light switches

English -ify and Telugu -inc have highly similar distributional profiles, both instantiating the ‘little light verb’ v of Chomsky (1995, 2000, 2001). But while -ify is heavily exploited as an intermediary between a Telugu verbal root and English verbal inflection in English/Telugu code switching, performing a switch at the light v level, ‘light switches’ in the opposite direction, employing Telugu -inc as an intermediary between an English verbal root and Telugu verbal morphology, are out of the question. The central question that this paper seeks to answer is why there should be such an asymmetry in English/Telugu code switching. Our analysis capitalizes on an independently establishable difference between English -ify and Telugu -inc: while occupying the same structural position in the tree (v), the two morphemes differ in that -inc is an incorporator, creating a complex X category which, in code-switching cases in which the verbal root is English, is not ‘language uniform’. This violates a general ban on head-internal code switching. We will show (following up on MacSwan 1997 but deviating from it in one important respect) that this ban on head-internal switches is a phonological constraint on code switching — specifically, a constraint (which we will show falls out straightforwardly from the theory) that rules out switching inside phonological words that are morphosyntactic heads (Xs). When ‘light switches’ featuring the ‘little light verb’ v are blocked by this constraint, switching with the aid of a ‘lexical light verb’ V (such as ‘do’ or ‘make’) is available as a last resort; when ‘light switches’ employing v are successful, economy considerations block the structurally more complex alternative of switching at the ‘lexical light verb’ level. 1 The background: Telugu causatives Classical Telugu (a South-Central Dravidian language) makes its causatives with the aid of the free-standing lexical verb cees ‘do/make’ , which takes an infinitival complement (ending in -a(n), the infinitival suffix) whose subject is marked accusative by the matrix verb (ECM) and whose object is ACC–marked by the infinitive (cf. (1b)). Informal modern Telugu instead employs the suffix -inc, which we will gloss as ‘DO’ (cf. Murti 1973, Krishnamurti & Gwynn 1985:202 for discussion of Telugu causatives). This suffix attaches to the transitive verb stem and gives rise to a faire-par type causative, with the causee marked with ceeta ‘INST, by means of’ (cf. (2b)). (1) a. paapa pustakamu-nu caduwu-nu child book-ACC read-AGR ‘a child reads a book’ b. siita paapa-nu pustakamu(-nu)caduw-a ceeyu-nu Sita child-ACC book-ACC read-INFIN make-AGR ‘Sita makes a child read a book’ (2) a. kamala niiLLu kaacindi Kamala water boil-PST-AGR ‘Kamala boiled the water’ b. raamu kamala-ceeta niiLLu kaay-inc-EEDu Ramu Kamala-INST water boil-DO-PST-AGR ‘Ramu made Kamala boil the water’ 2 Shoba Bandi-Rao & Marcel den Dikken — Light Switches 1 The form pooj preceding cees in (3a) is seemingly a bare stem. In actual fact, however, what we are dealing with here is a phonological reduction of pooja cees to pooj cees; in syntax, therefore, we do in fact have an infinitive in the complement of cees. Three things support this interpretation of the facts. First, when adverbial material follows the lefthand verb, as in (i), we see the infinitival suffix -a showing up overtly. Secondly, those (few) consonant-final words that serve as input to the cees construction, such as bajan ‘religious chant’ , do indeed show up in their infinitival form (cf. (ii)). And thirdly, in Kannada the phonological reduction of the counterpart of Telugu pooja to pooj does not take place in the context at hand: cf. pooje ‘worship’ and poojemaaDide ‘to worship’ (not *poojmaaDide). (i) pooja baaga ceesiri worship well make-3PL-PST ‘they worshiped well’ (ii) bajanceesiri chant-make-3PL-PST ‘they chanted’ 2 On -is, the Kannada counterpart of Telugu -inc, as the spell-out of the Chomskyan light verb v, see Lidz (1998). We have glossed neither cees nor -inc as causative elements because neither is in fact intrinsically causative: both cees and -inc can be used, alongside their causativizing uses, as light verbs serving as hosts for inflection in Sanskrit loans, as illustrated in (3a,b).1 (3) a. pooja ‘worship’ 6 pooj cees ‘to worship’ b. preema ‘love’ 6 preem-inc ‘to love’ Both cees and -inc are multi-purpose light verbs, therefore — the former differing from the latter in taking an infinitivally inflected complement. Selection of a full-fledged infinitival complement is a property of lexical verbs — cees, therefore, is a V, with both ‘heavy’ and ‘light’ incarnations. In its ‘heavy’ guise, cees behaves like garden-variety transitive verbs and is capable of taking a nominal complement (as in Siita vankai palyam cees-indi ‘Siita eggplant dish make-3PST’ ); as a ‘light’ verb, it behaves essentially like the types of verbs discussed by Grimshaw & Mester (1988), such as English make, give