José-Luis, I’d like to thank you for writing a detailed comment on one of my recent posts (on differential object marking) on your own blog (Philosophy of Linguistics). I’d like to discuss some of the general issues in more detail.
Thank you very much, Martin, for your interest in my opinions.
I agree with you that in addition to understanding “why” (e.g. why object marking tends to be differential in a particular way in languages), scientists also want to understand “how”, and this probably applies throughout biology and the behavioural sciences in much the same way. Following a famous 1963 paper by the ethologist Niko Tinbergen, scientists often distinguish between ultimate (adaptive and phylogenetic) explanations and proximate explanations (mechanistic and developmental/ontogenetic). So would you say that the distinction (between “why” and “how”) that you drew in your blogpost can be understood in this way? Can generative grammar be said to answer the question about the mechanisms of language in Tinbergen’s sense?
In short, yes. It’s very interesting that you raise this question in relation to Tinbergen, given that the origin of Generative Grammar (GG) has a direct relation with the ethological program of the 1950s. In fact, Chomsky was inspired by Lorenz and Tinbergen to attack Skinner’s behaviourist program and his assumption that behaviour is a response to external stimuli. This ethological programme is one of the sources of inspiration for the idea that the Faculty of Language is a human mental organ that results from the interaction between an innate predisposition (an instinct) and external stimuli. In fact, Chomsky (1988, among others) usually presents the task of linguistics (from a cognitive perspective) by asking certain questions that closely resemble those that Tinbergen proposes to organize the research strategy for ethology in the article you mention, which is clearly not accidental (for ease of reference I identify each question with a proper name and complete Chomsky’s original list with Darwin’s problem):
- What is knowledge of language? (Humboldt’s question)
- How is that knowledge acquired? (Plato’s question)
- How is that knowledge put to use? (Descartes’ question)
- How is that knowledge implemented in the brain? (Broca’s question)
- How did that knowledge emerge in the species? (Darwin’s question)
It is important to note that these questions are conceived from the cognitive perspective. This implies that the object of study of GG is not languages conceived as social or cultural objects, but rather it is the knowledge of language (that is, the “language organ” or the language faculty of people). The knowledge of language is not social or cultural; it is individual, it is a property of each person (pathologies apart). The cognitive perspective involves constructing an explicit model (a theory) of what is the structure of these knowledge systems that we call languages. The central question of linguistics (from the cognitive perspective) is, therefore, question 1 (which would coincide with what you term “mechanistic explanation”). Without a developed answer to this question (at least partially), reasonable progress in any of the others is impossible.
Yes, this is one way of looking at Human Language, and I mostly agree that these are interesting questions. But the phenomena that most linguists (including most generative linguists) look at on an everyday basis are social and cultural. For example, when we study differential object marking in Persian, we study the social conventions of Persian speakers (which are part of their overall cultural knowledge). Generative linguists often say that they really study the linguistic knowledge of an “idealized Persian speaker”, but this boils down to the same thing, because this idealized speaker must know the conventions of the speech community. These conventions have a reality outside of language organs, so we can learn them from books, for example. In your recent paper on UG (Mendívil-Giró 2018a), you distinguish between a biological aspect of language and a cultural aspect, which you call the “lexical interface”.
Yes, I think that every language has a biological (or natural) component and a cultural component. The cultural component is the result of linguistic change in historical time, while the biological component is universal, that is, invariable in (historical) time and in space. Obviously, the biological component is in turn the result of our species’ evolution (in geological time), and Darwin’s question (question 5) refers to it. But the historical change of the cultural component is not part of evolutionary theory in the strict sense. In this sense, I do not believe that we can affirm that languages evolve, but simply that they change. I mean that we have no arguments to say that linguistic changes are part of the evolution of the faculty of language, although this is the opinion of those who believe that languages are purely cultural objects.
