Structural analyses of grammatical patterns are often non-unique, as was classically observed by Chao (1934): Different analyses often work equally well, to all appearances. This would not be a problem if linguists did not typically work with the background assumption that there is only one true analysis, which will lead us to deeper insights about Human Language in general.

Some linguists do not make the uniqueness assumption (see below on the description-comparison approach), but among qualitative linguists (especially of the generative grammar tradition), it seems to be still very widespread. So let me reflect a little on the possible reasons for clinging to the uniqueness assumption. But before that, let us consider a concrete example of non-uniqueness, namely the possible existence of “theme vowels” in French verb inflection. There is some recent discussion in Pomino & Remberger (2022), who give the following overview table of different analyses of infinitival forms (consisting of stem, plus possibly theme vowel, plus infinitival suffix). Every practicing linguist knows that this situation is very typical, whether in phonology, morphology or syntax: Phenomena that have been studied widely have usually been given a variety of different analyses.

Why should one think that there is a single analysis in the first place? I do not know, but let me speculate. Over the course of the 20th century, structural linguistics (building on de Saussure 1916 and Bloomfield 1926) replaced historical linguistics, and most of its early practitioners were primarily trained in historical linguistics. History’s task is to find out the true course of event – there is only one history, and of two incompatible alternative accounts, only one can be the right one. Historical linguists were used to arguing back and forth about particular etymologies and particular sound laws, and maybe they simply carried this practice over to the synchronic analyses that they were more and more interested in since the 1920s.

In 1934, the linguist Yuen-Ren Chao (趙元任,1892-1982) published a paper in a relatively obscure Chinese journal (the Bulletin of the Institute of History and Philology of the Academia Sinica, an institution that had been founded only six years earlier, in 1928), in which he pointed out that American structural linguists seemed to make the “tacit assumption” that there was a unique solution in phonemic analysis (Chao 1934: 363):

On the basis of a discussion of Chinese and English phonology (with additional reference to a few other languages), Chao observed that there are multiple solutions, and he generalized this:

“Given a language, there is not necessarily one unique solution for the problem of reducing its sounds into elements.” (Chao 1934: 397)

Chao was not only concerned with theoretical phonology, but in the context of the practical problems of a country that was modernizing and trying to join the Western mainstream, he was also concerned with developing a scientifically grounded romanization system for Chinese (his system is known as Gwoyeu Romatzyh). Thus, his paper also contained various practical considerations that would not be considered relevant in theoretical phonology these days. (At the time, the differences between theoretical and applied phonology were probably not felt so strongly; after all, the first textbook on phonemics, Pike (1947), had the subtitle: A technique for reducing languages to writing.)

Chao’s paper has been cited again and again, but the uniqueness assumption was not dropped by theoretical linguists. Or maybe it was dropped, and was revived by Chomsky? The latter suspicion is fed by Labov’s (1972) paper on linguistic methodology. Labov  describes the situation of the 1960s and seems to say that Chomsky’s mentalist approach revived the uniqueness idea:

“Beginning with the elegant paper of Chao (1934)… we observe a quietistic tendency to claim that all our theories are notational variants of one another, that each is true in its own way… Among the other innovations which Chomsky brought with him was a note of high seriousness in this respect. He is clearly interested in the structure of human language and the capacities of the mind which learns it, not in different ways of looking at the matter.” (Labov 1972: 98)

Thus, the unique truth that Chomsky and his followers tried to find was no longer a single solution to a problem of phonemic analysis, but the true mental grammars of the users of a language.

However, it does not seem to have occurred to most linguists that the assumption that all users of a language have the same internal generalizations is not warranted: Linguistic behaviour is uniform for social reasons, and linguistic well-formedness judgements are judgements of social appropriateness (as I noted in an earlier blogpost). Thus, it could well be that different French speakers make different generalizations about their verbal inflection (to get back to the theme vowels of Pomino & Remberger’s paper): Some speakers may extract theme vowels (e.g. part-i-r, val-oi-r), some may attach these vowels to the stems (parti-r, valoi-r), and some may attach them to the suffixes (part-ir, val-oir). Of course, some of these choices lead to redundant representations, but there are no good reasons to think that storage of linguistic knowledge is redundancy-free. Mental grammars are perhaps best thought of as strategies to get the desired social output, and to extract grammatical knowledge from utterances (and vice versa, to construct utterances from mental grammars). I don’t think it would be very misleading to compare them to cognitive strategies for multiplication (e.g. different strategies for getting from 24 x 18 to 432), where it is clear that different people prefer different strategies.

