Philosophical issues keep coming up in linguistics, and although I have no particular training in this field, I have sometimes felt that I should make general comments about the nature of linguistic theorizing (e.g. in my 2010 paper on “framework-free grammatical theory”). Now I feel honored that the prominent philosopher Peter Ludlow has devoted a lengthy blogpost to the issue of how p-linguistics relates to g-linguistics and why framework-free science is not possible, in which he reacts to my earlier blogpost about p-linguistics and g-linguistics, where I had discussed his 2019 paper about the philosophy of generative linguistics. Here I continue this conversation.
Basically, Ludlow’s blogpost reiterates his 2019 point that “all observation must be for or against some [theoretical] view if it is to be of any service” (a quotation from Darwin), and he keeps comparing linguistics with several fields from the natural sciences such as geology, mineralogy and zoology. He finds the distinction between p-languages and g-language as irrelevant as the distinction between “p-rocks and g-rock” in geology, and he wonders how one can study things “in their own terms” at all:
“If you are studying something in “its own terms”, are you even studying? After all, one way to understand the study of the natural order is to understand how it fits into the greater scheme of things. That is, to study something is to learn how it is integrated with the natural order.”
Ludlow thus looks at the really big picture, and unlike in his 2019 paper, he does not really engage with questions of linguistics here. He tries to explain the difference between his and my view by suggesting that I do not properly distinguish between facts/phenomena and the object of study. Facts and phenomena are not the same as the object of study, he says, but provide evidence for it:
“For a generative linguist, facts about languages… provide evidence for the structure of the language faculty (UG).”
This is uncontroversial and is actually the same for a comparative linguist who is not committed to a particular (“UG-style”) view of linguisticality, but I did not understand why Ludlow does not admit that someone might be interested in, say, Colombian Spanish for its own sake. Why couldn’t a p-language be the object of study in one kind of scientific endeavor (p-linguistics), while at the same serving as a factual basis for another kind of endeavor (g-linguistics)? Ludlow seems to contradict himself when he says, on the one hand, that “other kinds of linguists may have other objects of investigation”, but on the other hand, that he doubts that “any serious scientific investigation into language would take p-languages … to be the ultimate objects of enquiry”. It may be that Ludlow is simply not aware that there are a large number of linguists who are primarily trying to figure out, in a scientific way, how a particular language works (e.g. how French works and how this can be laid out on 2600 pages).
In my earlier blogpost, I had emphasized that the structural uniqueness of languages makes it necessary to describe each language in its own terms, without squeezing it into the straitjacket of a pre-existing framework. Now Ludlow compares linguistics to biology and asks:
“The question of course is what does “studying things in their own terms” entail? For example, had Darwin studied barnacles in their own terms, would he have exorcised all thinking about evolution?”
Ludlow is certainly right that Darwin arrived at his findings about evolution after doing more than studying each species in its own terms – he was also, and perhaps primarily, engaged in worldwide comparative biology. His interests were thus similar to those of comparative linguists, and he even paid attention to comparative linguistics, as has often been remarked upon (e.g. Pagel 2017). But he also devoted himself to the details of barnacles, over many years, as recently described by biologist Lindell Bromham in a discussion of comparability in linguistics and biology (Bromham 2020: 4):
“Darwin spent eight years dissecting and describing barnacles, not only his own specimens, but also those sent to him by a worldwide network of corre- spondents. To produce a taxonomic description requires a heroic attention to detail, and Darwin’s barnacle work was all-consuming. His study at Down House became a shrine to barnacles, with boxes of borrowed specimens, pages of notes and detailed sketches, and the finest microscope a man of his comfortable means could acquire.”
This does not sound as if Darwin had a ready-made framework through which he described the different barnacle species, but it sounds more like the meticulous work of comparative linguists who first need to describe their languages in some detail before they can compare them. Darwin surely did not work with a maxim like the following (styled on Chomsky’s (2001: 2) Uniformity Principle):
Uniformity Principle for Barnacles
In the absence of compelling evidence to the contrary, assume barnacle species to be uniform, with variety restricted to easily detectable surface properties.
