Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos

Acta Linguistica Lithuanica LXI (2009), 61 – 111 Diathesis Alternations from the Perspective of Research in Different Theoretical Models KRISTINA LENARTAITĖ Institute of the Lithuanian Language The article gives an overview of the approaches offered in different theoretical frameworks (Generative Grammar, Relational Grammar, the Lexical Rule Models, Construction Grammar and the Lexical-Constructional Approach) to account for alternations in argument realisation (diathetic alternations) such as the active vs passive alternation, the ‘spray/load’ type alternations, dative alternations etc. The final part of the article presents certain considerations speaking in favour of the derivational view. 0. INTRODUCTORY REMARKS The article discusses the following phenomena of diathesis alternation1: active and passive constructions, applicative alternations, and other cases of alternation in argument expression (for example, locative alternations, etc.). Thus, how- 1 The concept of “diathesis” was introduced by Cholodovich in the early 1970s in order to describe the category of voice universally. In discussions of voice, this term often describes a change in the subject’s semantic function; for example, with an active verb the subject is an agent, whereas with a passive verb it is a patient or “experiencer” (undergoer) (Matthews 2007: 104–105). However, diathesis can also be understood more broadly—as a pattern of the distribution of grammatical functions (sentence constituents) in relation to semantic roles, with voice being only one of the grammatical means that make it possible to change diathesis (Holvoet, Semėnienė 2004: 35–36). Diathesis is understood in this broader sense in this article as well. In linguistic literature, syntactic alternations may refer to a wide variety of processes that change the valency of a predicative word, realized with or without an overt marker of morphological derivation. These phenomena can all be further classified according to their functions: valence-increasing (valence-increasing) processes (for example, causative constructions in relation to the corresponding non-causative ones), valence-decreasing (valence-decreasing) processes (for example, passive constructions are usually intransitive, whereas their active counterparts are transitive); alternations can also be divided into meaning-preserving (for example, active/passive alternations, applicative alternations) and meaning-changing ones (for example, causative constructions) (Kroeger 2005: 270–287). 61 KRISTINA LENARTAITĖ One of the cases under consideration involves two or more constructions in which, with the same verb (with or without an overt marker of morphological derivation), two or even several different diatheses are realized alternatively and are therefore regarded as alternating with respect to that verb. The theoretical models applied to explain diathesis discussed in this article can broadly be divided into two groups: (i) constructional models (only two of these are discussed here—Goldberg’s Construction Grammar and Iwata’s lexical-constructional model) and (ii) other models, such as generative grammar, Relational Grammar, and the so-called “lexical rules” models (of the latter, the article reviews the position of Lexical-Functional Grammar on diathesis alternations, as well as the studies of Levin and Rappaport Hovav and Pinker). At first glance, it might seem that these groups of models differ in whether they posit a derivational relation linking the alternating constructions (where one construction, or its lexical-semantic representation, is considered basic, and the other or others are derived from it), or deny any derivation of one construction from another and claim that the relevant constructions arise independently of one another (the position adopted in constructional models). However, as will become clear later, this distinction—derivational versus non-derivational—is not the only essential criterion distinguishing constructional models from the other models discussed here. Moreover, when the phenomena under study are viewed from the perspective of constructional models, the term “alternations,” although traditionally used by proponents of these models, seems misleading, since these models recognize no derivational relation between constructions at either the syntactic or semantic level2. Thus, it is clear that, had I viewed the phenomena under study from the perspective of constructional models, I probably would not even have defined the object of study in this way; the very title of the article therefore reveals my commitment to non-constructional models. I attempt to justify my theoretical preferences regarding these phenomena and my corresponding choice of research direction at the end of the article. 2 Pavyzdžiui, Goldberg konstrukcijų gramatikoje (1995, 2006) kiekvienas tokių sakinių yra laikomas skirtingų apibendrintų konstrukcijų konkrečiomis realizacijomis, arba instanciacijomis. Kel laiko konkrečių realizacijų atsiradimą čia iš esmės nulemia ne konkretaus veiksmažodžio reikšmė, o jos galėjimas būti „sulieta“ (integruotus) su skirtingomis apibendrintomis konstrukcijomis, ir kiekvienu realizacijos atveju svarbu tik tai, kaip ta konkreti reikšmė dera prie atitinkamos apibendrintos konstrukcijos reikšmės, o klausimas, koks santykis sieja dvi ar kelias vienos veiksmažodžio reikšmės pagrindu atsiradusias konstrukcijas, šiam modeliui nėra aktualus. 62 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos 1. OVERVIEW OF THE PHENOMENA UNDER INVESTIGATION The starting point for this study was various cases of argument expression alternation, such as locative alternations, also known as “spray / load” type alternations (spray / load alternation), “swarm” type alternations (swarm alternation), and so on3. The attempt to define this phenomenon—that is, to establish its functions and identify its place in the language system—prompted comparison with active / passive and applicative alternations. The cases of argument alternation mentioned above and others, as well as active / passive alternations, occur in one form or another in many of the world’s languages, whereas applicative constructions (applicatives) are characteristic of non-Indo-European languages (they are found in many African languages, some Indonesian languages, and others). All the phenomena under discussion, with or without overt markers of verbal morphological derivation (in the case of passivization and applicative constructions), make it possible to redistribute the grammatical functions of subject and direct object without changing either the number of predicate arguments or their semantic roles. Moreover, it is often claimed that the alternating constructions in question have the same propositional meaning (Mel’čuk 1993: 11), since they convey the same conceptual event or the same underlying information (Foley 2007: 362–370). This does not in any way mean that alternating sentences, for example, an active sentence and its corresponding passive, are absolutely synonymous—they differ first and foremost markedly in their unmarked information structure, and some of the phenomena under discussion also exhibit other pronounced differences in meaning. For example, in the case of locative alternations, the predicate in one of the sentences often has an additional holistic seme, indicating that the argument expressed as the direct object is completely affected by the action. Each of the phenomena under investigation is discussed separately below. 3 In this article, I use “argument expression alternations” or simply “alternations” to refer only to phenomena that meet the criteria set out in section 1 1., although in linguistic literature more generally the term may also be used for other phenomena (cf. footnote 1). For example, Levin, one of the best-known researchers of argument expression alternations in English, understands argument alternation as all the diathesis variants permitted by a single verb: this includes cases of variation in the expression of argument positions (subject, direct and indirect object) as well as adjunct-term positions. Furthermore, in English the same verb, as far as its form is concerned (i.e. without any overt markers of morphological derivation), can also realize argument configurations that could be expressed in Lithuanian only with derivatives of the corresponding verbs; compare: a. Janet broke the crystal ‘Janet broke the crystal’ and b. Crystal breaks at the slightest touch ‘The crystal breaks at the slightest touch’ (Levin 1993: 26). 63 KRISTINA LENARTAITĖ 1.1. Argument Expression Alternations I formulated the definition of this phenomenon and the criteria for distinguishing it primarily on the basis of Lithuanian examples (cf. Lenartaitė 2007), but they also apply to analogous phenomena in other languages. Here, too, the discussion will rely mainly on Lithuanian examples; in addition, it will review the so-called “dative” alternations in English and locative alternations in German. Since argument expression alternations have been most thoroughly studied in European languages (first and foremost, of course, in English), one might get the impression that this phenomenon is characteristic only of Indo-European languages. This is not the case: the phenomenon is also found, for example, in Japanese (cf. Iwata 2005, 2008). Thus, only cases that meet all of the following criteria are considered argument expression alternations here: (1) the alternating sentences have the same lexical composition, i.e. the predicate involves the same terms, bearing the same semantic roles; (2) changes in predicate structure are necessary to realize the alternating constructions; (3) changes in argument expression must affect at least one of the major grammatical functions—subject and / or direct object (i.e. different arguments must be realized in at least one of these positions); (4) the relationship between the alternating constructions can be understood as paraphrastic4, and, with respect to the truth conditions of the propositions, often even as equivalent. The various cases of argument expression alternation that meet these criteria are most conveniently classified by reference to the syntactic functions of subject and direct object, to which different terms may be assigned: (i) alternations in the expression of subject and direct object: (1) a. The children surrounded their mother. b. The mother surrounded herself with children. (ii) alternations in subject expression: (2) a. The courtyard smelled of lilacs. b. The yard smelled of lilacs. (iii) alternation in the expression of the direct object: 4 Dėl šios sąvokos ir jos taikymo plg. Chafe (2003: 107–108). 64 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (a) alternations in the expression of both the direct object and the adverbial denoting the location of the action5: (3) a. The girls tuck flowers into their braids. b. The girls adorn their braids with flowers. (b) alternations in the expression of the direct object and the indirect (recipient/beneficiary) object: (4) a. Mom pulls a hat (down/over) onto the child. b. Mom covers the child with a hat. (c) alternation between an adverbial denoting the location of the action/impact and the direct object: (5) a. The dog ran (across) the road. b. The dog crossed the road at a run. (6) a. The child kicked at the ball. b. The child kicked the ball. Now, drawing on specific examples of alternations, we can discuss each of the criteria presented above in greater detail. With regard to the first, it should be emphasized that the requirement for the alternating constructions to have the same lexical composition applies primarily to the identity of the terms (arguments) and adjuncts realized in the alternating sentences, whereas the predicates in such constructions may differ in form. That is, sometimes one of the alternating constructions can be formed only with a prefixed verb, while the other uses the corresponding unprefixed verb (cf. 5ab), or the verbs in the alternating constructions may have different prefixes (cf. 4ab). However, constructions realized with verbs bearing different prefixes are equivalent in all other respects (with regard to all criteria applied to alternations) to constructions formed with predicates that are identical in form; cf.: (7) a. The president (pre-)gave the students books. b. The president rewarded the students with books. (8) a. Mother (lays-)puts apples around the goose. b. Mother covers the goose with apples. 5 This summarizes the content of the various semantic functions performed by oblique terms, such as location, direction, or goal (goal). 65 KRISTINA LENARTAITĖ The lexical meaning of verbs determines that, for some predicates, a change in argument structure can be realized without changing the form of the verb (cf. sentence pairs (1), (2), (3), (6)), whereas for others, changes in semantic structure must be expressed (signaled) by an appropriate marker—in Lithuanian, a prefix (cf. sentences (5b), (7b), and (8b)). However, the processes involved in all the cases of alternation listed above can be explained and described in the same way. Moreover, in many cases, a prefix in one of the alternating constructions is merely preferred, but not required (cf. sentences (3ab))—this is another argument for treating all these cases as instances of the same phenomenon—alternations—and analyzing them in the same way. It is also useful here to mention German locative alternations, where one of the alternating sentences can only be realized with a verb bearing the prefix be-6, cf.: (9) a. Die Randalierer spritzten Farbe auf das Auto. ‘The vandals sprayed paint onto the car’ b. Die Randalierer bespritzten das Auto mit Farbe. ‘The vandals covered the car with paint’ (Examples from Iwata 2008: 153.) German locative alternations can also be invoked as an additional argument for treating Lithuanian alternating constructions uniformly, whether they are realized with verbs identical in form or with the corresponding prefixed verbs. Moreover, the so-called applicative variant of German locative alternations (sentence (9b)) (see also footnote 6), like, for example, applicative constructions in African Bantu and other languages (discussed in more detail in section 1.3), is used by many researchers as an argument for positing a derivational relationship between alternating constructions (derivation may be recognized at the level of constructions or at the level of verbs’ lexical-semantic representations): the argument is that, formally, the prefix be- 6 Since this process in German, whereby an oblique term is promoted to direct-object position and which is consistently accompanied by be- prefixation, is very similar to applicative constructions characteristic of non-Indo-European languages, where analogous processes of promoting terms to the position of the clause’s primary object (primary object) are likewise obligatorily accompanied by the addition of an applicative suffix to the verb, the corresponding German construction is also called applicative (cf. Michaelis & Ruppenhofer 2001). 