. Language 49.413-438. ------ 1978. The fate of morphological complexity in language death: ' evidence from East Sutherland Gaelic. Language 54.590-609. ----- . 1981. Language death: the life cycle of a Scottish Gaelic dialect. Philadelphia: University of Pennsylvania Press. ----- . 1982. Language loss and maintenance in language contact situations. The loss of language skills, ed. by Richard D. Lambert and Barbara F. Freed, 44-59. Rowley, MA: Newbury House. Ebbinghaus, H. 1885. Ueber das Gedfichtnis. Untersuchungen zur experimentellen Psychologie. Leipzig: Duncker & Humbolt. Edwards, G. 1976. Second-language retention in the Canadian Public Service. Canadian Modern Language Review 32.305-8. ----- . 1977. Second-language retention in the Public Service of Canada. Ottawa Public Service Commission of Canada. Eysenck, Michael W. 1982. Incidental learning and orienting task. Handbook of research methods in human memory and cognition, ed. by C. Richard Puff, 197-228. New York: Academic Press. Fakhri, Ahmed. 1985. Linguistic outcomes of foreign language loss: short-term changes in the interlanguage of American learners of French. DAI 46.964A. Gardner, Robert. 1982. Social factors in language retention. The loss of language skills, ed. by Richard D. Lambert and Barbara F. Freed, 24-43. Rowley, MA: Newbury House. ----- ; R.N. Lalonde; and J. MacPherson. 1985. Social factors in second language attrition. Language Learning 35.519-40. Gass, Susan M. 1987. Introduction. Studies in Second Language Acquisition 9.129-32. ----- . 1988. Integrating research areas: A framework for second language studies. Applied Linguistics 9.198-217. ----- , and Carolyn Madden (eds.) 1985. Input in-second language acquisition. (Issues in second language research.) Rowley, MA: Newbury House. Geoghegan, B. 1950. The retention of certain secondary school subject matter over the period of the summer vacation. Ph.D dissertation. Fordham University. 260 Ginsberg, Ralph B. 1986. Issues in the analysis of language loss: methodology of the language skills attrition project. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els, 19-36. Dordrecht: Foris. Godsall-Myers, Jean Elizabeth. 1981. The attrition of language skills in German-classroom bilinguals: A case study. DAI 43.157A. Gonzo, Susan, and Mario Saltarelli. 1979. Monitoring, pidginization, and immigrant languages. The acquisition and use of Spanish and English as first and second languages, ed. by Roger Andersen, 153- 163. Washington, DC: TESOL. , . 1983. Pidginization and linguistic change in emigrant languages. Pidginization and creolization as language acquisition, ed. by Roger Andersen, 181-197. Rowley, MA: Newbury House. Gregg, Kevin R. 1984. Krashen's monitor and Occam's razor. Applied Linguistics 5.79-100. ------ 1986. Review of The input hypothesis, issues and implications, by Stephen D. Krashen. TESOL Quarterly 20.116-22. Haastrup, Kirsten. 1989. The learner as word processor. Vocabulary acquisition, ed. by Paul Nation and Ron Carter, 34-46. (AILA Review 6.) Hansen, Lynne. 1980. Learning and forgetting a second language: The acquisition, loss and re-acquisition of Hindi-Urdu negative structures by English-speaking Children. DAI 42.193A. Hatch, Evelyn. 1978. Apply with caution. Studies in Second Language Acquisition 2.123-43. ------ 1983. Simplified input and second language acquisition. Pidginization and creolization as language acquisition, ed. by Roger Andersen, 64-86. Rowley, MA: Newbury House. Henning, Grant H. 1973. Remembering foreign language vocabulary: acoustic and semantic parameters. Language Learning 23.185-196. Higgs, Theodore (ed.) 1982. Curriculum, competence, and the foreign language teacher. (ACTFL foreign language education series 13.) Skokie, IL: National Textbook Company. ----- . 1984. Teaching for proficiency, the organizing principle. (ACTFL foreign language education series 15.) Lincolnwood, IL: National Textbook Company. ----- . 19858. The input hypothesis: an inside look. Foreign Language Annals 18.197-203. 261 ------ 1985b. Language acquisition and language learning: A plea for syncretism. Modern Language Journal 69.8-14. ------ 1985c. Teaching grammar for proficiency. Foreign Language Annals 18.289-96. Hosenfeld, Carol. 1977. A learning-teaching view of second-language instruction: the learning strategies of second-language learners with reading-grammar tasks. DAI 38.6581A. Jarest, Jackie, and Marsha Robinson. 