THE VERBAL CONTEXT PARADIGM Thesis for the Degree of Ph. D. MICHIGAN STATE UNTVERSITY SANDRA JEAN WOOLUM 1970 Date This is to certify that the thesis entitled THE VERBAL CONTEXT PARADIGM presented by Sandra Jean Woolum has been eccepted towards fulfillment of the requirements for. Major prof sor § \ 4 LIB}? 4RY V Michig; 13m: Unm .8127; T I (ti-Ya mm 7/2 fizz/V ABSTRACT THE VERBAL CONTEXT PARADIGM - BY Sandra Jean Woolum The context paradigm is a study of reading that focuses on the sentence level of analysis, particularly on the way that sentences reveal the meaning of unknown embedded elements. Essentially this method requires that a subject first guess what a nonsense word represents and then modify that guess on the basis of the information presented. The study most similar to this is Werner and Kaplan's (1950) word-context test. Using similar sentences, they discovered a developmental trend in children's abilities to abstract the meaning of an embedded element. However, they did not attempt to differentiate between sentences. Here the emphasis has been on measuring the differ- ences between sentences so that sentences that tend to reveal the meaning of unknown elements can be separated from those that do not. The method utilized is called the PUJOY technique. Subjects first read a nonsense word and Sandra Jean Woolum make a guess in the form of a word or a drawing. They then read a sentence containing that word and guess again. Thus the first page of a booklet may say, "Draw a PUJOY," and the second may say, "A PUJOY is green. Draw a PUJOY." The number of subjects who change their guesses after reading a sentence is taken as a measure of the information in that sentence. When the correct concept is known the number of changes to a correct concept is taken as a measure of the xalgg of that sentence. Several pilot studies were conducted. The first one required that subjects draw PUJOYs and GIDAKs and then modify the drawings on the basis of further information. Subsequent studies allowed subjects to either make a new guess or write, "same." "SALOP for sale," a story written about a SALOP, demonstrated that generally more "sames" are written as the story progresses. Order of presentation thus interacts with information. In an attempt to separate information from order, "What they said about the SALOP," counterbalanced sentence order and demonstrated that information can be measured independently. It also revealed strategies of concept formation which may be similar to conservative focusing and focus gambling. A final pilot study compared the PUJOY measure of information to information-theory information. The first major study used adjectives, selected according to their frequencies in English, in the sentence Sandra Jean Woolum frames A TOSOY is __;___and A TOSOY is not _____. By testing each adjective in both a positive and a negative sentence, it was possible to determine that the positive sentences were generally more informative than the negative ones. There was some indication that the difference between positive and negative sentences was greater for infrequent adjectives than for frequent ones. The other major study investigated the relationship between information and value. Information considers all changes but value considers only changes to a correct concept. For this experiment six concepts were chosen and subjects were asked to write sentences about these concepts. Some of these sentences were then presented to a second group of subjects and from their guesses a measure of the information in each sentence was obtained. Sentences ranking high, middle, and low in information were then selected for each concept. These sentences were presented to a final group of subjects who tried to guess the original concepts. As predicted, subjects who read high information sentences made more correct guesses than subjects who read low information sentences. An analysis of variance demonstrated a significant difference due to information for both recall and recognition tasks. However, the concepts without labels seem to be operating differently than those with labels. THE VERBAL CONTEXT PARADIGM BY Sandra Jean Woolum A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1970 ACKNOWLEDGMENTS I would like to thank my advisor, Dr. D. M. Johnson and the members of my committee, Dr. A. M. Barch, Dr. J. L. Phillips and Dr. J. S. Uleman for their contri- butions to this thesis. Although their comments have been helpful, I am equally appreciative of things unsaid. It is easy to criticize something that is new and unsophisti- cated but if the PUJOY technique has made any progress in these three years it is principally because it has not been prematurely barraged by reminders of what it is not. I would also like to thank some of those who distracted me from this work, especially my students and the members of the 151 seminar who continually reminded me that the world is real and people who happen to enjoy PUJOYs shouldn't forget it. Special thanks goes to Leslie Smith who created the original SALOP, to the 1,000 subjects who participated in this research, and to my family for their continued support and encouragement. ii TABLE OF CONTENTS Page INTRODUCTION 0 O O O O O O O I O O O O O 1 Instructions . . . . . . . . . . . . 5 LITEMTURE O O O I O O O O O O O C C O C 6 DEFINITIONS . . . . . . . . . . . . . . 20 Operational Definitions . . . . . . . . . 21 Procedure. . . . . . . . . . . . . . 22 PILOT STUDIES--PART I o o o o o o o o o o o 24 Effects of Order . . . . . . . . . . . 25 Strategies . . . . . . . . . . . . . 30 Information-Theory Inf rmation. . . . . . . 39 POSITIVE AND NEGATIVE INFORMATION--PART II . . . . 43 Method. . . . . . . . . . . . . . . 44 SUbj ects O O O O O O I I O O O O O 4 4 Materials . . . . . . . . . . . . 44 800k lets O O O O O O O O O O O O 45 Conditions . . . . . . . . . . . . 46 Results . . . . . . . . . . . . . . 47 Discussion . . . . . . . . . . . . . 52 INFORMATION AND VALUE--PART III. . . . .' . . . 55 Part A--Obtaining the Sentences . . . . . . 56 Part B--Measuring Information . . . . . . . 60 Part C--Measuring Value . . . . . . . . . 64 Results . . . . . . . . . . . . . . 67 Discussion . . . . . . . . . . . . . 80 PROSPECTS O O O O O O O O O O O O 0 O O 84 SWRY I O O O C I O O O O O O O O O O 88 iii REFERENCES APPENDICES Appendix A. . B. . C. . D. . E. . F. . iv Page 92 95 102 111 115 127 130 IIIT II II will [ 'III III 10. LIST OF TABLES Number of Concept Changes for Each Sentence of "A SALOP For Sale." . . . . . . . Sentences Ranked in Order of Decreasing Information in "What They Said About the SALOP O n O O O O O O D I O O O 0 Information and Frequency Analysis of the PUJOY Test . . . . . . . . . . . Design to Counterbalance Information and concept 0 O I O O O C O O O O 0 Analysis of Variance for Recognition Test, Using Number Correct After the Last sentence 0 I O O O I O O O I I 0 Analysis of Variance for Recall Test, Using Number Correct After the Last Sentence. . Number Correct on Recognition Task for Each Concept at Each Information Level . . . Number Correct on Recall Task for Each Concept at Each Information Level . . . Number Judged Correct on Recall for REPOL and HETAB, After All 10 Sentences. . . .. . Analysis of Variance for Individual Concepts. Page 28 31 41 66 72 73 75 76 77 77 Figure 1. 2. 10. 11. 12. LIST OF FIGURES A GIDAK Holds Water . . . . . . . . Number of Concept Changes with Sentences Presented in Fixed Order. (A SALOP For sale) 0 O O O O O O O O I O 0 Number of Concept Changes with Sentences Presented in Counterbalanced Order. (What They Said About the SALOP). . . . . E Example of Conservative Focusing. . . E Example of Focus Gambling . . . . . An Example of Focus Gambling That Didn't Pay Off 0 I O O O I O O I O O 0 Information in Positive and Negative Sentences With Adjectives of Various Frequencies in English (Group I). . . . . . . . Information in Positive and Negative Sentences With Adjectives of Various Frequencies in English (Group II) . . . . . . . A Possible Relationship of Information to Frequency: Data Points From Adjectives Group I and Group II are Considered in Relation to Straight Lines for Positive Negative Information. . . . . . . Scatter Diagram of Positive and Negative Information for Adjectives of Group I . Objects Used to Represent TEMOS and LUREK. Summary of Procedure for Selecting Sentences . . . . . . '. . . . vi of and Page 26 27 29 32 34 37 48 50 51 53 57 63 Figure Page 13. Effects of Information and Number of Sentences on Recognition . . . . . . . . . . 68 14. Effect of Information and Number of Sentences on Recall. . . . . . . . . . . . 69 15. Number of "Sames" Reported on Recall Test in Relation to Information and Number of Sentences for Repol and Temos Data combined 0 O O O O I O O O O O O 79 vii INTRODUCTION This is a study of how reading generates new concepts. Generally stated the research question is, "How does one word come to tell the meaning of another word?" The notion that words somehow reveal the meanings of other words or that reading increases understanding is not at all new. This is the foundation of educational systems. For certainly many words are introduced to us without nonlinquistic reference. They are said to be verbally defined. How many schizophrenics do we see in our first course in abnormal psychology? About as many schizophrenics as unicorns. And yet we are supposed to learn the meaning of the term "schizophrenic" from reading and listening to many sentences in which the word occurs. How can this be? (Brown, 1956, p. 297) Certainly, the acquisition of concepts through context is not a homogeneous process. It is relatively easy for an educated person to pick up new labels. Psy- chology articles often provide a new term of some type, philosophy papers generally provide a score of them, often with subscripts to indicate five or six distinctions for a single word. But the child who does not know the meaning of the "renaissance" is not likely to acquire it through any short set of discourse. Indeed children may not be able to utilize this information at all. Some "concepts" are merely new labels for old, others are new collections of pre-defined attributes, but some are genuinely new and some define new boundaries for the world and transform the unimaginable into the familiar. The importance of this research question must be emphasized because this study will be concerned with a very limited aspect of the problem. The focus here is on sentences as units that are capable of revealing the meaning of one of their own members. Thus, "A GIDAK hobbled down to the river for a swim," indicates that a GIDAK is something that can hobble and swim, perhaps an animal. The sentence as a unit of analysis and the requirement of an embedded label are arti- ficial constraints which make this phenomenon easier to study. In practice this process may extend over long periods of time, involve great collections of discourse and no label may be specifically provided. Here the emphasis will be with measuring information in sentences and with determining variables and co- variables of information. This concern stems from an earlier study as well as from the obvious conclusion that sentences are not equally informative. The term "information" is used in the common, everyday sense as in "information booth." Since it will be operationally defined it is not a loose definition but it does not coincide with the use specified by information theory. References to this specific usage of the term will always be labeled, "information-theory information." Since context is not always an effective means of communication, operational definitions will also be provided for such terms as "meaning" and "concept." The method used here is called the context paradigm or the PUJOY method. Whatever its theoretical problems may be, it is in practice very simple. A subject (g) is given a nonsense word and told to guess its meaning. Guesses are solicited in several forms, labels, drawings or choices on multiple-choice tests. Then the S’is provided with a sentence using this nonsense word and subsequently told to guess again. He may either make a new guess or write "same." The number of new guesses across Se is assumed to represent the relative amount of information in the pre- ceding sentence. Sometimes guesses are also scored as being either correct or incorrect. In this case the number of correct guesses across Se is assumed to represent the value of the preceding sentence. The relation of infor- mation and value is unknown but will be here pursued. When more than one sentence is presented, the matter is more complex. This too will be studied. The study is presented in three parts. The first part presents pilot data on the variability of sentences, on order, on Ss' strategies and on possible relations to information-theory information. The second part tests for the influence of positive and negative information and for frequency of adjectives used to complete the sentence "A PUJOY is ." The third section tests for the relation of information and value. The following example from "What they said about the SALOP," provides its own introduction. Draw a SALOP. An attractive blonde walked up to the SALOP and said, "My dear you really should consider a nose job." Draw a SALOP. The guide explained, "This is the only SALOP in the museum; it is probably our most controversial piece." Draw a SALOP. A woman remarked thoughtfully, "The SALOP is constructed totally of cardboard, I could make one like this for less than a dollar." Draw a SALOP. When a man spotted the SALOP beside the more serious works of art he said, "How did that get here?" Draw a SALOP. A man who had inspected the SALOP carefully said, "It looks exactly like my mother-in-law." Draw a SALOP. A woman said, "I like the SALOP, it reminds me of a Picasso." Draw a SALOP. Instructions In this experiment you will be asked to draw a SALOP. The word SALOP has been made up but it represents something that is real. Your job is to figure out what a SALOP is. Make a guess at first and then turn the page and read what it says. If you decide to change your guess draw another picture. If your first guess is still your best guess write "same" on that page and then go on to the next page. Whenever you change your mind about what a SALOP is draw a new picture. If you do not change your mind write "same" on that page. Please do not look ahead; do one page at a time. Remember that a SALOP is something that really exists and that all of the things that are said about it could be true. 4"— LITERATURE This is a phenomenon with many names. The early educational studies refer to vocabulary development: some writers are concerned with concept formation and others with conditioning and association. The theoretical splits may be even wider than the terminology implies, but the search here is for either evidence or insight and since both are somewhat limited it does not help to be particu- lar. As theoretical differences are not particularly relevant to this study all analyses will be considered as instances of the same phenomenon--that words somehow cell the meanings of other words. Early acknowledgment of this type of learning may be found in Ogden and Richards (1930, p. 213) It is rare for words to be formed into contexts with non-symbolic experience directly, for as a rule they are learnt only through other words. We early begin to use language in order to learn language. Dewey (1933, p. 240) writes on increasing the vocabulary. This takes place, of course, by wider intelligent contact with things and persons and also vicariously, by gathering the meanings of words from the context in which they are heard or read. References to early studies of vocabulary develop- ment by reading may be found in Carrol (1964) and Woolum (1968). Generally, these studies indicate that context is not a very effective means of increasing vocabulary but evidence varies and a recent study (Johnson & Stratton, 1966) demonstrates that context is just as effective as learning by synonyms, definitions or classification. Acknowledgment of concepts learned through context as a distinct kind of concept formation may be found in Bourne (1966), Davis (1966), and Carroll (1964). Carroll outlines the differences between school concepts and the types of concepts that are studied in the psychology labo- ratory. There is a