The reason Darwin went on his 5-year trip around the world was precisely that he was open-minded and was willing to be surprised by novelty. In actual fact, there existed something like a universalist framework of biology, sometimes called “transcendental anatomy”, which argued that there is an “ideal plan” or “archetype” in the animal and plant kingdom. Well-known representatives of this movement (of the first half of the 19th century) were Richard Owen, Étienne Geoffroy St. Hilaire, and the German poet and naturalist J.W. von Goethe. (Goethe even created a picture of an Urpflanze, showing the archetype of a plant.) In this approach, which was opposed to the earlier “functionalism” or “adaptationism” of William Paley and Georges Cuvier, the idea was that each plant or animal was a slight variation upon a basic underlying form, reminiscent of Chomskyan “universal grammar” and its parameters (I wrote an earlier blogpost about the intriguing parallels between the debates among 19th century biologists and the debates of late 20th century linguists). And at the time when Darwin set out on his voyage on the Beagle, the Linnean “System of Nature”, which had provided a prestigious framework for decades, was still very much alive. After Darwin’s proposals had been confirmed over and over again, and widely adopted by biologists, the oppsosition between the transcendental anatomists and the adaptationists disappeared, because Darwin showed how his theory could explain both adaptation and homology.
So was it because he worked in a “framework-free” way (rejecting “transcendental underlying forms”) that Darwin made such phenomenal progress, in contrast to his predecessors and contemporaries? I wouldn’t go so far. Darwin’s innovations were possible also because of a better understanding of deep time (and a reduced influence of biblical thinking), and he was primarily interested in worldwide comparison and in deep understanding, not so much in describing particular species. And we should not forget: His success was first and foremost in understanding phylogenies and adaptation – the sorts of things that are being studied in historical linguistics (e.g. Lass 1997) and in functional linguistics (e.g. Givón 2010). The still deeper understanding that modern biology has achieved through the study of genetics, especially at the molecular level, was not yet available in Darwin’s time.
Elsewhere I explained why p-linguistics should be framework-free, but in the present context it is interesting to ask, following Ludlow’s blogpost, whether this applies to other fields as well. I would answer that it mostly does not apply to natural sciences such as geology, mineralogy and zoology, because these fields work with natural kinds – they engage in “naturalistic inquiry” in this narrow sense. In my 2018 paper on comparative concepts, I contrasted comparison via natural kinds with comparison via comparative concepts. Ludlow simply ignores comparative concepts and makes it clear that what he understands by scientific inquiry is the identification of natural kinds, and this is of course the generative approach. (I’m glad that Ludlow agrees with me here, because I had an interaction with another philosopher, Matthew Spike, who did not agree with my framing of the generative approach in his 2020 paper.)
But in addition to natural kinds, the sciences also sometimes work with comparative concepts, such as “planet” (a concept famously sharpened in 2006 in such a way that it no longer includes Pluto). And especially in behavioural biology, we may sometimes want to describe phenomena in their own terms, without presupposing the existence of a particular set of natural kinds. For example, Cheney & Seyfarth’s famous study of alarm calls in vervet monkeys (Seyfarth et al.1980) did not assume that particular sets of calls are possible or impossible, in advance of studying these animals. Quite generally, I am not aware of any other branch of behavioural biology that is like linguistics in that it suggests that animal behaviours should be described in terms of pre-established frameworks (along the lines of Government-Binding, Relational Grammar, RRG, or LFG).
