SEIAITIC GENERALIZATION AS A FUNCTION OF CONNOTATIVE SIIILKRITY AND [WARENESS OF STIMULUS RELATIONSHIPS By Jean Judson Smith AN ABSTRACT Submitted to the School of Advanced Graduate Studies of Michigan Btnte University of Agriculture and Applied Science in pertiel fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology Year 1956 lpprovod fl' @( 03%, 1 Jean Judson Smith various nonpvcluntary responses, when conditioned either to verbal or to non-verbal stimuli, exhibit generalization to other stimuli which are semantically related to the conditionp ed stimulus. Several studies indicate that this semantically mediated generalization frequently occurs in the absence of awareness by the subject of the relationship between the coup diticned and unconditioned stimuli. In the present study, the galvanic skin response was conditioned to a verbal stimulus and.tested for generaliza- tion to a number of semantically related stimuli. The gener- alization of extinction of this response was also investiga— ted. Forty-nine experimental subjects were employed, 32 of ‘whcm were aware of the conditioned stimulus-unconditioned stimulus relationship. The study was designed to test three hypotheses. Hypothesis I predicted.that semantically mediated gener- alisation.would'be negatively related to the degree of conp notative difference between the conditioned stimulus and the several generalization stimuli, as established by Osgocd's Semantic Differential. Hypothesis II predicted that the up were subjects, as a result of the mediating function of their ability to verbalise the conditioned stimulus-unconditioned stimulus relationship, would exhibit greater generalization 2 Jean Judson Smith- than would the unaware subjects. Hypothesis III predicted that, for the same reason, the aware subjects also would ex- hibit greater generalization of extinction effects than would the unaware subjects. The results clearly indicate the Hypothesis I is un- tenable. A non-hypothesized finding was that the degree of generalization was positively correlated.with the magnitude of wordpasscciation frequency between the conditioned stimup Ins and the generalization stimuli. Hypotheses II and III are supported by the general trend of the data, althoughzit is doubtful that a decisive test of the hypotheses was af- forded by the present study. The study raised a problem for future investigation of the degree to which word association frequency may be uti- lized.in_predicting the degree of semantic generalization be- tween verbal stimuli. Problems were also raised.with respect to the relationship between conceptualization and condition- 108. The results appear to warrant the following conclusions: 1. No relationship exists between connotative differences between words, as measured by the Semantic Differential, and semantic generalization to such words. 8. Semantic generalization may occur as a func- tion of wordpassociation frequency, as mea- sured by the Kent-Rosanoff lord Association Test. 3. Awareness of the signal-function of a con- ditioned stimulus may increase speed and strength of conditioning. 3 Jean Judson Smith Awareness of the signal-function of a con- ditioned stimulus may increase the gener- alization of responses conditioned to such st imuli. SEIANTIC GENERALIZATION AS A FUNCTION OF CONNOTATIVE SIIILARITY AND AWARENESS OF STIMULUS RELATIONSHIPS By Jean Judson Smith A THESIS Submitted to the School of Advanced Graduate Studies of Michigan State University of Agriculture and Applied Science in l in partial fulf went of the requirements for the degree of I DOCTOR OF PHILOSOPHY Department of Psyshology Year 1956 ACKNOILEDGEIENT The writer is greatly indebted to Dr. H. R. Denny, of the lichigan State University Psychology Department, for his indiepensible guidance and encouragement during the completion of this study, and to Dr. A. B. Dibner, now at Clark University, for providing part of the apparatus with which to carry it out. Dr. D. J. Montgomery, of the Iichigsn State University Physics Department, and.lr. A. J. Dascher, of the Electrical Engineering Department, also deserve sincere thanks for their generous help with the electronic apparatus. TABLE OF CONTENTS LIST 0’ T‘BLEBO O O O O O O O O O O O O O O O 0 LIST OF FIGURES . . . . . . . . . . . . . . . . Imomorlon O O O O O O O O O O O O O O O O 0 Semantic Generalization Theoretical Significance of Semantic Generalization a: man STUDY 0 O O I O O O O O O O 0 O O O The Semantic Differential Hypothesis I Semantic Generalization of Anxiety Hypothesis II Hypothesis III mHOD O O O O C O O O C I O O O O I O O O O 0 Selection of Stimuli The Semantic Differential Subjects Apparatus Experimental Procedure Gontrol Groups Egan’s.COOOOOOOOOOOOOOOOOO Experimental Groups lethod of Analysis Conditioning Semantic Generalization and the Semantic Differential Relative Generalization of the here and Unaware Groups Generalization of Extinction Effects flnmrlon 0' RESULTS 0 I O O O O O O O O O O O SUIIARY AND CONCLUSIONS . . . . . . . . . . . . MIDI: ‘ O O O 0 O C e O O O O O O O O O 0 O mmn B O 0 O O O O O 0 0 O O O O O O O O O mun c O O O O O O O O O O O O O O O O O O Blmmmm O O O O O O O O O O O O O O O O O 111 Page iv 18 Table 1. 10. 11. 13. 13. 14. LIST or nears. Inter-group Comparison of Generalization-test Diffuerences Between later lords and Non—water or O O O O O O O O O O O O O O O O O O O O 0 Inter-group Comparison of Differences Between Extinction-test Decrements to later lords and tolon-waterlords .............. Comparison Between 3m; and Pro—river lords for ”at. mcup O O O O O O O O O O O O O O O 0 Comparison Befleen mm and Pre-river lords formuSGroup....o.......... Average GSR of Experimental and.Ccntrol Groups to First Presentation of lord List 11 (Genera- lization Test). . . . . . . . . . . . . . . . . Comparison Between later lords and Ion-water lords for Aware toup (Generalization Test) . . Comparison Between later lords and Ion-water lords for Unaware Group (Generalization Test) . Comparison Beheen later lords and Ion-water lords for Generalization-control Group (muuization TOCU) e e e e e e e e e e e e e Comparison Between lean Generalization-test cm and lean Extinction-test ass for Aware Group. . Comparison Between lean Generalization-test can and lean Extinction-test cm for Unaware Group. Comparison Betwun lean Generalization-test cm and lean Extinct ion-test CBS for Extinct ion- 00nt101GrOupeeeeeeeeeeeeeeeee Comparison Between Extinction-test Decrements to ater lords and lion-water lords for Aware “cup 0 O O 0 O O O O O O O O O O O O O O O O 0 Comparison Between Extinction-test Decrements to later lords and Hon-water lords for Unaware Group..................... Co arison Between Ext incticn-test Decrements to ater lords and lion-water lords for Extinction-control Group . . . . . . . . . . . iv Page 54 69 7O 71 78 73 74 75. 76 76 77 78 79 LIST OF FIGURES Figure Page 1. Paradigm for lediated Generalization . . . . 3 2. Paradigm for Semantic Generalization of Aware and unaware Subjects . . . . . . . . . 37 3. Schematic Circuit Diagram of Electronic Ohmmeter for Recording GSR . . . . . . . . . 34 4. Schematic Diagram of Disposition of Experimental and Conrcl Subjects . . . . . . 45 5. G83 to River and Pre-river lords in the Aware and Unaware Groups During the Conditioning Series . . . . . . . . . . . . 47 IITRODUCTIOI Sines the early reports by Russian investigators (16, la, 19, 51, 54) that responses which had been eon- diticned to auditory and visual stimuli could be elicited by words referring to the conditioned stimuli, interest in suantio conditioning and generalisation has inereased steadily. Sasran (39, 40, 41, as, 43, M, 45) has re- ported a series of tnvestigatioss of semantic generli- sation of ecnditicned salivaticn. lie results, as well as those of other authcrs (1, s, a, s, s. v, s. 9, io, 1?, Bl, 33, 36, 46, 47, 55)1 have demonstrated the variety of the semantie relationships, and of the kinds of in- voluntary responses, whieh are susceptible to semantie generalisation. Among the responses which have been uti- lized in these investigations are , besides the salivary response, vasoecnstrietion (89), acceleration of heart rate (21, as), the pupillary response (19), retardation of the heart rate (30), and the galvanic shin response (3, 5, 17, 31. as, as). The term, ssmantie generalization, refers to a spe- eial ease of mediated generalization. The possible sig- nificance of mediated generalization for furnishing an account of many complex behavioral phenomena has been dis- euseed at length by Cofer and Foley (1), Bollard and lil- ler (4), and by Osgood (35). The conventional paradigm, essentially that used by each of these authors, is shown in Figure 1. It s r———)s—->B 5.. Figure I. Paradigm for Ied-iated Generalization 3:. 82. s3, . . . an here represent stimuli which have become attached, through a' learning process, to the same cue-producing response, r. As a consequence of thte learned association, they have become, tn varying degrees, functionally equivalent for elicting r, and therefore for eliciting E. sum response. Gradients of generalisation theoritieally are determined by two factors. First, where 1: represents a unitary response, gener- alisat ion theoretically will depend upon the relative strengths of the associations between 83, 813, . . . In, and r. If each is equally strongly associated with r, genera alisation should be complete; that is, each of the stimuli, 81, I3, S3, . . . Sn, will be equally effective in evoking E. On the other hand, where the strengths of association vary, the effectiveness of each stimulus in elioting R will be a function of the degree to which it has previously functioned in eliciting r. lowevcr, r may represent a response complex, rather than a unitary response. In this case, the elements of the response complex may have become conditioned to 81, 83, S3, . . . Snin varying degrees or combinations. Gen- cralisation might then be expected to follow a gradient which corresponds to the number of elements elicited by each stimulus. If 81 is utilised as a conditioned re- sponse in a new learning situation, it is expected that stimulus gsneralisat ion for the new response will be me- diated by the already-existing relationships between 32, 83, . . . 5n and r in one of the ways just outlined. mm We In the special case of semantic generalisation, 81 may consist of a verbal stimulus, an object, or a physical event. 83, 83, . . . as may be verbal stimuli which refer to 81, words bearing various semantic relationships to S1, transistor cventsrzrefcrred to by 81. An operational description of semantic generalisation is afforded by the following quotation from Cofcr and Foley! According to the data of Diven I , Rasran liess [4.6], and lylie [55], synonymous words, when presented, will likewise elicit a response, though in less degree. low P may have only one synonym, or it may also have several synonyms; in the latter case these several synonyms may not bear to P the same degree of semantic relationship, 1. e., one word (synonym) may be more completely in- terchangeable with P" than the others although all A given word P‘, is presented cinforcedE 35 39], must bear some degree of interchangeability with P. (In this connection it should be noted that language authorities seldom assert that two words have precisely the same meaning. It seems likely, therefore, that the foregoing statement is sound.) Therefore, we may have a series of words, vary- ing in semantic similarity to P, which would be likely to elicit different degrees of response as a function of the degree of their similarity to P. (5. p. 535). The following paragraphs will review the studies which have been concerned with semantic generalizat ion and with the nature of the stimulus gradients along which it occurs. Early Russian studies, as summarised in the W m W have demonstrated that responses which have been conditioned to colored lights and to bells exhibit sti- mulus generalisation to the names of the stimuli. Iapustnik (16), first showed that children's responses may generalise from a name to a visual or auditory stimulus, or vice versa, and that extinction of the conditioned response leads to extinction of the generalised response. Traugott (54), next conditioned a response to a bell, and extinguished the same response to the bell paired with a blue light (conditioned inhibition). le'thcn found that the word, m was more effective than the word, ‘35, a variety of neutral words, or a red light, for inhibiting the response to the hell. he also reports that the word, m was more effective than the blue light for inhibiting other