Our book on Russian islands is now published
Russian Islands in the Light of Experimental Data
For reference: D. Belova, J. Demina, A. Gerasimova, E. Lyutikova, E. Morgunova, D. Petelin, K. Studenikina, A. Voznesenskaya. Russkije ostrova v svete eksperimental’nyh dannyh [Russian islands in the light of experimental data]. E. Lyutikova, A. Gerasimova (eds.). Moscow, Buki Vedi. 2021.
The volume “Russian islands in the light of experimental data” presents the first step in the full-scale study of island constraints in Russian. From the descriptive point of view we recognize the dearth of inquiry in the domain: theoretically oriented papers contain contradictory data (e.g. on extraction out of chto- and kak-clauses, on complementizer-trace effect being operational or on superiority effects in multiple wh-fronting in Russian) and several important issues (e.g. parameters of weak islands, intrusive pronouns in island and non-island structures) are not covered at all in the literature. The volume aims at creating an integrated description of island effects in Russian and estimating the degree of intralingual variation, comparing Russian “island profile” to the profiles of more studied languages. The theoretical modeling of the island constraints is performed using experimental methodology, which allows to collect data about acceptability of syntactic movement within different combinations of relevant grammatical parameters.
To sum up, the volume contributes to understanding the phenomena of island effects, summarizing relevant grammatical parameters of islands in Russian, defining whether the island constraints result from grammatical restrictions or processing overload.
Contact us if you are interested in purchasing a hard copy!
Ekaterina Lyutikova, Anastasia Gerasimova
This chapter presents an introduction to the issue of island constraints in Russian language. The chapter aims at providing readers with a definition of island structures and an overview of basic approaches to explaining unacceptability that arises when island constraints are violated. There are two possible sources of unacceptability: various types of island constraints can be caused either by grammatical rules or by an excessive cognitive load during the utterance processing. This opposition, known as the dichotomy between grammatical and reductionist approaches to islands, is explored using experimental methods that are supposed to confirm or refute the predictions of reductionism.
The chapter addresses the problem of studying islands in Russian. From the descriptive point of view, the problem of island constraints in Russian has clearly not been studied enough: conflicting data was used in theoretical research, and many issues did not receive any coverage. The chapter presents a research project that aims at studying island restrictions in Russian using experimental methods. Within the framework of this project this monograph was prepared. The chapter defines the range of research questions and provides a brief overview of nine chapters devoted to the study of various island configurations in the Russian language.
This chapter is an overview of the island constraints that exist in Russian. Islands are constructions which disallow extraction out of them. It is shown in this chapter that the inventory of island constructions in Russian is much alike to that in other languages. For instance, in Russian we find examples of various strong islands — structures that ban the movement out of their internal structure regardless of the type of the constituent being extracted. These include Coordinated Structure Constraint, Complex NP Constraint, Subject and Adjunct Islands, Left Branch Extraction and Definiteness Island. Russian also exhibits some weak island constraints, i.e. islands that only restrict extraction of certain types of arguments. Among those we examine Negation Island, Embedded Question Island and Factive Island. In addition to that, some typologically interesting phenomena related to island constraints are also reviewed. For instance, we look into the difference between the island properties of finite and non-finite subjects. Another asymmetry worth mentioning is the difference between the extraction of wh-words and regular DPs out of weak islands. In conclusion, Russian presents a myriad of data points worth examining for the benefit of studying island constraints in the typological perspective.
The purpose of this chapter is to justify the use of experimental methods in syntax. In particular, the place of experimental syntax in relation to other linguistic disciplines is determined. We consider various experimental methods, compare these methods with each other, highlight their strengths and weaknesses. In addition, the chapter provides an overview of the main methodological issues that arise with the use of linguistic experiments.
We show that experimental methods differ in their “sensitivity”: the production methods more subtly capture possible deviations from the grammatical pattern, and the acceptability judgment methods allow to eliminate the element of randomness and assess the prevalence of a construction in the linguistic community. Besides, it should be noted that experimental methods, especially in combination with corpus ones, can be used for the actual linguistic assessment of the language status. This rationale makes it possible to consider experimental techniques as a powerful tool for studying language in general, intralingual variation and its parameterization.
