Desktop version

Home arrow Language & Literature arrow Word and paradigm morphology

Exponence relations

Examples of this kind do not show that formatives cannot be associated with properties, nor that a biunique property-form correspondence cannot be established. What they do show is that these positions cannot be maintained simultaneously. No conflict arises in classical models, which preserve biuniqueness but do not associate properties with formatives.[1] As with Latin rexistt above, there is a biunique relation between the word form elelykete and the set of properties associated with the ‘2nd person plural past perfective indicative active’ paradigm cell of the verb lyo. Although the relation between individual properties and formatives is not similarly biunique, there is still a correlation between these elements. Following Matthews (1972, 1991), this looser property-form correlation is termed (morphological) exponence. The principal types of exponence relations are depicted in Figure 3.3.

The ‘Morphemic’ pattern in Figure 3.3 exhibits the biunique property-formative correspondence associated with the Post-Bloomfieldian morpheme. There is nothing to exclude this pattern within a WP account, and the analyses of the formatives e- and -le- in Figure 3.2 conform to the morphemic ideal. However, there is

Types of exponence (cf. Matthews 1991

Figure 3.3 Types of exponence (cf. Matthews 1991: i7off)

Table 3.1 Regular conjugation classes in Spanish

I

II

III

Infinitive

amar

vender

partir

1sg Future

amare

vendere

part ire

‘love’

‘sell’

‘leave’

also nothing normative about this configuration, and the remaining patterns in Figure 3.3 all deviate from it in one way or another. The fusional or ‘Cumulative’ pattern in Figure 3.3 is illustrated by each of the remaining exponents in Figure 3.2, from the root ly through the agreement suffix -fe.Thefissionalor‘Ext ended’ pattern in Figure 3.3 is illustrated by the spell-out of the perfective, active and past features in Figure 3.2.

The ‘Empty’ and ‘Zero’ patterns represent the two extremes in which there is no exponence relation. A formative is empty when it is not associated with any properties and properties are realized by zero when they are not associated with any formatives. These patterns are symmetrical in that they both involve a complete dissociation of properties and morphs. However they are also very different in character. Incontrovertibly empty morphs appear to be comparatively rare, whereas zero exponence is a pervasive and systematic phenomenon. Theme vowels in Romance are among the most commonly-cited examples of empty morphs (Hockett 1947:337, Anderson 1992:54). In Spanish, verbs are normally assigned to conjugations based on the final vowel in the infinitive. The first conjugation contains regular verbs with infinitives in-ar, the second conjugation contains verbs with infinitives in -er, and the third conjugation contains verbs with infinitives in -ir. Verbs preserve the conjugation vowel in future (and conditional) forms, which consist of an infinitive stem and an ending corresponding to a reduced present (or conditional) form of the auxiliary haber ‘to have’. Table 3.1 illustrates these patterns.

Although these elements do not carry any clear ‘denotative’ meaning, they nevertheless serve as markers of ‘conjugational class’ and thereby convey morphological information about the inflectional patterns followed by other forms of an item.[2] Thus the variation in infinitive endings in Table 3.1 is largely preserved in the present indicative paradigms in Table 3.2.

Table 3.2 Present indicative paradigms in Spanish

Sg

i

Plu

Sg

ii

Plu

Sg

iii

Plu

1st

amo

amamos

vendo

vendemos

parto

partimos

2nd

amas

amais

vendes

vendeis

partes

partimos

3rd

ama

aman

vende

venden

parte

parten

Table 3.3 Linking elements in German (Duden 2005)

-n-

-s-

Nom Sg

Spinne

Rakete

Arbeit

Liebe

Nom Plu

Spinnen

Rakaten

Compound

Spinnennetz

Rakatenstufe

Arbeitsamt

Liebesbrief

‘spider (web)’

‘rocket (stage)’

‘job (center)’

‘love (letter)’

Formatives that occur in derived forms and compounds are also frequently regarded as empty morphs (or, equivalently, as unanalyzed parts of ‘stem allo- morphs’). This type of account goes back at least as far as the treatment of ancient Greek in Bloomfield (1933:225). Bloomfield analyzes Greek nouns as having a “kernel” that underlies their inflectional paradigm and a ‘stem allomorph’ that functions as a “deriving form” and as a “compounding form”. On this analysis, the noun hippos ‘horse’ has the kernel hipp and the stem allomorph hippo, which occurs in hippote:s ‘horseman’ and hippokantharos ‘horse-beetle’. Significantly, the Bloomfieldian account contrasts whole stems without assigning any individual properties to the formative -o-:

Thus, we distinguish between the kernel [hipp-], which actually... appears in all the [inflected JPB] forms and the stem [hipp-o-], which underlies the further derivation. (Bloomfield 1933:225f.)

