Does Universal Grammar Theory Imply that Language Are All the Same?—Response to Vyvyan Evans (Part 2)
In the previous post, I began laying out a response to Vyvyan Evans’ article in Aeon and more generally in his book, The Language Myth: Why Language Is Not an Instinct. That earlier post focuses on whether language is in instinct. In this post, I address what Evans considers “a more important problem”; he writes:
“If our knowledge of the rudiments of all the world’s 7,000 or so languages is innate, then at some level they must all be the same. There should be a set of absolute grammatical ‘universals’ common to every one of them. This is not what we have discovered. Here’s a flavour of the diversity we have found instead.”
He then proceeds to catalogue several points of differences across languages, to which I shall return below. This is an odd sort of counter-argument for Universal Grammar (UG) theory. If what its opponents have discovered is that languages differ from each other in sounds, words, meanings, and grammatical patterns, that is not much of a discovery! After all, diversity of languages is “the most striking thing about the study of languages”, as Ferdinand de Saussure, dubbed “the father of modern linguists”, observed in his Course in General Linguistics. Differences among languages are so conspicuous that people must have noticed them very early on. “Even savages grasp them”, de Saussure remarked, “thanks to their contacts with other tribes that speak a different language”. Ancient Greeks recorded linguistic diversity, although they were more concerned with Hellenic dialects than with the various languages of “barbarian” outsiders. Virtually all ancient and medieval works on language either describe one particular language or highlight differences across several tongues. Most modern language textbooks, dictionaries, and grammar descriptions focus on the dissimilarities as well. Real science begins when things contrary to common wisdom are detected, such the fact that the Earth rotates around the Sun and not vice versa. In this sense, UG theory made a radical progress, compared to what came before.
An even bigger problem with the way Evans structures his argument is that he misrepresents his opponents’ theory as making a claim so silly that it doubtlessly must be wrong. Consider, for example, his description of UG theory in The Language Myth; Evans writes (boldface mine):
“This idea is often referred to as Universal Grammar: all human languages, no matter the variety we happen to end up speaking, are essentially the same. Whether someone learns English, Japanese, Swahili, Tongan or whatever, when you get down to it, they are all alike. Sure, each of these languages has different vocabularies. And each language makes use of a different, although partially overlapping, set of sounds. But underneath it all, the essential ingredients of language – our grammar – is pre-programmed in the human genome…” (p. 2)
Even a casual observer of language cannot help noticing that it is not only vocabularies and sets of sounds used to produce the words that differ from language to language, as grammars vary widely too. So do the proponents of UG theory claim that all human languages have “essentially the same” grammar? Not at all! Some grammars require the verbs to appear in the beginning of a sentence (Welsh), others at the end (Japanese), and yet others somewhere in the middle (English). (Image on the left is taken from here, but note that Russian is incorrectly classified as SOV.) Some grammars require “weather” sentences (e.g. ‘It rains’ or ‘It snows’) to have a meaningless subject ‘it’ (English and French), while others do not (Spanish and Italian). Some grammars require all question words (the “who”, “what”, “where” etc.) to be placed in the beginning of a question (Russian and Serbo-Croatian), other grammars require (and allow) only one question word to be in the beginning (English), and yet others do not require any question words to appear at the beginning of the sentence (Japanese and Mandarin). Some grammars place the verb before adverbs meaning ‘often’, ‘always’, ‘sometimes’ and the like (French and Italian), other grammars place verbs after such adverbs (English and Russian). It
will surprise Evans’ readers to find out that these and other differences across languages were actually discovered and described in minute detail by proponents of UG theory, not its opponents. I would go as far as suggest that the idea that all languages are “essentially the same” is so absurd that only someone hell-bent on fabricating a disingenuous critique of UG theory can claim that that is what “Universal Grammar” means. Needless to say, this strategy does nothing to promote an honest discussion and is at best misleading.
Of all the grammatical universals postulated by advocates of UG theory, Evans considers just one: recursion, which allows the creation of potentially infinite strings out of a finite set of elements. Evans writes:
“It […] remains unclear whether it is really universal among human languages. A number of researchers have suggested that it might in fact have emerged rather late in the evolution of human grammar systems, a consequence, rather than a cause. And In 2005, the US linguist-anthropologist Daniel Everett has claimed that Pirahã – a language indigenous to the Amazonian rainforest – does not use recursion at all.”
