Desktop version

Home arrow Language & Literature arrow Word and paradigm morphology

Rule blocks and disjunctive ordering

The agreement markers that Anderson discusses occur at ‘positions’ 2 and 11 in the schematic morphotactic structure proposed in Hewitt (1995):

The following are the morphemes that may occur in a Georgian verb-form, though not necessarily simultaneously: 1. Preverb(s), 2. Pronominal Agreement-Prefix, 3. Version- Vowel, 4. Root, 5. Causative Suffix(es), 6. Inceptive/Passive Marker, 7. Thematic Suffix, 8. Perfect/Stative Marker, 9. Imperfect Marker, 10. Mood-Vowel, 11. Pronominal Agreement- Suffixes). (Hewitt 1995:117)

The perfect/evidential form dagixatavt ‘we have (apparently) painted him/her/ them’ in Table 6.4 contains formatives in roughly half of the slots identified by Hewitt (1995). Anderson (1986, 1992) suggests that the rules that introduce these formatives are organized into linear blocks, and asserts in particular that the agreement rules in block 2 interact with those in block 11.

The evidence for cross-block rule interaction comes from the distribution of the agreement suffix -t. Consider first the future indicative paradigm of xat’va ‘paint’ in Table 6.5, in which the agreement markers are in bold. Table 6.6 shows that -t may mark agreement with 1pl and 2pl subjects. Table 6.7 further shows that that -t may mark agreement with 2pl objects. However, as Tables 6.5, 6.6, and 6.7 indicate, there are no contexts in this paradigm, or more generally in the conjugational system of standard Georgian, in which -t unambiguously marks 1pl subject agreement.

Agreement with 3pl subjects is marked by -en in Table 6.5, and by different 3pl markers in the other conjugational series in Georgian. In the Present and Aorist series, verbs do not mark a number contrast for 3p objects. However, paradigms in the Perfect series (along with paradigms of ‘indirect’ verbs) do exhibit a contrast between 3sg and 3pl indirect object agreement features, as a consequence of an

Table 6.4 Block stem structure of dagixatavt

Table 6.5 Future indicative paradigm of xat’va ‘paint’ (Tschenkeli 1958: §31)

Subj

1sg

1pl

Obj

2sg

2pl

3

1sg

dagxat’av

dagxat’avt

davxat’av

1pl

dagxat’avt

dagxat’avt

davxat’avt

2sg

damxat’av

dagvxat’av

daxat’av

2pl

damxat’avt

dagvxat’avt

daxat’avt

3sg

damxat’avs

dagvxat’avs

dagxat’avs

dagxat’avt

daxat’avs

3pl

damxat’aven

dagvxat’aven

dagxat’aven

dagxat’aven

daxat’aven

Table 6.6 ipl and 2pl subject agreement marking by-t

Subj

Isg

Ipl

2sg

2pl

Objlsg

damxat’av

damxat’avt

2sg

dagxat’av

dagxat’avt

3

davxat’av

davxat’avt

daxat’av

daxat’avt

Table 6.7 2pl object agreement marking by-t

Obj

Isg

Ip!

2sg

2p!

Subjlsg

dagxat’av

dagxat’avt

2sg

damxat’av

dagvxat’av

3sg

damxat’avs

dagvxat’avs

dagxat’avs

dagxat’avt

Table 6.8 ‘Inverted’ pluperfect paradigm of xat’va ‘paint’ (Tschenkeli 1958)

Iobj

Subj

1 sg

1 P!

2 sg

2 p!

3 sg 3 pl

1 sg

dagexat’e

dagexat ’et

davexat e

ipl

dagexat’et

dagexat ’et

davexat et

2 sg

damexat’e

dagvexat’e

daexat e

2 pl

damexat’et

dagvexat’et

daexat’ et

3

damexat’a

dagvexat'a

dagexat a

dagexat at

daexat a daexat at

inversion alternation that maps logical subjects onto indirect objects and promotes logical objects to surface subjects.10 11 The pluperfect paradigm in Table 6.8 exhibits a contrast between 3sg and 3pl object marking, but only in forms with 3rd person subjects. In these forms, -t marks 3pl indirect objects, as well as ipl and 2pl subjects and 2pl objects.[1] [2]

The aim of Anderson’s analysis is to formulate a single ‘elsewhere’ realization rule that introduces -t in the non-natural environments in which it seems to occur. The analysis incorporates two main claims. The first is that -t is neither a dedicated subject nor object plural marker, but instead “simply a perfectly general marker of the presence of a plural argument” (Anderson 1986:12). The second claim is that in every plural context where -t fails to occur, it is blocked by some other rule. If we restrict attention to the future paradigm in Table 6.5, the distribution of -t is described in Table 6.9. The dashes again identify cells realized by reflexive forms which do not contain the relevant markers. Cells with no plural features to realize

Table 6.9 Occurrence and blocking of-t

Obj

1sg

1pl

2sg

2pl

3

Subjlsg

-t

1pl

-t

-t

-t

2sg

0

2pl

-t

-t

-t

3sg

-s

-t

3pl

-en

-en

-en

-en

-en

are also marked by dashes. The remaining cells are indicated by the suffix realized in that cell.

