Perceptaal alzd Motor Skills, 1977,45, 1311-1314.
@ Perceptual and Motor Skills 1977
VISUAL MASKING BY TRANSLATION EQUIVALENTS I N BILINGUALS1 WARD O'NEILL
Summary.-Masking thresholds for common words as a function of the relationship between target and mask were assessed using French-English bilingual subjects. Five target-mask combinations were employed. Two control conditions involved unrelated pairs presented in either the same or both languages. The experimental conditions employed masks that were subordinate to the target, e.g., animal-horse, masks that were the translation equivalents of the targets, and masks which were homophones of the target. Homophones were superior masks compared to same-language, unrelated masks which were in turn superior to subordinately related masks. Translation equivalents were less effective masks than different-language unrelated masks but were similar to subordinate samelenguage masks. The results suggest that the relationship between the representations of translation equivalents is similar to that between same-language superordinate and subordinate words.
The bulk of recent evidence in bilingual information processing research suggests that bilinguals store verbal items in either of their languages in a common store; for a review see McCormack (1976). However, the nature of the relationship between the representations of translation equivalents has yet to be firmly established. Dalrymple-Alford and Aamiry ( 1970) postulated a relation of synonymity between translation equivalents while others (Liepman & Saegert, 1974) have argued for identity of representations at the semantic level. MacLeod (1976) employing a savings paradigm developed by Nelson ( 1971) found a pattern of savings with transfer between translation equivalents which does not conform with the lack of savings earlier reported for transfer between synonyms (Nelson, 1971). This led him to conclude that translation equivalents share a common abstract representation. However, the possibility remains that representations may be associated in a non-synonymous relationship. Interestingly MacLeod reports an unpublished study by Nelson and Fehling ( 1974) in which savings were reported when transfer was between words in a superordinatesubordinate relation although not for other types of associates. The purpose of this experiment was to examine the hypothesis that translation equivalents may function like superordinate-subordinate associates using a masking procedure. Jacobsen (1973) demonstrated that target words are less effectively masked by their associates than by other unrelated words. The finding has T h i s research was supported in part by a grant from the Conseil de Recherches, Universitk de Moncton. The author would like to thank Robert Huot for assistance in collecting data.
1312
W. O'NEILL
been replicated and a guessing interpretation of the effect shown to be questionable (Jacobsen, 1974, 1976). In the present experiment the basic Jacobsen paradigm is repeated using the superordinate-subordinate relationship as the basis of the associative condition. Three bilingual conditions have been added. In one French-English translation equivalents served as targets and masks. The second is a control condition using unrelated French-English targets and masks. The third condition involves using bilingual homophones (Phone-faune). The purpose of the latter condition was to attempt to replicate Jacobsen's (1976) finding that homophones are relatively effective maskers. To interpret the results of the bilingual conditions, it was felt important to replicate both the associative and homophonic masking results reported by Jacobsen. It was hypothesized that superordinate-subordinate words would produce less masking than unrelated words while homophones would produce greater masking. The central hypothesis was that translation equivalents would function similarly to the unilingual superordinate-subordinate items.
METHOD The subjects were 5 male and 5 female French-English bilinguals, between 19 and 22 yr. of age, who volunteered to participate in the study. Subjects were classified as "balanced bilinguals based on a language self-report scale (Macnamara, 1967). The criterion used was no more than one unit deviation on any of four scales between either language. The words serving as targets and masks were all common words in both their original language and translated forms, e.g., foot, pied. Masks were always of the same or greater length than corresponding targets and no more than one identical letter occurred in corresponding positions in the target and mask. Different targets were employed in each condition to eliminate effects of familiarity. Targets and masks were inscribed in white Lettraset (No. 724) on black cards. Presentations were made by means of a Gerbrands Model 300 threechannel tachistoscope. The subjects' eyes were 92 cm from the stimuli which were 1.5 cm in height. Subjects viewed five different target-mask combinations. Four experimental conditions, two unilingual and two bilingual consisted of eight pairs each. The unilingual conditions were the superordinate-subordinate pairs, and the unilingual unrelated pairs. Four of each group of pairs were English and four French. The two bilingual conditions were translation equivalents and unrelated pairs, with language of target and mask counterbalanced. Four pairs were used in the homophonic condition all linguistically mixed but counterbalanced in terms of language of target and mask.