66 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos the direction of derivation marked by the prefix coincides with the derivational processes assumed to take place in lexical semantics or at a deep syntactic level. However, as is typical of most linguistic phenomena, German alternations also have exceptional cases. For example, in very rare cases German allows alternating sentences to be expressed with an unprefixed verb form, alongside a possible prefixed form (for instance, with the verb streichen ‘spread, smear’) (see Iwata 2008: 153–156 for more detail). This situation is very similar to that of Lithuanian alternations. Furthermore, in German one can also find cases where the prefixed verb in an applicative construction has no unprefixed counterpart; for example, the use of beampeln ‘install traffic lights’ (literally, ‘traffic-light’) is attested in an applicative construction, but there is no corresponding sentence with the verb *ampeln. This can be explained as follows: since the verb is derived from the noun Ampel ‘traffic light’, its meaning incorporates the base noun as an object, so it can only be used in a construction in which an oblique term bearing the semantic role of location is promoted to direct-object position (and, in a passive construction, correspondingly to subject position), cf. Die Kreuzung Arsenalstraße wird beampelt ‘The Arsenalstraße intersection will be equipped with traffic lights’ (lit. ‘traffic-lighted’). (Example from Michaelis & Ruppenhofer 2001: 19.) This example poses difficulties for proponents of derivational models: the derivation is inconsistent here, since one derivational step is skipped, cf.: Ampel → *ampeln → beampeln. To conclude the discussion of the criterion of identical lexical composition in alternating constructions, it should be added that cases in which one of the terms cannot be realized in one of the constructions are not considered alternations, cf. The girl wiped the crumbs off the table and The girl wiped the table (?? of crumbs). These sentences are probably realized with verbs of different meanings: in the first case, ‘to remove something from somewhere by wiping’; in the second, ‘to clean (make clean, dry) something by wiping it’. The second criterion—that changes in predicate structure are required for alternating constructions—may be considered redundant: changes in predicate structure can only be identified from changes in its argument structure, which are defined by the third criterion. However, this criterion was intended to distinguish alternations from cases involving variation in the expression of oblique terms (cf. Bears sleep through the winter and Bears sleep in winter—changes in oblique terms do not alter the pred- 67 KRISTINA LENARTAITĖ (of the predicate structure) and from other cases resembling alternations (cf. They tore the child away from the mother and They tore the mother away from the child — here, too, the predicate structure remains unchanged, since in both cases the predicate opens up identical positions, which are merely occupied by different participants in the situation). The third criterion—that only cases involving variation in the realization of the subject and/or direct object count as syntactic alternations—draws a clear boundary between the two purely structural syntactic positions, subject and direct object, and the partly syntactic, partly semantic position of the indirect object7. The grammatical function of the indirect object correlates with a compact group of semantic roles (recipient / benefactive / experiencer / addressee), and no cases have been found in which several different terms selected by a single verb compete to occupy the syntactic position of the dative object. A term realized as a dative object, like other oblique terms realized in the position of more oblique objects or adjuncts, can at most be promoted to the direct-object position (sentences (4), (7)); by contrast, it is inconceivable that an adjunct term realized in a lower adjunct position could be promoted to the dative-object position, let alone that an argument realized in the subject or object position could be demoted to the dative-object position. In stark contrast to the grammatical function of the indirect object just discussed, the subject and direct object, although prototypically understood as correlating with the semantic roles of agent and patient, respectively, also allow adjunct terms bearing any other semantic roles to be promoted to these positions. When discussing the role of Lithuanian dative objects in alternations, it would be useful to review the “dative” alternations in English, also known as “dative shift” (dative shift), cf.: (10) a. John handed the scarf to Mary. “John handed the scarf to Mary” b. John handed Mary the scarf. “John handed Mary the scarf” (Examples from Shaumyan 2006: 236.) Sentence (10b) is also called the double-object construction (double object), or ditransitive- 7 Here, the term “indirect object” is given a narrower meaning than in traditional Lithuanian grammar, referring only to dative objects with a recipient / benefactive, etc. semantic role. 68 Diatesės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos a construction, and these constructions occupy a distinctive position in English grammar, since they play a crucial role in distinguishing recipient/beneficiary objects from other oblique sentence elements (oblique) and assigning them the grammatical function of indirect object. In other words, were it not for ditransitive constructions, in which a term bearing the semantic role of recipient/beneficiary is realized as the primary object of the sentence (primary object) and therefore loses its prepositional marking (cf. Mary in sentence (10b) and to Mary in sentence (10a)), there would be no substantial reason to distinguish the term realized by a prepositional construction in sentence (10a) from other oblique terms, also expressed with prepositions. Thus, in English it is easier than in Lithuanian to draw a distinction between the core grammatical functions—subject and direct object—and obliques: in English, everything realized by prepositional constructions is considered an oblique. As already noted, the only exception here is prepositional constructions that correspond to recipient/beneficiary objects. Continuing with English “dative” alternations, it is also worth mentioning that they involve verbs whose meanings are broadly assigned to the semantic class of “giving” verbs8. If we were to compare English “dative” alternations with Lithuanian alternations involving terms bearing recipient (cf. sentence (7)) or beneficiary (cf. sentence (4)) roles from the perspective of verb meaning, we would first observe 8 “Dative” alternation ditransitive constructions in English and other languages are most often realized with verbs denoting physical transfer (transfer), such as ‘give’, ‘grant’, ‘sell’, ‘return’, and so on. Such verbs encode an event in which an agent causes an object to pass into the possession of another animate object (i.e. a recipient, receiver, or prospective possessor). In many languages, benefactive constructions (benefactive) are very similar to ditransitive constructions. Verbs denoting benefactive situations (i.e. actions performed for someone’s benefit) are often difficult to distinguish from the physical-transfer verbs just discussed. For example, the English ditransitive construction She brought me a coffee ‘She brought me a coffee’ can also be paraphrased with a “dative” construction, She brought a coffee to me ‘She brought a coffee to me’, or with a benefactive construction, She brought a coffee for me ‘She brought a coffee for me’. As we have just seen in English, and likewise in many other languages, verbs denoting benefactive situations can be realized in ditransitive constructions, which often makes it difficult to differentiate the semantic roles of recipient and benefactive (beneficiary). Unlike in the case of verbs denoting physical transfer, however, a participant bearing the semantic role of beneficiary can also be realized with intransitive verbs, for example, She sang for me ‘She sang for me’ (Malchukov, Haspelmath & Comrie 2007). 69 KRISTINA LENARTAITĖ we would see that Lithuanian cases with a benefactive dative should be distinguished from English cases, since the beneficiary role borne by the terms here is not associated with “giving” semantics, and alternations of this type in Lithuanian are better viewed as a subtype of locative alternations, cf.: (11) a. The woman (ap-, už-) pulled the coat onto the girl /??onto the little girl /?onto the mannequin / onto the dummy. b. The woman (ap-) dressed the little girl / the mannequin in a coat. (12) a. Mother (ap-) gave the shoes to the child / *onto the child / onto (little) feet / *(little) feet. b. Mother (ap-) shod the child / little feet with shoes. In sentences (11) and (12) (cf. also sentence (4)), the animacy feature of the terms marked by the dative (mer-g ytei, vaikui) prevents them from being expressed by a prepositional construction, as is characteristic of locatives. The fact that the semantic role of beneficiary is essentially related to the animacy hierarchy of nouns (Silverstein 1976: 112–171) is also evidenced by the fact that in sentence (11) an inanimate object can also be marked by the dative (manekenui), or that a living being (but not a human!) can be marked in two ways, cf. Strike a horse with a blanket on it / a blanket over the horse. Moreover, it would hardly be possible to assign verbs that allow such alternations, such as vilkti, a uti, mut uriuoti and so on, to the semantic class of “giving” predicates. Of the Lithuanian examples mentioned, only sentence (7) can be compared in meaning with English “dative” alternations: its verb (pa-, ap-)dovanoti undoubtedly has “giving” semantics; cf. also (pa-)dalinti obuoliu vaikams ir (ap-)dalinti vaikus obuoliais. Nevertheless, for now we will distinguish such Lithuanian cases from “dative” alternations in English and other similar languages, if only because in those languages ditransitive constructions that reveal an indistinct distribution of object functions (cf. Holvoet 2009: 48–55) are closely associated with fixed word order in the sentence (Rappaport Hovav & Levin 2008: 160–163). Moreover, closer examination shows that the group of English verbs permitting “dative” alternations is itself heterogeneous; for example, we can distinguish here the semantic classes of give ‘give’ and throw ‘throw’: verbs of the first class do not emphasize the semantic aspect of a change in the location of the argument bearing the theme role, so both alternating sentences are always associated with a caused-possession meaning, whereas verbs of the second class do emphasize that semantic aspect, so their alternating sentences can have either 70 Di atezes alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos of caused motion, or caused possession (Rappaport Hovav & Levin 2008; Jackendoff 1990: 197–200). Lithuanian cases of recipient-term alternation with respect to verb meaning need to be investigated more thoroughly, so for now we will not compare them with English “dative” alternations. The fourth criterion—the paraphrasability or even logical equivalence of propositions expressed by alternating constructions—is the most immediately striking and salient property of the cases under study, and, as will soon become clear, is no less important in distinguishing alternations from other phenomena; cf.: (13) a. Jonas sprayed all my paint (onto the wall of the house). b. *Jonas sprayed the wall of the house with all my paint. c. Jonas sprayed the wall of the house with my paint. (14) a. Jonas sprayed my paint onto all the walls of the house. b. Jonas (completely, off-)sprayed all the walls of the house with my paint. In sentence (13a), the verb išpurkšti means ‘to use up, to exhaust something by spraying’, whereas the verb in sentence (13c) means ‘to cover a certain surface by spraying some liquid or the like onto it’—so the verbs in these sentences have different meanings, and therefore the sentences cannot possibly be paraphrases of one another; consequently, they cannot be regarded as alternations. Moreover, the direct-object quantifier compatible with the verb meaning in sentence (13a) (all) does not allow the corresponding realization with the “surface covering” meaning; cf. sentence (13b). Sentences (14ab) can be understood as paraphrases, but they are asymmetrical in terms of the “entailment” relations between their meanings; cf.: it follows from Jonas (completely, off-)sprayed all the walls of the house with my paint that Jonas sprayed my paint onto all the walls of the house, but not vice versa. This asymmetry in the “entailment” relations between the meanings of alternating sentences is due to the holistic seme characteristic of one of the sentences (for example, 14b), which indicates that the object is wholly or at least largely affected by the action. This is characteristic of the great majority of locative alternations (i.e., alternations in which, broadly speaking, an adjunct term bearing the semantic role of location is promoted to subject position (2b) or direct-object position (3b, 5b, 6b, 8b, 9b, 14b)). In cases where it is not a locative object but a benefactive / recipient object that alternates, the sentences can usually be judged equivalent, probably because here the effect of the holistic seme, if one can even be discerned, differs from that in locative alternations; cf.: if it is true that Mama put a coat on the child, then it is true that Mama dressed the child 71 KRISTINA LENARTAITĖ with a coat, and vice versa—if it is true that Mama put a coat on the child, then it is true that Mama put the coat on the child. Thus, to refine this criterion, it should be added that in the case of locative alternations, the sentences may be regarded as paraphrases, but their semantic relation (because of the holistic seme) is asymmetric: one of them (3b, 5b, 6b, 8b, 9b, 14b) entails the information of the other (3a, 5a, 6a, 8a, 9a, 14a, respectively), but not vice versa. Consequently, logical equivalence between propositions is characteristic of not all cases of argument-expression alternations; however, as we have seen, the paraphrase relation between alternating sentences is a very important feature of the phenomena under discussion. Finally, it should be emphasized once again that a sentence pair classified as an alternation phenomenon must meet all four criteria. 1.2. Passivization This section merely defines the phenomenon of passivization, that is, it discusses its place in the language system and its most characteristic features; discussion of such questions as the relationship between active and corresponding passive constructions, or atypical cases of passivization, is deferred to the sections reviewing the positions of different theoretical models regarding voice alternation phenomena. First, it should be noted that passivization is to be regarded as a subcategory with respect to the supercategory of voice (Mel’čuk 1993). In general, voices9 can be described as an operation that, through certain oppositions (for example, active—passive constructions or active—middle 9 Žvelgiant iš kalbų tipologijos perspektyvos, rūšies kategorija apima funkcijų atžvilgiu labai panašius, tačiau realizacijos būdų (arba skirtingų realizacijos lygmenų) atžvilgiu labai įvairius reiškinius, kaip antai Filipinų kalboms būdinga fokuso rūšis (focus voice), kuri nepakeičia termų sintaksinio statuso, tačiau žymi sakinio informacinės struktūros (t. y. temos ir remos) pokyčius: atitinkamas veiksmažodinės leksėmos žymiklis koindeksuoja fokuse esančio termo semantinį vaidmenį; inversinė rūšis (inverse voice) turinčiose kalbose argumentų paskyrimo loginiu subjektu ir loginiu objektu alternacijos orientuojamos pagal predikacijos referentų ontologinio iškilumo / neįskilumo hierarchiją (kurios subhierarchijomis laikytinos potencialaus agentyvumo ir gyvumo hierarchijos); įvestinė rūšis (derived voice), skirtingose kalbose apimanti įvairias žymėjimo strategijas, leidžiančias alternatyviai priskirti vieną aukščiausią — subjekto — poziciją skirtingiems argumentams (pavyzdžiui, aktyvinės — pasyvinės alternacijos); pamatinė rūšimi (basic voice) vadinamos aktyvinės — medialinės alternacijos (Klaiman 1991). 