unrelated conditioned responses. lmolenskays (51). conditioned a bulb-squeezing response as a discrimination between a blue and a green light, and between a rcdpwhite-ycllow and a yellow—white-red se- quence of lights. He found that the discriminations thus produced could be elicited by verbal presentation of the corresponding color names, while neutral words had no effect. Eotlisrevsky (19, 20), showed that both the pupillary re- flex and pulse retardation could be elicited by the word, m1, aftcr having been conditioned to the tone of a bell. Easran (39), in the earliest study in this country, used himself as a subject to study semantic generalisation of sslivation. He compared the amounts of salin elicited by thinking of a series of nonsense syllables, and of the various equivalents of the word, m, in English, German, French, Spanish, and Gaelic. He found s relationship be- tween the amount of saliva and the familiarity of the word, as measured by the double criterion of his speed of associ- ation to the word and his reading speed in the lengusge. Utilizing the salivary response again in a later study, Easran (40), compared the amount of generalisation of sali- vaticn to homophones and synonyms of words to which it had been conditioned. He found generalizat ion to both, with synonyms regularly eliciting a greater response than did homophones. This suggested to him that mediated generali- sat ion is stronger than primary generalizat ion. The latter study was repeated by both Riess (46), and lylie (55), utilising the galvanic skin response. lhile Riess's results ayeed with those of Razran, lylie found contradictory results, homophones eliciting greater generalisation than synonyms. Hie relates this discrepan- cy, however, to her failure to mislead her subjects as to the nature of the experiment, with the result that several subjects reported that they expected reinforcement of the homophoned, but not of the synonyms. She found also that the G51, when conditioned to nonsense syllables, exhibited generalization to paired associates learned earlier. lylic's argument about the influence of set in deter- mining her results is supported by further results of Rasran (42). He compared the influence of various facilitory, in- hibitory, and cognitive sets upon the degree of semantic generalisation, and found that the gradients of generalisa- tion were modified by his instructions to the subjects. This study involved oondit ioning both to tones and to words, and yielded further evidence that semantic generalisation may be greater than primary generalization. Further knowledge as to what kinds of semantic dimen- sions may mediate generalisation has resulted from studies by Rasran (45), and by Diven (3). Easran investigated the dependence of generalisat ion upon meaning changes by condi- tioning salivation to verbal propositions such as, 'Pover- ty is dcgrading.‘I He then reversed, and otherwise changed, subjects, predicates, and comics, and tested for genera- lisation to the new forms. he found greater generalisation to 'propositionally"similar than to 'sententially' si- milar forms of the sentences. Diven's study, which*will be discussed more fully in a later section of this paper, revealed generalisation of the GSR from the conditioned stimulus, 23:3, to rural words, and from t.g1,tc urban words. This added a class-relation dimension to those al- ready demonstrated. liess (43), studied the genetic aspects of semantic generalisation by comparing four different age groups for the degree to which they exhibited generalisation to ho- mophones, antonmms, and synonyms. le found that young children (mean age, 7 yrs., 9 mos.) showed most generalissp tion to homophones. next most to antonyms, and least to synonyms. An older group (10 yrs., 8 mos.) generalized most to antonyms, then to homophones, and least to synonyms. Finally, his two oldest groups (14 yrs., 0 mos.. and 18 yrs., 6 mos.) showed the greatest generalisation to syno- nyms, and least to homophones, with antonyms in between. Ec concludes: 'The present experiment has demonstrated that the relative strength of the semantic gradients does not depend on any a prior quality of the language, but upon the way-in which the whole organism ut - lises language in its development.... In other words, semantic conditioning does not depend solely upon any attribute of the stimulus as such, but the stimuli must be interpreted as part of large situap tionally experiential wholes within‘which the or- ganism exists and has learned to use and interpret verbal stimulation’. (47, p. 151) Rasran has contributed further evidence of the im- portance of learned relationships in mediating semantic generalization, as well as of the variety of these relation- ships. In one study (43), he compared the generalization on various semantic dimensions (partowhole, contrast, as- tion-agent, etc.) to that on several phonetcgraphic dimen- sions (pseudo-derivative, common elements). lhile he found a crude phonetic gradient, which was a function of the num- ber of common elements, he also found semantic gradients. The greatest semantic generalisation was to contrasts, co- ordinated, and subordinates. The importance of previously learned associations is sharply illustrated by lasran's findings when the stimli were elements. of compound words, «oh I» mm mails. he found that greater generalisat-ion occurs when the conditioned stimulus is E9211: and the generalisation stim- ulus is m than vice versa. Strong learning exists in English-speaking subjects which would lead to the elicita- tion of Mg, by presentation of m but not vice ver- as. Consequently, a greater probability exists that m presented as a generalisation stimulus, will lead to free- tional or implicit responses capable of mediating responses conditioned to m, than the reverse. A later study (44), supports this interpretation, and leads Barren to a similar conclusion. In this study, he 9. first had different subjects practice controlled association with various different verbal relationships, for example, part to whole, whole to part, sub- to superordinate, etc. , He then tested for semantic generalization to words bearing these relationships to the conditioned stimuli. He found that the scout of generalization along a specific semantic dimension was facilitated or inhibited by the set resulting from the subject's previous practice. For example, greater generalisation occurred from m to m when the sub- ject had practiced associating tum-dog, than when he had practiced dag-mm. In the same study, subjects were given varying amounts of training in the meaning of Russian words. Responses were then conditioned to certain words, and the degree of gener- alisation to other words was studied. lhcn the subjects had no knowledge of the meaning of the words, generalisation was greater to homophones than to synonyms, but, in consonanoe with the results of Biess's genetic study, there was an in- crease in generalisat ion to, synonyms with increasing know- ledge of their meaning. lasran concludes from this study that semantic generalisation is '...rooted in the learned activities of the individual.' (p. 256) Only two reported studies, by Eisen (5) and Keller (17), have failed to demonstrate cement ic generalisation. Eisen investigated the influence of set on semantic gener- alisation. He gave experimental subjects practice in pairing words having various semant io relationships to each other, and then compared them for degree of generalization 10 with subjects who had been infcraed of the relationships between the conditioned stimulus and the generalisation st imlus. Using a conditioned can, he found no significant difference between set conditions. testing the hypothesis that the greatest generalization should occur to words with the highest fhorndike—Lorge word-count frequency, he found no significant gradient. lramination of his criterion of conditioning, however, reveals that lisen's measure may have been too insensitive to reveal either conditioning or generalisation, if they did occur. He considered conditioning to have ocoured when the conditioned stimulus elicited a aeaaurable can, and the neu- tral stimulus did not. Since nearly a, Itilulus can elicit a measurable 03! in the presence of a sensitive aeasuring de— vice, these results raise a question as to whether lisen's apparatus was of sufficient sensitivity. Keller conditioned the can to a picture of a Boy Scout hat and tested for generalization to a picture of a fire- nan's hat and to the word, “1. Although generalisation occurred to the picture of the fireman's hat, none was found to the label, bli- leller interprets this as evidence that simple labeling did not aediate the generalisat ion between the two pictures. However, whether the observed generalisa- tion was primary or not, it is to be expected, in the light of the many other studies cited above, that m also should 11 have exhibited generalization. The discrepancy between these results and those of other investigators remains un- explained. In summary, these studies on semantic generalization, with two exceptions indicate that a variety of voluntary and involuntary responses, when conditioned to either verbal or nonpverbal stimuli, exhibit semantic generalization. The relationship existing between the conditioned stimulus and the generalization stimuli need not be one of physical or sensory similarity, but may be any of a variety of learned semantic and conceptual relationships. There is definite evidence that the effectiveness of these relationships in mediating generalization is readily altered by situational learning or by set. Furthermore, several of the studies sug- gest that semantic factors may be more effective than sen- sory or physical similarity in mediating generalization. wwmmw Cofer and roleyand their associates (1, 3, 7, a, 9) were the first authors to elaborate the theoretical possi- bilities for semantic mediation of complex behavior patterns. In the first of a series of articles on the topic (1) they summarise previous studies of generalisation from object to sign, from sign to object, and from sign to sign, present- ing an operational formulation of mediated generalization similar to the paradigm discussed above. They also outline 18 an elaborate theoretical scheme of higher-order generali- zation which ut ilized both homophones (non-mediated genera- lisation) and synonyms Qmediated generalisation). The theoretical possibilities of their scheme are illustrated by the following example from their discussion of the medi- ation of emotional behavior. In their example, elements of the feelings aroused by an anxiety-provoking experience, associated with the act of lighting a cigarette, may generalise to the verbal label, llight, tron this, generalisation may occur to the synonym, m, and thence to the homephone, m. l'urther generali- sation may occur to another synonym, hang, finally,rthefa- . rousal of the label, hang, upon sight of the actual object may serve to redintegrate a portion of the original anxiety, resulting in a subjectively inexplicable phobia. The occurrence of higher-order chains of generalization effects such as those hypothesised by these authors remains to be demonstrated, and in fact at least one author (83) has reported failure in an attempt to demonstrate it. It is apparent, nevertheless, that no such complex chains need occur in order for semantic factors to play a significant part in mediating many aspects of human behavior in which symbolic-language processes are involved. In their introductory article, Cofer and Foley outline possible applications of the concept of semantic generalisa- 13 tion to problems of perception, set, learning, memory, in» telligence, reasoning, free association, and emotional be— havior. Subsequently, they and their associates published a series of investigations of certain homophone and synonym gradients (7, 9), antonym gradients (a), and of the influ- ence of differences in professional training upon word asso- ciation (8). lash of these studies utilised transfer of verbal learning as the experimental technique. The first such study tested for positive transfer to a.word list from several other lists which the subjects had studied. rive learning lists were employed. Two of these consisted.of homophones of the words in the test list, one ‘was comprised of synonyms of the test-list words, and the fourth contained.words which were synonyms of the words in the first synonym list, but which bore no semantic relation to the words in the test list. A control list was made up of words not semantically related to the contents of