5.1. On the Nature of the Morphology-Syntax Interface
LI can be useful when treated as a hypothesis, as various titles suggest (e.g. Extended Lexicalist Hypothesis (Jackendoff 1972:13), Generalized Lexicalist Hypothesis (LaPointe 1985[1980]:8) etc.), as opposed to an inviolable constraint. After a careful evaluation of the predictions a maximally strong version of LI makes (i.e. violations) across a wide range of syntagmatic and interpretive phenomena, theoretical assumptions, and typological considerations, it can be concluded that LI, as either a theoretical parameter or a conceptualization of the morphology-syntax interface, is incorrect.
The tenability of LI is at least superficially challenged by the difficulty of defining its corresponding predictions (i.e. violations) in any universal manner; as noted in Section 1.2 (and Section 2, f.n. 17), certain potential violations can be recharacterized depending on whether a morphological or syntactic analysis is chosen. However, assuming distinct classes of LI violations (Section 2), the correlations between each violation (sub-)type and specific theoretical parameters observed in Section 3.2 present a greater challenge to LI and traditional assumptions of grammar. In particular, the relationship between STRONG and WEAK LEXICALIST assumptions and LINEAR MODEL grammatical architectures suggest theories inheriting such parameters (e.g. Lapointe 1985[1980], Aronoff 1994, etc.) provide the least ideal characterization of the morphology-syntax interface. This is also true for theories with STRONGLY LEXICALIST – MODULAR (in correspondence) assumptions (e.g. LFG), since a STRONGLY LEXICALIST viewpoint typically forces one to consider challenges to LI as strictly morphological or syntactic phenomena, assigned to distinct (parallel) syntactic and morphophonological components.
Furthermore, the correlations between each LI violation type and particular typological traits in Section 4.2 help identify which types of empirical phenomena a proper theory of the morphology-syntax interface should accommodate and reveals the narrow typological foundations of LI and its associated theoretical parameters. Certain MANIPULATION (movement) and phrase-level ORDER violations, which correlate with analytic tendencies (e.g. typically low degrees of semantic density and syntactic methods of grammatical expression), demonstrate that any descriptive and explanatory theory of the morphology-syntax interface should yield to such instances of apparent bidirectional interaction. More importantly, however, the interrelationship between each MANIPULATION, ACCESS, and ORDER violation type and subtype, specific synthetic typological properties (e.g. higher semantic density and concatenative morphology), and diachronic effects on morphosyntax shows a fully descriptive and explanatory grammatical theory cannot be based on select analytic typological characteristics alone, but must be cross-linguistically applicable.
Synthesizing across the typological survey observations (Section 4.2), the specific theoretical parameter(s) underlying each violation type (Section 3.2), and the development and current state of LI (Section 1.1 and 1.2) reveals the basic problem with, and circular nature of, LI. English served as the basis for Chomsky’s (1957, 1965) original conceptualization of grammar as a syntactically driven system; specifically, particular variations in English word order were taken as the method by which to understand linguistic “competence”, or “the underlying system of rules [i.e. grammar] that has been mastered by the speaker-hearer and that [s]he puts to use in actual performance” (1965:4). As a result, largely analytic typological characteristics were considered in both the formation and continued development of the theory and its various parameters (a derivational FRAMEWORK and LINEAR model, by virtue of syntactic transformations being the theoretical foundation). When particular synthetic properties of English were subsequently taken into account, a pre-syntactic word formation component was proposed (Chomsky 1970), resulting in the gradual formulation of LEXICALISM and LINEAR MODELS. As additional synthetic and analytic typological traits were considered, the lexicalist spectrum emerged, along with its various modular implementations. STRONG LEXICALISM and TYPE (2) LINEAR MODELS were developed in order to maintain syntax and morphology as distinct components obeying separate grammatical rules – the syntactic component based on analytic typological properties, particularly word order, and the morphological component (and lexicon) based on synthetic tendencies in English and other select languages (e.g. Di Sciullo and Williams (1987), Lapointe (1985[1980]). Unsurprisingly then STRONGLY LEXICALIST – TYPE (2) theories are presented with the most serious threat when evaluated in terms of potential LI violations, since intermodular morphosyntactic interaction, and synthetic typological characteristics, were not viewed as integral to the theoretical goals in the first place.
