William Croft’s approach to constructions in comparison with the system of constructions in the Russian Constructicon

Толық мәтін

1. Introduction

The monograph by William Croft “Morphosyntax. Constructions of the World’s languages” [2022] describes various language phenomena from a constructivist perspective. The book takes on an ambitious task to discuss very different phenomena, their function, form, and semantics in a typologically-oriented manner.

The monograph has a twofold aim that combines language theory and linguistic typology. The author seeks to review the main domains of grammar, to describe questions, ambiguities, and trouble spots they pose for a researcher and, at the same time, to show the variance of formal organization in the given domain across languages.

A comprehensive understanding of a language is impossible without discovering the language-specific patterns of its syntax, morphology, and semantics. The idea of understanding a language on its own terms dates back to the structuralist tradition (cf. [Boas 1911]). A reasonable question, however, is whether it is possible to create a cross-linguistic classification that would be universally applicable and meaningful. William Croft [2022] has made an attempt to provide such a universal inventory of Comparative Concepts (CCs) that could be employed to describe the morphosyntax of any language and, therefore, would provide grounds for cross-linguistic comparison of constructions in the world’s languages. He takes a Comparative Concept to mean “a concept that can be used to compare the morphosyntactic structure of different languages” [Ibid.: 676].

In this paper, we describe the general principles adopted in the book and its specific features and compare its approach with the practically-oriented framework used in the Russian Constructicon.¹

2. Organization and characteristic features of the book

The book covers both well-described topics that have become the subject of descriptive, theoretical, and typological studies and problems and areas of grammar that are rarely considered to be relevant for this domain. The former include complex clauses and voice alternations, while among the latter there are speech act constructions (their relation to modality is often dealt with in discussions of grammatical semantics but not of morphosyntax), as well as nonpredicational clauses, as in (1) and stative complex predicates, as in (2).

(1) Tiwi (isolate, Australia)

Purukupaɹli maɹntina

Purukuparli boss

‘Purukuparli is boss.’ [Croft 2022: 295]

(2) We painted the door red. [Ibid.: 432].

The language phenomena and semantic domains analyzed in Croft’s book are divided into several groups that correspond to the parts of the monograph.

Part II “Argument phrase structure: reference and modification” describes mechanisms used to refer to an entity and to its specific properties: referential accessibility; specificity; types of modification; classifiers; anaphora. In Part III “Clause structure: predication and arguments” the author focuses on phenomena that are usually regarded as simple clause phenomena: voice and transitivity; argument coding and valency; nonverbal predication; complex predicates. Among topics that are rarely included in the discussion of simple clauses are speech act typology (exclamatives, hortatives, and so on), salience of the participants of the situation, and information packaging. Finally, Part IV “Complex sentences” deals with not only canonical problems related to complex sentences and clause combining (syntactic typology of clauses: relative, complement, and adjunct clauses; the opposition of balanced and deranked clauses), but also topics regarding semantic and discourse relations in complex sentences (reference tracking, zero expression of arguments).

Apart from these phenomena-oriented descriptions, the monograph contains an introductory Part I consisting of two chapters: the first one outlines the structure of the book and the author’s approach to cross-linguistic comparison, and the second one discusses the constructivist orientation of the book.

A special feature of the book is the absence of a clear boundary between main levels / areas of language, such as semantics, syntax, morphology, and discourse. As a consequence, topics generally believed to be mainly syntactic (transitivity, clause combining) are analyzed side by side with semantic (motion events, stative verbs) and discourse phenomena (reference, salience of participants). This approach has its strengths and weaknesses. On the one hand, it allows the author to demonstrate links between the various levels of language, usually identified only for individual language categories (cf., e.g., studies by Hopper and Thompson [1980] and Tsunoda [1985], which present semantic grounds of syntactic transitivity opposition). However, this approach makes the monograph heterogeneous: for instance, the rather narrow issue of stative predicates is given a separate chapter, while another chapter focuses on the complement clause typology, which is a much broader topic.

The reader finds some lacunas in the set of topics. For example, the monograph does not contain a detailed discussion of tense and aspect typology: the grammaticalization of tense, aspect, modality, and polarity is only touched upon in sections about complex sentences and complex predicates. Similarly, categories like nominal case and number are not analyzed in detail. The reason seems to be that the author, due to the constructivist approach, adopts (implicitly and, to some extent, also explicitly) the following hierarchy of linguistic description:

1) the semantic / functional domain that the phenomenon represents;

2) its impact on the syntactic relations / the whole construction;

3) properties of elements of the constructions.

To illustrate, the domain of voice is given much attention because voice meanings affect the whole sentence, including information structure types, discourse relevance of participants, and syntax of the clause / sentence (2) above). By contrast, case is a category marking a dependent element, but it does not affect the features of the head. This is why case is, so to say, “not very morphosyntactic” (that is, not directly related to the syntax of the whole constituent) and is peripheral to the approach adopted in the monograph.

This choice, resulting from the perspective taken by the author, is not free from disadvantages: some phenomena that could be relevant for the syntax of languages are either not analyzed or scattered across different parts of the book. For instance, case has been shown (e.g., by Wierzbicka [1980] and Kittilä [2002]) to be relevant for categories of the whole sentence, such as transitivity — which, in turn, have to do with constructions and information structure — however, in Croft’s monograph, case, argument structure, and information structure are considered in different chapters, and their relations are not explicitly mentioned. This is why, in some cases, it would have been useful (especially for readers specialized in typology of grammatical categories) to emphasize in each chapter relations between close phenomena or phenomena with similar functions, including head- and dependent-related categories.