VISUAL MASKING IN BILINGUALS
1313
Trials began with the presentation for 3 sec. of a fixation point followed by the target for 20 msec. and a 10-msec. delay interval. The mask was presented for 120 msec. following the delay interval. Illumination was adjusted during practice trials for each subject to ensure adequate masking. Masking thresholds were obtained by increasing the delay interval in steps of 5 msec. until the subjects correctly identified the target. The method probably underestimates true thresholds but does not seem to affect the ordering of data ( C . F. Jacobsen, 1976). The sequence of 36 trials was completely randomized for each subject.
RESULTSAND DISCUSSION The mean masking thresholds were calculated for each condition and averaged over subjects. The mean thresholds were for the unilingual superordinatesubordinate words, 77 msec.; for the translation equivalents, 93 msec.; for the unilingual unrelated words, 125 msec.; for the bilingual unrelated words, 149 msec.; and for the homophones, 165 msec. An analysis of variance yielded a significant conditions effect (F4,36= 8.94, # < $005) . A Duncan's multiple-range test showed that the masking threshold for superordinate-subordinate words was significantly lower than that for unilingual unrelated words ( p < -05). This extends Jacobsen's earlier finding with associates. The mean masking threshold for homophones was significantly greater than for unilingual unrelated words ( p < .05), confirming the greater masking power of homophones relative to unrelated words also reported by Jacobsen. The difference between the masking thresholds for translation equivalents and bilingual unrelated words was also significant (fi < .01) , demonstrating that the facilitating effect of semantic relatedness on masking thresholds exists even in the presence of language mixing. The difference between the unilingual unrelated words and the translation equivalents does not approach acceptable significance so translation equivalents and unilingual subordinate masks function similarly. The experimental hypotheses have been supported, and it appears that considering translation equivalents in the framework of superordinate-subordinate hierarchies may prove fruitful. The fact that translation equivalents and unilingual unrelated masks did not differ significantly suggests that language mixing introduces an element of processing not present in unilingual associative masking. The fact that bilingual unrelated masks and homophones did not differ significantly reinforces this possibility. One hypothesis is that language mixing introduces the necessity of a process of language-tag recognition (McCormack, 1976) which may be independent of semantic recognition. Two cautions are in order in the interpretation of the present results. The
1314
W. O'NEILL
first consists of the possibility that the facilitation due to semantic relatedness may be a function of guessing. Arguing against this interpretation is the finding that homophones do not produce comparable facilitation (see also Jacobsen, 1976) but rather are more efficient masks. Although this seems contrary to a guessing interpretation it remains possible that the homophone results are caused by an interference process. The true test of the guessing interpretation lies in the use of more sophisticated psychological techniques such as a yes-no procedure (Green & Swets, 1966). A further caution is that in this study only the superordinate-subordinate relationship was examined. It is possible that other associative descriptions such as a set-theoretic approach (Meyer, 1970) may best describe the nature of the relationship between translation equivalents. However, the masking paradigm appears to be a useful one for investigating these possibilities. REFERENCES DALRYMPLE-ALFORD, E. C., & AAMIRY, A. Word associations of bilinguals. Psychonomic Science, 1970, 21, 319-320. GREEN,D. M., & SWETS,J. A. Signal detection and psychophysics. New York: Wiley, 1966. JACOBSEN, J. 2. Effects of association upon masking and reading latency. Casadkn Joarnal of Psychology, 1973, 27, 58-64. J. 2. Interaction of similarity to words of visual masks and targets. Joarnal JACOBSEN, of Experimental Psychology, 1974, 102, 431-434. J. 2. Visual masking by homonyms. Canadiala Joarnal of Psychology, 1976, JACOBSEN, 30, 174-177. LIEPMAN, D., & SAEGERT, J. Language tagging in bilingual free recall. Joarnul of Experimental Psychology, 1974, 103, 1137-1141. MACLEOD, C. M. Bilingual episodic memory: acquisition and forgetting. lournu1 of Verbal Le~laingand Verbal Behavior, 1976, 15, 347-364. MACNAMARA, J. How can one measure the extent of a person's bilingual proficiency. In L. G. Kelly (Ed.), Description and measzcrement of bilingzlalism: an internatiolzal semiw. Toronto: Univer. of Toronto Press, 1967. Pp. 79-97. MCCORMACK, P. D. Language as an attribute of memory. Calaadian Joarnal of Psychology, 1976, 30, 238-248. MEYER,P. E. On the representation and retrieval of stored semantic information. Coglzitive Psychology, 1970, 1, 242-300. NELSON,T. 0. Savings and forgetting from long-term memory. Joarnal of Verbal Lem~ingand Verbal Behavior, 1971, 101, 568-576. Accefited November 15, 1977
@ Perceptual and Motor Skills 1977
VISUAL MASKING BY TRANSLATION EQUIVALENTS I N BILINGUALS1 WARD O'NEILL
Summary.-Masking thresholds for common words as a function of the relationship between target and mask were assessed using French-English bilingual subjects. Five target-mask combinations were employed. Two control conditions involved unrelated pairs presented in either the same or both languages. The experimental conditions employed masks that were subordinate to the target, e.g., animal-horse, masks that were the translation equivalents of the targets, and masks which were homophones of the target. Homophones were superior masks compared to same-language, unrelated masks which were in turn superior to subordinately related masks. Translation equivalents were less effective masks than different-language unrelated masks but were similar to subordinate samelenguage masks. The results suggest that the relationship between the representations of translation equivalents is similar to that between same-language superordinate and subordinate words.
