H H HM “:1 w | , L I u | W H ‘ & W } 7__ ,7 ._. * _,,, t N 3%: “WM VARiABLES IN SEMANTIC SATIATiON Thesis for {he Dames cf M. A. MECHEGAN SYATE Ui‘i’s‘l‘RSi‘iY flurfcn L. AEperson “M4 THESIS W (W M 3 1293 01099 7348 1! I WT? HHWIWWIIL‘ LIBRARY Michigan State University PLACE II RETURN BOX to roman this checkout from your record. TO AVOID FINES mum on or bdoro date duo. DATE DUE DATE DUE DATE DUE MSU Is An Nflrmutivo Action/Equal Opportunity Institwon Wanna-m ABSTRACT VARIABLES IN SEMANTIC SATIATION by Burton L. Alperson Semantic satiation (the decrease in meaning of words with prolonged attention) has been studied by various investigators using decreases in the polarities of words on the Semantic Differential, decreases in the popularity of word associations emitted by §s and other associative techniques, andig's intro- spective reports of when the meaning disappears as evidence for the phenomenon. Independent variables have included having gs repeat, stare at, repeat and stare at, and listen to the words. Due to the proliferation of methods, the term."semantic satiation" has almost lost its meaning. There has been no serious empirical attempt yet reported to relate these vari- ables, and they are held together by little more than a common name. The various dependent variables all suffer from a number of weaknesses. The introspective techniques may bias results with suggestion and do not allow the use of non-satiated control groups. There has been no satisfactory replication of the results of the Semantic Differential studies reported outside of the labor- atories of the originators of this technique, although several attempts to replicate the original studies have been reported. Burton L. Alperson The word association techniques seem to be replicable, but the size and stability of the reported effects have generally been quite small. The present study is an attempt to determine: (1) if any combination of independent and dependent variables yields signi- ficant results; (2) if any of these combinations are significantly better than others; and (3) if those combinations which are found to be significant are correlated. The Semantic Differential technique and three word associa- tion techniques (popularity of the first response, total number of responses, and number of responses in the first five seconds) were investigated in a "before-after" design. Five treatments (visual, repetition, visual a repetition, listening, and non- satiated control) were employed with each dependent variable, the Semantic Differential and word association groups being indepen- dent of one another. The Semantic Differential group received 90 seconds of total satiation time per word (divided among six 15 see. periods). The word association group received 15 seconds of total satiation time per word. In order to determine the effect of making the association group more nearly approximate the Semantic Differential group, a small independent study was run using the word association variables under the repetition and control conditions with a 60 second satiation time. No evidence was found for satiation using the Semantic Differential technique. It was suggested that differential Burton L. Alperson reliabilities of words may have confounded the results. Whatever the reason, the failure of the present study, and other studies to find satiation with this technique brings its adequacy as a measuring instrument into question. 0f the word association techniques, only the popularity variable with 60 seconds of satiation time yielded evidence for satiation. Consideration of the data from the main study and the 60 second time control study suggests that an increase in the size of the independent variable will produce stronger sati- ation effects as measured by this technique. The data do not permit generalizations concerning the superiority of one independent variable over another. M 7/13/(/ f/W VARIABLES IN SEMANTIC SATIATION By Burton L} Alperson A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1964 I,» ‘4 _ J ACKNOWLEDGMENT The author wishes to express his gratitude and appreciation to his advisor, Dr. Paul Bakan, whose helpful suggestions and assistance throughout the various stages of this research enhanced the fulfillment of this project. ii TABLE OF CONTENTS ACKNOWLEDGEMENT . . . . . . . . . . . . . LIST OF TABLES . . . . . . . . . . . . LIST OF APPENDICES . . . . . . . . . . . Chapter I 0 INTRODUCTION 0 O O O O O O O O IIonODoesoososqooso Subjects Procedure III. RESULTS. 0 o o o o 9 o o O O s 0 Semantic Differential Polarity Word Association Satiation Time Control Study Word Order Correlations IV. DISCUSSION . . . . . . . . . . V. CONCLUSIONS . . . . . . . . . . REFERENCES APPENDICES C O O O O O O O O O O I O O O O O O O O O O O O O O O 0 iii Page ii iv 14 23 29 31 33 Table 10. LIST OF TABLES Stimuli and their Initial Polarities on Relevant scales 0 0 O O O O O O O O O O O O O O O O O O O O 0 Treatment Means, Standard Deviations, and Dunnett's t Statistic for Multiple Comparisons with a Control Word Means and Standard Deviations . . . . . . . . . . Analysis of Variance for Semantic Differential Pelarities O O O 0 O I O 0 O O O O O O O O O O 0 O O Reliabilities (r) of Polarity Scores (Semantic Dif- ferential Polarities: Control Group). . . . . . . . Analysis of Variance for Semantic Differential (Control Group Only) . . . . . . . . . . . . . . . . Analysis of Variance for Word Association (Popularity) 0 O O O O O O O O O O O O O O O O O O 0 Analysis of Variance for Word Association (Total numbe r ) O C O O O O O O O O O C C O O O O O O O O 0 Analysis of Variance for word Association (Number in late 5 secs) 0 o o o o o o s s O o O s o o s o o 