Some descriptions of specific phenomena also call for questions and comments, for example, Chapter 6 on transitivity and argument structure. The point of view that transitivity should be described as a prototype and a set of peripheral phenomena that do not fully correspond to the prototypical description is adopted by many linguists. However, many current studies also seek to answer another question: why is it the transitive construction (and not, for instance, the one with a dative object) that is so important for our description of the syntactic structure? It seems that the answer to this question requires a plurifunctional approach that takes into account the (i) semantics of events (e.g., breaking / change of state) and of participants (see the treatment of animate / inanimate participants by Hopper and Thompson [1980] and Tsunoda [1985]); (ii) formal properties of the transitive construction; (iii) behavior of the transitive construction, including available voice alternations. In other words, Croft’s chapter on transitivity and argument structure lacks some theoretical depth that is needed to describe the class of transitive verbs and their role in the grammatical theory.

The last point, the properties of constructions and / or their elements, is not always given sufficient attention throughout the book. A case in point is Chapter 10 “Nonprototypical predication and nonpredicational clauses”. Since the type of clauses under analysis lacks an explicit predicate form, an a priori classification of nonpredicational clauses is problematic. At the same time, Roy [2013] and Testelets [2008] show that some aspects of syntactic behavior differentiate predicational, presentational, thetic, and other types of nonpredicational clauses. Croft, however, mainly takes into account semantic and contextual features and suggests a classification mainly based on (im)possibility of a particular interpretation in the given context. Of course, it is impossible to consider the whole range of syntactic tests used for delimitation of types in one study. However, if some examples of syntactic contrasts were given, this would have made the classification of nonpredicational clauses more empirically founded.

A notable feature of the book is the wide use of quantitative data, see the statistics on word order in Chapter 6 “Event structure and argument coding: semantics, transitivity, and alignment” and Table 8.4 on argument expression in Huallaga Quechua in Chapter 8 “Argument coding and voice: discourse factors”. On the one hand, the prominence of statistics gives the language data a new perspective and distinguishes the book from many previous typologically-oriented descriptions. On the other hand, it seems that the statistical data are not always accompanied by an explanation.

Another important feature of Croft’s monograph is that the domains of grammar are defined on different grounds. While some of them are defined semantically (for instance, Chapter 15 “Temporal and causal relations between events” illustrates various means for expression of the same semantic and pragmatic relations, see also Chapter 12 “Speech act constructions” and Chapter 14 “Stative complex predicates, including manner”), others are based on formal grounds (in Chapter 10 “Nonprototypical predication” the absence of a verb or another type of predicate serves as the sole defining factor). This asymmetry is, to a large extent, an integral part of today’s syntactic and semantic typology. It results from the core-periphery structure that current theory assigns to grammar. For example, the core (canonical) sentence has a verb / predicate, and this is why the presence of the predicate goes without a comment; verbal sentences are semantically and syntactically classified into stative and dynamic, transitive and intransitive, and so on. By contrast, the absence of a predicate makes a clause / sentence noncanonical: this formal feature serves as the basis for further description of semantic or pragmatic distinctions, e.g., between ‘possession’, ‘qualification’, ‘identification’. Roughly speaking, when we classify grammatical phenomena in various frameworks (including Croft’s), we tend to classify them by the most visible and most unusual properties, even if these properties belong to different aspects of grammar and levels of language (semantics, syntax, illocutionary force, and so on).

We turn next to the concepts and principles proposed by the author and assess their strengths and weaknesses.

3. Сonstructivist orientation

A key feature of the book, as mentioned above, is its constructivist orientation. This perspective makes the task carried out by the author even more ambitious because constructions are more complicated to compare across languages and to distinguish from each other than, for example, grammatical markers.

It should be noted that Croft regards the construction as a basic concept of language organization and a basic instrument for language description. One of the key assumptions he puts forward in the introductory part is that “the proper unit for grammatical analysis is a (morphosyntactic) construction, such as the numeral modification construction” [Croft 2022: 4–5]. Word classes are also defined on constructivist grounds: the author states that “word classes are defined not by their semantics but by their occurrence in constructions — more precisely, a word’s occurrence in a particular role in a construction” [Ibid.: 8]. This strength of the constructivist approach is that Croft emphasizes the importance of empirical analysis of basic elements of grammar (such as words and word classes) on the basis of contexts they are compatible with, instead of postulating an a priori classification or using vague semantic formulations, such as ‘adjectives denote properties’. At the same time, this broad notion of a construction as a basic element of grammar does not provide results significantly different from notions of ‘verb class’ or ‘syntactic relation’ of standard grammatical typology. For instance, instead of saying that an adjective is a word that occurs as a modifier in a noun-modifier construction and in the predicate position in a copula construction, we can formulate it in a standard syntactic way: “An adjective can be a modifier of a noun or pronoun or the second argument of a copula verb”. In this sense, Croft’s construction is an instrument of description which does not provide results crucially different from those provided by formal or functional syntactic analysis.

Another possible problem of Croft’s type of constructional analysis, is the lack of clear boundaries between different constructions. To illustrate, in Modern Standard Arabic, the numeral wāħidun ‘1’ agrees in case and gender with the noun, the numeral talātun ‘3’ subcategorizes for genitive case and takes the gender opposite to the gender of the noun, and khamsūna ‘50’ does not have gender and requires the accusative case form (see [Grande 2001; Aoun et al. 2010] for a detailed discussion). Does this mean that ‘1’ belongs to one word class and ‘3’ and ‘50’ to another one (because the latter, but not the former, assign case to the noun) or that all three numerals are of different classes?

As mentioned above, Croft [2022] elaborates on the idea of comparative concepts and provides an arguably exhaustive list of them based on the typological analysis of constructions on the level from phrases to clauses and complex sentences. Let us repeat Croft’s definition: a comparative concept is “a concept that can be used to compare the morphosyntactic structure of different languages” [Ibid.: 676]. The taxonomy of Croft’s CCs includes several hundred CCs that are divided into four types: construction (cxn), semantic content (sem), information packaging (inf), and strategy (str). This four-dimensional distinction corresponds to the basic structure of a construction proposed by Croft. The notion of comparative concepts, i.e. theoretical constructs for typological comparisons, was introduced by Haspelmath [2010], who suggested differentiating comparative concepts from language-specific descriptive categories. The reason behind this differentiation is that “crosslinguistic comparison of morphosyntactic patterns cannot be based on formal patterns (because these are too diverse), but has to be based on universal conceptual-semantic concepts” [Ibid.: 665]. This idea is not new and follows the Greenbergian tradition of typological research, cf. [Greenberg 1963; Comrie 1989; Croft 2003].

