The relationship between verbs and constructions

Goldberg (1995: 24) explores the nature of the relationship between verbs and constructions by posing three questions which we discuss here in turn. We then consider a number of other issues relating to the relationship between verbs and the sentence-level (verb argument) constructions that they fill.

What is the nature of verb meaning?

 Goldberg argues in favour of a Frame Semantics view of verb meaning (e.g. Fillmore 1977, 1982). As we saw in Chapter 7, this account of word meaning holds that the rich and detailed meaning of individual words is under stood against the background of a particular conceptual frame (or domain, in Langacker’s terms). Goldberg argues that an account like this is necessary, among other reasons, for explaining the distribution of adverbial expressions. Consider the examples in (5).

Goldberg argues that a frame provides the basis of our understanding of the nature and manner of the motion involved, which explains why slowly can be felicitously applied to stagger but not to bound.

 What is the nature of constructional meaning?

 Within the speaker’s knowledge of language or constructicon, Goldberg argues that constructions form a network. Within this network, constructions have related and sometimes overlapping meanings. This means that constructions are not individually represented with unique fixed meanings, but that they interact with other constructions in a rather fluid network of relationships (section 20.1.4). This view predicts that constructions, just like words, will exhibit polysemy. Consider the examples in (6).

Observe that all the examples in (6) are instances of the ditransitive construction. While example (6a) implies SUCCESSFUL TRANSFER of a kiss to George, example (6b) only implies intended transfer (it’s possible that Lily will suffer a crisis of confidence and George will never see the jumper). In example (6c), it is also unclear whether George will ever receive the money, or indeed whether Lily even intends to repay it. According to Goldberg, SUCCESSFUL TRANSFER (6a) represents the central or prototypical sense of the ditransitive construction, while the other examples share aspects of the prototypical sense (TRANS FER) while departing from it in other respects (the TRANSFER may only be intended or potential). These examples also effectively illustrate the contribution of both the construction and the verb itself to the overall meaning of the sentence. While the construction determines what the possible meanings are (TRANSFER, successful or otherwise), the verb determines which of these possible meanings is realised. According to Goldberg, the central or prototypical sense associated with a construction is salient because it represents a basic aspect of human experience. She captures this view by positing the scene encoding hypothesis:

According to this view, a basic ‘scene’ of experience involves TRANSFER of an enitity from one person to another. This is a scene that we participate in and witness scores of times every day, which therefore represents a basic and fundamental aspect of human experience.

 When can a given verb occur in a given construction?

In explaining what governs the interaction of particular verbs with particular constructions, Goldberg argues that while verbs are associated with participant roles, constructions have argument roles. In other words, the frame semantics of a given verb means that it is associated with frame-specific participants. For example, the verb buy might be associated with the participant roles BUYER, SELLER and GOODS, while the verb sing might be associated with the participant roles SINGER and SONG. As these examples illustrate, participant roles are associated with rather specific meanings that are related to their underlying frame or domain of experience. Furthermore, Goldberg adopts Langacker’s (1987) view that a particular verb profile particular participants within the frame or conceptual domain that underlies the meaning of that verb. Recall from Chapter 7 the distinction in profiling between buy and sell, for example. Goldberg discusses a similar example, which we saw in Chapter 5, comparing the verbs rob and steal. Consider the examples in (7) and (8).

While rob obligatorily profiles THIEF (George) and TARGET (Lily), steal obligatorily profiles THIEF (George) and (metaphorical) GOODS (hope). While either verb may optionally represent the third participant as a peripheral prepositional phrase (7a; 8a), the sentences become ungrammatical if this optional participant is represented as the direct object (7b; 8b). Goldberg (1995: 45) rep resents the profiling properties of the two verbs as follows.

The relatedness of the two verbs is captured by the fact that each is associated with the same set of participant roles by virtue of being associated with the same (or similar) frame or conceptual domain. The difference between the two verbs is captured in terms of their profiling properties, represented in bold type.

