4.2 Semifunctional items in Chinese
4.3.1 Root support and analyticity
4.3.1.3 Is there an analyticity parameter?
Huang (2015) studies the high analyticity of Mandarin in comparison with English, Old Chinese, and some Modern Chinese dialects. He proposes an analyticity macroparameter that consists of a cluster of microparameters, each responsible for an analytic property. Huang’s analytic properties include the following (among others).
Table 4.10 Huang’s (2015) analytic properties
V domain N domain T/C domain
light verb construction numeral classifier wh-in-situ
pseudo noun incorporation localizer discontinuous wh-the-hell construction verbal atelicity discontinuous preposition no negative quantifier
no verbal coercion positive degree marker no reciprocal pronoun
Kaynean word order par excellence absence of canonical gapping
Huang explains these properties by two formal features, [+epp] and [+strong], with the for- mer triggering phrasal movement and the latter head movement, as in the quote below:
[T]he difference in the degree of analyticity can be characterized … within the Probe- Goal system on the basis of the nature of a probing head: a head with [+epp] or a [+strong] affixal feature leads to movement and synthesis, while [−epp] and a [−strong] head preserves analyticity by leaving elements in situ. While each case of the variation may be described in terms of a micro-parameter associated with a particular probing head, the clustering of parametric values across the board gives rise to … the analytic-synthetic macro-parameter. (Huang 2015: 34)
Huang’s proposal relates analyticity to movement in an even tighter way than my proposal. On the one hand, the [+strong] feature is more or less a syntacticized version of my PF-oriented reasoning (and that of Borer). For instance, Huang explains the lack of noun-to-light-verb in- corporation in Mandarin as follows:
[T]he light verb is phonetically non-null, thus blocking noun incorporation. In par- ticular, the light verb is phonetically independent, thus at least not an affix, and not [+strong] in the relevant sense. (Huang 2015: 11)
Huang does not specify the correlation between “affix” and “strong.” In one place he seems to equate the two:
An F feature may be “strong” (or affixal) or “weak.” A [+strong] feature F marks a strongly defective head that requires licensing by overt head movement (Move), while a [−strong] feature may be licensed simply under an appropriate matching configuration (Agree). (Huang 2015: 10)
However, if strong means affixal, then what is the ground for the syntacticization of such an ap- parently morphophonological property? After all, Borer’s work demonstrates that it is feasible to let the affixal property remain morphophonological. In addition, the ontological distinction between [+strong] and [+epp] and its consequences are unclear;41 we know from Huang’s de- scription that the former is responsible for head movement and the latter for phrasal movement, but what information must the syntax be granted to implement this division of labor? And how feasible it is for children to acquire two movement features? Overall, resorting to purely formal features to account for a highly lexically oriented phenomenon like analyticity seems to bring about more puzzles than it resolves.
On the other hand, [+epp] or phrasal movement is entirely absent from the traditional per- spective on analyticity, which treats it as a dimension of morphological typology. Huang is more 41This is not only Huang’s problem but a general problem in the field.
interested in “syntactic analyticity,” though again the term is not explicitly defined. Insofar as the properties in Table 4.10 (especially those in the T/C domain) are concerned, it seems quite differ- ent from the traditional perspective, and whether those properties are all indicators of analyticity is debatable. For instance, wh-in-situ also exists in Japanese and Korean, and the Japanese/Ko- rean counterpart of wh-the-hell is discontinuous too, as in (80). Since Japanese and Korean are quintessentially agglutinative in the T/C domain, properties like wh-in-situ and discontinuous
wh-the-hell may not be good indicators of high analyticity.42
(80) a. ittai [Japanese] the hell kare-wa he-top nani-o what-acc katta bought no? question
‘What the hell did he buy?’ (adapted from Huang & Ochi 2004)
b. daeche [Korean] the hell neo you nugu who ya? question ‘Who the hell are you?’
Faced with issues like the above, I choose not to parameterize analyticity in terms of formal features, nor do I treat it as a cluster property. Instead, I view morphological typology in the traditional sense as a matter of tendency; under GRS it can be conceived as a tendency in the application of a universally available structure-building strategy, root support. Being a tendency, analyticity may spread or shrink within a language and wield areal influence across languages at its own pace (i.e., independently from other dimensions of language change, such as the change in word order). Since root-supported heads are lexicalized, this is ultimately a lexicon-centered variation, though it differs from the classical BCC in not being encoded in formal features. In relation to this, I find the following quote from Borer (2005a) appealing:
[A]ll variation, both within a language and across languages, is reducible not only to the properties of range assigners to functional open values, but [also] to their morpho-phonological properties. (Borer 2005a: 264)
Based on this I tentatively propose an extended Borer-Chomsky conjecture: (81) Extended Borer-Chomsky conjecture
All parameters of variation are attributable to the lexicon, either to differences in the for- mal features of functional heads or to differences in their morphophonological nature. Thus, if analyticity were parameterizable, it would be more appropriately parameterized along the morphophonological path and independently from the formal feature path.
42Ian Roberts (p.c.) points out that Huang’s properties are not meant to be considered indicators of analyticity in
isolation but only codefine it when they cluster. While I am sympathetic to this perspective (which is essentially based on Biberauer & Roberts’ 2015 parameter hierarchy), to what extent the empirical phenomena in Huang (2015) lend themselves to this perspective requires more careful evaluation.