Theoretical Persuasions concerning the Semantic Extension among Different Senses of the Polysemy

In the recent trends of studies on semantic extensions concerning polysemy, there are four widely discussed approaches: Frame Semantics, Construction Grammar, the theory of Subjectivity operating with inference and Conceptual Blending. Some brief introduction and correlations among the three approaches are given below.

2.1.1 Frame Semantics

Frames refer to a structured background of experiences or beliefs, constituting a conceptual prerequisite for understanding the meaning of a word (Fillmore and Atkins 1992). In the infrastructure of FrameNet (http://framenet.icsi.berkeley.edu/), which is

constructed by University of Berkeley providing a frame-based analysis of the English lexicon based on the theory of Frame Semantics (cf. Fillmore 1968, 1976, Fillmore & Atkins 1992), polysemous⁶ words are put into different frames which highlight different frame elements, that is, participant role or arguments. The relations among the frames, and thus word senses, are not connected to one another directly, but are associated with the amount of shared frame elements and with the manner in which each frame highlights particular frame elements. For instance, if A frame inherits most of the core frame elements from B frame, the relation between A and B is called Inheritance. In addition to Inheritance, there are seven other frame-to-frame relations in FrameNet (Ruppenhofer et al. 2006:8, 103-111), namely using, perspective_on, subframe, precedes, inchoative_of, causative_of and see_also (cf.

Appendix I). Following the observation of Liu and Chiang (2008), semantic inheritance exists from top to bottom in the hierarchical structure. To account for multiple inheritances of a frame from other source frames, e.g., a given micro-frame may inherit features from a number of basic, larger-scope frames, the inheritance relation could be realized as syntactically revealed specifications of frame attributes (role-internal specifications of frame elements). For instance, the Suggest Frame is inherited by Statement Frame and Request Frame since the Speaker makes a suggestion and intends the Addressee to perform certain action by following the suggestion. Hence, the complement of the verb suggest is given the tag of frame element ‘Message+act’ to mark the attribute of its semantic role. The ‘inheritance’

relation discussed above will be the most relevant to account for the sense relatedness, background prerequisites and semantic extensions from one frame to another.

6 The relations of polysemous word corresponding to different frames is then named as ‘One sense, one frame’ principle by Liu and Chiang 2008.

The Frame Semantics provides sound semantic background for the analysis of the argument structure of verbs; however, in order to show a uniform representation for the meanings of words, sentences, and texts, Construction Grammar was later developed by Fillmore and his colleagues (Fillmore 1976:28). Petruck (1996) claimed that the connections between Frame Semantics and Construction Grammar go beyond the matter of representations. In Construction Grammar, the semantic frame associated with a lexical item provides some semantic information required for the sentence interpretation. As with lexical items and texts, semantic descriptions and explanations of grammatical constructions often depend on frames for background information about the scene schematized by the construction. The relations between the two complementary theories bring us to the section of Construction Grammar.

2.1.2 Construction Grammar

Construction Grammar (CG) (Fillmore, Kay & Connor 1988, Fillmore & Kay 1993, Jackendoff 1997, Goldberg 1995, 2006) is in fact partially derived from Frame Semantics. CG differs from other semantic theories in that in addition to word meanings, it emphasizes the meaning derived from the syntactic pattern, i.e.

construction, itself. Constructions which are defined as form-meaning correspondences that exist independently of particular verbs carry meaning themselves, independently of the words in the sentence. Take the intransitive verbs sneeze and bake for instance, they are coerced to become transitive verbs due to the placement in the following construction (34-35):

(32) He sneezed the napkin off the table. Æ CAUSE Y to MOVE Z by sneezing (33) She baked him a cake. ÆINTENDS TO CAUSE Y to HAVE Z by baking

The advantage of CG is that it can be applied to account for the dynamic or temporary meaning (Huang et al 2003) coerced by the construction itself. However, as what Fillmore, Kay and O’Connor (1988) had mentioned, Construction Grammar aims at accounting for pragmatic knowledge besides the knowledge of words, grammatical rules and semantic interpretation. To grasp the pragmatic knowledge, according to Sacks, Schegloff, Jefferson (1974) and Sacks (1992), pragmatic methods, especially conversation Analysis ought to be emphasized. There will be more description concerning the data collection in chapter 3.