gap between the findings of psychologists on the conditions under which very simple concepts are learned in the psychology laboratory and the "for real" concepts that are contained in the curriculum of the schools. It is not self-evident that there is any continuity at all between learning "DAX" as the name of a certain geometrical shape of a certain color and learning the meaning of the word ”longitude" (Carroll, 1964, p. 179-80). He notes that the attributes on which school concepts are based are frequently verbal, depending on elements of meaning that cannot be represented in terms of simple sensory qualities. He also provides an analysis of some difficult concepts and presents a verbal formula to define the word "tort.” Beyond acknowledgment the topic is rarely given any attention. Mowrer (1954, p. 663), however, defines a sentence as a unit of conditioning and thereby relates this phenomena to a large set of experimental literature. Let us explore now, instead, the proposition that in communication we are not transferring meanings from person to person so much as we are transferring meanings from sign to sign within a given person, within a single mind. In Mowrer's analysis of the sentence, "Tom is a thief," a fractional component of the meaning of "thief" is con- ditioned to the meaning of "Tom." Mowrer explains how words may serve as mediators and presents analogies from animal research. He assumes that basic meanings for "Tom" and ”thief" are already known however. This prevents the two words from becoming synonymous. Osgood (1963) notes that Mowrer's analysis will not explain (1) the possibility of understanding the sentence but not believing that Tom is a thief, (2) the influence of grammatical structure so that "Tom is a perfect idiot" does not condition "Tom" to both "perfect" and "idiot," and (3) the possible modification of the predicate rather than the subject. Osgood then defines two types of phrase analysing processes with different types of cognitive interaction. Homeostatic interactions exist with qualifying phrases like ”listless nurse," or "sincere scientists." Multiplicative interactions exist with quantifying relations like ”somewhat evil" or "very charming." Meanings of the phrases can then be predicted from meanings of components using semantic differential scales. Osgood refers to evidence. Making an analogy to computer storage of infor- mation he also notes that individual sentences do not permanently "repunch" meaning cards thus leaving meaning in constant flux. In the long run however, sentences may bend or even repunch meaning cards. In this respect nonsense words may be considered to be blank cards with no resistance to being bent, eager indeed to be punched. Some related experimental evidence is given by M. G. Johnson (1967). He presented nonsense syllables in subject and predicate positions in active and passive sentences. Thus, "The XEK hit the WUQ,” "The RIW was comforted by the FOJ.” He then obtained semantic differ- ential ratings which demonstrated that subjects were rated as more animate than objects with active subjects being more animate than passive subjects, and passive objects being more animate than active objects. Thus, the position a word occupies in a sentence influences the rated meaning of that word. From a similar theoretical framework, Staats (1968, p. 517) presents an example and an analysis. "He was a very hufod animal, always running here and running there, never still for a moment. Hufod animals are the healthiest, however. Choose one that is hufod and you will always be glad you did. Hufod ones are more fun besides." The first sentence has the effect of pairing the new word "hufod" with "always running" and "never still." The meaning response elicited by these words would be conditioned to "hufod." On such an occasion, the reader will ordinarily also have learned to go back and give himself repeated trials so 10 that the meaning responses are strongly conditioned to ”hufod." Staats thus takes the notion of repeated trials with the same material quite seriously. He explains that the reader may look in quick succession at such pairs as "hufod-- always running," and ”hufod--never still." Taking Staats literally, this phenomena might be studied by manipulating exposure time and presentation rate but it will not be here included. At an earlier date, and in an entirely different framework, Werner and Kaplan (1950) demonstrated a develop- mental trend in children's ability to abstract a word from its context and discover its lexical concept. The 125 children studied were between the ages of eight and thirteen. The word-context text provides twelve sets of six sentences. A set describes a lexical concept by em- bedding an artificial word in each of the sentences. Wérner and Kaplan were not particularly concerned with sentences. They wrote their own, placing sentences that they considered to have the most definitive clues last in the sequence. Some of the concepts included in the test had single English labels, others did not. 1. If you eat well and sleep well you will HUDRAY . 2. Mrs. Smith wanted to HUDRAY her family. 3. Jane had to HUDRAY the cloth so that the dress would fit Mary. 4. You HUDRAY what you know by reading and studying. 5. To HUDRAY the number of children in the class there must be enough chairs. ll 6. You must have enough space in the book case to HUDRAY your library. HUDRAY for example has no single English substitute but means grow, increase or expand. All of the sentences are relatively complex, some are real riddles. "The more you take out of a CONTAVISH the larger it gets." Werner and Kaplan differentiate several types of possible errors, with younger children tending to fuse the meaning of the word with that of the sentence. For example one child responded to number six above, "HUDRAY means that you should have a good library." The conclusion that children under age eight or ten cannot make use of this type of data seems quite strange. Textbooks written for this age level and for even younger children often define new concepts internally, but the materials provided are much simpler than the ones used by Werner and Kaplan. Thus a third-grade science textbook provides this explanation of "algae." The green color comes from tiny plants called algae. Some algae were on the aquarium plants when you got them. Like all green plants algae take in carbon dioxide. They make food with it. They grow. They give off oxygen. The algae are not harmful. Snails eat some, and so do some fish (Schneider & Schneider, 1965, p. 62). Of course this does not mean that the children really understand algae but it presents an interesting research question. Syntax, particular words like "called," prior knowledge (certainly relevant with "oxygen” and "carbon 12 dioxide" in this example) and sentence information may be important variables. Brown (1956) presents a criticism of the Werner and Kaplan study, noting that the children sometimes used the phonemic attributes of the artificial words and knowledge of English sentence structure as possible clues. He also considers possible "strategies" for this type of concept formation, defining the task as one of determining con- ditional probabilities between sentences. He notes that some of the children's guesses had higher probabilities for individual sentences than did werner and Kaplan's solutions. Woodward (1963, 1968) developed a test of structural meaning, (that derived from word form and arrangement oper- ating in a system) and tested normal hearing and deaf children of ages 13-15. Thus: The igging nabe alped into the yost. a. What was the nabe doing? b. Which nabe went somewhere? c. What do you think "alped" means? A gab was lanning in the carf. He gumphed a trig. The gab sempt the trig lan. a. What did the trig do? b. What does this story tell us? What happened to the gab. What the trig did to the gab. What the gab did to the trig. 13 The hearing children were more successful at this task than the deaf children. All of the children are above the ages that werner and Kaplan tested but this test might also be revealing with younger children. Woolum and Johnson (1968) studied the effect of number of presentation on concept acquisition. Stories were written which used real English words of low frequency. Each word was used once in each of twelve sentences of a story. Beside the complete story, presentations of O, l, 2, 4 and 8 sentences were also created by randomly se- lecting sentences from the story. Realizing that sentences differ in their communication ability, but not knowing what to do about it, the author attempted to equate the sentences by intuition. Although clear trends in the effect of number of instances were shown, it was equally clear that some sentences were considerable more powerful than others. In a story using the word "marmoreal" (which means marble- like) random selection probably provided the most infor- mative sentence for inclusion in the one—sentence con- dition. This sentence alone was more valuable than any other condition, although the eight- and twelve-sentence conditions included the same sentence. Deese (1967) also investigated concepts learned from sentences. A definition was given to the first §_who wrote ten sentences which were given to the next S and so on for four gs. Each § wrote both a definition and ten sentences but only the sentences are passed on. Deese 14 investigated concepts that were readily coded into fa- miliar English words and concepts that could not easily be coded. Definitions and sentences were then rated for acceptability and appropriateness respectively. Little change in appropriateness was found for the sentences con- sidered across §s but definitions of uncoded concepts for the third and forth Se in the chain hit almost the bottom of the acceptability scale while ratings of definitions for familiar concepts did not change at all. The difference between concepts with English labels and those without is interesting, and will be considered in this study. The relation of this study to information theory is tricky. Within the information theory framework some investigators have studied the degree of statistical con- straint exerted upon a given word by its surrounding context. Shepard (1963) randomly deleted words from passages containing one to forty words and measured both uncertainty and production rates. Uncertainty was measured by having the §s guess each letter of the missing word. gs were also asked to produce as many words as possible to go into the blank space during five-minute intervals. He found that average rate of listing words dropped from twenty words per minute when no context was provided to about two words per minute when the maximum context was provided. Individual sentences are not of particular 15 concern to Shepard but he does note that different contexts of the same length do not exhibit the same degree of con- straint. Aborn, Rubenstein and Sterling (1959) dropped one word from sentences of six, eleven or twenty-five words. They also considered position in the sentence and word class. Significant differences were found in relation to all three variables. Word classes in order of ease of prediction were (1) function words, (2) verbs, (3) pronouns, (4) adverbs, (5) nouns, and (6) adjectives. Greater con- straint was found for words in middle positions than for words at either the beginning or the end of the sentence, and predictability increased with sentence length. In some ways this task is similar to the PUJOY method when words with real English labels are used. But the concern here is specifically with the differences between sentences and the results have never been scored according to the variety of guesses provided, though that could be done. Furthermore the PUJOY method sometimes uses unfamiliar concepts with no English labels and often more than one sentence is provided. It is however, not sur- prising that sentence length is important to both measures. These results are presented in Part III of the study. Information theory may also be related to the PUJOY method if the notion of ”having information" is conceived of as predictability. Whatmough (1964, xvii) writes: 16 Words used with the greatest frequency are low in semantic range and content, important as they may be grammatically and functionally (the, and, or, not); content is enhanced by relative rareness (antidishes— tablishmentarianism) as is affectivity (obscene or profane words); cliches or small talk may be common enough, but as we say "they mean little or nothing." Lyons (1968, p. 415) also discusses the quantifiability of "having meaning." In any case, it is certainly in accord with general, everyday usage to say that the "meaningfulness" of utterances, and parts of utterances, varies in inverse proportion to their degree of "expectancy" in context. He further notes however that this quantification would depend on our ability to identify the contextual features which determine "probabilities of occurrence" and that it is very unlikely that we will ever quantify "having meaning” in this precise sense. Lyons also writes that "what" meaning an element has in a given context may be unrelated to "how much" meaning it has in relation to the elements with which it has been compared. Fortunately, such controversies may be left to the linguists. The quantification of meaning is difficult if not ridiculous. However, the notion of predictability may be considered in relation to the variable of adjective frequency in Part II of this study. It does not prove to be a very powerful variable. Information theory really has little to say about semantic information. Bar-Hillel and Carnap (1953, p. 503) write: 17 Theory of information, as practiced nowadays, is not interested in the content of the symbols whose infor- mation it measures. The measures as defined for instance by Wiener and Shannon, have nothing to do with what these symbols symbolize, but only with the frequency of their occurrence. They then present a theory of semantic information, based on Carnap's theory of inductive probability. The contents of the symbols will be decisively involved in the definition of the basic concepts of this theory and an application of the concept and the theorems concerning them to fields involving semantics thereby warranted. But precaution will still have to be taken not to apply prematurely these concepts and theorems to field such as psychology and other social sciences, in which users of symbols play an essential role. They go on to define two measures of semantic information and provide interesting alternative formulas for the combi— nation of information but their system is obviously in another world, one with logically true, logically false and logically disjunct statements. It does seem that empirical formulas for the combi- nation of two sentences could be discovered, even with methods as crude as the one proposed here. In particular synonymous and independent sentences should be detectable. This possibility will be considered in a section on future prospects. Ghosts of information theory pursue PUJOYs in yet another realm. It is an artificial world where things exist in strict multivariate dichotomies. Here it is not the probability of sentences but the probability of concepts that must be considered. This relationship will be in- vestigated in Part I of the study. 