Диссертация (Experimental study of several core concepts of theoretical morphology (on the material of russian) - regularity, syncretism, markedness), страница 4

Accusative plural forms coinciding with the nominative plural were shown to bethe most effective attractors.These results have implications for different models of attraction and for other discussionsin morphology concerning ambiguity processing, different approaches to syncretism and theproblem of lexical insertion. In particular, they show that syncretism does not simply boost theeffect of the plural feature — it is an independent factor. The paper argues that the results cannotbe explained as local coherence phenomena and suggests an explanation in terms of the cuebased retrieval approach to agreement attraction, with two notable modifications. First, thestudy shows that retrieval cues can tap into alternative feature sets of syncretic forms.

Notably,unlike in the previously discussed cases, this is possible although no parallel syntactic parsesare considered and no reanalysis is attempted. Second, it demonstrates that retrieval should relyon compound cues such as {nominative+plural} instead of a set of individual cues.3.2. Gender agreement attractionPaper selected for the defense: (Slioussar, Malko 2016).13As we noted in the previous section, in all studied languages, attraction effects were found tobe asymmetric.

They can be observed when the head is singular, and the attractor is plural, butare much weaker or virtually non-existent in the opposite configuration. In the majority ofagreement attraction studies, this asymmetry is explained in terms of feature markedness. Pluralis assumed to be the marked value of number feature, and the asymmetry is attributed to the tothe fact that attractors with a marked feature are more disruptive.However, many approaches to markedness have been proposed in the literature: the markedvalue of a certain feature can be determined based on frequency, presence of a non-zero affix,default use of a form (e.g., in impersonal sentences), various semantic tests etc.

An overviewcan be found in (Haspelmath 2006). It is impossible to evaluate which approach ispsycholinguistically relevant looking only at singular vs. plural — other features should bestudied. We conducted one production and three self-paced reading experiments on genderagreement attraction in Russian (Slioussar, Malko 2016). There were very few productionexperiments (Badecker and Kuminiak 2007; Franck et al.

2008; Martin et al. 2014; Vigliocco,Franck, 1999) and no previous comprehension studies looking at gender.We compared sentences with subject NPs with different gender combinations, as in (5a-b).Badecker and Kuminiak’s (2007) study on Slovak was especially relevant for us. Both Slovakand Russian have three genders. Masculine is the most frequent, while neuter is used asgrammatical default (in impersonal sentences). Badecker and Kuminiak found that inproduction, feminine behaves as more marked than masculine and neuter, and masculine asmore marked than neuter. Our production experiments observed the same pattern in Russian.(5) a.

oknov pole / vo dvorwindowN to fieldN to yardMb. vyxod v les / v poleexitM to forestM to fieldNNumber agreement attraction studies found parallel patterns in production and comprehension,but our comprehension experiments revealed a different pattern: masculine, the most frequentgender, triggered weakest effects. Thus, our study shows that different approaches tomarkedness are psycholinguistically relevant. There was another important finding: in thecomprehension experiments attraction was observed for all dependent noun genders, but onlyfor a subset of head noun genders. This goes against the traditional assumption that the featuresof the dependent noun are crucial for attraction, showing the features of the head are more14important. We demonstrated that this approach could be extended to previous findings onattraction and that there existed other evidence for it.3.3.

Prototypicality of inflections in agreement processingPaper selected for the defense: (Slioussar 2018b).The experiment presented in this section (Slioussar 2018b) is also dedicated to subject-predicategender agreement processing, but studies it without attraction, focusing on two factors: thegender of the noun and its inflectional class, or declension. Russian nouns are inflected for caseand number, and, depending on the set of their inflections, are divided into several declensions.We will rely on the most widely accepted system presented in Table 1. As Table 1 shows, mostconsonant-final nominative singular forms (having zero inflection) are masculine, and mostfeminine nominative singular forms end in -a/ja, with 3rd declension consonant-final femininenouns (3D-F) being less usual, or non-typical.Declension andgenderPercentage of nounsin the RNC1st1 decl.