72 Di atezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos constructions) realize a different distribution of arguments across syntactic positions (in other words, voice) without changing propositional (i.e., truth-conditional) meaning. Another feature important for defining voice phenomena is that all voice changes occurring here are marked in the verb structure. Marking strategies may differ across languages; thus, generalizing about the options for marking passive constructions, it may be observed that languages can have a strictly morphological passive (where a passivizing affix is attached directly to the verb (for example, in Japanese), realized by an infix (in Tagalog (tagalog, Philippines) or by vowel alternation (in Hebrew and Arabic)) or a periphrastic passive, realized by an auxiliary verb and a special form of a transitive verb (for example, the so-called passive participle) (for example, in Lithuanian, German, and English). Sometimes a language may have several marking strategies (for example, in Russian, alongside productive periphrastic passive forms, there is another marking strategy using the reflexive affix (-sja), cf. Doma strojat-sja rabočimi ‘Houses are being built by workers’). Only very rarely is the process of passivization unmarked in the verb structure by any special marker, being indicated solely by the verb’s agreement with a subject of the relevant noun class (for example, in Swahili (swahili, a Bantu language)) (Keenan & Dryer 2007: 332–339). Next, we will discuss the functions of passivization and the processes accompanying this phenomenon. Passivization, like all phenomena of the voice category, preserves the proposition’s meaning (its so-called semantic content); therefore, the differences realized by active and corresponding passive constructions are primarily pragmatic (i.e., they reorganize the sentence’s information structure). However, unlike most other means of marking information-structure differences, which are considered sentence-level operations—for example, logical sentence stress, inversion of word order in the sentence, and detached topics (see Holvoet 2003: 79–98)—passivization is a predicate-level operation (Keenan & Dryer 2007: 325–328). In general, the realization of passive constructions involves the following main processes: shifting the most prominent (thematic) argument into the background (backgrounding), that is, into an oblique position, thereby demoting (demotion) its syntactic status, or bringing a non-thematic argument into the foreground (thematic) position in the sentence (foregrounding), thereby promoting (promotion) its syntactic status (i.e., realizing it in the syntactic subject position), cf.: 73 KRISTINA LENARTAITE (15) a. Jonas built a house. b. The house was built (by Jonas). (16) a. John kissed Marry. “Jonas kissed Merė1 b. Mary was kissed (by John). “Merė was kissed (by Jonas)1 In both the Lithuanian (15b) and the English (16b) passive sentences, compared with their corresponding active variants, both types of operation take place: the detopicalization of the most prominent (agentive) argument (Jonas, John), realized in an oblique syntactic position, and the topicalization of the second-most prominent (patientive) argument (house, Marry), realized in subject position. However, both languages allow the agentive argument to be expressed or not in passive sentences, as required. Here, too, it is important to answer the question of which passive construction should be taken as basic: the one with an expressed agent phrase, i.e. the “long” passive construction (long / agented passive), or the one without an agent phrase, the “short” passive construction (short / agentless passive). Another question related to the latter is which of the processes of passivization—the promotion of a participant in the encoded situation to a foregrounded (thematic) position, and the corresponding raising of its syntactic status, or its demotion from the most prominent position, and the corresponding lowering of its syntactic status—constitutes the essence of this phenomenon. Here, too, there are differing opinions. For example, Givón considers passivization to be a process that promotes a non-agentive argument to the primary thematic position. And because languages have coding devices that identify the primary thematic argument as the subject, the promotion procedure in a passive construction may also involve subjectivization (Givón 1979: 186). Evidence in favor of this operation could also be the fact that promoting a non-agentive argument to subject position necessarily involves demoting the agentive argument, but not vice versa—thus, demotion could be regarded as a consequence of the promotion procedure. However, there are quite a few languages, such as Latvian (17) and the Taba language of Indonesia (18), whose passive constructions do not allow the realization of a demoted agentive argument at all; cf.: (17) Es tieku mācīts (*no mātes). 1SG become-1SG teach:PART.PASS.PRAES “I am being taught1 74 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (18) a. i n=bes niwi. 3SG 3SG-peel coconut nut “She peeled a coconut” b. niwi ta-bhes do. coconut nut DEAGENT-peel REALIS “A coconut was peeled” (Examples from Keenan & Dryer 2007: 331.) In sentence (18b), the prefix marking the passive predicate indicates, in a more general sense, the absence of an agent; when attached to semantically transitive verbs, it functions as a marker of passivization while also showing that the sole argument of such a verb corresponds to the patient of the active sentence (ibid., 331). These facts, together with impersonal passive constructions (cf. the Latin impersonal passive construction (b): a. Boni cives (NOM.PL) legibus (DAT.PL) parent “Good citizens obey the laws” and b. Legibus (DAT.PL) (a bonis civibus) paretur “The laws are obeyed (by good citizens)” (Keenan & Dryer 2007: 347), as well as cases of passivization of intransitive verbs (cf. the German impersonal passive construction with an intransitive verb Gestern wurde getanzt “There was dancing yesterday” (ibid., 346)), attest to the value of backgrounding the agentive argument (by lowering its syntactic status)10. This position will also be adopted here; therefore, passive constructions without an expressed agent phrase are taken as basic (cf. Keenan & Dryer 2007: 328–342, Mel’čuk 1993: 21, Desclés, Guentchéva & Shaumyan 1985: 26–27, 42–46). 1.3. Applicative constructions As already mentioned, these constructions are characteristic of non-Indo-European languages; for example, they are highly developed in the Bantu languages of Africa. In brief, this 10 Such atypical cases of passivization may also plausibly be taken as additional evidence in favor of the demotion operation, where a patient argument is also realized in the direct-object position in a passive construction, cf.: a. U nej třex teljat (ACC) zarezano (PART.PASS.SG.N) “lit. at her three calves slaughtered” (Mel’čuk 1993: 24) and b. The puppy still hasn’t been given anything today (Jablonskis 1997: 140). In the first case (a), the passive construction is impersonal, while in the second (b), a recipient argument has been promoted to subject position; in both cases, however, the position of the patient argument remains unchanged compared with the corresponding active sentences—it remains in object position. 75 KRISTINA LENARTAITĖ This phenomenon is described as the promotion of an oblique term to the direct-object (or primary-object) position. This process is always marked in the verb structure by an applicative suffix. Let us consider several examples from Chichewa (chichewa), a Bantu language11: (19) a. Ndi-na-tumiz-a kalata kwa mfumu. 1SG.SUBJ-PRAET-send-ASP letter to-chief “I sent a letter to the chief” b. Ndi-na-tumiz-ir-a mfumu kalata. 1SG.SUBJ-PRAET-send-APPL-ASP chief letter “I sent the chief a letter” (20) a. Msangalatsi a-ku-yend-a ndi ndodo. entertainer(1) SUBJ(1)-PRAES-walk-ASP with stick “The entertainer walks with a stick” b. Msangalatsi a-ku-yend-er-a ndodo. entertainer(1) SUBJ(1)-PRAES-walk-APPL-ASP stick “lit. The entertainer walks the stick” (Examples from Kroeger 2004: 64.) In sentences (19a) and (20a), oblique terms realized by prepositional constructions (kwa mfumu, ndi ndodo) acquire a different syntactic status in sentences (19b) and (20b)—here they are direct objects. This process of promoting an oblique term to the direct-object position is always marked in the predicate structure by an applicative suffix—in Chichewa, the suffix is (-ir- ~ -er-). One might ask whether the terms mfumu and ndodo in sentences (19b) and (20b) really are objects. We will soon see that the objecthood of these terms is evidenced by word order, verb agreement with the object, and passivization diagnostics. In sentence (19a), the term expressed without a preposition (i.e., unmarked in form) (kalata) occupies the position immediately after the verb, whereas in sentence (20b) this position is occupied by the term (mfumu), which is likewise realized without a prepositional construction. Sentence (20b) is ditransitive (or, in other words, a double-object construction), since it contains two unmarked object complements; however, in applicative constructions (i.e., those whose verb has an applicative suffix), the promoted object, which has the semantic function of recipient/beneficiary, is realized first 11 Duomenys apie čičeva kalbos aplikatyvines konstrukcijas pateikiami iš Kroeger (2004: 63–71). 76 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos in a postverbal position before another object complement that is also unmarked in terms of expression and has the semantic function of theme. In most languages with double-object constructions, the primary object is also realized in the first postverbal position (cf. the English ditransitive construction (10b)). Furthermore, in Chichewa the verb must agree in gender12 with the subject, whereas agreement with the object is optional. If the verb does not have an object-agreement marker, the object must be realized immediately after the verb. However, when the verb has an overt object-agreement marker, the object need not be realized in the first postverbal position. In Chichewa applicative constructions, the promoted (or applicative (applied object)) object, if the verb does not have an overt agreement marker, is realized immediately after the verb; if the object-agreement marker is overt, the object may also be realized elsewhere or left unexpressed altogether, cf.: (21) Amayi [a-ku-mu-tumb-ir-a mtsuko]VP woman(1) SUBJ(1)-PRAES-OBJ(1)-pour-APPL-ASP cup(3) (mwana). child(1) “The woman pours a cup for him (the child).” (Example from Kroeger 2004: 67.) Here in (21), the verb shows gender agreement with the primary object (mwana), which is realized outside the verb phrase or may even be left entirely unexpressed. The passivization data discussed below also confirm that the promoted arguments of applicative constructions are direct (or primary) objects, cf.: (22) a. Ndodo i-ku-yend-er-edw-a. stick(9) SUBJ(9)-PRAES-walk-APPL-PASS-ASP “The stick is walked with.” b. *Ndodo i-ku-yend-edw-a (ndi). stick(9) SUBJ(9)-PRAES-walk-PASS-ASP with (“The stick is walked with.”) 12 Chichewa has a complex gender-class system consisting of 18 classes; in glosses, the gender class is indicated in parentheses by Arabic numerals. 77 KRISTINA LENARTAITĖ In the passive sentence (22a) (the passive verb is marked by the suffix -idw- ~ -edw-), the subject is the argument ndodo, which has the semantic role of instrument; however, the verb in this sentence also has an applicative suffix (-er-), indicating that ndodo is a promoted object, cf. sentences (20a) and (20b), in which the argument in question is, respectively, an oblique and an object. The verb in sentence (22b) does not have an applicative suffix indicating that ndodo is the promoted object, and therefore its promotion to subject position in the passive sentence is impossible. The predicate of applicative ditransitive sentences, such as (19b), would allow only an object with the semantic role of recipient, but not one with the semantic role of theme (i.e., the primary object), to be realized as the subject in a passive construction. Thus, to summarize: the purpose of applicative constructions is to promote an oblique argument to the syntactic position of object. This process is always accompanied by changes in predicate structure, which in Chichewa are marked by an obligatorily expressed applicative suffix. The German locative alternations discussed in section 1.1 are very similar to applicative constructions (see also footnote 6). In this article, applicative constructions are invoked not only to attest to the wide variety of diathesis alternation phenomena, but primarily because various theoretical models often use them to support their hypotheses about other argument alternation phenomena. 1.4. Concluding remarks Having reviewed all the phenomena under study, we can summarize and compare how the following aspects are reflected in them: (i) changes in the meaning and functions of alternating sentences, (ii) which syntactic positions are affected in terms of expression, and (iii) how a change in argument status is reflected in verb structure. Looking across all the cases under study, the paraphrastic relationship between alternating sentences is immediately apparent; taken together, these phenomena can be said to function as grammaticalized means of expressing changes in information structure. In typical cases, active/passive alternations involve changes in the expression of both major syntactic positions—the subject and the direct object. Applicative alternations involve variation only in the expression of the direct object. The various cases of argument-expression alternation discussed in section 1.1 allow variation in the expression of the subject and/or the direct object. Moreover, if changes in the syntactic status of arguments in active/passive and applicative alternations are always, respectively 78 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos are marked in the verb structure, other argument-expression alternations behave very differently in this respect: for example, the vast majority of German locative alternations are marked by the verbal prefix be-; in English alternations, the verb never has any external marker; Lithuanian behaves inconsistently here—often the same verb permits an alternation both without any markers and with the corresponding prefixed derivative, but there are also many cases in which one of the alternating constructions can be formed only with the corresponding prefixed verb. Different theoretical models treat these variations in the marking of alternation processes in the verb structure differently. It is clear that all the phenomena under investigation have a great deal in common, but is this really sufficient grounds for comparing them and seeking common principles that govern these processes? There can be no single answer to this question either. However, if we choose the concept of voice as the basis for defining all the phenomena under investigation, we can view them as different configurations of the syntactic positions occupied by the semantic variables (terms) selected by a single verb. Consistently maintaining this position further allows us to treat all these phenomena as subcategories of a single supercategory—the category of voice—if voice is defined as an inflectional13 (inflectional) category specifying modifications of the basic voice of a verbal lexeme (Mel’čuk 1993: 11). It should be noted here that, in this case, one voice is treated as basic (basic), and the others as its modifications, which are most often expressed by a corresponding marker in the verbal lexeme14. 