 Argument roles

In contrast to the relative specificity of participant roles, the argument roles that are associated with sentence-level constructions in Goldberg’s model are of a more general semantic kind, and are familiar from a range of approaches to sentence structure that assume semantic roles. We have already encountered semantic roles in Part III of the book (recall our discussion of grammatical functions in Chapter 17, for example). As we have seen, this type of approach rests upon the semantic partition of the clause into predicate and arguments. Recall that this sense of the term ‘predicate’ is different from the traditional grammar sense, in which the predicate is everything in a clause apart from the subject (that is, the verb and any objects or modifiers it may have). In the semantic roles sense, the predicate is usually a word-level unit that can be thought of as the semantic ‘head’ of the sentence. This word expresses the action, event, property or relation that the clause describes. Prototypically, the predicate of a clause is the lexical or content verb, which explains the central status of the verb in many approaches to explaining the relationship between grammar and meaning. As we saw in Chapter 17, the predicate can be a predicative adjective, a predicate nominal or a preposition in sentences with a copular verb.

Depending on the semantics of the predicate, it will take a certain number of arguments which are the participants or entities that the predicate requires in order to complete its meaning: a verb like die only involves a single participant, while a verb like love involves two and a verb like put involves three. The number and type of arguments that a predicate requires is traditionally referred to in terms of valence, as we saw in Chapter 17; argument structure is an alternative term for valence. Parts of the sentence that are not required by the predicate, but that provide ‘incidental’ or circumstantial information (typically, expressions of place, manner, time and so forth), fall outside the argument structure of that predicate, which explains why expressions like this are optional.

The semantic roles approach goes beyond the number of arguments required by a predicate and also looks at the types of arguments required in terms of their semantic properties. For example, the verb die requires a participant capable of living in the first place, while the verb love requires at least one of its participants to be a conscious and sentient being. On the other hand, it is difficult to say that you love someone or something ‘on purpose’, while purpose and intention are certainly involved if you slap someone. In order to try and capture these semantic restrictions, various proposals have been put forth concerning the semantic roles played by these arguments or participants, some of which are familiar from the discussion in previous chapters. Another name for semantic roles is thematic roles. Some examples are given in (10).

The idea of semantic roles has been very influential in modern linguistics, and both formal and cognitive models rely upon this notion in terms of addressing the nature of the relationship between grammar and meaning. In transformational generative approaches like Principles and Parameters Theory, semantic roles are listed in the lexicon as part of the lexical entry of a predicate. In the cognitive model, of course, this partition between lexicon and grammar is not admit ted. As we saw in Chapter 17, semantic roles play a crucial role in Langacker’s Cognitive Grammar account of the grammatical functions subject and object via their participation in the prototypical action chain model, where AGENT is conceived in terms of ‘energy source’ and PATIENT in terms of ‘energy sink’. This model underlies unmarked active declarative sentences as well as explaining the properties of passive constructions on the basis of marked coding or TR-LM reversal. In this respect, Langacker’s model is rather similar to Goldberg’s, in that AGENT and PATIENT are not linked directly to individual verbs but to some underlying representation that structures the clause. However, while Langacker focuses on the cognitive model that underlies the clause, Goldberg focuses on the grammatical construction itself that arises from this cognitive model.

In the remainder of the chapter, we will see how semantic roles also play a crucial role in constructional approaches to grammar. As we have seen, it is important to emphasise the difference between the ‘standard’ view of argument structure and Goldberg’s view. In most frameworks that assume semantic roles, these are associated directly with a particular lexical item, usually the verb. In Goldberg’s model, semantic roles or argument roles are associated instead with the sentence-level construction. Thus, while a verb is conventionally associated with its own participant roles, a sentence-level construction has its own independent argument roles. This idea is represented by Figure 20.1.

Constructional profiling

 While each verb determines which of its participant roles is lexically profiled or conceptually highlighted, sentence-level constructions also profile their argument roles. However, the constructional profiling of argument roles is more flexible. Goldberg suggests that only the argument roles that are linked to a grammatical function (subject, direct object or indirect object) are con structionally profiled. As we saw in the case of examples (7a) and (8a), other argument roles may optionally be present in the sentence but represented as prepositional phrases, sometimes called oblique objects. In Goldberg’s sense of the term, these are not constructionally profiled: ‘Every argument role linked to a direct grammatical relation (SUBJ, OBJ or OBJ 2) is constructionally profiled’ (Goldberg 1995: 48). This reveals the distinction between lexical profiling and constructional profiling in Goldberg’s model. Lexical profiling relates to the aspect of an expression’s meaning that is made explicit by some expression (recall our discussion of profile and base in Chapter 15). In other words, in the sentence George bought some champagne, the expressions George and some champagne lexically profile (express in language) two participant roles relating to the semantic frame of the verb buy (BUYER and GOODS, respectively). Constructional profiling in Goldberg’s model relates to the realisation of argument roles in terms of core grammatical relations. This means that other arguments may be explicit (lexically profiled) yet not construction ally profiled.