1986. Charlemos un poco. Boston: Heinle and Heinle. Jaspaert, Koen; Sjaak Kroon; and Roeland Van Hout. 1986. Points of reference in first-language loss research. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els, 37-49. Dordrecht: Foris. Jenkins, James J. 1974. Remember that old theory of memory? Well, forget it! American Psychologist 29.785-95. Jordens, Peter; Kees De Bot; and Henk Trapman. 1989. Linguistic aspects of regression in German case marking. Studies in Second Language Acquisition 11.179-203. Judd, E.L. 1978. Vocabulary teaching and TESOL: A need for reevaluation of existing assumptions. TESOL Quarterly 12.71-6. Kellogg, Gillian S., and Michael J.A. Howe. 1971. Using words and pictures in foreign language learning. The Alberta Journal of Educational Research 17.89-94. Kennedy, L.R. 1932. The retention of certain Latin syntactical principles by first and second year Latin students after various time intervals. Journal of Educational Psychology 23.132-46. Kohn, Susan E.; Arthur Wingfield; Lise Menn: Harold Goodglass; Jean Berko-Gleason; and Mary Hyde. 1987. Lexical retrieval: The tip-Of- the-tongue phenomenon. Applied Psycholinguistics 8.245-266. Koster, Cor J. 1987. Word recognition in foreign and native language. Dordrecht, Netherlands: Foris. Kramsch, Claire. 1986. From language proficiency to interactional competence. Modern Language Journal 70:366-70. ------ 1987. The proficiency movement: Second language acquisition perspectives. (Review article.) Studies in Second Language Acquisition 9.355-62. Krashen, Stephen D. 1982. Principles and practice in second language acquisition. Oxford: Pergamon. 262 ------ 1985. The Input Hypothesis: Issues and implications. New York: Longman. ----- , and Tracy D. Terrell. 1983. The Natural Approach: Language acquisition in the classroom. Oxford: Pergamon. Lamadrid, Enrique E.; Laurel A. Briscoe; Robert Phillips; Enrique R. Lamadrid. 1984. Communicating in Spanish: A first course. 2nd edn. Boston: Houghton Mifflin. Lambert, Richard D., and Barbara F. Freed (eds.) 1982. The loss of language skills. Rowley, MA: Newbury House. ----- , and Sarah J. Moore. 1984. Recent research on language skill attrition. ERIC/CLL News Bulletin 8.1-8. , . 1986. Problem areas in the study of language attrition. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els, 177-86. Dordrecht: Foris. Lantolf, James P., and William Frawley. 1985. Oral proficiency testing: A critical analysis. Modern Language Journal 69.337-45. Laufer, Batia. 1989. A factor of difficulty in vocabulary learning: Deceptive transparency. Vocabulary acquisition, ed. by Paul Nation and Ron Carter, 10-20. (AILA Review 6.) Levenston, E.A. 1979. Second language vocabulary acquisition: Issues and problems. The Interlanguage Studies Bulletin 4.147-160. Leyen, Ina Angelika. 1986. Native language attrition: A study of vocabulary decline. DAI 46.2602A. Lightbown, Patsy M. 1983. Exploring relationships between developmental and instructional sequences in L2 acquisition. Classroom oriented research in second language acquisition, ed. by Herbert W. Seliger and Michael H. Long, 217-45. Rowley, MA: Newbury House. ----- . 1984. The relationship between theory and method in second language acquisition research. Interlanguage, ed. by Alan Davies, C. Criper, and A.P.R. Howatt, 241-52. Edinburgh: Edinburgh University Press. ----- . 1987. Classroom language as input to second language acquisition. First and second language acquisition processes, ed. by Carol Wollman Pfaff, 169-87. Cambridge:.Newbury House. Lockhart, Robert S.; Fergus I.M. Craik; and Larry Jacoby. 1976. Depth of processing, recognition and recall. Recall and recognition, ed. by John Brown, 75-102. New York: John Wiley. Long, Michael H. 1983. Does second language instruction make a difference? A review of research. TESOL Quarterly 17.359-82. 263 Lowe, Pardee, Jr. 1982. The U.S. government's foreign language attrition and maintenance experience. The loss of language skills, ed. by Richard D. Lambert and Barbara F. Freed, 176-90. Rowley, MA: Newbury House. Ludwig, Jeannette. 1984. Vocabulary acquisition as a function of word characteristics. Canadian Modern Language Review 40.552-62. Lyons, John. 1977. Semantics, vol. 2. Cambridge: Cambridge University Press. McLaughlin, Barry. 1979. The Monitor Model: Some methodological considerations. Language Learning 28.309-32. ----- . 