the test list. live groups of subjects each studied one of the learning lists, and were then tested for positive transfer to the test list. Bach of the first four listed, including the list of second-order synonyms, led to significantly greater positive transfer than did the control list. The greatest transfer, however, was to homophones, in conflict withthe findings of all other investigators but lylie. Osgood (35) has made an important criticism.of the de- 14» sign of this study, which holds as well for the rest of the series by these authors. This criticism may also account for the discrepancy between their findings and those of Ra- san, et al., with respect to the relative order of homo- phones and synonyms in eliciting generalisation. Osgood points out that the experimental technique in the above study allowed the subjects to recognise the nature of the relationships between word lists, and to reconstruct the test list on the basis of this knowledge. Any subject who recognised, during the presentation of the test list, that it consisted, for example, of synonyms of the list which he had previously studied, need then only recall the items of his learned list and manufacture synonyms, in order to a- chieve a high score on the test list. The second study in this series utilised the same transfer technique to demonstrate semantic generalisation from a learning list to a list of antonyms, yielding posi- tive results. The third study used a group of subjects who had studied Spanish and a group who had not. They were com- pared for degree of positive transfer from a list of Spanish words to a list of lnglish synonyms of the Spanish words. The Spanish students exhibited significantly greater trans- fer. lcwever, both of these studies are equally susceptible to Osgood's criticism. Bollard and Killer (4), have presented an elaborate a- nalysis of complex human behavior, including reasoning, pro- 15 blem solving, formation of neurotic symptoms, conflict, and psychotherapy, in terms of the mediating function of language. A central process in their analysis is that of verbal labeling. They point out how verbalisations may function as cue-producing responses to a wide variety of stimulus situations, therebp mediating the generalization of previously learned responses to any new situation which may elicit these verbalisations. For example, they state . that: Attaching the same cue-producing response to no distinctive stimulus objects gives them a certain learned equivalence increasing the extent to which instrumental and emotional responses will general- ice from one to the other. (p. 101) Ihen . . . fear is attached to response produced cues, any new stimulus that becomes able to elicit the response producing these cues will arouse the fear they elicit. For example, a young man may have fear attached to the cues produced by the first incipient responses of sexual excitement. Then if a previously indifferent girl is labeled 'sexy,‘ she may arouse incipient responses of sex- ual excitement which in turn elicit fear. (p. 163) Although much of their exposition consists of untest- ed hypotheses, it appears to afford a rich potential for explaining many complex processes in terms compatible with the concepts of general experimental psychology. Their emphasis on the mediating role of language serves to point up the potential importance of semantic generalization as a mechanism by which this may be accomplished. Osgood (36), has elaborated the paradigm of mediated 16- generaliaation into a learning hypothesis which embraces not only instrumental learning but also problem solving, concept formation, and language behavior. lith respect to the last of these, he states that: In terms of its cental relevance to general psy— chological theory and its potential applicability to complex social problems, no other area of ex- perimental psychology so greatly demands atten- tion as language behavior-and.in the past has re- ceived so little. (p. 737) An interesting example of the degree to which seman- tic generalization may be implicated in the complex mental processes is furnished.by a recent experiment by Eisman (6). She first trained four groups of subjects to apply the labels, gag, m, and .33, differentially to differently shaped whiteblocks. Each group was also trained to choose one of the shapes in preference to the others, by associa- tion of the shape with a reward. The second step was for the subjects to learn to apply each white block's label, 3.3, And!» or ggg, to a'black, yellow, or green rectangu- lar block. That is, the subjects had.now learned to apply the label, 13g, for example, to a white triangular block and.a green rectangular block, and so on. In the final step of the experiment each group was treated differently. Group one‘was tested to determine whether they exhibited preference for any of the colored blocks, and its members showed a significant preference for the colored block bearing the same label as the rewarded 17 white block. Group “0, when tested for preference between black, yellow, and green objects which they had not yet soon, also exhibited a signifith tendency to choose the color which previously. had been labeled the same as the re- warded block. Group three, when presented with new white blocks of different shapes from the originals, but bearing the same labels, tended to choose the block bearing the re- warded label. A fourth group, presented with a hypotheti- cal situation in which they were to choose whether to in- vite a black, a yellow, or a green group of children to a party, tended significantly to prefer the group whose color had been associated with the rewarded label. Rewarded la- bels and colors were randomly distributed among members of each group. lisman states, 'The investigation was designed to test the general hypothesis that a color-preference response (positive attitude) could be developed through mediated gen- eralisation, and that this preference could be demonstrated in four situations differing in context, complexity, and social significance.‘ (p. 333) She concludes that the hy- pothesis is supported by her evidence. m FRESH“ STUDY The aim of the present study is mic-fold: l) to in- vestigate the existence of semant io generalisat ion along a semantic dimension which has not yet been explored, and 3) to shed some light upon the possible role of semantic generalization in mediating anxiety. These two aims will be discussed separately in the following paragraphs. MWW Previous studies of semant io generalisat ion have been concerned with semantic relationships of a culturally deter- mined nature, based on the comon denotative functions of the words involved. In contrast, the first aim of the pre- sent study is to determine whether relationships of a more idiosyncratic, connotat ive nature may also mediate genera- lisation. Osgood (34, 36, 37), has recently described a tech- nique, the Semantic Differential, for scaling the differ- ence in connotative meanings of words. Subjects are re- quired to rate words, concepts, or objects in terms of a series of paired polar adjectives, such as good-bad, large- small, strong-weak, etc. Comparison of the ratings of different concepts or words on such a series furnishes a 19' measure of their connotative difference.1 lore speci- fically, the objective of the present study is to test the following hypothesis. W I. A response which has been conditioned to a ver- bal stimulus should exhibit generalization to an. other verbal stimulus inversely to the degree of 22223231113%§‘§§§°32;.§2§Z°B§é‘3.33.§?’“’ “ lore, the Semantic Differential is seen as estab- lishing the degree to which I; (conditioned stimulus) and la, 83, . . . Sn (generalisation stimuli) may be func- tionally equivalent for eliciting r in the paradigm. 81 ‘will be conditioned to a.response (3)3 the Semantic Dif- ferential will be utilised to order 83. 33. . . . in) and the degree to which this ordering may be used to predict the order of Sz, 83, . . . an as elicitors of the condi- tioned response will then‘be determined. lrevious authors, with the exception of Eisen (5), andalazran (39), have been concernedywith the qmantifioa- ‘Zw 1Ordinarily, a low difference score is considered to imply similarity. However, since a low difference score be- tween two worde on the Semantic Differential may be the re- sult either of similarity of the two words or of the inap- plicability of the scales to the particular words in ques- tion, it is incorrect to mahe this assumption. Consequent- ly, the Semantic Differential will be referred to through- out this paper as a measure of difference, but not of . 11311“ it’s 20 tion of semantic generalization only to the extent of deter- mining the relative degree to which different semantic rela— tionships mediate generalisation. In the event that a rela- tionship is found to exist between the Semantic Differential and degree of generalization, it is foreseen that Semantic Differential ratings may be useful for predicting the amount of generalisation between many different kinds of stimuli. manila W 91. mm lhile few psychologists would hold that autonomic re- sponses, per se, are identical with emotions or anxiety as conceived in humans, numerous authors (1, 4, .33, 38, 49, 50, 52), offer justification for regardingelearning of these re- sponses as a paradigm for learning of emotional responses. A passage from Osgood contains the essence of this view: Stimulus-objects typically elicit a complex pattern of reactions from the organism, some of which are dependent upon the sensory presence of the object for their occurrence and others of which can occur with-out the object being present. . . . 'hen other stimuli occur in conjunction with the stimulus ob- ect, they tend to be conditioned to the total pat- ern of reactions elicited by the object; when la- ter presented without support of the stimulus-ob- ject, these stimuli elicit only the 'detachable' reactions. (36. p. 398) The studies reviewed in the first section of this pa- per have been largely concerned with semantic generalisa— tion of such 'detachable' autonomic responses. The evidence of the readiness with which these responses are conditioned 21 and generalized, together with the evidence (16, 29), that they are often extremely resistant to extinction, is sug- gestive of the role they may play in clinical manifesta- tions of anxiety. If an individual experiences frequent elicitation of a few such 'detachable' reactions, it is readily seen how he may appear in the clinic with subjec- tively inexplicable complaints of 'nervousness,' headache, or a variety of psychosomatic symptoms. At the same time, as a result of repression, imperfect discrimination, or other factors, he may be unable to specify, either to his therapist or to himself, the nature of the situations which elicit his symptoms. The second aim of this study is related to these con- siderations. Bollard and Killer (4), speak of psycho-ther- apy as a labeling process. Insofar as psycho-therapy deals with symptoms of the sort mentioned above, semantic genera- lisation, mediated by verbal labels, should play a part not only in the appearance of such symptoms, as outlined by Cofer and Foley (1), but also in their effective removal. Several studies have demonstrated learning and semantic ge- neralisation of autonomic anticipatory responses in situa- tions where many of the subjects were unaware of the nature of the stimuli to which they responded. The following par- agraphs will discuss these studies briefly, and the hy- pothesis deriving from them will then be presented. 32 - Diven (3) utilised the SSH, finger tremor, and differ- ential recall of words to show that learnigg and generaliza- tion may take place without the subject's awareness. He presented his subjects with a word list in which the recur- ring word, M was followed by an electric shock. In an alternate form of the list, the critical word was gagi, Both lists contained a series of rural words, such as plgg, m etc., and a series of urban words such as m, gtrggtggg, etc. Subjects were instructed to respond to each word by free—associating until told to stop. The length of free association varied from one group of sub— jects to another, and the shock which followed either m cr.taxiuwas delivered at the end of this interval. Diven reported generalisation not only to the words, 1299. and 12.1121: which always preceded m and 5.2.1.: re- spectively, but also to rural words in.the case of the piggy shook subjects, and to urban words for the tgxiyshock subjects. One of the most significant findings of Diven's study con- cerns the fact that, of his 52 subjects, 21 were unable to report at the conclusion of the experiment what word had served as a signal for the shock. Although examination of Diven's published data reveals that these unaware subjects may not here conditioned as well as the others, they appear to have generalised in the same way as did those who were an were of the pggprshock relationship. 