Fusion Havingsetoutthesemanticandstructuralpropertiesthattheindividualverband the grammatical construction each bring to the sentence, questions naturally arise concerning how the two are integrated or fused, In Goldberg’s terms. Goldberg posits two principles that govern the association of a verb’s participant roles with a construction’s argument roles: (1) the Semantic Coherence Principle; and (2) the Correspondence Principle. These are reproduced below:

The Semantic Coherence Principle: Only roles which are semantically compatible can be fused. Two roles r1 and r2 are semantically compatible if either r1 can be construed as an instance of r2 , or r2 can be construed as an instance of r1 . . .Whether a role can be construed as an instance of another role is determined by general categorization principles.

The Correspondence Principle: Each participant role that is lexically profiled and expressed must be fused with a profiled argument role of the construction. If a verb has three profiled participant roles, then one of them may be fused with a nonprofiled argument role of a construction. (Goldberg 1995: 50)

The Semantic Coherence Principle works by matching a participant role with an argument role and seeing if the two overlap sufficiently for one to be construed as an instance of the other. For example, general categorisation principles enable us to determine that the THIEF participant role of the verb steal overlaps sufficiently with the argument role AGENT, because both share semantic properties such as ANIMACY, INTENTION, CAUSATION and so on.

 The Correspondence Principle states that profiled argument roles are obligatorily matched with profiled participant roles, but builds some flexibility into the system by allowing that one of the participant roles may or may not be constructionally profiled in the case of a verb with three participant roles. Equally, a ditransitive construction can contribute a third role to a two-participant verb. These ideas are illustrated by Figure 20.2, which represents the CAUSE-RECEIVE ditransitive construction. In this representation of the construction, ‘Sem’ represents the semantic structure of the construction in terms of argument roles, and ‘Syn’ represents the syntactic structure of the construction in terms of how the grammatical functions subject and object(s) realise the argument roles. ‘PRED’ represents the potential for any given verb to be mapped onto the construction, and the empty angled brackets represent the potential for that verb’s participant roles to be fused onto the argument roles of the construction. The dotted line represents the argument role that may or may not be constructionally profiled in the case of a three-participant verb, or the argument role that can be contributed by the construction in the event that this third participant is not part of the verb’s independent specification. This means that two-participant or three-participant verbs can be inserted into the construction (because the construction obligatorily profiles AGENT and PATIENT, strict one-participant verbs are not compatible with this construction). Consider the examples in (12) which illustrate how this works.

In example (12a), the three participant roles of the verb send (SENDER, SENDEE and SENT) are mapped onto the three argument roles of the ditransitive construction (AGENT, RECIPIENT and PATIENT, respectively). In this case, all three profiled participant roles are constructionally profiled. In (12b), on the other hand, only the SENDER and SENT participant roles are mapped onto argument roles; the SENDEE role is optionally represented as a PP, which means that it is not constructionally profiled because it is not represented as a direct object nor as an indirect object. These possibilities are represented in Figure 20.3. Observe that the construction also rules out (12c), on the ungrammatical interpretation that Lily is the RECIPIENT (George sent Lily something). Because AGENT and PATIENT roles are obligatorily profiled, if one of these fails to be realised, the result is ungrammatical. Observe that (12c) is grammatical on the interpretation that Lily is the PATIENT (George sent Lily somewhere).

