In reconstructing protoforms and linguistic histories, many methods can and have been employed with varying degrees of effectiveness (cf. Bonfante, 1945, for a general discussion). For the specific purpose of establishing the time sequence of phonological changes according to which sound systems have evolved, traditionally two procedures have been most frequently followed: one is to sift the historical records for clues to the dates of linguistic innovations, and the other is to infer from the systematic correspondences between protoforms and their modern reflexes the internally motivated order of diachronic rules.

There has been a great deal of discussion recently concerning the rôle that morphological and paradigmatic considerations play in the phonological structure of languages (e.g. Kiparsky, 1972, 1974; Anderson, 1974, 1975; Hooper, 1974; Harris, 1973; Kisseberth & Abasheikh, 1974). However, although it has generally been accepted that such factors play a part in determining the phonological properties of morphemes and words, the precise characterization of the interplay between phonology and morphology is not at present very clear. In this paper, I will examine two of the most interesting problems of Modern Greek (MG) that fall on the boundary line between morphology and phonology and thus provide a test case in the present debate. Moreover, since these problems have been previously examined both outside the transformational generative model (Householder, 1964) and within it (Warburton, 1970a, 1970b; Newton, 1972a, 1972b, 1973; Adams, 1972), the facts are fairly clear and, consequently, they offer excellent ground for arguing theoretical points.

I. In this paper I will examine the systematic alternation in grammatical relations (or functions) of noun phrases in the semantic rôles that I will refer to as ‘content’ and ‘container’ for that class of English verbs which appear in the syntactic frames given in (I):

(I)(a)…V NP1 IN/ON(TO) NP2

…load the hay on the waggon

(b)…V NP2 WITH NP1

…load the waggon with the hay.

This class of verbs, exemplified by load, spread and spray is particularly interesting because of the claim made in Anderson (1971) that the noun phrase which stands in the grammatical relation of direct object is semantically interpreted as being wholly involved in the action or wholly in the state indicated by the verb. Thus, for verbs that may appear in both of the frames in (I), if Anderson's analysis is correct, the semantic interpretation of structures like (a) and (b) will be distinct in that in (a), NP1 should be interpreted holistically and NP2 should not, while in (b), NP2 should be interpreted holistically, and NP1 should not. After first examining some general characteristics of this class of verbs, I will point out some problems which are presented by Anderson's analysis.

I. Chains are familiar in diachronic phonology as sets of shifts of the form A>B, B>C, C>D, etc. Well-known examples include the following (cf. Martinet, 1955: Ch. 2):

(I)(a) Grimm's Law: bh>b>p>f

(b) Azores Portuguese: a>doubt:>о>u>y

(c) Great English Vowel Shift: æ>doubt:ēī>ai

(d) Hauteville dialect: ề>doubt:ε>a>doubt:

(e) High German Consonant Shift: d>>ts/s

(f) Latin>Italian: kui>kwi>ki>tfi

Martinet (1952, 1955) gives a ‘functional’ interpretation of this phenomenon, distinguishing between ‘push’ and ‘drag’ chains; King (1969), arguing from a generative point of view, rejects the ‘push’ variety, and considers ‘drag’ chains to be a case of rule simplification.