from The Teacher's Wvrd Book vI30.0UO Homophones as ambiguous stimuli in research: Words (Thorndike & Lorge, 1944). From these basic pools. the 36 homophone and A comparison of the associative meaning 36 nonhomophone stimuli werc selcctcd 10 reflect various combinations of frequcllcy of homophones and nonhomophones of occurrence in the language and of major grammatical dass. This was done to control for possible effects of these variables on RUSSELL FOOTE* responses to both homophones and University of Tennessee, Knoxville, Tenn. 37916 nonhomophone stimuli. The Thorndike-Lorge frequency count Thirty-six homophones and 36 nonhomophones were presented randomly to 30 Ss in a (Thorndike & Lorge, 1944) was uscd to single response, free association task. The greater dispersion of the response distributions determine frequency for all stimulus 10 homophones than to nonhomophone stimuli as weH as the longer response latencies to words. Although these frequencies are homophones suggested thaI the associative meaning of homophones is more ambiguous derived from written rather than from than other elements of the vocabulary and, consequcntly, that homophone units may be spoken discourse, they were used in lieu of fruitfully employed in research necessitating the usc of such ambiguous stimuli. Several any of the few sources of spoken word such potential uses were discussed along with possible research aimed towards c1arifying frequencies (i.e., Howes, 1966) because the the phenomenon ofhomophones per se, apart f,om their use as research tools. latter sources provide gene rally inadequate frequency da ta for homophone words, What are commonly referred to as any differently to homophones than to especially those of lower frequency. homonyms or-more accurately­ other units of the vocabulary having Grammatical (form) dass for all stimulus homophones, for some time have m u I t i pIe me anings. Consequen tly, a words was determined by unanimous interested historical linguists concemed research question is posed which must be agreement of three independent judges-all with sound change and with the effects of answered if homophones are to be used undergraduate students-that a particular the semantic environment on meaning fruitfully in research as described above. stimulus word belonged primarily in one (Lehman, 1962). These pairs (or, more The current experiment is addressed to form dass. Only when all three judges rarely, triplets) of words whose individual this question. The meaning of both agreed was a word designated as a member members are pronounced the same but homophone and nonhomophone stimuli of a particular form dass and induded in differ both in spelling and meaning form a employed in this study was defined in the stimulus list. There was no difference relatively sm all subset of the vocabulary of terms of several measures of Ss' responses in the degree of judges' agreement on American English and are, in fact, to these words as they were employed as homophone and nonhomophone words. nonexistent in some other Indo-European stimuli in a single-response, word F 0 llowing the above outlined languages. Due to their special association task. That is, the homophones procedures, 36 nonhomophone items were phonemic-semantic properties, and nonhomophones used in the study obtained such that 12 were of AA (high) homophones have attracted the attention were compared on several indices of frequency, 12 were of A (medium) of a number of psychological researchers as associative me aning (Deese, 1965). frequ ency, and 12 were of low convenient stimulus items for use in studies Further, the study was aimed at correcting frequency-with each of the frequency dealing spe cifically with linguistic a possible methodological difficulty in the groups containing four nouns, four verbs, ambiguity or, more gene rally, with research Galbraith and Taschman investigation of and four adjectives. The 36 homophone requiring ambiguous verbal items for the relative strength of the separate words items were selected so that the six possible stimulus material. within homophone units. In the free combinations of the three frequency Galbraith & Taschman (1969), in association task used by Galbraith and groups (high, medium, low) with two research generally concerned with Taschman only homophones were used as component units of each homophone pair establishing normative data, have pointed stimuli and consequently, it was uncertain had six individual homophone items out some of these previous and poten haI whether or not a set for homophones may representing them. Thus, there were six research uses of homophones. However, have been produced in Ss, creating possible homophone items where one sense of the whether the focal point of the contarninating effects on their associative homophone was of high frequency and the accomplished or anticipated research is responses. other sense also of high frequency, six perceptual encoding-decoding