The Fugenelemente ‘linking elements’ that occur in West Germanic compounds are also commonly treated as empty formatives (Wiese 1996, Booij 1997). Table 3.3 illustrates two patterns involving feminine nouns in German. Regular feminine nouns ending in schwa form their plural in -n, as illustrated by the pairs Spinne^Spinnen and Rakete^Raketen. When nouns of this class occur as the first element of a compound, they also take the n-form, almost always obligatorily, and irrespective of the number features of the compound.

Compound bases in -n thus represent a simple case of ‘parasitic’ (Matthews 1972) or ‘morphomic’ (Aronoff 1994) syncretism, as they share the form but not the features of the corresponding plural.[3] The compounds based on arbeit and liebe in Table 3.3 illustrate the opposite pattern, in which the linking element -sis fully dissociated from the inflectional paradigms of these nouns. The formative s occurs in two places in common noun declensions, as a general plural form, and as a genitive singular marker for strong masculines. The linking element -s- cannot be assigned either analysis in Table 3.3. Both nouns are abstract here and have no real plural, though the the plurals of corresponding count-noun senses would in any case be be formed in -(e)n. As feminines, these nouns are also uninflected in the singular (as described in more detail in Table 4.9). Hence -s- is a pure ‘linking element’ in these compounds and has at most an analogical connection to plural or genitive markers.

As with conjugation vowels, the classification of linking elements as ‘empty’ reflects what may just be an excessively narrow conception of ‘meaning’. As Carstairs-McCarthy (1994) argues in a different context, class affiliation can be regarded as the meaning of a conjugation vowel. Realizational models adopt this expanded notion of meaning when they treat theme vowels (or other types of elements) as the ‘spell-out’ of class ‘features’. From a classical WP perspective, the same basic idea can be expressed in terms of the implicational relations between theme elements and class-specific patterns of form variation. In either case, the ‘meaning’ carried by these elements is system-internal, providing information about other forms or patterns within the system. Bloomfield’s ‘deriving’ and ‘compounding’ forms can be viewed in a similar light, as conveying information about ‘construction type’.

In some systems, elements may occur in a range of contexts, inflectional and derivational. Whitney (1889) describes a pattern of this kind in Sanskrit:

All the simple vowels come to assume in certain cases the aspect of union-vowels, or insertions between root or stem and ending of inflection or of derivation.

a. That character belongs oftenest to i, which is very widely used: 1. before the s of aorist and future and desiderative stems... 2. in tense-inflection, especially perfect... 3. in derivation ... (Whitney 1889: 86)

These types of cases present acute difficulties for an analysis that seeks to assign a common meaning that spans the contexts in which elements like “union-vowels” occur. It is of course conceivable that these formatives are separate elements, with only a residual diachronic connection. However, it is also possible that the choice of union vowels, or their patterns of alternation and distribution, may play a role in identifying lexical or morphological classes in a language. Because a classical account does not treat formatives as signs in the sense of Saussure (1916), there is no expectation that they will be associated with discrete meanings. Hence where the analysis of words into smaller units isolates elements that cannot be associated with any discernible meaning or function, this can be regarded as a symptom of morphological ‘overextraction’, i.e., the attempt to associate properties with submeaningful units.

Anderson (1992: 89) identifies the /t/ in Menominee ke-t-os ‘your canoe’ as a genuinely empty formative, which he characterizes as a “morphologically conditioned concomitant of the addition of a possessive marker to a vowel-initial stem”. However, even ‘conditioned’ formatives of this kind will not be entirely

‘empty’ if they serve to discriminate possessive constructions from other forms that begin with the same initial sequences of phonemes. More generally, the communicative function of language leads one to expect that ‘morphological noise’ will be comparatively rare and unstable, as speakers and learners assign some kind of interpretation to patterns of form variation. In many cases, an understanding of the function of ‘empty’ elements will also require statistical analysis of paradigmatic or syntagmatic distribution.

