slioussar resume EN (Experimental study of several core concepts of theoretical morphology (on the material of russian) - regularity, syncretism, markedness), страница 2

2006; Desai et al. 2006; Indefrey et al. 1997; Jaeger et al. 1996; Joanisseand Seidenberg 2005; Marslen-Wilson and Tyler 1998; Münte et al. 1999; Newman et al. 2007;Oh et al. 2011; Sach et al. 2004; Sahin et al. 2006; Ullman et al. 1997). Inflectional morphology5in morphologically richer languages like Finnish, Polish and Arabic was examined in a numberof neuroimaging studies, but these studies did not focus on the regular vs. irregular distinction.This is why we decided to address the problem of regularity in an fMRI study of Russian, alanguage with rich and diverse morphology.The Russian verb system is very complex, and there are several approaches to dividingverbs into inflectional classes. These classes differ in type frequency, and four or five of them(depending on the chosen classification) are productive.

The most frequent productive AJ classwas observed to behave as the default class in several previous behavioral studies (Gor andChernigovskaya 2001, 2003). Thus, there is no obvious division into regular and irregular verbs,and, if any distinctions between different classes are found, we may be able to explore the natureof regularity and to find out which properties (type frequency, productivity or defaultness) arerelevant for production and processing.In our first experiment, we decided to look at the two poles of the verb class system,comparing verbs from the most frequent and productive AJ class to verbs from smallunproductive classes (we reasoned that if any differences between these two groups were found,we could compare them to other verbs in subsequent studies). Participants were asked togenerate present tense forms from different visually presented real and nonce verbs and topluralize real and nonce nouns.

First we performed a subtractive analysis of the data (Slioussaret al. 2014), then a ROI – whole brain voxel-wise analysis of context dependent changes infunctional connectivity (PPI analysis) (Kireev et al. 2015).We demonstrated that morphological regularity and processing difficulty effects can beteased apart. Activity of the left inferior frontal gyrus (LIFG) was greater for production ofirregular verbs (compared to regular ones) and real verbs (compared to nonce ones), so thispattern was explained by processing difficulty (Slioussar et al. 2014). But the functionalconnectivity of the LIFG with temporal lobe was relatively increased when regular verbs wereproduced (compared to irregular ones).

Nothing similar was found for the real/noncedistinction, so this was concluded to be a genuine regularity effect (Kireev et al. 2015).Let us discuss the PPI results in more detail. Firstly, we found that functional connectivitybetween the LIFG and bilaterally distributed clusters in the superior temporal gyri wassignificantly greater in regular real verb trials than in irregular ones.

No other comparisons gavesignificant results. Secondly, we observed a significant positive covariance between the numberof mistakes in irregular real verb trials and the increase in functional connectivity between LIFG6and the right anterior cingulate cortex in these trails as compared to regular ones. Thus,regularity and processing difficulty effects could be dissociated not only using differentmethods of analysis, but also by the PPI analysis alone.We found only one previous published PPI study of inflectional morphology (Stamatakiset al. 2005).

In this study, functional connectivity between functionally predefined ROIs wasassessed during the same/different judgment task. Stimuli were orally presented pairs ofEnglish words and nonce words (e.g. jumped – jump, thought – think, jade – jay). Thus, themethod and materials were very different from ours. Our first finding was similar to whatStamatakis et al. reported, which shows that the observed regularity effect is very robust, beingvalid crosslinguistically both for production and comprehension.As for the second finding, Stamatakis et al. have similar results going in the oppositedirection. This is also true for the subtractive analysis of their data reported in (Tyler et al.2005).

We hypothesized that the processing difficulty went in the opposite directions in the twostudies. Tyler et al. and Stamatakis et al. looked at stimulus pairs like stayed – stay vs. taught– teach. In regular pairs, the first stimulus was morphologically complex and the second wasnot, while in irregular pairs, both stimuli were morphologically simple. Thus, regular verb trialsinduced more processing load.

In our study, all verb forms participants read or produced weremorphologically complex. But irregular verbs involved various alternations in the stems etc.,so irregular verb trials induced more processing load.The two verb groups used in the first experiment differed by many properties (typefrequency, productivity, defaultness), so it was impossible to determine which one wasresponsible for the observed regularity effect. Therefore we conducted a second fMRI studyadding a third verb class — the I class (very frequent, but less frequent than the AJ class,productive, but not default). The results have already been presented at several conferences, buthave not been published yet.In this experiment, we also wanted to find out whether the observed effect would bereplicated in comprehension.