83. A later study, modeled in part on Diven's was con- ducted by Haggard (lO).He used shock to condition the 089 to the word, gggggg which was always preceded by pagan, and embedded in a 43aword list. Since he was concerned with the effects of various situational factors upon the extinction of such a response, he did not test for seman- tic generalisation. However, utilising the same free associ- ation method as did Diven, with an interval of 10 to 12 se- conds, Haggard found that, of 18 subjects, only 9 were aware, at the and of conditioning, of the relationship between “3. mg, and shook. Two subjects reported that shock was preceded by a*war word, while 7 subjects had no aware— ness of the relationship. He reports that there was no sig- nificant difference between his aware and unaware subjects 'with respect to their general reaction to conditioning, al- though the aware subjects exhibited a greater 693 to gharp, and.ggg;d,than did the unaware subjects. Haggard states that ';.. . several Se in the Aware group mentioned . . . that they felt mild emotional distur- bances whenever they heard or thought of the stimulus words, ghgrp, or Iggrg, while the Unaware Se spoke rather of a gen— eralised.feeling of apprehension or anxiety.‘ (p.872) He hypothesizes that his unaware subjects were more disorga- nized by anxiety, and therefore less able to develop the necessary figure-ground relationship. He supports this by the observation that his unaware subjects reacted more strongly to the shock than did the aware subjects, and show- 24' ed less adaptation. Haggard evaluates his results as fol- lows: Is there any basis for assuming the occurrence of 'unconsoious emotions' in the a participating in this study? The answer rests on the definition of the term. If (a) we define 'unconscious emotion'l as a attern of observable reactions of an indivi— dual of which n; is unaware) to stimuli which are known to be associated with objects or events which do elicit 'conscious eiiiotion'l of which he a able to report, and if (b) the observable reactions (of which he is unaware) are not qualitatively differ- ent from those of individuals who, under similar circumstances do report having experienced affec- tive disturbances, then the present data indicates the presence of such I'unoonscious emctions.‘ (p. 375) A third study, by Lazarus and locleary (35), condi- tioned the cm to nonsense syllables. Ihen the syllables were subsequently presented tachistoscopioally at speeds which precluded correct identification, incorrectly per- ceived shook-syllables elicited a greater mean cm than did incorrectly perceived non-shock syllables. The authors con- clude: 'There seems little doubt that subjects can make au- ‘ tonomic discriminations when they are unable to report con- scious recognition.‘ (9. 118) J. I. Lacey and Smith (21), have reported a study si- milar to those of Diven and Haggard. They presented 31 sub- jects with a word list which contained 6 repetitions of the words, 991 and mg, S different ruralwords, and 11 non- rural words. Half of the subjects were shocked after 15 seconds free association to 991, while half received shock after mu. Gal, digital blood flow, and heart rate were 35 recorded, although the authors report only the heart rate da- ta. After one presentation of the word list, the subjects were interviewed in order to determine their degree of aware- ness or unawareness of the m-shock or mpg-shock relation- ship. Ot thefil subjects, 6 were aware of the stimulus rela- tions, 3 reported that a farm word had preceded shock, and 29 were unaware. Generalization was studied only for the 83 unaware subjects. The m—shock subjects developed significantly great- er responses (aocelerated heart rate). to rural than to non- rural words, while the “peg-shock subjects did not. During a single extinction presentation of the word list, both groups exhibited significant extinction to their respective critical words. Bacey and Smith conclude: . . If a wordpeign becomes the siglal for a pain- ful stimlus, without the subject being able to ver- balise this relationship, an anticipatory autonomic response will ensue. This unconsciously formed anx- iety reaction, moreover, will appear to other word- signs meaningfully related to the conditioned word. (p. 1051) They further conclude that the difference between 391-- shock and mgr-shock subjects in generalization to rural words may mean that such generalisation depends upon seman- tic relationships previously formed in the subject's life experiences, rather than in the immediate situation. They also state that there was some evidence that the generali- sation response was stronger and more reliable than the con- ditioned response. A second publication, by Lacey, Smith, and Green (23), compares the responses of these same 82 unaware sub- ‘ jects to those of so subjects who were informed at the be- ginning of the experiment which word would be the signal for shock. They found that, except for the initial conditioning trials, when the informed subjects responded much more strongly to the conditioned stimulus, the two groups were ap- proximately equal with respect to degree of conditioning. The unaware subjects appeared to generalize slightly more than did the aware subjects. The second aim of the present study, to investigate semantic generalisation of anxiety, stems directly from these studies. The fact that some individuals may exhibit condi- tioning and generalization of autonomic responses without forming a verbalizable concept of the relationship between the unconditioned and conditioned stimuli muggests a test- able hypothesis about the role of mediating factors in such behavior. To return to the paradigm of mediated generali- zation presented earlier, the degree to which Sa, 83, . . . Sn function as elicitors of'R is seen as dependent upon the degree to which they function in common with 91 as eli- citors of r. It would seem reasonable to argue, in the case of the aware and.umeware subjects in the above experi- ments, that r represents a more extensive complex for the 27 aware subjects, since it includes whatever response pat- terns may be assumed to underlie the formation of the ver- bal concept, I'wordeleadsmtomshcck." On this basis, the se- quence illustrated in figure 3(a) should lead to greater generalization than should that in Figure 2(b). (e) ‘ :2 ~ 3 ‘r a- . \\\ 1 \,—sn(csa) I . ’, .s-e r3(concept)-es I (b) . r as -—-)a(cm) Figure II. Paradigm for Semantic Generalization of Aware and Unaware Subjects That is, the aware subjects, through the mediation of their conceptual response to the situation, might be expec- ted to show greater generalisat ion to words related semanti- cally to S1 than should unaware subjects. This hypothesis is not in agreement with the results of Lacey, et a1. How- ever, it may be argued, since their aware group's advance 38 information that the critical word would signal shock im- plied that no other word would, the discrimination between SldandSSg, 83, . . . 8n was facilitated by the instructions. Thus the generalization of the aware group was minimized. The present study will utilize as aware subjects those who become able spontaneously to conceptualize the stimulus relationship in the course of conditioning. These subjects will be compared to the unaware subjects in order to test the following hypothesis: W3 Individuals who are able to verbalise the fact that a particular verbal stimulus always precedes the unconditioned.shock stimulus will tend to ex- hibit greater generalization to verbal stimuli bearing a semantic relationship to the conditioned stimulus than will subjects who are not able to verbalise this sign function of the conditioned stimulus. If it is assumed that the same mediating factors func- 'tion in the generalization of extinction effects, a similar hypothesis may be formulated for extinction. The following hypothesis will be tested: W 1.11 Following experimental extinction of a conditioned response to a verbal stimulus, the ability to verba- lize the relationship between the conditioned and un- conditioned stimuli will be accompanied by greater semantic generalization of the effects of extinctions in the form of a proportionally greater decrement of response to the generalization stimuli. Positive evidence with respect to Hypothesis III will furnish support of the hypothesis of Dollard and miller that 29; proper labeling of situations facilitates the formation of appropriate responses to them. It may be seen that, if evidence from the present study supports Hypothesis III, the acquisition of appropriate labels during psychotherapy may he expected to result in heightened generalization of more adequate responses to extra-therapeutic situations. METHOD The procedure of this study represents a modifica- tion of that of Diven. It differs in one major aspect from that of Diven, Haggard, or Lacey, et. al. In these studies, the generalization stimuli were included in the initial con— ditioning list of words. In order both to take advantage of the incubation indicated by Diven's results, and to study generalization in a situation slightly removed from the im- mediate conditioning situation, the generalization stimuli were included in a separate list, and presented only after the completion of conditioning. Warmers; The conditioned stimulus, giygg, was chosen from the lentARosanoff Iord Association Test (18). Inzorder to pro- vide a.rough continuum of similarity, part of the generali- zation stimuli were chosen from the Ient-Rosanoff frequency table of associations to giygz, Ten words were chosen which occurred more than once as associations to rizgr, and fire were chosen which occurred once only. The exact Kent- Rosanoff frequencies of these words are give in Appendix I. Ten more words, which did not appear in the Kent-Rosanoff table, were chosen at random form among the words in the Thorndike-Lorge word frequency table (53),‘which have a frequency of 50 per million or more. ill the generalization Si stimuli meet this criterion, affording a rough control for the effects of language frequency. Seven words, gm, 2:22;. strata. an. fined. later, and mi. all concep- tually related to 1.119;, were included in order to provide a basis for comparing the Aware and Unaware groups for de- gree of generalization. MWW In order to minimize, for the subjects, the connec- tion between the conditioning experiment and their ratings of the stimuli on the Semantic Differential, the ratings were carried out separately from the rest of the ex- periment. Two psychology classes were asked to rate m; and the SS generalization stimuli on 80 of Osgood's scales. Subsequently, volunteer subjects were recruited from these classes, without being told of the connection between the earlier ratings and the experiment in which they were now being asked to participate. The scales were chosen to represent equally Osgood's (34. 