Information packaging in Croft’s terms is contrasted to strategy. Strategy refers to the set of specific means used to convey a given meaning. By contrast, packaging refers to the role of information in discourse. As Croft [2022: 13] puts it, “any semantic content can be packaged in any information packaging function”. The main packaging functions are reference, modification, and predication. For instance, the property ‘tall’ can be predicated in That tree is tall, used to modify a referent in a tall tree and used as a referent in the tallness of the tree.

According to Croft [2022: 5], each construction is a combination of form and function. The form refers to the morphosyntax of the construction (i.e., strategy), while the function is further subdivided into semantic content (i.e., meaning) and information packaging, suggesting that any meaning can be packaged differently, see Figure below. Thus, four types of CCs correspond to the three end nodes (inf, sem, and str) and the root node (cxn) in the structure.

 

Рис. Основная структура сооружения [Croft 2022: 5]

 

Croft’s broad notion of a construction calls for a comparison with other, more practically oriented approaches to constructions. Recently, linguists’ interest in constructions has acquired a new, applied dimension, namely, the creation of constructicons (databases of constructions that contain their semantic and structural description, as well as annotation of relations between them, both formal and semantic). This is why the monograph under analysis is also important as a potential starting point for creating an inventory of grammatical constructions (as well as grammatical markers) and compiling new constructicon resources.

Constructicons are created for a limited number of languages, among others, for English, German, Japanese, Swedish, Brazilian Portuguese, and Russian. While the majority of constructicons are based on corresponding FrameNet resources (cf. [Fillmore 2008])², the Russian Constructicon is built independently from FrameNet and uses its own set of syntactic and semantic types to describe the constructions. In what follows, we compare Croft’s approach to that adopted in the Russian Constructicon.³

4. Constructions in Croft’s account and in the Russian Constructicon: a comparative account

The construction is a notion that is, on the one hand, very useful in current linguistic studies, and, on the other hand, rather challenging. Almost everyone recognizes the importance of constructions, at least in some of the senses used in linguistics and to some degree. At the same time, the great diversity of approaches and definitions points to the fact that different authors may employ the same notion differently — and, thus, several construction-oriented descriptions may in fact be technically and theoretically different.

A comparison across various approaches to constructions is of particular relevance because Lyngfelt et al. [2022] propose the application of Croft’s CCs to the emerging field of multilingual constructicography, which aims to compare constructions across various languages. However, cross-linguistic comparison of constructions presents a challenge, particularly for typologically distinct languages (cf. an exploration of the alignment of constructions in English and Japanese via interactional frames in [Ohara (forthcoming)]). This difficulty extends even to typologically close languages, as differences in the choice of linguistic material across constructicon resources and variations in the granularity of grammatical constructions pose obstacles to cross-linguistic comparison. CCs are supposed to map constructions in different languages to conceptual domains, but that does not mean finding constructional equivalents [Lyngfelt et al. 2022].

But can we apply the set of CCs developed by Croft to the constructions in the Russian Constructicon? Are the two systems compatible? What is the conceptual overlap of the two classifications and how are they different? We address these questions below.

It is necessary to note that the notion of constructions used by Croft is not identical to that used in the Russian Constructicon. The main difference is that for the Constructicon, a construction is an empirical object, while for Croft’s description of morphosyntax, a construction is a theoretical object. The following formulations illustrate each approach:

Croft: the main instrument for language description are constructions, and many aspects are easier to analyze in the construction-based description. The constructional mechanism is equally used in the description of the core grammatical mechanisms (e.g., complement clauses) and narrow quasi-grammatical phenomena (e.g., copula constructions).

Russian Constructicon: the class of constructions is heterogeneous, some of them are lexicalized (semantically and syntactically close to lexemes and / or peripheral grammatical markers with narrow combinational properties), and others belong to a core grammatical domain.

Of course, many language elements can be described as constructions in both frameworks. For instance, the Russian èto-construction that marks the real participant of the situation as contrasted to other, hypothetical, participants is a construction in both senses:

(3) Èt-o Vasj-a razbi-l okn-o.

this-nom.n.sg Vasja-nom.sg break-pst.m.sg window-acc.sg

‘It was Vasja who broke the window.’

In this case, not only the presence of èto, but also its linear position (it typically appears in the leftmost position), and the intonational emphasis on the argument following èto mark the contrast between Vasja and other possible participants. The construction functions as a complex multiword unit and thus calls for a Construction Grammar analysis. The same seems to be the case with stative complex predicates, discussed in Chapter 14 of [Croft 2022], as well as by [Goldberg 1995; Himmelmann, Schultze-Berndt (eds.) 2005], among many others.

However, many constructions, such as transitive clause constructions; constructions with adjunct embedded clauses; complement clause constructions, and so on, are covered in the monograph but definitely feel to be less lexicalized and closer to the grammatical level in the proper sense. The reason is that Croft regards constructions as a convenient theoretical instrument for describing various types of constituents, mainly because elements of almost any constituent type are interconnected: for instance, though complementation is very often marked by a dedicated syntactic element (complementizer) or morphological form of the embedded verb (infinitive or another embedded form), its description also requires appealing to the main clause properties: Which classes of main predicates take complement clauses? Why do some quasi-synonymous predicates differ in ability to take complement clauses or in the type of complement clauses they take? How are the main and the embedded event in the complement clause construction situated in time?

By contrast, in the Russian Constructicon constructions are regarded as an empirical phenomenon. In this sense, Croft’s broad notion of a construction is too theoretical to be directly applied to a constructicon. The Russian Constructicon team considers constructions a heterogeneous class of elements that include ‘core grammar’ as well as lexicalized constructions (‘quasi-grammar’) and lexicon.