Everett’s claims have raised a heated controversy, some of which has been discussed in my earlier posts (see here, here, and here). The latter post, in particular, deals with the issue of recursion; for a more technical discussion, the reader can turn to Nevins, Pesetsky & Rodrigues (2009).
Let’s now turn to some of the other differences across languages listed by Evans as potential arguments against UG theory. His first point of difference is that “spoken languages vary hugely in terms of the number of distinct sounds they use, ranging from 11 to an impressive 144 in some Khoisan languages (the African languages that employ ‘click’ consonants)”. This difference, however, pertains not to grammar as such, but to the range of “building blocks” on which a grammar operates. But it is interesting to note that even in regards to sound inventories, there are some universal patterns. All languages use both consonants and vowels. All languages employ more consonants than vowels. All languages allow syllables that have a consonant followed by a vowel (CV), but not all languages allow the reverse: VC syllables. If one assumes that there can be no innate universals about language, as Evans does, such generalizations must be brushed off as accidental (unless they can somehow be explained as consequences of our particular anatomy and physiology).
Evans’ second point of cross-linguistic differences concerns word order; he writes:
“[Languages] also differ over the word order used for subject, verb and object, with all possibilities being attested. English uses a fairly common pattern – subject (S) verb (V) object (O): The dog (S) bit (V) the postman (O). But other languages do things very differently. In Jiwarli, an indigenous Australian language, the components of the English sentence ‘This woman kissed that bald window cleaner’ would be rendered in the following order: That this bald kissed woman window cleaner.”
The description is correct, but several important points have been glossed over. While Evans’ claim that languages “differ over the word order used for subject, verb and object” is true, it is also true—though ignored by Evans—that all languages distinguish these elements in the first place. For starters, consider Evans’ claim that “Straits Salish, an indigenous language spoken in and around British Columbia, gets by without nouns and verbs”. Indeed, claims have been made that Salish languages do not distinguish nouns and verbs. But even such claims do not amount to saying that Salish languages lack both nouns and verbs. Rather, it has been suggested that these languages have one category, which can be labeled either “nouns” or “verbs”, depending on one’s analysis. (Baker 2003: 89 suggests that these words are better conceptualized as verbs than as nouns, but see pp. 175-189 of his book for a more detailed discussion of lexical categories in Salish.) More generally, having defined what a verb is, Baker (2003: 88-94) asks if there are any verb-less languages. The “most serious candidate that [he] ha[s] found for a verbless language is the Australian language Jingulu”, originally described by Rob Pensalfini. After a careful consideration of Pensalfini’s arguments and data, Baker concludes that even Jingulu has a lexical category of verbs. (Evans also claims that some languages lack adjectives; this possibility is considered and ultimately rejected by Baker 2003: 238-263.)
In addition to distinguishing lexical categories such as nouns, verbs, and adjectives, all known languages distinguish grammatical functions of “subject” and “object” in that at least some rules in each language treat them differently. This non-interchangeability of subjects and objects is called “configurationality”. In some languages, like English, configurationality is easily observed from word order: subjects must precede the verb, and objects must follow it. (I am leaving aside some peculiar structures that need additional treatment.) The sentence John devoured Brussels sprouts is fine, but the reverse—Brussels sprouts devoured John—is not (unless we imagine some nightmarish scenario in which Brussels sprouts is the subject and John is the object). The bottom line is that in English word order rules make reference to subjects and objects as distinct categories.
But this is not the case for all languages: for example, indigenous Australian languages like Jiwarli (mentioned by Evans) and Warlpiri are known for their free word order. For example, in Warlpiri ‘The child is seeing me’ can be rendered with the same Subject-Auxiliary-Verb-Object word order as in English (S-Aux-V-O: Kurdu-ngku ka-ju nya-nyi ngaju). But the order of subject and object can also be flipped: O-Aux-V-S Ngaju ka-ju nya-nyi kurdu-ngku. (These examples are taken from Baker 2001: 409.) This observation led some linguists to claim that Warlpiri, Jiwarli, and other similar languages are non‑configurational, namely that their grammars do not distinguish subjects and objects but rather treat them as one and the same.