35), three semantic factors, Ivaluation, Potency, and ictivity. These scales and their factor loadings are given in Appendix 3. mm The subjects in this study were 31 male and 34 female students from an introductory course in psychology. Since they were assigned to the Aware and Unaware groups on the 32 basis of their ability to verbalise the nature of the 03-338 relationship at the end of the conditioning trials, it was impossible to match these groups in any respect. serum The unconditioned stimulus for the can was shock, supplied by a pair of 1% volt dry cell batteries attached to a Harvard inductorium which was set at a scale reading of four. The shock was delivered to the right forearm by a pair of electrodes, secured to a web belt, and fitted snugly just below the elbow. A pair of finger plungers was affixed to the end of the right arm of the upholstered chair in which the subject was seated. These were operated by the subject's right in- dex and ring fingers. The index-finger plunger was attached to a mercury switch which activated an electric timer. Slight movements of this plunger in any direction alternate- ly started and stopped the timer, resulting in rapid series of audible clicks. The galvanic skin response was recorded by . means of an electronic chnmeter, which was designed to provide a di- rect current of approximately two milliaaperes through the subject's skin. The instrument was equipped with a variable resistance switch, calibrated to yield a reading of the sub- jects basal resistance level. Skin resistance was read from a n.0, aisro-ammeter, which was calibrated through the en- 33 tire range of the instrument, from 10 thousand to 110 thousand ohms. The instrument was operated in series with a voltage stabilizing transformer, which provided a constant line vol- tage. Figure 3 contains a schematic diagram of the instru. ment's construction. Polarization of the zinc palmar electrodes was com- pensated by means of a reversing switch installed in the electrode circuit. The electrodes were mounted in a small block of plastic, which was fixed in the subject's palm by means of a heavy rubber band passed over the back of the hand. A small amount of electrode jelly was placed on the electrodes in order to insure good contact. mm The experiment was conducted in a.semi-soundproof‘win- dcwless room, which measured approximately six by nine feet. The subject was seated in an upholstered.chair, facing the closed door of the room. All apparatus was located on a tap ble directly behind the subject. The experimenter sat at this table in such a position as to be able to observe both the apparatus and the subject's face and right hand. 34 d8 mgmoog How 85950 oaqohpooao Ho Scams? passage oflpgomomttm .mfim K m K m ‘< mum 83 j 35 Then the subject was seated, before the electrodes for either the can or the shock apparatus were put in place, the following explanation , adapted from Diven's, was given: This is an experiment in psychology investiga- ting muscle coordination. It has nothing to do with intelligence or personality, only your a- bility to coordinate your fingers. In order to register these coordinations electrically it is necessary to put your hand in an electrical cir- ouit with the apparatus. You have probably heard of the galvanic skin response. That is what I am measuring. It is perfectly safe, so thereis nothing to be alarmed about. The current is only a ew housandths of an ampere, and you won't even be able to feel it. On your other arm, though, I want to put some electrodes which.will give you a slight shock occasionally during the course of the experiment. It will definitely be uncomfortable, but it is not dangerous, since the current comes only from these three volts, so it can't possibly hurt ycuu I wonder how you feel about it? Only one subject refused to undergo the procedure as presented above, although one other who went through the conditioning series refused to remain for the rest of the experiment, saying she had another appointment, and a third quit after receiving two or three shocks. After the subject‘s consent had been obtained, the SSS electrodes were placed in the palm of the left hand, the shock electrodes were placed on the right forearm, and the subject's fingers were placed on the plungers. The experimenter then took his place at the table, out of sight of the subject, and proceeded as follows: So youwwill know what to expect, let's make a 36 few trials with the shock now. First, just to see how sensitive you are to it, I will give you a very mild shock, which I really doubt that you will feel at all. At this point , the doorbell button which activated the shock circuit was pressed. This caused the circuit breaker of the Harvard inductorium to emit a buzzing sound. However, since the secondary coil was withdrawn from the primary, no shock was delivered. Although three subjects reported feel- ing a barely perceptible shock, the majority said that they felt nothing. The secondary coil was then set up to about one quarter of the first interval on the inductorium scale, and the following instructions were given, followed by a slight shock. low I am going to give you a slightly stronger one, in order to see how strongly you react to it. The secondary coil was then set up to the fourth inter- val on the scale, and, following the instructions below, shock was administered for approximately one second. I am going to set this up now to the strength of shock which I have been using with all my other subjects, and give you one trial with it. If you think it is too strong, I can out it down a little, but most of the others have been able to put up with it a few times. Only three or four subjects reported that the shock was more than they thought they could tolerate, and for these the coil was moved back about a half-interval. The next in- structions were as follows: low, in a little while I am going to call out a series of signal or stimulus words. As soon as you hear the word I speak, I want you to call back to me the first related word that comes to mind, at the same time giving a light, even pressure on the plunger with your ring finger. Only the ring finger is to be moved. Keep the index finger quiet on its plunger. Then you move the inger, you will hear a click. Try to keep from making any clicks by moving only your rin finger and keeping the index finger still. This is the coordination I am interested in. Keep right on giving me words and pressures, then, until I say stop. Don't try to make any sense out of the task. I don't care what the words are that you say. All I want is for you to keep on giving as words and pressures, without moving the wrong finger, until I say stop. For example if I say 93:13:, and the first word you think of is t on say and press with your ring inger. Then if the 232:3: 1221:?“3: é: ease-rater; stop, then wait for the next word. The subjects were then asked to find a comfortable po- sition, to try to relax, and to refrain from unnecessary movement as much as possible. After a short wait, in order to adjust the ohwueter and to allow it to stabilize, the basal level of the sub- ject's skin resistance was recorded, and Iord List I was presented. The same order was used for all subjects. The subject was allowed twelve seconds for free association be- fore the tel-bal signal to stop. Following the signal to .L 3The finger coordination task outlined in these in- struct ions was purposely more complex than that of pre- vious experimnters, in order to increase the proportion of unaware subjects. 3’ stop, an interval was allowed to elapse before presentation of the next word, in order to allow the 08! to subside. This interval varied from a thirty second minimum to two or three minutes, depending on individual differences in ' the time required for the skin resistance to stabilize. Iord List I one union cause red first lamp two circle circle river hang river five large bed better weather hat seven river plow bright circle above school kiss river college river cook river horse circle circle full oil After each presentation of the conditioned stimulus, riygg, the signal to stop was immediately preceded by a shock. A record was made of each subject's resistance le- vel at the time of presentation of the critical word, as well as of the lowest level reached during the first post- stisulus excursion. The difference between these two 1e- vels was then taken as the GER. The initialeescursicnhonly *msserecorded in order to avoid confounding the response to the stimulus word with responses to subsequent associations. 'here previous experimenters preceded the critical word with the same word on each presentation, each presen- tation of the critical word in the present study was prece- ded by a different word. This was done in order to furnish an internal control for conditioning. cm was recorded for each pre-river word in the same way as for {$131, One presentation of Iord List I (six presentations of 39 the conditioned stimulus) comprised the gonditigninglggrlgg, lash subject was now asked to tell what he had noticed dur- ing the first part of the experiment, in order to discover whether he could verbalise the riygzrshock relationship. The subject's spontaneous responses to this question were noted, and further questioning was postponed until the end of the experiment. At the end of approximately ten minutes' rest, during which the subject was allowed to walk around, smoke if he desired, get a drink of water, etc, he was once more seated in the armchair. The electrodes were again adjusted, and the subject was again requested to relax and to remain as motionless as practicable except for his finger movements. He was instructed, a laniven, 'To complete the word list, do just what you did before.‘ lord List II, the generalisation list, was now presen- ted, following the same procedure as with lord List I, except that the SSB‘was recorded for each‘word of List II. and no shock*was adminstered. This presentation of List II‘will be referred to below as the W 1911. In order to control for possible position effects upon.generaliza- tion, two different random arrangements of List II were uti- lized, each arrangement being presented to one—half of the subjects. The next step consisted of two extinction presentations Iord List 113 lord List IIb sugar mountain current brook flow flow body line tree lake demand island ocean sand brook body stream demand ice current dinner sky land table lake desert desert tree line land hunt ice sand dinner boat flood island stream mountain sugar fish fish table 'water flood hunt 'water ocean sky boat of Iord List I in which mg was not reinforced by check. This yielded an extinction series of twelve trials. These presentations differed.from those of the conditioning ser- ies only in that for economy of time the first four and last four words were omitted.from the final presentation. 683 was again recorded both for ;;1§;_and.fcr the word pre- ceding it on each trial. Immediately following the extinction series‘lord List II was once more presented.in order to test for generaliza- tion of extinction effects. The final presentation of List II will be referred to below as the gztipgtigputggt, Each 45 subject received the same form of List II as he had in the generalization test. At the conclusion of this test trial, the subject was once more questioned about the original conditioning series in order to determine more definitely whether he could ver- balise the xizgrrehock relationship. Subjects who had ver- balised this relationship clearly, following the condition- ing series, were not questioned further. The following questions were asked in order: Did any of the words in the list you heard be- fore the rest period seem to be any more import- ant than the others? Do you.remember any of the words better than the others? ‘lere any of the words more disturbing than the others? ‘las there any word which seemed to stand.out more than the others on any basis other than repetition? Did there seem to be any system to the shocks you.were getting? Then did.you expect the shock? A few of the subjects who appeared unaware of the rig? g-shock relationship were told at the conclusion, that shook had always followed rgygg, In each case, these subjects denied having known this. In no case did a subject who, on the basis of questioning at the end of the conditioning ser- ies, had appeared to be unasare of the stimulus relations reveal such knowledge in the later questioning. All subjects 43 who failed specifically to verbalise the relationship, rgzggrleads-to-shook, were assigned to the Unaware group, including the partially aware subjects. 92mm .222”: In order to furnish a basis for estimating the effects of the extinction series on generalization, as well as the pro-existing stimulus value of the generalization.wcrds, two small control groups were utilized. One of these, designa- ted the lxtincticn—ccntrcl group, was treated in exactly the same way as the experimental groups, except that the extinc- tion series was omitted. In order to preserve all possible unaware subjects for the experimental group, however, only the last six aware subjects were utilized as controls. Since these subjects received identical treatment with the experimental subjects throughout the conditioning series and the generalization test, their results were included with those of the Aware group in analyzing the results of this part of the experiment. The second control group, the Generalizationpccntrcls, were merely instructed that it was desired to obtain their reactions to some words. They were seated, given an expla- nation of the GER electrodes, and instructed to associate to the words in the same way as the experimental groups. The finger plungers and the shock electrodes were omitted. Iord List II was then presented once, and the subjects' re- “a sponses recorded in the same way as for the other groups. an: emits WM Of the total of 55 subjects employed in the main body of the study, 35 became aware of the riygnrshock relationp ship during the conditioning series, 15 failed to conceptua- lise this stimulus relationship, and 5 exhibited a vague realisation of it, making some such statement as, 'I got shocked every time I said “people' in response to 'river.