18 In summary, it is obvious that the PUJOY method was originally conceived in blissful ignorance of information theory but post hoc analogies are many: (1) within sentences contextual constraint may determine ease of guessing the missing word, (2) probability of occurrence may be considered in relation to ”what sentence is likely to come next," (3) a theory of semantic information such as that proposed by Bar-Hillel and Carnap may have some re- lation to the interaction of sentences, and (4) probability of occurrence may be considered in relation to the par- ticular concept being described when highly artificial worlds are considered. It seems that semantics too would have something to say about information in sentences, but linguists generally talk about competence not performance (Chomsky, 1965) and the psychological snatching of linguistic concepts is often unwelcome and unwarranted. In relation to the PUJOY method two tempting analysis may be considered. The reduction of long complex sentences into kernel sentences a la Chomsky would be innocent enough and the subsequent combination of information might be amusing though this would be surface, not deep structure. A test of phrase markers, or supposed phrase markers a la Katz and Fodor (1963) seems more serious and less revealing though the minimum number of such sentences needed to uniquely define a given concept might be determined. They are talking about a theoretical 19 analysis of meaning, not an empirical test of communi- cation. In conclusion,--PUJOYs traipse on broken ground, hufod animals and igging nabes have come before them, not to mention Tom the thief and Mrs. Smith who wanted to HUDRAY her family. But this is an unsettled land with plenty of room for newcomers. DEFINITIONS The definitions provided here are presented to clarify the method involved. They make no claims to being the best possible definitions but are not intended to deviate significantly from common usage. Specification, however, tends to involve restriction. Concepts are hypothetical constructs that enable individuals to make consistent decisions about subdi- visions of universal sets or events. Just as a person may be certain or uncertain about a decision a concept may be yaggg_or well defined. (These terms are not used in the sense of fuzzy around the edges vs not so fuzzy, but in the sense of certain and uncertain.) When one person believes that he knows a concept and another person seems to have the same concept, the first person judges that concept as correct. The process by which a concept becomes well-defined and the one by which it approaches a correct concept may or may not coincide. Meaningful words are concepts with labels. Nonsense words are labels which have not been paired with concepts. Two other possible combinations of the label- no label, concept-no concept dichotomy exist. Though they 20 21 will not be of concern here, there might be concepts with no labels (free floating concepts) and distributions, of things or ideas perhaps random, with neither labels nor concepts (double nonsense). Sentences may or may not contribute to a well defined and/or a correct concept. Sentences that au- thorities believe to be related to the correct concept are relevant, others are irrelevant. Sentences that tend to contribute to new concepts (whether they are better de- fined or not) are relatively informative, others are up: informative. Sentences that tend to contribute to a cor- rect concept are relatively valuable. Operational Definitions In this situation an individual's decision in the form of a response is taken to represent his concept. Though this representation may be loose, a new decision is taken to be indicative of a new concept. Vague concepts are relatively sensitive to information, well defined concepts are relatively insensitive to information. Thus if information is held constant, concepts which change may be called vague relative to concepts which do not. It is suspected that vague concepts are closely related to (newly formed from) nonsense words. A concept is judged as correct if the response provided matches the response that an authority would provide. Authorities are indi- viduals who are believed to know a concept because it is 22 a meaningful word in a language with which they are fa- miliar or because they have been presented with an instance of this concept. Nonsense words are randomly selected sequences of letters that have not been previously paired with a concept. (This refers to individuals, no super-organic world of concepts is assumed.) Here nonsense words are words which do not occur in English. Meaningful words are those which have previously been paired with a concept. Here this includes common English words, as well as words which are derived from nonsense words, (by being paired with a concept). Sentences written by authorities who have been instructed to write about the concept are relevant. Sentences that tend to precede new concepts (across gs) are relatively informative. Degree of informativeness is quantified as information. Sentences that tend to precede correct concepts (across gs) are valuable, quantified as value. Procedure A g is given a nonsense word and told to guess its meaning. Once he has made this decision he has a concept, however 23223, and the word is a meaningful word. Then he reads a sentence which may or may not be relevant and makes a new decision or writes "same." The number of new decisions across Se is an index of 23 information in the proceeding sentence, the number of cor- rect decisions across §s is an index of the yglpg_of that sentence. On the first guess (before any sentences are pre- sented) all concepts are assumed to be equally vague, after a sentence is presented any decision may be an interaction of vagueness and information. In effect this means that order of presentation is important. For all of the studies presented in Part I and Part II, (with the exception of the one on information- theory information) no correct concept is defined, the concern is only with information. In Part III both infor- mation and value are studied. Sentences which are known to be irrelevant are not included in this study but the re- lation of information and value is discussed in relation to relevant and irrelevant information. P ILOT STUDIES--PART I For the first study, each § drew a picture on each page of a small booklet. The first page read simply, "Draw a PUJOY," or "Draw a GIDAK," but subsequent pages read, "A GIDAK is purple. Draw a GIDAK," or "A GIDAK is ugly. Draw a GIDAK." Several results from this study were important for later developments of the method. PUJOYs are rounder than GIDAKs so the nonsense words themselves cannot be ignored. Many gs didn't consider PUJOYs or GIDAKs to be real things and drew anything at all. Instructions were later modified to indicate that PUJOYs and such really exist. Some gs drew new pictures on each page, others seemed to draw exactly the same picture repeatedly. -The "same" alternative was introduced in later studies to make the identicalness of pictures the §fs own decision. Finally, the order of sentences appeared to affect the end result. This effect was not specifically examined in later studies so an example will be presented here. The sentence, "A GIDAK holds water" presented initially tended to yield a vase which was later modified to a cup with an ugly face. But one § had already drawn a person, made it purple and put 24 25 on an ugly face. The sentence, "A GIDAK holds water" did not change things very much (Figure 1). Effects of Order At this point the number of new pictures drawn after each sentence (as opposed to the number of "sames") was taken as an approximation of the information in the preceding sentence, and the influence of order of pres- entation was tested. For the first study, "A SALOP for sale," 38 Se drew SALOPs after each of seven sentences, always presented in the same order. The general trend indicated that information decreases as the story progresses but there are reversals with two of the sentences (Figure 2, Table 1). It is apparently more informative to know that a SALOP reminded the man of his mother-in-law than it is to know that the man had decided to buy the SALOP. In an attempt to separate the information in a sentence from the order of presentation a story was written so that sentences could be presented in any order. The sentences were then counterbalanced so that each occurred in each position first through sixth in one of the six versions of ”What they said about the SALOP." The results from 90 Se, 15 in each condition are presented in Figure 3. It is also possible to analyze results for individual sentences across order. The sentences and aver- age number of new pictures they generated are presented in 26 Figure l.--A GIDAK holds water. Number of Concept Changes 27 38-- . . \. .1. .T. o 304)- {I- i- 0 4r \. 0 «w 2041 ¢r C 101 T ‘ 1. 0 tr e: : Le a: e: r47 2+; 1 2 3 4 5 6 7 8 Order of Presentation of Sentences Figure 2.--Number of concept changes with sentences presented in fixed order. (A SALOP for sale.) 28 TABLE l.--Number of Concept Changes for Each Sentence of “A SALOP For Sale.” 1. Draw a SALOP. 38 2. One day an artist took time off from his more serious work to make a SALOP. Draw a SALOP. 35 3. He made it of cardboard and paint rather than using metal because he did not want to waste much time with the SALOP. Draw a SALOP. 24 4. When all the parts were finished he assembled the SALOP and stood it on the floor. Draw a SALOP. 23 5. Everyone who came to the studio laughed at the SALOP which had a long cardboard nose sticking out of its head and one eye on each side of its nose. Draw a SALOP. 30 6. One day a man who was looking at some more expensive statues said that the SALOP reminded him of a Picasso. Draw a SALOP. 23 7. He said that he had always liked Picasso's work but that he had decided to buy the SALOP for a different reason. Draw a SALOP. 4 8. He told the painter that he wanted the SALOP because it looked exactly like his mother-in-law. Draw a SALOP. 12 Average Number of New Concepts H c U1 % 29 H U1 i o Hdb N! U .5 U" as (i— \l Order of Sentences Figure 3.--Number of concept changes with sentences presented in counterbalanced order. (What they said about the SALOP.) 30 Table 2. Again it seems logical to say that there is more information in a sentence that says a SALOP is made of cardboard than in a sentence, presented in this context, that says a SALOP is controversial. Although a control booklet that said, "Draw a SALOP," on every page might duplicate the effects of order, simply due to fatigue of drawing new pictures, it would not duplicate the effects of sentences independent of order. This seems to indicate that both order and sentences are important variables. It is easy to obtain sentence infor- mation independently of order if only one sentence is presented. This plan was followed in some of the subsequent studies. The interaction of information and order is com- plex and depends on the independence of information. It might best be studied with two sentences. Strategies Before proceeding to the other considerations of information it may be noted that two strategies of concept formation were apparent from these drawings and that they seem to be analogous to conservative focusing and focus gambling strategies discussed by Bruner, Goodnow and Austin (1956). With conservative focusing the S retaines the original notion but modifies it enough to accommodate the additional information. An example is presented in Figure 4. With focus gambling the § changes the whole thing each time. Figure 5 demonstrates how different this 31 TABLE 2.--Sentences Ranked in Order of Decreasing Infor— mation in "What They Said About the SALOP." Average Number of Sentence Concept Changes 1. Draw a SALOP.* 15.0 2. A woman remarked thoughtfully, "The SALOP is constructed totally of cardboard, I could make one like this for less than a dollar." Draw a SALOP. 9.3 3. An attractive blond walked up to the SALOP and said, "My dear you really should consider a nose job." Draw a SALOP. 9.0 4. A woman said, "I like the SALOP it reminds me of a Picasso." Draw a SALOP. 7.3 5. A man who had inspected the SALOP carefully said, "It looks exactly like my mother-in-law." Draw a SALOP. 7.1 6. The guide explained, "This is the only SALOP in the museum; it is probably our most controversial piece." Draw a SALOP. 7.0 7. When a man spotted the SALOP beside the more serious works of art he said, "How did that get here?" Draw a SALOP. 5.3 *All of the sentences were presented in counter- balanced order except for the intial sentence--Draw a SALOP. Draw a SALOP. 5.1%. ‘flmsw a The guide explained, "This is the only SALOP in the museum; it is probably our most controversial piece." Draw a SALO', S ACOP TV 32 An attractive blond walked up to the SALOP and said, "My dear you really sho ld consi- der a nose job." ‘ Draw a SALOP 2. . A "-man remarked thoughtfully, "The SALOP is constructed totally of cardboard, I could make one like this for less than a dollar." Draw a SALOP. 4. Earner Figure 4.--An example of conservative focusing. When a man spotted the SALOP beside the more serious works of art he said, “How did that get here?" Draw a SALOP. A woman said, "I like the SALOP, it reminds me of a Picasso.” Draw a 33 A man who had inspected the SALOP carefully said, "It looks exactly like my mother-in-law." Draw a SALOP. 34 1. Draw a SALOP. vmen a man spotted the SALOP ”" beside the more serious works of art he said, "How did that get here?" Draw a SALOP 2. A woman said, “I like the SALOP The guide explained, "This is it reminds me of a Picasso." the only SALOP in the museum; it is probably our most con- Draw a SALOP. troversial piece." 3. Draw a SALOP Figure 5.--An example of focus gambling. 35 An attractive blond walked up to the SALOP and said, "My dear you really should consider a nose job." Draw a SALOP 5. T A man who had inspected the SALOP carefully said, "It looks exactly like my mother-in-law." Draw a SALOP. 7. Sam, A woman remarked thoughtfully, "The SALOP is constructed totally of cardboard, I could made one like this for less than a dollar." Draw a SALOP. 6. 36 technique is but the st final picture fits all of the descriptions. Focus gambling is not always so successful however. In Figure 6 the § seems to retain the general idea of a SALOP until the very last picture. The box drawn on the last page meets few criteria except that it is cardboard. It would be hard to imagine it as controversial, looking like a Picasso, needing a nose job or reminding anyone of his mother-in-law. These are drawings, not concepts, so it may be unwise to get carried away with analogies. The hypothesis that §s will consistently follow one mode or the other is intriguing however and it was investigated with high and low information sentences written for REPOL and HETAB in Part III of this study. Eight gs read low information sentences and ten read high information sentences. "Same" responses were not permitted for this study so a new picture followed each of the ten sentences for each concept (REPOL and HETAB). Booklets with just the pictures, no sentences, were then scored according to number so that HETABs and REPOLs drawn by the same g could not be identi- fied. The four judges, working independently, simply counted the number of times that the picture on one page seemed to incorporate all or part of the picture on the previous page. This is a reasonable simple discrimination and reliability was quite high, .94 for REPOL and .86 for HETAB. Correlations of the number of times that incor- poration occurred were then obtained by pairing the number Draw . SALOP. An a tr <- -1. ond walked up to the SALOP and said, "My dear you really should consider nose job." Draw a SALOP. 3. 9% a 37 A man who had inspected the SALOP carefully said, "It looks -xactly like my mother-in-law." W N ) A woman said, "I like the SALOP, it reminds me of a Picasso." Draw a SALOP. 4. Figure 6.--An example of focus gambling that didn't pay off. 38 The guide explained, "This is the only SALOP in the museum; it is probably our most contro- versial piece." Draw a SALOP 5. Sim A woman remarked thoughtfully, "The SALOP is constructed totally of cardboard, I could make one like this for less than a dollar." Draw a SALOP. When a man spotted the SALOP beside the more serious works of art he said, "How did that get here?" Draw a SALOP 6. 9m 39 for REPOL to the number for HETAB for each S. The corre- lation was .68 for low information sentences and .46 for high information sentences which may indicate that strate- gies play a more important role when information is low and the S is left to his own imagination. This conclusion, obviously speculative at this point, is testable at least to the extent that drawings represent concepts. Further- more comparisons could be made with the strategies used by the same Ss for a traditional concept formation task. Information-Theory Information In the final pilot study, Ss were asked to consider a simpler world in which there are only plants, animals, machines that locomote and machines that do not locomote. The categories may be distinguished by two bits of information—theory information. There are things that live and things that do not. There are things that locomote and things that do not. Lives + _ m +' Animals Machines that g locomote o E ' Plants Machines that do not locomote If a PUJOY is an animal and the S draws a snake on the first trial the instruction, "Draw a PUJOY" is scored as having two bits of information. If the S draws a 40 refrigerator the instruction, "Draw a PUJOY" has zero bits of information. The following diagrams represent changes in categories and the bits of information that would be attributed to the sentence in each case. .