The bulk of recent evidence in bilingual information processing research suggests that bilinguals store verbal items in either of their languages in a common store; for a review see McCormack (1976). However, the nature of the relationship between the representations of translation equivalents has yet to be firmly established. Dalrymple-Alford and Aamiry ( 1970) postulated a relation of synonymity between translation equivalents while others (Liepman & Saegert, 1974) have argued for identity of representations at the semantic level. MacLeod (1976) employing a savings paradigm developed by Nelson ( 1971) found a pattern of savings with transfer between translation equivalents which does not conform with the lack of savings earlier reported for transfer between synonyms (Nelson, 1971). This led him to conclude that translation equivalents share a common abstract representation. However, the possibility remains that representations may be associated in a non-synonymous relationship. Interestingly MacLeod reports an unpublished study by Nelson and Fehling ( 1974) in which savings were reported when transfer was between words in a superordinatesubordinate relation although not for other types of associates. The purpose of this experiment was to examine the hypothesis that translation equivalents may function like superordinate-subordinate associates using a masking procedure. Jacobsen (1973) demonstrated that target words are less effectively masked by their associates than by other unrelated words. The finding has T h i s research was supported in part by a grant from the Conseil de Recherches, Universitk de Moncton. The author would like to thank Robert Huot for assistance in collecting data.
1312
W. O'NEILL
been replicated and a guessing interpretation of the effect shown to be questionable (Jacobsen, 1974, 1976). In the present experiment the basic Jacobsen paradigm is repeated using the superordinate-subordinate relationship as the basis of the associative condition. Three bilingual conditions have been added. In one French-English translation equivalents served as targets and masks. The second is a control condition using unrelated French-English targets and masks. The third condition involves using bilingual homophones (Phone-faune). The purpose of the latter condition was to attempt to replicate Jacobsen's (1976) finding that homophones are relatively effective maskers. To interpret the results of the bilingual conditions, it was felt important to replicate both the associative and homophonic masking results reported by Jacobsen. It was hypothesized that superordinate-subordinate words would produce less masking than unrelated words while homophones would produce greater masking. The central hypothesis was that translation equivalents would function similarly to the unilingual superordinate-subordinate items.
METHOD The subjects were 5 male and 5 female French-English bilinguals, between 19 and 22 yr. of age, who volunteered to participate in the study. Subjects were classified as "balanced bilinguals based on a language self-report scale (Macnamara, 1967). The criterion used was no more than one unit deviation on any of four scales between either language. The words serving as targets and masks were all common words in both their original language and translated forms, e.g., foot, pied. Masks were always of the same or greater length than corresponding targets and no more than one identical letter occurred in corresponding positions in the target and mask. Different targets were employed in each condition to eliminate effects of familiarity. Targets and masks were inscribed in white Lettraset (No. 724) on black cards. Presentations were made by means of a Gerbrands Model 300 threechannel tachistoscope. The subjects' eyes were 92 cm from the stimuli which were 1.5 cm in height. Subjects viewed five different target-mask combinations. Four experimental conditions, two unilingual and two bilingual consisted of eight pairs each. The unilingual conditions were the superordinate-subordinate pairs, and the unilingual unrelated pairs. Four of each group of pairs were English and four French. The two bilingual conditions were translation equivalents and unrelated pairs, with language of target and mask counterbalanced. Four pairs were used in the homophonic condition all linguistically mixed but counterbalanced in terms of language of target and mask.