1987. Theories of second-language learning. Baltimore: Edward Arnold. Marton, W. 1977. Foreign language vocabulary learning as problem number one of language teaching at the advanced level. The Interlanguage Studies Bulletin 2.33-57. Meara, Paul. 1980a. Syntactic structure and memory span in second language learners. Interlanguage Studies Bulletin 5.31-54. ------ 1980b. Vocabulary acquisition: A neglected area of language learning. Language Teaching and Linguistics: Abstracts 13.221-46. ----- (ed.) 1983. Vocabulary in a second language, vol. 1. London: Centre for Information on Language Teaching and Research. ----- . 1984. The study of lexis in interlanguage. Interlanguage, ed. by Alan Davies, C. Criper, and A.P.R. Howatt, 225-35. Edinburgh: Edinburgh University Press. ----- (ed.) 1987. Vocabulary in a second language, vol. 2. London: Centre for Information on Language Teaching and Research. ----- . 1989. Matrix models of vocabulary acquisition. Vocabulary acquisition, ed. by Paul Nation and Ron Carter, 66-74. (AILA Review 6.) Melka Teichroew, Francine J. 1982. Receptive vs. productive vocabulary: A survey. Interlanguage Studies Bulletin 6.5-32. Moorcroft, R., and Robert C. Gardner. 1987. Linguistic factors in second-language loss. Language Learning 37.327-40. Murdock, Bennet B., Jr. 1982. Recognition memory. Handbook of research methods in human memory and cognition, ed. by C. Richard Puff, 1-26. New York: Academic Press. Murray, David J. 1986. Characteristics of words determining how easily they will be translated into a second language. Applied Psycholinguistics 7.353-72. 264 Nation, I.S.P. 1982. Beginning to learn foreign vocabulary: A review of the research. RELC Journal 13.14-36. Neisser, Ulric. 1984. Interpreting Harry Bahrick’s discovery: What confers immunity against forgetting? Journal of experimental psychology: General 113.32-5. Obler, Loraine K. 1982. Neurolinguistic aspects of language loss as they pertain to second language attrition. The loss of language skills, ed. by Richard D. Lambert and Barbara F. Freed, 60-79. Rowley, MA: Newbury House. ----- , and M.L. Albert. 1984. Language in aging. Clinical neurology of aging, ed. by M.L. Albert, 245-53. New York: Oxford University Press. Olshtain, Elite. 1986. The attrition of English as a second language with speakers of Hebrew. Language attrition in progress, ed. by Bert Weltens, Kees De Bot, and Theo Van Els, 187-204. Dordrecht: Foris. ------ 1989. Is second language attrition the reversal of second language acquisition? Studies in Second Language Acquisition 11. 151-65. Omaggio, Alice C. 1984. The proficiency-oriented Classroom. Teaching for proficiency, the organizing principle, ed. by Theodore V. Higgs, 43-84. (ACTFL foreign language education series.) Lincolnwood, IL: National Textbook Company. O'Malley, J. Michael; Anna Uhl Chamot; and Carol Walker. 1987. Some implications of cognitive theory to second language acquisition. Studies in Second Language Acquisition 9.287-306. Ostyn, Paul, and Pierre Godim. 1985. RALEX: An alternative approach to language teaching. Modern Language Journal 69.346-53. Oxford, Rebecca L. 1982a. Research on language loss: A review with implications for foreign language teaching. Modern Language Journal 66.160-9. ----- . 1982b. Technical issues in designing and conducting research on. language skill attrition. The loss of language skills, ed. by Richard D. Lambert and Barbara Freed, 119-37. Rowley, MA: Newbury House. Paivio, Allan. 1976. Imagery in recall and recognition. Recall and recognition, ed. by John Brown, 103-130. New York: John Wiley. Palmberg, Rolf. 1987. Patterns of vocabulary development in foreign- language learners. Studies in Second Language Acquisition 9.201-19. 265 Pan, Barbara Alexander, and Jean Berko-Gleason. 1986. The study of language loss: models and hypotheses for an emerging discipline. Applied Psycholinguistics 7.193-206. Pimsleur, P. 1967. A memory schedule. Modern Language Journal 51.73-5. Postovsky, Valerian A. 1974. Effects of delay in oral practice at the beginning of second-language learning. Modern Language Journal 58.229-39. Pratella, W.C. 1969. The retention of first and second year Spanish over the period of the summer vacation. Ph.D. dissertation. Fordham University. Puff, C. Richard. 1982. Handbook of research methods in human memory and cognition. New York: Harcourt, Brace, Jovanovich. Rivers, Wilga M. 1986. Comprehension and production in interactive language teaching. Modern language journal 70.1-7. Robison, Robert Eugene. 1985. The effect of the summer vacation on language attrition in secondary school students of first-year Spanish. DAI 46.636A. Sapir, Edward. 