__, l . \ J J 2 bits 1 bit -1 bit 0 bits Twenty Ss completed this task, results were scored ac- cording to the usual number of "sames" measure of infor- mation and according to information-theory information as outlined above. As may be noted in Table 3, two different rankings of the sentences are obtained. The word PUJOY evidently provides too much information-theory information by itself. A chance distri- bution would give .5 bits of information to the statement, "Draw a PUJOY." The statement "A PUJOY is green" then provides negative information, presumably because of the rarity of green animals. The ordering of sentenCes is again important. The statement "A PUJOY is alive," is by definition a 1- bit sentence, but the sentences preceding it, the sound of PUJOY or the Ss preference for drawing animals reduced the (information-theory) information. The results provided by this analysis present something different than the other analysis and it is important to keep in mind the difference between forming 41 TABLE 3.--Information and Frequency Analysis of the PUJOY Test. Number of Bits of Sentence New Pictures Information 1. Draw a PUJOY. 20 1.45 2. A PUJOY is green. Draw a 3. A PUJOY is alive. Draw a PUJOY. 6 .15 4. A PUJOY flies. Draw a PUJOY. 18 .35 5. A PUJOY eats worms. Draw a PUJOY. 10 .30 Total 2.00 E concept and forming the particular concept the experi- menter has in mind. The st reaction to "A PUJOY is green" is reasonable and in a probabilistic sense it would often have yielded progress toward the correct concept. This distinction also appears with considerations of information and value. An information theory analysis could conceivably be used to differentiate overlapping from independent infor- mation. Thus, saying that a PUJOY breathes includes the information that a PUJOY is alive. Saying that a PUJOY is alive, however, does not tell whether or not a PUJOY breathes. It might be possible to define "information sets" which overlap, include or exclude other information sets. 42 runs on \ctricity The requirement of artificial worlds with multivariate dichotomies is extremely difficult to meet however, and it may be possible to study the same phenomena with the measure of information proposed here. This consideration will be reserved for a section on future prospects. In conclusion, there is no shortage of topics to investigate with the PUJOY method and early returns demon- strate a number of interesting variables, including sentence information, order of sentences, and nonsense syllables used. Considerations of strategies and information-theory information are also possible but somewhat speculative. POSITIVE AND NEGATIVE INFORMATION-- PART II In the reception paradigm negative instances of a concept are generally less valuable than positive instances. Hovland (1952) noted that negative instances generally provide less information but that even with controls for information, negative instances were less valuable. The present study sought to demonstrate that negative statements had less information in the context paradigm and to determine whether or not it would be possible to control for information with positive and negative statements. It was also hypothesized that frequent words would provide less information than infrequent ones. Thus sentences using words like, "A TOSOY is big," seem to convey less about a TOSOY than do sentences like, "A TOSOY is brocade." The design was further complicated by concerns for reliability and for the effects of the nonsense sylla- bles themselves. The method for measuring information was the previously described PUJOY method. Several changes were introduced however. In order to eliminate effects of order 43 44 and interaction, the number of sentences was reduced to one. For ease in obtaining responses and in order to eliminate effects of preferences for certain types of drawings, verbal responses were sought. Method Subjects The Ss were 184 students from introductory psy- chology classes at Michigan State University. With the elimination of a few incomplete tests and random elimi- nation to equate groups there were 20 Ss in each of eight groups. Materials The verbal materials consisted of 42 nonsense words of the type CVCVC and two lists of 42 adjectives each. The nonsense words were created by randomly drawing each vowel and consonant from a pool that contained the vowels and consonants according to their frequencies in in English (Newman & Gerstman, 1952). Adjectives were defined according to the classification of Fries(l952). Any word was accepted if it would fit into the sentence frames, "A TOSOY is " and "There was a TOSOY." Only the first sentence frame was actu- ally used in the study. The adjectives were drawn randomly from the Thorndike and Lorge (1944) list such that there were six adjectives from each of seven frequency groups. 45 Frequency groups were I (1—10 occurrences per million), II (ll-20), III (21-30), IV (31-40), v (41-49), VI (A=50-100), and VII (AA=100+). Each adjective was paired with another adjective according to frequency. Nonsense words were randomly paired, randomly matched with an adjective pair and randomly ordered. Booklets Subjects were initially instructed to make a guess for each of the nonsense words and examples were provided. Guesses were restricted to nouns but could be either concrete or abstract nouns. l. PAKEB l. checker board 2. SOFON 2. rug The following pages contained a list of the forty-two nonsense words and space for guesses. When this section was completed, Ss read additional instructions which told them to compare each guess with the sentence provided and to make a new guess if their first guess was not their best guess. Examples were again provided. 1. (A-AN) PAKEB is flat 1. same 2. (A-AN) SOFON is religious 2. minister The following pages listed the nonsense words in the same order but each was used in the sentence frame. (A sample booklet is included in Appendix A.) 46 Conditions There were eight groups of Ss. Groups §_through S saw the first set of adjectives; S and S saw the second set. For each group half of the adjectives were in a positive sentence, half were in a negative sentence. S is the complement of S, S_of S, S of S, and S,of S. For example, consider the pair of adjectives "old" and ”busy," and the pair of nonsense words YEMAT and HEMIB. 5 21. (A-An) YETAH is old. 36. (A-An) HEMIB is not busy. B 21. (A-An) YETAH is not old. 36. (A-An) HEMIB is busy. S and S are the same as §_and S but the word is paired with the opposite nonsense syllable. S 21. (A-An) YETAH is not busy. 36. (A-An) HEMIB is old. D 21. (A-An) YETAH is busy. 36. (A-An) HEMIB is not old. and I!!! I'll are exact replicates of S_and S. Since the nonsense words were randomly ordered and the sentences were arranged to match the nonsense words, the change from SfS to 9:2 creates a difference in the adjective matched with a given nonsense word and a differ- ence of the adjective position in the total sequence. 47 S and S are like A and S but new adjectives are used and the nonsense syllables are re-paired and re- ordered. Results Correlations presented represent scores obtained by counting the number of "sames" for each adjective or its complement--the number of new guesses. The possible number of "sames" was 20: the actual range was 0-16. Within this range the distribution of scores appears to be ap- proximately normal. (Correlations are thus for adjectives not for Se as the number of Se making a new guess provides the measure.) Correlations of Q_with S and Q_with S provide a reliability estimate for the tests. The correlations obtained were .67 and .72. These figures seem reasonably high. Correlations of S with S and S with S_provide an estimate of reliability independent of the nonsense syllables and their order in the list. The correlations obtained were .54 and .61. This may not represent a sig- nificant decrease from the reliability coefficients but there probably is some difference due to the nonsense words. The difference between the positive and negative sentences, considered for 97S and 27S combined is presented in Figure 7. A t-test for the difference between a 20T 184 TV l6“ l4 4)- \. 12 ‘- 10+ I Average Number of New Concepts 0 L._.._.._.-.+_ ....- .._.-_....+..._. "1-... . ...._- I II 0-10 11-20 Frequency of Adjectives Positive Mean Standard Deviation 13.4 12.4 4.5 5.5 Negative Mean Standard Deviation 9.3 8.8 4.0 4.4 21-30 48 ./ \ .k .-.....-.... -. III 31-40 14.2 15.9 6.6 3.6 10.1 12.3} 7.0 6.9 \x. /\r;>< »—~~ 4— --~ I V VI VII 41-50 50-100 100+ (Occurrences Per Million) 12.1 12.0 10.7 5.6 4.9 2.6 9.9 9.2 11.0 5.9 4.3 3.1 Figure 7.--Information in positive and negative sentences with adjectives of various frequencies in English (Group I). 49 sentence in its positive and negative sense (therefore a matched pair) is significant at the .001 level. The correspondence of the two curves is apparently due to the positive correlation between the information in the positive and negative sense of the words. For SfS and 272 combined this correlation is .54. Correlations for the groups considered separately were somewhat lower (57S = .39; ng = .50 and £72.: .21). Although a positive correlation seemed reasonable the shape of the curves in Figure 7 is quite irregular. In order to clarify this, groups ETE considered an entirely new set of adjectives. With SfS_there was a negative instead of a positive correlation (r = -.28) and again the overall curves reflected this (Figure 8). A t-test for the difference between positive and negative sentence was again significant at the .001 level. Thus it is not the general case that positive and negative information are positively correlated but it is the case that positive sentences generally have more information than do negative sentences. An analysis of variance would probably show sig- nificant effects due to frequency for both samples but if the effect is really due to individual words that analysis would be meaningless. It is also possible to consider both samples as indicative of random variations from a real trend. Figure 9 show the variation of data points from converging 50 20? 18« 14. . /\./\\\/ \ 10 ii» /.\. \ . \‘y\. 8 .\ . 12 ‘T Average Number of New Concepts L J J A l A l v o ur__r"_1“-n 1 _r_ , ,__1 I II III IV V VI VII 0-10 11-20 21-30 31-40 41-50 51-100 100+ Frequency of Adjectives (Occurrences Per Million) Positive Mean 11.7 13.8 12.5 13.3 12.5 12.3 12.5 Standard Deviation 3.4 2.6 2.6 3.4 3.7 2.8 2.5 Negative Mean 9.0 8.2 10.7 10.1 9.0 12.1 10.0 Standard Deviation 1.0 2.9 1.4 1.0 3.2 2.9 3.5 Figure 8.--Information in positive and negative sentences with adjectives of various frequencies in English (Group II). Average Number of New Concepts 20... 18.. 16-: MT 12.. 10” 51 -—————___________m. Positive (0) é. x ‘X %~ x. X1 Negative (x) X A I II III IV V VI VII 0-10 11-20 21-30 31-40 41-50 50-100 100+ Frequency of Adjectives (Occurrences Per Million) Figure 9.--A possible relationship of information to frequency: Data points from adjectives of Group I and Group II are displayed in relation to least squares lines for positive and negative information. 52 straight lines. Such a trend might actually exist though it is purely speculative at this point and the variation due to individual words seems to have a much more powerful effect than any underlying regularities. Part of the problem with frequency is its inexact correlation with probabilities of things in the world. The probability is low that if someone says, "I am thinking of something" that he will be thinking of something ebony but practically anything could be constructed of ebony, at least in minature. (This does not seem to be the case with corduroy.) On the other hand the world "disproportionate," is rare but the number of things in the world that are disproportionate is great. To consider these words individually a scatter diagram is presented in Figure 10. Such data would make it possible to select words in order to control for infor- mation. Note that either "ripe" or "not ripe" yield many new concepts while either "productive" or "not productive" yield few. Discussion It is only fair to admit that the method has problems. Higher relibilities would certainly be desirable, the influence on the nonsense syllables is cumbersome though it may be more than a procedural error. Much more serious is the problem of assuming that people form concepts having heard a single word totally out of context Number of Concepts Changed Following Positive Sentence 53 40 ” .Ripe .lulOiSt 35 ‘ Corduroy Stormy T' ht Respogsible Pearl ' ' lg .Curious .Canvas .S.oer J . . D. 0 016 ro ortionate ‘ 3 p p. . .Forgetful .Rap'd ChBral Humble .Supreme .Gay Empty ' .Intellectua 25 . Colonial . .Spiritual .Silly .Modegflsy B Ifiarr°w Motionless . o Cheer ’ . Cheap Hee ass 20 + . Ridiculous' .Ebony Perfect Glorious .Beautiful Ideal. .: doss .Natural Secure 'roductive 15 ” 10 d» 0 t if I T I = i 10 15 20 25 30 35 40 Number of Concepts Changed Following Negative Sentence Figure 10.--Scatter diagram of positive and negative infor- mation for adjectives of Group I. 54 since that is impossible. But the method is interested in the change of concepts with the presentation of information and a beginning must be made somewhere. Writing "same or making a new guess is a very crude measure of an individual's concept formation. The drawings are better for delineating subtle changes in a concept. A S may consider his first guess to be extremely unlikely but still possible after reading the sentence. Although there is nothing wrong with the analysis in a probabilistic sense it might be better to place more emphasis on obtaining the correct concept. On the other hand, the method is reasonably re- liable and seems to be capable of demonstrating some notions that are logically sound (positive and negative instances) while demonstrating extreme variability with a notion that seemed relatively simple (frequency). INFORMATION AND VALUE-- PART III As mentioned previously, the relation of infor— mation to value is unknown. Information is here defined by the PUJOY method as the number of pgw concepts across Ss. Value may be measured by a similar task when the cor- rect concept has been defined so that responses may be scored as either correct or incorrect. The interaction of relevance to value and infor- mation is also unknown but it is suspected that irrelevant sentences could be of high information and low value. With relevant sentences the relationship should be positive with high information sentences being of high value and low information sentences being of low value. There is no reason to suspect that the relationship would be perfect, even with sophisticated measurement, since information reflects the reader's knowledge of what a PUJOY is not and value reflects his knowledge of what a PUJOY is. At this stage of precision, even the demonstration of a consistent relationship is difficult. Relevance could be defined by author variables rather than by reader variables so the interaction of 55 56 relevance with information and value could be assessed. Here, however, all sentences are supposedly relevant. The study took place in three parts, essentially a communication chain. The first group of Se encountered instances of the correct concept and wrote the relevant sentences. The second group provided the measure of infor— mation. Sentences were then grouped according to infor- mation level, (high, middle, and low) and presented to a third group so that value could be measured. Both recall and recognition were tested and ten sentences were presented for each information level so the cumulative effects of information could be assessed. In order to make the task as representative of this type of concept formation as possible, six different concepts were used. Two of the concepts had actual physi- cal referents as well as common English labels (a slide projector and a head of cabbage). The second pair had actual physical referents but no common English labels (the original SALOP--a cardboard statue, now called LUREK; and a piece of perception lab apparatus made in the Michigan State University shop and used for fragmentation, called TEMOS, Figure 11). The third pair had no physical referents but had common English labels (freedom, gravity). Part A--Obtaining the Sentences Each of the four physical referents with matching labels (PROJECTOR, CABBAGE, LUREK, TEMOS) and the two 57 Isms (alias HETAB) LUREK (alias REPOL) (Also known as a SALOP) Figure ll.