13 The term “inflectional category” should be understood somewhat differently here than usual (i.e., not as a paradigm of full morphological oppositions), since, for example, in Lithuanian the opposition between active and passive voice is characteristic only of participles, not of finite verb forms, while differences in voice meaning are expressed by a verbal phrase; this is one reason why the Lithuanian active–passive voice oppositions are better regarded as a semantic rather than an inflectional category. Moreover, in Lithuanian it is not possible to establish a strict paradigmatic correspondence between active forms and their corresponding passive forms (cf. Holvoet & Semėnienė 2004: 36–41). The “inflectional nature” of voice here should be associated with the corresponding marking in the verb structure that most often accompanies modifications of the basic (basic) verb diathesis (i.e., the configuration of the semantic roles and syntactic functions of arguments) (cf. Mel’čuk 1993: 5–13). 14 The fact that in some cases of alternation in some languages the verb has no additional marker is not considered grounds for excluding such cases from those that do have the corresponding markers (cf. Mel’čuk 1993: 25–29). 79 KRISTINA LENARTAITĖ From this perspective, comparing the phenomena under investigation is indeed justified. 2. OVERVIEW OF THEORETICAL MODELS Before turning to an overview of the positions taken by different theoretical models on diathesis alternations, it should be emphasized that the main focus here will be on cases of argument-expression alternations (i.e., the phenomena discussed in Section 1.1), while active/passive and applicative alternations will generally be invoked for purposes of argumentation. As noted in the introductory section, all the theoretical models discussed can be divided into two groups—constructionist models and all the others, for which it would be difficult to find a single common label. 2.1. Generative, Relational Grammar, and “Lexical Rule” Models What all these theoretical models have in common is that, at one level or another—for example, at the sentence level or at the level of the predicate’s semantic structure, i.e., lexical-semantic representations—they seek to define the relationship linking alternating constructions. Let us first consider models that recognize several levels of syntactic-representation organization, such as generative and relational grammars. Generative grammar distinguishes two levels of syntactic structure—deep structure (deep structure) and surface structure (surface structure)—the former provides the basis for the semantic interpretation of a sentence, while the latter provides the basis for phonological representation. Transformational rules operate between these two levels. For example, it is argued here that the corresponding active and passive constructions share a deep syntactic structure, but that when the passive construction is realized, special transformational rules for noun-phrase15 movement (NP movement) are applied before surface structure is organized (cf. Radford 2002: 420–435). Relational grammar is similar to this theoretical model, except that, unlike generative grammar, where the predicate and 15 Here, the term is interpreted in a purely syntactic sense and includes only the complements of a verb (words or phrases). 80 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos The relations of terms are defined in terms of their structural configuration (for example, a noun phrase directly dominated by the sentence node bears the subject relation, while a noun phrase directly dominated by the verb phrase bears the object relation); in relational grammar, the relation a term bears with respect to the predicate—subject, object, and indirect object—functions directly in syntactic description, that is, as a primitive notion, not derived from anything and not defined on the basis of any other elements. In relational grammar, the grammatical functions of terms are identified at each level, or stratum, of syntactic representation (stratum), and the rules defining changes in the corresponding grammatical functions relate the current representation to an earlier, underlying representation (underlying) and reassess the grammatical functions of terms if the current representation is not at the initial level, that is, if it comprises more than one stratum (Blake 1990). For example, passive constructions are claimed here to consist of at least two strata; cf.: (23) a. John sent the letter to Mary. “John sent the letter to Mary” b. John sent Mary the letter. “John sent Mary the letter” c. The letter was sent to Mary by John. “The letter was sent to Mary by John” d. Mary was sent the letter by John. “lit. Mary was sent the letter by John” (Examples from Perlmutter & Postal 1986: 92–93.) Only sentence (23a) can be regarded as monostratal, in which the term John functions as the subject, letter as the direct object, and (to) Mary as the indirect object. Sentence (23b) is bistratal. In the final stratum, represented by the sentence under discussion, its terms have the following grammatical functions: John is the subject, Mary the direct object, and letter the chômeur (French for ‘unemployed’); however, the underlying stratum of this representation is sentence (23a), reflecting the different distribution of grammatical functions among terms discussed above. Sentence (23c) is the passive counterpart of sentence (23a), while sentence (23d) is the passive variant of sentence (23b). Sentence (23c) is bistratal because its underlying representation is the monostratal representation of sentence (23a); the functions of its terms are distributed as follows: letter is the subject, (to) Mary the indirect object, and (by) John the chômeur. The passive sentence (23d), however, comprises three 81 KRISTINA LENARTAITĖ strata, since its underlying representation, (23b), was already bistratal; the functions of the terms in (23d) are distributed as follows: Mary is the subject, letter the chômeur, and (by) John the chômeur (cf. ibid., 92–95). The pair of sentences (23a) and (23b) exemplifies “dative” alternations (cf. sentences (10) in section 1.1); thus, in relational grammar, sentences realizing both “dative” and active/passive alternations are treated alike—one is underlying with respect to the other (derived) sentence. However, one important aspect seems not to be defined in relational grammar: namely, how the position of the relevant predicate changes across representations of different strata: whether it remains the same (although this seems unlikely when the grammatical functions of terms change), or whether changes in the predicate are assumed to be revealed automatically by changes in the grammatical functions of terms. Moreover, the morphological changes in predicate structure expressed in the corresponding alternating constructions (for example, passive-form markers and the like) seem to be entirely disregarded here—presumably they are abstracted away, as is linear word order in the sentence (cf. Blake 1990: 2). Returning to how generative grammar treats the diathesis alternations at issue, it should be noted that a derivational relation is unequivocally established here only between active/passive alternations, whereas in the case of so-called locative and “dative” alternations, different researchers offer different interpretations even when working within the same theoretical model. For example, in the version of the grammar under discussion known as Government & Binding, the Projection Principle (Projection Principle) stipulates that all lexical properties of a verb be encoded at the outset of the derivation, that is, already in deep structure. The subcategorization properties of each verb (i.e. its complements, their type (whether realized as a prepositional phrase (PP) or without a preposition (NP)), and the order of the complements in the sentence (Radford 2002: 339–361)) can be predicted from the thematic properties of its arguments; thematic roles (for example, agent, theme, etc.) therefore play a part in organizing the syntactic projection of the verb. In the case of alternations in argument realization, a fundamental question arises: are there several verbs, each specifying thematic roles associated with different syntactic structures in a different way, or is there just one verb associated with a single syntactic structure, with the other structure regarded as derived from the first? Since the phenomenon of alternation encompasses many different cases, it is unsurprising that one of these strategies is applied to some cases and the other to others. As for locative alternations (for exam- 82 Diатеzės alternacijos iš skirtingų teorinių modelų tyrimų perspektyvos ple, a. I loaded hay onto the truck ‘I loaded hay onto the truck’ and b. I loaded the truck with hay ‘I loaded the truck with hay’), the more common view is that such realizations are based on different verbal lexemes (cf. Radford 2002: 372–384). There is no comparable consensus regarding “dative” alternations (cf. (10), (23ab))16, so let us examine this case in more detail. When “dative” alternations are recognized as related by a derivational relation, it is the double-object construction (10b, 23b) that is considered derived from the “dative” construction (10a, 23a). When the cases under discussion are treated as lexical alternations (i.e. as syntactic projections of different verbal lexemes, not linked by a derivational relation), the following points are often cited in support: (i) derivational operations are generally characterized by regularity and productivity, whereas “dative” alternations exhibit neither (these alternations involve only verbs from certain semantic groups, but even for those specific groups it is difficult to define the constraints that permit or disallow the alternations); (ii) constructions related by derivation are incompatible with the operation of the Projection Principle (see above). Proponents of a derivational relation argue that without recognizing derivation, it would also be impossible to account adequately for such regular and productive alternations as applicative alternations (see section 1.3.); indeed, the latter are invoked to explain the processes by which “dative” alternations are realized. We have already seen that German locative alternations were compared with applicative alternations: one of the constructions is necessarily realized with a verb bearing the prefix be- (see section 1.1 and cf. footnote 6). Thus, in order to support the transformational shift of the dative in double-object constructions, it was proposed that these constructions in English and other Germanic languages be analyzed as realizations of a zero applicative suffix. In other words, the predicate of double-object constructions is derived from the predicate of the corresponding “dative” construction, but is realized with a zero morphological derivation marker—that is, the derivational marker is not overtly expressed. First, this makes it possible to encompass analogous phenomena across different languages and explain them on the same basis; moreover, this interpretation helps solve certain problems arising in generative analysis: when a prepositional phrase is realized in a “dative” construction (cf. 10a, 23a), 16 The differing treatment of locative and “dative” alternations seems to have been due in considerable measure to the different positions of their constructions with respect to the aspect of paraphrasability; for a fuller discussion, see section 1.1. 83 KRISTINA LENARTAITĖ And prepositions, it is claimed, are lexicalized in deep structure and do not allow their governed nouns to be moved by transformation into the direct-object position. This seems to contradict the transformational origin of the double-object construction, since here it is precisely that noun which is realized in the direct-object position (cf. 10b, 23b) and which, in the corresponding prepositional “dative” construction (10a, 23a), is expressed as the prepositional complement (its dependent). This problem can, however, be resolved if the derivation of the double-object construction incorporates a null predicate affix that replaces the directional preposition in the “dative” construction and bears the semantic feature PATH17, while the corresponding transformation in the double-object construction raises the indirect object of deep structure (the prepositional complement in the “dative” construction) to the direct-object position. Moreover, the applicative effect of the null affix can be compared with “conversion” in derivational morphology, whereby, for example, in English lexical adjectives are likewise “converted” into verbs through the application of a null affix; cf. brown ‘brown’ and brown ‘make brown’. In this way, the problem of the non-productivity of “dative” alternations is shifted to the lexical level (see Emonds & Whitney 2006: 73–144 for a fuller discussion). It is not difficult to imagine how this interpretation could be applied to other types of alternations, such as locative alternations. At this point, it would also be useful to discuss a somewhat different phenomenon, likewise associated with a change in the status of prepositional phrases in alternating constructions—atypical passive constructions, or prepositional passive constructions (prepositional / non-canonical / peculiar passive / pseudopassive), cf.: (24) a. This spoon has been eaten with. ‘This spoon was eaten with’ b. This pub hasn’t been smoked hash in before. ‘Hashish had not been smoked in this tavern before’ (Examples from Kageyama 2006: 88–90.) c. This pot has been cooked in (by generations of Smiths). ‘This pot was boiled in (by [entire] generations of the Smiths)’ d. This bed hasn’t been slept in (by anyone). ‘This bed had not been slept in (by anyone)’ (Examples from Anderson 1997: 202.) 17 There is no consensus on the precise position of the null affix in configurational structure, but it is agreed that it should be directly dominated by the V node (Emonds & Whitney 2006: 73–144). 