This chapter discusses island properties of finite embedded clauses in Russian and the results of the experiment aimed at studying acceptability of constituent extraction out of them. The goal of the study was to check two claims previously made in the literature. The first claim is that the movement out of the embedded clause with the complementizer chtoby is more acceptable than the movement out of the clause with the complementizer chto. The second claim is that there is an asymmetry between the extraction of different types of constituents out of the chto-clause. Namely, distant scrambling is said to be more acceptable than wh-movement. An acceptability study using Likert scale was conducted. The experiment employed factorial design; clause type and move- ment type were used as factors.
The results indicate that there is a significant difference between chto— and chtoby-clauses in regard to their transparency for extraction. This, in turn, supports the claim about the two types of embedded clauses having different syntactic structures. The experiment also showed that scrambling out of finite embedded clauses is in fact rated lower than wh-movement, which contradicts the data in other papers. Finally, the study shows that the finite embedded clause in Russian is more island-like compared to the similar structures in other languages. This outcome is important both for the syntactic typology and the future research in Russian syntax.
This chapter examines the islandhood of yes/no indirect li-questions in Russian. Indirect questions are supposed to be weak islands. For an indirect li-question there are two possible configurations with different syntactic structures: (a) li attaches to a verb, V+li move to C, and Spec, CP is empty; (b) li attaches to some XP, which in turn is in Spec, CP. According the theory of Relativized Minimality (RM), configuration (a) should allow movement out of the indirect question, and configuration (b) should ban such movement. Besides, it is assumed that indirect questions exhibit different properties for interrogative and relative pronouns movement, that is, the type of movement affects its acceptability. In addition, it is known that in weak islands the type of constituent affects the possibility of movement: subject cannot move out, while direct object can.
The study tests the following hypotheses:
(i) the choice of the constituent preceding li (XP li or V li) affects the possibility of movement out of indirect li-questions;
(ii) the type of A′-movement (relative/wh/scrambling) affects the acceptability of movement out of indirect li-questions;
(iii) the type of moving constituent influences acceptability of movement (subject vs. direct object).
To answer these questions, we conducted two experiments, in which respondents evaluated acceptability of sentences using Likert scale. The experimental results confirmed the hypotheses. This result supports the predictions of RM: the absence of another A′-element in the Spec, CP of the indirect question enables the movement. However, the differences between different types of A′-movements are still to be accounted for.
This chapter investigates an argument-adjunct asymmetry in Russian wh— questions. Such asymmetry is described for many languages. In current relevant literature there are different approaches to wh-movement in Russian, but some questions are still unanswered, for example: is there a general argument-adjunct asymmetry in different types of questions?
In the present study, we conduct three acceptability judgment experiments in an attempt to clarify the empirical picture regarding argument-adjunct asymmetry. According to the results, the island constraints hold in Russian for chto-clauses and indirect questions. Chtoby-clauses do not demonstrate island effects. However, we also observe that there exists a slight preference for adjunct extraction in all types of indirect questions.
This chapter describes the research of the transparency of subjects in Russian depending on two factors: a structural type of the verb (unaccusative, unergative, transitive) and position of the DP before or after the verb. The former factor is interesting because it can show the influence of the base position of a constituent as an internal or external argument on its opacity. The latter one is important to investigate since the information structure affects the transparency of subjects according to studies of different languages: the topic position is often opaque to subextraction.
A similar study for the Russian language based on long-distance subextraction was conducted in [Polinsky et al. 2013b]. At the same time, the contrast of long-distance and short-distance movements is known to be relevant for subextraction in some languages. Therefore, the study of the subject-object asymmetry in root clauses is relevant.
Our study consists of two experiments. The first — pilot — one aimed to investigate the acceptability of non-basic word orders in Russian (VS for intransitive verbs, OVS/VOS/VSO for transitive verbs where V stands for Verb, S stands for Subject and O stands for Object). During an experiment in which a respondent sees only stimuli without context, sentences with a focal subject may be perceived as less acceptable. The second experiment, or the main one, was intended to test the acceptability of subextracting elements from subject and object noun phrases (by wh-movement). Lexicalizations in both experiments were the same. The respondents rated the stimuli using Likert scale from 1 to 7.
The results of the pilot experiment showed that all non-basic word orders without context are rated significantly lower than the basic SV(O), even the second most common OVS.