For the ‘quasi-grammatical’ and ‘lexical’ subclass of constructions, the two key features that are often described are the absence of markers and / or non-compositional use of markers and lexicalization. A construction may, for example, (i) not contain an element whose main or only function is to connect other (meaningful) elements and (ii) the link between elements may be stronger than in most language structures of similar type (as is the case with the depictive construction: the link between the depictive and the main clause is stronger than between the clauses in a complex sentence, which leads to non-predictable selective restrictions). To illustrate, the construction to i delo ‘often’ with the meaning of frequency is semantically very close to a grammatical iterative marker. Furthermore, it contains several word forms and is not interpreted compositionally, based on the meaning of the words to ‘that’ and delo ‘business’.

By definition, such a class of lexicalized / quasi-grammatical constructions cannot contain subtypes of “Transitivity expression” or “Voice expression”. In general, voice mechanisms and transitivity features are systemic phenomena that characterize the language system as a whole, are marked in a standard way (e.g., voice markers, direct object case markers), and are far from being lexicalized. At the same time, some transitivization patterns, such as those with the word svoё ‘one’s’ (on svoё otbegal ‘he cannot run anymore’, lit. ‘he his own out-ran’) can be included in construction databases due to the peculiar and non-compositional use of the transitive verb in combination with the reflexive possessive pronoun.

Some of the phenomena discussed by Croft should be assigned a peripheral position in a constructicon, because they show no sign of constructionalization in the narrow sense (they are interpreted compositionally, and their elements are productive grammatical markers or syntactic elements). This means that the conditional construction with if in English or the purpose construction with čtoby in Russian is not a core element of the constructicon (they are elements of ‘big’ syntax, namely, principal types of conditional and purpose subordinate clauses). On the contrary, constructions where grammatical elements are used in a noncanonical way, such as the “order” construction with čtoby (Čtoby ja tebja zdes’ bol’še ne videl ‘I order you to never come here again!’ (lit. ‘that I would never see you here!’)) are core elements of a constructicon: čtoby is here neither a subordinator, nor a marker of the purpose meaning, and the properties of this construction type (such as the absence of the main clause) characterize the construction as a whole, rather than the word čtoby, which serves here as a quasi-grammatical marker of imperative.

Let us now describe the general makeup of constructicons (based mainly on the example of the Russian Constructicon) in more detail and compare it to Croft’s proposal. The Russian Constructicon is an electronic linguistic database that consists of more than 2200 grammatical constructions of Russian (https://constructicon.github.io/russian/). The constructions in the resource are supplied with elaborate semantic, syntactic, and stylistic annotation, definitions translated into English and Norwegian, corpus-based example sentences, common fillers for construction slots, and constructional equivalents in English and Norwegian. Although constructicons are often described as “dictionaries of grammatical constructions” [Lyngfelt et al. 2022: 101], a constructicon is emphatically not a list but a structured network of constructions that are linked to each other [Fillmore 1988: 37; Goldberg, Herbst 2021: 286] or even a “set of networks” [Langacker 2008: 237]. Identifying the links between constructions makes it possible to model horizontal and vertical relations between them and to group constructions into families, clusters, and networks. The elaborate description of the semantic classification of constructions in the Russian Constructicon as well as the family-based expansion of the constructicon is available in [Janda et al. 2023] and [Endresen et al. (forthcoming)].

4.1. Bottom-up approach to classification

The semantic, syntactic, and stylistic classification of constructions in the Russian Constructicon has resulted from a bottom-up approach and emerged together with the expansion of the database. In practice, this means that constructions were added and annotated at the same time. Oftentimes the decisions about the annotation had to be reconsidered several times during the process when more data were added and constructions were analyzed in groups.

The bottom-up approach means collecting a maximally inclusive inventory of constructions and analyzing them on their own terms, thus allowing patterns to emerge from the data. The bottom-up approach is opposed to the top-down approach that presupposes a predefined system in which more granular distinctions are formed later.

This bottom-up approach means that the starting point for a Constructicon is always the data of an individual language. Not only may the items and meanings of a given semantic zone be very different across languages, but also the set of domains and subdomains in different languages can also be incomparable, meaning that its analogues cannot be found in other languages. This makes the “cross-constructicon” comparison problematic, which means that the items collected for each language and each domain have to be further checked for their membership in one or another small domain. In this sense, typological research based on the data of constructicons is “post-hoc”, built after and upon the individual-language descriptions. On the contrary, Croft’s typology and typological studies, e.g., by Dahl [1985], Comrie [1989], Shopen [2007], from which Croft inherited many systemic features is mostly pre-hoc: they are built on broad (but not so finely distinguished and described) semantic zones and syntactic phenomena types that are often assumed rather than emergent from language data.

4.2. Semantic classification

Constructions are defined in Construction Grammar as conventionalized form-meaning (or function) pairings [Goldberg 2006: 5] (see also [Goldberg 1995; 2013]). Semantic types identified in the Russian Constructicon characterize the latter aspect of constructions, i.e. their meanings. As mentioned above, Croft’s types are not always defined on semantic grounds: very often, their classification is based on salient syntactic and / or morphological features. The semantic classification in the Russian Constructicon comprises 55 major semantic types and 182 more specific subtypes. Semantic types of constructions form larger units: subclasses and classes. All major semantic types are illustrated in Table 1 as bullet points. If a semantic type has subtypes, it is illustrated with a plus symbol next to the type. For example, the semantic type Addressee has four subtypes (Core addressee, Audience, Beneficiary, and Maleficiary), while the semantic type Possession has no subtypes. The description of all semantic types and subtypes is available on the Instructions page of the Russian Constructicon website⁴ and in [Janda et al. 2023]. The larger units (classes and subclasses) are illustrated with numbered black and blue categories in the table respectively.