-_Objects used to represent TEMOS and LUREK. 58 labels without referents (FREEDOM, GRAVITY) were presented one at a time to a group of 41 students at Michigan State University. The Ss were given a pad of paper and these written instructions. In this experiment you will be asked to write sentences which include a particular word. It may be a real word but in some cases it will be a nonsense word. You may also be shown something that the word represents. For instance, if you were shown a design and told that it was called a ROSUB you would then write sentences using the word ROSUB and referring to the design. If the word or objects is something that you al- ready know about you may use your knowledge of that thing or idea. The point is to write sentences that are relevant and to write as many different sentences as possible within the five minute time limit. Be sure that you actually use the word in each sentence and that each sentence is in some way related to the idea or object presented. Thus there was no indication that these sentences would be used to communicate anything to anyone. It was felt that a wide variety of sentences would be necessary to obtain the necessary range of information and a wide variety of sentences were obtained. A total of 2,106 sentences were written, an average of 8.6 sentences per subject per five minutes. The most sentences were written about PROJECTOR (10.1) the fewest about GRAVITY (7.3). (PROJECTOR was the first concept presented but GRAVITY was the second so this may indicate something besides writer's cramp.) Sentence length ranged from two words (CABBAGE spoils) to 96 words (a very long sentence about GRAVITY). The second longest sentence had 41 words. 59 A small prOportion of the sentences referred to the experiment itself rather than to the word or object presented. "I wonder what the girl next to me is writing about CABBAGE," or "I hope that I have the FREEDOM to smoke, as I am now going to light a cigarette." Sometimes sentences referred to an object presented previously as well as the one that was currently presented. "FREEDOM and CABBAGE are a lot alike, both are ahead of their time," or “I wonder if LUREK thinks my friends are CABBAGES." These sentences were excluded from inclusion in the sample but practically everything was included. Some sentences were specific to Michigan State University, "They serve lousey CABBAGE in Brody," or "You can also learn about gravity in HPR 105." Some were thoughtful, "FREEDOM, said Sartre, is anguish," or "SDS will attack CABBAGE as being capitalistic, imperialistic and half way normal.” Others were imaginative, "Five days in each week are devoted to the betterment of LUREK's figure, but on the other two her eating habits are so outrageously piggish that she is constantly gaining weight," or incorrect, "The HETAB can shoot laser beams that can shoot down any enemy plane traveling under 100,000 mph." But some were descriptive, "LUREK looks like a tree with a fish on top of it" or "HETAB looks like a new sinister weapon straight from the pure, pristine ivory tower walls of untainted scientific research." 60 Somehow, these sentences are not like the ones that psychologists write to put in experiments, nor are they like textbook sentences. Perhaps many of them are not really relevant but since these differences cannot be systematically defined it seemed unwise to discriminate between them. It was, however, necessary to preserve some degree of independence between the sentences used in the final part so the sentences that seemed to say essentially the same thing were included in content categories ("PRO- JECTORS are expensive," "PROJECTORS cost a lot" or "The American flag is a symbol of FREEDOM," "Red, white and blue represents FREEDOM to many people.") Sentences were then drawn randomly with the restriction that one and only one draw come from each content category. The remaining draws came from sentences that had not overlapped with other sentences, the idiosyncratic group. A total of 360 sentences, 60 for each concept were thus provided for inclusion in Part B. Part B--Measuring Information Three hundred sixty nonsense words were devised by randomly choosing consonants and vowels. Each nonsense word was randomly paired with one of the sentences chosen for inclusion in Part B. The instructions were similar to those used for Part II of this study, Ss first guessed a meaning for each nonsense word and then read the sentences and modified the guess or wrote "same" (Appendix B). 61 In order to measure information for all 360 sentences without requiring that any S read them all, the sentences were sorted into 12 sets. Each set had 30 sentences, five for each of the six concepts. Booklets were then prepared for each set of sentences and given to 30 Ss. This required a total of 360 Ss. They were tested in small groups. It had previously been decided to eliminate any sentence if more than four Ss actually guessed the correct concept. Since concern in Part C would be with the cumu- lative effects of information, sentences which revealed the correct concept by themselves would not be wanted. A surprising number of sentences did this. They are included in Appendix C. It is possible that some incidental learning occurred since the Ss did read five sentences about the same concept, though there was no indication that any of the sentences would be about the same thing. Ideally more concepts would be used and fewer sentences about each one would be required so that each booklet would contain only one sentence about each concept. Such were not the demands of this study. At any rate these sentences were eliminated from further inclusion so the result was simply that more sentences were lost, which may have restricted the infor- mation range. The number of "sames" was tabulated for each remaining sentence and frequency counts of these numbers 62 were made considering each concept separately. The ten sentences with the fewest number of "sames" were then selected for the High information group, the ten with the fewest number for the Low information group and the ten falling closest to the median for the Middle information group. In cases where several sentences having the same rank exceeded the number needed, sentences were randomly eliminated. Thus from the 360 sentences used in Part B, 180 sentences were chosen for inclusion in Part C. The procedure for selecting sentences is summarized in Figure 12. The sentences included in Part C and their information scores, in number of new guesses are included in Appendix D. The possible range for number of new guesses would be from O to 30, the actual range observed was from 5 to 30 (5-29 for PROJECTOR, 10-29 for CABBAGE, 13-29 for FREEDOM, 15-29 for GRAVITY, 11-28 for LUREK, and 11-29 for TEMOS). There were at least 15 Ss difference in all cases (50 per cent). It is important to note that selection was not based on correct concepts. Some sentences that were correctly identified by one or two Se in Part B actually ended up in the Low information group because the majority of Se did not guess the correct concept and did not change their original concepts. One important finding may be concluded from these results without reference to the results on value. 63 PART A SENTENCES GENERATED FOR EACH OF SIX CONCEPTS 360 sentences randomly selected for Part B (60 for each concept) PART B TEST OF INFORMATION (360 sentences divided into 12 sets of 30 for inclusion in booklets, 30 Se tested on each type of booklet) number of correct responses / tabUIated \ number of "sames" sentences with more tabulated for than four correct responses remaining sentences eliminated from study (see Appendix C) 10 High, 10 Middle and 10 Low ranking sentences chosen for each concept \ PART C TEST OF VALUE (30 sentences for each concept, 120 total, included in booklets--see Design Table 4) Figure 12.--Summary of procedure for selecting sentences. 64 Information was positively correlated with sentence length, r = .30. The correlations for each concept were, CABBAGE .10, TEMOS .23, PROJECTOR .31, GRAVITY .36, REPOL .37 and FREEDOM .46. This finding agrees with the findings of information-theory information presented earlier (Aborn, Rubenstein & Sterling, 1959; Shepard, 1960). Part C--Measuring Value Ten high, ten middle and ten low information sentences for each concept were thus obtained for the test of value. Each concept was randomly paired with a nonsense word, including the two that already had nonsense labels. (SEYAH, projector; RULAT, gravity: REPOL, LUREK: DOFITH, cabbage, KEHET, freedom and HETAB, TEMOS.) The recall test for objects without English labels was the usual, "Draw a (HETAB or REPOL)." For the concepts with English labels the recall test was "The English word for is ." (With SEYAH, RULAT, KEHET or DOFIH provided in the first blank.) The recognition test for concepts without English labels used real objects similar in some ways to the objects originally described. Thus, the "Art" objects used to match LUREK (REPOL) were a collage in the shape of a shield, a stitchery, a high-contrast negative photograph, a statue of Degas "dancer," a statue of the Macabees, a stuffed hippopotamus, a cardboard dog made out of ice cream cartons, an op flower pot made out of a TV antenna and a glass owl. 65 The "Lab" objects used to match TEMOS (HETAB) were part of a T-scope, a pair of distorted lenses, a small T-scope, a metronome, a color wheel, a distance perception box, a light box, a retinal perimeter stand and a complication clock. The recognition test for concepts with English labels used English words (Appendix D). Each of the 240 Ss read 10 sentences for each of the six concepts and took recognition tests on three of the six and recall tests on the other three. The concepts were originally paired according to labels and referents so one was used for recall when the other was used for recog- nition. The design also counterbalanced information level (Table 4). Within each of these letters (A, A', etc.) there were ten sentences. These were counterbalanced so that each sentence occupied each position first through tenth. With 40 Se in a group there were thus four Se in a cell (having exactly the same test material.) There were ten subgroups for each of the six main groups. The order of presentation of concepts within booklets was randomized. In a pilot study, booklets alternated information and test, the sentence on one page and the alternatives or the question on the next. These booklets had 134 pages which had to be assembled in exactly the right order, never the same order for any two booklets. Besides the time required to make these booklets, the Ss were confused 66 TABLE 4.--Design to Counterbalance Information and Concept. Recognition Test Recall Test Group H M L H M L I A B C A' B' C' II B C A C' A' 8' III C A B B' C' A' IV A' B' C' A B C V C' A' B' B C A VI 8' C' A' C A B A and A' = concepts with labels and referents (A - Projector: A' - Cabbage) B and B' = concepts with labels (B - Gravity; B' - Freedom) C and C' = concepts with referents (C - REPOL; C' - HETAB) H = High information M = Middle information L = Low information by this plan and tended to respond on the page with the sentence, sometimes inappropriately, like drawing KEHET. This was especially difficult to prevent since some of the "sentences" were in fact questions (”Did Andy Warhol invent the REPOL?"). The task was already complex and the hope of making a booklet that was easier to assemble convinced the S to start over. Booklets were then made in two parts, one part for information, containing the 11 sentences and one part for test, containing the 11 matching response pages. Ss then turned one page in each booklet and it was easier to confine responses to the real questions. Instructions are presented in Appendix E. Still, this was an exceedingly 67 cumbersome task of compilation, involving over 30,000 little pages. The Ss were again tested in small groups, which did not coincide with the group numbers except that Groups I to III were distinct from IV to V1. This was due to the necessity of using real objects for the recognition tests for the concepts without labels. Art objects were used for LUREK (REPOL) and lab objects were used for TEMOS (HETAB). Each S took a recognition test for either REPOL or HETAB so one set of objects was required whenever a test was given. The objects were displayed on desks at the front of the room, always available for inSpection and numbered from 1 to 10 to facilitate responding. Results The number of correct responses to the recognition test after each sentence was tabulated and totals were computed. The results show the general trends for the three information levels (Figure 13). For the overall curves the hypothesis is thus confirmed, more information leads to more correct guesses. i The number of correct guesses on the recall test for the four concepts with labels may be tabulated in a similar fashion. Wider differences are apparent (Figure 14). With LUREK and TEMOS the correct guess on the recall task is not so easy to assess, but the final pictures were judged as either correct or incorrect by six Average Number Correct 68 40<» , High 4b -— /. 35 o/' ' ' Middle / ’.- -." "" ‘- 30.. ,,»~0 ' ,O’ Low I l 25.. '///I ,0’ ’. ' . . o-—o’ I 0 O O / - 0.. 0.. T T r I fi 1 —I T T 2 3 4 5 6 7 8 9 10 Number of Sentences Figure 13.--Effects of information and number of sentences on recognition. Average Number Correct 69 40., 35¢ High 30+ , ,." o/ ./. 25., ,/ Middle o/ O- -o 201% .,// I , / ’o— -O' ’. o ., 15‘. / / ’ LOW ,0 O .’ 10 ii. / / / e I. . 0 5 4)- . / . . . ..o . o . 0 EL¢99L__, +» 4 e +7 e ; ; g l 2 3 4 5 6 7 8 9 10 Number of Sentences Figure 14.--Effect of information and number of sentences on recall. 70 judges who saw the original objects. In general the pic- tures were quite different from the objects, most were entirely unrelated and no judges saw them as correct, about 10 per cent were close and the judges opinions varied. Computing reliabilities by counting the number of pictures in each subgroup that each judge considered correct gives a total reliability of .89. This was obtained by a coef- ficient of concordance ranking REPOL and HETAB together. If ranks are considered separately REPOL has a reliability of .46 and HETAB has a reliability of .96. The judges seemed to take a single criterion as indicative of a correct HETAB; pictures that showed some kind of viewer were generally considered correct. In truth there was not much agreement about individual pictures but the reli- abilities given represent the subgroup data. Thus one judge might have considered one member of a subgroup cor- rect and another judge might have chosen another member of that subgroup, but they generally agreed about how many to pick from each subgroup. The problem of judging pictures is serious. Though higher reliabilities might be obtained by instruction of judges these reliabilities probably represent the real circumstances, these pictures are simply very different from the objects. If you look at a peach and a pear and an orange it is very difficult to decide which one is the apple. If three or more judges scored a picture as correct, it was considered correct for 71 purposes of completing the REPOL and HETAB categories in the design. An analysis of variance for a 3 x 3 latin square with repeated measures (Weiner, 1962) was then computed for groups I, II and III and for groups IV, V, and VI for recognition and for recall. It is necessary to keep these analyses separate because the two groups read about differ- ent concepts for recognition and recall. Thus for recog- nition tests groups I, II and III read sentences about gravity, projector and REPOL; groups IV, V, and VI read sentences about freedom, cabbage and HETAB. For the recall test groups I, II and III read sentences about freedom, cabbage and HETAB; groups IV, V and VI read sentences about gravity, projector and REPOL. It would be possible to compare recognition to recall within the same analysis but the number of correct answers is distinctly different and the tasks are quite different. These analyses are pre- sented in Tables 5 and 6. Effects of information (Factor A), concept (Factor B) and group are significant in all cases and the interaction of concept and information is significant in two cases. This interaction is represented as (AB)' instead of AB because this latin square design would compound the AB interaction with the effect due to groups. The (AB)' interaction is the within—subject component (Winer, 1962, p. 539). The effect of concepts is due to numerous factors, the concept, the sentences and TABLE 5.