Anderson proposes that the 3pl subject marker -en preempts -t because the 3rd person plural features realized by the corresponding rule are more specific than the person-neutral plural features realized by the rule that introduces -t. This leaves the absence of-t in 1pl object forms as the sole remaining gap. As Anderson notes, the occurrence of the 3sg subject suffix -s in the realization of 1pl object and 3sg subject features cannot be attributed to competition with -t, given that -t preempts -s in the realization of 2pl object and 3sg subject features. The absence of -t also cannot be attributed to any suffixal competition in the unsuffixed form that realizes 1pl object and 2sg subject features.

The solution that Anderson proposes is that the absence of -t in forms realizing 1pl object features is due to competition with the 1pl prefixgv-:

Clearly the gv- rule is more specific than the t-rule... and thus the operation of the former excludes the latter by disjunctive ordering. (Anderson 1986:10)

To allow prefixes to block suffixes in Georgian, Anderson proposes the principle in

  • (6.3) , which extends a disjunctive ordering condition across blocks:
  • (6.3) “Elsewhere” Principle: Application of a more specific rule blocks that of a later more general one. (Anderson 1992:132)

The principle in (6.3) is a highly subversive ‘extension’ of a disjunctive ordering condition, since it undermines the coherent division between the intrinsic ordering that applies within blocks and the extrinsic ordering that applies across blocks. Moreover, the the descriptive usefulness of this extension depends on a number of stipulations that are independently open to question. For the rules introducing prefixes to take priority over those introducing suffixes, it is critical that the prefixal block precedes the suffixal block. Yet there is no motivation provided for this assumption. However, it is perhaps suggestive that the two sets of agreement markers occur outside the derivational exponents, and show the uniform distribution and interpretation characteristic of inflections. The agreement suffixes are the final element in the template identified by Hewitt (1995), corresponding to the agreement termination -te in Figure 6.21. On the left flank, the agreement prefixes follow only the preverb. Hence on almost any stem analysis, the agreement blocks

Table 6.10 Block stem structure of dagixat’avt

da

g

i

xat’

av

t

Linear

1

2

3

4

7

11

Nested

A

C

B

A

B

C

General plural rule in Georgian (Anderson 1986:12,1992:132)

Figure 6.22 General plural rule in Georgian (Anderson 1986:12,1992:132)

would be peripheral, as in Figure 6.10, outside the blocks that introduce the root, version vowels, thematic suffixes, and other derivational exponents.12

In principle, a block structure that mirrored a traditional stem structure could extend the domain of disjunctive ordering in precisely the way required for an analysis that treats -t as a general plural marker whose distribution is constrained by competition with agreement prefixes and suffixes. Anderson (1992:131) alludes to this alternative when he notes that “blocks are (in part) a reconstruction of the traditional notion of position classes in morphological structure” and goes on to acknowledge that the general conception of rule block is flexible enough that “a rule of prefixation and a rule of suffixation could potentially belong to the same disjunctive block”. Hence, if one were to assume that -t is a general plural marker whose distribution is constrained by competition with more specific prefixes and suffixes, the most natural implementation of an analysis on which gv- blocks -t would adopt a nested block structure. So, even accepting that prefixes can compete with suffixes provides no support for the claim that disjunctive ordering needs to span rule blocks.

Conversely, there are compelling grounds for questioning both the assumption that -t is a general plural marker and the claim that its distribution is due to disjunctive ordering. The argument that Anderson (1986,1992) develops against a morphemic analysis hinges on the claim that -t is introduced by a single rule. The formulations of this rule from Anderson (1986) and Anderson (1992) are repeated in Figure 6.22.

