SUMMARY OF PRINCIPLES IN PAPER ON CLITICS IN THE EUROPEAN SCIENCE FOUNDATION “EUROTYP SERIES” VOLUME

The full paper is J. Emonds (1999) "How Clitics License Null Phrases: a Theory of the Lexical Interface," in H. van Riemsdijk, ed., Empirical Approaches to Language Typology: Clitics in the Languages of Europe, Mouton de Gruyter, Berlin.

Peer reviews by M. Baker, M. den Dikken, F. Dreikonigen, H. Haider, M. Haverkort and T. Teraldsen, along with my responses, follow the paper. Since it would seem controversial that I entirely deny the existence and necessity of clitic climbing, it is perhaps surprising that only a couple of the reviews question my overall conclusions. My responses claim to answer all but perhaps one of the reviewers’ empirical objections.

Abstract

The paper defends a general “clause mate” or more precisely “phrase mate” condition between Standard French, Italian and Spanish verbal clitics and their phrasal sources. It argues that all five instances in the literature of “clitic climbing” have been mis-analysed. (i) Franco-Italian genitive en/ne is argued to crucially involve right-dislocated constituents (sometimes null) as its source; and (ii) complements of adjectives are argued to be optionally generated outside AP as sisters to higher linking verbs. I claim then that only sisters to such V projections are the source of the clitics. There is no “climbing” from inside DPs, NPs or APs.

Then, initial grammatical verbs V_x with non-finite complements V_c , i.e., (iii) auxiliaires, (iv) restructuring verbs and (v) causative verbs, are argued to instantiate mono-clausal “flat structures.” Hence, clitics don’t “climb” to V_x; the complements and adjunct phrases of V_c are rather deep sisters of some projection of V_x as well. In particular:

(a) There are obligatory flat structures for the “empty” or copular auxiliaries avoir/ être (French), avere/ essere (Italian), and haber/ ser (Spanish) constructed with past/ passive participles.

(b) There are optional flat structures for closed classes of verbs with infinitival and gerundive complements in both Italian and Spanish, so-called “restructuring verbs.”

(c) There are no flat structures for the vast majority of verbs in all three languages, nor for French counterparts to Italian and Spanish restructuring verbs. Open class items have phrasal complements.

The paper develops a theory of lexical insertion so that (a)-(c) follow from the levels at which lexical items enter a derivation. These levels in turn follow from the nature of the syntactic features in lexical entries.

Main Principles Defended and Results Claimed in the Paper

I have used bold for statements whose formulations still seem adequate (Oct, 2002) and which in addition rest on ample empirical justification. Numbers refer to the numbering in the paper and so serve to locate that statement in that text. My current comments are enclosed in […].

(37) Genitive Phrase Mate Hypothesis. The Franco-Italian clitic en/ne on V_i is related only to (possibly dislocated) PP, DP or NP that are sisters to some projection of that V.

[ Defense of (37) with more carefully elaborated discussion is the focus of a later publication in French (the logic is the same): “La relation entre la Dislocation à Droite et le clitique franco-italien en/ne.” Journal of the Linguistic Society of Japan 119, 1-32, 2001. ]

(50) En/ne are well-formed if co-indexed with a dislocated XP_i sister to V^k, provided they are not also co-indexed with a definite DP_i within the clause.

(56) Sisterhood. If W and Z are sisters, W dominates X, and X dominates the only lexical material under W, then X and Z are sisters. [ I later call (56) “extended sisterhood.”]

(57) Phrase Mate Hypothesis. Romance clitics on V_i are related only to XP sisters to some projection of that V (V_i^k), where XP = DP, IP, PP, NP, AP, VP.

(59) Canonical Realisation. UG canonically matches a few syntactic features F to each syntactic category B. These features F contribute to LF only in these "canonical positions" on B, and appear elsewhere only via language-particular lexical stipulation.

(60) Alternative Realisation ("AR"). A syntactic feature F matched in UG with category B can be realised in a grammatical morpheme under X⁰, provided X^k is a sister of [B, F].

[ My later Lexicon and Grammar: the English Syntacticon extends AR to any projection of [B, F]. For example, the suffix –ever in the D⁰ whatever can alternatively realise a feature F of the C if, since DP is a sister of the C¹ projection of [C, F]. Similarly, applicative suffixes on V alternatively realise P, etc.]

(61) Invisible Category Principle. If all marked canonical features F on B are alternatively realised by (60), except perhaps B itself, then B may be empty.

(70) Deep Lexicalisation (DL). Items associated with non-syntactic, purely semantic features f satisfy lexical insertion conditions (just) before transformations apply to domains containing them. Such f occur only on N, V, A and P.

(71) Phonological Lexicalisation (PL). Items specified solely in terms of contextual and other non-interpretable features are inserted subsequent to any operation contributing to LF.

(74) Extended Classical Subcategorisation (tentative). @ , X , +___Y is satisfied if and only if Y⁰ is the lexical head of a (maximal projection) sister to X⁰.

