FEATURE ASSIMILATION

There are independent syntactic reasons for postulating the existence of the Have and Be hierarchies: consider the following sentences.

(9)    (a)   (i) the boy’s knee aches

                (ii)*the statue’s knee aches

                (iii) the statue’s knee broke

         (b)   (i)  the man’s neck itches

                (ii)*the man’s book itches

                (iii)the man’s book convinced me

The sentences in (9 a) are both well-formed with respect to the Have hierarchy since both ‘boy’ and ‘statue’ stand in the same relation to ‘knee’. But one observes that (9 ai) has a conceptual status which differs significantly from sentence (9 aii). The difference is a reflection of the fact that a ‘statue’ is normally both inanimate and not living and, hence, cannot be subject to the set of events in which nouns that are both animate and living can participate. The descriptive problem concerns the fact that the anomalous interpretation of the phrase ‘knee aches’ is dependent upon information which is not included in the lexical entry for the noun ‘knee’. The information that a ‘knee’ belongs to a noun marked either ‘ + animate, + living’ or ‘ — animate, —living’ is necessarily dependent upon some mechanism which assigns these features to ‘knee’ on the basis of the underlying phrase structure configurations. This mechanism provides a formal object which can be systematically interpreted by the semantic component.

The mechanism which assimilates the relevant features in sentences (9 a) explicitly refers only to the lexical Have hierarchy rather than to any sentence with ‘have’. This becomes clear on consideration of the fact that such assimilation does not apply in the event that the head noun does not stand in an inalienable Have relation to its modifier, as in (9 bii). Underlying (9 bii) is the string ‘the man has a book’ in which the interpretation of ‘have’ is one of ‘possession’. As was assumed earlier, there are independent reasons for assuming that the non-lexical ‘have’ of ‘possession’ is distinct in certain ways from the lexical ‘have’ of ‘inalienability’. This hypothesis is confirmed with respect to feature assimilation. The hypothesis that feature assimilation involves the lexical Have hierarchy is supported further by the fact that sentences in which the Have hierarchy is used to its fullest extension have properties identical to those in (9 a), as demonstrated in (10):

(10) (a) the toe of the foot of the leg of the boy aches

(b) *the toe of the foot of the leg of the statue aches

There is a similar additional motivation for the Be hierarchy as a necessary component of the grammar: consider (11).

(11) (a) a gun which is a cannon fires cannonballs

        (b) *a gun which is a six-shooter fires cannonballs

The descriptive problem posed by the sentences in (11) exactly parallels that posed by the sentences in (9a). The peculiarity of (11b) is reflected by the fact only a gun which has the properties of a ‘cannon’ can ‘fire cannonballs’. The difference between the two sentences is formally characterized by an assimilation rule which assimilates the features of the modifying noun to the ‘head’ noun: in (11) the features of ‘cannon’ and ‘six-shooter’ are assimilated to ‘gun’. Such a feature assimilation provides the information requisite to the semantic interpretation of the sentences in (11) which is not systematically given by the underlying structure.¹ In particular, it is necessary to know what kind of ‘gun’ is involved in order to determine the semantic well-formedness of the predicate.

¹ It is conceivable that the assimilation rules are to be incorporated into the interpretative semantic cycle itself rather than to provide a derived semantic structure upon which the interpretative component operates. Factors bearing on this decision will not be explored in this paper. At the time when we first proposed the problem raised by sentences like those in (9) several of our friends and teachers argued that a syntactic solution would be forthcoming. This has been incorporated as part of the analysis of adverbs proposed by G. Lakoff. Briefly (9 a) would be analyzed as ‘the boy aches in his leg’, and ‘the statue aches in his leg’. The anomaly of the second would be explained as a function of the fact that ‘the statue aches’ is anomalous. (9bii) would be anomalous because there is no sentence ‘the man itched in his book’, while there is a sentence corresponding to (9 ai)‘ the man itched in his neck’. Whether or not one is convinced by this analysis (which we are not), the fact remains that those nouns which can act as agents with verbs, like ‘ache’, ‘itch’, ‘hurt’, etc. are just those which are represented in the lexical Have hierarchy as inalienably part of the head noun. For example, there is a sentence ‘ I hurt ’ and ‘ I hurt in my arm ’, but not ‘ I hurt in my carburator ’. Similarly words like ‘ book ’, ‘ idea ’, or ‘ thought ’ are not entered lexically. Thus, either form of syntactic analysis of these examples presupposes a lexical Have hierarchy. (Note that this does not preclude a ‘ creative ’ grammatical component which would allow for new ad hoc Have and Be relations to be produced as needed. As we continually emphasize, the ontogenesis of lexical structures is not treated in this paper.)