VISUAL MASKING IN BILINGUALS
1313
Trials began with the presentation for 3 sec. of a fixation point followed by the target for 20 msec. and a 10-msec. delay interval. The mask was presented for 120 msec. following the delay interval. Illumination was adjusted during practice trials for each subject to ensure adequate masking. Masking thresholds were obtained by increasing the delay interval in steps of 5 msec. until the subjects correctly identified the target. The method probably underestimates true thresholds but does not seem to affect the ordering of data ( C . F. Jacobsen, 1976). The sequence of 36 trials was completely randomized for each subject.
RESULTSAND DISCUSSION The mean masking thresholds were calculated for each condition and averaged over subjects. The mean thresholds were for the unilingual superordinatesubordinate words, 77 msec.; for the translation equivalents, 93 msec.; for the unilingual unrelated words, 125 msec.; for the bilingual unrelated words, 149 msec.; and for the homophones, 165 msec. An analysis of variance yielded a significant conditions effect (F4,36= 8.94, # < $005) . A Duncan's multiple-range test showed that the masking threshold for superordinate-subordinate words was significantly lower than that for unilingual unrelated words ( p < -05). This extends Jacobsen's earlier finding with associates. The mean masking threshold for homophones was significantly greater than for unilingual unrelated words ( p < .05), confirming the greater masking power of homophones relative to unrelated words also reported by Jacobsen. The difference between the masking thresholds for translation equivalents and bilingual unrelated words was also significant (fi < .01) , demonstrating that the facilitating effect of semantic relatedness on masking thresholds exists even in the presence of language mixing. The difference between the unilingual unrelated words and the translation equivalents does not approach acceptable significance so translation equivalents and unilingual subordinate masks function similarly. The experimental hypotheses have been supported, and it appears that considering translation equivalents in the framework of superordinate-subordinate hierarchies may prove fruitful. The fact that translation equivalents and unilingual unrelated masks did not differ significantly suggests that language mixing introduces an element of processing not present in unilingual associative masking. The fact that bilingual unrelated masks and homophones did not differ significantly reinforces this possibility. One hypothesis is that language mixing introduces the necessity of a process of language-tag recognition (McCormack, 1976) which may be independent of semantic recognition. Two cautions are in order in the interpretation of the present results. The
1314
W. O'NEILL
first consists of the possibility that the facilitation due to semantic relatedness may be a function of guessing. Arguing against this interpretation is the finding that homophones do not produce comparable facilitation (see also Jacobsen, 1976) but rather are more efficient masks. Although this seems contrary to a guessing interpretation it remains possible that the homophone results are caused by an interference process. The true test of the guessing interpretation lies in the use of more sophisticated psychological techniques such as a yes-no procedure (Green & Swets, 1966). A further caution is that in this study only the superordinate-subordinate relationship was examined. It is possible that other associative descriptions such as a set-theoretic approach (Meyer, 1970) may best describe the nature of the relationship between translation equivalents. However, the masking paradigm appears to be a useful one for investigating these possibilities. REFERENCES DALRYMPLE-ALFORD, E. C., & AAMIRY, A. Word associations of bilinguals. Psychonomic Science, 1970, 21, 319-320. GREEN,D. M., & SWETS,J. A. Signal detection and psychophysics. New York: Wiley, 1966. JACOBSEN, J. 2. Effects of association upon masking and reading latency. Casadkn Joarnal of Psychology, 1973, 27, 58-64. J. 2. Interaction of similarity to words of visual masks and targets. Joarnal JACOBSEN, of Experimental Psychology, 1974, 102, 431-434. J. 2. Visual masking by homonyms. Canadiala Joarnal of Psychology, 1976, JACOBSEN, 30, 174-177. LIEPMAN, D., & SAEGERT, J. Language tagging in bilingual free recall. Joarnul of Experimental Psychology, 1974, 103, 1137-1141. MACLEOD, C. M. Bilingual episodic memory: acquisition and forgetting. lournu1 of Verbal Le~laingand Verbal Behavior, 1976, 15, 347-364. MACNAMARA, J. How can one measure the extent of a person's bilingual proficiency. In L. G. Kelly (Ed.), Description and measzcrement of bilingzlalism: an internatiolzal semiw. Toronto: Univer. of Toronto Press, 1967. Pp. 79-97. MCCORMACK, P. D. Language as an attribute of memory. Calaadian Joarnal of Psychology, 1976, 30, 238-248. MEYER,P. E. On the representation and retrieval of stored semantic information. Coglzitive Psychology, 1970, 1, 242-300. NELSON,T. 0. Savings and forgetting from long-term memory. Joarnal of Verbal Lem~ingand Verbal Behavior, 1971, 101, 568-576. Accefited November 15, 1977