(75) Lexical Head/Projection. Let Y⁰ be the highest lexically filled head in Z^j, where Y⁰ = N,V,A,P. Then Y⁰ is the lexical head of Z^j, and Z^j is a lexical projection of Y⁰.

(78) Extended Classical Subcategorisation. @ , X , +___Y is satisfied if and only if Y⁰ is the lexical head of a sister to X⁰. [ Generalised to (107) below ]

(79) Economy of Representation. At a given level of lexical insertion, satisfy subcategorisation features with as little phrasal embedding as possible.

(82) The subject DP of an infinitive is structurally distinct from the DP' subject of its governing verb, across the languages under consideration.

(93) LF Subjecthood. A P^DP sister of a lexical projection of V_i may be interpreted as LF subject of V_i, (i) if VP_i is the sister of I, and (ii) P is empty at s-structure.

Note 59: With the recursive definition (56) of extended sisterhood, (93) simplifies drastically to (ii):

(ii) LF Subjecthood. A DP sister of a lexical projection of V_i may be interpreted as the LF subject of V_i if VP_i is the sister of I.

[ The topic of determining subjecthood as late as LF is handled with more simplicity in my “The Lower Operator Position with Parasitic Gaps,” Features and Interfaces in Romance. Essays in honor of Heles Contreras, J. Herschensohn, E. Mallén, and K. Zagona (eds), John Benjamins, Amsterdam, 2001, 85-106. ]

(95) Empirical range of clitic climbing in French causatives: "...neither the direct object nor the circumstantial complements are subject to the Specified Subject Condition [JE: they can cliticise on the causative verb V_x ]. The only complements that remain subject to the SSC [JE: they cannot cliticise on V_x ] are...the subcategorised nondirect objects..."

J. Aoun (1985), A Grammar of Anaphora, MIT Press, Cambridge, 128

(96) Logical Subject Condition. In ...X...Z...Y..., where X c-commands Z and Z c-commands Y, clitic co-indexing may not involve X and Y if Z is a Logical Form subject.

[ Note: (96) crucially makes no superfluous mention of any clausal domain containing Z and Y but not X. I maintain that this has been the essence of problems with Kayne’s classical SSC account of clitic climbing in French causatives and its successors. Cf. R. Kayne (1975), French Syntax, MIT Press, Cambridge. ]

(107) Generalised Subcategorisation. @ , X , +___Y is satisfied if and only if Y⁰ is the lexical head of a complement within a lexical projection of X.

(117) Head-driven Phrase Mate Hypothesis. Romance clitics are related and co-indexed only with XP sisters of a projection of a lexical head V_i. (This is possibly too general)

(118) Dislocation Convention. A phrase @ adjoined to a clause co-indexed with an identifying @' inside the clause requires that @' be phonological if possible.

Revised Dislocation Convention. A phrase @ adjoined to a clause requires, if possible, a unique phonological co-indexed constituent @' inside the clause. [from Note 22 in the article. The article on French/Italian en/ne cited above perhaps improves on (118) and on this revision. ]

(119) Economy of Derivation. Among alternative derivations from the same deep structure, prefer the derivation with the fewest insertions of free morphemes.

Some of the arguments presented at the end of the paper for a delay in assigning subject status to any NP within a Romance causative complex predicate in which clitics are on the first (grammatical) verb:

(99) Whether the interpreted subject of a lexical V_c complement of a causative may occur in an à-phrase is determined by the varying subcategorisation frames for complements of the causative/ perception verbs V_x, as shown in J.Herschensohn (1981), "French Causatives: Restructuring, Opacity, Filters and Construal," Linguistic Analysis 7, 217-280.

(100) Attempts to block movement over the subject of a lexical verb V_c by assigning it external argument status postulate construction-specific devices to explain why both the direct object and adjuncts of V_c are unaffected by its presence. In particular, although adjuncts usually resist extraction, their cliticisation is unrestricted; cf. (95) above.

(101) Kayne (1975, Ch. 4) establishes that many dative DPs in the faire...à construction are subcategorised base arguments of faire. Nothing prevents such arguments from serving as subjects of V_c, provided some general interpretive principle, e.g. (93), assigns them subjecthood late in a derivation.

(102) Kayne (1975) establishes that the subject of a complement V_c never results from "raising to subject"; that is, NP movement never uses this putative external argument for a landing site. But in all other cases, it is rather internal arguments that cannot serve as landing sites for NP movement. Thus, the subject of V_C is not a structurally like a subject during a derivation.

(103) In complex examples given by Kayne (1975, Ch. 6 and its footnotes), word orders predicted by assigning underlying external argument status to the interpreted subject of V_c are completely unacceptable. It rather acts like a base-generated complement.

(104) Depending on whether or not it receives accusative case, the DP subject of V_c precedes or follows (respectively) other subcategorised internal arguments of V_c, This follows naturally if this DP has internal argument status itself during the derivation.