--Analysis of Variance for Recognition Test, Using Number Correct After the Last Sentence. SS DF MS F W Between 83 Groups 5.4 2 2.7 18.0** gs within groups 17.9 117 .15 Within SS A (information) 5.8 2 2.9 24.2** B (concepts) 1.7 2 .85 7.1** (AB)' 8.9 2 4.4 36.7** Error within 27.3 234 .12 Groups IV, V, and VI Between 85 Groups 2.0 2 1.0 5.9** §s within groups 20.2 117 .17 Within 83 A (information) 1.6 2 .8 5.4** B (concepts) 3.0 2 1.5 10.0** (AB)' .3 2 .15 1.0 Error within 35.1 234 .15 **Significant at the .01 level. Note: On Recognition, Groups I, II and III were tested on Gravity, Projector and REPOL: Groups IV, V and VI were tested on Freedom, Cabbage and HETAB. 73 TABLE 6.--Analysis of Variance for Recall Test, Using Number Correct After the Last Sentence. SS DF MS F Groups I, II and III Between Ss Groups 2.9 2 1.4 7.4** gs within groups 22.4 117 .19 Within 83 A (information) 3.3 2 1.6 9.4** B (concepts) 9.7 2 4.8 28.2** (AB)' .2 2 .10 .6 Error within 40.8 234 .17 Groups IV, V and VI Between 85 Groups 1.1 2 .55 5.0** §s within groups 12.3 117 .11 Within 85 A (information) 26.6 2 13.3 120.9** B (concepts) 11.4 2 5.7 51.8** (AB)' 3.2 2 1.6 l4.5** Error within 26.1 234 .ll **Signifcant at .01 level. Note: VI were tested on Gravity, Projector and REPOL. On Recall, Groups I, and III were tested on Freedom, Cabbage and HETAB; Groups IV, V, and 74 the test given for recognition or recall and is not particularly revealing. The group effect is partially due to this and is similarly uninteresting. The effect of information however, is the hypothesis under test and has been, in general, demonstrated. The two interactions indicate that some concepts may not be contributing to this effect. Considering the concepts separately, the three information levels are not always in the hypothesized direction (low information yielding low value). Complete results are presented in Tables 7, 8 and 9. With recog- nition Projector, Freedom and Gravity lie in the hypothe— sized direction, Cabbage shows no real differences, and REPOL and HETAB are in the wrong direction. The results are similar for the Recall test but Freedom is in the wrong direction and TEMOS is in the right direction. The results of an analysis of variance on the final number correct for each concept (using subgroups l to 10 in order to have interval data) is presented in Table 10. It was originally thought that the number of "sames" could also be analyzed in this study and trends similar to those found with the SALOP but differentiated by information level discovered. Thus High information sentences would have fewer "sames" than Low information 75 TABLE 7.--Number Correct on Recognition Task for Each Concept at Each Information Level. Sentences 0 1 2 3 4 5 6 7 8 9 10 Information Low FREEDOM 3 10 13 16 23 26 25 24 29 29 28 GRAVITY 3 10 15 20 24 24 23 28 30 28 27 PROJECTOR 3 3 3 2 2 3 3 2 4 4 5 CABBAGE l 12 21 23 29 30 33 35 36 36 37 TEMOS 6 12 14 17 17 18 16 16 21 22 20 LUREK 4 7 15 22 22 28 30 30 33 35 37 Total 20 54 81 100 117 129 130 135 153 154 154 Average 3 9 14 17 20 22 22 22 26 26 26 Middle FREEDOM 3 11 15 17 18 23 27 27 26 28 29 GRAVITY 0 16 23 24 28 30 32 35 34 35 33 PROJECTOR 7 20 21 24 27 28 28 31 35 35 35 CABBAGE 2 16 22 21 27 27 30 33 30 30 31 TEMOS 12 19 24 28 27 32 33 32 32 31 31 LUREK 4 16 22 22 23 25 31 33 34 30 30 Total 28 98 137 136 150 165 181 191 191 189 189 Average 5 16 23 23 25 26 30 32 32 32 32 High FREEDOM 3 19 28 30 34 33 37 39 38 38 38 GRAVITY 7 27 30 36 33 36 38 39 37 36 38 PROJECTOR 7 24 30 34 39 38 36 37 39 40 39 CABBAGE 6 13 25 26 29 30 29 34 36 36 38 TEMOS 6 16 21 29 27 26 28 27 28 30 28 LUREK 4 11 18 18 19 18 21 25 27 24 27 Total 33 110 152 173 181 181 189 201 205 204 208 Average 6 18 25 29 30 30 32 34 34 34 35 76 TABLE 3.-—Number Correct on Recall Task for Each Concept at Each Information Level. Sentences 0 l 2 3 4 5 6 7 8 9 10 Information Low FREEDOM 0 2 6 7 10 13 14 -15 15 16 20 GRAVITY 0 0 0 0 2 2 1 2 2 3 2 PROJECTOR 0 0 0 0 0 0 0 0 0 0 0 CABBAGE 0 0 0 0 2 3 4 5 6 10 13 Total 0 2 6 7 14 18 19 22 23 29 35 Average 0 0 2 2 4 4 5 5 6 7 9 Middle FREEDOM 0 3 5 8 8 ll 13 13 14 13 15 GRAVITY 0 1 5 7 10 ll 12 11 11 13 13 PROJECTOR 0 3 5 8 12 15 18 21 23 25 27 CABBAGE 0 3 14 20 17 23 24 23 22 27 23 Total 0 10 29 43 47 60 67 68 70 78 78 Average 0 2 7 11 12 15 17 17 18 20 20 High FREEDOM 0 7 19 22 23 26 30 31 31 31 32 GRAVITY 0 4 14 20 18 23 27 33 36 33 36 PROJECTOR l 3 ll 16 24 27 29 29 31 33 33 CABBAGE 0 l 1 5 6 6 6 9 9 9 10 Total 1 15 45 63 71 82 92 102 107 106 111 Average 0 4 11 16 18 20 23 26 27 27 28 77 TABLE 9.--Number Judged Correct on Recall for REPOL and HETAB, After All 10 Sentences. Information Low Middle High HETAB 2 8 l3 REPOL l 7 3 TABLE lO.--Ana1ysis of Variance for Individual Concepts. Recognition Recall Right 819- Right Sig— Order nificant* Order nificant* PROJECTOR yes yes yes yes CABBAGE no no no no FREEDOM yes yes no yes GRAVITY yes yes yes yes REPOL no yes no no TEMOS no yes yes yes *Significance is at the .05 level. 78 sentences and this effect would interact with order of sentences something like this. 401 Low Information u m_ 30* g 3 Middle Information 25 20— men 3‘ .1 avg 10 /////”’flfl,.—v—“"'—F High Information > 42 T l I I I l l T T r O l 2 3 4 5 6 7 8 9 10 The number of correct responses to items with labels was not predicted, however. This compounds the "same" analysis because "same" may be written after either a correct con- cept or an incorrect one. If High information sentences reveal the correct concept they will then have more "sames" than Low information sentences. It is, however, possible to do this analysis with the recall task for REPOL and TEMOS as the correct answer is never really obtained. The re- sults are presented in Figure 15. This is close to the hypothesized curve but more variable and the Low infor- mation curve falls below the Middle information curve by one point for the final sentence. Unfortunately it is not possible to do an analysis of variance that considers the order of presentation since the results from one sentence are not independent of those from the next. A repeated-measures design might be de- vised but that would require random ordering rather than Average Number of "Sames" 79 401’ 35" 30(L ,! Low 25? ..-- ~~~~ ""4' ..---' ...-»°-'—-—-"’ ' 20 (L .." w, I“ Middle . .’,o-" /.——-—o-——-0 15 Fr 75/7 /°\ ./. High f;/ o ' 10 ‘ .°o % f/ /// f/ o 5‘.f/ ./// V 0 J. % ‘L_ «1.1.- _....+..-.,.....-+.._____+_. ...... +_..._ 1 ..___4 Number of Sentences Figure 15.--Number of "sames" reported on recall test in relation to information and number of sentences for Repol and Temos data combined. 80 counterbalancing. The interaction of sentences and order really needs to be considered at a level which does not exceed current statistical theory especially since this is a very unsophisticated technique at this point. As previously stated, the E is now convinced that this can best be done with two sentences. Discussion To some extent this study has demonstrated the relationship of information and value with relevant sentences. The variability of relevance has, however, been defined very vaguely and has not been experimentally manipulated. Certainly it would interact with the value information relationship. Irrelevant sentences could be obtained simply by borrowing ones written for other concepts but the systematic manipulation of instructions to obtain varying degrees of relevance would be more interesting. This could probably be done by telling one group of gs that the sentences were intended for communi- cation. Meanwhile it may be assumed that the sentences in this study varied somewhat in relevance and that control of this factor might reduce variance between sentences. The possible operational distinction of relevance, value and information is probably the most important result of this experiment. Though these terms are included as definitions and though a more clever E might have made this distinction prior to doing the research, they were, here, the result not the cause of the study. Indeed the PUJOY 81 method has followed this same procedure for most of its development. But relevance does not explain the differences between the concepts. The sentences about cabbage produced almost no differences in value though differences in infor- mation were reasonably high. If the lack of difference were only on the recognition test the test itself might be suspect as all food words are used and thus some potential information was eliminated. (Note that other recognition tests also selected items reasonably close to the correct alternative.) But the sentences for cabbage do not dif- ferentiate on the recall test either. Recall scores were quite low; many gs, for example, thought that DOFIH represented "lettuce." Recognition scores on the other hand were quite high. Four sentences, one at each infor- mation level produced four correct responses out of four on the first trial. Sooner or later all gs encountered one of these sentences hence final results were near the ceiling for all groups. Such sentences existed for the other concepts too but never in the "low information" group. In truth it is not particularly surprising that cabbage came out differently. Concepts and sentences do vary. It could have been for any of a number of reasons, but it cannot now be determined why or for which ones. 82 The recognition data for REPOL and TEMOS, however, were significant but in the wrong direction. These nonsense words represented objects which do not have real English labels and it is quite possible that an entirely different phenomenon is operating. The difference between information (which is scored by changes, which indicate knowing what an object is not) and value (knowing what an object is) may be more crucial with concepts that are unfamiliar. With familiar concepts progressive clues about what to eliminate would eventually provide the correct selection (given a finite set of known concepts). With concepts that are undefined, however, elimination would not yield the correct concept. Perhaps this explanation is too sophisticated for current methods or results. The differ- ence might be much simpler, relating to the choice of alternative objects, or specific clues in specific sentences. The recall data do come out in the correct direction for TEMOS. Concepts with labels are similar to the "guess the missing word" game of information theory. Syntactic clues may be helpful as well as specific associations. Deese (1967) also found differences between concepts with labels and those without. Whatever the reason, currently mys- terious, it is clear that concepts with labels should be distinguished from those without in future studies. 83 Similarly, recall and recognition tests seem to indicate different things and this distinction must be preserved. It now seems clear that the extent of the relation- ship between information and value could best be examined in a study that used many concepts, many gs and only one sentence for each concept. A simple correlation between the measure of information and the corresponding measure of value for each sentence could then be obtained. PROSPECTS Determining what makes a sentence work is a diffi- cult task. The procedure always raises many more questions than it answers but perhaps that is because the problem is becoming more clearly defined. This level of analysis, at least, seems valuable. Sentences are neither too atomistic to be meaningful not so molecular that they cannot be systematically controlled. Although some general laws may have been discovered at this point it is probably more important that the differences between sentences have not been hidden. New ways to distinguish between sentences, such as relevance and value, continue to appear. The next step should be to determine the inter- action of sentence information and the relation of infor— mation to the real world. A discussion of concepts should ultimately have some relation to their referents. Using two sentences and assuming that the nonsense syllable produces a random sample of concepts, the following relations may be defined. 84 85 Synonymipy occurs when Sentence A followed by sentence B, A = x information and B = 0 information. and Sentence B followed by sentence A, B = x information and A = 0 information. (x must be some quantity of information greater than zero.) Independence Sentence A then sentence B, A = x and B = y and Sentence B then sentence A, B = y and A = x Inclusion (B included in A) Sentence A then sentence B, A = x and B = 0 and Sentence B then sentence A, B = y and A = 2 Obviously there are still measurement problems. Other research might introduce a response of "impossible" into the PUJOY method. Just as there are possible words that don't exist (like SALOP and TEMOS) there are things that could exist but don't happen to. However, new concepts expand an individual's knowledge of 86 things or relationship that exist and notions which seem impossible to an individual may be rejected in spite of evidence to the contrary. Further evidence may then make these concepts more feasible. If someone describes a rare African beast that looks like a lion but has the horns of a moose, an indi— vidual might question the source but ultimately add this lion-moose to his personal classification of animals. On the other hand, if someone describes a moose that can climb trees like a lion, the individual might consider it logically impossible. Broadly considered, anything at all could exist but some possible concepts are more likely than others. This is individually defined and it is important to the notion of an education ”expanding the world” because it defines the limits of that world. For example, con- sider this concept and these responses: A KADAP is portable. ok A KADAP is useful. ok A KADAP is aluminum. ok A KADAP is transparent. no--doesn't exist-- but seems possible. A KADAP is edible. no--impossible. Within this cognitive world, transparent aluminum is possible but edible aluminum is not. It might however, be possible to convince someone that aluminum was now being manufactured according to a special process that makes it edible. 