0 Analysis of Variance for Word Association (Time Control Study: Popularity) . . . . . . . . . . . . . . . iv Page 15 16 17 18 18 19 20 20 21 Appendix A. B. LIST OF APPENDICES Task Instructions . . . . . . Semantic Differential werd Association Word Norms . . . . . . . . . Page CHAPTER I INTRODUCTION Semantic satiation (the decrease in intensity of the meaning of a word after it has been given prolonged attention) has been studied in a wide variety of contexts. Early studies in the area were introspective and were primarily concerned with the selection of independent variables. In these studies some satiation procedure was employed during which trained ob- servers (usually the eXperimenter and one or two cohorts) were to report when and how the meaning of the word disappeared. Using this technique, it was reported that satiation effects could be observed by presenting stimuli visually (Severance & Washburn, 1907) or by instructing §s to repeat the word (Bas- sett & Warns, 1919) until the meaning disappeared. It was also noted that visual presentation seemed to produce stronger sati- ation effects that did repetition (Don & Weld, 1924) and that the words tended to break up between syllables (Severance & Washburn). These early studies seem to regard the semantic satiation effect as an isolated laboratory curiosity. Later studies might be characterized as attempts to increase the generality of the phenomenon. Mason (1941), for example, found a significant 1 correlation between the time at which a word lost meaning and fluctuations in galvanic skin response. Wertheimer and Gillies (1958), working within the framework of Kbhler's studies of figural after effects, reported that two syllable words satiated faster than one syllable words, six letter words satiated faster than four letter words, and that the second half of the experi- mental session produces more satiation than the first half. Word frequency and the objectivity of the referent, however, did not prove to be significant variables. In another study, Wertheimer (1958) found a tendency for words that were rated as "fitting" (i.e. whose sounds were, in some sense, similar to their meaning) to be more difficult to satiate than words rated as "non-fitting". All of the above studies used either the original intro- spective technique or some close variant of it. In 1956 Smith and Raygor introduced a word association technique for measuring satiation. After gs fixated on a word which was presented on a screen, they were asked to write down their first association to the next word that was flashed on the screen. They reported that words yielded significantly more uncommon associations after they had been satiated than they did without satiation. Subjects classified as "impermeables" (introverts) seemed to be more prone to satiation than gs classified as "permeables" (extraverts). This technique was later used by Paul (1962) who modified it slightly by the use of Noble's (1952) continued associations technique. The use of Noble's system.allowed for the examdnation of a number of new dependent variables, and Paul reported that satiation was reflected in " . . . a reduced number of associ- ations early in the response period, a reduced number of popular first associations, and an increased latency for the emission of popular associates." (p. 166). He also presented evidence for the generalization of satiation, although no regular generalization gradient was observed. At McGill University, Lambert and Jakobovits (1960) intro- duced yet another dependent variable. Using a reduction in polar- ity scores on the Semantic Differential ratings of words as a measure of satiation, this group has attempted to link the effect both to Osgood's (1953) representational mediational theory of meaning and to Noble's (1952) associational system. ‘Lambert and Jakobovits (1960; Jakobovits and Lambert, 1961, 1962 b), who feel satiation is a cognitive form of reactive inhibition, find support for Osgood's mediation hypothesis in satiation ef- fects. They also feel that satiation effects are a function of the central rather than the peripheral nervous system. Kanungo, Lambert and Mauer (1962; Kanungo and Lambert, 1963), on the other hand, view satiation as a gradual reduction in 3 components, such that the word being satiated begins to elicit itself as an association. They invoke this hypothesis to explain the effects of satiation on paired—associate learning, and a satiation study in which a technique similar to Paul's is used, but in which the dependent variable is Noble's Eh This brief history should include some reference to Warren's (1958, 1961, 1962) studies employing auditory repetition. When §s hear a word repeated on a tape loop, the word they are listen- ing to seems to change into other words. Warren (1962) feels that part of this process consists of a loss of meaning of the word which he relates to the satiation studies. In teams of the satiation phenomenon itself, all of the above studies suffer from one or more of a number of weaknesses. In the first place the term "semantic satiation" is almost meaning- less. Due to the proliferation of methods, when one speaks of semantic satiation he may be speaking of any combination of four or more independent variables with three or more dependent vari- ables. Since there has been no serious empirical attempt yet reported to relate them, these variables are held together by little more than a common name. Secondly, those studies employing timed introspection as a dependent variable (Severance and Washburn, 1907; Don and Weld, 1924; Wertheimer, 1958; Wertheimer and Gilliss, 1958; and Mason, 1941) must face the serious question of how much of the