2.1.3 Subjectivity

To deal with the issues of grammaticalization and semantic change, five pragmatic mechanisms are often utilized: metaphor, inference (or the conventionalization of implicature), generalization, harmony, and absorption of contextual meaning (Bybee, Perkins and Pagliuca 1994). Those five mechanisms are neither arbitrary nor exclusive. They may occur in different stages during the process of meaning development or when some or one of them is in operation. Among the five mechanisms, ‘inference’ has drawn my special attention due to its correlation with the notion of subjectification, proposed by Traugott (1999), which operates with

‘subjectivity.’ Traugott suggests that subjectification is able to account for the overall tendency for meanings to become more subjective with the passage of time:

propositional > textual > interpersonal (expressive). Tragott (1999) proposes three semantic-pragmatic tendencies, listed below:

1) Semantic-pragmatic tendency 1:

Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) situation.

Spatial > temporal, e.g. going to future

Object > space, e.g. body-part metaphors such as head, back, foot, face

2) Semantic-pragmatic tendency 2:

Meanings based in the described external or internal situation > meanings based in the texture situation

e.g. the emergence of connectives such as since.

3) Semantic-pragmatic tendency 3:

Meanings tend to become increasingly situated in the speaker’s subjective belief state/ attitude towards the situation:

e.g. may (ability > permission > possibility)

Among the three semantic-pragmatic tendencies Tragott proposes, the third one, which demonstrates the development of epistemic modality by the grammaticalization of may, is most relevant to the focus of this study and found adequate for explaining the cognitive mechanism behind the semantic extension of the target classes of verbs.

2.1.4 Conceptual Blending

The theory of Conceptual Blending (Fauconnier 1985, 1994, 1997, Fauconnier and Turner 2002) is at work in many areas of cognition and action, including metaphor, and conceptual change. It is suggested by this theory that meaning is not understood compositionally but by the formal expression promoting hearer and reader to assemble and develop mental spaces operating by conceptual blending.

The mental space can be understood as a frame, a socio-cultural scenario or a skeletal conceptual structure set up by grammatical constructions, i.e. the meaning coerced by the construction itself. Mental spaces can be introduced either by grammatical expression (so-called space-builder) or by means of pragmatic information projected onto the input. In oral languages, based on Pascual 2002, in addition to meaning carried by construction, space-builders may take on a variety of grammatical forms, such as: i) subject verb complexes (e.g. Ursula wants, Sally believes, The child thought); ii) prepositional phrases (e.g. in her dream, in cold

climates, behind bars); iii) adverbials (e.g. perhaps, actually, theoretically); iv) conjunctions (e.g. if, when, because); and v) some morphemes such as evidential particles (e.g. -la in Haviland 1987, 1996).

These input either open up a new mental space, corresponding to the ‘frame convergence’ which is accompanied with role coercion that allows the different frame elements in the same position get transferred and conflated into a new role (Liu &

Hsieh 2008) or shift the focus to some existing ones, i.e. the semantic extension to another mental space. Figure 6 below illustrates how blending operates.

Figure 6: Conceptual Blending (Fauconnier 1997)

In Conceptual Blending Theory, the four-space model is typically made use of.

The four spaces are: two ‘input’ spaces, a ‘generic’ space, and the ‘blend’ space.

Blending is generally described as involving two input mental spaces that, according to a given structure mapping, will generate a third one, called Blend. For instance, as shown in Figure 6, the Input I1, carrrying the space builder, the caused motion construction, fuses with the Input I2, carrying the frame of causal sequence to yield a new space, i.e. the blend, leading to the transitive use of a prototypical intransitive

Junior sped the car around the Christmas tree.

Caused Motion Construction:

NP V NP PP

John throws the ball to Susan.

A causal sequence :

[[a ACT] CAUSES [b MOVE to c]]

ACT: speed

verb speed in the caused motion construction. It’s noticeable that the new blend maintains partial structure from the both inputs and adds emergent structure of its own.

This is compatible with the analysis of multiple inheritances of frames suggested by Liu & Chiang (2008). Therefore, the four-space model of Conceptual Blending theory is adequate for integrating the above three theoretical foundations, i.e. Frame Semantics, Construction Grammar and the theory of subjectivity. The space-builders, such as metaphorical extension, inference and the grammatical coding means (e.g.

affix) can be incorporated into the model as the link between each mental space.

In summary, the four approaches shed light on different aspects of polysemy and sense extension and may adequately account for some facets of the surface syntactic coding of verbs. Hence, in this study, Frame Semantics will be taken as the major framework for the consideration of argument structure. With the contribution of Construction Grammar, the dynamic meaning coerced by a construction can be accounted for. Finally, the four-space model of Conceptual Blending will be applied to incorporate the inputs of meanings from frames and construction and the space-builders, such as the inference induced by subjectification and metaphor as a possible cognitive base for explaining the semantic extension among the target verb classes.

2.2 Classification of P-C-U, Manipulation and Modality Verbs in English