processes, or SUBJECTS items where one sense was of high phonetic, semantic, subject, or contextual The Ss were 30 students enrolled in frequency and the other sense of medium \::riable effects. the assumption (often introductory psychology courses at the frequency, and so on. Ideally, each of these implicit) is that homophones are more University of Tennessee. All Ss were frequency groups was to have consisted of ambiguous than nonhomophone words. volunteers. The median age was 19 and equal distributions of nouns, verbs, and Such an assumption has not, in fact, been both sexes were about eQually represented. adjectives. However, the lirnited set of supported empirically. As Galbraith & STIMULI homophones in the language precIuded all Taschman point out in their artide, words A 72-word list composed of 36 possible combinations of form cIasses and with multiple meanings are common in homophones and 36 nonhomophones was frequency. However, each of the major English, possibly the rule rather than the used. Homophones were defined as pairs of form cIasses used in this study was well exception. In terms of the prevalent indices words (or, infrequently, as triplets: represented among the 36 homophones. In of word meaning employed by where-wear-ware) that were identical addition, one sense of some homophones psychologists-such as semantic differential phonemically but different in both spelling occasionally belonged to a form cIass other ratings and associative norms-there is as and meaning. An init.ial pool of 200 than the major three used in the yet no general evidence that Ss respond homophones was selected randomly from a experiment. Dictionary 01 American Homophones and Those homophones that incIuded past *Now at the Univcrsity ofWyoming. laramic. Homographs (Whitford, 1966) and 200 verb lenses or contractions as one or two Wyo.82070. nonhomophones were similarly selected of their different sen ses were generally

Psychon. Sci., 1970, Vol. 21 (4) 249 Table 1 Thorndike·Lorge Frequencies and Primary Form Class for Homophones and ~onhomophones and !\"umber 01' Associations 10 Different Homophone Senses Homophones :\onho11l0phonc,

T·L Primary :\umbcr of Homophone l"nits Frequeneks Form Cl;ssa Assodations Primary T·L Forn1 .-\ B A B A B A B \\·ord Frequencies Class" l. meat-meet* AA AA N V 24 4 l. la\\' AA N 2. wear-where* AA AA V 0 27 0 2. voice AA N 3. son-sun* AA AA N N 7 22 3. ball AA N 4. scene-seen AA AA N V 15 10 4. heaven AA N 5. write·right AA AA V Aj 9 20 5. red AA Aj 6. whcle·hole* AA AA Aj N 4 18 6. old AA Aj 7. week·weak* AA A N Aj 4 26 7. heavy AA Aj 8. plain·plane* AA A Aj N 6 22 8. deop AA Ai 9. sail·sale AA A N N 17 11 9. open AA V 10. same-sum AA A Ai N 20 10 10. consider AA V I!. be-bee* AA A V N 4 20 11. travel AA V 12. bear·bare AA A N Ai 16 11 12. build AA V 13. him·hymn* AA other N N 21 7 13. anger A N 14. wastc·\\'aist* AA other N N 23 4 14. dawn A N 15. beat·beet* AA other V N 28 0 15. religion A N 16. break·brake* AA other \" N 29 0 16. net A N 17. dear·deer* AA other Ai N 5 23 17. slaw A Ai 18. so-sew* AA other 0 V 2 27 18. wet A Ai 19. tail-tale* A A N N 21 4 19. loud A Aj 20. principle-principal A A N N 7 20 20. sad A Ai 21. warn-warn A A Ai V 11 16 21. hit A V 12. steal-steel* A A V N 22 6 22. jump A V 23. stare·stair* A A V N 27 2 23. glow A V 24. wheel·\\'e'll A A N 0 27 0 24. hide A V 25. heel·heal A other N V 18 9 25. candy other N 16. birth·berth* A other N N 28 0 26. insect other N 17. aunt-ant* A other 1\ N 3 26 27. divorce other N 28. mail-male A other N N 10 16 28. snai! other N 29. bore-boar A other V N 17 7 29. sour other Ai 30. load-lode' A other N N 25 0 30. somber other Ai 31. altar-alter other other N V 14 15 3l. swift other Ai 32. bridal-bridle other otlwr Aj N 16 12 32. innocent other Ai 33. dual-duel other other Ai N 8 20 33. chase other V 34. mantle-mantel* other other N N 4 23 34, rob other V 35. loan-lone other other N Ai 21 8 35. starve other V 36. foul·fowl other other Aj N 13 15 36. decorate other V

os = 'ZOll/l. V = l'erb, Aj = adjectil'e, 0 = other */ndicates signi[icant difference in proportions of responses receil'ed, p < .05. a voided. Similarly, homophones were were asked to respond with one word, the columns under the heading Number of avoided in which one or both senses were first word which came to mind, as rapidly Associations in Table 1 present the number obscure words or proper nouns. The pair as possible. Responses were made by Ss of times that the judges agreed wheel·we'll, however, was incJuded as an into a microphone which was visible to Ss unanimously that S's response to a exception. Also, despite preexperimental and which was attached to a Wollensak homophone belonged to one or the other proofreading of the stimulus list by the stereophonie tape recorder. After Ss left of the homophone's two major senses. author and at least six other people, a the experimental room, the recorded When there was not unanimous