87 Finally, some consideration should be given to the great subtitles of language that are way beyond any current means of research. For example Wiener and Mehrabian (1968, p. 2) treat the differences between such sentences as "These people need help" and "Those people need help," or "When you and I spoke of " vs "When we spoke of In fact, careful consideration will show that what appears to be the same thing (i.e., the same content) said with different words can be a basis for inferring quite different feelings or attitudes on the part of the speaker. Language is, after all, complex; therein lies the beauty and the intrigue. SUMMARY The context paradigm is a study of reading that focuses on the sentence level of analysis, particularly on the way that sentences reveal the meaning of unknown embedded elements. Essentially this method required that a subject first guess what a nonsense word represents and then modify that guess on the basis of the information presented. The study most similar to this is Werner and Kaplan's (1950) word-context test. Using similar sentences, they discovered a developmental trend in children's abilities to abstract the meaning of an embedded element. However, they did not attempt to differentiate between sentences. Here the emphasis has been on measuring the differ- ences between sentences so that sentences that tend to reveal the meaning of unknown elements can be separated from those that do not. The method utilized is called the PUJOY technique. Subjects first read a nonsense word and make a guess in the form of a word or a drawing. They then read a sentence containing that word and guess again. Thus the first page of a booklet may say, "Draw a PUJOY," 88 89 and the second may say, "A PUJOY is green. Draw a PUJOY." The number of subjects who change their guesses after reading a sentence is taken as a measure of the information in that sentence. When the correct concept is known the number of changes to a correct concept is taken as a measure of the gglpg_of that sentence. Several pilot studies were conducted. The first one required that subjects draw PUJOYs and GIDAKs and then modify the drawings on the basis of further information. Subsequent studies allowed subjects to either make a new guess or write, "same." "SALOP for sale," a story written about a SALOP, demonstrated that generally more "sames" are written as the story progresses. Order of presentation thus interacts with information. In an attempt to separate information from order, "What they said about the SALOP," counterbalanced sentence order and demonstrated that information can be measured independently. It also revealed strategies of concept formation which may be similar to conservative focusing and focus gambling. A final pilot study compared the PUJOY measure of infor- mation to information-theory information. The first major study used adjectives, selected according to their frequencies in English, in the sentence frames A TOSOY is _____ and A TOSOY is not _____, By testing each adjective in both a positive and a negative 90 sentence, it was possible to determine that the positive sentences were generally more informative than the negative ones. There was some indication that the difference between positive and negative sentences was greater for infrequent adjectives than for frequent ones. The other major study investigated the relationship between information and value. Information considers all changes but value considers only changes to a correct concept. For this experiment six concepts were chosen and subjects were asked to write sentences about these concepts. Some of these sentences were than presented to a second group of subjects and from their guesses a measure of the information in each sentence was obtained. Sentences ranking high, middle, and low in information were then selected for each concept. These sentences were presented to a final group of subjects who tried to guess the original concepts. As predicted, subjects who read high information sentences made more correct guesses than subjects who read low information sentences. An analysis of variance demonstrated a significant difference due to information for both recall and recognition tasks. However, the concepts without labels seem to be operating differently than those with labels. It has thus been demonstrated that the PUJOY technique can provide reliable data and that measuring information can give an index of which sentences are 91 likely to be valuable for regenerating the concept originally described. Two future prospects are explored. It is hoped that the PUJOY technique used with two sentences will make it possible to measure synonymous and independent sentences and distinguish concepts that a subject con- siders impossible. REFERENCES REFERENCES Aborn, M., Rubenstein, H. & Sterling, T. D. Sources of contextual constraint upon words in sentences. Journal of Experimental Psychology, 57, (1959), 171-180. Bar Hillel, Y. & Carnap, R. Semantic information Communi- capion Theory. Edited by W. Jackson. Butterworth Scientific Pub. London, 1953. Bourne, E. Humgn conceptual behavior. Boston: Allyn & Bacon, Inc. 1966. Brown, R. Language and categories Appendix to A Study of Thinking. Bruner, et a1. Wiley, N.Y., 1956. Bruner, J. S., Goodnow, J. J. & Austin, G. A. A study of thinking. Wiley, N.Y., 1956. Carroll, J. B. Words, meaning and concepts. Harvard Educational Review, 34, (1964), 178-202. Chomsky, N. Aspects of the theory of syntax. MIT Press, Cambridge, 1965. Davis, G. A. A note on two basic forms of concepts and concept learning. Journal 9f Psychology, 62, (1966), 249-254. Deese, J. Meaning and change of meaning. American Psyf chologist, 22, (1967), 641-651. Dewey, J. How we think. Boston: D. C. Heath & Co., 1933. Fries, C. C. The structure of English. New York: Harcourt Brace, 1952. Hovland, C. I. A "communication analysis" of concept learning. Psychological Review, 59, (1952), 461—472. Johnson, D. M. & Stratton, R. P. Evaluation of five methods of teaching concepts. Journal of Edu- cational Psychology, 57, (1966), 48-53. 92 93 Johnson, M. G. Syntactic position and rated meaning. Journal of Verbal Learning and Verbal Behavior, 6' 11967), 240-2460 Katz, J. J. & Fodor, J. A. The structure of a semantic theory. Language, 38, (1963), 170-210. Lyons, J. Introduction to theoretical linguistics. Cambridge, The University Press, 1968. Mowrer, O. H. The psychologist looks at language. American Psychologist, 9, (1954), 660-694. Newman, E. G. & Gerstman, L. J. A new method for analyzing printed English. Journal quExperimental Psye Ch01092' 44' (1953), 114-125. Osgood, C.'E. On understanding and creating sentences. American Psychologist,.18, (1963), 735-751. Ogden, C. K. & Richards, I. A. The measupgment of meaning. London, Routledge & Kegan Paul, 1930. Schneider, H. & Schneider, N. Science far and near. Boston, D. C. Heath 8 Co., 1965. Shepard, R. N. Production of constrained associates and the informational uncertainty of the constraint. American Journal of Psychology, 76, (1963), 218-228. Staats, A. W. Learning, language and cognition. New York, Holt, Rinehart and Winston, i968. Thorndike, E. L. & Lorge, I. The teacher's word book of 30,000 words. New York: Columbia University Publications, 1944. Weiner, M. & Mehrabian A. Language within language. Appleton Century Crofts, 1968. Werner, H. & Kaplan, E. The acquisition of word meanings. Monograph of the society for research in child development, 15, (1950?? 3-120. Whatmough, J. Introduction to Semantics, studies in the science of meaniwg, (M. Breal.) New York, Dover Pub. Inc., 1964. A Winer, B. J. Statisticalyprigciples in experimental design. New York, McGraw Hill, 1962. 94 Woodward, H. M. E. A nonsense test of structural meaning. Journal of verbal Learning and Verbal Behavior, 7' (1968*), 313-32. Woodward, H. M. E. The structural component of linguistic meaning and the reading of normally hearing and deaf children. Unpublished M.A. Thesis, Washington University, St. Louis, 1963. Woolum, S. J. & Johnson, D. M. Concept acquisition as a function of frequency of occurrence in verbal context. Psychonomic Science, 14, (1969), 200. Woolum, S. J. The learning of concepts from verbal context. Unpublished M.A. Thesis, Michigan State University, 1968. APPENDICES APPENDIX A Instructions On the following page is a list of nonsense words. The words are made up but they represent things that are real. Your job is to guess what each word represents and write your guess beside the nonsense word. For example: 1. PAKEB 1. Chain £022ch 2. SOFON 2. 115219, 3. TUPID 3. W 4. NAGET 4. Zea/my Please note that all guesses should be nouns. They may be either concrete nouns (rug, checker board) or abstract nouns (fame, justice). Do not guess the same thing twice but be sure to make a guess for every nonsense word given. When you have finished with the first page, turn to the next page and read the rest of the instructions. 95 10. 11. 12. 13. 14. 15. 17. 18. 19. 20. 21. 22. 23. FINOH KADAL HODOW MOKET HEKOL FITAH HAFET TOSOY SELEN NUHIL WADEN NAHET RATAV NEHIT GARIF RATEN DEWAP TATEM KIHET LEBEW YETEH WALAN VATAH 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 96 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. DOFOD TOREF NUSOB TANAR REGOK YEMAT GIDAK SETID HEBOD RATAS DETEF HURAB TAHEL PUNAR 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 97 98 Instructions - Part II On the following page each of the nonsense words will be used in a sentence. For example: . (A-AN) PAKEB is flat. . (A-AN) SOFON is religious. . (A—AN) TUPID is not furry. . (A-AN) NAGET is green. bump-I The parenthesis around the (A-AN) simply indicates that your sentence may or may not require an article. (Thus, "Fame is, green," but {A rug is green” and {An apple is green.") After reading the sentence you may want to change your guess about the meaning of the nonsense word. If your first guess is still ok write "same" beside that sentence. If it is not ok, make another guess. (It will prObably be easier to com- pare the sentences with your original guesses if you detach the pages and place each word beside its sentence.) For example: First Page Second Page 1. PAKEB i 1 1. (A-AN) PAKEB is flat. 1. 2. SOFON 2; 9.. (A-AN) SOFON 1's religious2., . 3. TUPID 3. 1 I 3. (A-AN) TUPID is not furry3. ' ILL :LOL . .4zzamc/' 4. NACET ”'1g2nnxu .,4. (A—AN) NAGET is green. 4. ; I'll. Ilall ,. ‘ r l 'I . I d l v o u A u 0 a . e o ”C. .‘1: 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. (A-AN) (A-AN) (A-AN) (A-AN) (A—AN) (A-AN) (A—AN) (A—AN) (A-AN) (A-AN) (A-AN) (APAN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (Pk-AW) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) FINOH KADAL HODOW HEKOL FITAH HAFET TOSOY SELEN NUHIL WADEN NAHET ble. RATAV is is is is is is is is is is is is is 99 hold. not sober. not rapid not heedless. forgetful. motionless. not beautiful. intellectual. not natural. not choral. narrow not responsi- not empty. NEHIT is not dis- proportionate. GARIF RATEN DEWAP TATEM KIHET LEBEW YETEH WALAN VATAH is is is is is is is is is not curious. not supreme. not ideal. ripe. silly. corduroy. busy. not glorious. cheap. 13. 14. 15. 16. 17. 18. 19. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A-AN) (A—AN) (A-AN) (A-AN) (A-AN) (15.-AN) (A-AN) (A-AN) NUSOB TANAR REGOK GIDAH SETID HEBOD RARET TEDON NORAY HEMIB LOTIW RATAS DETEF HURAB TAHEL PUNAR is is is is is is is is is is is is is is is is is is is 100 gay. ebony. secure. not tight. not colonial. dim. spiritual. queer. not gross. canvas . 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. not productive34. not moist. not old. not perfect. modern. ridiculous. pearl. not stormy. humble. 35. 38. 39. 40. 41. 42. 101 1. What do you think the experiment was trying to discover? 2. Did you use any particular strategy for making your original guesses? If so, what was your strategy? APPENDIX B 1, II! I fill lllllll 'I' I ll’ ll‘l'll' II (I! Instructions On the following page is a list of nonsense words. The words are made up but they represent things that are real. Your job is to guess what each word represents and write your guess beside the nonsense word. For example: 1. PAKEB LeerQDpL 2. SOFON 2. roux 3 . NAGET 3 . Jed/mi, Please note that all guesses should be nouns. They may be either concrete nouns (envelope, rug) or abstract nouns (fame). In this part, do not guess the same thing twice but be sure to make a guess for every nonsense word given. Please do not look ahead in the booklet. When you have finished with one page turn to the next. 102 10. 11. 12. METIF LATIG TEJUH YOSEL KIWEX TIVER BOQOH WIYER TUNEP EAZAC NOLUN LILIX 10. ll. 12. 103 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. NAKAT MEGED RECUV SONIL DOGIG SARER PETIR SEPIW TATEN DUDAY JIHOM SOBAG 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 104 IV ll:[[.i||[rl‘1ll| ‘.lill[‘lllllli“!llj]i‘l 25. 26. 27. 28. 29. 30. WEDIZ NESOT RITAR CEGOJ NUTEP HIRAY 25. 26. 27. 28. 29. 30. 105 IV \I ill'llllllillll‘llll‘lll‘lln'llll ll!:llll’.lilllrllllJ \v \III III 106 Instructions -- Part II On the following page each of the nonsense words will be used in a sentence. For example: 1. The PAKEB lived by the sea. 1. 2. It was a large SOFON 2. 3. NAGETs always remind me 3. of Thanksgiving. After reading the sentence you may want to change your guess about the meaning of the nonsense word. If your first guess is still ok write "same" beside that sentence. If it is not ok, make another guess. (It will probably be easier to compare the sentences with your original guesses if you detach the pages and place each word beside its sentence.) For example: First Page Second Page 1. PAKEB l. grudLOPi, 1. The PAKEB lived by 1. '0wa the sea. 2. SOFON 2. Coca 2. It was a large SOFON. 2. bcumi, 3. NAGET 3. «Ea/m2, 3. NAGETs always remind BWXEHF, me of Thanksgiving. 3 Please do the sentences in order and do not change a guess after it is made: your first impression is fine. Do not leave any item blank. 10. 11. 12. 107 I wouldn't be surprised if METIF were something photographic. Anyone can make a LATIG. A fat person has more TEJUH pull on his body than a skinny person. YOSELS show different pictures. KIWEX stands for alot of things. TIVER is usually hand-picked. The BOQOH is almost white on the bottom. I'd like to see the rest of WIYER. TUNEP seems to be some kind of optical test for the imagination. FAZAC has sloped shoulders. When entropy or NOLUN is lost from a system, energy is used to cause this. LILIX is a reward. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 108 Most people like to eat NAKAT. MEGED has a rounded top. RECUV looks like a strange walking creature. SONIL would look great sitting on my bookshelf, she's quite a conversation piece. DOGIG is great. SARER is being able to do as you please as long as you don't hurt anyone. PETIRS are sometimes amazing. SEPIW has a hard core. TATEN has an electric cord attached to it so it must need electricity to function. Lack of DUDAY can be seen in outer space. A ball falls to the ground because of JIHOM. Galileo experimented with SOBAG. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. IV 25. 26. 27. 28. 29. 30. 109 Maybe tomorrow I'll win a WEDIZ. It would be difficult to assemble a NESOT. RITAR represents a group of people. The CEGOJ can shoot laser beams that can _ shoot down any enemy plane traveling under 100,000 mph. The NUTEP needs sunlight to grow. There is very little HIRAY on the moon. 25. 27. 29. 30. IV 110 1. What do you think the experiment was trying to show? 2. While you were reading the sentences did you notice any two or more sentences that seemed to describe the same concept? If you did, give an example of the sentences or the concept described. (Please don't look back). 3. Please indicate sex. Male Female Thank you for your participation. II: J] ll. lllll|||..flllllu .‘II' ll‘l.i [Il’ APPENDIX C Sentences that were eliminated because more than five subjects guessed the correct concept after reading this sentence (Part III B). Number out of 30 who guessed Sentences correctlyy Gravity* Gravity is the force pulling you toward the center of the earth. 27 A ball falls to the ground because of gravity. 26 Without gravity we would float in the air in every which way and begin to spin around and around 26 If there were no gravity we would be thrown out into space. 25 Did you ever see pictures of astronauts flying around in a room without gravity pulling them down. 22 Isaac Newton was the first to come out and comment openly on gravity. 20 He began his study of gravity when an apple fell from a tree. 18 Gravity, of course, is part of the idea that "whatever goes up must come down." 16 Gravity is what keeps the moon in the sky. 15 It is gravity which accounts for the arc in which an object falls when projected hori- zontally. l4 *In the test booklets, various nonsense words were substituted for this word. 111 112 Number out of 30 who guessed Sentences correctlyy Gravity (cont.) The moon has less gravity pull than the earth. 14 Galileo dropped the balls simultaneously from the Tower of Pisa to see on which one gravity had the greatest effect. 13 The value of gravity (acceleration due to gravity) constant is about 32 ft sec2 in the English system, or about 9.8 meters/sec2 in the metric system. 11 These men of the Apollo 10 mission are now under the influence of the moon's gravity. 11 Without gravity there would be no atmosphere surrounding our planet. 11 Actually gravity is the concept that all bodies have an attraction for each other. 11 Gravity affects motion. 8 All planets have a field of gravity. 8 Is gravity dependent on rotation of a body? 7 Gravity is like a magnetic force. 6 Freedom* The American Flag supposedly stands for freedom. . 18 Freedom is necessary, but so is the protection of the rights of others. 