84 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (25) a. Urvai yra žvėrių gyvenami. (Example from Ambrazas, ed., 1997: 322.) b. This room was inhabited by him. c. Šitas šaukštas tavo valgomas. d. This road is no longer traveled (by people) now. e. The puppy is still not given anything today. (Examples from Jablonskis 1997: 132, 140.) These examples from English and Lithuanian show that, in atypical cases, recipient/beneficiary (25e), instrumental (24a, 25c), or locative (24b–d, 25abd) oblique terms can be realized as the subjects of passive constructions. At first glance, it may seem that the subject NPs of English passive constructions (24) (cf. this spoon, this pot) correspond to prepositional phrases in active constructions (for example, with this spoon, in this pot), and that the preposition should prevent its complement from being moved into subject position by a transformational process (as in the case of the “dative” alternations discussed above); furthermore, an NP (noun phrase) can be moved into the subject position of a passive construction by transformation only from the first postverbal NP position (which is typically occupied by the direct object). But this is not so: the complement of the preposition functions as the subject in the corresponding passive construction, and, moreover, the preposition in such a passive construction usually does not disappear. These cases of passivization can be explained by appealing to a reanalysis rule: in deep syntactic structure, after the lexicalization process (that is, after both the preposition and its complement have also been lexicalized), the reanalysis rule allows the relationship between a simple verb and its indirect complement (oblique term), marked by a prepositional phrase, to be reinterpreted. For example, the constituent structure that was [V eat] [PP with [NP spoon]]] before reanalysis becomes [V eat with] [NP spoon] after reanalysis (Radford 2002: 427–432). Thus, after such reanalysis, a simple verb becomes a complex verb, and there is no longer any obstacle to the noun phrase in the first postverbal position being moved into subject position by transformation. But how should the English sentence (24b) be explained, where the noun pub is moved into subject position not from the first postverbal position (where the direct object hash is realized even in the passive construction), but from an oblique position, and how should analogous Lithuanian constructions be interpreted? In Lithuanian, all subjects of passive sentences in the corresponding active sentences 85 KRISTINA LENARTAITĖ In these sentences, they would be marked with oblique cases (locative – (25ab), instrumental — (25cd), dative — (25e)). In generative grammar, oblique cases can be interpreted as prepositional phrases with a null preposition (Emonds & Whitney 2006: 78); thus, it seems that an analysis of Lithuanian atypical passive constructions would not be possible without null affixes, which would have to be postulated in the predicate structure of the passive construction itself. It is also worth noting that in the English sentences discussed (with the exception of sentence (24b)), after reanalysis, the noun phrase formerly governed by a preposition is newly interpreted not only as an independent NP occupying the first postverbal position, but also as a direct object (which is generally identified in relation to the position it occupies). In the case of Lithuanian sentences, it is more difficult to demonstrate that the subject noun phrases of atypical passive sentences, before being moved into this position, would have been interpreted as direct objects¹⁸. Here it is also worth recalling that argument-expression alternations which allow variation in the expression of the direct object accordingly realize different arguments as subjects in passive constructions as well, cf.: (26)(= 4) a. Mother puts a hat on the child. b. Mother hats the child. (27) a. The hat is put on the child by (the mother). b. The child is hatted by (the mother). The passive counterpart of sentence (26a) is (27a), and that of sentence (26b) is (27b). The possibility of realizing both terms (both the hat and the child) as subjects of passive sentences merely confirms that they are direct objects; however, another conclusion could also follow from this—that the subjects of passive sentences must be the direct objects of the corresponding active constructions¹⁹ (regardless of whether this status was present from the outset... 18 Although in Lithuanian it would theoretically be possible, for example, to coin for sentence (25a) the lexeme gyventi-kame ‘live-in-where’, which governs a direct object that is promoted to subject position in a passive construction, this would nevertheless not be a direct object in the usual sense of the term (Holvoet 2009: 14). On the transitivity of Lithuanian predicates that allow atypical passive constructions, and on the assumption that the subjects of such passive sentences should be understood as direct objects before being realized in this position, cf. Lenartaitė (2009: 80). 19 The passive variants of English “dative” alternation constructions are also different; compare sentences (23). Generative grammar takes the position that passivization of the indirect object (not marked by any dative case) is possible 86 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos initially, or only after promotion). But how, in this case, can sentences (25e) and (24b) be explained, where the deep-structure indirect (recipient/benefactive) object (šunelis) and an oblique (locative) (in this pub) are realized, respectively, in subject position, but the terms nieko, hash? are also expressed in direct-object position in passive constructions? It seems that when the corresponding term is removed from the prominent (subject) position, not only the one realized in direct-object position, but also other terms may be eligible for it (cf. also footnote 10). We have seen that in both models discussed, the diathesis alternation phenomena of interest to us are explained as surface structures (or, in the case of Relational Grammar, new strata relative to the initial one) derived from a single underlying syntactic structure (in Generative Grammar—from deep structure, or base; in Relational Grammar—from the initial stratum (initial stratum)). In Generative Grammar, one of the major advantages often attributed to the transformational analysis of passivization and other alternations is that selectional restrictions for individual predicates need to be specified only once—in deep structure. If alternating sentences were not generated on the basis of a single base, separate restrictions would have to be specified for active/passive and other alternating sentences. However, derivation in these models, as well as in the “lexical rules” models to be discussed shortly, faces a problem: while the direction of derivation is uncontroversial for active/passive and applicative alternations, other cases of argument-expression alternation are sometimes less clear-cut in this respect. Lexical rules, as an alternative to transformational rules, establish relationships among individual lexical items; that is, these rules determine how a direct object is formed by moving the relevant noun phrase (NP) only from the position to which it has already been promoted1 (Emonds & Whitney 2006: 96–98). Regarding other factors underlying the emergence of atypical passive constructions, such as the semantic nature (internal structure) of their predicates, it is observed that the predicates of the active counterparts of such passive constructions are usually intransitive; moreover, it is assumed that only unergative verbs (i.e. intransitive verbs whose surface-structure subject is also the deep-structure subject, for example, mąstyti, kalbėti, atsistoti, etc.), but not unaccusative verbs (i.e. intransitive verbs whose surface-structure subject is the deep-structure object, for example, skęsti, degti, mirti, etc.), allow such passive constructions (cf. Blake 1990: 29–34). However, not all unergative verbs in English allow prepositional passives (Levin 1993: 86). 87 KRISTINA LENARTAITĖ words of different classes (for example, verbs and their nominalizations) may be related to one another at the phonological, syntactic, and semantic levels. There are two possible strategies for applying lexical rules, according to which the so-called “lexical rules” models can be classified: (i) lexical rules may link lexical items that have one and the same lexical-semantic representation but different syntactic realizations (for example, as in Lexical-Functional Grammar); (ii) lexical rules may define a relation between lexical items that have different but systematically related lexical-semantic representations (for example, in the work of Levin and Rappaport Hovav and of Pinker). In Lexical-Functional Grammar (Lexical-Functional Grammar), the argument structure of each predicate represents the semantic functions of its possible n arguments that can be interpreted grammatically, and a grammatical function is assigned to the lexical form of each predicate argument. The semantic functions represented by a predicate must be compatible with all the semantic interpretations of the subjects, objects, and other grammatical arguments associated with that predicate. This is a necessary semantic compatibility requirement, which in Lexical-Functional Grammar replaces the selectional restrictions of Generative Grammar (see above). Let us now consider how this model treats the phenomena of interest to us. For example, in typical cases, active/passive alternations retain the same predicate argument structure (and thus the same selectional restrictions); the only difference is that, compared with the active predicate, the grammatical functions assigned to the arguments of the passive predicate change—passivization redistributes the grammatical functions associated with the same arguments. Thus, the requirements of semantic compatibility are not violated. Lexical-Functional Grammar applies selectional restrictions to the semantics of predicate argument structure, whereas Generative Grammar treats these restrictions as a set of semantic features defined as relevant to the grammar (for example, [abstract], [animate], etc.), on the basis of which the restrictions are established. Furthermore, the Functional Completeness (Functional completeness) principle requires every predicate argument to be linked to a functional argument of the type specified by the lexical form of the verb (i.e. a grammatical function); for example, in English, the sentence *John admired ‘John admires’ is impossible because the functional argument linked to the second argument of the lexical form ‘ADMIRE ((SUBJ), (OBJ))’—the object—is not expressed. In addition, the Functional Cohe- (Functional cohe- 88 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos rence) principle requires every meaningful functional argument of the type specified by the verb’s lexical form to be obligatorily linked to a predicate argument. For example, the sentence Fred sent John Mary ‘Fred sent Mary to John’ is functionally coherent because all three functional arguments—the subject Fred, the object John, and the second object Mary—in the lexical form denoting ‘SEND ((SUBJ), (OBJ), (OBJ2))’ can be linked to the corresponding predicate arguments (Bresnan 1982a: 39–49). Thus, in Lexical-Functional Grammar, active/passive alternations are regarded as two different syntactic projections of a single lexical-semantic representation, while the corresponding active and passive verbs are regarded as separate lexical items. Moreover, a passive sentence can be generated directly, independently of the corresponding active sentence, provided that the relevant passive verb exists in the lexicon. The functions of the passivization rule are to derive a passive verb from an active verb and, at the same time, to express the systematic relationship between these two lexical items. Schematically, the passivization rule for typical cases can be represented as follows: (28) (SUBJ) → (OBLagt) (OBJ) → (SUBJ). This rule is to be interpreted as follows: each reference to the object (OBJ) relation in the base lexical item is replaced in the derived lexical item by a reference to the subject (SUBJ) relation, while each reference to the subject (SUBJ) relation in the base lexical item is simultaneously replaced by a reference to the oblique (OBLagt) relation (Bresnan 1982a: 2–86; Kroeger 2004: 53–63). “Dative” alternations are explained analogously in this model: two different sets of grammatical functions are assigned to a single predicate argument structure whose semantic roles are agent, theme, and goal (goal), cf.: (29) a. Fred handed a toy to the baby. ‘Fred handed a toy to the baby’ b. Fred handed the baby a toy. (Examples from Bresnan 1982b:150–151.) ‘Fred handed the baby a toy’ c. ‚HAND ((SUBJ), (OBJ), (TO OBJ))‘ d. ‚HAND ((SUBJ), (OBJ), (OBJ2))‘. 89 KRISTINA LENARTAITĖ Here (29c) is a representation of the grammatical functions of sentence (29a), and (29d) of (29b). Thus, in Lexical-Functional Grammar, the systematic relationship between the NP_to-NP (29a) and NP–NP (29b) constructions is established by the corresponding lexical rule, cf.: (30) (OBJ) → (OBJ220) (TO OBJ) → (OBJ). This rule stipulates that a lexical unit with (OBJ) and (TO OBJ) (i.e., specifications for an object governed by the preposition to), as in sentence (29a), can be replaced by another lexical unit in which (OBJ) is replaced by (OBJ2), and (TO OBJ) is replaced by (OBJ) (Kaplan & Bresnan 1982: 173–180). Applicative alternations are explained analogously in this model (see Section 1.3.): the corresponding lexical rules ([X]v → [X-APPL]v; (OBL) → (OBJ); (OBJ) → (OBJ2)) establish the systematic relationship between a non-applicative verb and the corresponding applicative verb, and the grammatical functions they specify (cf. Kroeger 2004: 68–69). To summarize: in Lexical-Functional Grammar, diathesis alternations are recognized as having separate, single-layer syntactic representations; however, the predicates of alternating constructions are indirectly related by a “kinship” relationship defined by the corresponding lexical rules. Thus, in this model, the derivational relationship between alternating constructions can be established only indirectly, that is, through the derivational relationship linking their predicates. Another variant of the so-called “lexical rule” models, in which lexical rules define the relationship between lexical units with different but systematically related lexical-semantic representations, includes the studies of argument expression alternations by Levin and Rappaport Hovav, as well as Pinker. Levin and Rappaport Hovav investigate cases of locative alternations. Their research is based on the assumption that alternating constructions reflecting systematic variation in meaning are realizations of different lexical-semantic representations, and that the corresponding lexical rules govern the mapping of a verb’s different lexical-semantic representations onto the correct argument realization patterns, cf.: 20 In ditransitive constructions in English and other languages, the primary object is the recipient/beneficiary object (as indicated by passivization effects, a strictly fixed position in the sentence, and other features), whereas the object with the semantic role of Theme in such sentences is considered the second object; in Lexical-Functional Grammar and other models, it is marked (OBJ2). 