Диссертация (Experimental study of several core concepts of theoretical morphology (on the material of russian) - regularity, syncretism, markedness), страница 26

Most nouns ending in -r’and -l’ are masculine, many of them have agentive suffixes like -tel’ or -ar’. Nevertheless, feminine nouns are not just singular cases in this group. Forms of nouns endingin other lingual consonants are either predominantly feminine or are evenly distributed between the two genders (but the number of feminine lemmas is still larger). Themajority of nouns ending in -t’, which are especially numerous, have the suffix -ost’.In our pilot experiment 2, we explored whether adult native speakers are sensitiveto these differences.1Cases where gender variation is observed are not taken into account (a corpus-based studyof such cases was conducted by [Savchuk 2011]).68961Slioussar N. A.Table 1.

The distribution of nouns among genders and declensionsin the grammatically disambiguated subcorpus of the RNC2Declension andgenderPercentageof nounsin the RNC1st decl. feminine1st decl. masculineEnding in Nom.Sg andprototypicalityExamples29% nouns end in -a/ja, ‘prototypical F’1% nouns end in -a/ja,‘non-prototypical M’46% nouns end in a consonant,‘prototypical M’18% nouns end in -o/e, ‘prototypical N’5% nouns end in a consonant,‘non-prototypical F’1% nouns2nd decl. masculine2nd decl. neuter3rd decl. feminineirregular andindeclinablezhena ‘wife’djadja ‘uncle’syn ‘son’,gel’ ‘gel’pole ‘field’mel’ ‘shallow’Table 2. Nouns in the grammatically disambiguatedsubcorpus of the RNC and in the GDRLRNC (Nom.Sg forms)Final consonant/b’//p’//v’//f’//m’//d’//t’//z’//s’//n’//r’//l’/MGDRL (lemmas)F34 (24%)3 (2%)313 (1%)00748 (51%)713 (5%)319 (49%)80 (14%)2,354 (45%)2,160 (76%)6,648 (71%)M1101691,44821670713,1843274912,8426772,653F1—1——1017751261771083111920235534143457112342152.

Previous experimental studiesTwo groups of experimental studies are relevant for the present paper: analyzinggender agreement and nouns with more or less morphologically regular inflections.2The counts are taken from [Slioussar and Samoilova 2015]. Substantivized adjectives werenot taken into account.3These three forms are rup’ (a reduced form of the noun rubl’ ‘ruble’).69062Gender, Declension and Stem-final ConsonantsThere are relatively few experimental studies of gender agreement in Russian.

In threeof them ([Akhutina et al. 1999, 2001]; [Romanova & Gor 2017]) adjectives were presented before nouns audially or visually. In congruent conditions, adjectives agreedwith the following nouns, in incongruent ones they did not, and some experiments alsoincluded a baseline condition where bare adjective stems without inflections or adverbswere presented. Several methods were employed, including lexical decision (answering whether the presented stimulus is a real word or a nonce word), grammaticalityjudgment (answering whether the presented fragment is grammatical) and cued-shadowing in which participants must repeat the second presented word (the target noun).However, the question was always the same: would participants answer significantly faster and more accurately in congruent conditions compared to incongruentones, and would there be any differences associated with the gender of the nouns?In experiments with a baseline condition, it was also possible to check whether the difference between congruent and incongruent conditions was primarily due to facilitation in the former, or to inhibition in the latter, or both effects were equally prominent.In brief, [Akhutina et al.

2001] observed significant facilitation and inhibition effectsfor feminine nouns, while for masculine nouns, only inhibition was significant, and forneuter ones, only facilitation was significant4. Results from other studies were similar.The explanations offered in these studies go along the same lines. Masculinegender as the most frequent is assumed to be unmarked, or default, while neuteris considered the most marked.

Thus, masculine is expected by default, and strengthening this expectation by a masculine adjective does not produce a big difference(hence no significant facilitation effects). Neuter is the least expected option, so priming a neuter noun with a neuter adjective has the largest effect compared to the baseline condition (hence facilitation effects for neuter nouns are larger than for femininenouns). Inhibition effects are explained by rechecking, which is especially costly formasculine nouns presented after non-masculine adjectives.None of these three studies looked at 3rd declension feminine nouns, while theexperiments by [Taraban and Kempe 1999] specifically focused on them. Tarabanand Kempe selected masculine and feminine nouns ending in a palatalized consonant(opaque condition) and in non-palatalized consonants or in -a/ja, which are unambiguously masculine or feminine (transparent condition).

They examined the role of suchtransparency for subject–predicate agreement using word-by-word self-paced reading and forced choice tasks. Participants were asked to read sentence beginnings like(1a) or (2a) and then to select one of the two verb forms in the remaining fragmentlike (1b) or (2b). In some conditions, sentence fragments contained adjectives. Participants were adult native speakers and L2 learners.