phenomenon is an artifact of the instructions. This method necessitates that the §s be told that they are expected to satiate, e.g. "Fixate [the word] steadily, and give it maximal attention until its meaning has disappeared" (Don & Weld, 1924, p. 44) or "L§s] were asked to judge the time at which meaning disappeared for each of 10 words . . . " (Wertheimer, 1958, p. 444). One may only guess at how much of the time difference is due to satiation and how much of it is due to suggestion. Another major defect of this method is the fact that it is impossible to use a non-satiated control group here. Thirdly, those studies which rely on decreases on Semantic Differential polarity (Lambert and Jakobovits, 1960; Jakobovits and Lambert, 1961, 1962; and Kanungo, Lambert and Mauer, 1962) are not unequivocal, although they do, at least control instruc- tional bias. Several recent studies suggest that the phenomenon may not exist in terms of this technique; or if it does exist, it may be a unique characteristic of the McGill population. Floyd (1962) in a partial replication of the Lambert and Jako- bovits study,found "very weak support" for the existence of the phenomenon. Yelen and Schulz (1963), after failing to find satiation effects in a partial replication of the same study, attempted an exact replication and still failed to find evidence for satiation. They present evidence for a generation effect (i.e. an increase in meaning) which seems to be roughly inversely correlated with initial meaning on the Semantic Differential. Das (1964) has also reported this effect. In attempting to find satiation using polarity scores, he is apparently forced to dis- cuss his results as representing degrees of "lack of satiation.” Reynierse (1963) failed to find any evidence for satiation and he suggests that the selection of Semantic Differential scales may influence the results. Finally, although the word association technique of Smith and Raygor has received relatively little attention, those studies which have been reported do not yield as firm support as might have been hoped. As measured by this technique (both decrease in the number of associates and a decrease in the popularity of the first response) Paul suggests that satiation effects are " . . . slight . . . transient and quick to dissipate and [seem] to be dependent on individual word characteristics.” (p. 165). Reynierse found no evidence for satiation using this technique. Kanungo and Lambert do report a decrease in relevant responses (g9 after satiating a word. However, since no mention is made of a blind scoring system to separate rele- vant from irrelevant responses, these results are open to ques- tion. In summary, it may be said that the literature of semantic satiation has not progressed a great deal since the 1920's when researchers were concerned with the search for significant in- dependent and dependent variables. The most meaningful approach would seem to be a return to these questions using the more sophisticated techniques that the recent literature has made available. Until more is known about the satiation phenomenon itself, and until a reliable means of assessing it is found, _attempts to relate this phenomenon to other variables are, at best, premature. The present study is an attempt to deal with three ques- tions. If all previously reported independent and dependent variables (with the exception of the introspective techniques) are employed with a given set of stimuli, and conditions are set up to allow the maximal possibility of satiation: Does any combination of independent and dependent vari- ables yield significant results? Are any of these combinations significantly better than others? Are those combinations, which are found to be significant, correlated? CHAPTER 11 METHOD Subjects Subjects were 150 volunteers from introductory psychology classes at Michigan State University. They were assigned to one of two major dependent variable groups: either Semantic Differ- ential or Word Association. Each §_was then further assigned to one of five treatment groups (Visual, Repetition, Visual & Repe- tition, Listening, or Control) within each dependent variable. Thus, there were 75 Se assigned to each dependent variable and 15 to each treatment. A separate study designed to control the length of the satiation period in the Word Association dependent variable group employed 18 additional §s from the Michigan State population. All assignments of Se to independent and dependent variables were random. Procedgre All §s were tested on 2 separate occasions. During the first session the stimuli (Table 1) consisted of 20 words. During the second session, one week later, 10 of the original 20 words were employed and were exposed by either §_or §, The 10 experi- mental words were selected for the following characteristics: (a) 2 syllables, and (b) 6 or more letters (Wertheimer & Gillies, 1958). A final consideration in the selection of these words was 8 to obtain a fairly representative sampling of Semantic Differential polarities (as calculated from Jenkins, Russell, & Succi, 1958 on the appropriate scales). The "filler" words (the words other than the 10 experimental words) were selected on the basis of having a polarity score relatively close to one of the experimental words. It was hoped that the similarity between the filler and experimental words would reduce the effects of memory of the ratings made in the first session. TABLE 1 STIMMLI AND THEIR INITIAL POLARITIES 0N RELEVANT SCALES “.11 fl ,____ ———r V—. Experimental Words Polaritya Filler Words Polaritya Bodkinb 2.26 Winter 4.53 Somber 4.18 Hungry 5.12 Bottom 4.80 Sweeping 5.68 Beggar 5.30 Graceful 6.70 Sunday 7.73 Hurried 7.27 Sister 8.47 Mallet 8.90 Abrupt 9.37 Sparkling 9.04 Sudden 10.07 Brother 9.16 Hammer 11.07 Success 10.91 Starving 11.51 Statue 11.01 aAs calculated from Jenkins, Russell, & Suci (1958). bNot included in any of the analyses due to the large number of subjects who looked the word up between the pre- and post-test. 10 Each‘g saw the experimental words in one of 15 predetermined random orders. The order of presentation of experimental stimuli was constant for each § during both sessions. During the first session, however, the first word presented and every other word thereafter was a filler word. The order of the filler words was constant for all gs. Semantic Differential Polarity.