90 Diateses alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (31) a. They loaded boxes onto the truck. “They (un-, loaded) boxes onto the truck” b. They loaded the truck with boxes. “They loaded the truck with boxes” c. Load: [x cause [y to come to be at z] / LOAD] “Load: [x causes [y to be on z] / LOAD]” b. Load: [[x cause [z to come to be in state]]] BY MEANS OF [x cause [y to come to be at z] / LOAD] “Load: [[x causes [z to undergo a state]]] IN SUCH A WAY THAT [x causes [y to be on z] / LOAD]” Here, sentence (31a) is associated with the semantic representation of the verb load (31c), while sentence (31b) is associated with (31d). It should be noted that only semantic representation (31d) inherits and incorporates semantic representation (31c), but in no way vice versa. Since in the semantic representation (31d) the location (z) is the first argument undergoing change, it is considered eligible to be linked to the syntactic object position in the syntactic argument realization model (31b). In semantic representation (31c), the theme (y) (the transported load) is the first argument undergoing a change (of location), and is therefore linked to the object position in (31a) (Levin & Rappaport Hovav 2005: 206–208). The main shortcoming of this account is identified as the unidirectionality of the derivation; as a result, the model cannot explain why some semantically similar verbs allow only one argument realization pattern. Moreover, the implication of one semantic representation in another seems to be based primarily on the supposed precedence of one event over another; that is, it seems to us that the loading event precedes the filling event, since when something is loaded into some container (for example, a wall into a cart), the things being loaded are usually carried (or transported in some other way) into that container in certain quantities, and it takes some time for it finally to become full, that is, loaded up—loading takes a while and seems to separate these closely related events. To a large extent, this also underlies the intuition that such two “successive”21 events are connected by a causal relation (without the loading activity, which takes that 21 Although we could reasonably think that in such cases the result of the action (a cart full of hay) was brought about by the loading action performed, on closer inspection both the loading (of the hay) and the filling of the cart took place in parallel; that is, both events were simultaneous right up to the end. 91 KRISTINA LENARTAITĖ length of time, the container will not be filled). However, this explanation is harder to apply to analogous cases where the simultaneity of events is more obvious, and consequently the causal relation linking the events does not stand out (cf. Iwata 2005: 375–376). In Pinker’s model, a corresponding lexical rule likewise regulates the conversion of a verb whose semantic structure core (core) is ‘x causes y to move onto / into z’ into a new verb whose semantic structure core is ‘x causes z to change state in such a way that y moves into / onto it’ (Pinker22 1989: 79). Compared with the account discussed above, what is new here is that, in the case of locative alternations, derivation can proceed in both directions: from the locative variant (31a) to the with variant (31b), and from the with variant to the locative one (this is a case of “inverse subordination”). The direction of derivation is determined by which argument cannot be omitted—that is, which argument can occur as the sole object complement: if it is the theme, derivation proceeds from the locative variant, whereas if it is the goal (goal), it proceeds from the with variant. However, this test proves unreliable, since there are cases in which complements in alternating sentences can both occupy the same syntactic position and both can be omitted (or neither can). According to Pinker, cases such as alternations with the verb stuff ‘stuff’ indicate derivation from the with variant (32b) to the locative variant (32a), cf.: (32) a. He stuffed the breadcrumbs into the turkey. “He stuffed breadcrumbs into the turkeyᵉ b. He stuffed the turkey with breadcrumbs. “He stuffed the turkey with breadcrumbsᵉ c. *He stuffed the breadcrumbs. d. He stuffed the turkey. Sentences (32c) and (32d) show that only turkey can occur as the sole object complement of the verb stuff, since sentence (32c) is ungrammatical; therefore, the underlying form is (32b). Here the semantic representation of the verb has a more complex event structure (‘x causes z to change state in such a way that y moves into / onto it (z)’), and when the corresponding rule deletes the main component (‘x causes z to change state in such a way that y moves into / onto it (z)’), the subordinate component can be promoted to a main component (Iwata 2005: 22 Referuojama iš Iwata (2005: 358). 92 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos 358–375). In Pinker’s model, thematic cores (thematic cores) are regarded as the “intermediaries” between form and meaning, that is, schematic representations of an event type or relation that constitutes the semantic core of the relevant verb class. Thematic cores such as “x affects y” or “x causes y to have z” are associated with the corresponding syntactic frames [NP V NP], [NP V NP NP]. Thus, a verb occurs in a given syntactic frame if its thematic core is associated with that syntactic frame (ibid., 363). Furthermore, Pinker classified English locative verbs according to whether they allow alternations: some, “content-oriented” (content-oriented) verbs necessarily express an argument denoting the relevant “thing put” (locatum), whereas others, “container-oriented” (container-oriented) verbs, necessarily express an argument denoting a location (location). According to Pinker, action types whose arguments bearing the semantic role of goal (goal) can more readily be construed as undergoing a change are more likely to realize their goal arguments as patients, and can therefore participate in alternations (cf. Arad 2006: 474–477). To summarize, it is worth noting that in both the models of “lexical rules” proposed by Levin and Rappaport Hovav and by Pinker, two different semantic representations of a single verb, linked indirectly by a derivational relationship, are associated with different argument realizations. Thus, although no direct derivational relationship between the two realizations can be established here, as in the case of lexical-functional grammar, the relationship linking the relevant lexical-semantic representations can nevertheless be traced indirectly, through lexical rules. The essential difference between the generative and relational grammar models discussed here and the “lexical rules” models is that the former explain the differences between alternating constructions at the syntactic level, whereas the latter, by contrast, explain these differences as resulting from processes that take place at the semantic level and are subsequently associated with different syntactic representations. Thus, with respect to syntactic representations, “lexical rules” models are monostratal, whereas generative and relational grammar models are multistratal. 2.2. Constructional models Constructional models can immediately be described as usage-oriented, that is, usage-based (usage-based), ling- 93 KRISTINA LENARTAITĖ uistic models that recognize gradual processes of language acquisition, which largely proceed by way of inductive cognition (i.e., from individual instances to generalizations), and that foreground generalizations as the primary mechanism driving various linguistic processes. Moreover, these models do not recognize a process important in other models, such as the derivation of some linguistic structures from others, because, unlike those discussed so far, constructional models regard constructions established in language use as the fundamental linguistic unit, and these cannot be further decomposed into any atomic primitives. Structures recognized as constructions—learned, concrete linguistic units with a characteristic form and meaning or discourse function (constructions here also include morphemes, words, and idioms)—are those whose formal or semantic aspects cannot be fully predicted from their individual components or from other constructions in the language. Furthermore, even a particular instance of a form–meaning pairing that is fully predictable may be recognized as a construction if it is used frequently enough (Goldberg 2006: 5). In the remainder of this section, we will discuss phenomena of diathesis alternations in relation to Goldberg’s Construction Grammar and Iwata’s lexical-constructional model. In Goldberg’s Construction Grammar, the following English argument structure constructions, for example, are distinguished as basic means of sentence expression: (33) a. Ditransitive construction “x causes y to receive z” Subj V Obj Obj2 Pat faxed Bill the letter. “Pat sent Bill the letter by fax” b. Caused-motion construction “x causes y to move to z” Subj V Obj Obl Pat sneezed the napkin off the table. “lit. Pat sneezed the napkin off the table” (Goldberg 1995: 3–4). Such argument structure constructions are associated more with a group of related meanings than with a single abstract meaning, although such a meaning could be identified in the inherited construction hierarchy (Goldberg 1995: 108–110). In Construction Grammar, as in other varieties of constructional models, constructions largely determine differences in sentence meaning, and 94 Dia tezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos The verb23 is only semantically integrated into the construction. Constructions exist independently of the verb, have their own meaning and specific syntactic structure. For example, in the caused-motion construction (caused motion), the argument roles x causes y to move to z—cause (cause), goal (goal), and theme (theme)—are directly associated with the construction, whereas the participant roles of the verb PUT—putter, thing put, and location where it is put—are associated with the verb. “Fusing,” or integrating, the argument roles with the action’s participant roles yields a composite syntactic-semantic structure; cf.: (cf. Goldberg 1995: 52) The process of fusing a construction’s argument roles with the verb’s participant roles is governed by the following principles: (i) the Semantic Coherence Principle (Semantic Coherence principle): only semantically compatible roles can be “fused”; (ii) the Correspondence Principle (Correspondence Principle): every participant role that is lexically profiled and expressed must be “fused” with the argument roles profiled by the construction. If a verb has three participant roles, one of them may be “fused” with a role that is not profiled in the construction’s argument structure (Goldberg 1995: 50). It should be noted here that lexically profiled participant roles are those that must necessarily be expressed in the sentence, whereas constructionally profiled argument roles (for example, cause (agent), goal, theme) are associated with the main syntactic positions—the subject and direct object. Cases of alternation in Construction Grammar are explained as involving one and the same verb meaning (i.e., an action understood on the basis of the corresponding semantic-interpretive frame and the participant roles it entails) 23 In constructional models, verb meaning is necessarily linked to the semantics of the corresponding semantic-interpretive frame, on the basis of which the constraints on the verb’s participant roles are determined (see Goldberg 1995: 25–31; 43–44). 95 KRISTINA LENARTAITĖ the ability to integrate with two or more different argument-structure constructions. For example, in the case of English locative alternations, also known as spray / load alternations, spray / load alternations (cf. sentences (31a) and (31b)), the same verb load can be integrated with the caused-motion construction (31a) or with the causative + prepositional with construction24; in other words, the sentences in question should be viewed as instantiations of different constructions: sentence (31a) instantiates the caused-motion construction, whereas sentence (31b) instantiates the causative + prepositional with construction. In this case, differences in meaning between the alternating sentences are determined by differences in the meanings of the constructions (cf. Goldberg 1995: 175–179; Goldberg 2006: 34–38). Active/passive alternations are likewise explained here as instantiations of different constructions. The purpose of the passive construction is to shade (shading) the highest-ranking participant role associated with the verb. Thus, a particular passive sentence should be viewed as an instantiation of a construction that shades the relevant participant role, i.e., the passive construction (Goldberg 1995: 56–57). The fact that such alternations as passivization or applicative constructions are generally accompanied by morphological derivation is explained in Construction Grammar as follows: the closed classes of grammatical morphemes associated with these phenomena, such as verbal passivization markers or applicative suffixes, are treated as argument-structure constructions25; therefore, the semantic integration of the relevant morpheme with the verb is analogous to the integration of a construction with a verb (Goldberg 1995: 22–23). It is interesting to consider how the Chichewa applicative constructions discussed in section 1.3 would be treated in this model: after all, integrating verbs with the same closed class of grammatical morphemes—more precisely, the applicative suffix (-ir- ~ -er-)—here yields entirely different constructions, for example, a ditransitive construction (19b) and a causative construction (20b). The rejection in Construction Grammar of the derivational relation that links the phenomena under discussion in one way or another (at the syntactic or semantic level) in other models is justified as follows: (i) it is better to generate A directly rather than 24 Such an argument-structure construction should be viewed as complex, i.e., as consisting of two independent constructions: a causative construction and a construction governed by the preposition with (Goldberg 2006: 36). 25 In constructional models, as in Cognitive Grammar, there is no strict distinction between lexical and grammatical units. 96 Di atezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos derive from C if there exists a model B with the same target syntactic properties as A, but which is clearly not derived from C; (ii) it is better not to derive A from C if there exists a model B with the same syntactic properties as C, but which is not an input from which A can be derived. (Goldberg 2006: 22–25). In this way, the problem faced by other models that recognize derivation at one level or another is eliminated: argument-structure constructions in this model function as surface generalizations surface generalizations with respect to their particular instantiations. A shortcoming of Goldberg’s Construction Grammar is said to be that it selects higher-level abstract constructions as the primary linguistic units, although child-language acquisition and other facts indicate that, in this respect, lower-level constructions associated with specific verb classes function as the primary units. Higher-level abstract constructions cannot exist without lower-level constructions, and therefore the former cannot be considered more adequate than the latter with respect to psychological reality26 (Iwata 2008: 123–129). Another shortcoming noted concerns the operation of the Correspondence Principle discussed above: first, as the omission test shows, identifying the roles of lexically profiled participants in an action—that is, the roles of participants that must be expressed—is problematic here; moreover, according to Iwata, this principle partly duplicates the Semantic Coherence Principle, and it could generally be dispensed with (that is, the Correspondence Principle could be dispensed with) (Iwata 2005: 382–387). It might also seem that information conveyed at different levels in Construction Grammar is repeated and redundant; for example, participant roles associated with the verb seem to provide sufficient information about the syntactic projections of arguments, yet all this is also repeated at the constructional level. However, this redundant information constitutes the essence of the instantiation mechanism, which is highly significant in Cognitive and Construction Grammars. The purpose (functional motivation) of more abstract constructions is to sanction lower-level, more concrete constructions, such as sentences, including newly formed ones. In this way, constructions perform the same function as rules in generative models. 26 For constructional models, as for usage-based usage-based models, it is very important that the grammatical units and claims they recognize be adequate with respect to other psychological and cognitive processes as well; that is, that they be psychologically real for the speaker / listener. 