The nature of functional categories is not clearly defined in current linguistic theory. There are many definitions available, and each one seems to pick out a distinct subset of those morphemes which have specifically syntactic representations. There is more detailed discussion below concerning the difficulties raised by such unclarity, in the context of an investigation of prepositions - a category which typically seems to be neither functional nor lexical, at least on existing definitions (chapter 4). Up to now I have been using such terms as lexical / functional classes, lexical / functional categories, function words, substantives, closed and open class items and so on in a rather vague and intuitive way. In my exposition of the aphasie deficit here under investigation, however, I make use of some such terms in rather specific senses.
‘Function words’ I take to be those functional categories with a free morphological, phonological and / or orthographic representation; function words can be heard, and are not bound to a substantive. This term therefore includes overt pronouns, though it excludes pro and PRO (assuming that these have any theoretical value under minimalist approaches - e.g. Manzini and Roussou 2000 argue that PRO and control can be reduced to properties of the operation Merge); it includes auxiliary verbs, though it excludes the realisation of tense and agreement on substantive verbs. ‘Function w ords’ is a useful term because it delimits the set of functional categories which can be used in single word reading tasks, which have formed an important part of the investigation
presented here (and also of other investigations of aphasie deficits). ‘Functional categories’, on the other hand, is a much broader term, and includes all those categories (pronounced or not, free or bound) which are directly relevant to the syntax. So this term includes AGR, TNS and C as well as auxiliary verbs, pronouns, conjunctions, negation, quantifiers and so on. It will be seen later that this term also includes some categories not always thought of as functional, including the argument-structure-changing morphemes PASS (associated with passive verbs) and UNACC (associated with unaccusatives - chapter 3), as well as prepositions (see chapter 4), and functional items often assumed to be modifiers rather than heads of functional projections - adverbs being a case in point (chapter 5). ‘Functional heads’ are again a subset of functional categories, and this term includes all those items (overt or covert) which head a projection in the functional domain. Functional categories generally have certain features in common with each other: they are the locus of parametric variation (Ouhalla 1991; W exler & Manzini 1987), and they often have zero alternations. ‘Substantives’ is the term I shall use to cover the major lexical classes: nouns, (most) verbs and adjectives.
1.6.2 Language faculty organisation
>
I also follow certain assumptions about the organisation of the conceptual system and the language faculty. I accept Fodorian modularity (Fodor 1983), inasmuch as modular systems must of necessity be informationally encapsulated, domain specific, fast and mandatory in their operations, and follow a distinct ontogenetic path. I also assume that modular systems have specific neural substrates, which seems to me to be a necessary assumption in the discussion of differential breakdown o f isolable systems within a particular cognitive faculty (such as language), when focal brain injury is the causal factor. However, I also accept that some systems with apparently modular organisation may be diffusely represented in the brain.
The lexicon is a case in point. Given Sperber and W ilson’s (1995) account of lexical entries as tripartite conceptual addresses, it seems very likely that a lexical entry for a single word may be represented in different ways and in more, than one brain system. To simplify, Sperber and Wilson view concepts as abstract psychological
objects which have ‘addresses’ in the central system. Under each conceptual address, three kinds of information are stored. The logical entry of a conceptual address is its meaning postulate (in the Fodorian sense) - a deductive rule which applies to logical forms in which that concept participates. For example, the logical entry under the concept CAT would be along the lines of a CAT-elimination rule, which permits the concept to form part of a basic inferential process along the lines of ‘if X is a CAT then X is an ANIM AL’. The encyclopaedic entry is a store of general and idiosyncratic information about the concept’s denotation - for example, under the concept CAT is represented information about different breeds, your own cat, and so on. Encyclopaedic entries by definition vary between individuals and can be changed or altered by experiences or reasoning processes; they therefore also form part of the central system. The lexical entry is a listing of the linguistic properties of the lexical item which encodes the concept. I view this as a kind of interface between the central system representations associated with a particular concept, and the mental lexicon which is internal to the language faculty. In a sense, then, the language faculty is not entirely modular, because the lexical entries of substantives interface with central system representations. This view is similar to that put forward by Smith and Tsimpli (1995).
Substantives, then, are those items with a direct mapping between their lexical representation and their conceptual representation, the latter being located in a ‘conceptual lexicon’ (to borrow a term from Smith and Tsimpli 1995) outside the language faculty. Functional categories do not have such a 1:1 relationship with their conceptual representations. Although I do not accept the view that functional categories are somehow ‘deficient’ with respect to meaning (e.g. Friedman 1995), it is clear that the mapping between functional categories and concepts is not as straightforward as I have assumed for substantives; there is no 1:1 mapping between functional categories and concepts. Tsimpli gives the realisation of Tense as an example of this mismatch: Tense certainly has an interpretive effect, so must map onto some conceptual representation(s). But it is well-known that Tense can be syntactically realised in many different ways - as bound morphology on V, as aspectual features, as adverbials, as auxiliary verbs - and therefore the correspondence between the interpretation of this category and its syntactic reflexes is a many-to-many relation, unlike that of a substantive with its meaning postulate.
Functional categories are part of the language faculty per se, on the assumption that they are the locus of parametric variation and are therefore relevant only for language-specific processes. Functional categories, then, are assumed to be represented in an innately specified part of the language faculty, the UG lexicon (or functional lexicon). I return to a more detailed consideration of the lexical / functional dichotomy in chapter 3; but the basic idea is that, as Uriagereka (1998:454) puts it:
‘[There is] a distinction between standard lexical items and grammatical functional items. Language has a fixed, very limited set of the latter. Suppose that the split between lexical and functional items means that they’re not even part of the same mental stock....’.
