Диссертация (Experimental study of several core concepts of theoretical morphology (on the material of russian) - regularity, syncretism, markedness), страница 5

In Finnish, as inmost other free word order languages, the vast majority of narrative sentences have a ‘givennew’ constituent order. So appropriate contexts presupposed such an order in targetconstructions, while inappropriate ones presupposed a ‘new-given’ order — i.e., essentiallyviolating the IS requirements of target sentences. Kaiser and Trueswell showed thatnoncanonical constructions in appropriate contexts were processed faster than those ininappropriate contexts, but were still slower than the canonical order.In our study on Russian (Slioussar 2011), we used the same methodology, but decided totake longer constructions and more extensive context sentences (making two out of threeconstituents in the target sentence given).

In the first self-paced reading experiment, wecompared S V IO DO (canonical), DO S V IO and DO IO V S constructions in appropriate vs.inappropriate contexts. Examples of context sentences are given in (7a-b), examples of targetsentences are given in (8a-b). (7a) was used as appropriate context for (8a) and as inappropriatecontext for (8b) and for another target sentence with the DO S V IO order.(7) a.

Na 8 marta Danja Kašin pozdravilMašu Smolinu.on March 8 [D.K.].NOM congratulated [M.S.].ACC(7) b. Na 8 marta Maša Smolina polučila pljuševogo slonaikonfety.on March 8 [M.S.].NOMreceived [plush elephant].ACC and [candies].ACC(8) a. Danja Kašin podaril Maše Smolinoj pljuševogo slona.S V IO DO[D.K.].NOM gave[M.S.].DAT[plush elephant].ACC(8) b. Pljuševogo slonaMaše Smolinoj podaril Danja Kašin. DO IO V S[plush elephant].ACC [M.S.].DATgave[D.K.].NOM2Sekerina (2003) also compared Russian PP S V IO DO PP and PP DO S V IO PP constructions in isolation andin the same context.17The context factor was significant, while the syntax factor was not. The less pronounced contexteffect evidenced in (Kaiser and Trueswell 2004) might be due to the use of shorter target andcontext sentences.

We also demonstrated that the slow-down starts at the first contextuallyinappropriate constituent, which shows that the information about context requirements is takeninto account immediately, but that it develops faster on preverbal subjects and postverbalindirect objects (occupying their canonical symtactic positions) than on preverbal indirectobjects (occupying a noncanonical position, or scrambled).

In the seconf self-paced readingexperiment, these findings were replicated for IO S V DO and IO DO V S orders. S V IO DOorders with a continuation were used to show that there is no additional effect of inappropriatecontext at the end of the sentence. Thus, processing of word forms in a sentence depends bothon their IS status and syntactic position.4.

Creating resources for experimental researchPaper selected for the defense: (Alexeeva, Slioussar, and Chernova 2017).Experimental studies identified a large list of word properties that play a role for speechproduction and comprehension. They include lemma and form frequency, word length, thenumber of syllables, stress pattern, whether the word has homonyms, homographs ororthographical neighbors, whether it has multiple senses etc. Of course, various grammaticalcharacteristics (part of speech, inflectional paradigm etc.) are also important. For severallanguages, databases with search tools were designed to take these properties into accountduring stimulus selection (e.g.

Balota et al. 2007; Colheart 1981; Davis 2005; Davis, Perea2005; Duchon et al. 2013; Heister et al. 2011; New et al. 2004).As for Russian, some required characteristics were not represented in electronic dictionariesand databases at all, while the others were scattered across frequency lists, grammatical orexplanatory dictionaries or others resources. Therefore, we created StimulStat – a lexicaldatabase for Russian in the form of a web application (http://stimul.cognitivestudies.ru).

Thedatabase contains more than 50,000 most frequent Russian words (1.7 million word forms).These words and forms are characterized according to more than 70 properties, includingfrequency, length, phonological and grammatical properties, orthographic neighbourhoodfrequency and size, homonymy and polysemy. Some properties were retrieved from variousdictionaries and presented collectively in a searchable form for the first time, the others werecomputed specifically for the database. In the paper describing the database (Alexeeva,18Slioussar, and Chernova 2017) we also present some interesting crosslinguistic differences thatcan be identified using such databases.5.

ConclusionsExperiments collected in this dissertation study how word forms are produced and processedin isolation and inside sentences on the material of Russian. Many experiments were the firstor among the first to look at a certain phenomenon (e.g. at the factors of syncretism anddeclension in agreement processing studies, at gender agreement attraction in comprehension)or to apply a particular method (e.g. the PPI method of fMRI data analysis in a linguistic study).The study shows that several notions discussed in theoretical morphology (includingmorphological regularity, markedness, syncretism) are psycholinguistically relevant and thatexperimental data can shed new light on them, demonstrating how theory and experiment caninteract fruitfully.Firstly, the studies discussing how word forms are represented and processed in the mentallexicon debate whether there is a categorical distinction between morphologically regular andirregular forms.

This is a repercussion of a major controversy in cognitive sciences whether ourcognitive architecture uses symbolic rules or not. In our fMRI experiments, we could tease apartprocessing difficulty and morphological regularity effects and went on to ask howmorphological regularity should be defined: whether it is related to type frequency, productivityor defaultness of the inflectional class.Secondly, we looked at agreement — one of the basic grammatical operations that governsusing word forms in a sentence. Many psycholinguistic works aim to find out how subject–predicate agreement is produced and processed relying on so-called attraction errors in numberand gender agreement.

Most studies are dedicated to number agreement attraction, while fourour experiments focused on gender, showing that this feature is interestingly different fromnumber. Number agreement attraction was examined in three other experiments, which testedthe role of different types of syncretism for this phenomenon. Both groups of experimentsprovide arguments for one of the alternative approaches to agreement attraction, the retrievalapproach, and call for certain modifications in it. We also examined how the prototypicality ofinflections (i.e. whether a particular inflection is more or less typical for a particulargrammatical feature, e.g. for a particular gender) affects agreement processing. This factor was19repeatedly demonstrated to play a role for processing of isolated forms, but only a couple ofrecent studies showed that it might also be important on the sentence level.Finally, we studied how different word forms are processed depending on the word orderin the sentence.

This question was addressed in many experimental studies, but the majority ofthem looked at sentences in isolation thus ignoring the context factor — in most languages thatallow for word order permutations they are used to convey different information structures andthus are felicitous only in particular contexts.

Our experiments took this factor into account anddemonstrated how it affects online processing.ReferencesAkhutina, T., Kurgansky, A., Kurganskaya, M., Polinsky, M., Polonskaya, N., Larina, O., Bates, E., Appelbaum,M. Processing of grammatical gender in normal and aphasic speakers of Russian // Cortex. Vol. 37. 2001. P. 295–326.Akhutina, T., Kurgansky, A., Polinsky, M., Bates, E. Processing of grammatical gender in a three-gender system:Experimental evidence from Russian // Journal of Psycholinguistic Research.

Vol. 28. 1999. P. 695–713.Alexeeva, S., Slioussar, N., Chernova, D. StimulStat: a lexical database for Russian // Behavior Research Methods,online first. DOI 10.3758/s13428-017-0994-3.Andonova, E., D’Amico, S., Devescovi, A., Bates, E. Gender and lexical access in Bulgarian // Perception andPsychophysics. Vol. 66. 2004. P.