97 KRISTINA LENARTAITĖ In Iwata’s lexical-constructional approach lexical-constructional approach, although it essentially follows Goldberg’s Construction Grammar, all its main aspects are modified somewhat: (i) Construction Grammar selects higher-level abstract constructions as the principal units of grammatical description, whereas the lexical-constructional approach selects as primary lower-level constructions associated with a specific verb class; (ii) in Construction Grammar, verb meaning can be reduced to participant roles in an action, that is, the participant roles in the action denoted by the verb are sufficient to represent its meaning, whereas in lexical-constructional grammar verb meaning is not limited to participant roles but is understood more broadly; (iii) in Construction Grammar, the relationship between a verb and a construction is determined by how the meaning of a particular verb is compatible with the meaning of the construction, whereas in the lexical-constructional approach the relevant question is how an entire utterance—for example, He sprayed the wall with paint ‘He sprayed the wall with paint’—is compatible with a construction associated with a specific verb class (cf. Iwata 2008: 123). The advantage of applying lower-level constructions has already been mentioned above, while the fact that parameters traditionally associated with a verb lexeme (for example, argument structure) cannot adequately reflect compatibility with a construction is demonstrated by cases in which this compatibility is clearly determined not only by the verb but also by other sentence elements; that is, a verb can occur in a particular construction with some terms but not with others, cf.: (34) a. *He spread the sleeping child with a blanket. ‘He covered the sleeping child with a blanket’ b. He spread the bread with butter. ‘He spread butter on the bread’ (Examples from Iwata 2008: 124.) Both of Iwata’s works discussed in this article (Iwata 2005; 2008) address problems concerning locative alternations, and the main questions raised with respect to these phenomena are: why do some verbs allow locative alternations while others do not? This question also gives rise to others: which aspects of the semantics of verbs participating in locative alternations determine this possibility, and why? Thus, it is clear that Iwata’s research is more oriented toward analyzing the semantics of locative alternations. In greater detail, 98 Diატезės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos having examined in detail Pinker’s and others’ claim that the so-called with variant of a locative-alternation pair (cf. (31b)), in which the constituent functioning as the location—more briefly, the locative—appears as the object, can be characterized as denoting a ‘(locative) change of state’ change of state with respect to the other alternating sentence, which denotes a ‘(theme) change of location’ (the onto / into variant, cf. (31a)), Iwata concluded that the ‘(locative) change of state’ characterization, both because it is overly general and because of the marked semantic differences between verbs that allow locative alternations and prototypical ‘change of state’ verbs such as break ‘break’, cannot adequately describe the semantic aspects of the phenomenon under study (Iwata 2008: 22–25). Consequently, in the lexical-constructional approach he proposes, Iwata maintains that only lower-level constructions, that is, constructions associated with a specific verb class, can adequately reflect the semantic and syntactic processes of locative alternations. The with variant of locative alternations (31b), in which the locative is the object, can be described as realizing the semantics of the ‘cover’ verb class (and, in some cases, the ‘fill’ verb class), whereas the into / onto variant (31a), in which the object that changes location (locatum) is the object, realizes the semantics of the ‘put’ verb class. A construction associated with a specific verb class should be regarded as a sanctioning schema27 with respect to the particular realizations that instantiate it, cf.: (35) a. The vandal sprayed paint onto the sculpture. ‘The vandal sprayed paint onto the sculpture’ b. The vandal sprayed the sculpture with paint. ‘The vandal sprayed the sculpture with paint’ (Example from Iwata 2008: 48.) 27 In usage-based theoretical models, constructions are nothing other than schematic form–meaning pairings abstracted from usage events (sentences). Constructions, understood as schemas established in usage that encompass commonalities among their specific realizations, are embodied in those realizations and sanction new realizations that match their specifications (Iwata 2008: 38). Both the specific meaning of the verb participating in alternations and the meaning of the construction are to be regarded as abstractions of a kind with respect to particular usage events. Furthermore, verb meaning is difficult to separate from construction meaning; that is, they overlap. Thus, the meaning of a particular verb may be conceived of as the minimum amount of information associated with that verb (ibid., 87–99). 99 KRISTINA LENARTAITĖ Sentence (35a) is regarded as an instantiation of a verb-class-level construction containing the semantic component “put,” while sentence (35b) is a specific realization of a verb-class-level construction containing the semantic component “cover” (ibid., 27–51). Alternations occur when the semantics of the same verb (the limits of its specifications are defined by the relevant semantic-interpretive frame) can be compatible (integrated) with two different constructions. Other significant semantic properties of verbs that permit locative alternations are the holistic nature of the action and its telicity (i.e., the presupposition of an endpoint to the action). In sentences such as (31b) and (35b), where the locative occurs as the object, the verb classes with the semantics of “cover” (35b) or “fill” (31b) that sanction these sentences entail the emergence of an additional semantic component—the holistic effect—that is, the perception that the object undergoing the action is completely affected by it. However, sentence (35b), for example, allows both a holistic and a non-holistic interpretation: with respect to a sculpture as a work of art, even the slightest streak of paint sprayed onto it may be regarded as a change in the sculpture’s condition, that is, its aesthetic appearance, and thus realization (35b) is possible. According to Iwata, however, only realizations with a holistic effect are sanctioned by constructions of verb classes with the semantics of “cover” / “fill”; thus, the same sentence (35b), when it allows a holistic interpretation, will be regarded as an instantiation of constructions of the aforementioned group, whereas when it does not, this means that it sanctions a construction belonging to some other verb class (for example, verbs of “disgust”). Moreover, if a particular sentence that selects a locative as its object does not allow a holistic interpretation, then, strictly speaking, it is not a genuine case of locative alternation, since it is not sanctioned by constructions of verb classes with the semantics of “cover” / “fill.” The reflex of the semantic component of telicity is explained similarly: if the action in sentences that select a locative as their object is telic (non-durative / bounded), they are sanctioned by constructions of verb classes with the semantics of “cover” / “fill,” whereas if it is atelic, they are sanctioned by a construction of some other verb class (ibid., 55–64). In examining cases of locative alternation in which one of the sentences is formed with a morphologically more complex verb, such as German locative alternations (cf. sentence (9)) or Japanese locative alternations, Iwata is inconsistent: sentences with German prepositional (be-) verbs realize the locative in object position 100 Di atezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (9b), in general, are not considered cases of locative alternation under strict scrutiny, because they are not sanctioned by constructions of verb classes with the semantics of “cover” / “fill,” allegedly because they lack a holistic interpretation (ibid., 2008: 152–156); whereas, for example, in Japanese locative-object sentences formed with the suffixal (-tsukusu) verb, since realizations with these verbs can be regarded as instantiations of constructions of verb classes with the semantics of “cover” / “fill,” and the suffix -tsukusu is not a grammatical morpheme, these cases of alternation are assessed in the same way as cases of English locative alternation (ibid., 2008: 188–192). The lexical-constructional model may be criticized on the following grounds: (i) apparently, any alternating sentences in which one has a clause constituent with the semantic role of locative (in the broad sense) in object position can be called “locative alternations,” yet Iwata tends to regard as locative alternations only those cases that are relevant to the semantic constraints he has established; (ii) it seems that the role of certain aspects of verb semantic structure—holisticity and telicity—is also overestimated here: holisticity is always relative—it is always assessed by the speaker—and the possibility of multiple (holistic / non-holistic) interpretations of sentences such as (35b) shows precisely that holisticity does not necessarily play as decisive a role in locative alternations as Iwata would like to ascribe to it (on this issue, cf. also Michaelis & Ruppenhofer 2001: 23–27). The holistic seme is, of course, characteristic of the great majority of locative alternations, but not all of them. Thus, although constructional models can undoubtedly explain argument alternation phenomena, the phenomena under discussion do not exist in these models as “alternations” in the literal sense of the term, because no distinctive relation connects them here. In constructional models, all realizations that can be formed with the same verb could generally be called alternations. 2.3. General Remarks We should now briefly comment on why all the theoretical models discussed here (in Sections 2.1 and 2.2) are not classified on the basis of derivation, that is, into derivational and non-derivational models. It has already been mentioned that, in generative and relational grammars and in “lexical rules” models, derivational processes 101 KRISTINA LENARTAITĖ take place at different levels—in the case of the former models, directly and in syntax, and in the case of the latter, indirectly (through lexical rules) and in semantics. Thus, calling all of them “derivational” is possible only with certain reservations. Constructional models, however, should not be called “non-derivational,” because the rejection of derivation is better understood here as a consequence arising from a more significant feature of these models—their orientation toward the speaker and instances of use. As we have seen, derivational approaches face numerous problems, such as the direction of derivation and, more generally, proving its existence in a particular case; hence, processes such as generalization seem more significant to proponents of constructional models, while also making it possible to assess grammatical processes in relation to other processes of human mental activity. We have already clarified the positions of the models discussed with respect to diathesis phenomena, but it remains necessary to emphasize briefly how these models stand in relation to one another. From the concise overview presented, it is of course difficult to identify common ground among generative grammar, relational grammar, and constructional models; they might meet somewhere at the level of the philosophical evaluation of linguistic phenomena, but not in grammatical description. “Lexical rules” models, however, can be compared with constructional models: both seek the causes of different realizations at the semantic level, and therefore recognize only single-layer syntactic realizations; both assign different meanings to different realizations (although “lexical rules” models associate those meanings with a particular verb, whereas constructional models associate them with the meanings of different constructions that sanction different realizations). Fundamentally, what distinguishes them is that in “lexical rules” models the different lexical-semantic representations of alternating constructions are systematically linked through lexical rules, whereas in constructional models the different syntactic representations are linked only by the same verb, whose role in constructional models is sharply reduced compared with all the others. 3. IN LIEU OF CONCLUSIONS The review in this article of theoretical models that have examined diathesis alternations makes clear that they all address the same or similar problems in different ways—each of the studies discussed (and many others not mentioned here), viewing the phenomena under investigation from a different perspective, criticizes the shortcomings of the others and attempts to remedy them. Thus, in closing, I will briefly discuss 102 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos the model I have chosen for studying diathesis alternations and attempt to justify this choice. To explain phenomena involving alternation in the expression of arguments, I apply Shaumyan’s (Shaumyan) Universal Applicative Grammar (Universal Applicative Grammar) (Shaumyan 2006; 1987), in which, as in lexical-functional grammar, for example, the relationship between alternating sentences is established at the semantic level: alternating constructions can be related only through the relation linking their predicates, which is defined by the relevant rules for transforming predicate structure. Thus, one of the first and most important tasks for the chosen model is to demonstrate that the predicate structure of one of the alternating constructions is to be regarded as basic relative to the predicate structure of the other construction. First, it is essential to note that the account in applicative grammar strictly distinguishes two levels of linguistic phenomena—the genotypic (genotype) and the phenotypic (phenotype). The genotypic level represents the essential universal properties of the world’s languages as systems of signs. This level includes functional units such as predicates and terms, as well as their abstract relations as operators, operands, or modifiers. Phenotypic grammar investigates phenomena at the genotypic level and operates with theoretical constructs formed on the basis of empirical data. The phenotypic level describes how phenomena at the genotypic level are marked by the morphological, phonological, and other means of a particular language (Shaumyan 1987: 97, 101, Desclés, Guentchéva & Shaumyan 1985: 5). Therefore, the analysis of diathesis alternations must also proceed in two stages: at the genotypic level—that is, universally—the rules and regularities governing these phenomena are defined, while at the phenotypic level, the means used in particular languages to express those genotypic rules and regularities are described. The predicate-structure changes that constitute the essence of diathesis alternations can be defined at the genotypic level only in terms of changes in the relations between predicates and their terms (since the morphological marking that accompanies these changes, as in passive constructions in various languages or, for example, applicative constructions in the Bantu language group, is already a phenotypic-level phenomenon). Because concepts defining the relations between predicates and their terms, such as “subject,” “direct object,” and “indirect object,” cannot be considered