• Keine Ergebnisse gefunden

Agreement in Sentence Production

3.3 Determinants of Attraction in Sentence ProductionProduction

3.3.4 Morphophonological Factors

3.3.4.1 The Role of Number Marking

The observation that the asymmetry between singular and plural in overt number marking—singular is unmarked, plural is marked by means of morphology—goes hand in hand with an asymmetry with regard to attraction—singular controllers are more vulnerable to attraction—gives us a first hint that morphophonological marking might play a role for agreement computation. To examine this suspicion we have to compare controllers that share the same feature specification but differ in the morphological transparency of that specification. As said above, we might expect that irregular and ambiguous number marking creates some uncertainty about the subject’s number specification. It might not only increase the likelihood of agreement errors in general but also increase the likelihood of interference by a distractor. And indeed, experimental data confirm this expectation; agreement errors and in particular attraction errors are more likely when the subject NP con-tains a number ambiguity either on the noun (for Italian see Vigliocco et al., 1995) or on the determiner (see Hartsuiker et al., 2003, for Dutch and German).13 For number ambiguity on the noun, compare the sentence fragments in (32) and (33) taken from Vigliocco et al. (1995). Most Italian nouns have distinct forms for sin-gular and plural such as gatto/gatti (‘cat/cats’) in (32). Some nouns, however, e.g., menu (‘menu(es)’), have identical forms in the singular and the plural. Usually, the determiner resolves the resulting number ambiguity, e.g., the definite article as in (33).

‘the cat on the roof(s)’

b. I

‘the cats on the roof(s)’

13For comparable findings regarding gender agreement see Franck et al. (2008); Meyer and Bock (1999); Vigliocco and Zilli (1999).

(33) a. Il

In a sentence completion task with complex subject NPs like the ones in (32) and (33), participants produced more agreement errors when the head noun was num-ber ambiguous as in (33), especially when the head noun was plural. In addition, and confirming the second expectation, the attraction rate was higher with num-ber ambiguous nouns in comparison to controller nouns which are marked for number.

In Italian, the determiner disambiguates a number ambiguous noun like menu (cf. (33)). Dutch allows us to investigate the reverse configuration - noun phrases with a noun which is unambiguous with regard number and a determiner which is either ambiguous or unambiguous in this regard. The Dutch definite article is het for neuter gender and de for common gender and plural. Thus, the determiner de is number ambiguous whereas the determiner het is unambiguously singular.

(34) a. de

‘the street(s) near the church(es)’

(35) a. het

‘the square(s) near the church(es)’

Hartsuiker et al. (2003) found attraction errors after preambles like (34a), but no attraction errors after preambles like (35). Thus, the number ambiguity of the determiner de gave way to attraction while the unambiguous determiner het helped to produce the correct number specification on the verb (cf. Hartsuiker et al., 2003)). Similarly, attraction is more common with feminine subjects in German (Schriefers and van Kampen, 1993; Hartsuiker et al., 2003). The femi-nine definite article die exhibits the same number ambiguity as the Dutch article de. But German die is not only number ambiguous, it is also case ambiguous.14 In

14This is also true for Dutch de but for het as well.

contrast to the number ambiguity, the case ambiguity is not resolved by the noun.

This might create some uncertainty about the case of the complex NP and in re-sult its syntactic function. Thus, the observed attraction effect could well rere-sult from the ambiguous case marking of the subject NP rather than from the number ambiguity of the determiner. Evidence against such an explanation comes from the observation that neuter NPs resisted attraction despite their case ambiguity.

Although the neuter determiner das is case ambiguous it is unambiguous with re-spect to number. Hartsuiker et al. therefore conclude that the attraction effect with feminine subjects is indeed forced by the number ambiguity of the determiner die.

While the evidence for morphophonological effects regarding the controller noun is overwhelming, the picture for distractor nouns is less clear. Most studies found no effect of morphophonological marking on the distractor noun (Bock and Eberhard, 1993; Vigliocco et al., 1995, Haskell and MacDonald, 2003, experiment 2), but there are also some hints of possible morphophonological effects (Haskell and MacDonald, 2003, experiment 3).

Consider first the case of pseudo-plurals, i.e. distractors which look or rather sound like plurals nouns but are actually singular nouns (e.g., cruise resembling the real plural noun crews). As Bock and Eberhard (1993) have demonstrated, pseudo-plurals do not elicit attraction errors. The error rate in sentences with a pseudo-plural distractor (e.g., the ship for the cruise was below the error rate in sentences with a homophonous plural modifier (e.g., the ship for the crews) and instead comparable to sentences with a singular modifier (the ship for the crew).

Bock and Eberhard (1993) report the same observation for nouns which are not homophonous but phonologically very similar to plural nouns (e.g., course resem-bling courts). The absence of attraction with pseudo-plural distractors indicates that a superficially plural-like appearance is not sufficient to cause interference during the computation of agreement. This in turn suggests that plural morphol-ogy on the verb is not triggered by plural-like ending on the subject (either a real plural allomorph or just a phonologically similar word offset) but is rather trig-gered by the processing of agreement relying on abstract feature specifications.

It would be interesting, however, to know whether agreement with pseudo plurals such as cruise is in general as stable as agreement with nouns that do not resemble plural nouns. Furthermore, it might be worth to examine nouns like news where the plural-like ending can be separated. For the moment, I note that attraction seems to be independent of the morphophonological shape of the distractor and rather sensitive to abstract grammatical feature specifications.

Further evidence for this conclusion comes from the observation that attrac-tion is hardly affected by the regularity of plural marking on the distractor noun.

Regular plural forms like rats irregular plural forms like mice elicit comparable attraction rates (see Bock and Eberhard, 1993; Haskell and MacDonald, 2003 for English and Vigliocco et al., 1995 for Italian). For illustration consider the

ex-ample below taken from Bock and Eberhard (1993). Error rates are provided in parentheses.

(36) a. The trap for the rat/rats (0/8) b. The trap for the mouse/mice (0/9)

The finding of almost identical error rates elicited by regular and irregular plural distractors has been replicated in Haskell and MacDonald (2003) who in addition report comparable response latencies for the two sentences types. The picture, however, changes when the controller noun is a collective noun. Collective nouns produce more plural verbs in general especially when modified by a plural NP.

Crucially, regularity of number marking has an effect resulting in a higher pro-portion of plural verbs when the distractor exhibits regular plural morphology compared to plural distractors with irregular morphology (Haskell and MacDon-ald, 2003). But the effect was only marginally significant when differences in the conceptual plurality were taken into account.15 If at all, the regularity of number marking of the distractor seems to play only a minor role. This conclusion is sup-ported by the finding that even number ambiguous distractors can cause attraction.

Moreover, Vigliocco et al. (1995) found almost identical error rates for ambiguous and unambiguous distractors.