This post was prompted by a recent paper by Peter Ludlow (a Michigan/Illinois-based philosopher) on “the philosophy of generative linguistics” (2019), where he targets a 2010 paper of mine for criticism, and (quite flatteringly) pits me against Darwin. But he confuses general linguistic theory with language-particular theory, and as this confusion seems to be more widespread, it probably deserves some discussion.
Particular languages (p-languages) are immediately observable, in the sense that we can hear people talk in them (i.e. we can observe speech produced in them). But the general phenomenon of Human Language (g-language) is a more elusive concept – even in the 18th century, not all intellectuals were sure that all human groups had a p-language (see Levelt 2018), and it takes some abstraction to say that every p-language is somehow a manifestation of a more general notion of “Human Language”. In other domains, the analogous difficulties are serious: For example, are Christianity, Shintoism and Confucianism manifestations of a more general phenomenon of “Human Religion”? (And what about anthroposophy and Neotantra?)
Of course, everyone nowadays agrees that in addition to the various p-languages, there is also a general phenomenon Human Language, and we surely want to study it. But how do we go about doing this, given that it is not directly observable? There is strikingly little discussion of this question in the language sciences – as if it were clear that the study of some particular language (whether Greek, which was Aristotle’s p-language, or German, Heidegger’s p-language, or English, Pinker’s p-language) will automatically yield insights into Human Language. Wouldn’t we first have to make sure that the features of the p-language we are discussing are truly general across the world?
For some reason, many philosophers of language, and more recently even linguists, have not been worried by this problem and have simply ignored the variation. Just as Arnauld & Lancelot (1660) based their general discussion of grammar on French, Chomsky (1957) based his general discussion of grammar on English. But why should this be a respectable way to proceed? That has never been quite clear to me, and I suspect that this shortcut has been accepted primarily because our desire to speculate about Human Language has long been greater than our access to comparative data on the world’s languages.
(Of course, one good argument is the poverty of the stimulus: We are able to learn regularities that are not apparent from the speech we hear, and we do not need to study different languages to draw some general conclusions from this about our learning abilities.)
General philosophical linguistics was long marginal in our field (and clever books like von der Gabelentz (1891) and Gardiner (1927) had little impact), but this has changed since the 1960s – the idea that the study of a p-language can contribute to g-linguistics is almost universal by now. But strangely, linguists do not really seem to know how and why this is possible. Of course, when we find something highly unusual or unique in a language, then we can argue that our research has expanded the range of possibilities of Human Language (but this leads to less restrictiveness, and thus not really to any further understanding). Maybe we can use a single p-language to argue for theoretical innovations (such as probabilistic grammar, or exemplar-based models), but eventually, we will have to show that the theoretical innovations are equally relevant for all languages – because many theories do not generalize beyond the language for which they were designed.
In syntax and morphology circles, a common way in which linguists think that they can contribute to g-linguistics on the basis of a p-language is by adopting a widespread framework and modifying it slightly. For example, they can adopt the DP framework for the description of nominal phrases (familiar from the MGG community), and argue that an empty D can have novel (but not very surprising) effects (as in Erschler 2019). But this makes sense only if the framework is more than just a metalanguage – it must be claimed to be an innate set of categories and architectures (= a hypothesis of what a rich UG looks like), because otherwise there would be no theoretical point in arguing for an analysis in terms of the shared framework.
In a (2010) paper entitled “Framework-free grammatical theory”, I argued that linguists wishing to analyze/describe a single language (i.e. “p-linguists”) should not adopt one of these innate frameworks, but should build a custom framework for the purposes of their language (in other words: “describe each language in its own terms”, as urged by fieldworkers since Boas’s time). It cannot be excluded that there is a rich UG and that all languages work in terms of UG categories, but we are not close to having found the right categories, so it is best to be agnostic with respect to innate categories.
