Desktop version

Home arrow Language & Literature

From word to paradigm

The patterns summarized in Section 7.1.1 illustrate one type of evidence that underpins the classical WP view that associating grammatical properties with word forms creates less uncertainty than associating properties with sub-word units. This does not of course entail that classical WP descriptions treat words as unanalyzable wholes or deny the significance of sub-word units. Rather, as mentioned earlier, these models treat sub-word variation as exhibiting contrasts that discriminate between larger meaningful units. This perspective is clearly expressed in Matthew’s summary of Priscian’s treatment of Latin stem syncretism (cited on p. 103 and repeated below). As Matthews notes, an inflected form of the future active participle (the nominative singular masculine) is obtained by replacing the -m ending of the supine by the ending -rus. This analysis does not assign any grammatical properties to -rus or -m, or explicitly classify -r either as a stem formative or inflectional exponent:[1]

Priscian’s rule would actually derive the Nominative Singular Masculine ... by the addition of rus to the u: form of the ‘Supine’ and would moreover be limited to verbs in which the ‘Supine’ is not deficient... (Matthews 1972: 85)

From a classical WP perspective, sub-word units may serve discriminative roles without functioning as signs. In the example considered above, the contrast between -m and -rus distinguishes the supine and nominative singular masculine form of the future active participle within the lexeme of amo. At the same time, the contrast between-tum and -re distinguishes the supine from the infinitive of amo. The contrast between ama- and moni- likewise distinguishes the supines (and other members of the supine series) of amo and moneO. In each case, sub-word variation distinguishes pairs of words that can be assigned stable grammatical properties. But the grammatical difference between those words is not assigned to the variants that distinguish them.

From a contemporary perspective, it might seem that the stability of a word- based description incurs a significant cost. In a Post-Bloomfieldian model, the association of properties with sub-word units is designed in part to facilitate the interpretation of novel forms. Although a particular form may be unfamiliar, it can be understood by combining the properties of previously-encountered parts. But if meaning resides wholly at the word level, how do speakers avoid the need to learn the meaning of each word individually?

The answer offered by proponents of classical WP models is that the relation between word forms and the properties of paradigm cells is constrained in various ways by the larger grammatical system in which they are embedded. Robins’s quote from p. 67, repeated below, provides a concise statement of the classical WP position:

In many ways, and quite apart from any phonological markers, the word is a unique entity in grammar, and not just a stage in the progression ‘from morpheme to utterance’. As a grammatical element the word is unique in its relative fixity of internal morphemic structure, its focal status in relation to syntactically relevant categories, and, in inflected words, the stability of its paradigms. All of these factors make it a strong basis for grammatical description, both morphological and syntactic. The assumption of a simple ascent in order of size from single morpheme to complete sentence, ignoring or blurring the distinction of morphological structuring and syntactic structuring, achieves its apparent simplicity at the cost of neglecting or distorting patent structural features of languages. (Robins 1959:137)

As Robins notes, the structuralist morpheme is embedded in a general unifor- mitarian model. At each level of linguistic analysis, minimal units are combined to form the minimal units of the next highest level. Moreover, fundamentally similar combinatoric operations apply across the levels. In contrast, a traditional WP model assumes that, as units increase in size in a grammatical analysis, the domains over which relations apply to those units scale up as well. At the scale of phoneme-sized units, alternations such as assimilation or dissimilation apply largely to adjacent segments and even suprasegmental patterns tend to operate over a contiguous domain. The paradigmatic pressures that are taken to determine the adaptive dispersion of vowel inventories (Liljencrants and Lindblom 1972) likewise repel elements that are similar along some dimension in a perceptual space. The domains relevant to word-sized units are larger and the relations are more complex, due in part to the role that words play as the primary point of entry for grammatical meaning. Along the syntagmatic axis, collocation relations may hold between nonadjacent words. Paradigmatic relations likewise operate over large sets of words, from inflectional paradigms, to lexemes and derivational families.

It is the affiliation with these larger sets of forms that principally constrains uncertainty in the association between individual word forms and grammatical properties. For each cell in a paradigm, there is uncertainty about which inflectional variant realizes the cell. This uncertainty correlates with the amount of allomorphy exhibited by the realization and the distribution of allomorphic patterns. For each form, there is likewise uncertainty that correlates with the number of cells that it realizes. For a simple illustration, consider the Russian noun paradigms in Table 7.5 (repeated from Table 4.1 on p. 72 above).