While we might describe send as a prototypical three-participant verb, it is not obvious that the verbs write and sing would also be described in this way. For example, both can occur in an intransitive frame (for example, George writes; George sings vs. *George sends), as well as in a monotransitive frame (for example, George wrote a novel; George sang the blues). As examples (12d) and (12e) illustrate, however, these verbs are licensed to occur with an ‘extra’ argument (the RECIPIENT) by virtue of their occurrence in the ditransitive construction. As these examples show, the construction contains the flexibility, while the verb determines which of the possibilities provided by the construction are realised. Furthermore, while the verb send permits both possibilities presented by the construction – in other words instances of the construction in which the recipient either is (12a) or is not profiled (12b) – a verb like hand permits only the first option.

The difference between the two verbs can be captured in terms of which participant roles they obligatorily lexically profile, as we saw above. The square brackets around the SENDEE participant role in the representation of send (14a) illustrates that this participant role is optionally lexically profiled, while all three of its participant roles are obligatorily lexically profiled by the verb hand (14b).

Of course, as well as explaining how the participant roles of particular verbs are mapped onto the argument roles of particular constructions, Goldberg’s model must also explain how the ‘right’ verbs are matched with the ‘right’ constructions in the first place. In other words, the model must explain how examples like (15) are ruled out:

An example like (15) might result if we were licensed to map the verb sadden onto the ditransitive construction, merging the three-participant semantics of sadden (X CAUSES Y TO BE SAD BY SOME MEANS Z) onto the three-role semantics of the ditransitive (X CAUSES Y TO RECEIVE Z). As Goldberg points out, it is necessary to restrict the linking of certain constructions to certain classes of verbs by explaining which aspects of a verb’s meaning license the linking. Although Goldberg does not state a specific principle that governs this licensing, she suggests that certain aspects of verb meaning are salient in this licensing process. For example, if a verb’s meaning denotes a subtype of the event type represented by the semantics of the construction, this will license the linking. For example, give, handand send are all subtypes of the CAUSE-RECEIVE event. Alternatively, a verb’s meaning might denote the means by which the event designated by the construction is brought about. This is illustrated in (16):

In each of these examples, the mapping of the verbs throw, roll and slide onto the ditransitive construction is licensed because the verb expresses the means by which George caused Lily to receive the tomatoes (by throwing, rolling or sliding them). This approach therefore goes some way towards explaining why strict one-participant verbs like die are not mapped onto the ditransitive construction: it is difficult to think of a strict one-participant verb that encodes the semantics of the TRANSFER event type, because this event type by definition requires the profiling of at least two participants, if not three.

اخر الاخبار

اخبار العتبة العباسية المقدسة

قسم صناعة الشبابيك يواصل أعمال الإدامة لأحد أبواب حرم مرقد أبي الفضل العباس (عليه السلام)

ضمن فعاليات ملتقى النورين.. المجمع العلمي يقيم ندوة فكرية في بابل

المجمع العلمي يحيي ليلة القدر الأولى في مقام الإمام المهدي (عجّل الله فرجه)

مركز السيد جعفر الحلي الطبي يقدم الاستشارات والخدمات الصحية في ليالي القدر المباركة

المزيد

الآخبار الصحية

خمسة مؤشرات صحية مهملة قد تنقذ حياتك

متى تكون الحمى طبيعية؟

التدخين الإلكتروني يهدد قلبك.. دراسة جديدة تكشف مخاطر ارتفاع ضغط الدم

دراسة تكشف فائدة بعض أنواع الشاي في محاربة الاكتئاب

المزيد

الآخبار التكنلوجية

الجهاز المناعي يظهر دورا غير متوقع في صحة العظام

ابتكار من جامعة الشارقة منخفض التكلفة للحماية من الزلازل

اكتشاف تأثير غير متوقع للكاكاو على وظائف الدماغ

باحثون يرصدون أثر تغيّر المناخ في دم البشر

المزيد

مواضيع ذات صلة

Advantages of a constructional approach to verb argument structure

Assumptions

Grammaticalization in other pidgins

On pidgins and other restricted linguistic systems Discussion

The rise of new functional categories

Goldbergs construction grammar

The architecture of construction grammars

The Generalisation Commitment

Comparing Construction Grammar with Cognitive Grammar

Construction Grammar: a broadly generative model

Kenya Pidgin Swahili (KPS)

On pidgins and other restricted linguistic systems