"Compound" and "Coordinate" Bilingualism: A Conceptual Artifact*

"Compound" and "Coordinate" Bilingualism: A Conceptual Artifact* Since the Fifties, linguists and psychologists have talked of compound and coordinate bilingualism as if such phenomena existed in identifiable form. The terms were conceived in the context of behaviorist learning theory and the Saussurean theory of signs, and apparently Ervin and Osgood were the first to use them in print.l The notion seems to be based on the belief that different manners of learning second languages will result in radically different grammars in the brain. Compound bilinguals, it is thought, do not have an independent grammar for their second language. It is asserted that people can learn a second language in such a way that it will always be dependent on (i.e., compounded to) the first language. A putative example would be the case of the student who is taught an English equivalent for every French word. This student might eventually become a balanced bilingual and his ordinary conversation might become indistinguishable from that of a native Frenchman. Yet it would be asserted by some psycholinguists that this compound bilingual, because of the way he originally learned French, would still be translating into English every time he heard French, and translating out of English every time he spoke French. Coordinate bilinguals, on the other hand, would be those people who learned two languages in separate contexts; therefore, the grammars of their two languages would be completely independent. It is even thought that coordinate bilinguals would have great difficulty in translating because of this separateness of their two languages. I argue that it is an error to think that there are two kinds of bilingualism that fit the labels compound and coordinate. First, compound and coordinate • This article was originally presented as a paper at the Forty-second Annual Meeting of the Linguistic Society of America in Chicago on December 29, 1967, and revised for publication in this journal. 1 Susan Ervin and Charles E. Osgood, "Second Language Learning and Bilingualism," in Psycholinguistics, ed. Charles E. Osgood and Thomas A. Sebeok (Bloomington, Ind., 1954), pp. 139-146. 254 "COMPOUND" AND "COORDINATE" BILINGUALISM 255 bilingualism are poorly defined; second, the experimental evidence does not support these concepts; and third, there are strong linguistic reasons why these concepts cannot stand. I First, the problem of definition: compound and coordinate bilingualism are not well defined. Perhaps as an effort to get around impreciseness, it is usually stated that bilinguals may be arranged along a continuum between pure compoundness and pure coordinateness, and that indeed there is no sharp dichotomy between the two kinds of bilingualism. Even worse, incompatible definitions have been formulated for compound and coordinate bilingualism. Let us take as an example a typical language student who studies either by the grammar-translation method in which he learns word lists or by the audiolingual method in which he memorizes dialogues with the aid of translations in parallel columns. Will this student become a compound or a coordinate bilingual? According to the definition of Ervin and Osgood,2 this person is a prime example of the compound bilingual. But Lambert3 has changed his definition to argue the opposite-he argues that everyone who learns a second language outside the home after ten years of age becomes a coordinate bilingual. Another example of incompatible definitions is the case of a man who has a Swedish-speaking mother and a Finnish-speaking father. Speaking both Swedish and Finnish in the home from childhood, he is equally proficient in both languages. Is he a compound or a coordinate bilingual? Ervin and Osgood and also Lambert would agree that he is a compound bilingual. Brooks, however, would take exception here. This man is a "true" bilingual, he would argue, and "true" bilinguals are coordinate bilinguals. Brooks has what we might call a "common-sense definition" of compound and coordinate bilingualism. For him, a compound bilingual is a language learner who is still in a decoding stage, the stage we were in after studying Latin in high school. Yet notice how close Brooks comes to saying that compound bilingualism is not truly bilingualism at all.4 Not only do various authorities use the terms compound and coordinate bilingualism with different meanings, but sometimes these terms are used with different meanings by the same authorities. Ervin and Osgood,s for 2 Ibid., pp. 139-140. 3 W. E. Lambert, "Psychological Studies of the Interdependencies of the Bilingual's Two Languages," presented to the Linguistic Institute of the Linguistic Society of America, mimeographed (1966), p. 14. 4 Nelson Brooks, Language and Language Learning (New York, 1964), p. 267. s Ervin and Osgood, p. 140. 