Ludlow (2019: 522) says that there are two different approaches of linguistic analysis: a bottom-up approach “that begins with the data and only then proceeds to theory construction”, and a top-down approach that collects data in the service of refining a theory. For the latter, he mentions a well-known comment made by at one point Charles Darwin (in a private letter):
“How odd it is that anyone should not see that all observation must be for or against some view if it is to be of any service!”
Ludlow thinks that in contrast to Darwin, I advocate bottom-up investigation, “beginning with theory neutral data collection”, and he implies that I want to be an “atheoretical observer”. This is wrong, and the mistake comes from overlooking there are two kinds of theories: p-theories (theories of particular languages) and g-theories (general theories of Human Language).
Darwin is of course right that observation must be in the service of theory, but what I am concerned with in the 2010 paper is p-theories: comprehensive descriptions of particular languages (such as my 1993 grammar of Lezgian). We may not often call grammatical descriptions/analyses “theories”, but they are certainly not atheoretical data collections. Everyone who has thought in some depth about p-linguistic analysis knows that it is a complex interaction of hypothesis formation, data collection for hypothesis testing, reformuation of hypotheses, and so on. P-linguistic analysis is an eminently theoretical enterprise, regardless of wider claims about g-linguistic implications that one might wish to make. Ludlow has one passage where this becomes clear: “Simply positing the existence of word boundaries assumes a lot of theory. Morphemes are not things that we nakedly observe; they are the product of scientific theorizing” (p. 523; this can be seen abundantly in my 2011 paper on word boundaries, where I discussed possible evidence for word hypotheses in detail).
Ludlow then goes on to discuss several specific parts of my paper, showing that they are not actually atheoretical, for example the categories “Predicate” and “Topic” that I make use of in the description of Tagalog clause structure (following Schachter & Otanes 1972). But I never said that they are “atheoretical” – what I say is that they are motivated by Tagalog facts, not taken from some pre-established framework that was originally motivated by different facts from different languages. The capitalization of the terms “Predicate” and “Topic” indicates that they are emic categories, specifically designed for the purposes of describing Tagalog. They are theoretical entities from a p-theory of Tagalog syntax. They are not taken from an innate framework of pre-established categories.
It seems that it never occurred to Ludlow that one might construct a theory of a p-language that does not necessarily have implications for g-language (even though I say it clearly in the 2010 paper). The only way in which he can conceive of what I am doing is as “atheoretical” – even though I explicitly say (in the title of the paper) that I am just as theoretical as everyone else in the volume (Heine & Narrog’s handbook of linguistic analysis).
Ludlow is not a prominent figure and I could simply ignore him, but my paper has been misunderstood by others (see here and here), so it is probably worth emphasizing again that all I’m saying is that p-theories should not make use of speculative ideas of what UG might possibly be like. And what is the source of the misunderstanding? I think it’s the sloppy failure to distinguish between p-linguistics and g-linguistics. One finds this in work by many of the best and most prominent linguists, e.g. in this quotation from Pinker & Jackendoff (2005: 202):
“The first [question] is which aspects of the faculty [of language] are learned from environmental input and which aspects arise from the innate design of the brain (including the ability to learn the learned parts). To take a clear example, the fact that a canine pet is called dog in English but chien in French is learned, but the fact that words can be learned at all hinges on the predisposition of children to interpret the noises made by others as meaningful signals.”
It is very odd to say that the word dog is part of the faculty of language, because this is a biological capacity, whereas the word dog is part of a tight set of linguistic conventions (namely the English language). The study of the capacity for language is part of g-linguistics, but the study of words like dog and chien is part of the p-linguistics of English and French, respectively. This is not controversial, but Pinker & Jackendoff’s sloppy formulation makes it appear as if there were no clear difference between biological capacity and particular conventions. If one talks like this, it is probably because one has forgotten that the study of language-particular coventions need not be relevant to general theories (but may require a specific p-theory, consisting of emic categories).
(P.S. The terminological distinction between “p-linguistics” and “g-linguistics”, first made in this earlier blogpost, may sound odd, but I increasingly feel that there is a real lexical gap in English, and that I need short terms for these two types of research to explain the source of misunderstandings.)