‘ As a result of apparatus failure, and the subtraction of some of the experimental subjects to serve as controls. the net number of subjects employed in various parts of the experiment was somewhat smaller than indicated above. Ii- gure 4 indicates schematically the disposition of subjects in successive stages of the experiment. hotbed of. Aselxaia A problem in dealing with GSR data has been to obtain a unit of measurement whose distribution would fulfill the assumptions of conventional statistical techniques. Studies by Haggard (11, is, 13), and by c. L. Lacey and Siegel (23. 24), indicate that this problem is best set by transformation of resistance data into units of conductance, the reciprocal of resistance. The formula for this transformation is as follows: “a" fit? 45 .Total Subjects 55 AW are ‘Unaware Generalization - controls conditioning 35 30 Series (3 discarded) {(3 discarded) Generalisa- 33 l? 10 tion Test 362 l? Ixtinction (S discarded)i(5 discarded) Series 30 lxperiaental lxtincticn 13 Test 6 lxtinction Controls Figure 4. Schematic Diagram of Disposition of Ex- perimental and.0ontrol Subjects. 46 H1 represents the resistance in ohms immediately pre- ceding the stimulus, and Hg is the lowest level of resis- tance reached in the initial excursion of the GSR follow- ing presentation of the stimulus. Hultiplication by a fac- tor of one million was carried out in order to eliminate decimal places. The unit thus obtained is the pm. All GSR data on the following pages will be expressed in terms of this unit. Significance tests were made by means of the 1-test wherever the appropriate assumptions (28) of normality and equal variance could be met. There these assumptions were not appropriate, the lann-Thitney n-teet (27, 31), was utilized. One-tailed tests were used wherever it was de- sired to test either of the hypothesis, H z 0 or f1 2 1'3, and two-tailed tests were used where the direction of the differences was not of concern Ci % O or 1'1- Is). The form of the hypothesis being tested will be specified below in 9.011 OtIOs W The difference between can to 31m and sea to the word immediately preceding each presentation of 3.1121 was taken as the criterion of conditioning. Tach subject's average response to the six conditioning presentations of m was compared with his average response to pre-river, words, and the mean difference was computed. These data 47' may be found in Tables 3 and 4, Appendix 0. ' lhile the Aware group, according to this criterion, exhibited a significant amount of conditioning (t = 3.07). the Unaware group did not (t I-O.13). The difference be- tween the twc groups, as given by the yktest, was signifi- cant at the .05 level. The response levels of the Aware and Unaware groups are compared graphically in Figure 5. 7 Ewarezriver M S _____ Aware =pre-river g 5 , . ' Unawarezriver 1: 4 ._ e. —-e Unaware3pre-river if} 3 3 2 1 l a: 33 4 s; é: 7 * TRIALS figure 5. GSR to R and Pre-river‘Tords in the Aware and Unaware Groups ing the Conditioning Series That the Unaware group not only showed a lower degree of conditioning, but also showed less overall reactivity during the conditioning series is suggested by the fact that the mean GER to pre-river words alone was 1.01 microm- hos for the Unaware group, while that of the Aware group *was 1.98 micrcmhos. This difference was significant at the .02 level (t = 2.62). mm W and its. mm W . In order to test Hypothesis I, the pro—experimental Semantic Differential ratings which had been obtained for the contents of Tord List 11 were correlated with the gen- eralization test 083s obtained to the same words. To obtain a measure of the ccnnotative difference be- tween giygg,and each of the words in List II, the difference, 1, between the rating of 1.1.1.12. and that of each other word on each single scale was obtained. These differences were squared and summed separately for each generalisation sti- mulus, and the square root of the sumnwas taken. The re— sult is Osgood's (34, 35, 37), difference statistic,‘2, The formula is given by D I» Zd3. ‘ Since with both GER and n a marked tendency was ob- served for the ranges of individual subjects to very wide- ly, all datawwere transformed into standard scores, based upon the means and standard deviations of individual sub- jects. The correlation betwun mean GS! and mean Q by sub- jects, was - 0.15 (I > .10), indicating that 083 and 2 are are independent measures. The Pearson product-moment correlation for individual cans for each word, plotted against the corresponding Q_val- ‘ues,‘was found to be 0.0l. For an I of 1218, based upon the product of the number of subjects and the number of respon- ses for each subject, this correlation coefficient was not 49 significant. In order to guard against the possibility that this correlation had been attenuated by inclusion of the Un- aware group, who apparently failed to condition or to gener- alize, (see below), a second correlation was computed with an I of 799, based upon the responses of the Aware group a- lone. In this group, where significant generalisation was shown to have occurred, the correlation between can and 11 was 0.05. The probability of obtaining this value by chance was greater than .05. These results represent evidence a— gainst Hypothesis I. ‘ Since the generalization stimuli had originally been chosen in part on the basis of their frequency of associa- tion with m it was felt that it would be of interest to compute a correlation coefficient for the relationship between word-association frequency and cm. Tord-associa- tion data were available not only from the Kent-Rosanoff tables from which many of the words were taken, but also from a more recent Minnesota (48), sample of word associ- ations. In the present data, the correlation between word.- association frequency and 088 may be spuriously high, be- cause of the fact that most of the words having high asso- ciation frequencies to m are also water words. However, it was felt that such an analysis might still yield some information. Data were utilized only from the Aware group for this analysis, since they were the only group which had 50 shown definite generalization. Since the frequencies of the generalization stimuli did not form a continuous distribution (see Appendix A), although it was assumed that such a distribution underlies association frequency, bi-serial ;.was employed for this comparison. Bi—serial ;_for Kenteaoeanoff frequency and individual GER was 0.27,'withH equal to 799. This was highly significant (P < .001). For Minnesota frequency and GER, the bi-serial zywas only 0.08, H equal to 799. This value was also significant, although not at so high a level as before (P < .05, ) .01). mwummmm Average overall GER of the Aware, Unaware, and Genera— lization-control groups to the first presentation of Tord List II was first computed. These group averages are pre- sented in Table 5, Appendix 0. Comparisons between groups re- veal no significant differences in overall level of response to Tord List II in the generalization test. It is of interest, however, to note that the Unaware group mean is lower than that of either of the other groups. This is consistent with the apparently lower overall reactivity of this group in the conditioning series. In order to determine whether the experimental groups exhibited differential semantic generalization effects, a El comparison was next made for each group between the average response to water words and that to non-water words. For this purpose, the average response to the six nouns, m 12min stream. Ian. 11229.. and m was compared with the average response to the rest of List II. Although pug- m had originally been included as a water word, it was excluded from the analysis of results. This was done in consideration of its obviously close relation to the uncon- ditioned shock stimulus. Several subjects, especially in the Unaware group, gave unusually large Gas to m at the same time that they associated shock with it verbally. It was felt that such Gals were cverdetermined by this asso- ciation, and the word was consequently rejected. Tables 5, 7, and 8, Appendix 0, present the mean indi- vidual and group differences between water word and non-wap ter word responses. Only the Aware group showed significant- ly more generalization to water words than to non-water words. (t 8 3.73, P (.0005), while the Gontrol group showed a slightly greater, non-significant, response to non-water words. Table I shows that the Aware group difference was signifi- cantly greater than that of either of the other two groups. ‘ The significance of this difference between the Aware and Unaware groups supports Hypothesis II, although the failure of the Unaware group to condition to a significant degree or to exhibit significantly greater generalization to 53. TABLE 1 IITER-GROUP COHPIRIBON O? GENERALIZATIOIBTEBT DIFFERENCES BETWEEN TATE IORDS AID ION-IATER IORDB consumes our. tfitl t x3) 9 Aw are-Controls l. 11 3. 86 <. 005 Unaware-Controls 0.84- *‘ n.s.' . Aware-Unaware 0.27 *‘ (.01 ‘="'lbt significant ‘0 U-test ‘water‘words makes it doubtful whether these data adequately test the hypothesis. Wfimm In order to determine the general effect of the extinc- tion series, the difference between overall generalizationp test GER and overall extinction-test GER‘was first computed for each of the experimental groups and for the Extinction- control group. As may be seen.fromLTab1es 9, 10, and 11, Appendix C, both the Aware and Unaware groups showed a sig- nificant overall decrement of GER between the first and sec- ond presentations of Tord List II. The Control group did not decline significantly. Then intergroup comparisons were made of these declines, however, there were no significant differences. The apparent conflict in these results sug— gests that the reduction.in overall response level to the 53 extinction test was slight. In order to afford a test of Hypothesis III, the dif- ference in cm between the generalization test and the ex- tinction test was computed separately for water words and for non-water words for each subject. The individual sub- ject differences between these decrements were then aver- aged by groups. This yielded a mean difference for each group between the extinction test decrement to water words and that to non-water words. These means were tested for the significance of their differences from zero, and inter- group comparisons were made. lxamination of Tables la, 13, and 14, AppendixC, re- veals that the Aware group showed a significantly greater average extinction decrement to water than to non-water words, while the Unaware and Control groups did not. The Control group showed a slightly greater non-significant, average decrement to non-water words. The intergroup com- parisons in Table 2 show that this differsncevms signifi- cantly greater for the Aware group than for either the Un- aware or the Control group, while the difference between the Unaware and Control groups was not significant. Hypothesis III is supported by these data, although, as in the case of Hypothesis II, the apparent failure of the Unaware youp to condition, or to generalize the effects of conditioning or extinct ion, makes it doubtful whether a 55 TABLE 3 I‘le-GROU! COHPARIEOH or DIFFmENCES arrears EXTIHCTIOT— rrsr nrcmrn'rs ro um TORDE AND ro non-warm woans W cmmson Diff. tau 3 12); a Aware-Controls 1.41 3.22 <.oos Unaware-Controls 0.94 " n.s.‘ aware-Unaware 0.47 " .085 3"‘Ibt significant “ ULtest good test of the hypothesis has been afforded. IVALUATIOI’OF'RESULTS Probably the most important finding of this study is the failure of the Semantic Differential to correlate with semantic generalization. 