For native speakers, transparencyand the presence of a gender-marked adjective did not play any role.(1) a. Dažeevenb. isčez/isčezladisappearedM/F4(obyčnaja)ordinaryFizfrommuka/sol’flourF.1D/saltF.3Dmagazinov.storesteper’…nowThis study also involved aphasiac patients, while [Romanova and Gor 2017] compared native speakers to second language learners, but we will not discuss these groups here.69163Slioussar N. A.(2) a.

Nakanunethe-day-beforeb. bolel/bolelahurtM/F(otёkšij)swollenMotfrompalec/lokot’fingerM.2D/elbowM.2Dudara.injurysil’no…strongly[Slioussar and Malko 2016] studied gender agreement attraction. To give an example, an attraction error is present in the English sentence “The key to the cabinetsare rusty”, where the verb agrees not with the head of the subject phrase, but withanother noun, termed attractor.

In production, such errors are more frequent thanagreement errors without attraction. In comprehension, they are missed more oftenand produce smaller delays in reading times and less pronounced ERP responses.Number agreement attraction is widely discussed in the literature, while genderagreement has been analyzed only in a few studies so far. Among other things, it wasnoted that both in production and in comprehension, attraction effects can be observed in the sentences with singular heads and plural dependent nouns (e.g., “Thekey to the cabinets...”), but not in the sentences with plural heads and singular dependent nouns (e.g., “The keys to the cabinet...”).

Almost all proposed explanations appealto feature markedness, although approaches to markedness may be very different,from representational to frequency-based. Looking for similar asymmetries in genderagreement attraction, several studies of Romance languages obtained controversialresults (e.g. [Acuña-Fariña et al., 2014]; [Anton-Mendez et al., 2002]; [Martin et al.,2014]; [Vigliocco & Franck, 1999]).

[Badecker and Kuminiak 2007] found that neuterbehaves as unmarked in a series of production experiments on Slovak, in which neuter is the least frequent gender, but is used in impersonal sentences, like in Russian.[Slioussar and Malko 2016] conducted one production and three comprehensionexperiments.

The results of the former were similar to the Slovak study, while in thelatter, masculine behaved differently from feminine and neuter. Namely, attractionwas observed for all dependent noun genders, but only for neuter and feminine heads.In other words, masculine heads were significantly more resistant to attraction: readers detected agreement errors irrespective of possible attractors’ interference5.This result can be reconciled with the observations made in [Akhutina et al. 1999,2001]; [Romanova & Gor 2017].

However, given that different patterns were observedfor production and comprehension, we cannot explain them by a particular singleproperty of gender features anymore. This reminds us that the notion of markednessusually invoked to explain all asymmetries between features is problematic becausesome studies rely on representational markedness (primarily counting the numberof positive feature values in formal morphological models), the others consider themost frequent value to be the default etc. From the representational point of view,neuter is the unmarked gender in most accounts, while if we rely on frequency, masculine is. Maybe, these approaches should be seen as complementary, because different properties of features appear to be relevant in different experimental tasks.5It is traditionally assumed that the features of the dependent noun are crucial for attraction, but both this study and some other findings suggest that the features of the head mightbe more important.

We will not discuss this problem here.69264Gender, Declension and Stem-final ConsonantsNow let us turn to the problem of inflections that are more or less typical fora particular gender. Many experimental studies used different tasks (for example, determining a noun’s gender or selecting an article) to show that nouns with more typical inflections are associated with faster and more accurate answers. This was demonstrated for Italian, French, Hebrew, Bulgarian etc. (e.g. [Andonova et al., 2004];[Bates et al., 1995]; [Gollan & Frost, 2001]; [Spalek et al., 2008]).However, fewer studies investigated the role of this factor in sentence processing. To give an example, Caffarra et al.

(2015) looked at Italian nouns with more andless typical endings presented in the same sentences. They were preceded by articles, which carried gender information. Nouns from the two groups elicited differentERP responses. [Franck et al. 2008] and [Vigliocco and Zilli 1999] demonstrated forItalian, Spanish, and French that heads with regular inflections are more resistantto agreement attraction. The same is true for number agreement attraction (e.g.[ Bock& Eberhard, 1993]; [Vigliocco et al., 1995]).As for Russian, only [Taraban and Kempe’s 1999] study discussed above addressed this problem. The authors found no differences between prototypical andnon-prototypical subject nouns.

However, the task they used required selecting a correctly agreeing verb form, which is a less immediate measure than simple readingtimes. For this reason, we decided to come back to this factor in the present study.3. Experiment 1This experiment was conducted together with Pavel Shilin, an MA studentat Saint-Petersburg State University.