 

Table 1

Semantic classification of constructions in the Russian Constructicon: 55 main semantic types included

1. Qualia
1.1. Situation structure	1.2. Major roles	1.4. Logical relations	1.6. Sets and elements
• Timeline + • Taxis + • Actionality + • Pluractionality + • Phase of action + • Result + • Actuality	• Addressee + • Instrument • Possession • Comitative • Caritive • Non-standard subject +	• Cause • Purpose • Consequence • Condition • Concession +	• Additive • Inclusive • Exceptive • Exclusive • Subset • Options • Quantification +
1.3. Situation modifiers		1.5. Properties	1.7. Magnitude
• Spatial expressions + • Temporal expressions + • Manner		• Salient property + • Temporary characteristics + • Comparison +	• Non-existence + • Measure + • Calculation +
2. Modality and its neighborhood	3. Subjectivity	4. Discourse	5. Parameters
2.1. Core modal meanings	• Assessment + • Attitude + • Polarity value + • Source of opinion + • Mirative	4.1. Discourse organization	• Degree of intensity + • Degree of accuracy +
• Root modality + • Epistemic modality +		• Discourse structure + • Intersubjectivity • Objectivity • Source of information
2.2. Neighborhood		4.2. Discourse clauses
• Volition • Causation + • Prohibition + • Threat • Request • Apprehension + • Curse		• Reaction to the previous discourse + • Routine +

 

Importantly, the principles of classification adopted in the Constructicon and in the monograph under analysis are different. Croft begins the classification from sentence types: for him, a construction is a ‘type of utterance / sentence’. This does not mean that only whole-sentence categories are discussed; however, for example, voice is more likely to figure in Croft’s study than, for instance, case-marking or prepositional expression of the argument. The semantic classification of the Russian Constructicon works as a system of tags, and a construction can be assigned more than one semantic type. This is different from the system adopted in constructicons based on FrameNet classification (i.e., of semantic frames), where semantic multiple motivation of constructions is not allowed. The system of tags adopted in the Russian Constructicon is useful to capture the complex semantics of a given construction and is fruitful for modeling the horizontal relations across constructions. The combination of semantic and syntactic classification reveals a structured network of constructions.

In the Russian Constructicon, the semantic types are identified based on shared semantic properties of constructions. The structure of the semantic type, therefore, resembles a radial category with more prototypical members and a semantic periphery. For instance, both constructions (4) and (5) express the cause of some action. Therefore, they bear the semantic tag Cause. However, while (4) introduces a cause, (5) is a question to the addressee about the cause of his / her actions. Thus, (5) can be considered less prototypical than (4). The construction in (5) additionally conveys the speaker’s dissatisfaction with some action by the interlocutor and is also tagged as Attitude with the subtype Dissatisfaction. This additional semantical component places the construction in (5) even further away from the prototype.

(4) ID 53 vvidu togó, čto Cl, Cl ‘due to the fact that…’

Izdatel’stv-o objazan-o vyplati-t’ avtor-u vs-e

publishing_house-nom.sg obliged-n.sg pay-inf author-dat.sg all-acc

100 % gonorar-a vvidu t-ogo, čto rukopis’

royalties-gen.sg due_to that-gen.sg that manuscript.nom.sg

otklonen-a uže posle odobreni-ja.

rejected-f.sg already after approval-gen.sg

‘The publishing house is obliged to pay the author 100 % of the royalties due to the fact that the manuscript was rejected after approval.’

(5) ID 353 čego (èto) NP-Nom VP? ‘why…’

Č-ego èt-o ty lež-iš’?

what-gen.sg this-nom.n.sg you.nom lie-prs.2sg

‘Why are you lying here?’

In Croft’s classification, (4) and (5) would definitely belong to different classes. While (4) and similar constructions are classified as a type of complex clauses (more precisely, adjunct clause constructions), (5) would be described as a question. Croft’s approach is in fact convenient for cross-linguistic comparison. The most visible and most accessible for formalization are aspects of language related to marking. While the utterance like ‘He left because he felt tired’ very often contains only a marker of cause, a question of the type ‘Why are you lying here?’ is often marked as a question (questions being in most languages more derived and more marked than affirmative utterances).

The radial structure of semantic zones, adopted in the Russian Constructicon, is not convenient for a typological study and is not used in his book. Croft’s semantic labels aim to be more clear-cut and well differentiated from each other for the sake of typological comparison. Furthermore, Croft’s book relies on the convenience of syntactic radial categories, for instance, for description of voice and transitivity.

Contrary to the Russian Constructicon, Croft’s study does not feature a detailed semantic classification. The wider the typological sample, the more the semantic description gets reduced to generalized domains and semantic types.

The semantic classification obtained using a bottom-up approach is, nevertheless, verified by typological studies, reflecting the types of meanings that are encoded grammatically in other languages. In the Russian Constructicon, many semantic types correspond conceptually as well as terminologically with typologically attested grammatical categories. For instance, the construction that indicates the absence of a secondary agent or the absence of an object possessed by the main participant of the situation bears the semantic tag Caritive borrowed from the typological literature, see, e.g., [Plungian 2011: 125]. The semantic approach to constructions in the Russian Constructicon, however, often distinguishes more subtypes and includes more values than the typological approach as it has to deal with complex multiword constructions and not just grammatical affixes. The semantic classification also includes semantic types that are not attested in the typological literature, such as Calculation or Curse.

4.3. Syntactic classification in the Russian Constructicon

Unlike the semantic classification that works as a system of tags assigned to a construction, the syntactic classification of constructions adopted in the Russian Constructicon resembles a system of boxes, into which the constructions are distributed during the classification process. This means that, as a rule, one construction can belong to only one syntactic type. However, the syntactic annotation is four-dimensional to capture the complex structure of multiword constructions.

First, there is a Syntactic type of construction. This annotation category describes construction as a whole. Second, Syntactic function of anchor describes the syntactic function only of the fixed elements of the construction. Third, the anchor itself can be a complex entity, and thus we need to have Syntactic structure of anchor category for annotation. Finally, all anchor elements of construction are annotated according to Part of speech of anchor.