WEAKLY LEXICALIST – TYPE (3) theories are the result of maintaining the syntactocentric theoretical foundations, but selectively integrating some morphological phenomena into the syntactic component (e.g. instances of FUSIONAL morphology and high semantic density in French and Italian verbal inflection (Rizzi 1997)), and otherwise working to treat distinct derivational processes morphologically, while assuming inflection to be part of syntax (e.g. Aronoff 1976). Other theoretical approaches were developed as reactions to, but still adopting specific assumptions made in early Transformational Grammar (e.g. Chomsky 1957, 1965, 1970). For instance, as Van Valin (2009:2) notes, “a leading idea in the development of [GPSG] was to determine whether natural language syntax could be adequately described in terms of a context-free phrase structure grammar” (cf. Chomsky 1957, 1964), while LFG involved “applying the formalism of unification grammar to natural language phenomena and showing that lexical rules were superior to transformational rules” (pp. 2). In the case of the former (as well as its descendant HPSG), the syntactocentric typological viewpoint was maintained by virtue of its focus on phrasal combination, and in the case of both GPSG (later, HPSG) and LFG, strong LEXICALISM persisted, with morphology treated via separate word formation rules (LFG) and generally considered realizational in nature (HPSG) (Nordlinger and Sadler 2016). Therefore, LI violations are variably correlated with MODULAR (in correspondence) theories, based on how a particular theorist treats the inherited (problematic) theoretical parameters in relation to instances of morphosyntactic interaction. Although still assuming a form of UNIFICATION-BASED – MODULAR (in correspondence) analysis, Role and Reference Grammar (RRG) was developed in recognition of diverse typological structures (e.g. Lakhota, Tagalog, Dyirbal, and Barai), and seeks to understand how the interaction between syntax, semantics, and pragmatics across such diverse typological profiles can be modeled and explained (Van Valin 2009:2; see also Van Valin and LaPolla 1997). Given the theoretical goals and general focus on the interrelationship between syntax, semantics, and pragmatics particularly, future work could explore the precise nature of LI and the morphology-syntax interface within specific implementations of RRG.
As noted in Section 3.2, those theories which remove the boundary between morphology and syntax fare the best when considered in terms of each LI violation type, with the question then being whether morphology is just an extension of a single syntactic component, or whether morphology and syntax truly form a single (unified) morphosyntactic system of interrelated constructions (or schemas). The extension of syntactic rules (such as head movement) to account for highly synthetic typological tendencies in FULLY SYNTACTIC – INCREMENTAL approaches stem from, for example, noun incorporation in Mohawk (Baker 1985a) and choice data examples from English, Tagalog, French, Dutch, and several other languages (Lieber 1992)51. Once we recognize that such theories assume the basic narrow conceptualization of grammar (Chomsky 1957, 1965) as strictly formulated on English’s tendency toward syntactic (analytic) means of grammatical expression, it does not seem appropriate to describe and explain all aspects of grammar – morphological, syntactic, or otherwise – using generative approaches derived from one typological trait. A FULLY SYNTACTIC – REALIZATIONAL theory such as DM52, which treats derived and inflected word forms as the result of syntactically manipulated feature bundles, that are then phonologically spelled-out via lexical insertion late in the syntactic derivation (Marantz 1995:379), suffers from the same fundamental issue as FULLY SYNTACTIC – INCREMENTAL theories; it seeks to model all aspects of linguistic expression in terms of theoretical assumptions based on very narrow typological considerations (i.e. predominantly analytic means of linguistic expression).
LI is no more or less tenable than the prevailing conceptualizations of grammar in which it was developed and subsequently maintained. The relationship between morphology and syntax is likely much more fluid than previously conceived, since the notion of LI and a modular morphology-syntax interface is (i) evidently violable, (ii) completely theoretically sensitive, and (iii) developed based on strictly ISOLATING typological characteristics. Accordingly, a proper grammatical theory should not simply seek descriptive and explanatory adequacy in a single linguistic domain, from which to extrapolate in an attempt to understand the system in its entirety; rather it should seek to characterize a given linguistic system in terms of all its relevant properties, whether phonological, morphosyntactic, semantic, pragmatic, or cognitive in nature. Current UNIFIED linguistic theories based on CxG (e.g. Goldberg 1995; Michaelis 2012) and PA (e.g. Jackendoff 1997, 2015), which assume no division between syntax and morphology, or between the grammar and the lexicon, and model both phrasal and productive morpheme combinations as signs licensed by constructions (combinatoric, derivational and inflectional), best address the considerations and observations outlined in this paper. Ultimately, if the goal is a proper characterization of the morphology-syntax interface, then one must abandon the notion of LI as well as the theoretical foundations from which it arose, and look instead to theories, like CxG and PA, which, rather than seeing syntax, semantics and lexicon as independent modules through which a sentence passes in the course of a derivation, seek a uniform format for the analysis of linguistic signs.
Section 5.1 Footnotes
51: The syntacticization of various morphological phenomena, whether through REALIZATIONAL or INCREMENTAL methods, is intimately related to the assumptions underlying “the Great Agglutinative Fraud” (Hockett 1987:82-84).
52: DM is based on insights drawn from non-concatenative, templatic morphological patterns (e.g. largely non-categorical roots) predominant among Semitic languages (Jackendoff, personal communication, September 16, 2017).