17 The liberty Bell is a symbol of freedom. 16 Freedom is being able to do as you please as long as you don't hurt anyone. 16 The American Revolution was a fight for freedom. 15 We try and advocate the freedom of speech, religion, want and the freedom from fear. 15 113 Number out of 30 who guessed Sentence correctlyi Freedom (cont.) Freedom to some is the right to burn buildings and hurt people. 14 Personal freedom is guaranteed by the U.S. Bill of Rights. 14 Freedom of speech is allowed. 13 To have freedom is to be able to do anything that you want. 11 Today freedom is really very limited in the fact that society is placing more and more restrictions on people. 11 Freedom means no barriers. 10 Michigan State University has an Academic Freedom Report. 8 America is said to be a land of freedom! 8 But to the black slaves in the old South, freedom was not so common. 7 Cabbage* Cabbage can be produced as sauerkraut. 26 Cabbage is primarily used in cole slaw. l3 Brussel sprouts are junior cabbages. 12 I use to get cabbage mixed with lettuce. 10 Stuffed cabbage is very good. 8 Cabbage goes well with corned beef. 6 114 Number out of 30 who guessed Sentence correctly Projector* For example, one would use a 33 mm projector to show film taken with a 35 mm camera. 21 Some projectors have a button that you can push and turn the slides automatically. 19 A projector may be used for showing home movies and scenes from vacation trips. 18 A projector is a device that shows slides in a larger view. 15 A projector shows moving pictures. 14 Usually when a projector is in use there is a narrator describing the pictures seen. 13 The projector was convertible for movies and slides. 12 In order to project an image from film onto a screen, one needs the use of a projector. 7 APPENDIX D GRAVITY-~RULAT Infor- Recog.* Recall* mation Value Value Low information 1. The force of RULAT is limited. 19 l 0 2. Do you think there is RULAT in heaven? 18 l 0 3. Thank goodness for RULAT. 15 0 0 4. Everyone takes RULAT for granted. 20 2 0 5. If it weren't for RULAT, there wouldn't be much of anything anywhere. 19 2 0 6. RULAT cannot be tasted. 20 0 0 7. RULAT can be dangerous. 18 0 0 8. RULAT is universal. 17 2 0 9. This RULAT he thought was related to the objects mass. 19 2 0 10. RULAT as a physical phenomenon is amoral. 20 0 0 Middle information 1. Life is a struggle against RULAT. 25 l 0 2. RULAT is a physics principle. 26 l 1 3. RULAT presents a problem to the space program. 26 l l 4. RULAT is strong. 25 3 0 5. You can also learn about RULAT in HPR 105. 25 0 1 6. All masses have RULAT. 25 0 0 7. I wonder if someone will ever invent anti-RULAT juice. 24 3 0 8. I first learned about RULAT in grade school. 23 l 2 9. RULAT is responsible for many of the phenomena which occur in nature. 23 2 1 10. RULAT is the ruling force of the universe. 25 4 1 High information 1. Without RULAT there could be no existing life on earth. 29 3 0 2. It is RULAT which determines the tides, orbits of objects around earth, etc. 29 4 1 *Number correct after this sentence alone. 115 116 GRAVITY--RULAT Infor- Recog. Recall mation Value Value High information (cont.) 3. RULAT enables us to stand, walk, sit or move as we want instead of having no control of movement. 29 3 0 4. We can escape from RULAT. 29 3 1 5. A fat person has more RULAT pull on his body than a skinny person. 28 2 0 6. Galileo experimented with RULAT. 28 3 l 7. Some formula describes the effects of RULAT. 28 2 0 8. Although man has been able to conquer RULAT he has not been able to exhibit complete control of this force. 28 2 l 9. Machines such as pumps operate under RULAT. 28 1 0 10. RULAT also determines our weight. 28 4 0 117 FREEDOM--KEHET Infor- Recog. Recall mation Value Value Low information 1. KEHET should be. 17 0 0 2. Does anyone really know what KEHET is? 15 0 0 3. KEHET is great. 18 2 0 4. KEHET is different things to different people. 13 l 0 5. Some countries do not have KEHET. 18 l l 6. KEHET is relative (this statement is as ambiguous as the word itself). 20 0 0 7. KEHET, said Sartre, is anguish. 18 l 0 8. We have a free university here, which means KEHET. l7 2 l 9. KEHET is supposed to be shared by all. 19 2 0 10. KEHET can be abused. l4 2 0 Middle information 1. Man has KEHET of the mind. 25 4 0 2. Man will have complete KEHET only at death. 24 l 0 3. KEHET is not. 23 0 0 4. When people are denied KEHET in this country they call it a responsibility. 26 3 2 5. You can't go around killing people just because you have the KEHET to do it. 26 0 0 6. KEHET stands for a lot of things. 26 0 l 7. Everyone longs for KEHET from something at one time or another. 26 2 0 8. KEHET is not a visible object. 24 0 0 9. KEHET is therefore guilt. 26 0 0 10. All believe they possess KEHET is some sense, which is ludicrous. 25 l 0 High information 1. Why should others fight for KEHET while some are born with it? 28 3 3 2. When entropy or KEHET is lost from a system, energy is used to cause this. 29 0 0 3. If all the orders I got were merely termed advice I would be on the road to attaining more KEHET. 29 3 l 118 FREEDOM--KEHET Infor- Recog. Recall mation Value Value High information (cont.) 4. Personal KEHET is walking barefoot on the beach with wind blowing through your hair. 29 4 2 5. KEHET is the most valuable gift man has. 29 2 0 6. A democratic state includes KEHET. 29 0 0 7. KEHET is no more possible than total anarchy. 29 2 0 8. KEHET means guaranteed rights. 28 2 0 9. Every country in the world wants to be known as the stronghold of KEHET. 28 1 0 10. Others give their children too much KEHET. 28 4 l 119 LUREK--REPOL Infor- Recog. mation Value Low information 1. REPOL is basically black and white. 2. I would say that most had never seen REPOL before. 3. REPOL hasn't made any sound. 4. I don't know why I see such sordid things in REPOL. 5. REPOL has no arms as human beings do. 6. REPOL wears clothes. 7. REPOL does not appear to be very sturdy. 8. One might hypothesize that REPOL is a visitor from another world. 9. REPOL could be described in many ways. 10. Did Andy Warhal invent the REPOL? Middle information 1. 2. owmfl ease High REPOL looks like a tree with a fish on top of it. REPOL seems to be some kind of optical test for the imagination. I don't care to meet REPOL. REPOL has dark skin, I believe she's then colored. REPOL would look great sitting on my bookshelf, she's quite a conversation piece. I'm not sure, but it seems that REPOL just might have some Freudian symbols about her. REPOL has a normal neck. REPOL might frighten some people. REPOL is shaped somewhat like a human. REPOL has funny looking lips; no wonder she doesn't have a beaux. information REPOL's nose is funny. I almost think that REPOL and her people are similar to hampsters. The Granny REPOL I know is not quite as pretty as the REPOL I now an observing. REPOL has sloped shoulders. 12 19 18 18 12 17 15 18 12 11 25 25 24 25 25 25 24 25 25 25 27 28 28 28 OI—‘OO l—‘N N OOH NOD-41b w 120 LUREK--REPOL Infor- Recog. mation Value High information (cont.) 5. REPOL is of a small family from a small town. Cardboard is what REPOL is made up of. REPOL is a placid person who is content with her life. Everyone loves REPOL because she is friendly and cheerful making no special demands from anyone. REPOL is very good at cheating on tests. If REPOL is related to a chicken or another type of bird, then it should be able to lay eggs. 28 l 27 2 27 2 27 0 27 1 27 0 121 TEMOS--HETAB Infor- Recog. mation Value Low information 1. 2. HETAB is associated with a peculiar instrument. Often times a HETAB can be found at an arcade. HETAB could also be a secret weapon. HETAB is an almost indescribable thing. To a Freudian, HETAB might also be an unconscious sex symbol. Using HETAB would be quite simple. HETAB has an opening at the front of it. HETAB has very interesting designs in the shape it forms. The other main wooden part might be a HETAB box. HETAB looks fairly stable for what job it may perform. Middle information 1. 2. 3. Ulub 6. 7. 8. 9. 10. High 1. 2. HETAB rests upon a stand that appears to provide very adequate support. How old is HETAB? HETAB is the mechanistic nightmare of modern life. HETAB's viewer is black. HETAB is relatively inexpensive to purchase, but brings forth good profits. HETAB looks like a loudspeaker. HETAB has a rounded top. HETAB unfortunately does not work. HETAB's opening resembles a pair of goggles. HETAB looks like a new sinister weapon straight from the pure, pristine ivory tower walls of untainted scientific research. information HETAB is partially made from a TV picture tube. The HETAB can shoot laser beams that can shoot down any enemy plane traveling under 100,000 mph. 19 15 17 11 17 14 18 18 18 23 22 22 23 23 23 24 23 23 23 28 27 P‘Obo H NI—‘N DUO MOP-'45P 122 TEMOS--HETAB Infor- Recog. mation Value High information (cont.) 3. 4. 5. 6. 7. 8. 9. 10. HETAB has nails in the side. Of course there might be girlie shows in a HETAB also. HETAB is made of wood, plastic and masonite. HETAB is a modern day replica of the early movie projectors that were run for 5¢ back in the early part of the century. HETAB looks like a strange walking creature. HETAB seems to be homemade also. HETAB has an electric cord attached to it so it must need electricity to function. HETAB can be used by a psychologist to watch peoples eyes. 27 28 28 29 27 28 27 26 123 PROJECTOR--SEYAH Infor- Recog. Recall mation Value Value Low information 1. A SEYAH can be used for many reasons. 15 0 0 2. The SEYAH was as old as silent movies. 11 2 0 3. SEYAH come in all sizes. 16 0 0 4. I wish I had a SEYAH like that. 10 0 0 5. SEYAH are very useful. 18 0 0 6. The SEYAH is square. 15 0 0 7. The SEYAH is broken. 16 0 0 8. It would be difficult to assemble a SEYAH. 16 l 0 9. The SEYAH was small. 5 0 0 10. The SEYAH is light and dark brown. 11 0 0 Middle information 1. SEYAH can be useful in bringing back old memories. 23 2 0 2. SEYAH are fun to use. 23 0 0 3. Usually the SEYAH has a reverse switch. 21 4 0 4. The slides, which he had intended to show with the SEYAH, were rather boring. 22 4 3 5. SEYAH are sually expensive. 22 0 0 6. Grandfather had an antique SEYAH. 22 0 0 7. SEYAH represent bridges which link the world of reality with the world of fantasy. 22 l 0 8. Did you ever see Superman on a SEYAH? 23 l 0 9. We use SEYAH in almost all classes. 21 4 0 10. Movie theaters also use SEYAH. 21 4 0 High information 1. The SEYAH has a fan on the inside to keep it from getting too hot. 26 4 0 2. SEYAH are even manufactured for the small child to use and play with. 28 l 0 124 PROJECTOR--SEYAH Infor- Recog. Recall mation Value Value 3. Not too long ago, a new technique for showing movies called cinerama came into use, utilizing three SEYAH stationed at three different locations giving a very wide screen effect for the movies. 27 3 1 4. A SEYAH has to be focused. 29 0 0 5. There are many companies which produce SEYAH: Bell & Howell, Kodak, etc. 29 4 0 6. SEYAH shine light through a photographic tranSparency. 27 l l 7. But, of course, different film is used in each SEYAH. 29 4 0 8. Anyone can make a SEYAH. 28 0 0 9. SEYAH can magnify a picture. 28 4 l 10. SEYAHs show different pictures. 27 3 0 125 CABBAGE--DOFIH Infor- Recog. Recall mation Value Value Low information 1. The DOFIH is wrapped in a plastic bag. 17 2 0 2. Rabbits like DOFIH. 17 4 0 3. I have also seen a red DOFIH. 10 1 0 4. You buy DOFIH in the super market. 17 0 0 5. Who cares if the DOFIH is kept sanitary? 16 0 0 6. You can do a lot of things with DOFIH. 11 0 0 7. DOFIH are round in shape. 17 l 0 8. The DOFIH is dirty when you first buy it. 19 2 0 9. In fact I've met many people that could have passed their brain for a DOFIH. 16 l 0 10. The DOFIH is almost white on the bottom. 16 l 0 Middle information 1. Home grown DOFIH is quite leafy. 25 4 0 2. DOFIH is good in a salad. 25 2 0 3. DOFIH is included in many salads and dishes in restaurants. 25 l 0 4. If DOFIH is measured by the head, then one head is on the stand. 25 4 0 5. DOFIH should be thoroughly mashed before eating it because the Birchers are spraying poison on it in order to kill off the left wingers, but they are getting a lot of middle of the roaders at the same time. 25 l 0 6. DOFIH is green and white (Yea MSU). 24 l 0 7. Most pets eat DOFIH. 25 0 0 8. Also if you cook DOFIH too long you lose any nutritional value you started with. 25 2 0 9. Most people like to eat DOFIH. 25 l 0 10. DOFIH is usually hand-picked. 25 0 0 126 CABBAGE--DOFIH Infor- Recog. Recall mation Value Value High information 1. DOFIH grows on the ground. 28 l 0 2. Did you ever smell a DOFIH field after its been cut. 30 l 0 3. A DOFIH will die if it gets too cold out. 30 l 0 4. One of my father's favorite meals is canadian ham, DOFIH and potatoes all boiled in the same pot. 28 4 0 5. On some days we'd cut over a hundred crates of DOFIH. 28 1 0 6. Many people eat DOFIH. 29 0 0 7. DOFIH is a vegetable widely used by cooks all over the world. 27 3 0 8. DOFIH is usually grown in a garden. 28 0 1 9. DOFIH is frequently used for meals. 29 0 0 10. DOFIH is produced by farmers. 27 2 0 APPENDIX E Draw a REPOL. Draw a HETAB. Which of the objects dis- Which of the objects dis- played is a REPOL? played is a HETAB? #1 #1 #2 #2 #3 #3 #4 #4 #5 #5 #6 #6 #7 #7 #8 #8 #9 #9 #10 #10 127 The real word for SEYAH is Which words SEYAH. 1. one of the following is the real word for Lamp(s) Toaster(s) Piano(s) Thermometer(s) Projector(s) Radio(s) Phonograph(s) Typewriter(s) Microscope(s) Television(s) 128 The real word for DOFIH is Which words DOFIH. 1. 2. one of the following is the real word for Liver Bread Cactus Banana(s) Cucumber(s) Cabbage(s) Grapefruit(s) Cereal Coke Pumpkin(s) The real word for RULAT is Which of the following words is the real word for RULAT? l. Diffusion Locomotion Electricity Friction Erosion Reflection Inertia Gravity Radioactivity Velocity 129 The real word for KEHET is Which of the following words is the real word for KEHET? Propserity Truth Beauty Emotion Love Justice Fate Pride Freedom Chance APPENDIX F Instructions This experiment deals with six concepts, each of which is represented by a nonsense word. The words have been made up but they represent real things or ideas. Your job is to discover what each word represents. You will be given two booklets, one has the nonsense words and the sentences, the other has questions for you to answer, or things for you to do. Sometimes you will be asked to draw the object, other times you will be asked to supply the real English word for the object or to choose which object is correct from a list of words or from a display of things. When you have a choice of answers, please circle the correct alternative. When you supply a word it should be a noun but it may be either concrete (caterpillar) or abstract (fame). Thus on the first page of the information booklet you might see the word GIDAK and on the first page of the question booklet it might say, "Draw a GIDAK." Even though you do not know what a GIDAK is be sure to make a guess before you turn the page. Always turn oneypage_in the question booklet for each page in the information booklet. 130 131 The next page will present a sentence about the idea that is represented. Thus a sentence might say: You can't buy GIDAK. This sentence may help you find out what GIDAK is. If your first guess for GIDAK is still your best guess write "same” after the next question. Please do not draw a new picture or select a new alternative if you do not change your guess: just write "same." If you do change your guess, then draw a new picture or select a new alternative. The object on display are relevant to the part of the study that says, "Which of the objects on display is a GIDAK?" They have nothing to do with the part that says, "Draw a GIDAK," so please do not use them for models. Remember: 1. The point is to discover what a GIDAK (or some such) really is. 2. After each new nonsense word you should make a guess and after each new sentence you should either make a ngy_guess or write "same." 3. It is very important that you do not leave any question blank. There should be 11 information pages and 11 question pages for each concept. If something seems to be incorrect please tell the experimenter. Please do one page at a time. Don't skip any and turn ahead and don't look back. MMMMM TUN u’ffl‘fimmfllflmfiu Irzflilfitlfiflflfljlfgfiml“ 3 12 0317