To illustrate, the example (4) can be analysed as shown in Table 2. The construction as a whole is a Connection construction because it contains a combination of words that perform as a conjunction. This combination of words is an anchor part of the construction and it functions as a Subordinator. Since this is a not a single word conjunction, vvidu togó, čto is annotated as a Multiword Conjunction. If we decompose this multiword conjunction, we will see that it consists of a preposition (vvidu), a pronoun (togó), and a conjunction (čto).

 

Table 2

Syntactic annotation of constructions ‘ID 53 vvidu togó, čto Cl, Cl’ and ‘ID 353 čego (èto) NP-Nom VP?’

Cxn	Syntactic type of construction	Syntactic function of anchor	Syntactic structure of anchor	Part of speech of anchor
ID 53	Connection construction	Subordinator	Multiword Conjunction	Preposition Pronoun Conjunction
ID 353	Clause	Modifier	Not applicable	Pronoun

 

Construction (5) is less structurally complex because its anchor part does not comprise a complex entity. Thus, as a whole this construction is a Clause, where the anchor čego functions as a Modifier. It has an optional element èto which performs an emphatic function. Structurally, čego and èto do not comprise one complex entity; therefore, the syntactic structure of an anchor is not characterized for this construction (tag ‘Not applicable’). Both čego and èto are Pronouns, which is reflected in the part of speech annotation.

While the semantic system is organized as a system of tags to capture the complex semantics of multiword constructions, the syntactic system is organized as a four-dimensional system of types in order to capture the complex syntactic structure of constructions.

The most complicated type of formal classification is the classification of constructional patterns. While within one language (and one constructicon) it is possible to describe the techniques a language uses to make constructions from what used to be free combinations, this cannot be achieved in the typological perspective and, thus, is not so crucial for Croft’s project. However, in some examples it is useful to describe constructional patterns even in typological studies. For instance, what Croft describes as constructions with stative predicates is in fact a result of the widespread process of constructionalization. Here belongs the resultative construction like We painted the door red [Croft 2022: 432], which, according to [Goldberg 1995], made its way to a construction through constructionalization of the combination of a verb like paint and an adjective denoting the final state, such as red.

4.4. Communicative types

Unlike the large semantic and syntactic annotation systems, the system of communicative types in the Russian Constructicon is rather small. This is due to the fact that this annotation system is only applicable to those constructions that form a clause or a combination of clauses. The system of communicative types has only four categories: Declarative, Interrogative, Exclamatory, and Interrogative/Exclamatory. The names of the types are mostly self-explanatory, with the exception of the Interrogative/Exclamatory type, which means that the construction combines two types at the same time. This generally refers to constructions that function as rhetorical questions and in fact express negative attitude of the speaker towards the actions of the addressee, as the construction represented in (5).

5. Two classifications combined

In this section, we describe differences between Croft’s model and the Russian Constructicon, applied to individual examples and generalized semantic domains. Croft’s four-part model might be expected to fit neatly with the annotation model of the Russian Constructicon since the latter is also divided according to form-function distinctions. Presumably, one would search for the associative links between

— the semantic types in the Russian Constructicon and the semantic content (sem) type of the CCs;

— the syntactic types in the Russian Constructicon and the strategy (str) type of the CCs;

— the communicative types in the Russian Constructicon and the information packaging category (inf) type of the CCs.

Construction (cxn), which is the fourth type of CCs, is a more complex comparative concept since it consists of both form and function.

Let us first apply Croft’s CCs to an already discussed example from the Russian Construction. If we look at the construction (5) from Section 4.2, we can characterize its semantic content as causal (sem), which Croft [2022: Appendix] defines as “the semantic relation between two events where one event causes the other”. An additional semantic CC that is applicable here is cause (sem), defined as “a semantic role including participant roles for a participant, usually an event, that causes the event expressed by the predicate” [Ibid.: Appendix]. As we can see, this CC is related to a semantic role of the subordinate element (the pronoun čego) in the construction. The annotation of the Russian Constructicon also includes a set of semantic roles. They are assigned to the constructional slots and are marked in the definition of construction and in example sentences. Example (4), therefore, is annotated in the resource in the following way (6), with a Situation and a Cause filling the slots:

(6) ID 53 vvidu togó, čto Cl, Cl ‘due to the fact that…’

[Izdatel’stv-o objazan-o vyplati-t’ avtor-u vs-e

publishing_house-nom.sg obliged-n.sg pay-inf author-dat.sg all-acc

100 % gonorar-a]_Situation vvidu t-ogo, čto [rukopis’

royalties-gen.sg due_to that-gen.sg that manuscript.nom.sg

otklonen-a uže posle odobreni-ja.]_Cause

rejected-f.sg already after approval-gen.sg

‘The publishing house is obliged to pay the author 100 % of the royalties due to the fact that the manuscript was rejected after approval.’

Syntactically, the most relevant CC from the list for construction (4) is adverbializer (str) defined by Croft [2022: 664] as “a morpheme that overtly expresses the semantic relation in an adverbial clause construction”. Croft has an extended understanding of a morpheme and lists, for example, English because as a morpheme. Therefore, it seems legitimate to extend this terminology to the Russian vvidu togó, čto complex conjunction. Adverbializer is a subtype of conjunction (str) CC, “a free morpheme or clitic that encodes the relation between the events denoted by the two clauses in a complex sentence construction” [Ibid.: 680]. In Croft’s terms, construction (4) would also be considered an instantiation of an adverbial clause construction (cxn), which is a CC defined as “a complex sentence construction with a figure–ground construal / information packaging of the relation between the events denoted by the two clauses” [Ibid.: 664]. As we can see, this CC is still predominantly form-oriented rather than function-oriented.

Zooming out from the level of individual constructions to the level of semantic types within the Russian Constructicon, a broader perspective reveals systematic correspondences between these semantic types and Croft’s Comparative Concepts. A comprehensive comparison at this level seems rather promising, as the two systems demonstrate a substantial level of compatibility because they refer to similar conceptual domains. Thus, 22 out of 55 semantic types in the Russian Constructicon fully overlap with the CCs identified by Croft and 10 semantic types exhibit partial overlap with the set of CCs. We deliberately compare the semantic types and the CCs on a conceptual level because they can be distinct terminologically. For instance, we consider Apodosis (cxn/sem) [Croft 2022: 668] to fully correspond to the semantic type Consequence in the Russian Constructicon as they both denote the consequences of an action. Overall, full and partial overlap with CCs is attested for approximately 58 % of the semantic types in the Russian Constructicon.