and do (cf. Siita pani cees-indi ‘Siita work make-3PST’ , where pani is a Telugu noun, not a Sanskrit loan). By contrast, -inc selects a stem — either a bare stem or a transitive stem. Whenever -inc is attached to a transitive stem, the output is causative. But when the stem hosting -inc is not a transitive stem, the result of -inc affixation is not causative; it may be transitive, but it does not have to be: (4a) is an unaccusative inchoative construction with -inc, itself eligible as input to causativization with the aid of cees, as shown in (4b). (4) a. nadi prawah-incu-nu river flow-DO-AGR ‘a river flows’ b. Ganga devi nadi-ni prawah-inc-a ceeyu-nu Ganges goddess river-ACC flow-DO-INFIN make-AGR ‘the Goddess Ganges makes the river flow’ What (4) shows very clearly, then, is that -inc is not itself a causativizer. As a matter of fact, its sole function in the example in (4a) is to serve as a bridge between the Sanskrit loan prawah ‘flow’ and the subject agreement inflection -nu. Similarly, in (4b) -inc attaches to prawah to mediate between it and the infinitival inflection, -a. This makes -inc a ‘light verb’ in the sense of Chomsky (1995, 2000, 2001) — a connective v between the predicative root and the inflectional structure of the clause.2 3 Shoba Bandi-Rao & Marcel den Dikken — Light Switches 3 All the code-switching data reported in this paper are based on native speaker judgements collected by the first author. The use of -ify as a ‘pivot’ in code switching is by no means peculiar to the English/Telugu case. It is in fact used profusely on the Indian subcontinent — there is evidence of the use of -ify in code-switching constructions featuring Kannada, Malayalam and Tamil as well (for their judgments, we thank Sudha Gowda, Prem Panikar and Latha Narayan, respectively), with -ify being necessary (as in the Telugu case) to link the verb to the English inflection. Bhatia (1989) argues that in ‘Filmi English’ , a specific form of English/Hindi code switching, English -ify (reduced to -fy and obligatorily separated from the stem by the vowel -o-, which Bhatia 1989:271 claims is a functional morpheme of sorts) attaches exclusively to nominal stems (cf. e.g. you mask-o-fied him ‘you jokeN-o-ify-PST (joked with) him’ ); when the Hindi stem is verbal, English verbal morphology may be added directly to the combination of the Hindi stem and the -omorpheme (cf. I manaa-o-ed her ‘I consoled her’ ). But for the English/Telugu -ify cases discussed in this paper, it is entirely clear that the Telugu root is verbal. (In fact, switches between Telugu nouns or adjectives and English -ify+INFL are impossible: the ‘lexical light verb’ cees will always mediate between Telugu N/A and English -ify — something which will follow if (i) -ify is a lexicalization of v (as we are claiming) and (ii) v must have a VP in its complement.) 4 The -iin forms like kal(i)pify is subject to elision, the syncopated forms being particularly common in the spoken language. The fact that, when it does show up, the vowel surfaces as -iprovides an interesting piece of evidence for the claim that the Telugu base verb and the suffix -ify form a phonological unit. Telugu has a vowel harmony process by which an /u/ in the final syllable of the stem changes to /i/ under the influence of a high-vowel suffix (cf. Babu 1981). Roots like kalupu ‘stir’ undergo this process not just in the context of an indigenous suffix (cf. kalip-indi ‘she stirred’ ) but also under the influence of -ify (cf. (7a–d)). Thus, we have identified two types of ‘light verb’ in Telugu, a lexical ‘light V’ (cees ‘do/make’ ) and a non-lexical ‘light v’ (-inc), the former selecting a full-fledged infinitival complement, and the latter a projection of the lexical verb, yielding a verbal but not necessarily transitive output. This is summarized in (5). (5) a. ‘light V’ (cees) – takes a full infinitival complement b. ‘light v ’ (-inc) – takes a projection of the lexical verb as its complement 2 The problem: A code-switching asymmetry 2.1 The use of -ify as a pivot in code switching English -ify is a close match of Telugu -inc as far as syntactic distribution is concerned (cf. (6a,b)): it likewise functions as a go-between for lexical roots and inflectional morphology producing an output that is verbal but not necessarily transitive. (6) a. they are trying to diversify/gentrify/pacify/... the neighborhood b. this neighborhood has diversified/gentrified/pacified/... dramatically over the past few years It is precisely the fact that -ify serves to connect things which could not otherwise host verbal inflectional morphology, such as the adjective diverse or the noun gentry, that makes -ify an ideal ‘pivot’ in code switching. From this perspective, it does not come as a surprise that it is -ify that helps out in code switches at the juncture between English verbal inflection and a Telugu root.3 Indeed, the apparently completely vacuous use of -ify is extremely common in this context. Some illustrat