For some reason, Ludlow seems to think that what I suggest implies “siloing a field of investigation off from broader theory and from integration with other sciences”. Worldwide comparison and larger questions about human language have long been my main interest. But I have questioned whether one can reach this goal by adopting one of the several currently popular frameworks, by applying it to some language data, by tweaking the framework a little, and then moving on to the next cycle. Language-particular work of the sort that I admire particularly (e.g. references grammars like those published in the CoGL series) is evidently theoretical but does not adopt an off-the shelf framework – hence I describe such work as working with “framework-free grammatical theory”. Such grammatical descriptions are not “siloed off”, and neither p-linguistics nor g-linguistics is prevented from integration with other sciences. But both are free to create their own concepts, rather than being forced to adopt a set of currently fashionable ideas about what the universal building blocks of the grammar blueprint might be. Eventually, we may find that some of them are indeed part of our innate “core knowledge”, but this is an empirical question, and we will not answer it if we start out by assuming that all languages are uniform (“in the absence of compelling evidence to the contrary”, Chomsky 2001).
References
Bromham, Lindell. 2020. Comparability in evolutionary biology: The case of Darwin’s barnacles. Linguistic Typology 24(3). 427–463. (doi:10.1515/lingty-2020-2056)
Chomsky, Noam A. 2001. Derivation by phase. In Kenstowicz, Michael (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: MIT Press.
Givón, T. 2010. The adaptive approach to grammar. In Heine, Bernd & Narrog, Heiko (eds.), The Oxford handbook of linguistic analysis, 27–49. Oxford: Oxford University Press.
Haspelmath, Martin. 2010. Framework-free grammatical theory. In Heine, Bernd & Narrog, Heiko (eds.), The Oxford Handbook of Linguistic Analysis, 341–365. Oxford: Oxford University Press.
Haspelmath, Martin. 2018. How comparative concepts and descriptive linguistic categories are different. In Van Olmen, Daniël & Mortelmans, Tanja & Brisard, Frank (eds.), Aspects of linguistic variation: Studies in honor of Johan van der Auwera, 83–113. Berlin: De Gruyter Mouton. (https://zenodo.org/record/3519206)
Lass, Roger. 1997. Historical linguistics and language change. Cambridge: Cambridge University Press.
Ludlow, Peter. 2019. The philosophy of generative linguistics: Best theory criteria. Current approaches to syntax: A comparative handbook, vol. 3, 521–547. Berlin: De Gruyter Mouton. (doi:10.1515/9783110540253-019)
Ludlow, Peter. 2021. On p-language, g-language, and Naturalistic Inquiry. Medium (https://peterludlow.medium.com/on-p-language-g-language-and-naturalistic-inquiry-d22a145f6287)
Pagel, Mark. 2017. Darwinian perspectives on the evolution of human languages. Psychonomic Bulletin & Review 24(1). 151–157. (doi:10.3758/s13423-016-1072-z)
Seyfarth, Robert M. & Cheney, Dorothy L. & Marler, Peter. 1980. Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Animal Behaviour 28(4). 1070–1094. (doi:10.1016/S0003-3472(80)80097-2)
Spike, Matthew. 2020. Fifty shades of grue: Indeterminate categories and induction in and out of the language sciences. Linguistic Typology. De Gruyter Mouton 24(3). 465–488. (doi:10.1515/lingty-2020-2061)
I think there’s a significant difference between linguistics and many areas of biology, which is that the latter are built on top of observations of entities which we are heavily pre-adapted to manage (since we live in part by eating them if they are tasty and catchable, not eating them if they are poisonous, and running away from them if they want to eat us and capable of doing that, and our ancestors have been doing this for a rather long time). All you need to be Darwin is a peculiar combination of the interests of many children, the steadiness of a real grownup, and enough $$ to have a big house to house your collection and travel to places to acquire it. Plus literacy to attach names, dates and places to the collected items.
To even perceive syntax, on the other hand, you need a kind of mathematics which is not very useful until technology has risen to a considerable level, though a version of it did manage to appear in Ancient India without anything that we would recognize as a useful application. Otherwise you just have an unintelligible mess of ‘this sequence of words means this, that sequence of words means that’, over and over again, which we are not preadapted to make explicit sense of.