However, we were worried that in case of isolated word forms,processing of inflectional morphology would be shallow because it cannot be meaningfullyinterpreted or used for syntactic parsing. The only published fMRI study where regular vs.irregular verbs were compared in comprehension was discussed above (Stamatakis et al. 2005;Tyler et al. 2005), and it is notable that an ingenious design was used to draw participants’attention to the morphological features of stimuli.7This design could not be replicated in Russian, so we came up with the following one. Inevery trial, participants first saw a real or nonce verb in the infinitive form and a pronoun ja ‘I’or on ‘he’ below it (600 ms).

After an interval, two present tense forms of the previously shownverb appeared on the left and on the right of the screen (1500 ms). One of them agreed with thepronoun, the other did not. Participants were asked to select the correct form. We preferred thisdesign to showing only one form (agreeing or not) because this task would involve agreementviolations and would focus participants’ attention on error detection.We analyzed BOLD signal changes associated both with the 1st stimulus (an infinitive anda pronoun) and the 2nd stimulus (two present tense forms). Based on these analyses we selectedROIs for the PPI analysis. All non-trivial results were associated with the 2nd stimulus. Weshowed that the effects of interest are the same in production and in comprehension.

Thesubtractive analysis showed that the activity of the LIFG gradually increased from the AJ classto I class and then to irregular verbs. The effect was analogous to the processing difficulty effectfrom (Slioussar et al. 2014).The PPI analysis revealed a connectivity pattern that was very similar to the ones reportedin (Kireev et al. 2015) and (Stamatakis et al. 2005). The fact that it was found in the twolanguages with relatively poor and relatively rich inflectional morphology and in the studiesusing three different tasks proves that this effect is reliable. Moreover, we found out whether itcan be associated with type frequency, productivity or defaultness.

The latter was true: anincrease in functional connectivity of the LIFG was observed for the AJ class (as opposed tothe I class and irregular verbs). Notably, this can be explained only in the dual route approachto inflectional morphology postulating a categorical distinction between the default class andthe other classes. In the other approaches, regularity effects, if present at all, are expected tocorrelate with type frequency and productivity.3.

Production and processing of word forms in a sentence3.1. Number agreement attractionPaper selected for the defense: (Slioussar 2018a).Three papers presented in this section are dedicated to production and processing of predicativeagreement. For this domain, the phenomenon of agreement attraction plays an important role.An example of an attraction error is given (1a): the verb agrees not with the head of the subjectNP but with a dependent NP (an attractor).8(1) a. *The key to the cabinets were rusty.b. *The key to the cabinet were rusty.Across languages, such errors in number agreement have been shown to arise more frequentlythan errors of the type exhibited in (1b), where no attraction is possible (e.g. Bock and Miller1991; Eberhard et al.

2005; Franck et al. 2002, 2006; Hartsuiker et al. 2003; Solomon andPearlmutter 2004; Staub 2009, 2010; Vigliocco et al. 1995, 1996). In comprehensionexperiments, attraction errors have been demonstrated to trigger more grammaticality judgmentmistakes and to provoke less pronounced effects in reading time and EEG studies than otheragreement errors (e.g. Clifton et al. 1999; Dillon et al. 2013; Pearlmutter et al. 1999; Tanner etal. 2014; Wagers et al. 2009).Two major approaches to agreement attraction can be identified in the literature:representational and retrieval approaches. According to the representational approach (e.g.Brehm and Bock 2013; Eberhard et al.

2005; Franck et al. 2002; Nicol et al. 1997; Staub 2010),agreement attraction takes place because the mental representation of the subject NP’s numberfeature is faulty or ambiguous. Some authors assume that the number feature can “percolate”from the embedded NP to the subject NP. Others, relying primarily on the Marking andMorphing model (Eberhard et al. 2005), argue that the number value of the subject NP is acontinuum. The more plural the subject NP, the higher the possibility of choosing a plural verb.This plurality depends on properties of the subject NP as a whole and of its head, such ascollectivity, distributivity etc. The retrieval approach (e.g.