Interestingly, for some semantic subclasses, Croft’s approach is more compatible with the account adopted in the Russian Constructicon than for others. For instance, the subclass Logical relations of the semantic class Qualia shows the highest degree of overlap between the system of the Russian Constructicon and Croft’s system (100 %), while the subclass Discourse clauses of the semantic class Discourse shows no overlap at all (0 %). The degree of overlap between the two systems is exemplified in Table 3. The table shows the semantic annotation in the Russian Constructicon (with types, subclasses, and classes). The dark green highlighting is used for semantic types that have a corresponding Comparative Concept in Croft’s classification. The light green highlighting shows that there is a partial correspondence between the classifications. For example, the semantic type Actuality in the Russian Constructicon has six subtypes: Attenuative, Cumulative, Distributive, Gradual development, Punctual, and Saturative, but in Croft’s classification only punctual (sem) is listed as a Comparative Concept.

The clearest difference between Croft’s classification and the Russian Constructicon is observed in semantic domains that contain many values not coded by grammatical markers. Compare, for instance, the aspectual domain: while the opposition of resultative and non-resultative is often coded by verbal inflectional markers, this is rare for saturative. This means that in Croft’s approach resultative or punctual has a greater chance to be classified as a typologically relevant constructional pattern, relevant for the whole-sentence level (punctuality is relevant for the semantic transitivity degree, cf. [Hopper, Thompson 1980], and resultative is closely related to passive). By contrast, saturative only has a chance to be represented in constructicons due to the existence of individual constructions like On napilsja v drova / v nol’ / … ‘He is very drunk’ (lit. ‘He got drunk to the firewood / to zero’, and so on).

 

Table 3

Overlap of the semantic classification of constructions in the Russian Constructicon with Croft’s Comparative Concepts

1. Qualia
1.1. Situation structure	1.2. Major roles	1.4. Logical relations	1.6. Sets and elements
• Timeline + • Taxis + • Actionality + • Pluractionality + • Phase of action + • Result + • Actuality	• Addressee + • Instrument • Possession • Comitative • Caritive • Non-standard subject +	• Cause • Purpose • Consequence • Condition • Concession +	• Additive • Inclusive • Exceptive • Exclusive • Subset • Options • Quantification +
1.3. Situation modifiers		1.5. Properties	1.7. Magnitude
• Spatial expressions + • Temporal expressions + • Manner		• Salient property + • Temporary characteristics + • Comparison +	• Non-existence + • Measure + • Calculation +
2. Modality and its neighborhood	3. Subjectivity	4. Discourse	5. Parameters
2.1. Core modal meanings	• Assessment + • Attitude + • Polarity value + • Source of opinion + • Mirative	4.1. Discourse organization	• Degree of intensity + • Degree of accuracy +
• Root modality + • Epistemic modality +		• Discourse structure + • Intersubjectivity • Objectivity • Source of information
2.2. Neighborhood		4.2. Discourse clauses
• Volition • Causation + • Prohibition + • Threat • Request • Apprehension + • Curse		• Reaction to the previous discourse + • Routine +

 

The second major finding from a comparison on the level of semantic types is that the semantic classification in the Russian Constructicon is often more granular than Croft’s CCs. For example, the Epistemic modality type in the Russian Constructicon has two subtypes: High degree of certainty and Low degree of certainty, while the corresponding CC called epistemic modality (sem) has no subtypes. One would expect this from a large-scale typological classification which aims to cover a fully inclusive inventory of meanings rather than elaborate on fine-grained nuances of meanings. A more fine-grained analysis is, however, beneficial for describing constructions of a particular language.

The difference in the nature of semantic classification allows us to describe Croft’s system as a grammar of constructions (not in the conventionalized sense ‘construction grammar’, but denoting ‘the system representing the most general semantic domains, relevant for grammatical marking and features of the whole sentence’) and a constructicon as a grammatical dictionary of constructions (that contains much information about the behavior of constructions but includes very fine classification, characteristic of lexical descriptions).

The system of communicative types in the Russian Constructicon exhibits almost full correspondence with Croft’s CCs, as shown in Table 4.

 

Table 4

Communicative types compared to information packaging CCs

Communicative type in the Russian Constructicon	Comparative Concept [Croft 2022]
Declarative	declarative (inf/cxn)
Interrogative	interrogative (or question) (inf/cxn)
Exclamatory	exclamative (inf/cxn)
Interrogative / Exclamatory	—

 

The description of parts of speech poses a problem in both approaches. The Russian Constructicon includes a special feature Part of speech of anchor — however, problematic cases where the part of speech is not obvious are not addressed in a systematic way. In Croft’s approach, parts of speech are partially covered by the description of strategies (str) for a given semantic domain. For example, in Chapter 4, numerals and quantifiers (Section 4.1.3) are considered separately from nominal modification constructions (Section 4.1.4), and the description of comparatives and equatives in Chapter 17 distinguishes strategies employing a particle (English than, Latin quam) and those employing a preposition, a postposition, or a case marker (Mundari ete ‘from’, Breton wid ‘for’). However, Croft’s monograph does not answer the question whether the Comparative Concept approach is sufficient to describe basic syntactic notions, such as parts of speech.

6. Possible relationships between the two approaches

The comparison between Croft’s monograph and studies used for building constructicons reveals a difference of approaches — both may be adequate, but their domains of use and the range of tasks they are appropriate for are different. In particular, Croft’s approach, which builds on the a priori classification of grammatical and discourse phenomena, is not very convenient for building the constructicon of a specific language. Moreover, using an a priori classification may result in missing some peripheral (for grammatical typology) language phenomena, which, however, are richly represented by lexicalized constructions.