256 KARL C. DILLER example, maintained that people would become compound bilinguals when learning a language through "vocabulary lists, which associate a sign from language B with a sign and its meaning in language A." Yet what about the man who has spoken Swedish and Finnish since infancy? They would say that he is a compound bilingual. But which language is basic for him? Does he translate from Swedish into and out of Finnish-or does it work the other way? Ervin and Osgood have already provided an answer: Neither Swedish nor Finnish would be basic, but he would have some third compromise language. He would have to translate both for Swedish and for Finnish. As Ervin and Osgood put it, "in this instance some compromise representational processes taken from both languages may be established, with neither having- pronounced dominance." 6 Ervin and Osgood, then, have at least two distinct types of compound bilingualism: one in which the second language is compounded on to the first and one in which two native languages are compounded to a compromise base. My first major point is, then, that the terms compound and coordinate bilingualism are not well defined. There exist in current usage at least three definitions which are incompatible with each other. Furthermore, the assertion that there is a continuum between pure compoundness and pure coordinateness only suggests that there is no clear distinction between these two supposedly different kinds of bilingualism. II Let us turn now to the question of whether there is any good experimental evidence to support the notion of two kinds of bilingualism. There are a number of experiments on the problem, the most notable of which are by Lambert and his associates at McGill University. Ervin and Osgood suggested that the Semantic Differential would be a good device to test whether "translation-equivalent signs" are identical or different for the two kinds of bilinguals. 7 But is the Semantic Differential a good test? After all, it has almost no bearing on what linguists think of as semantic matters; it furnishes no information that could be used by a dictionary maker. When a person rates a word on the Semantic Differential, he tries to analyze his emotional reaction to the word or to its underlying concept. The scale is a continuum between two opposing adjectives, and the word to be rated is a noun or a noun construction. For example, on a Semantic Differential test we might be asked whether a house is hot or cold. One can say that it is halfway between hot and cold, or three-fourths of the way to hot, and so on. Then we might be asked whether houses are 6 Ibid. 7 Ibid., p. 141. "COMPOUND" AND "COORDINATE" niLINGUALISM 257 beautiful or ugly, angular or rounded, small or large. Surely the dictionary definition of house depends in no way on whether houses are beautiful, rounded, large, or even hot. Therefore, we cannot use this test to learn whether bilinguals have compound or coordinate dictionaries in their brains. It becomes evident that one's ratings on this test are affected more by one's experience with houses than by the manner in which he learned his second language. Carroll suggests that the Semantic Differential "might better be termed an experiential differential." s Lambert's experiments with the Semantic Differential underline the problems we have just considered. He found that, as expected, his "compounds" gave similar ratings to such pairs as house and maison, but his "coordinates" divided themselves into two groups according to their responses. Dissimilar ratings were given by the "coordinates" who had learned their two languages in two different cultural settings, such as Paris for French and Montreal for English. However, those who had learned both of their languages in the same place (in this case, Montreal) behaved as if they were compounds.9 This result supports the view that the experiential context, not the manner oflearning the second language, is reflected in Semantic Differential ratings. The Semantic Differential tests the wrong variables, and we can get no evidence about compound and coordinate bilingualism from this experiment or any other which uses the Semantic Differential, such as Lambert's experiment on the "satiation of meaning." There are some-Diebold, for example-who assert that "the most compelling evidence" for compound and coordinate bilingualism comes from case histories of aphasic bilinguals.10 The evidence is not really at all compelling. Diebold credits Lambert and Fillenbaum 11 with "the most significant discovery" in this area. Those authors, however, are much more cautious in interpreting the results of their pilot study of aphasia among bilinguals. They studied the cases of fourteen bilingual aphasics in Montreal a_nd compared them with twenty 8 John B. Carroll, rev. of The Measurement of Meaning, by Charles E. Osgood, George J. Suci, and Percy H. Tannenbaum, Language, XXXV (1959), 72. 9 W. E. Lambert, "Behavioral Evidence for Contrasting Forms of Bilingualism," Report of the Twelfth Annual Round Table Meeting 011 Linguistics and Language Studies, ed. Michael Zarechnak, Monograph Series on Languages and Linguistics, Georgetown University, No. 14 (Washington, D.C., 1961), p. 76. to Cf. A. Richard Diebold, Jr., "The Consequences of Early Bilingualism in Cognitive Development and Personality Formation," prepared for the symposium "The Study of Personality: An Interdisciplinary Appraisal," Rice University, Houston, Texas, mimeographed (1966), p. 15. II W. E. Lambert and S. Fillenbaum, "A Pilot Study of Aphasia among Bilinguals," Canadian Journal of Psychology, XIII (1959), 28-34. 