Most people that study language change (without adopting a biological perspective) don’t use the term “faculty of language”…
OK, but if we accept that languages have a biological/natural component, then the only thing that is subject to change and variation is the cultural component. Although it may be a simplification, following the expository strategy that I have used in other works (Mendívil-Giró 2014, 2018a,b), let’s assume that the only variable part in languages is morphology and phonology (in fact, this is a common hypothesis in current GG), while the rest (semantics, syntax and phonetics) is common to the species, since it’s biologically (or naturally) conditioned. Let’s call the cultural or variable part of languages the lexical interface. This label makes sense if we consider that this component has the function of connecting (i) the syntactic derivations that the computational system produces in interaction with the conceptual-intentional system with (ii) the sensorimotor system, which translates them into sounds (or visual signs).
In your blogpost, you compare functional-adaptive explanations in linguistics to Darwin’s theory of selection and adaption in biological evolution. I liked this analogy, because I made this comparison, too, in an earlier paper (Haspelmath 2004).
Yes, once we have the design space of the lexical interface clearly delimited, it is possible to consider whether the use of languages for communication and thought may have an influence on which features of them are more or less likely to be transmitted in a certain context (that is, adaptation). It is reasonable to think that since the lexical interface of languages changes from generation to generation, the same principles we use to explain the evolution of biological organisms can be analogously applied: inheritance, mutation and isolation. I myself have dealt with this matter in detail (Mendívil-Giró 2006; 2014). So yes: it is possible to provide functional-adaptive explanations for some properties of languages, but always within a design space bounded by the invariant part of the language faculty (see Mendívil-Giró (2018b) for a more detailed discussion of those limits within the context of a research topic on the adaptive value of languages).
OK, so we seem to agree then that a system of cultural linguistic conventions (= “a language”, in my parlance) can be functionally adapted, at least in principle. And from my (admittedly subjective) perspective, such functional adaptations are actually the most interesting part of general linguistics, because I feel that it is possible to make substantial progress. I’d like to add this to the five questions listed above, maybe as follows: “6. Why are languages the way they are? (Zipf’s question)”. Would you agree that this question could be added as an important kind of question for science?
Yes, definitely. This sixth question, in my opinion, would be complementary to question 1 (or even a part of it, given that a person’s language knowledge necessarily includes a lexical interface). It is the domain of historical linguistics and the study of linguistic change. Undoubtedly, an essential chapter of the science of language.
But I’m not so happy with the name you assign to it: I would prefer to call it Schleicher’s question (although Humboldt’s multifaceted work could cover the two questions).
In my (2004) paper I also observed that it was apparently possible to hit upon the right theory in biology without having a good account of the molecular mechanisms (which Darwin knew nothing about, as you note). So do you think that we could find the right (ultimate) functional-adaptive explanations in linguistics without having settled the question of the mechanics?
Of course, we do not need to wait for a complete answer to Humboldt’s question (question 1) to be able to continue progressing in the other questions, if only because it is possible that what we find out working on the other questions could also inform about the answers to question 1. In fact, Plato’s question (question 2, which corresponds to Tinbergen’s developmental/ontogenetic explanation) has also been central in GG (as is normal in an innatist theory), and progress has been made in interaction with work on question 1. The same can be said with part of Descartes’ question (question 3), which is very complex, since it includes two areas: the study of language processing in real time (what psycholinguistics usually studies, and which has also advanced in interaction with the work on questions 1 and 2), and the area of intentionality and free will, in which we are more in the dark (Chomsky tends to say that this is more of a mystery than a problem).
But progress on questions 4 and 5 depends crucially on progress on the previous ones. As vision scientist David Marr noted, the first objective of the cognitive scientist has to be the computational level, that is, the equivalent to question 1. Without a cognitive model of how vision works, it is impossible to discover how it is implemented in the brain (and therefore how it has evolved). A theory about the knowledge of language should serve as a guide in the study of the neuronal implementation of the cognitive organ of language and, of course, in the study of how that organ could emerge evolutionarily. If we go back to Tinbergen’s model (mentioned in your first question), I would venture to say that linguists, qua linguists, are limited to proposing proximate (mechanistic and developmental/ontogenetic) explanations. Only biologists with a solid background in linguistic theory (let’s call them biolinguists) could undertake ultimate explanations to Broca and Darwin questions. I believe that the time of biolinguists has not yet arrived, but that is a different and controversial issue.