The most obvious problem with these formulations is that neither expresses a realization rule. Instead, both describe a family of rules. As the paradigms in Tables 6.5 and 6.8 show, Georgian verbs may mark the agreement features of subjects and direct or indirect objects. A bundle representing the features of a Georgian verb must likewise represent the agreement features of multiple arguments. In order to spell out the features of such bundles, a realization rule [3]

Table 6.11 Object marking by-I

Isg

2sg

3

Obj

Ip!

dagvxat’av

dagvxat’avs

2pl

dagxat’avt

dagxat’avt

must in turn specify which of these feature sets it interprets. On the intended interpretation, a specification like ‘[... +plural... ]’ will be satisfied whenever plural features are present in any of the agreement specifications of a bundle. Hence this ‘rule’ is just a concise expression of a disjunction over the more specified rules that would actually apply in each of the satisfying cases. The later formulation has the same character, where “in the context of [+Plural]” does not specify an exponence relation but describes a class of rules that spell out the [+Plural] features of a subject, object and/or indirect object.

However, even if one were to assume that -t could be introduced by a single rule, it is far from clear that the distribution of this marker is attributable to disjunctive ordering. Consider the status of -t as a marker of ipl and 2pl object features. The relevant forms from Table 6.5 are repeated in Table 6.11. The forms in the second row show that -t can mark 2pl object features.

The only evidence that -t can mark the plurality of ipl objects is that it is blocked by ipl gv-. And the only evidence that 2pl gv- can apply across blocks to constrain the occurrence of suffixal markers is that it blocks -t. Outside this loop of mutually- dependent assumptions, there is no evidence that -t ever marks the plurality of ipl objects. In a case like this, it is surely much simpler to assume that -t never marks ipl object agreement, and hence doesn’t need to be blocked in the environments where it never occurs. The ultimate motivation for an appeal to blocking here is the claim that -t is a general plural marker. But given that the descriptions of the distribution of -t in Figure 6.22 describe disjunctions of realization rules in any event, the realization of -t as an object marker can be largely restricted to rules specifying 2p objects.

The final pattern that Anderson presents as support for a general plural analysis of -t involves the interaction of 3pl subject markers and -t:

since [dagxataven] can refer to either a singular or a plural second person object, we conclude that the en-rule introducing the third person plural suffix -en belongs to the same disjunctive block as the s-rule and the t-rule, and in particular that it takes precedence over the t-rule within that block. (Anderson 1986: 8)

The implicit claim is that when a rule that spells out subject agreement features takes priority over a rule spelling out object features, the second rule must realize general, argument-neutral features, since there is no principled reason why the subject agreement features of one rule should be disjunctively ordered with respect to the object agreement features of another. This claim might support a general plural analysis of -t if rules realizing different argument agreement features did not interact in Georgian. But in fact they do, in both the prefixal and suffixal blocks. The prefixal competition is illustrated by the 2pl form dagxat’avt in Table 6.11, where, as Anderson (1986: 6) notes, “we see that instead of the combination of v- and g- which we might expect, only the g- appears”. The suffixal competition is illustrated by dagxatavt, again from Table 6.11, where as Anderson (1986:8) remarks, “the suffix -t blocks the otherwise expected appearance of the suffix -s marking third person singular subject”. In the first case, a 2p object prefix takes priority over a ip subject prefix, and in the second case a 2pl object suffix takes priority over a 3sg subject. Both patterns involve agreement features of distinct arguments, and neither can be attributed to a specificity-based disjunctive ordering condition.

Various solutions to these challenges have been proposed within the theoretical literature.[4] However, before proceeding to consider some of the general issues raised by the phenomenon of ‘slot competition’ it is worth recalling that the Georgian patterns summarized above were initially presented as evidence for a blocking analysis, and specifically for an analysis in which disjunctive ordering relations could apply across rule blocks. Even if one accepts the main claims of this analysis, there is no evidence that disjunctive ordering does in fact apply across blocks. As noted above, there also appear to be compelling grounds for rejecting the main claims of this analysis. Moreover, when viewed in terms of blocking, the competition between agreement markers appears chaotic. Far from supporting the role of disjunctive ordering in a realizational analysis, agreement markers in Georgian highlight the limitations of specificity-based strategies for describing paradigm-internal variation.

  • [1] This treatment of inversion is assumed in grammars such as Tschenkeli (2958) and developedmore explicitly in Harris (2982, 2984) and Blevins (2025b).
  • [2] On the assumption that the indirect object is a ‘non-canonical’ grammatical subject in inversion,Anderson concludes that -t does not mark 3pl objects. However this assumption is independent of theother components of his analysis, and merely introduces a complication into a highly general agreementrule (Anderson 2986: 22).
  • [3] It is immaterial whether the preverb is analyzed as forming a discontinuous stem with the root,as in Figure 6.21, or treated as somehow combining with an inflected form.
  • [4] See, for example, the analyses in Carmack (1997) and Stump (2001).
 
Source
< Prev   CONTENTS   Source   Next >

Related topics