These forms exhibit the many-many associations between cells and exponence patterns that characterize languages with inflection classes. The majority of cells are realized by at least two different patterns across different classes, as shown in Table 7.6. Conversely, the majority of exponents realize multiple cells. This is shown in Table 7.7, in which the exponents all happen to be suffixes that follow an inflectionally invariant stem form ‘X’.

Nevertheless, each paradigm cell has a single realization in Table 7.5. The generality of this pattern underpins the use of tables to exhibit paradigms (and

Table 7.5 Inanimate ‘soft stem’ declensions in Russian (Timberlake 2004)

I (Masculine)

II (Feminine)

III (Feminine)

Sg

Pl

Sg

Pl

Sg

Pl

Nom

slovar’

slovari

nedelja

nedeli

tetrad’

tetradi

Gen

slovarja

slovarej

nedeli

nedel’

tetradi

tetradej

Dat

slovarju

slovarjam

nedele

nedeljam

tetradi

tetradjam

Acc

slovar’

slovari

nedelju

nedeli

tetrad’

tetradi

Inst

slovarem

slovarjami

nedelej

nedeljami

tetradju

tetradjami

Loc

slovare

slovarjax

nedele

nedeljax

tetradi

tetradjax

‘dictionary’

‘week’

‘notebook’

Table 7.6 Cell-exponent ambiguity in Table 7.5

Nom

Gen

Dat

Acc

Inst

Loc

Sg

2

2

3

2

3

2

Plu

1

2

1

2

1

1

Table 7.7 Exponent-cell ambiguity in Table 7.5

Exponent

Xi

X

Xu

Xa

Xe Xej

Cells

5

3

3

2

2 2

Exponent

Xem

Xam

Xami

Xax

Cells

1

1

1

1

the definition of paradigm functions in Stump (2001)). More strikingly, nearly every form of first declension slovar’ ‘dictionary’ receives a single analysis. The sole exceptions are the syncretic accusative and nominative forms, which reflect a general pattern for inanimate nouns in Slavic.[2]

The contribution that paradigms make to reducing uncertainty is exhibited by the contrast between the relatively stable form-property associations in Table 7.5 and the variation displayed in Tables 7.6 and 7.7. In isolation, most of the forms of slovar’ are potentially ambiguous. This ambiguity cannot be resolved by disassembling these forms, since it derives ultimately from the fact that recurrent inflectional endings are associated with different cells across classes. It is instead the structure provided by the paradigmatic affiliations of these classes of interdependent forms that effectively constrains variation.

The disambiguating function of paradigms also clarifies where realizational approaches depart from classical WP models. In a WP model, locating words within a larger set of interdependent forms constrains the association between the words and grammatical properties. Class membership is not a property of forms, but is exhibited by patterns of associations between properties and forms.[3] By encoding class membership by means of class ‘features’, Matthews (1965) explicitly represents the organization that remains implicit in the ‘exemplary paradigm and principal parts’ formulation of a classical grammar. Yet this strategy collapses the paradigmatic structure of a classical WP model. The implicational structure exhibited by classes is reduced to an essentially neo-Bloomfieldian relation of ‘selection’ that governs the compatibility of the stems and exponents modified or introduced by realization rules.

  • [1] This analysis is repeated by Gildersleeve and Lodge (1895) on p. 103 above.
  • [2] The increase in ambiguity in the second and third declensions is attributable to other generalsyncretisms. The paradigm of second declension nedela ‘week’ contains a distinctive accusativesingular form but exhibits a syncretism between dative and locative singular and between genitivesingular and nominative plural. The paradigm of third declension tetrad’ ‘notebook’ preserves thesyncretisms from the first and second declension and, in addition, collapses the genitive and dativesingular forms.
  • [3] The contrast between specifying class features and exhibiting class affiliation is reminiscent of thedistinction between Sagen and Zeigen in Wittgenstein (1921).
 
Source
< Prev   CONTENTS   Source   Next >

Related topics