258 KARL C. DILLER European cases reported by Leischner. In the Montreal cases, "both languages are damaged but retain the order of dominance which was developed prior to aphasia." 12 In a number of the European cases, however, a childhood language did not appear to survive as well as the currently used second language. If one assumes that the Montreal aphasics were compounds and that the European asphasics were coordinates, then the data would indicate that compounds and coordinates have different recovery responses to aphasia. But Lambert and Fillenbaum point out that there is an alternative explanation: Montreal is a bilingual city, and the person recovering from aphasia there has the freedom to use both his languages. Yet in most of the European cases, the bilingual was in a monolingual environment and had little opportunity to use his childhood language. The convalescent's environment may explain the differences in the case histories, and this is probably the better answer. I know of no cases where aphasia has obliterated one of a bilingual's languages, leaving the other language untouched. The reason is that aphasia is not the loss of language; it is a physiological interference with language. Lenneberg has formulated this distinction well: "careful observation of the recovery process during the critical post-morbid period, makes it very plain that the patient does not start with specific lexical or grammatical lacunae, but that some basic physiological processes relating to activating, monitoring, or processing of speech are deranged." 13 If we accept Lenneberg's viewpoint, we are almost forced to accept Lambert's alternative explanation, which stresses the difference between a monolingual and a bilingual environment for the person recovering from aphasia. We have seen here, with the Semantic Differential tests and with aphasia, two cases in which various groups of bilinguals have behaved in statistically different ways. In neither case can we link this difference in behavior to compound and coordinate bilingualism. The converse, however, is just as vacuous. Failure to find statistically significant differences in behavior between groups of bilinguals does not mean that compound and coordinate bilingualism do not exist. Kolers reports on a word-association experiment in which subjects "with nearly identical linguistic histories respond differently, while others with different histories respond similarly".14 This finding, he says, "casts some doubt on the usefulness of these terms [compound and coordinate]." It is impossible, however, to find crucial evidence against such poorly defined concepts. Since it is not at all clear why 12 Ibid., pp. 31-32. 13 Eric H. Lenneberg, Biological Foundations of Language (New York, 1967), p. 207. 14 Paul A. Kolers, "Interlingual Word Associations," Journal of Verbal Learning and Verbal Behavior, II (1963), 299. "COMPOUND" AND "COORDINATE" BILINGUALISM 259 compound and coordinate bilinguals should behave differently on word-association tests, we cannot say that Kolers' experiment does more than "cast doubt." If the compound/coordinate distinction has any validity, it should be corroborated by translation tests: the compounds should be better translators. However, in four experiments involving translation, Lambert found negative results in three. As he has pointed out, difficulties in experimental design keep these negative results from being conclusive.15 The same can be said for the experiment that had positive results. My second major point in this article is that no good experimental evidence supports the distinction between compound and coordinate bilingualism. One should not be surprised, then, at Lambert's statement, made at the 1966 Linguistic Institute, that although he is encouraged by his current bilingual experiments with the Stroop color-word test, his "confidence in the compound-coordinate matter is only luke-warm."16 Ill Now let us consider the theoretical difficulties of compound and coordinate bilingualism from a more linguistic viewpoint. Weinreich wrote of people who have "merged [linguistic] systems." 17 This phenomenon surely exists for many language learners. Yet is it possible for a proficient bilingual to have merged linguistic systems? The answer would seem to be no, because, as Weinreich points out, merged systems cause linguistic interference. Let us take an example from syntax. Native speakers of French will frequently try to form the English past tense by using a construction with have and a past participle. That, of course, gives them the present perfect tense. They say, "Yesterday I have eaten lunch at 12 o'clock." One can easily understand the mistake. We might say that this bilingual has a merged or compound system and that he is trying to speak English as if it were French. But then compound grammatical systems are a cause of mistakes: no two languages are grammatically similar enough to be compoundable. The same is true with regard to vocabulary. A glance through a good bilingual dictionary should show the impossibility of a compounded vocabulary. The vast majority of words in French, for example, have to be 1s Lambert, "Behavioral Evidence," p. 78. 16 Lambert, "Psychological Studies," p. 41. 