OK, you have now talked about ultimate questions of the biological component of Human Language. But we agreed earlier that there is also a cultural component (which you call the “lexical interface”), and that this component, too, can be studied under the perspective of mutation and inheritance (what many people call “cultural evolution” these days, though you may not like this term, as you said that you don’t want to say that languages, or lexical interfaces, “evolve”). If so, then we can also have functional adaptation at this level, and my question was whether we need to settle the (biological) mechanics before we can provide functional explanations at the level of cultural evolution.
Of course this is not necessary. As I pointed out in the original post, Darwin did not have to wait to know the structure of DNA to formulate the theory of evolution. But here I want to make an important remark. Although I speak of a cultural component, I do not want to imply that languages are cultural objects (such as political institutions, religious creeds or financial systems). In my view, a language’s “lexical interface” is the equivalent of an organism’s genome. Note that all organisms share the same DNA molecule, but organize it into different genes. Similarly, all languages share the same structure but organize it (externalize it) into different lexical interfaces. So, I do not think we can say that phenomena like Persian DOM or Basque ergative alignment are social conventions, but they are “mutations” in the mental organs that pass from one generation to another.
Of course, these phenomena could have a functional/adaptive explanation, but the cognitive perspective we adopt requires taking other factors into account. As I point out in the papers mentioned, just as there are two ways of looking at natural evolution with regard to the role of adaptation in the final design of organisms (which we could popularly formulate as the Dawkins vs. Gould controversy), there is the same issue in the study of the history/evolution of languages (let’s call it the Haspelmath vs. Lass controversy). I’m sorry to say it, but I align myself with Gould/Lass on the idea that adaptation has a limited capacity to explain the structure of evolutionary objects (whether they are natural organisms or languages).
Well, the argument in my blogpost was that DOM phenomena are sufficiently explained in terms of cultural evolutionary adaptation. While you are right that Kalin (2018) cited the earlier functionalist work, I think she did not understand that the earlier work actually provides a sufficient explanation, and that therefore no “mechanistic” explanation is needed. My impression is that she simply did not think of the possibility of functional-adaptive explanations, because she took an exclusively biological perspective, as if cultural evolution had no explanatory value.
My impression is that Kalin is concerned with how to account for the linguistic knowledge of speakers of DOM languages: that is, what aspects of their mental grammars are different from those of speakers of languages without DOM, whereas she does not consider the matter of why there are languages with DOM and languages without it. What historical changes have led to the grammars towards one configuration or another is not discussed in her article, although obviously it is an important issue.
Do you have an idea how we could make progress in avoiding talking past each other, as has happened so often in the past (as you also note in your 2018a paper)?
I think it would be possible to explain many theoretical disagreements in our discipline as the result of a misunderstanding about what we mean by the word language (both in the mass and the count readings). For a generativist, a language is a biologically determined knowledge system with a variable cultural component (the lexical interface that serves for the externalization of language). That is why generativist authors cannot accept the assertion that languages are learned by imitation, that they vary without limit, that they adapt to the communicative needs of speakers, etc. But all these remarks would be much more acceptable if we talk about the lexical interface. Of course, for a functionalist a language is essentially a cultural object at the service of communication. But note that this definition coincides with a part of what a language is from the point of view of GG. Indeed, the lexical interface that is part of each language is a cultural object, which is learned inductively, varies profoundly, etc. As I point out in the (2018a) paper, if generativists reading a functionalist paper interpret the word language as ‘lexical interface’, they would find it much more interesting, and if functionalist authors reading a generativist paper understand the word language as ‘something that integrates a cognitive system deeper than the surface forms’, they would also find it more acceptable. For this reason, I would say that, to a large extent, generativist and functionalist research programmes are more complementary than incompatible, although many (most?) of their practitioners ignore it.