0n the assumption that the Semen. tic Differential affords a reliable continuum of ccnnotative difference for a series of verbal stimuli, the present re- sults indicate that such a continuum bears no relationship to semantically mediated generalization. This, coupled with the finding of a possible relationship between semantic gen- eralisation andnwordpassociation frequency, suggests that the mediating relationships in semantic generalisation are those which are built into the individuals habit patterns by repeated linguistic usage. The individually unique, effec- tive similarity which is presumed to be measured by the Ee- mantic Differential appears to be of secondary importance, “any. The finding of a significant correlation between semanp tic generalization and word-association frequency data from two independent large samples appears to justify a more care- ful study of this relationship. In the present study, the words hating the highest wordpassociation frequencies to ggyggywere all water words, a relationship which is itself sufficient to mediate a high degree of generalization. The correlations found betweennword-association frequency and generalization may be entirely due to this fact. However, further investigation is required in order to determine whether this is the case. Hypothesss II and III are supported by comparison be- tween the Aware and Unaware groups with respect both to me- diated stimulus generalization and to generalization of exp tinction effects. However, as pointed out in the previous section of this paper, the failure of the Unaware group to exhibit significant conditioning or generalisation raises doubt that these two hypothesss were adequately tested in the present study. Reference to Tables 6—8 and 13-14, Appendix G, will reveal, however, that in both the generalization test and the extinction test all differences between groups are in the hypothesized direction. In both the generalization test and the extinction test, the mean difference between GER to water words and to nonmwater words is negative for the Control groups. For both the Aware and the Unaware groups these differences are positive, while in each case the flare group shows significantly more generalization than does the Unaware group. These regularities would seem to warrant further research along these lines, once the factors responsible for the failure of conditioning in the Unaware group have been clarified. The results of the present study are in sharp conflict with those of studies by Diven, Haggard, and Lacey, et al., 52 with respect to conditioning of the unaware subjects. lhile examination of Diven‘s published data suggests that his Una~ ware subjects may not have conditioned to the same extent as did his aware subjects, Haggard states specifically that no clear-cut difference occurred in his data. Lacey and.his collaborators report that their unaware subjects showed a lo- wer level of conditioning, but an equal degree of generaliza— tion, in comparison with their aware subjects. The failure of the unaware subjects to condition in the present study is doubtlessly attributable to procedural dif- ferences between this and other studies. Two such consis- tent differences exist. First, the distraction afforded by the finger plunger taskwwas more complex in this than in pre- vious studies. This may have interfered*with the rate of con- ditioning. However, a probably more significant difference lies in the fact that Diven, Haggard, and Lacey all preceded the critical word with the same word, at each presentation. This was not done in the present study. Instead, the criti- cal word, rgygg, was preceded by a different word at each presentation, in order that the pro-river words might serve as a control for conditioning. It is likely that this had the effect of decreasing the discriminability of the critical word for the unaware subjects. Another problem, common to all the studies in this area, is that of why approximately 40% of the subjects in an experi- ment of this type fail to become aware of the wordpshock re- 58 lation, while others volunteer the concept as early as the second conditioning trial. That this difference in concept formation is not due to intelligence differences is suggest- ed by the fact that the Aware and Unaware groups in the pre- sent study were found to have almost identical average col— lege entrance examination scores. A possible answer is suggested by an examination of some of the hypotheses put forward by the unaware and par- tially a... subjects. when asked, at the end of the ex- periment, what had determined the occurrence of shock, se- versl subjects appeared to have satisfied themselves with incorrect hypothesis, so that no further attempt was made to determine when the shock would occur. For example, Subject 24 volunteered after the fourth Ling-shock presentation that she was shocked every tine the experimenter said M. This hypothesis was repeated during later questioning. Another subject, 8. 53, although he failed to exhibit appreciable conditioning, responded quite differently to wa- ter words and non-water words in both the generalization and extinction tests. This subject stated that he thought he was receiving shock whenever his response to Lin; or m; was m. i'ypical Unaware group hypotheses were that the shock followed a regular time interval, that it occurred af- ter aparticular number of st imuli had been presented, that it occurred'. . . every time I paused,' or '. . . whenever you let me say 13 words before you said .192. ' 5,3 Nearly all subjects appeared to be attempting to formu- late hypothesis which would enable them to anticipate the shock. However, some subjects who initially struck on incor- rect or only partially correct explanations apparently were afforded insufficient opportunity, within six conditioning trials, to test the accuracy of these. That a lack of accurate conceptualization does not pre- clude generalization, however, is revealed by examination of the GER averages of Subjects 24, 53, and 56 in Tables 4, 7, and 13, Appendix C. lhile these subjects failed to show any marked evidence of conditioning to :izgr, all three appear to have generalised more to water words than to nonswater words. On the other hand, at least one aware subject, 8. 7, ‘who realised.readily that ziygz,led to shock, gate a greater average GER to pre-river words than to zizgg, in the condi- tioning series. The apparent complexity of the relationship between conditioning, generalisation, and formation of the. concept, 'tjzgz,leads to shock,‘ suggests the desirability of research into the relationship between the formation of anti- cipatory responses and the formation of verbal concepts. One further observation invites discussion. The most effective inter-stimulus interval for conditioning is usually found to be approximately. .5 second (36). This interval has also been found to hold for ordinary GER conditioning (30). However, strong conditioning repeatedly has been shown to co- 90 our in the present kind of situation when the inter-stimulus interval is as great as 30 seconds. That this conditioning is not dependent upon whether the subject forms a conscious set, the unconditioned stimulus to follow the conditioned stimulus, has been shown by studies already cited. Consider- ation of possible factors responsible for this unusual ex- tension of the inter-stimulus interval suggests that not on» ly may generalization be mediated by verbal processes, but also that these processes may mediate conditioning. Inves- tigation of this aspect of verbal mediation appears to be lacking in the literature. SUMMARY AND CONCLUSIONS Various non-voluntary responses, when conditioned either to verbal or to nonpverbal stimuli, exhibit generalization to other stimudi which are semantically related to the condition. ed stimulus. Several studies indicate that this semantically mediated generalization frequently occurs in the absence of awareness by the subject of the relationship between the con- ditioned and unconditioned stimuli. ‘ I In the present study, the galvanic skin response was conditioned to a verbal stimulus and tested for generaliza- tion to a number of semantically related stimuli. The gener- alization of extinction of this response was also investiga- ted. Forty-nine experimental subjects were employed, 38 of whomnwere'aware of the conditioned stimulus-unconditioned stimulus relationship. ‘The study was designed.to test three hypotheses. I Hypothesis I predicted that semantically mediated gener- aliaationuwould be negatively related to the degree of cone notative difference between the conditioned stimulus and the several generalisation stimuli, as established by Osgood's Semantic Differential. Hypothesis II predicted that the ap ‘ware subjects, as a result of the mediating function of their ability to verbalize the conditioned stimulus-unconditioned stimulus relationship,‘would exhibit greater generalisation SS than'would the unaware subjects. Hypothesis III predicted that, for the same reason, the aware subjects also would ex- hibit greater generalization of extinction effects than would the unaware subjects. The results clearly indicate the Hypothesis I is un- tenable. A non-hypothesized finding was that the degree of generalization was positively correlated.with the magnitude of wordpassociation frequency between the conditioned stimup Ins and the generalization stimuli. Hypotheses II and III are supported by the general trend of the data although it is doubtful that a decisive test of the hypotheses was af- forded by the present study. The study raised a problem for future investigation of the degree to which word association frequency may be uti- lized in predicting the degree of semantic generalization be- tween verbal stimuli. Problems were also raised.with respect to the relationship between conceptualization and condition. ing. The results appear to warrant the following conclusions: 1. no relationship exists between ccnnotative differences between words, as measured by the Semantic Differential, and semantic generalization to such words. 8. Semantic generalization may occur as a func- tion of word-association frequency, as mea- sured by the IsntARosanoff lord.Association Test. 3. Awareness of the signal-function of a con- ditioned stimulus may increase speed and strength of conditioning. 4. Awareness of the signal-function of a con- ditioned stimulus may increase the gener- alization of responses conditioned to such stimuli. 6.3.. lPPEHDIX l nur-aosnorr wonmassoc n'rIon rarqurnc IE8 3mm crmnxzmon suuuu mm later Stream Lake Flow Brook Boat Ocean Mountain Current Fish Island Body Tree Hunt Line Sugar flood Dinner Demand Ice Table Desert OOOOOOOOOOHHPHHUO 65 APPENDIX B FACTOR LOADINGB OF SEIANTIO DIFFERENTIAL SCAKIB 11. 12. 13. 14. 15. 16. 17. 16. 19. 20. (Evaluation) good-bad .88 large-small .08 beaut iful-ugly . 86 hard-soft -.48 sweet-sour .83 strongeweak .19 deep-shallow .27 sharp-dull .33 ferocious-peaceful -.69 heavy-light -.36 relaxedptense .55 hot-cold -.04 nice-awful .87 treble-bass .33 angular-rounded -.17 fragrant-foul .84 honest-dishonest .85 active-passive .14 fast-slew .01 rugged-delicate -.4S Factor Il Factor II (Potency) .05 .62 .09 .55 -.14 .62 .46 .07 .17. .62 .13 -.06 -.4? .08 -~04 .07 .04 .00 .60 Factor 111 (not iv ity) -.09 .34 .01 .16 -.09 .20 .14 .52 .41 -.11 -237 .46 .06 .45 -.u -.02 .59 .70 .26 APPENDIX C TABLES OF INDIVIDUAL AID mOUP‘f—‘REBPOIBIB WRIIG CONDITIONING SEES, GERALIZATIOI 7181', AID IHIDOTIOI TEST rash: 3 59 OOIPARIBOH BETWEEN AND FEE-Rm IORDS FOR E GOUP . lean river: lean non- Diff. Gm riter 088 5 f1 f2 11 - Is 2 a. 6? n 5 1. 00 0.68 0. 32 7 3.64 6.80 -3. 16 8 1.07 1. 23 -0.16 9 10 48 1. 01 -0. ‘7 ll 0. 71 1. 34 -0. 63 13 1. 42 1.19 0.23 14 2.14 1.04 1.10 '15 1.04 0.09 0.95 16 3.90 1. 30 2. 60 17 2.68 2. 45 0.21 19 2.54 0.90 1.64 20 4.93 0.92 4.01 22 ‘ 1.79 1.03 1.76 23 0.80 0.32 0.48 25 0.72 1.13 -0. 41 26 5. 61 - 0. 35 5. 26 31 2.94 2.12 03.82 32 3. 75 1.50 2.25 33' 3. 69 2.90 0'. 79 35 3.84 1.16 2.68 36 10. 43 5.27 5.16 37 16. 37 4. 3O 12. 07 39 4.14 2. 39 1. 75 ‘0 1.58 3.05 -Oe ‘7 42 6. 51 4.08 2. 43 45 2.25 2. 32 -0. 0? 4'3 3.92 3.17 0. 75 48 2.25 1.17 1.08 49 6.29 1.84 4.45 54 2. 25 2. 30- -0. 04 tCi-D‘O). e 3.07 r< .005 “5t cbtsIned use“ COMPARISON Blflflfl R AID FRI-RIVER IOHDS FOR UN GROUP lean river Hean pre- Diff. GSH river GE! 6 I], f3 11 - f3 3 0.87 1.16 -0. 29 4 1.01 1. 40 p -O. 39 6 0.70 O. 80 -0. 10 12 1. 16 1. 35 -0. 20 18, 0.74 0.74 0.00 21 0. 54 0.14 0.40 24 2. 02 0.33 1.69 28 0.91 0.77 0.14 38 0.96 0.94 0.02 41 0.46 1.73 -1. 27 44 1.53 1.10 0.43 so 0'. 66 o. 69 O. 07 51 2. 21 2. 56 -0. 35 52 2.12 1.31 0,81 55 0. 56 0. 92 -0. 36 56 2. 63 2. 31 0. 32 in: 0.03 gaeouu TABLE 5 71 AVERAGE 666 OF EXPERIMENTAL AND CONTROL GROUPS T0 FIRST PRE- SIITATIOI OF IORD LIST II (GENERALIZATION TEST) A'ARE GROUP URINARE GROUP GEIERALIZATIOH— ' C ITR L UP 8 Mean 631 S Hill cm 9 H5811 Ga 2 1.06 3 3.77 64 3.63 5 1.19 4 1.33 66 3.93 7 3.14 6 0.61 67 3.31 9 1.76 _ 18 1.92 as 0.51 11 1.35 31 0.73 70 3.30 13 3.59 34 1.69 71 3.36 14 1.59 38 4.33 73 1.90 15 0.79 36 0.66 73 1.30 16 3.40 41 1.14 74 3.84 17 3.39 44 3.30 75 1.96 19 3.29 50 3.74 30 1.13 51 1.34 33 1.03 53 1.81 23 0.69 53 5.13 35 1.36 55 0.41 36 3.54 56 3.38 31 3.39 33 3.78 33 3.31 35 1.01 36 3.62 37 5.65 39 3.93 40 3.75 43 4.06 43 3.93 45 1.50 47 1.53 48 3.45 49 0.80 54 6.07 T: 2.32 1:1.77 Y=2.45 o—z - 1.80 ,2 = 2.18 52 = 1.16 - TABLE 6 72 COMPARISON BETWEEN HATER IORDS AND NON-IA‘I'ER WORDS FOR AIAR! GROUP (GENERALIZATION TEST) lean water Difference lean non- QSR watfr GSR __ __ S 11 3 11 - 13 2 2. 