Desktop version

Home arrow Language & Literature

Special morphs

Other fundamental challenges arose in cases where no analysis could be made to conform to the strictures of the IA model. From the earliest formulations of this model, it was clear that the notion of ‘morph’ would have to be more abstract than just a sequence of segmental phonemes. Harris (1942) identifies three initial classes of non-segmental morphs, (or, again, “morpheme alternants”):

It is useful to generalize this definition of morpheme alternant by taking sequence to mean not only additive sequence (the addition of phonemes), but also zero (the addition of no phoneme), negative sequence (the dropping of a phoneme), and phonemic component sequence (the addition of a physiological feature of phonemes). (Harris 1942:170)

The introduction of ‘zeros’ is the most conservative of these generalizations, as it just extends a uniform morphotactic analysis to forms in which a morpheme has no segmental realization. The notion of a ‘phonetically null’ morph appears to originate with the Post-Bloomfiedians, as it contrasts with the interpretation of zeros as literal absence in Bloomfield:

His [Bloomfield’s JPB] use of zero is apparently the classic use found in Sanskrit grammar, namely the removal of something, and its replacement by nothing, rather than the distribu- tionalist version, in which a ‘zero element’ is present. (Fought 1999:13)

The most restrictive treatment of zero elements was suggested by Bloch (1947:402), who proposed that “no morpheme has zero as its only alternant”. This constraint disallowszeroMORPHEMES, suchasthe zero singular sometimesposited in analyses of English nouns. However, it allows zero morphs as one of the realizations of a morpheme. Hence it is compatible with an analysis of plural sheep in terms of a stem sheep and a zero plural marker ‘0’.

Reinterpreting zeros as ‘phonetically null segments’ also creates the challenge of arranging these ‘segments’. As Anderson (1992) notes, the large number of possible arrangements of zeros leads to pervasive indeterminacy: the assumption that any information which is not overtly signalled nonetheless corresponds to some zero morpheme leads to the formal problem of assigning a place in the structure (and linear order) to all of those zeros. (Anderson 1992: 61)

The other generalizations create even more vexing difficulties. Under “phonemic component sequences”, Harris subsumed the suprasegmental properties that Bloomfield (1933) had treated as ‘modulation’ (and which Firth (1948) termed ‘prosodies’). Classifying these properties as extended types of morphs is merely the first tentative step away from a method based on ‘dividing expressions into sequences of phonemes’. Treating ‘subtraction’ as another type of ‘alternant’ is a more decisive step in the same direction. To illustrate a subtractive pattern, Harris (1942:110) cites the formation of Hidatsa imperatives in (2.3a). The Papago forms in (2.3b) provide another familiar case, in which perfectives lack the final-m of the corresponding imperfectives.[1]

(2.3) Truncation in Hidatsa and Papago

a. cicic ‘he jumped’ ~ cic ‘jump!’, ika-c ‘he looked’ ~ ika ‘look!’

b. him ~ hi ‘walking’, hihim ~ hihi ‘walking.PL’ (Zepeda 1983)

However, it is the treatment of ablaut and other stem alternations as ‘replacements’ that signals the abandonment of a general method of analysis based on dividing expressions into segmental material. To extend an IA analysis to ablaut patterns in English, Harris (1942) proposes a complex “negative-additive” element that ‘drops’ one vowel and ‘adds’ another:[2]

In took we have two morphemes: take, and /ej/~/u/ ‘past time’. The latter occurs also in shook as compared with shake. It is a combination of negative and additive sequences: dropping /ej/ and adding /u/. Another negative-additive morpheme is /a/~/e/ ‘plural’, which occurs in men as compared with man. (Harris 1942:171)

The use of subtractive and replacive morphs introduces non-segmental ‘items’ that cannot be brought into a linear ‘arrangement’ with segmental material. As Matthews (1972:59) emphasizes, “the attempt to disguise ‘replacement’ as a segment” merely confounds segments and processes:

The notion of ‘replacement’... is one which is quite foreign to the Item and Arrangement view of language. What is involved is not a certain segment in a certain position... but the process by which the segment arrived in such a position; to speak of this process as a

‘morph, or as the ‘allomorph’ of a particular morpheme, would be a blatant conceptual error. (Matthews 1972: 59)

Process morphs also create seemingly intractable difficulties for the notion of ‘arrangement, as reflected in the deliberations of Hockett (1947):

Men is therefore morphemically {man} + {s}. But—so runs the argument that would set up alternation morphs—men and man resemble each other in phonemic shape, both containing m-n ... One morph in men is man. The other is the alternation a~e. Or - arguing now for a zero morph - men ... consists of an alternant men of {man} plus an alternant /0/ of {s}. (Hockett 1947:340)

On an analysis on which “[o]ne morph in men is man. The other is the alternation a~e”, it is unclear how to order the morphs man and a~e. The second alternative considered by Hockett adopts the solution that Bernard Bloch had developed in his analysis of stem allomorphyin English verbs. Bloch (1947:404) treated ablauted preterites such as took as consisting of a tense-neutral stem allomorph took and a zero inflectional marker. The plural men could be handled similarly in terms of a number-neutral stem allomorph of the morpheme {man}, followed by a zero allomorph of plural {s}.

The indirect treatments of truncation and stem ablaut were necessitated by the agglutinative bias of the IA model. The model reduced all form variation to affixation by encapsulating alternations in ‘items’ that could be concatenated. The ‘primary’ items consisted, as expected, of segmental material. Non-affixal alternations were expressed by ‘secondary’ (or ‘trojan horse’) items that could induce changes in morphotactic structure. Yet the unification achieved by treating these items as ‘morphs’ was never more than terminological:

the new definition of‘morph’ is no longer that with which we began; perhaps, therefore, it would be advisable to distinguish terminologically between, say, ‘primary morphs’ (those of overt phonemic content) and ‘extended morphs’ (including primary ones and morphs of the zero, replacement, or subtraction types). (Hockett 1947:240)

  • [1] Subtraction tends to be invoked in cases in which the truncated or remnant unit can be definedprosodically but not segmentally, so that a single subtraction process describes an alternation requiringmultiple, segmentally distinct, affixal patterns.
  • [2] Although Harris (1942) refers to this element as a ‘morpheme’ he seems to mean ‘morphemealternant, corresponding to what Hockett (1947) calls ‘replacive’ morphs.
< Prev   CONTENTS   Source   Next >

Related topics