1921. Language. New York: Harcourt, Brace & World. Savignon, Sandra. 1985. Evaluation of communicative competence: the ACTFL Provisional Guidelines. Modern Language Journal 69.129-34. ----- , and Lyle F. Bachman. 1985. The evaluation of communicative language proficiency: A critique of the ACTFL oral interview. Modern Language Journal 70.380-9. Scarcella, Robin, and Leroy Perkins. 1987. Shifting gears and Krashen's input hypothesis. (Review article.) Studies in Second Language Acquisition 9.347-53. . Scherer, George A.C. 1957. The forgetting rate in learning German. The German Quarterly 30.275-7. Schmidt, Richard W. The role of consciousness in second language learning. (to be published in Applied Linguistics) Schouten-Van Parreren, Caroline. 1989. Vocabulary learning through reading: Which conditions should be met when presenting words in texts? Vocabulary acquisition, ed. by Paul Nation and Ron Carter, 75-85. (AILA Review 6.) Schulz, Renate. 1986. From achievement to proficiency: some caveats. Modern Language Journal 70.373-9. Seliger, Herbert W., and Michael H. Long (eds.) 1983. Classroom oriented research in second language acquisition. Rowley, MA: Newbury House. 266 Sharwood Smith, Michael. 1981. Consciousness raising and the second language learner. Applied Linguistics 11.159-168. ----- . 1984. Discussion of ‘The study of lexis in interlanguage.’ Interlanguage, ed. by Alan Davies, C. Criper, and A.P.R. Howatt, 236-9. Edinburgh: University of Edinburgh Press. ----- . 1986. Comprehension vs. acquisition: The ways of processing input. Applied Linguistics 7.239-56. Sheldon, Amy, and Winifred Strange. 1982. The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede speech perception. Applied Psycholinguistics 3.243-61. Smythe, P.C., G.C. Jutras, J.R. Bramwell, and Robert C. Gardner. 1973. Second language retention over varying intervals. Modern Language Journal 57.400-5. Spinelli, Emily, and Marta Rosso-O'Laughlin. 1988. Encuentros. New York: Holt, Rinehart, Winston. Spitze, K., and Susan D. Fischer. 1981. Short-term memory as a test of language proficiency. TESL Talk 12.32-41. Spolsky, Bernard. 1985. Formulating a theory of second language learning. Studies in Second Language Acquisition 7.269-88. Stevick, Earl W. 1976. Memory, meaning, and method: Some psychological perspectives on language learning. Rowley, MA: Newbury House. Swain, Merrill. 1985. Communicative competence: some roles of comprehensible output in its development. Input in second language acquisition, ed. by Susan Gass and Carolyn Madden, 236-56. Rowley, MA: Newbury House.‘ ----- , and Sharon Lapkin. 1989. Canadian immersion and adult second language teaching: What's the connection? Modern Language Journal 73.150-9. Terrell, Tracy David. 1986a. Acquisition in the Natural Approach: The binding/access framework. Modern Language Journal 70.211-27. ----- . 1986b. Recent trends in research and practice: Teaching Spanish. Hispania 69.193-201. ----- . 1989. The role of grammar instruction in a communicative approach. NE Conference Newsletter. 22-5. ----- ; Magdalena Andrade; Jeanne Egasse; and Elias Miguel Mufioz. 1986c. Dos mundos. New York: Random House. 267 Tulving, Endel. 1976. Ecphoric processes in recall and recognition. Recall and recognition, ed. by John Brown, 37-74. New York: John Wiley. Tversky, Barbara. 1973. Encoding processes in recognition and recall. Cognitive Psychology 5.275-87. Twaddel, F. 1973. Vocabulary expansion in the ESOL classroom. TESOL Quarterly 7.61-78. VanPatten, Bill. 1984. Communicative values and information processing in L2 acquisition. 0n TESOL ’84. A brave new world for TESOL. (Selected papers from the annual TESOL convention.) ------ 1987. On babies and bathwater: Input in foreign language learning. Modern Language Journal 71.156-64. ------ 1989. Can learners attend to form and content while processing input? Hispania 72.409-17. Weltens, Bert. 1987. The attrition of foreign language skills: A literature review. Applied Linguistics 8.22-38. ----- ; Kees De Bot; and Theo Van Els (eds.) 1986. Language attrition in progress. Dordrecht: Foris. ----- ; Theo Van Els, and Erik Schils. 1989. The long-term retention of French by Ontch students. Studies in Second Language Acquisition 11.205-15. Wingfield, Arthur, and Dennis L. Byrnes. 1981. The psychology of human memory. New York: Academic Press. Winitz, Harris (ed.) 1981. The comprehension approach to foreign language instruction. Rowley, MA: Newbury House.