--For each of the 75 Se in this group §_exposed the 20 stimuli one at a time during the first session. Each word was typed in capital letters on a 3"x5" index card. Following each word presentation §_rated the words by point- ing to a position on one of 6 Semantic Differential scales (GOOD- BAD, CRUEL-KIND, ACTIVE-PASSIVE, EAST-SLOW, WEAKFSTRONG, or HARD- SOFT) which appeared on a memory drum. This procedure was repeated until all 20 words were rated on all 6 scales (for a total of 120 ratings per S). The order of words and scales maximized the dis- tance between the re-occurrence of any particular word or scale. The instructions given to each §_were substantially the same as those by Reyniesse (1963) (Appendix A). Word Association.--During the first session, each of the 75 SS in this group took a word association test for each of the 20 stimulus words. Words were presented on 8 1/2” x 11" mimeographed sheets in a format similar to that used by Noble (1952). Specifically, each word appeared on a separate sheet; it appeared 30 times on the sheet, followed each time by a blank line on which S was to write his 11 responses to the word. Repetitions of the stimulus word were numbered from 1 to 30 on each sheet. Instructions given to each §_(Appendix A) were identical to those used by Noble. gs were allowed 60 sec. to write down their responses, and each sheet was followed by a rest period lasting about 15 sec. Visual Presentation.--During the second session, §s were given a "deck" of 3"x5" index cards containing the 10 experi- mental words. They were instructed to stare at the top word and concentrate on it until told to stop (15 sec. later). Following this, they either rated the word on a Semantic Differential scale or were given a word association sheet for the word, depending on which dependent variable group they had been assigned to the previous week. §s then stared at the next card in the deck and repeated the procedure. The order of presentation of words and scales remained constant in the two sessions for each §. Pro- cedures for all other independent variables follow this pattern. Exceptions are noted below. Repetition.--§.presented each stimulus word for about 1 sec., following which §_was instructed to repeat the word in a monotone at a rate of 2-3 times per sec. until he was told to stop. Visual & Repetition.--Stimuli were exposed by § who was instructed to stare at the word while be repeated it in a mono- tone at a rate of 2-3 times per sec. until he was told to stop. 12 Listening.--§s in this condition listened first to gfs voice (recorded monophonically) over a set of sterophonic head- phones. They were instructed to set a balance control to a point at which fife voice sounded as though it were coming from the center of their heads. E then presented the words to them on the 3"x5" cards. Following a 1 sec. visual presentation of each word by‘g they listened to the same word which was repeated at a rate of 2-3 times per sec. for 15 sec. by means of a tape loop. They were then tested on the appropriate dependent vari- able. Although intensity level was constant for all words for each §_it was not controlled between §s. They were instructed to set the volume control of the tape recorder to a point at which they could hear "clearly and comfortably." Non-Satiated Control.--Experimenta1 words were presented in the same way in which they had been presented in the first session. Satiation Time Control Study.-~It may be noted that the total satiation time for a word for §s in the Word Association group was 15 sec., in contrast to Se in the Semantic Differential Polarity group who received 90 sec. of total satiation time per word (divided among six 15 sec. satiation periods). In order to determine the effect of making the Word Association group more nearly approximate the Semantic Differential group, a small inde— pendent study was run. This study was identical to the Repetition and Control conditions of the Word Association variable in the 13 main study. The only difference was that a smaller sample (N = 9 per group) and a 60 sec. satiation time were employed. CHAPTER III RESULTS Semantic Differential golgrity The scoring system for this group is parallel to that used by Lambert & Jakobovits (1960). An overall polarity score was calculated for each §_on a given word (scale positions designated: -3-2;1704—Lf2+3). This procedure was repeated on the ratings made in the second session. A polarity difference score was then cal- culated by subtracting the polarity scores of the first session from those of the second. Thus an §tho rated a word +3, +l,+0, -1, +2, -1 on the six scales in the first session and then rated the same word +1,‘+l, +1, -2,*'1, 0 in the second would receive a polarity difference score of (T+T+l+2+l) - (3+1+1+3'1)i= -2 for that word. The data in Tables 2 and 3 for the Semantic Differential group have been increased by a constant of 10 to eliminate negative signs for convenience. A mean less than 10 may be interpreted as a difference in the direction of satiation (i.e. a decrease in polar- ity), while a mean above 10 may be interpreted as evidence for a generation effect (i.e. an increase in polarity). The only signifi- cant value in Table 4, the summary analysis of variance for these scores, is the main effect for Words (p 4.01). This method of obtaining polarity difference scores will ad- mittedly distort or cancel ratings which change sign between the 14 15 N Mammfi -- asa.x «mm. see. use. a .uom n no .02 cm. «a. as. no. no. a coausaoosma ea.m ma.ca NH.oH ma.oe ~o.oe m. who: -- has. Has. amm. ems. u .oz Hence mm.