agreement, homophone, red-read, appeared among responses were used to actuate the pen of a when a response was uninterpretable, or those designated as nonhomophone items. single-channel Techni-rite heat writing when there was no response by S, no Since all Ss' responses to this stimulus were oscillograph (Model 4) with chart paper assignment was made. Of 1.080 (30 X 36) related to the color sense and none to the moving at the rate of 10 mm/minute. After possible responses to homophones, only 38 sense of the verb's past tense, and since the the recording had been made on chart could not be assigned to one or the other responses indicated no temporal pattern paper, the tape was replayed to coordinate major senses of a homophone because of distinct from other nonhomophones, it was each deflection of the oscillograph pen failure to achieve unanimous agreement feIt that the item could be effectively with an associative response by S. among judges. As an example, Iable 1 regarded as a nonhomophone stimulus. Consequently, a temporal as weIl as a indieates that 24 Ss were judged to have Table I presents the 72 .stimulus items verbal re cord of each S's response was given associative responses to the meat and includes information regarding obtained. sense of the first homophone pair shown in frequency and form c1ass as weIl as RESULTS AND DISCUSSION the table, 4 Ss were judged to have information to be discussed in the results Three independent judges (not the same responded to the meet sense, leaving two below. individuaIs involved in fonn cJass responses not assigned by the judges to PROCEDURE judgments described above) assigned Ss' either sense of the homophone. All Ss were presented the 72 stimulus responses to homophones to one of a Of the 23 homophone pairs used in both items-in a different random order for each homophone's two major senses or, if the present experiment and the work S-in a single-response free-association task, conditions to be described below prevailed, reported by Galbraith & Iaschman (1969), Words were presented aurally to Ss who to neither sense. The numbers in the two inspection of the proportion of

250 Psychon. Sei., 1970, VoI. 21 (4) frcc-associativc responses givcn tu each did nonhomophones, this result reached rurther experimental investigation. It is far sense or these pairs revcals that only in thc significance only at p< .20, x2 (1) = 3.75. from clear why one semantic sensc of a c:asc of sccllc-scen are thc proportions Neverthelcss. this result, taken togcther homophone should overshadow thc olhcr obtained in the present research (.60 and with thc rcsult concerning differellces in S's immediate responding to these AO. rcspcctively) opposite to that obtained between homophones and nonhomophones stimuli. Possibly, an experimental task in the earlier experiment (.43 and .53, on the percentage of the primary response, which allows Ss to givc continuous respectivcly). Sincc neithel 01' these seems to indicate adefinite tendency in thc association to homophones may clarify thc proportions differs significantly [rom r esponsc distributions to homophone relationship between elements of .50-.50, however, this diffcrcnce may be stimuli toward more dispersion than in thc homophone pairs and help to explain why attributed to chance. In the othcr 22 pairs, response distributions to nonhomophones. one sense should consistently dominate the whcre the direction of the proportions are The evidence for differences in the other when a homophone is encountered thc same in both experiments, the degree of ambiguity of the associative out of linguistic context. Certainly, magnitudes of the differences in meaning between homophones and semantic contextual effects are the most proportions between the present study and nonhomophones is furthered by Jikely source of disambiguity of the rcsults of Galbraith and Taschman may consideration of the marked difference in homophones in day-to-day speech. likewise be attributed 10 chance or to Ss' response latencies to homophones and Nevertheless, as the wide occurrence of consistent differences in experimental nonhomophones--where response latency linguistic puns will attest, the ambiguity of procedure (Le., use of homophone stimuli was defincd as the time between offset of homophones can be comfortably exclusively vs a mixed the stimulus word and onset of S's entertained and even courted in speech. homophone-nonhomophone list or response When individual Ss' mean Thc question arises that un der wh at differences in judgments in assigning Ss' response time to the 36 homophones was conditions does a lang1.lage community responses to one or the other sense of the compared to their mean response time to tolerate the existence of a small subset of mutually used homophone stimuli). In the 36 nonhomophones, it was found that its total vocabulary which allows the general, however, the results are quite 26 of the 30 Ss took longer, on the possibility of