The main experiment shows that (i) all arguments, including the object, are more transparent in the pre-verbal position, (ii) in the pre-verbal position, subjects of all three types are equally transparent and the object is more opaque, (iii) in the post-verbal position both subjects of intransitive verbs are equally transparent, both arguments of the transitive verb are the least transparent. Firstly, these results differ from the ones of [Polinsky et al. 2013b], therefore, the length of wh-movement from a dependent or a matrix clause affects the subject-object asymmetry. Secondly, the results are not predicted by any of the three traditional theoretical approaches to the subject island.
This chapter examines the interaction of superiority effects and d-linking in Russian multiple wh-questions. Multiple questions in Slavic languages differ in the way whether they have the restrictions on the order of wh-words (i.e. superiority effects) or not. Factors which make multiple wh-questions immune to superiority include discourse linking (d-linking) of wh-words. If a wh-word is d-linked, an answer is chosen from the set established in discourse, and superiority effects may not show up.
Previous studies of Russian multiple wh-questions contain contradictions both regarding the theoretical approach and the empirical data. They demonstrate different opinions about superiority in Russian and its interaction with d-linking. In view of the diversity of opinions on this issue, we have conducted acceptability judgment experiment (Likert scale, 1 to 7) on the interaction of superiority effects and d-linking. The following factors have been tested: word order (direct/reverse), d-linking of the wh-subject, d-linking of the wh-object.
The results show that the difference between scores for direct and reverse word order is statistically significant for bare wh-words as well as for bare subject and d-linked object. Acceptability of reverse word order is possible in two cases: firstly, for d-linked subject and bare object and secondly, for d-linked subject and object.
The experimental study demonstrates that superiority effects are present in Russian multiple wh-questions with bare wh-words. The influence of d- linking on superiority effects is the following: the effects show up when the wh-subject is bare and may not show up when the wh-subject is d-linked.
The chapter compares the syntax of Russian multiple wh-questions in matrix and embedded clauses. When analyzing multiple wh-questions, one should consider not only matrix questions, but also questions with a long-distant wh-movement, correlatives, questions with a topicalized constituent and indirect questions. Certain studies of multiple wh-movement postulate the absence of superiority effects in matrix questions and their presence in the other configurations, whereas other approaches assume that superiority effects do not show up in any configurations in Russian. As for the interaction of superiority and d-linking, it was not examined for any configurations besides matrix questions in previous studies.
We aimed to compare the structure of Russian multiple wh-questions in matrix and embedded clauses by conducting acceptability judgment experiment (Likert scale, 1–7). The experimental design included three factors: word order, d-linking of wh-subject, d-linking of wh-object. The results for embedded clauses were the same as for matrix clauses: only direct order is acceptable for bare (not d-linked) wh-words as well as for bare subject and d-linked object; both direct and reverse word orders are allowed for d-linked subject and bare object as well as for d-linked subject and object. Thus, the superiority effects show up for bare subject and may disappear for d-linked one.
The experimental study of multiple wh-questions in matrix clauses also demonstrated the presence of superiority effects in Russian; thus, the above mentioned approaches cannot be applied for the analysis of Russian multiple wh-questions. Our goal was to examine whether the strength of superiority effects is the same in matrix questions and in embedded clauses. Since the statistical analysis revealed no statistically significant differences between the scores for each condition in matrix and in embedded clauses, the Russian multiple wh-questions in both configurations have the same properties in relation to superiority and d-linking.
This chapter deals with the syntax of superiority effects in Russian multiple wh-questions ignoring subject-object asymmetry. The previous experimental study of multiple constructions has shown that superiority effects are observed in Russian multiple wh-questions with bare (not d-linked) wh-words and d-linking affects word order in nontrivial ways. The mentioned study investigated multiple wh-questions with wh-subject and wh-object. Although the subject-object contrast shows the presence of superiority effects, it does not allow to determine the nature of the effect, whether the order of wh-words is determined by the asymmetry between the subject and other grammatical relations, or whether the determining factor is c-command.