At the same time, it seems that adopting Croft’s approach can be useful on some stages of building a сonstructicon — not as the only source of language phenomena classification, but rather as a starting point. The most adequate approach seems to be to switch between the typological approach and individual language studies and, then, to analyze their results in a cross-linguistic perspective. A possible strategy is demonstrated below.

1) Croft’s approach can be used as a source point for constructicons, as it identifies a possible range of phenomena that could presumably be used for a constructicon. The list of notional categories and language elements proposed in the monograph is necessary, but not sufficient: domains of grammar mentioned by Croft should be included in any constructicon (or at least checked for being really ‘constructional’ in the language under analysis), though there are many domains and elements missing from Croft’s system.

2) Each of the domains proposed by Croft can be tested against the data of a particular language. This testing makes it possible to add new elements to any domain. Due to the fact that many of Croft’s domains are labeled formally (e.g., “Nonpredicational clauses”), many resulting groups and entries of a constructicon may get a two-part description: for instance, the construction u + np-gen + np-nom (U Pet-i gripp-Ø [by Petja-gen.sg flu-nom.sg] ‘Petja has the flu’) can be semantically classified as a construction that marks a person’s physical state. On the other hand, on the basis of syntax, it can be described as a construction with a zero copula (nonverbal predication). Similarly, the construction praedic + np-acc + v-inf (Neobxodimo èt-o sdela-t’ [necessary.praedic this-acc.n.sg do-inf] ‘It is necessary to do it’) is semantically a modal construction, and syntactically, a nonverbal predication with a predicative neobxodimo.

7. Conclusions

Let us sum up our conclusions based on the comparison of the two approaches. Analysis of the semantic and syntactic classification in the Russian Constructicon suggests that it exhibits considerable overlap with the system of Comparative Concepts introduced by Croft. Conceptually, the structure of annotation in the Russian Constructicon is compatible with Croft’s system, suggesting that Semantic types predominantly correspond to Semantic content (sem) type of CCs, Syntactic types represent the Strategy (str) type of CCs, while Communicative types correspond to Information packaging (inf) type.

However, the semantic classification in the Russian Constructicon is much more detailed than that in Croft’s system, while the communicative types are less developed compared to Croft’s information structure types. Although this comparison provides a preliminary overview, further investigations are required for a more comprehensive analysis of the compatibility of two systems, especially on the level of individual constructions.

A bottom-up classification based on data from a single language and a full-scale typological classification based on multiple genetically distinct languages arrive at similar results, thus cross-validating each other. The complementary categories that are present in one system and absent from the other are a fruitful source for the expansion of the constructicon resource as well as for updating the list of Comparative Concepts. This also suggests that a closed universal list of categories may not be sufficient to capture the semantic types of many language-specific constructions. However, a list of Comparative Concepts can indeed be a useful first step for a multilingual mapping of constructions.

What is definitely missed by Croft’s analysis or a similar approach built on Comparative Concepts is the difference between constructions organized as core grammatical phenomena (for instance, adjunct and complement clauses, voice, and so on) and constructions that include a closed set of elements and are closer to lexical (or quasi-grammatical) phenomena, such as saturative (on napilsja ‘he is drunk, he has drunk enough’) or special types of terminative (on svoё otbegal ‘he cannot run anymore’, lit. ‘he his own out-ran’). Comparative Concepts are not suitable to distinguish between these levels of constructional description. The bottom-up approach to classification makes it possible to differentiate between ‘grammar’ and ‘quasi-grammar’ / ‘lexicon’ in the description of constructions.

Although some technical and theoretical solutions proposed in Croft’s monograph are not uncontroversial, the book will be of great use both for descriptive studies and for linguistic typology and theory. The monograph covers a great deal of linguistic phenomena that either have not received much attention or have rarely been systematically described, with phenomena of different types and different languages put together. In this sense, Croft’s monograph is not only a significant contribution to Construction Grammar and linguistic typology, but also a potential starting point for many studies and projects, lying at the crossroads of grammatical description, typology, and construction grammar.

The combination of Croft’s Comparative Concepts and the constructicon technique developed by the Russian Constructicon provides a way for further development of constructional approaches. One of the questions that arise is whether a two-faceted approach is possible: on the one hand, constructions can be classified on the basis of argument marking and semantic roles of arguments (as well as a more fine-grained semantic classification), on the other hand, labels of the general constructional typology can be used.

A more ambitious task is to elaborate an opposition of assertive / non-assertive or head / non-head semantic components that form the meaning and the formal make-up of constructions. For instance, how formally and semantically different are constructions that are marked as questions and those that are questions only in the discourse level, but not formally marked as questions? What can help us to decide whether most features of a given construction result from it being ‘a question construction’ or ‘a causal construction’? This research problem is waiting to be addressed.

ABBREVIATIONS

2 — 2^nd person

acc — accusative case

dat — dative case

f — feminine

gen — genitive case

inf — infinitive

m — masculine

n — neuter

nom — nominative case

praedic — predicative (nonverbal predicate)

prs — present tense

pst — past tense

sg — singular

 

¹ The resource is available online at https://constructicon.github.io/russian/about/.

² Available online at https://framenet.icsi.berkeley.edu/fndrupal/.

³ The Russian Constructicon is the result of many joint efforts; among the major contributors are Anna Endresen, Daria Mordashova, Ekaterina Rakhilina, Laura Janda, Olga Lyashevskaya, as well as over 40 students from both UiT The Arctic University of Norway and HSE University in Moscow.

⁴ Available online at https://constructicon.github.io/russian/semantic-types/.

Авторлар туралы

Valentina Zhukova

UiT The Arctic University of Norway

Хат алмасуға жауапты Автор.
Email: valentina.a.zhukova@uit.no
Норвегия, Tromsø

Alexander Letuchiy

HSE University; Vinogradov Russian Language Institute, Russian Academy of Sciences

Email: alexander.letuchiy@gmail.com
Ресей, Moscow; Moscow

Әдебиет тізімі