such ambiguity? An comparabJe. average, to respond to homophones. A interesting suggestion offered by some To investigate the hypothesis that two-tailed sign test showed this difference historical linguists and dialect geographers homophones, when used as stimuli in a to be significant, p< .00]. The overall (Lehman, 1962) is that such members are word-association task, tend to elicit mean response latency to homophones was t olerated in the most widely used patterns of responses indicative of more 1.85 sec and for nonhomophones was vocabulary of a language community at a stimulus ambiguity than do 1.52 sec; the SDs were .11 and .10, particular time to the extent that their nonhomophones, analyses were made respectively. For those four Ss whose mean different semantic senses are part of comparing the response distributions to response latency was higher for sharply distinct semantic environments: A homophones and nonhomophones used in nonhomophones, the magnitude of the homophone whose two senses refcrred to the present experiment. Two frequently difference was never larger than .11 sec. two closely related parts of the body, for employed measures of the degree of For those Ss who took longer to respond example, would have little chance of ambiguity of the associative meaning of to homophones, the magnitude of this survivaI; a synonym for one or the other verbal stimuli are percentage of the greater mean latency ranged from .11 to sense, some linguists have pointed out, primary (most frequently occurring) .81 sec for the 26 Ss. generally comes to take the place of one of response in response distributions to such The results of the present experiment the senses of the homophone. Clearly, stimuli and the number of different suggest that it is feasible to regard research which would clarify the roll of the (n on repeated) responses occurring in homophones as relatively ambiguous semantic environment in the establishment, distributions of total responses to such stimuli for use in research calling for such perpetuation, and disambiguation of stimuli (Cramer, ]968; Deese, 1965). To ambiguous stimulus material, at least 10 the homophones is called for. assess differences between homophones extent that the associative meaning of and nonhomophones on the percentage of homophones reflects the general semantic primary responses produced, a median ambiguity of these word pairs and triplets. chi-square test was perforrned. For Together with the normative data collected REFERENCES homophones the percentage of the prirnary by Galbraith & Taschman (1969), this CRAMER, P. Word association. New York: response in the separate response finding should allow wider use of these Academic Press, 1968. distributions to 13 homophone stimuli relatively unique linguistic structures as DEESE, J. The structure o[ associations in exceeded the overall stimulus list median useful research tools. In addition to the language and thought. Baltimore: Johns Hopkins Press, 1965. pr: iIlary response percentage of 34.2, and potential uses mentioned by Galbraith and GALHRAITH, G., & TASCHMAN, C. Homphone the primary response percentage of the Taschman and in the beginning of this u nits: A nonnative and methodological other 23 separate homophone response paper, these stimuli seem particularly well investigation of the strength of component distributions fell below this median. The suited for research on such phenomena as elements. J oumal of Verbal Learning & Verbal Behavior, 1969, 8, 737-744. results were 22 and 14, respectively, for category c1ustering when they are used as HOWES, D. A word count of spoken English. nonhomophones. Tbe results of this test stimuli in continuous association and recall Joumal of Verbal Learning & Verbal Behavior, yielded a X2 (1) of 4.50, p < .05. Similarly, tasks. Because only homophones, among 1966,5,572-604. a median chi-square test was made all words of multiple meaning in the LEHMAN, W. Historical linguistics: An introduction. New Y ork: Holt, Rinehart, & comparing homophone and English vocabulary, have their distinct Winston, 1962. nonhomophone stimuli in terms of the semantic senses demarcated graphemically THORNDIKE, E., LORGE, I. The teacher's word number of different responses in each of into two or more different words, an book of 30,000 words. New York: Bureau of the distributions of total responses to each additional, unique research potential for Publications, Teachers College, Columbia of the stimulus words. Although homophones is suggested in relating spoken University, 1944. WHITFORD, H. A dictionary o[ American homophones tended to yield more language and its written represcntation. homophones and homographs.. New York: differen t responses per distribution Along with these research uses of Bureau 01' PUblications, Teachers College, (median 01' 14.5 vs median of 11.5) than homophones, the homophone per se merits Columbia University, 1966.

Psychon. Sei., 1970, Vol. 21 (4) 251