~ ae.~ mo.m ma.~ mo.~ m coausauosm< ea.oa no.aa ea.oa eo.aa eo.aa m. etc: -- new. oNA. «an. new. u suesmaaaom an. on. as. as. mm. m eoausaoomme mo.oa am.a um.m am.m ma.m m use: -- sum. was. Ham. «on. u «usuanueom N~.m ma.m ma.m na.m oe.m m Hmeueuuumuea am.m No.m Ne.oa ao.oH a~.m m. uauemaom aonueoo masseuse; scapegoaom eouuauuaue aussa> assesses amamw> uoooeoaun eoeazoo a zoo: mzomaeemzoo mamaaaez mom oHamHafim o «Buzzes $2 .monafi>B 333m .332 Egg 16 Ho. mm. no. am. no. me. mm. mm. mm. m no.oH o~.o~ NH.oa no.0H no.o~ o~.oH mo.od NH.oH mo.oH .m mm.~ em.~ ma.m <¢.N em.N mo.m na.~ unavoomon E mZOHHma QM“.01 Two additional analyses were performed in the Word Association group. A decrease in the total number of responses from the first session to the second as well as a decrease in the number of responses written during the first 5 sec. of the writing period (as recorded by E) would be considered as evidence for satiation. These analyses are reported in Tables 8 and 9 and Tables 2 and 3. Once again, a constant of 10 has been added to the scores in Tables 2 and 3. A significant main effect for Words (p .4 .01) was observed in the "total number" variable (Table 8), while none of the values in the "number in the 1st 5 sec." variable reached significance (Table 9). 20 TABLE 8 ANALYSIS OF VARIANCE FOR WORD ASSOCIATION (Total Number) m Source gf. MS F Treatments 4 20.709 .826 Error 8 70 25.050 Words 8 16.123 2.999* TxW 32 6.457 1.201 Error b 560 5.376 Total 674 * p4¢:.01 TABLE 9 ANALYSIS OF VARIANCE FOR WORD ASSOCIATION (Number in lst. 5 sec.) Source g; MS F Treatments 4 1.086 2.281 Error a 55 .476 Words 8 .040 .085 TXW 32 .468 1.002 Error b 440 .467 Total 539 The reduction in sample size here is due to the mis-recording of data for 3 SS. To equalize N, 3 corresponding §s were discarded in all groups. An S was discarded if he had received the same order of words as one of those whose data had been mis-recorded. 21 Satiation Time Control Study Total number of responses and number of responses in the first 5 sec. of writing were observed, but the complete analysis was not carried through since t-tests between the Non—Satiated Control and Repetition conditions yielded p values well above the .05 signifi- cance level. The summary analysis of variance for the Popularity data in this study is presented in Table 10. The main effect for Treatments is significant (p.4L.05), and is in the expected direc- tion for a satiation effect. The main effect for Words did not reach significance in this analysis. TABLE 10 ANALYSIS OF VARIANCE FOR WORD ASSOCIATION (Time Control Study: Popularity) Source g§_ MS F Treatments 1 2.176 6.40* Error a 14 .340 Words 8 .676 1.32 TxW 8 .455 .89 Error b 112 .514 Total 143 1*p‘LOS Word Order Correlations The "total number" variable in the main study, the ”total number” variable in the Satiation Time Control study, the polarity difference 22 scores, the initial Semantic Differential polarities, and the initial total number of associations (2') were all compared with each other in terms of word order. None of these correlations reached statis- tical significance. However, the correlation between the main study and the Satiation Time Control study on the "popularity" variable was significant (r = .676; p‘.025). CHAPTER IV DISCUSSION Lambert & Jakobovits (1963) advance a number of reasons for Yelen & Schulz's failure to replicate their original study (they do not deal with the other failures). They note that the discrepan- cies might be accounted for by a change in the scoring system, a different task set, and differences in the samples of Se employed. These arguments will be considered in turn. The scoring system in the present study does not differ in any significant way from the Lambert & Jakobovits method. The only dif- ferences are in the addition of a constant to the scores, and the use of scores per-word-per-subject rather than scores per-word-per- subject-per-scale. Neither of these changes will affect the overall analysis and both scoring systems will always agree on the direction of change. Lambert & Jakobovits feel that Yelen & Schulz may have intro- duced a difference in task set " . . . by their mode of explaining the rating procedure or by their directions to repeat the word as rapidly as possible" (p. 6). Since Lambert & Jakobovits' (1960 & 1963) descriptions of their instructions parallel the instructions used in the present study almost word for word, this is not a rele- vant criticism in this case. 23 24 The present Michigan State sample was composed of volunteer ‘Ss as were the McGill samples. This sample was predominantly lst and 2nd year students while the McGill samples consisted of 2nd, 3rd, and 4th year students. However, until such time as Lambert & Jakobovits can demonstrate reliable differences in satiation within the span of approximately 18 to 23 years of age, criticism of the present study is seriously weakened on these grounds. There is one major procedural difference between the Lambert & Jakobovits studies and the present experiment. Lambert & Jakobo- vits obtained pre- and