The pervasive phenomenon of ‘zero exponence’ can be regarded as an extreme case of pattern-based meaning, in which speakers assign interpretations based on the absence of elements that occur elsewhere in a system. Noun paradigms in English provide a simple example. Plural nouns in English are marked by the suffix /z/ (represented orthographically as -s). There is, however, no marking of singular number, and none is needed, given that a singular noun is unambiguously identified by the lack of a plural marker. A zero morph adds no information to what speakers can already deduce from the absence of any realized exponent.[4] Similar patterns are even more typical of more intricate paradigms. As Anderson (1992) notes, Georgian verbal paradigms provide a striking illustration of the fact that that “information may sometimes be conveyed not by constituents that are present in the structure of a given word, but precisely by the fact that certain other material is absent”:

Consider the Georgian Verb form mogk’lav in [Table 3.4], for example,... This form represents agreement with a first-person singular Subject and a second-person singular Direct Object... But while an overt affix (/g/) is present to signal agreement with the second- person Object, no affix marks the fact that the Subject of this Verb is (and must necessarily be) first-person singular. This agreement can be inferred from the following information. The Subject cannot be second person, because if it were, the sentence would be reflexive— but reflexive forms in Georgian are grammatically third person, and this Verb has a second- person Object. Similarly, the Subject cannot be third person, since, if it were, there would be a suffix (/-s/) at the end of the Verb. Thus... the Subject must be first person. But it must be singular, rather than plural, since a first-person plural Subject would trigger the introduction of a suffix /t/ at the end of the Verb. We know therefore that the Subject of this Verb must be first-person singular, but this fact is not signaled by the presence of any overt affix in the word. (Anderson 1992: 87)

The interpretation of an unmarked form need not invoke this kind of‘process of elimination’ deduction each time that the form is encountered. Instead, a classical model can treat this deduction as part of an acquisition process in which speakers place or locate forms within the paradigm of an item. It is the lack of this larger

Table 3.4 Future indicative paradigm of k’vla ‘kill’ (Tschenkeli 1958: §31)

Subject

1 Sg

1 Pl

Object 2 Sg

2 Pl

3

1 Sg

-

-

mogk’lav

mogk’lavt

movk’lav

1 Pl

-

-

mogk’lavt

mogk’lavt

movk’lavt

2 Sg

momk’lav

mogvk’lav

-

-

mok’lav

2 Pl

momk’lavt

mogvk’lavt

-

-

mok’lavt

3 Sg

momk’lavs

mogvk’lavs

mogk’lavs

mogk’lavt

mok’lavs

3 Pl

momk’laven

mogvk’laven

mogk’laven

mogk’laven

mok’laven

paradigmatic context in Post-Bloomfieldian models that creates the need for ‘zero morphs’ to record the properties of unmarked forms.

In sum, words maybe divided into formatives in a WP model, but this division does not in general yield individually meaningful parts. Words remain the smallest Saussurean ‘signs’ of a language. As the passage from Robins (1959) on p. 6 above makes clear, this treatment of words rests on the empirical claim that grammatical properties are more reliably associated with whole words (or larger expressions), than with component formatives.8 A WP approach is still able to represent the internal structure of word forms and even, via exponence relations, express any systematic correlations between this structure and grammatical properties. The use of analogical strategies to interpret or extend morphological patterns in WP approaches also attributes a measure of internal structure to the words that define patterns. This structure may remain implicit, as in the “patterns of modification” given in classical grammars (Matthews 1991:191), or it may be expressed more explicitly, as in proportional analogies (Paul 1920). But in either case, analogical strategies exploit the predictive value of word structure without mediating those predictions through the assignment of properties to individual formatives.

  • [1] Indeed, classical variants of the WP model do not even recognize morphological units betweenwords and phonemes/graphemes (Matthews 1994; Law 1998).
  • [2] It is worth bearing in mind that ostensibly ‘pure’ morphological classes may show statisticaltendencies to align with syntactic or semantic classes.
  • [3] Morphomic patterns are discussed in more detail in Chapter 5.
  • [4] As noted in Chapter 2, some Post-Bloomfieldian accounts introduced a ‘zero’ singular marker insuch cases (Harris 1942:110), and the practice survives in contemporary work. Hockett (1947:230)recognized that this type of element had “a very dubious status”, and the coherence of a ‘zero marker’ isquestioned by Matthews (1991:124). From a classical WP perspective, the appeal to ‘zeros’ can be seenas a means of compensating for the rigidly syntagmatic character of the Post-Bloomfieldian model,which has no provision for any type of paradigmatic comparison or deduction.
 
Source
< Prev   CONTENTS   Source   Next >

Related topics