Languages exhibit hugely more diverse phenomena than are displayed in well-studied European families. However, citing a collection of exotica does not prove Evans & Levinson's (E&L's) claim that “it's a jungle out there” (sect. 3, para. 17). Examining languages more closely, or at a higher level of abstraction, often reveals critical similarities which superficial descriptions can obscure. Moreover, languages frequently employ distinct grammatical strategies to achieve parallel outcomes; thus, the universal is the end result, not the means of achieving it. Finally, unrelated languages often “choose” the same strategy, despite the lack of a single universal solution, suggesting that homogeneity is widespread.
Lexical category distinctions (sect. 2.2.4)
Certainly, there is no invariant set of lexical or functional categories. But it remains to be demonstrated that a language may lack any distinctions between lexical categories, or, more specifically, may lack a noun/verb distinction. E&L note that languages of the Pacific Northwest Coast are frequently claimed to have no noun/verb distinction, illustrating with Straits Salish. Similar claims have been made for a nearby, unrelated family, Southern Wakashan (e.g., Makah, Nuuchahnulth). Here, nouns can function as predicates (i.e., not only arguments) and bear predicative inflections, including tense, aspectual, and person/number marking, and verbs can function as arguments (i.e., not only predicates) and bear nominal inflections, including determiners; (1) and (2) give Nuuchahnulth examples from Swadesh (Reference Swadesh1939):
1. mamuuk-maa quu
as-iwork-3s:indic man-the
“The man is working.”
2. quu
as-maa mamuuk-iman-3s:indic work-the
“The working one is a man.”
Thus, nominal and verbal roots cannot be identified either by distribution or morphology. Additionally, essentially any lexical root in Nuuchahnulth, including (the equivalents of) nouns, adjectives, and quantifiers, can take verbal inflectional morphology, superficially suggesting that all words are predicative, and thus that there is no noun/verb distinction. Immediate evidence against this (Braithwaite Reference Braithwaite2008) is that verbs only function as arguments when a determiner is present, whereas nouns function as arguments even without a determiner.
Close inspection reveals further behavioral differences between noun and verb roots (Braithwaite Reference Braithwaite2008). For instance, proper names can take nominal inflections, such as the definite -
i, shown on noun and verb stems in (1) and (2), but cannot take the third singular indicative verbal inflection -maa:
3. *Jack-maa
Jack-3s:indic
(“He is Jack.”)
Names, a subclass of nouns, therefore cannot be predicates, clearly distinguishing them from verb roots.
Moreover, although both nominal and verbal predicates can bear possessive markers, nominal predicates with possessive morphemes display a systematic ambiguity in terms of which argument an accompanying person marker is understood to refer to, whereas verbal predicates display no such ambiguity. A similar ambiguity arises in tense marking. Verbal predicates in Nuuchahnulth display a past tense suffix: -(m)it:
4. mamuuk-(m)it -(m)aħ
work-past-1s.indic
“I was working.”
This suffix also appears on nouns. Even nonpredicative nouns, including names, can bear tense morphology, apparently supporting the lack of a noun/verb distinction:
5.
aħaaa qaħši–'a mista-(m)itand.then die-eventive Mista-past
“Then (the late) Mista died.”
qaħšiThe past-tense marker -(m)it on the name conveys the specific meaning “former”; since names cannot be predicative in Nuuchahnulth, as (3) shows, this is evidently not a nominal predicate. However, past-tense markers also attach to nominal predicates, which are then interpreted in one of two ways: (6) shows a past-tense nominal predicate, exactly parallel to (4), except with a noun root; (7) displays a predicate nominal in which -(m)it bears the alternative “former” meaning:
6. quu
as-(m)it-(m)aħperson-past-1s.indic
“I was a man.”
7.
uunuu ani uumiik-(m)it-qabecause that whaler-past-subordinate
“because he was a former whaler”
Critically, -(m)it on a verbal predicate never exhibits the “former” meaning but is always interpreted simply as past tense. In sum, careful investigation such as that of Braithwaite provides ample evidence for a noun/verb distinction in Wakashan languages, despite superficial appearances.
Constituent structure (sect. 5)
As E&L note, “non-configurational” languages display free word order and discontinuous constituents: in (8), from the Australian language Kalkatungu, the underscore shows the components of the ergative subject, and italics show the (nominative) object:
8. Tjipa-yi tjaa kunka-(ng)ku pukutjurrka lhayi nguyi-nyin-tu.
this-erg this branch-erg mouse kill fall-participle-erg
“The falling branch hit the mouse.” (Blake 2001, p. 419)
E&L state that “the parsing system for English cannot be remotely like the one for such a language” (sect. 2, para. 3), because case-tagging indicates relationships between words, rather than constituency and fixed word order. But, in fact, the parsing system for English is well used to non-local dependencies – that is, to relating items not contiguous in the string. Note the discontinuous constituents in the following examples, and that the dependency even occurs across a clause boundary in the second instance: A student sauntered in wearing a large fedora; Which girl did you say he gave the books to __?. Parsing in Kalkatungu (or Latin) therefore utilizes a strategy also found in languages which do have clear constituents. Moreover, completely unrelated non-configurational languages like Kalkatungu and Latin share the same method of signaling relationships between words (case-marking). All this is hardly indicative of the jungle E&L assume; rather, it is evidence that very few solutions are available, and that languages make differential use of options from a small pool of possibilities.
Furthermore, certain non-configurational Australian languages (e.g., Wambaya; Nordlinger Reference Nordlinger2006) actually have one strict word order requirement, namely that the auxiliary is in second position, thus either second word, (9), or second constituent, (10) (Hale Reference Hale, Anderson and Kiparsky1973 outlines the parallel requirement in Warlpiri):
9. Nganki ngiy-a lurrgbanyi wardangarringa-ni alaji
this.erg 3sf-pastgrab moon-erg boy
“The moon grabbed (her) child.”
10. Naniyawulu nagawulu baraj-bulu wurlu-n duwa.
that.dual.nom female.dual.nom old.person-dual.(nom) 3.dual-prog get.up
“Those two old women are getting up.”
Crucially, the auxiliary cannot appear as, say, third word within a four-word noun phrase. Contra E&L, this demonstrates the psychological reality of word order and of constituent structure in such languages. Moreover, while by no means universal, second-position phenomena occur widely (e.g., Sanskrit, Celtic, Germanic), demonstrating remarkable formal homogeneity cross-linguistically.
Finally, E&L claim linguistic diversity is not characterized by “selection from a finite set of types” (sect. 8, para 9, their thesis 3). Case-encoding systems are few indeed, and familiar strategies (such as ergativity) even occur in language isolates such as Basque.