F29% nouns1st decl. M1% nouns2nd decl. M46% nouns2nd decl. N18% nouns3rd decl. F5% nounsirregular+indeclinable 1% nounsEnding in Nom.Sgand typicalityend in -a/ja, ‘typical F’end in -a/ja, ‘non-typical M’end in a C, ‘typical M’end in -o/e, ‘typical N’end in a C, ‘non-typical F’Exampleszhena ‘wife’djadja ‘uncle’syn ‘son’, gel' ‘gel’pole ‘field’mel' ‘shallow’Table 1. Declension and gender in Russian nouns (excluding substantivized adjectives).Many experimental studies found differences between nouns with more and less typicalinflections in a variety of languages (e.g. Andonova et al.

2004; Bates et al. 1995; Gollan andFrost 2001; Spalek et al. 2008). However, these studies usually looked at the processing ofisolated nouns. Among the few sentence-processing studies, Caffarra et al. (2015) looked atItalian nouns with more and less typical endings presented in the same sentences. They werepreceded by articles, which carried gender information. Nouns from the two groups eliciteddifferent ERP responses.

Franck et al. (2008) and Vigliocco and Zilli (1999) demonstrated forItalian, Spanish, and French that heads with regular inflections are more resistant to genderagreement attraction.1Slioussar and Samoilova (2015) identified the frequency of nouns with different grammatical characteristics inthe grammatically disambiguated subcorpus of the Russian National Corpus (http://www.ruscorpora.ru).15Only one paper (Taraban and Kempe, 1999) addressed this problem in Russian, and nodifferences between 2nd declension masculine and 3rd declension feminine nouns were foundfor native speakers. The task involved selecting a correct verb form for different subject nouns.We decided to come back to this question using the self-paced reading method, which couldallow us tracing earlier processing stages.

Materials were sentence sets in six experimentalconditions like in (6).(6)Xalat / kurtka / šinel’ + byl / byla potrepannym / potrepannoj + ot mnogoletnej noski.robe2D-M / jacket1D-F / overcoat3D-F + wasM/F shabbyM/F + from years-long wearFirstly, the influence of the two factors of interest could be detected only in the sentences withagreement errors, i.e. no gender or declension is intrinsically more difficult to process (at least,in the sentence context).

Secondly, declension played a role at a very early stage and its effectwas very short-lived, while the role of gender became visible later and its effect was morepronounced. Namely, agreement errors were noticed significantly later with non-typical 3D-Fnouns than with typical 1D-F or 2D-M ones. But error-related delay on subsequent words wassignificantly more pronounced for M subjects than for F ones (both 1D and 3D).The latter result suggests that predictions we make about predicate gender are stronger forM subjects.

This finding agrees with Slioussar and Malko’s (2016) study presented in theprevious section, as well as with experimental studies of adjectival agreement in Russian(Akhutina et al. 1999, 2001; Romanova and Gor 2017). The former result can be explained bythe fact that 3D-F nouns essentially look like masculine. But an alternative explanation is alsopossible: all agreement errors (in M or in N) may be harder to detect after 3D-F nouns, i.e.

theirgender can be in general harder to retrieve. A follow-up experiment is conducted to tease apartthese two scenarios. Another current experiment looks at non-typical masculine nouns likepapa ‘dad’.3.4. Processing different word ordersPaper selected for the defense: (Slioussar 2011).The paper presented in this section (Slioussar 2011) studies how different word forms areprocessed depending on the word order in the sentence and on its information structure (IS),i.e. what information is given or new.

Russian has flexible constituent order. Available ordersare used to encode IS, which means that they have different context requirements: for example,an OVS sentence is felicitous when the object is given and the subject is new, but not in zerocontext.16However, most experimental studies of flexible constituent order in different languageshave focused on its syntactic properties, ignoring the context factor.

Canonical andnoncanonical orders were compared in zero context, and the latter were found to be moredifficult to process (e.g., Bader and Meng, 1999; Erdocia, Laka, Mestres-Misse, and RodriguezFornells, 2009; Frazier and Flores d’Arcais, 1989; Hyönä and Hujanen, 1997; Miyamoto andTakahashi, 2002, 2004; Stojanović, 1999; Vasishth, 2002). This was explained by the increasedsyntactic complexity of noncanonical orders.

But an alternative explanation is also possible:noncanonical orders are infelicitous in zero context, which violates their IS requirements.Context-sensitive studies of scrambling are very rare (Bornkessel and Schlesewsky 2006;Bornkessel et al. 2003; Kaiser and Trueswell 2004).2 Kaiser and Trueswell (2004) comparedFinnish SVO and OVS structures in appropriate and inappropriate contexts.