It seems reasonable to interpret the notion that lexical and functional items are of different mental stock as suggesting that the two different kinds of category are respresented in isolable lexica; but such an assumption brings with it the requirement for an articulation of the possible relationships between the substantive and functional lexica, and their respective conceptual (interpretive) representations. Smith and Tsimpli (1995) propose that the lexical representations associated with entries in the ‘conceptual’ lexicon (items which can participate in reasoning and inference processes in the Language of Thought (Fodor 1975) directly) are available at the level of an independent morphological component, which therefore provides an interface between the central systems and the language faculty. Tsimpli (1996) takes a somewhat different view of the relationships between different lexica in the language faculty, assuming the existence of a UG lexicon but also the presence of a language-internal substantive lexicon. Both lexica have the ability to map onto conceptual representations, which remain in the central system.
I am inclined to the latter view, for two main reasons, the first being conceptual. If lexical entries of substantives are assigned to concepts at the level of morphology, then a great deal of syntactic information is also assumed to be represented in the morphological component - including theta grids and information about the extended projection (in the sense of Grimshaw 1991). This, then, seems tantamount to simply moving the lexicon, not to proposing the existence of a separate morphological component proper. Secondly, there is empirical evidence which suggests that the presence of an extended projection can actually compensate for the deficient operation of a damaged morphological component, at least when a linguistic representation enters
the syntactic derivation (see chapter 2). This suggests that even when the morphological component is not working well, detailed information about the lexical representation of substantives may still be available. This is an argument against the notion that linguistic lexical representations of conceptual entries are assigned in the morphological component, so it seems advisable to maintain the existence of a distinct morphological component.
This, then, is the picture of the language faculty’s organisation within which I will attempt to locate and account for the patterns of sparing and impairment observed in the case of MC. I posit the existence of a UG lexicon, within the language faculty, which is largely innately specified (hence the categories contained therein are realised universally in surprisingly uniform ways) and which contains lexical information about functional categories. These categories are the locus of parametric variation and follow a separate ontogenetic path from the syntactic component, which is strong evidence for their innate specification and also their representation in an isolable component of the language faculty. Functional categories are also distinguished from substantives by the different relations between categories and conceptual representations, which is largely 1:1 for substantives, but is more complex for functional categories. On this view of two distinct lexica, I posit that bound morphemes also have lexical representations, with inflectional morphemes being represented in the UG lexicon and derivational morphemes in the substantive lexicon. Morphological realisations of lexical categories from both lexica are assigned at an independent level of morphological representation. Because morphology is in this sense an interface between substantives and functional categories, I will again follow Smith and Tsimpli (1995) and refer ofteri to this level of representation as the morphological interface.
The morphological interface, then, is the level at which various combinatory operations hold between substantives and derivational morphology, or substantives and inflectional morphology, or functional categories and inflectional morphology^. It is also the level at which morphological realisations are assigned to these combinations, and to individual items from both lexica. The assignment of a morphological realisation to a lexical entry is essentially a process of recognition of categorial features and translation o f those into a format which can be utilised by the syntactic component.
^ The combination of functional categories and derivational morphology is ruled out on the grounds that derivational affixation tends to be a category-changing operation.
It will be noted that I do not subscribe in particular to any one of the morphological theories outlined above (in section 1.1). Nevertheless, I do follow some assumptions shared by most of the frameworks mentioned there. I assume that information relevant to other components of the grammar is provided as part o f the lexical representation of a category, so that the morphological, phonological, syntactic and semantic properies of a category are lexically specified (cf., for example, Sadock 1988, 1991). I follow Lieber (1983, 1992) in assuming that affixes as well as stems are morphemes, and as such can participate in morphological operations. On such a view, morphological operations are responsible for word formation, and therefore include affixation, derivation, compounding and other operations which may have direct effects on the argument structure of the stem. Unlike the morphological theories outlined all too briefly above, however, I will propose (following Smith and Tsimpli 1995) that the morphological component is isolable - a distinct module of the language faculty. My view also differs from some of those above in that I assume both inflection and derivation to be essentially identical processes which hold at the morphological interface, with the differences between them arising as a direct consequence, not of particular properties of the syntax (e.g. Anderson 1988), but of the distinct lexical properties of the inflectional and derivational morphemes themselves - to the extent that inflections are part of the UG lexicon, whereas derivational affixes are represented in the substantive lexicon.
One final, but important assumption which I will follow: as standard in cognitive neuropsychology, damage to a particular component of the mind/brain is presumed to yield patterns of performance which allow us to infer the nature of the damaged component in the normal brain - compensatory mechanisms do of course exist, and can confound the issue, but essentially a damaged system does not result in radical re organisation (see, e.g., Caplan 1987; Shallice 1988).
1.7
Summary
In this chapter I have outlined various theoretical approaches to the role and functional location of morphology in the language faculty, and considered some of the implications of these varying ideas for empirical evidence from language pathology and
language acquisition (normal and disordered). Previous attempts to capture syntactically-based deficits in aphasia within theoretical frameworks (Government and Binding and Minimalism) show that there is great variation between patients who fall under the broad diagnostic category of ‘agrammatism’, rendering the search for a unitary explanation futile. A broader view of agrammatism as an umbrella term for a variety of acquired language disorders which involve some aspect of the (morpho)syntactic component is adopted, avoiding the issues inherent in the syndrome classification controversy, and allowing for the derivation and testing of wider empirical predictions from constructs of syntactic theory.
A view of the language faculty as a partly modular system with an independent morphological component and separate lexica for functional categories and substantives has been put forward. In the next chapter, I present evidence in support of this view of the language faculty from a case of acquired language disorder.