Desktop version

Home arrow Language & Literature

'An agglutinating system gone wrong'

Much of the appeal of the IA model derived from the apparently simple and intuitive treatment that it offered for concatenative patterns. A regular plural like

hoofs could be broken down into a stem form, hoof, which realizes lexical meaning, and an inflectional form, -s, that realizes plurality. Yet complications arise in extending this type of analysis even to the full range of inflectional patterns of English. More complex inflectional systems raised further challenges. To preserve the letter of the IA model, the Descriptivists explored a variety of mainly technical solutions. Mismatches between ‘units of meaning’ and ‘units of form’ are repaired through the postulation of more “special kinds of morphs”:

The alteration [of Harris (1942)] by which the number of morphemes in an utterance fails in some cases to coincide with the number of morphs consists of recognizing two special kinds of morphs: empty morphs, which have no meaning and belong to no morpheme; and portmanteau morphs, which belong simultaneously to two (or, theoretically, more) morphemes, and have simultaneously the meaning of both. (Hockett 1947:236)

As examples of empty morphs, Hockett (1947) cites theme vowels in Spanish conjugations. To illustrate portmanteau morphs, he gives the example of the French contraction au, which occurs in place of the preposition and article sequence a le. Yet these additional morphs merely created greater analytical indeterminacy, without addressing the source of meaning-form mismatches.

Portmanteau morphs provide an alternative analysis of patterns that can be described in terms of zeros. In the comments that precede the discussion of men on p. 31 above, Hockett (1947:339) acknowledges that “one solution, and certainly the most obvious one, is to regard men as a single portmanteau morph, representing the morpheme sequence {man} + {s}”. Thus the IA model makes available at least the three analyses in Figure 2.2: (i) man and a replacive plural allomorph, (ii) men and a zero plural allomorph, and (iii) portmanteau men, without providing criteria for choosing between them.

More generally, allowing empty and portmanteau morphs largely rescinds the ‘principle of total accountability’ on page 24 above. Whenever there are more morphs than morphemes, the excess morphs can be treated as ‘empty’; if there are more morphemes than morphs, the shortfall can be made up by invoking portmanteau (or zero) morphs. Taken together, these refinements amount to the abandonment of the central empirical claims of the IA model. Worse yet, the strategy of introducing a special morph to handle each type of mismatch between a ‘unit of meaning’ and a ‘unit of form’ did not address the source of these mismatches, so much as treat their symptoms.

Languages of the ‘flectional’ type, expose the severe limitations of this strategy. As Matthews (1972:1326!) shows, the Latin verb re:ksisti: ‘you had ruled’ (rexistl in the standard pedagogical orthography) exhibits the many-many feature-form

Candidate IA analyses of English men

Figure 2.2 Candidate IA analyses of English men

Morphological analysis of Latin re:ksisti (Matthews 1972:132)

Figure 2.3 Morphological analysis of Latin re:ksisti (Matthews 1972:132)

Many-to-many feature-form associations

Figure 2.4 Many-to-many feature-form associations

relations displayed in Figure 2.3. The ending -ti: exhibits what is usually termed a ‘fusional’ pattern, in which a single formative realizes the features perfective aspect, 2nd person and singular number. At the same time, the perfective feature exhibits a converse ‘fissional’ pattern, as it is realized by each of the formatives -s-, -is-, and -ti:.

From the standpoint of a classical grammar, Figure 2.3 is a case of ‘overextraction, in which meaning is associated with sub-meaningful units of form. In the conjugational paradigm of rego, there is a unique (in this case, biunique) correspondence between the second person perfective active cell and the word form re:ksisti:. However, no such correspondence can be established between the features contained in the cell and the formatives that make up the word form.[1] Instead, further decomposition of the features of the cell and further disassembly of the form produce the non-binunique pattern in Figure 2.4.

As in the simpler patterns discussed by Harris and Hockett, there are technical solutions for the challenges posed by flectional languages. The ‘extended expo- nence’ (Matthews 1972) illustrated by the multiple perfective markers in Figure 2.4 can be accommodated by designating one formative as ‘primary’ and excluding any other, ‘secondary’ exponents in the determination of morphemic biuniqueness.12 Indeed, as the subsequent history of the IA model shows, an item-based analysis can be extended to nearly any recalcitrant pattern through the introduction of special morphs or construction-specific conventions. However, the original goal of the IA model was not solely to maximize the use of items, but to bring out the correspondence between items and units of meaning. If, as Matthews (1972) notes, this correspondence does not obtain, the central motivation for items and for the IA model itself is undermined:

One motive for the post-Bloomfieldian model consisted, that is to say, in a genuinely factual assertion about language: namely, that there is some sort of matching between minimal ‘sames’ of ‘form’ (morphs) and ‘meaning’ (morphemes). Qua factual assertion this has subsequently proved false: for certain languages, such as Latin, the correspondence which was envisaged apparently does not exist... One is bound to suspect, in the light of such a conclusion, that the model is in some sense wrong. (Matthews 1972:124)

  • [1] Parallel remarks apply to more ‘morphological’ relations. The perfect indicative active is one ofthe four ‘principal parts’ of a Latin verb identified by pedagogical descriptions such as Hale and Buck(1903). Although it is conventional to cite the perfect indicative in the lsg, the 2sg form re:ksisti: isequally ‘diagnostic’ of a third conjugation verb like rego. As it happens, the consonant-final stem re:kis also diagnostic of the third conjugation, though this reflects the fact that inflections in the perfectactive series do not exhibit class-specific variation in Latin and is not true of stems in general. See Carstairs (1987) and Noyer (1992) for accounts that develop this contrast, and Harris (2009)for a comprehensive discussion of extended exponence.
< Prev   CONTENTS   Source   Next >

Related topics