Another big difference, related to the first, is that biology supervenes on many other disciplines, such chemistry, physics, and thermodynamics, as well is the original ‘common sense’ abilities that enabled us to catch our dinners and avoid those who wished to dine upon on us, so that if a colony of lizards is discovered in an outcropping that is proposed to be surrounded by a suburb, we know that the critters need to eat, so that if they are to survive, there must be a sufficient population of things to eat there, either self-sustaining, or capable of flying in, and we also know that the lizards can’t either fly over the suburbs to hunt or live elsewhere, or tunnel under it. At a more advanced level, it might be possible to ascertain whether the population has been stably isolated for a long time, or dependent on gene flow from nearby ones.
In syntax, none of this is possible: the only useable fact we get from the rest of science is that grammar must be transmitted via the child’s experiences in their surroundings rather than by telepathy; as Chomsky has noted, we can’t even assume that grammars fall into a nice computational complexity class, although it is certainly worth exploring the possibility that they do (maybe people just avoid the intractable cases, like map-makers for computer games such as DooM in the 1990s, who learned to avoid making maps that would make trouble for the (intractable) algorithms that made them playable.
So I don’t think that framework-free syntactic description is possible, although what is possible is not to believe that your chosen framework is da best and that anybody who chooses something different is a perverse idiot. A slightly different way of putting it is that syntactic work needs *a* framework which requires some explicit construction, without much help from elsewhere, whereas biology can draw on a vast range of preexisting frameworks, some of which have a strong innate component due to the lifestyle of our ancestors, others developed by people ranging over varied timespans of up to millennia in the case of aspects of mathematics.
Many thanks for these interesting reflections, Avery! Yes, syntax is harder than phenomenological biology, which is also why Aristotle made a lot less progress on it than on biology. And yes, we need very abstract (quasi-)mathematical concepts for syntax, and we can call them a “framework”. But recall that I actually don’t say that we should do syntax *without a framework*, but merely that we should not expect a *universal framework*. So I think that all you say is compatible with what I say. In HPSG, there is explicit recognition that the technical machinery is primarily a very sophisticated notation and is not meant to be innate. It’s different with some LFG writings, e.g. Dalrymple (2001: 3): “the abstract syntactic structure of languages obeys universal principles of functional organization and draws from a universally available set of possibilities” – this does sound like the “innate grammar blueprint” idea. But you probably wouldn’t subscribe to it but adopt a point of view that is more like HPSG.
A bit more; something that would make syntax more like geology or biology would be bigger corpora, but there’s a problem of a mixed empirical and theoretical nature that we don’t really know how big a corpus has to be to illustrate all of the constructions of a language, and thereby be somewhat comparable to a rock that can be put in a drawer in a museum and later studied in comparison with other rocks.
Speculating wildly, I note that Hart and Risley (1995) proposed that upper SES children get 3 million ‘words’ per year, lower SES about a third of that. This ‘word gap’ has been challenged (I’ve lost the reference to a criticism of it that seemed very cogent at the time), so let’s assume 3mw/year for all, and also that a child can learn 3 syntactically different languages quite thoroughly by the age of 10, but toss the first hear of input as irrelevant to syntax, giving us a suggested figure of corpus size=30mw, a bit more than twice as big as CHILDES English, which contains few or zero examples of some reasonably interesting things that are clearly English. Which would need to have reasonably decent translations and situational annotations (traditional story told in traditional setting, rant on facebook, academic position paper, horror movie subtitles, Scandinavian detective story , …). That is far out of reach for most languages. But maybe we could justify a lower minimum size for a corpus to be ‘complete’.
For example one thing that I found to be missing in CHILDES is triple possessive recursions such as “John’s mothers’ dog’s name”, but if we accept a UG that predicts this from the existence of double recursive possessive such as “John’s mother’s dog” and “John’s dog’s name”, of a form that does occur in CHILDES. Then the absence of the triples doesn’t mean that CHILDES isn’t big enough.