post-satiation ratings from their §s in one experimental session, while in the present study there was a 1 week delay between pre- and post-satiation ratings. It is possible that this change accounts for the failure to find satiation effects in the present study. However, if one is to accept the theoretical explanation and the claims for the generality of the phenomenon, as measured by the Semantic Differential, which are advanced by the McGill group, it becomes difficult to understand why a one week delay should have any effect. It might also be pointed out that Kanungo & Lambert (1963) report satiation effects with a 24 hour delay between pre- and post-satiation ratings (using word associ- ation techniques). There seems to be some confusion in the literature regarding the importance of the selection of stimuli in satiation studies. Reynierse, Yelen & Schulz, and Wertheimer feel that stimulus selec- tion may be an important variable while Lambert & Jakobovits insist that it may or may not be of importance. This confusion makes 25 interpretation of the significant main effect for Words in Table 4 difficult and tentative. There are two converging arguments, how- ever,which suggest that stimulus characteristics are relevant vari- ables in satiation studies. Intuitively it seems reasonable that some words are more meaningful (i.e. reliable) than others. Although the filler words did not appear during second session, there is some justification for regarding the control group as a reliability study of polarity difference scores. Inspection of Table 5 reveals that different words do seem to have different reliabilities. If this finding can be replicated, the implication would be that the probability of obtaining large random differences for words with low relia- bilities would be increased, while it would be more difficult to obtain large difference scores for highly reliable words. The effect of these differences upon satiation is an empirical problem which cannot be studied with the data collected in the present study. It would be of interest to know whether or not there is any correlation between the reliabilities of words and their po- larity difference scores in those studies which report finding satiation. This argument may be further strengthened by consider- ation of Table 6. Even in the absence of independent variables, there are reliable differences between the polarity difference scores of different words. The only evidence for satiation, using polarity difference scores, are the introspective reports of the §s. Not only were there no significant treatment effects, few of the differences 26 obtained were in the predicted direction (Tables 2 and 3). If there were any satiation effects, the Semantic Differential tech- nique was not sensitive enough to measure them. Total number of associates and the number of associates in the first 5 sec. of responding seem to be of little usefulness in satiation studies. The "Total Number" row in Tables 2 and 3 essen- tially replicates Kanungo & Lambert's (1963) finding that the number of associates increases from the first to the second session in this design. Their explanation of this finding, that memory facili- tates responding, seems plausible. Whatever the explanation, this increase limits the validity of this measure for satiation studies. In the same study, Kanungo & Lambert report that the number of rele- vant responses decreases with satiation. This type of analysis was not included in the present study because of the difficulty of separ- ating relevant from irrelevant responses. Under the best of circum- stances this decision is an arbitrary one (in spite of Noble's criter- ia). The lack of a blind scoring system, in addition to the fact that one of the ES apparently made these decisions considerably weakens the evidence for satiation in this study. Number of associations in the first 5 see. is a very crude latency measure. The speed with which an S can write limits his score to 0, 1, 2, or in rare instances, 3. This restriction of range coupled with the discrete nature of the observations and the possi- bility of systematic errors in scoring by,§ lessens the sensitivity and validity of this measure. Response latency is an obvious measure 27 for satiation and it should be examined with a more sensitive instrument than the one employed in the present study. Table 10 summarizes the only data in the present study which may be interpreted as evidence for satiation. There is some evi- dence in the literature for an increase in satiation effects with an increase in satiation time. Paul's (1962) satiation effects on a popularity of the first response dependent variable were obtained with a 30 sec. satiation interval. The only study which has examined the effect of different satiation times is the Smith & Raygor (1956) study. Using a popularity of response measure, they reported that satiation effects increase markedly with a shift from three 7 sec. satiation periods to one 20 sec. period. If the results in the present study do represent satiation effects, they are fairly im- pressive since they were obtained with a considerable loss of power due to the reduced sample size. It should also be pointed out that in the main study, although none of the differences reached statistical significance, the popu- larity of associations measure was the only one in which all of the differences for treatments, and most of the scores for words were in the expected direction for a satiation effect. Thus, there is reason to suspect that a longer satiation interval in the main study might have produced clearer satiation effects. This hypothesis is slightly embarrassed by the fact that the Words