However, subsequent research has shown that subjects and objects are not indistinguishable even in languages of the Warlpiri/Jiwarli type, which allow flipping them around as far as the word order is concerned. While word order rules may be “blind” to the subject/object distinction in these languages, some other grammatical rules, for example, those concerning anaphora and coreference, do make that distinction. To understand what is meant by “anaphora and coreference”, let’s consider some English examples first. Certain English words, such as himself or each other, depend in their interpretation on other words in the sentence. For example, if I say John praised himself, it means that John is being praised, but if I say Bill praised himself, it means that Bill is being praised. The same word, himself, changes its meaning depending on other words in the sentence, in this case the subject.* Words like himself are called “anaphors” and their meaning dependence on some other word in the sentence is called “coreference” (in particular, himself co-refers with either John or Bill in the examples above.)
Now we come to the crucial observation: anaphors in English can be objects co-referring with subjects, but not vice-versa: John praised himself is fine, but *Himself praised John is a word-salad. (The technical term is “ungrammatical” and this is marked by an asterisk in front of the bad sentence.) One might expect *heself to be a subject anaphor form (by analogy with him/himself, he/*heself), but that form does not exist—because a subject could never be an anaphor in English. Thus, rules governing the distribution of anaphors in English “care about” which is the subject and which is the object.
Guess what? When it comes to anaphors and coreference, the same asymmetry between subjects and objects obtains in the allegedly non-configurational languages, including Warlpiri and Jiwarly, whose grammars are presumed to be “blind” to the subject/object distinction! For example, the Warlpiri analog of *Heself sees the man is as ungrammatical as the English sentence (Baker 2001: 411). This means that the grammar of Warlpiri is not, after all, insensitive to the distinction between subjects and objects.
Similarly unhelpful to opponents of UG theory are languages like Inuktitut, which Evans describes as “build[ing] ‘sentences’ by creating giant words from smaller word-parts”. In this language, an entire sentence such as ‘Do you have any tobacco for sale?’ is expressible in one word, he claims: tawakiqutiqarpiit. Languages such as Inuktitut are called “polysynthetic” because they “synthesize” (i.e. put together) “poly” (many) morphemes. It is not true, however, that in polysynthetic languages “each word is an entire sentence”, as Evans claims: in addition to one-word sentences such as tawakiqutiqarpiit, there are multi-word sentences in Inuktitut. But even with multi-word sentences, Inuktitut and similar languages (e.g. Mohawk) do not “use word order to signal who is doing what to whom”. Subjects, objects, and verbs can appear in any order imaginable. Like Warlpiri and Jiwarli, Inuktitut and Mohawk were once considered non-configurational languages, namely languages where the subject/object distinction does not apply.
But it turned out that even in polysynthetic languages some grammatical rules are sensitive to the subject/object distinction. Recall that the defining property of these languages is that they can “lump” (technical term: “incorporate”) nouns into verbs creating “giant words”. Well, only objects—but crucially not subjects!—can be “lumped” into the verb. As an example (taken from Baker 2001: 419), consider how the sentence ‘The stone broke the window-pane’ can be expressed in Mohawk. Besides, a three-word sentence, where both the subject and the object are independent words, this meaning can be expressed by a two-word sentence that translates roughly as “STONE INDEED-IT-PANE-SHATTER-CAUSED-ONCE” (hyphens indicate that these are morphemes of one giant word). Note that the noun meaning ‘window pane’ (“PANE”) is inserted into the giant verb word, between a bunch of prefixes—including a factive prefix (which, for purposes of illustration, I rendered as “INDEED” though that hardly does it justice) and an agreement prefix (which I rendered as “IT”)—and the verb root ‘shatter’. The factive prefix (“INDEED”), the causative suffix (“CAUSED”), and the punctual suffix (“ONCE”) tell us that this giant word is indeed a verb, and the agreement prefix (“IT”), expressing agreement with a neuter singular subject—and no agreement with an object—tells us that it is an intransitive verb. The subject, “STONE”, is expressed by a separate word.