15 0. 73 1. 43 5 1. 62 1. 03 0. 59 7 3. 66 2. 92 0. 74 8 1. 92 1. 71 0. 21 9 1. 71 1. 78 -0. O7 11 1. 64 1. 12 0. 52 13 5. 46 2.99 2. 47 14 2. 30 1. 26 1. 04 15 1. 74 0. 46 1. 28 16 3. 13 2. 10 1. 03 17 2. 40 2. 21 O. 19 19 2. 67 2. 10 O. 57 20 1. 60 0.94 0. 66 23 0e 33 Oe 81 .0e 43 25 1. 45 1. 25 O. 20 26? 5. 49 1. 52 3. 97 31 2. 53 2. 29 0. 24 32 2. 84 2. 66 0. 18 33 4. 25 2.89 1. 36 35 1. 43 O. 90 0. 53 36 4. 10 3. 37 O. 73 37 6.97 5. 20 1. 77 39 3. 44 2. 71 0. 73 4O 3. 18 2. 53 0. 65 42 4. 40 3. 19 1. 21 43 2. 45 4. 51 -2'. 06 45 1.73 1. 44 0.29 47 1. 88 1. 44 O. 44 48‘ 4. 18 1. 90 2. 28 - 49 0. 77 0. 79 -0. 02 m are: 6....91 019...... In :- 0. as ,2 = 1. cc “TQM-3.73 1.: < .0005 COHPARImH BETWEEN IATER IORDS AND NON-IATER 'ORDS FOR UEAIARI warm 7 GROUP (GENERALIZATIOI TEST) 73 #1 I' L lean water lean non- Difference can water 08% 8 11 Ta 31 - Is 3 2.93 4. 12 -1. 19 4 1. 22 1. 42 -0. 20 6 0. 72 0. 61 0. 11 13 3. 76 3. 78 -0. 02 18 1. 44 2. 03 -0. 59 31 0e 87 Ge 74 -0e 07 24 2. 72 1. 24 1. 48 28 4. 97 4. 05 0. 92 38 0. 66 0. 93 -0. 27 u 2. a 3. 25 -0. 5? 50 3. 09 2. 31 0. 78 51 1. 19 l. 33 -0. 14 53. 9. 55 3. 32‘ 6. 23 55 0. 07 0. 53 -0. 46 56 4. 36 2.88 1. 50 In II"O.42 5-2 e. 3.77 Hi"! 0)." 1.01 r > .05 74 TABLES conmlaon BE‘l'llEN um loans AID non-um IORDS ton ammun- IZATION—COHTROL GROUP (GENERALIZATION TEST) Keen Inter Kenn non- Difference 68 water GER 8 :1 I; :1 - Ia 64 4. 36 3. 25 1.11 66 3. 48 2. 62 0.86 67 1. 78 2. 19 -0. 41 68 2. 70 4.88 -2. 18 69 0.60 0. 43 ' 0.17 70 1. 32 2. 49 -1.17 72 1. 42 2.02 -0.60 73 1. 30» 1. 21 0.09 75 0.73 2. 24 -1. 51 In 3-0. 42 . 0'2 t 0.96 “I; M41.” P“) .10 . ‘75 rum: 9 conmataon amrn nu GENERALIZAT IOU-TEST am an an tx- rmcuon-ns'r 033 ton um: mom» W lean lean lxt incticn Generalizat ion Ext incticn teat teat can teat Gm decrement 8 11 13 T; - 13 3 1. 06" 1. 13 -0. o7 5 1.19 1. 25 '-0.06 7 3. 14 1. 76 1. 38 8 1. 76 1. 21 0. 55 9 1. 76 0. 71 1.05 11 10 35 O. 73 0e 52 13 3. 59 0. 77 2.82 14 1. 58 0. 50 1. 08 15 0. 79 O. 27 0.52 16 2. 40 1.86 0.54 17 2. 29 2.62 -0. 33 20 1.12 0. 59 0. 53 23 0. 69 0. 28 O. 41 25 1. 26 O. 60 0. 66 31 2. 39 3. 48 -1. O9 33 3. 21 1.23 1. 98 36 3.62 2.00 1.62 37 5. 65 3. 22 2. 43 40 2. 75 1. 06-1 1. 69 42 4. 06 0. 35 3. 71 TABLE 10 76 COMPARISON BETWEEN MEAN GENERALIZATION-TEST GSR AND MEAN £19 TINCTION-TEST GSR FOR UNAIARE GROUP lean general- __1_ Mean ext inc- F Extinction teat ization test tion teat decrement 0m , can 3 11 13 . 11‘. fig 3 3.77 1.91 1.86 4 1.33 0.66 0.67 6 0.61 0.16 0.45 12 3.84 3.66 0.18 21 0.73 1.01 -0.28 24 1069 O. 35 1. 34 88 4’. 23 ‘. 28 -05 06‘ A 38 0.86 0.89 -0.03 50-' a. 7‘ 1. 68 1. 06 52. 1.81 1.33 .0.01 53 5.13 10 36‘ 30 7? 56: 3.28 1. 79 1.49 i}, n- 0.87 0'3 a 1.32 an i 0) .. 3.03 ,.. < .005 TABLE 11 OOIPARIml RETIRE! HEAR @NERALIZATIOI—TEST Gm AID “Al EX- TINCTION-TEST GER FOR EXTINCTIOI—COHTROL GROUP M ‘ lean genera1- lean extinc- Extinction teat irat 133““ t ion teat cm decrement a ll Ta !1 - Ta 43 3. 93 2. 85 O. 98 ‘5 1. 5° 1. 65 -0. 15 47 1 . 52 2. 79 .1. 27.» 48. 2. 45 3. 49 -1. 04 49 O. 80 0‘. 79 O. 01 54 6'. 07. 3. 37' 2'. 70 “3360). 1.0.81 r > .05 77 7191.: 12 COIPARIBOI BITWEEN EXTINCTION—T33! DECREMENTS TO WATER FORDS AND IOHATE'R IORDS FOR AWARE GROUP ' __:-: l! Keen 0901-9919111: Diff. “between to non-water words water and non- teter decrements Keen De0rement to veto: word- ! Du Du Du - Bu 2 0.71 -o.15 0.95 5 O. 32 -0.” 0. 4'1 7 2.97 1.22 1.45 9 0.79 0.51 0.37 - 9 0.72 1.20. -0. 27 11 0.95 0.24 0. 21 12 4.90 2. 09 2. 92 14 2. 02 0.59 1.27 15 1.20 0.24 1.02 20 0. 79 0.29 0.29 22 -o.12 0.50 -0. 72 25 1.45 0. 42 1.02 21 -0. 90 -1.17 0.27 22 2.59 1.42 2.44 29 2.52 1.19 1.24 27 2.29 2.20 1.19 40 1.95 1.71 0.24 42 2.59 2.99 0.90 1"- 0.92 .12 a. 0.92 “fl 0)» - 2.92. P < .0005 ‘ 79 I 7121.: 15 COHPARIBOH WEED EXTINCTION-TEST DECRHENTS 1'0 HATER WORDS AND HON-IA!“ IORDS FOR UNA'ARE GROUP Dirt. between lean Decrements to Inter words to non-Inter words water and non- weter decr emente Keen Deoremen’a 8 0: D5 Du - Bu 2 1.59 2. 05 .0.47 4 0. 47 0.94 .0.27 5- 0.59 0.45 0.14 12 0.05 0.21 -0.29 21 -0.71 .0.14 -0. 57 24 1.94 1.05 0.79 29 -1.24 1.99 .2 90 50 1.91 0. 52 1.09 53 10‘s -0.03 10“ 52 9.59 2.99 5.59 59 2.90 1.00 1.90 Yu- 0. 25 g-z .. 4.22 “Tim-some r > .10 79 TABLE 14 COHPARIHDH BEHEEI EXT INCTION—TEST DECREMENTB TO WAT!!! WORDS AID NOI—‘ATER IORDB FOR “TING? ION—CONTROL GROUP W Keen Deoremant Keen Decrement Diff. between to water worde to non-water words water and non- wetar decrements a n. 9:1 n. - Do 43 O. 35 1. 35 ul. 10. 47 -1. 60 -1‘. 15 -O. 45 48. -1. 68 - . 58’ -1.07 49 O. 03 O. 06 .0e 04 54 8. 96 8. 73 O. 39 'x' ._ 0.59 a-a :5 0. 25 HfLo) a 0.22 r > .10 1. 10. 11. 80 BIBLIOGRIBBY cher, 0. 1L, and Foley, J. P., Kediated generalization and the interpretation of verbal behavior: I. Prelego- mena. 23123391. 531., 1943, 49, 513-540. _________, Janis, K. 0., and Howell, K. K., Kediated generalisation and the interpretation of verbal behea- vior: III. Experimental study of antonym gradients. 1. .21. 2119201., 1943, 32, 266-269. Diven, L, Certain determinants in the conditioning of anxiety reactions. 1. Pexchgl. , 1937, 32, 891-308. Bollard. J.. and liner. 5. 1.. 25112051111. mm:- therapy. ll. 7.: KcGrav-Hill, 1950. lieen, N. 8., The influence of set on semantic general- ization. l. m. mg. 2119391., 1954, 49, 491-488. Iisman, B. 8. , lttitude formation: The development of a color preference through mediated generalisation. i. seem. 211.929.]... 1955, 50, 221.22 Foley, J. 9., and Gofer, c. 1., Kediated generalization and the interpretation of verbal behavior: 11. Bxperi- mental study of certain homophcne and synonym gradients. 1. :12. 2119231., 1943, 32, 168-176. 3 and M111“. 8. In} Kediated generaliza- tion ancf the interpretation of verbal behavior: 7. 'Free association” as related to differences in professional training. 1. m. m” 1943, 33, 299—310. . and Mathews, K. 1., Kediated generalization and the interpretation of verbal behavior: 17. Experi- mental study of the development of interlinguistic sync- nym gradients. ,1. m. Zuchgl,, 1943, 33, 188-300. Bag d, E. 1. Experimental studies in affective processes: I. 53:. effects of cognitive structure and active partici- pation on certain autonomic reactions during and following experimentally induced stress. ,1. m. MRI.” 1943, 33, 8579884. , lxperimental studies in affective pro- cesses: 1!. 0n the quantification and evaluation of 'meaeured‘ changes in skin resistance. .2, 3x2. szchgl., 1945, 35 46-56. 12. 13. 14. 15. 16. . 17- 18. 19. 8°. 21. 81 , 0n the application of analysis of vari— ance—to GER data: I. The selection of an appropriate measure. g. m. Pughglu 1949, 39, 378-391. r , 0n the application of analysis of vari- ance to 088 data: 11. Some effects of the use of in- agproggiate measures. 1. exp. Psycho)", 1949, 39, 8 1-8 s . and Garner, I. 3., An empirical test of a derived measure of changes in skin resistance. :1. 9.32- 1419021.. 1943. 36, 59-70. Hudgins, C. 7. , Conditioning and voluntary control of 3119 3321331119” light reflex. l. m. 23m” 1933, a e Kapustnik, 0. P.. (The interelation between direct cone ditioned stimuli and their verbal symbols.) Tpudz ab rat rii tiziglggii zygghgyflgggyggy, t n s ngyELQ—ngg-L'fi W Pgdagogighgsgkgflig tiftfiut Qgrtzgng, 1930, 2, 11-28. Paychgl. b§§;., 1934, 8. 153 18. Keller, K., Kediated generalization:‘ The generaliza— tion of a conditioned galvanic skin response established th: fictured object. “:1. 1. P5101191... 1943, 56, 4 4 . Kent, 0. 1., and Rosanoff, A. J., A study of associap tion in insanity. .4291. i. 13933., 1910, 87, 37-96, 317-390. llsc see Rosanoff, A. J., M1 91 m- 5112. l. 7.: liley and Bans, 1938. Iotliarevsky, L. 1., (The formation of pupillary condi- tioned reflexes and of a differentiation in response to both direct and verbal stimuli.) ABLE: W 313% 5 uk 1925, 29 (2), 477-499. 49].. hm” 1 27, ll 8 , 13724) 1483. . (Cardio-vascular conditioned re- flexes to direct and to verbal stimuli.) l' 21 . Q- 2. a. a. 1929 20 229-242. 29193191. b . 1923'; 12, (4049) 111.’ ’ Lacey, J. 1., and Smith, R. L., Conditioning and gen- eralization of unconscious anxiety. $33394, 1954, 120, 1045-1052. 32. 33. 34. 35. 35. 37. 31. 33. 33. 34. 35. 92 , and Green, 1., Use of con- ditioned autonomic responses in the study of anxiety. W. Med... 1955, 17, 308-317. Lacey, 0. L., An analysis of the appropriate unit for use in the measurement of level of galvanic skin resis- tance. g. m. Psychgln 1947, 37, 449-457. . and Biegel, P. 8., in analysis of the unit of measurement of the galvanic skin response. .1. 3.32. 21101191. , 1949, 39, 133-137. Lazarus, R. 8., and KcCleary, R. 1., Autonomic discrimina- tion without awareness: A study of subcepticn. KEY—9.9.9.1.- m., 1951, 58, 113-182. Kgldtsmsn, 1., and Brooks, L. 0., A iailure it; find sec- 0 or er semantic generalization. . m. zchgl. 1956, 51, 413-417. ’ Kann, H. 8., and Ihitney, D. 12., On a test of whether one of two random variables is stochastically larger than - the other. an. math. my,” 1947, 19, 50-90. Kclemar Q. Wfitatistigg. H. 7.: .1197 and 8011; , 1949. . Kenzies, R. , Conditioned vasomotor responses in human subjects. 1. Psychgln 1937, 4, 75-130. Kceller, 0., The 08 - 008 interval in cm conditioning. ,1. 9.32. 21193331., 1954, 48, 169-166. Kcses, L. 1., Non-parametric statistics for psychological research. 21121311. 11311... 1958, 49, 189-143. Kowrer, 0. 8., l stimulus-response analysis of anxiety . and its role as a reinforcing agent. mu. £51... 1939, 45, 553-565. Kyasishechev, 7., (The psychcgalvanic reflex and its sig- nificance in the study of personality.) £91. - £11101. 311,1... 1929, 3‘, 333-357. 21192.01. b951,, 1930, 4, 4304 4. Osgood, C. 5., Report on development and application of the semantic differential. Unpublished paper. The nature and measurement of meaning. m 2211-. 1953. 49. 197.237. 36. 37. 45. 43. 4?. 83 Osgood. c. r.. I! h asiihamiaknazimamm- m. U. 1.: xford Univ. Press, 1953 . and Susi, G. J., A measure of relation. ditermined by both mean difference and profile infor- mation. Pugh“. m” 1952, 49, 251-262. Backle‘y, L. E. ,1 Thetlialood pressure ind galvagic reflex as n ca ore 0 cm onal states. . an]... M. mw,u,m%ma , ’ Rasran, G., Salivating and thinking in different langu- ages. 1. 2119321., 1935, 1, 145-151. A qualtitative study of meaning by a condi- tfoned sajfivary technique (semantic conditioning). m 1939, 90, 89-90. . A simple technique for controlling subjective attitudes in salivary conditioning of adult human sub- Jects. m 1939, 89, 160-161. Attitudinal determinants of conditioning and c? neraliz ation of conditioning. l. :51. 2119991- 194 , as, sac-ass. ' . Bemantic and phonetographic generalizations 07 salivary conditioning to verbal stimuli. w. Pughglu 1949, 39, 642-652. Bose psychological factors in the generalisa- tio; c? salivary conditioning to verbal stimuli. M. . Sentential and propositional generalizations Ersalivary conditioning to verbal stimuli. i 19‘9. 109’ ‘47.4G0 Rises, B. 7., Semantic conditioning involving the gal- vanic skin reflex. 1. 131. m, 1940, 26, 238-240. :93 . Genetic changes in smantic conditioning. rm. 2112121.. 1946. as, 143-153. Russell, I. 4., and Jenkins, J. J., Mammal; m: The complete Minnesota norms for responses to 100 words from the Isnt-Rosanoff lord Associ- ation Test, Technical Report lo. 11. Univ. of llinn., August, 1954. 49. 50. 51. 58. 53. 54. 55. 84 Shaffer L. F. The problem of psychotherapy. £23.- W i947. 2. 459-467. mew, l'. J., A stimulus-response analysis of repression ggd 3:23.“ in psychotherapy. 21111131,. 3 V . 1943. , . imolenskaya, I. P., (Verbal symbols of conditioned and differential stimuli.) _l_a_ t ; Izuch. vxsshygh r 111 so .a . 1934. 3 4.315. W'. m. "3L1 35.—£11139, 1 s) 31: ' ' 8togdill R. l. Neurosis as learned behavior. 27.121.191.- m, 1934. 413497.507. Thorndike I. L. and Lcrge 1. Th; 131929.?! l M g; gém I. . H. Ti: éureau 0 Publications, Teachers ollege, olumbia Univ. , 1944. Traugott, l. 8., (The interrelations of immediate and symbolic projections in the process of the formation of conditioned inhibition. l Putxggg L_£§2§Q, zxgggzgg, 87%0303. 41.9.1191- £2911. . r di . 3gb. 1934 135%, 9. 1166) 131: ' lylie, R., Generalization of semantic conditioning of the galvanic skin response. Unpublished I. A. thesis, Univ. of Pittsburgh, 1940. 15 3175 0 III KIWI?! llJlllWll