universal28, in order to define diathesis alternation and other pro- 28 For example, proponents of Role and Reference Grammar (Role and Reference Grammar) believe that syntactic functions can be recognized only in languages where there are 103 Kristina Lenartaitė cesses, these concepts are divided in applicative grammar into two classes of primitive concepts: a syntactic class—first, second, and third terms—and a semantic class—agent, patient, and non-[agent/patient]. The subject of accusative languages is the first term, which denotes an “agent” in transitive and ditransitive sentences and an “agent/patient” in intransitive sentences. The direct object of accusative languages is the second term, denoting a “patient” in transitive and ditransitive sentences. In ergative languages, the main grammatical functions are usually marked by the absolutive and ergative cases, respectively. The absolutive marks the first term, which has the role of “patient” in transitive and ditransitive constructions in ergative languages and the role of “agent/patient” in intransitive constructions. The ergative marks the second term, which has the role of “agent” in transitive and ditransitive constructions in ergative languages. The concept of “indirect object” is defined in the same way for both accusative and ergative languages; however, the semantic role of the third term is heterogeneous (“recipient,” “beneficiary,” “addressee,” etc.), so its meaning is defined as “non-[agent/patient].” Thus, the indirect object is the third term, which has the semantic role “non-[agent/patient]” in ditransitive constructions in both accusative and ergative languages (Desclés, Guentchéva & Shaumyan 1985: 3, 66–67, 79–80). Next, let us compare the following cases of diathesis alternation in Lithuanian: examples of their existence independently of semantic or pragmatic functions. And, on the basis of data from languages such as Acehnese (acehnese / achehnese / achinese, North Sumatra) or yumas (Papua New Guinea), where the distinctive argument function in a construction is apparently related to its semantic role (“actor,” “experiencer”) rather than to syntactic properties, proponents of this grammar regard the grammatical functions of subject and object as non-universal and do not use them in their account at all. Instead, they introduce generalized semantic macroroles (macroroles), which designate generalized argument types that play significant roles in grammatical structure—that is, much like arguments in logical structure, they are the first to be targeted by grammatical rules. The generalized agent-type role is called the “actor” (actor), and the patient-type role the “undergoer” (undergoer). In addition, this grammar distinguishes another level of syntactic-structure organization: the level of the “syntactic pivot” (pivot). The “syntactic pivot” functions as a controller of syntactic and semantic processes in a construction; for example, in English, agreement, the omission of a coreferential element, and other operations are controlled by a single type of noun phrase, called the subject in other models (Van Valin & LaPolla 1997: 139–141, 274–285; Foley 2007: 388–402). 104 Dia tezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos (36) a. The piglets swarmed over the pig. b. The pig was swarmed over by piglets. (37) a. The piglets swarmed over the pig. b. The pig swarmed with piglets. Sentences (36) are active/passive alternations, whereas sentences (37) are one of the cases of alternation in argument realization (cf. the alternations in subject and direct-object realization in Section 1.1). Sentences (36a) and (37a) are transitive, and here syntactic positions typically correlate with the corresponding semantic roles: in both sentences, the noun phrase piglets, realized in the position of the first term, or subject, bears the agent role; likewise, the noun phrase pig, realized in the position of the second term, or direct object, bears the patient role in both sentences. Sentences (36b) and (37b), by contrast, are intransitive, and here the noun phrase pig, realized in the position of the first term, or subject, is the patient, while the noun phrases bearing the agent role—piglets in (36b) and piglets in (37b)—are realized here in the corresponding oblique positions and perform the grammatical function of predicate modifiers. The predicate of an intransitive sentence can realize a noun phrase bearing the patient semantic role in subject position (see the definition above), but what is unusual here is that the noun phrases bearing the agent semantic role, which would normally be realized as subjects, are adjuncts in sentences (36b) and (37b). Let us consider another example: (38) a. The girl is scattering sand onto the path. b. The girl is scattering the path with sand. This is a case of the so-called locative alternations, and in sentence (38a) we see the typical distribution of the syntactic positions of terms and the semantic roles they bear: the agent (girl) is in subject position, the moving affected object, which can be understood in general as a subtype of the patient role (sand), is the direct object, and the locative (onto the path) is an adjunct. In sentence (38b), by contrast, we see the locative (path) realized in direct-object position, and the patient (sand) in an oblique (instrument) position. Thus, to sum up, in the case of diathesis alternations, one of the alternating sentences realizes the typical configuration of the syntactic positions of terms and the semantic roles they bear, whereas the other (or others) sen-29 29 Here, of course, difficulties may arise concerning the classification of semantic roles and the criteria for determining them in general. Moreover, according to Dowty, variation in argument realization, i.e. the dif- 105 KRISTINA LENARTAITĖ tences—the atypical one. Atypical realizations of terms reflect changes in the semantic structure of the predicate, which can be defined universally at the genotypic level by corresponding rules (or, in other words, constraints) for changing predicate structure; these are determined by, and defined through, changes in the syntactic status of the predicate’s terms (cf. Shaumyan 2006: 234–237). Consequently, only the typical distribution of syntactic positions and roles realized by a predicate can be regarded as direct³⁰ (for example, in sentences (36a), (37a), and (38a)); all others are secondary and are realized by a predicate whose semantic structure has changed (cf. sentences (36b), (37b), and (38b)) in relation to the corresponding predicate with an unchanged structure (for more on the possibilities of applying applicative grammar to explain argument alternation phenomena, see Lenartaitė 2007). To define the essential features of diathesis alternation phenomena universally, it is important to set aside the morphological-level phenomena of individual languages that accompany the processes under discussion. It is clear that neither Goldberg’s abstract argument-structure constructions, nor, still less, Iwata’s constructions associated with a particular verb class, nor, for example, generative grammar’s passivization transformations involving the NP movement operation (NP movement), which is oriented toward the linear arrangement of words in the sentence of a particular language, can universally characterize diathesis alternation phenomena. However, a comprehensive study of argument alternation phenomena in particular languages requires clarifying how rules and regularities formulated at the genotypic level accord with data at the phenotypic level. Of course, data from phenotypic-level studies of particular languages can also be compared, and typologically significant generalizations may even be found; however, these phenomena can be defined universally only strictly through the genotypic level the realization of arguments in subject or direct-object position, which is possible when neither argument has prototypical agent or prototypical patient properties (cf. Levin & Rappaport Hovav 2005: 203). Nevertheless, in all cases of diathesis alternation, corresponding semantic roles can undoubtedly be assigned to the terms of particular predicates (as in sentences (38ab), where the term sand is assigned the semantic role of a moving affected entity, i.e. a theme, which can generally be understood as a subtype of the patient role, while the term path is assigned the locative semantic role), even though they often lack all the properties characteristic of the prototype. These semantic roles remain unchanged despite changes in the verb’s diathesis. 30 For the relationship between different diatheses of the same verb, such as the basic diathesis (basic) and its modifications, see also Mel’čuk (1993: 9–10). 106 Dia tezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos theoretical constructs, as is also done in universal applicative grammar. ABBREVIATIONS APPL — applicative verbal suffix ASP — aspect DEAGENT — deagentive form N — neutrum, neuter gender OBJ — marker of agreement (gender) with the primary object PART — participium, participle PASS — passivum, passive voice PL — pluralis, plural PRAES — praesens, present tense PRAET — praeteritum, past tense SG — singularis, singular SUBJ — marker of agreement (gender) with the subject LITERATŪRA AMBRAZAS, V., red. 1997: Dabartinės lietuvių kalbos gramatika. Vilnius: Mokslo ir enciklopedijų leidybos institutas. ANDERSON, J., M. 1997: A Notional Theory of Syntactic Categories. Cambridge: Cambridge University Press. ARAD, M. 2006: The Spray-Load Alternation. In: EVERAERT & VAN RIEMSDIJK, eds., vol. 4, 2006, 466–478. BLAKE, B., J. 1990: Relational Grammar. London: Routledge. BRESNAN, J. 1982a: The passive in lexical theory. In: BRESNAN, ed., 1982, 3–86. BRESNAN, J. 1982b: Polyadicity. In: BRESNAN, ed., 1982, 149–172. BRESNAN, J., ed. 1982: The Mental Representation of Grammatical Relations. Cambridge, MA: MIT Press. 107 KRISTINA LENARTAITE COMRIE, B. & M. POLINSKY, eds. 1993: Causatives and Transitivity. Amsterdam–Philadelphia: John Benjamins. CHAFE [ČEJF], W., L. 2003: Značenje i struktura jazyka, Moskva: URSS. DESCLÉS, J.-P., Z. GUENTCHÉVA & S. SHAUMYAN 1985: Passivization in Applicative Grammar. Amsterdam–Philadelphia: John Benjamins Publishing Company. EMONDS, J. & R. WHITNEY 2006: Double Object Constructions. In: EVERAERT & VAN RIEMSDIJK, eds., vol. 2, 2006, 73–145. EVERAERT, M. & H. VAN RIEMSDIJK, eds., 2006: The Blackwell Companion to Syntax, vols. 1–5. Malden–Oxford–Victoria: Blackwell Publishing. FOLEY, W. 2007: A typology of information packaging in the clause. In: SHOPEN, ed., vol 1, 2007, 362–446. GIVÓN, T. 1979: On understanding grammar. New York: Academic Press. GOLDBERG, A., E. 1995: Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. GOLDBERG, A., E. 2006: Constructions at work: The nature of generalization in language. Oxford–New York: Oxford University Press. HOLVOET, A. 2003: Aktualiosios skaidos problemos. In: HOLVOET & JUDŽENTIS, red., 2003, 79–98. HOLVOET, A. 2009: Argumentų hierarchijos ir gramatinės funkcijos. In: HOLVOET & MIKULSKAS, red., 2009, 1–36. HOLVOET, A. 2009: Difuziniai subjektai ir objektai. In: HOLVOET & MIKULSKAS, red., 2009, 37–68. HOLVOET, A. & A. JUDŽENTIS, red. 2003: Sintaksinių ryšių tyrimai. Vilnius: Lietuvių kalbos institutas. (Lietuvių kalbos gramatikos darbai, 1.) HOLVOET, A. & R. MIKULSKAS, red. 2009: Gramatinių funkcijų prigimtis ir raiška. Vilnius: Vilniaus universitetas & Asociacija „Academia Salensis“. (Acta Salensia, 1.) HOLVOET, A. & L. SEMĖNIENĖ, 2004: Rūšies kategorija: mediumas ir pasyvas. In: HOLVOET & SEMĖNIENĖ, red., 2004, 35–60. 108 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos Holvoet, A. & L. Semėnienė, red. 2004: Gramatinių kategorijų tyrimai, Vilnius: Lietuvių kalbos institutas. (Lietuvių kalbos gramatikos darbai, 2.) Iwata, S. 2005: Locative alternation and two levels of verb meaning. Cognitive Linguistics 16–2, 355–407. Iwata, S. 2008: Locative Alternation: A Lexical-Constructional Approach. Amsterdam–Philadelphia: John Benjamins Publishing Company. Jablonskis, J. 1997: Lietuvių kalbos gramatika, Vilnius: Mokslo ir enciklopedijų leidykla. Jackendoff, R. 1990: Semantic structures. Cambridge MA: MIT Press. Kageyama, T. 2006: Property description as a voice phenomenon. In: Tsunoda & Kageyama, eds., 2006, 85–114. Kaplan, R., M. & J. Bresnan 1982: Lexical-Functional Grammar: A formal system for grammatical representation. In: Bresnan, ed., 1982, 173–281. Keenan, E. L. & M. S. Dryer 2007: Passive in the world’s languages. In: Shopen, ed., vol. 1, 2007, 325–361. Klaiman, M. H. 1991: Grammatical voice. Cambridge: Cambridge University Press. Kroeger, P. R. 2004: Analyzing Syntax: A Lexical-Functional Approach. Cambridge: Cambridge University Press. Kroeger, P. R. 2005: Analyzing Grammar: An Introduction. Cambridge: Cambridge University Press. Lenartaitė, K. 2007: Lietuvių kalbos argumentų raiškos alternacijos ir aplikacinės gramatikos priemonės joms aiškinti. Acta Linguistica Lithuanica 57, 17–44. Lenartaitė, K. 2009: Tiesioginiai objektai ir ditranzityvinių konstrukcijų klausimas lietuvių kalboje. In: Holvoet & Mikulskas, red., 2009, 69–98. Levin, B. 1993: English Verb Classes and Alternations: A Preliminary Investigation. Chicago–London: University of Chicago Press. Levin, B. & M. Rappaport Hovav 2005: Argument Realization. Cambridge: Cambridge University Press. 109 Kristina Lenartaite Malchukov, A., M. Haspelmath & B. Comrie 2007: Ditransitive constructions: a typological overview. http://email.eva.mpg.de/~haspelmt/Ditransitive.html Matthews, P., H. 2007: The Concise Oxford Dictionary of Linguistics. Oxford–New York: Oxford University Press. Mel’čuk, I. 1993: Voice: Toward a rigorous definition. In: Comrie & Polinsky, eds., 1993, 1–46. Michaelis, L. & J. Ruppenhofer 2001: Beyond Alternations: A Constructional Model of the German Applicative Pattern. Stanford: CSLI Publications. Perlmutter, D., M. & P. M. Postal 1986: Toward a universal characterization of passivization. In: Perlmutter, ed., 1986, 3–29. Perlmutter, D. M. & P. M. Postal 1986: Some proposed laws of basic clause structure. In: Perlmutter, ed., 1986, 81–128. Perlmutter, D., M., ed. 1986: Studies in Relational Grammar 1. Chicago–London: The University of Chicago Press. Pinker, S. 1989: Learnability and Cognition: The Acquisition of Argument Structure. Cambridge MA: MIT Press. Rappaport, M. & B. Levin 1988: What to do with theta-roles. In: W. Wilkins, ed., Syntax and Semantics 21: Thematic Relations. New York: Academic Press, 7–36. Rappaport, M. & B. Levin 2008: The English dative alternation: The case for verb sensitivity. Journal of Linguistics 44, 129–167. Radford, A. 2002: Transformational Grammar. Cambridge: Cambridge University Press. Shaumyan, S. 1987: A Semiotic Theory of Language. Bloomington–Indianapolis: Indiana University Press. Shaumyan, S. 2006: Signs, Mind and Reality: A Theory of Language as the Folk Model of the World. Amsterdam–Philadelphia: John Benjamins. Shopen, T., ed., 2007: Language Typology and Syntactic Description. Vol. 1: Clause Structure. Cambridge: Cambridge University Press. 110 Diatezės alternacijos iš skirtingų teorinių modelių tyrimų perspektyvos SILVERSTEIN, M. 1976: Hierarchy of features and ergativity. In: R. M. W. DIXON, ed., Grammatical Categories in Australian Languages. Canberra: Australian Institute of Aboriginal Studies, 112–171. TSUNODA, T. & T. KAGEYAMA, eds., 2006: Voice and Grammatical Relations. Amsterdam–Philadelphia: John Benjamins. VAN VALIN, R. D. & R. J. LAPOLLA 1997: Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Kristina Lenartaitė Lietuvių kalbos institutas P. Vileišio 5, LT-10308 Vilnius [email protected] 111