That sounds good, but “lexical interface” is a bit unwieldy. If we spoke French, could we say that the biological component is langage, and the cultural component is langue (= lexical interface)? And if so, couldn’t we simply say that Human Language (= le langage) is a biological capacity (subject to Darwin’s question), while particular languages (= les langues) are cultural objects that are subject to Zipf’s question (“Why are languages the way they are?”)?
This formulation is very attractive to me. Unfortunately, carrying out it in English is very difficult. And it could be misleading in general, that is, a return to the conception of languages as social institutions. Let me return to the analogy with biology to clarify in what sense I share your approach. If language is life, languages are organisms (e.g., animals). Organisms are forms of life, and languages are forms of language. The goal of biology is to explain what life is; The goal of linguistics is to explain what language is. But life does not exist in itself, it only exists in the form of organisms. And the same happens with language: nobody speaks language, we all speak languages. The biologist cannot study life without studying a particular organism, and the linguist cannot study language without studying a particular language. That is why, as you said above, all linguists (independently of our theoretical persuasion) spend 90 percent of our time studying specific languages (that is, ‘cognitive language organs with different lexical interfaces’). For both the biologist and the linguist it is crucial to distinguish which properties of the “individual” they study are general, and which are the contingent result of history. But in both cases the ultimate goal is to understand the mechanisms of life/language – although it is also very possible that the goals and interests of a molecular biologist are different from those of a zoologist. Of course, the first is not better than the latter, they are just different.
Many thanks for this interesting conversation – till next time!
Chomsky, Noam A. 1988. Language and problems of knowledge: The Managua lectures. Cambridge, MA: MIT Press.
Haspelmath, Martin. 2004. Does linguistic explanation presuppose linguistic description? Studies in Language 28. 554–579. doi:10.1075/sl.28.3.06has.
Kalin, Laura. 2018. Licensing and differential object marking: The view from Neo-Aramaic. Syntax 21(2). 112–159. doi:10.1111/synt.12153.
Mendívil-Giró, José-Luis. 2006. Languages and species: Limits and scope of a venerable comparison, in The Biolinguistic Turn: Issues on Language and Biology, eds. J. Rosselló, and J. Martín (Barcelona: PPU), 82–118.
Mendívil-Giró, José-Luis. 2014. What are languages? A biolinguistic perspective. Open Linguistics 1: 71–95.
Mendívil-Giró, José-Luis. 2018a. Is Universal Grammar ready for retirement? A short review of a longstanding misinterpretation. Journal of Linguistics, 54/4: 859-888. https://doi.org/10.1017/S0022226718000166
Mendívil-Giró José-Luis. 2018b. Why don’t languages adapt to their environment? Frontiers in Communication/Language Sciences 3: 24.
Tinbergen, Niko. 1963. On aims and methods in ethology. Zeitschrift für Tierpsychologie, 20: 410–433. https://doi.org/10.1111/j.1439-0310.1963.tb01161.x
It appears wrong to limit the value of studying the structure of variation among organisms/languages, which is what typology is, to merely functional explanation. Correlation among various traits and their distribution in a species can tell us about the pressures from the environment it adapts to, but it can also tell us about the mechanisms that produce or are produced by these traits (of course, it tells us many things about the species’ history as well).
In my view, whose assumptions might not be lighter depending on your perspective, the mechanism of language is functional. That is, the speaker actively adapts his language ability to communicating with other speakers, which consists of both learning the received language on one side, and doing things with language and meaning things on the other.
I imagine neural architecture that enables language use is developed and tuned by a collection of neurons attempting to minimize the prediction error of received stimulus. I could imagine that Universal Grammar is a structure that inevitably arises in any neural network attempting to parse any naturally occurring language, but I am sceptical of this as well.
Thus I don’t consider learning a language to be necessarily qualitatively different from learning to compose music, swim, do advanced mathematics, play tennis or a video game, although presumably language has a more complex neural structure which develops to parse (predict) it.
If there is indeed a “next time”, then I would like to see José-Luis quizzed on the following passage …
“let’s assume that the only variable part in languages is morphology and phonology (in fact, this is a common hypothesis in current GG), while the rest (semantics, syntax and phonetics) is common to the species, since it’s biologically (or naturally) conditioned. “