main effect reached signifi- cance in the main study (Table 7), while it did not reach significance in the Satiation Time Control Study (Table 10). One possible explana- tion is that the satiation effect is strong enough to wipe out differences 28 between words. There is some evidence for the same (or at least similar) process operating in both the Satiation Time Control study and popularity of response experimental groups in the main study. The correlation for word order between the experimental groups of the main study and the experimental group in the Satiation Time Control study is significant, suggesting that the difference be- tween these studies is one of degree. The above arguments are speculative, but they do suggest that it would be profitable to replicate the Popularity variable under the conditions of the main study, using a longer satiation time. CHAPTER V CONCLUSIONS There is little doubt that the semantic satiation phenomenon exists. The problem is to find a reliable means of producing it and a measuring instrument sensitive enough to assess it. The data of the present study do not lend themselves to interpretations of the superiority of one independent variable over another. However, some generalizations concerning the various dependent variables which have been employed are possible. The failure of the present study and a number of other studies to find satiation effects using the Semantic Differential technique leads one to question its adequacy in this situation. The only con- clusions seem to be that either all of the investigators who have attempted to replicate the McGill studies have overlooked some pro- cedural detail, or that only Canadian subjects satiate on this vari- able. The first alternative must remain a possibility. The second is rather unpalatable to all concerned. The Total Number of Associates variable is of little use due to the possibility of a memory factor reducing the size of the effect. Number of Associates in the First 5 seconds did not prove to be sensitive enough to measure satiation. However, due to the conceptual fidelity of a latency measure to the definition of satiation (i.e. a 29 30 decrease in the strength of meaning), research on a similar but more sensitive measure than the one employed in this study is definitely in order. Based upon the previous literature and the present study, a Popularity of Associations variable seems to hold the greatest promise for satiation studies, at present. Although no significant satiation effects were found on this variable in the main portion of the present study, the results of the Satiation Time Control study and the compara- tive success reported in other studies with this variable suggests that satiation may be fairly reliably measured by its use. An in- crease in the size of the independent variable should lend more weight to this measure. REFERENCES Bassett, M. F., & Warne, C. J. On the lapse of verbal meaning with repetition. Amer.fiJ. Psychol., 1919, 66, 415-418. Das, J. P. Hypnosis, verbal satiation, vigilance, and personality factors: a correlational study. J. abnorm. soc. Psychol., 1964, 66, 72-78. Don, V. J., & Weld, H. P. Lapse of meaning with visual fixation. Amer. J. Psychol., 1924, 66, 446-450. Floyd, R. L. Semantic satiation: replication and test of further implications. Read at Midwestern Psychological Association, Chicago, May, 1962. Jakobovits, L. A. & Lambert, W. E. Semantic satiation among bilin- guals. J. exp. Psychol., 1961, 62, 576-582. Jakobovits, L. A. & Lambert, W. E. Semantic satiation in an addition task. Canad. J. Psychol., 1962, 62, 112-119. Jakobovits, L. A. & Lambert, W. E. Mediated satiation in verbal transfer. J. exp. Psychol., 1962, 66, 346-351. Jenkins, J. J., Russell, W. A. & Suci, G. J. An atlas of semantic profiles. Amer. J. Psychol., 1958, 1;, 688-699. Kanungo, R. & Lambert, W. E. Semantic satiation and meaningfulness. Amer. J. Psychol., 1963, 26, 421-428. Kanungo, R. N., Lambert, W. E., & Mauer, S. M. Semantic satiation and paried-associate learning. J. exp. Psychol., 1962, 66, 600-6070 Lambert, W. E. & Jakobovits, L. A. Verbal satiation and changes in the intensity of meaning. J. exp. Psychol., 1960, 69, 376-383. Lambert, W. E. & Jakobovits, L. A. The case for semantic satiation. 1963, McGill University (Mimeo.). Mason, M. Changes on the galvanic skin response accompanying reports of changes in meaning during oral repetition. J. gen. Psychol., 1941, 26, 353-401. 31 32 Noble, C. E. An analysis of meaning. Psychol. Rev., 1952, 62, 421-430. Osgood, C. E. Method and theory ip_experimental psychology. New York: Oxford Univer. Press, 1953. Paul, C. Generalized inhibition of verbal associations. J. verb. Learn & verb. Behav., 1962, lg 162-167. Reynierse, J. H. Semantic satiation and generalization. Unpublished MA thesis, Michigan State University, 1963. Severance, E. & Washburn, M. F. The loss of associative power in words after long fixation. Amer. J. Psychol., 1907, 16, 182-186. Smith, D. E. P. & Raygor, A. L. Verbal satiation and personality. J. abnorm., soc. Psychol., 1956, 65, JZJ-J‘U. Warren, R. M. & Gregory, R. L. An auditory analogue of the visual reversible figure. Amer. J. Psychol., 1958, 1;, 612. Warren, R. M. Illusory changes of distinct speech upon repetition-- the verbal transformation effect. Brit. J. Psychol., 1961, 62, 249-258. Warren, R. M. Illusory changes in repeated words: Differences between young adults and the aged. Amer. J. Psychol., 1961, 16, 506-516. Wertheimer, M. The relation between the sound of a word and its meaning. Amer. J. Psychol., 1958, 11, 412-415. Wertheimer, M. & Gilliss, W. M. Satiation and the rate of lapse of verbal meaning. J. gen. Psychol., 1958, 69, 