17 Uriel Weinreich, Languages in Contact: Findings and Problems (The Hague, 1964), pp. 8-11. 260 KARL C. DILLER defined by at least two English words, and vice versa. Sometimes the dictionary will take half a page to give the equivalents of a word in the other language. Ervin and Osgood suggested that the learning of word lists would foster compound systems,1B but few word lists are that simple. Take, for example, the third lesson of an elementary Greek textbook: agathos means 'good' or it means 'brave'; ethe/6, 'wish' or 'be willing'; kai, 'and', 'also', or 'even'; kalos, 'beautiful', 'honorable', or 'fine'; luo, 'loose', 'break', or 'destroy'. Nearly half of the words in that vocabulary list have two or three English glosses, and that is a simplified list. If a person tries to compound his vocabulary, he can expect trouble. A Frenchman who liked his steaks saignant asked "for a "bloody beefsteak" in London. The waiter replied, "I suppose you want seme God-damned mashed potatoes too." My argument here is that insofar as people have compound or merged linguistic systems, they will make mistakes in their second language. No two languages are similar enough to allow morpheme-by-morpheme or rule-by-rule correspondences.19 My second linguistic argument questions whether "coordinate bilinguals should be poorer translators than compounds are." Does a bilingual exist whose languages are stored so separately in his brain that he cannot speak in one language about things which he has learned in the other? Obviously not. Conversations with foreign students about certain aspects of their native countries would be impossible if that were the case. Surely one can expect to find his best translators among people who have the least interference between their languages. It has been argued that compound bilinguals should be better translators since the translation equivalents arc already compounded in the bilingual's brain. Yet it seems quite unlikely that a person stores a verbal definition with each word in his brain. A child might be told that a bungalow is a 'small house'. However, he does not thereby make a compound sign with bungalow and small house. He knows that such definitions are approximate and that bungalows are probably 18 Ervin and Osgood, p. 140. 19 J. C. Catford (A Linguistic Theory of Translation [London, 1965]) has provided a great deal of evidence on the noncompoundability of language pairs. Even two dialects of the same language frequently have noncompoundable subsystems. He cites this example on p. 37: Standard English has a two-dimensional system of demonstratives (this, that/ these, those), but NE Scots has a unidimensional system (this, that, yon) in which plurality is irrelevant. These two systems cannot be compounded. On the other hand, there are example of dialects which differ only in a few phonological rules, such as Pig Latin and English. Here a compound system might be possible but would be superfluous. The Pig Latin form is always predictable from the English word. If a new word is brought into English, e.g., Sputnik, there is no doubt about what the Pig Latin equivalent will be. "COMPOUND" AND "COORDINATE" BILINGUALISM 261 different from such small houses as cottages and cabins. Why, then, would a person make a compound sign when he is told that a bungalow is a 'petite maison'? Translation is actually quite similar to the process of paraphrasing. Just as it seems senseless to talk of a compound monolingualism which fosters paraphrasis, it is likely that there is no such thing as a compound bilingualism which facilitates translation. In conclusion, let us summarize our observations. First, we saw the problems in defining compound and coordinate bilingualism. At least three contradictory definitions are currently used. Sec::ond, we saw that the experimental evidence does not confirm a distinction between two kinds of bilingualism. Third, we pointed out that what might be taken as compound systems are a cause of mistakes in the grammar and vocabulary of a bilingual's second language. Therefore, compound systems are incompatible with bilingualism, and the expression compound bilingualism is a contradiction in terms. Yet neither can one hold to the notion of a strict coordinate bilingualism in which the languages are so separate that translation is inhibited. All bilinguals can translate, and the manner in which a bilingual has learned his second language does not seem to inhibit his ability to translate. In 1956, Haugen gave every benefit of the doubt to psycholinguists who were trying to solve the problem of compound and coordinate bilingualism. He noted, however, that "little has been done so far beyond the elaboration of a terminology."20 What I have been arguing here is that compound and coordinate bilingualism will never be more than an "elaboration of a terminology." The terms are empty. Compound and coordinate systems do not exist in identifiable form in bilinguals who are proficient in their second language. Department of English Harvard University Cambridge, Massachusetts 02138 20 Einar Haugen, Bilingualism in the Americas: A Bibliography and Research Guide, Publication of the American Dialect Society, No. 26 (University, Ala., 1956), p. 69.