If Mohawk were a truly non-configurational language, whose grammar treats subjects and objects the same, we would expect that the same meaning could be expressed by the reverse of the sentence described above, namely with the subject “STONE” lumped into the verb and the object “PANE” forming an independent word: “PANE INDEED-IT-STONE-SHATTER-CAUSED-ONCE”. However, this sentence is ungrammatical in Mohawk.** The overall generalization is that “objects but not subjects can be incorporated into the verb in Mohawk” (Baker 2001: 419). Thus, we have seen that the structural notions of subject and object appear to be relevant even for the languages like Warlpiri or Mohawk, which do not wear their configurationality on their sleeve, so to speak.
Readers of Evans’ article may be surprised to find out that the generalizations described above—that in Warlpiri anaphors cannot be subjects co-referring with objects and that in Mohawk objects and not subjects can incorporate—were discovered and described not by opponents of UG theory, who seem to put cross-linguistic diversity on a pedestal, but by advocates of the theory: Ken Hale and Mark Baker. Moreover, these facts would have probably gone unnoticed if advocates of UG theory had not been looking for evidence of configurationality in such languages where they least expected to find it, such as free-word-order languages of the Australian aboriginal type and polysynthetic languages of the Americas. In this respect, UG theory is methodologically advantageous in that it prompts linguists to discover facts that would otherwise remain overlooked.
Ironically, opponents of UG theory (including Evans) seemingly like to emphasize cross-linguistic diversity, mentioning lesser-known languages and their unusual features at every chance they get. Yet, in reality, they almost always ignore the nitty-gritty details of such seemingly “exotic” patterns as the distribution of anaphors in Warlpiri or incorporation in Mohawk, described above. They have little use for these facts, or for the more general observation that even the seemingly non‑configurational languages turn out to be configurational after all. The anti-UG position does not inspire one to find such universal truths about language. But even if opponents of UG could stumble upon such a generalization accidentally, they would probably shrug their shoulders (“It’s an accidental coincidence, meaning nothing!”) or brush it under the carpet as an inconvenient truth.
Finally, I would like to challenge another conclusion a lay reader may draw from this debate, namely that proponents and opponents of UG theory are like optimists and pessimists, one side seeing the proverbial glass as half-full and the other as half-empty. Evans and like-minded challengers of UG theory focus on diversity among languages and either deny or ignore the existence of grammatical universals (or diminish their importance), whereas supporters of UG theory seemingly focus on the similarities across languages rather than the differences. In reality, UG theory embraces both similarities and differences. Its main conjecture is that languages are different in a constrained way. As much as languages differ from one another, certain types of “languages” one could imagine or construct artificially are not found among the nearly 7,000 languages spoken today or in the scores of languages known to have been spoken in the past. An important premise of UG theory is that these gaps in linguistic diversity—that is “languages” that are imaginable but do not actually exist—occur for a reason, as they are outside the bounds of what is possible for natural human languages. It is those bounds that we call “Universal Grammar”. According to this key hypothesis, natural languages cannot violate the laws of Universal Grammar any more than physical objects can violate the laws of physics. But within the bounds of UG, a broad range of variation is possible—and indeed observable. Figuring out exactly what the bounds of UG are is one of the main goals of the advocates of UG theory.
_____________
*Similarly, The professors praised each other means that professors are being praised, and The students praised each other means that students are being praised. Thus, the interpretation of each other also depends on other words in the sentence, here the subject.
**Actually, like the English “Brussels sprouts” example above, this sentence is grammatical with a bizarre meaning that the window pane broke the stone.
Additional Sources:
Baker, Mark C. (2001) The Natures of Nonconfigurationality. In: Mark Baltin and Chris Collins (eds.) The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell. Pp. 407-438.
Baker, Mark C. (2003) Lexical Categories. Verbs, Nouns, and Adjectives. Cambridge, UK: Cambridge University Press.
Nevins, Andrew; David Pesetsky, and Cilene Rodrigues (2009) Pirahã exceptionality: A reassessment. Language 85(2): 355-404.