79-85. Yelen, D. R. & Schulz, R. W. Verbal Satiation? J. verb. Learn & verb. Behav., 1963, 1, 372-377. APPENDICES 33 APPENDIX A 34 35 TASK INSTRUCTIONS Semantic Differential The purpose of this study is to measure the meanings of certain words by having you judge them against a series of descriptive scales. The scales and the words which are to be judged, will be presented individually. You are to point, with this pointer, to the position on the scale which, in your estimation, most nearly agrees with the meaning of the word. Please make your judgements on the basis of what these words mean to you. Here is how you are to use these scales. If you feel that the word is very closely related to one end of the scale, you should point as follows. If you feel that the word is quite closely related to one or the other end of the scale (but not extremely) you should point as follows. If the word seems only slightly related to one side as opposed to the other side(but is not really neutral) then you should point as follows. The direction toward which you check, of course, depends upon which of the two ends of the scale seem most characteristic of the word you are judging. If you consider the word to be neutral on the scale, both sides of the scale equally associated with the word, or if the scale is completely irrelevant, unrelated to the word, then you should point to the middle space. Important: Point in the middle of the spaces. Do not point to the boundaries. Do not try to remember how you checked similar items earlier in the test. Make each item a separate and independent judgement. 36 Make each judgement as rapidly as possible. It is your first im- pression, the immediate "feelings" about the words that we want. 0n the other hand, please do not be careless because we want your true impressions. Are there any questions about your task? If not, let us try a sample item so that you can get a feel for what is required. Let's begin. Word Association This is a test to see how many words you can think of and write down in a short period of time. You will be given a kgy word and you are to write down as many 66665 words which the key word brings to mind as you can. These other words which you write down may be things, places, ideas, events, or whatever you happen to think of when you see the key word. For example, think of the word, KING. Some of the words or phrases which KING might bring to mind are written below: queen kingdom King Cole England ruler imperial Sky-King kingfish No one is expected to fill in all the spaces on a page, but write as many words as you can which each key word calls to mind. Be sure to think back to the 551 word after each word you write down because the test is to see how many other words the key word makes you think of. A good way to do this is to repeat each key word over and over to yourself as you write. APPENDIX B 37 ABRUPT 38 WORD NORMS1 1.2788-19 sudden 1.2041-16 quick(ly) 1.0414-11 stop 07782? .6990- .3010- .0000- 6 short 5 halt, sharp 2 end, ending, fast 1 glad, nasty, now, pause, statement, unceasing, volcano BEGGAR (illegible: 1) 1.2788-19 poor 1.1761-15 thief .9031- .7782- .4771- .3010- 00000- 8 bum 6 money 3 hunger(ry), tramp 2 alms, hobo, rich man 1 beaver, beg, book, cup, dirt, drunk, food, king, man, penurious, poor man, poverty, rogue, scarce. BODKIN (no response: 5) .6990- .6021- 03010- .0000- 5 body 4 nothing 2 blank, book, Brooklyn, don't know, elf, gremlin, man 1 author, baby, bare, bark, batch, bay, book, boot, bridge, Buddhist, catkin, chemistry, child, clothes. cow. delta, l The appears in format of these norms is as follows: the stimulus word capital letters. Below it are the log frequency and frequen- cy of the responses. 39 BODKIN (con't.) e 0000- 1 BOTTOM 1.6232-42 .6021- 4 .4771- 3 .3010- 2 .0000- 1 HAMMER 1.5315-34 .6990- 5 .6021- 4 .4771- 3 .3010- 2 SISTER 1.6721-47 .4771- 3 .0000- 1 doll, dress, sat, experiment, gadkin, ghoul, hat, horse, jacket, jersey, John, Kathy, mannequin, nonsense, nun, nut, odds, odkin, pin, Prof., Baken, pumpkin, relative, sodkin, surry, stockings, stupid, toboggan, wagon, what, wisdom, word. top lake end, pit, up floor bare, bell, black, boat, bottomless, bottoms up, button, butt, cliff, deep, down, empty, inside, low ocean, pot, seat, ups. nail(s) pound, saw song mallet anvil, head, hit, Peter, Paul & Mary, sickle, tools, beat, bell, chisel, crush, head, John Brown, noise, record, shark, sledge, wood, workshop. brother girl, none big sister, cold, confidant, Dianne, four, Johanna, Judy, Karen, little, love, Mary, Mary Ellen, me mother, my, niece, Rosanne, Sally, sibling, sister-in-law, smaller, three. SOMBER 1.0792-12 1.0414-11 .8451- 7 .3010- 2 .0000- 1 SUDDEN 1.3010-20 1.0000-10 .8451- 7 .6990- 5 .4771- 3 .3010- 2 .0000- 1 STARVING 1.5798-38 1.2304-17 .3010- 2 .0000- 1 SUNDAY 1.5135-33 1.3617-23 4O sad quiet, sleep sober dull, drunk, serious bland, bleak, bomber, color, dark, dread, dreary, dry, funeral, gentle, grey, happy, man, moody, morbid, night, noisy, peaceful, person, polite, puritan, severe, silent, solemn, still, surly, swift, winter. quick stop abrupt fast now, start, suddenly death, surprise all, at once, fright, halt, happening, hit, jump, last, life, movement, noise, panic, right, second, sharp, short, slam, soon, still, swift. hunger(ry) food dying, man, people, skinny animal, children, eating, hamburger, malnutrition, now, pain, poverty, set stomach, thirst, tomorrow. church Monday 41 SUNDAY (con't.) .6990- 5 school .4771- 3 sabbath .3010- 2 weekend .0000- 1 day, dinner, ice cream, never on, relax, Saturday, skip, supper, week. 599%. {Bi six... men use cm. 1/1 [:9 s" . 5’; (.3 U l '1- ; w."ij (Nile? L9H "Ilili‘lll‘lflll“Elli“