(Some readers may wonder whether this distinction between “mental grammars” and social regularities is the same as the one made by Quine (1972) between “guiding rules” and “fitting rules”. I am not entirely sure if I understood Quine correctly, but I think it is basically the same. Quine pointed out that two quite different sets of guiding rules can lead to extensionally identical outputs; some discussion can be found here.)

So why don’t get linguists ever tired of proposing new alternative analyses for well-known phenomena, as in the case of French theme vowels? Why isn’t it enough to have one reasonable analysis that gets the facts right? I don’t really know (because I have not understood what makes people think that all speakers of a language have the same mental grammars), but one factor is surely Chomsky’s idea that the innate universal grammar helps learners select the correct grammar. As noted by Lidz & Dudley (2017), „the vast majority of surface evidence is compatible with multiple grammatical descriptions“, and the idea of generative linguistics is that children can select the true grammar on the basis of their innate knowledge.

If this is the case, then by finding the one true mental grammar of a language, linguists can contribute to understanding the architecture of the language faculty – and in this way, they can contribute to general linguistics by analyzing one phenomenon of one language. A study that appears to be purely p-linguistic (e.g. a study of French verbal inflection) can aspire to g-linguistic significance. For example, Pomino & Remberger (2022) appeal to architectural claims of Distributed Morphology (DM) in making their proposal about French theme vowels. They analyze French in terms of universal (and presumably innate) categories such as v° and T°, so if they find the true analysis making use of these categories, this is further evidence for an innate blueprint of this kind.

However, as we saw above, children need not converge on the same internal generalizations – they merely need to arrive at extensionally identical outputs in order to conform to the social conventions. For this, they may use rather different generalizations, so there is no real justification for the uniqueness assumption. Chao’s (1934) observation about phonemic analysis is as relevant now as it was then. Moreover, general claims about innate architecture (such as DM’s) and about innate categories are typically speculative and very hard to substantiate (because of the many moving parts).

We end up with a dilemma for theoretical p-linguistics: As it is quite likely that different speakers use different mental strategies, finding the one true mental grammar is probably not a realistic or even desirable goal. Structural linguists can propose competing analyses, but these may all be equally true, or none may be true (because conceivably no speaker is as intent on extracting generalizations as linguists are). Thus, the theoretical contribution of an analysis that does not cover more ground than earlier analyses is unclear.

So how can p-linguistics contribute to the larger goal of understanding Human Language? I think that general linguistics (of the structural type) must be based on universals, so we need to adopt a description-comparison approach. This means that we identify universal trends on the basis of comparative concepts (as in the World atlas of language stuctures, for example), not on the basis of language-particular analyses. To compare languages, we need to get the facts right, and there is no need to arrive at the right analysis, or for in-depth analyses.

Thus, Chao’s (1934) point is still important today: The uniqueness problem has not been solved, and our p-linguistic analyses are rarely definitive, so we cannot build too much on top of them. General linguistics is best served by rigorous and broad comparison (including more and more languages), rather than by revisiting earlier analyses over and over.

References

Bloomfield, Leonard. 1926. A set of postulates for the science of language. Language 2(3). 153–164. https://doi.org/10.2307/408741.

Chao, Yuen Ren. 1934. The non-uniqueness of phonemic solutions of phonetic systems. Bulletin of the Institute of History and Philology Academia Sinica 4. 363–397. https://www2.ihp.sinica.edu.tw/file/4990FcvKyNN.pdf

de Saussure, Ferdinand. 1916. Cours de linguistique générale. (Ed.) Charles Bally, Albert Sechehaye & Albert Riedlinger. Lausanne & Paris: Payot.

Labov, William. 1972. Some principles of linguistic methodology. Language in Society 1(1). 97–120. https://doi.org/10.1017/S0047404500006576.

Lidz, Jeffrey & Rachel Dudley. 2017. Evaluation measures and the study of language acquisition. Language Acquisition. Routledge 24(2). 81–84. https://doi.org/10.1080/10489223.2017.1312850.

Pike, Kenneth L. 1947. Phonemics: A technique for reducing languages to writing. Ann Arbor, Michigan: University of Michigan Press.

Pomino, Natascha & Eva-Maria Remberger. 2022. Does French have theme vowels? Glossa: A journal of general linguistics 47(1). https://doi.org/10.16995/glossa.8501.

Quine, W. V. 1972. Methodological reflections on current linguistic theory. In Donald Davidson & Gilbert Harman (eds.), Semantics of Natural Language (Synthese Library), 442–454. Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-94-010-2557-7_14.

---

---