In order to determine the nature of the superiority in Russian, we conducted an experimental study that allows to consider syntactic phenomena in multiple wh-questions ignoring the subject-object asymmetry between wh-words and regarding only c-command. The stimuli were sentences with an embedded infinitival clause and two direct wh-objects: the matrix clause object and the embedded clause object. Since the main and embedded clause objects attest the same grammatical relation, this configuration allows to find out how the c-command itself affects superiority. According to the experimental data, superiority effects were not found in multiple questions with bare wh-objects, in contrast to questions with bare wh-subject and wh-object. In other configurations, there also have been differences between the pairs of two wh-objects and wh-subject+wh-object. Thus, this experiment shows that the acceptability of various orders of multiple wh-elements depends on several factors: grammatical relation (subject vs. object), animacy, length of the wh-phrase, c-command, d-linking.
This chapter examines the acceptability of intrusive pronouns in Russian. Intrusive pronouns are pronouns that spell out the trace when a constituent moves. There is evidence that such pronouns can “repair” sentences in which movement was illicit — for example, out of island structures. Such studies exist for English, but this phenomenon has not been studied in Russian. The purpose of this chapter is to determine whether the phenomenon is attested in Russian using, experimental methods. As experimental methods, two Likert scales were chosen — with 5 and 7 divisions. In addition, we used a forced choice method. Such a set of methods was chosen to compare them with each other — in the literature there is contradictory evidence about which of them are capable of capturing the influence of intrusive pronouns on acceptability, and which are not. The type of island structure and the presence of the intrusive pronoun were the factors. We also consider the influence of animacy of the constituent being moved. The results of the experiments show that intrusive pronouns not only do not increase the acceptability, but even decrease it, which makes Russian different from English. In addition, the distinction between animate and inanimate elements allows us to determine the islandhood of the construction: whether it is a strong island, a weak island, or not an island at all.
This chapter discusses complementizer-trace effect in Russian. Complementizer-trace effect is an effect of ungrammaticality that arises when a subject of an embedded clause is moved out to the matrix clause when the complementizer is present. It is not entirely clear whether Russian exhibits this constraint as well, since the data provided in the literature has been based on elicitation of individual speakers; moreover, different researchers provide different judgments for the sentences in question. The goal of our study was to determine whether complementizer-trace effects are indeed present by means of experimental methods. We looked at two types of embedded clauses: clauses with the complementizer chto and clauses with the complementizer chtoby. We also tested whether a high adverb affects this phenomenon, as suggested by data from other languages. We constructed an acceptability judgment experiment (using a 1–7 Likert scale) with the factorial design.
There were two factors: the type of the embedded clause and the type of the extracted argument. The results indicate that subject extraction is in fact less acceptable than object extraction. Moreover, the type of the embedded clause did not affect the judgments on subject extraction, unlike in case of object extraction. This contradicts the data provided in some earlier papers. We also found that a high adverb did not influence the speakers’ judgments either. In this matter, Russian differs from other languages, such as English; thus, this piece of data should also be considered when modeling complementizer-trace effects. Finally, we studied how the presence of the context that preceded the experimental sentences affected the participants’ judgments. It has turned out that in some cases the context can in fact influence the ratings. We hope that this result will be beneficial for constructing future experiments.
This chapter provides an overview of procedures that were used to process the results of the experiments presented in this monograph. We describe and justify the view developed in the project on experimental results processing, primarily judgment data analysis.
The chapter examines the practical use of acceptability judgments. We analyze in detail the problem of attributing Likert scores to an ordinal or interval measurement scale, discuss the meaning of z-score transformation, validate the use of grammatical and ungrammatical fillers and their comparison to the target conditions.
One of the goals of this chapter is to draw on the experience gained from working on the project and make recommendations that would help those readers who are not familiar with experimental syntax conduct their own experimental research. In this regard, we describe the necessary steps for preliminary data systematization and statistical processing, we compare various types of visualization in accordance with the researcher’s goals. The chapter also contains an overview of the statistical criteria that were used in the project.
It is expected that the researchers’ transparency with regard to decisions that determine the design, procedure and statistical processing of an experiment will help strengthen the idea of reproducible research in the research community, including traditional humanities.
The chapter provides a glossary of terms and concepts of formal syntax, which aims at contributing to a better understanding of theoretical approaches to island constraints in Russian by less prepared readers. Grammatical phenomena associated with movement, as well as theoretical concepts involved in the analyses of island constraints are discussed.