UCSD vs. MIT on Frames in semantics

From: Grant Callaghan (grantc4@hotmail.com)
Date: Sat 22 Feb 2003 - 16:33:30 GMT

  • Next message: Keith Henson: "Re: UCSD vs. MIT on Frames in semantics"

    >> The cognitive linguists (Turner, Fauconnier, Lakoff, etc.) seem to be
    referring to the frame concept of Charles J. Fillmore; cf. his widely cited article “Frame Semantics” (The Linguistic Society of Korea (ed.) 1982: Linguistics in the Morning Calm. Seoul: Hanshon Publishing Co.). As far as I can discern, Fillmore’s frames are more or less the same as Fauconnier’s mental spaces (1994: Mental Spaces. Cambridge UP; 1997: Mappings in Thought and Language. Cambridge! UP).

    >About the computer-mind metaphor that is entirely unacceptable to those
    >people, as anyone can see from scanning Lakoff and Johnson. In a sense,
    >their aversion to MIT/Chomskyan linguistics defines the identity of
    >cognitive linguistics.

    Differences between the UCSD and MIT approaches to frames:

    Frame Semantics Fillmore (1982) suggests that there exist many, many words which rely for their understanding on speakers' experience with the scenarios and social institutions they presuppose. For example, words such as ``Tuesday'' cannot even be defined without providing a certain amount of background information about the more general concept of the organization of the week. Similarly the significance of the distinction between ``week'' and ``weekend'' arises because its meaning is motivated by the practice of the 5-day work week. Thus Fillmore defines a frame as a system of categories whose structure is rooted in some motivating context. Words are defined with respect to a frame and perform a categorization that takes the frame for granted. Fillmore (1982) emphasizes how meanings grow out of these motivating experiences and thus lexical semantics requires an account of how and why people use words, as well as a characterization of the scenarios they presuppose. Fillmore construes frame semantics as a far-reaching research program with implications for lexical semantics, meaning change, the creation of novel words, and even the assembly of the overall meaning of a text. In lexical semantics, for example, the aim is to characterize the motivating context for a particular word and explain how the word's meaning relates to that context. Thus words are defined with respect to frames, and are used to evoke them. Fillmore (1977) shows how a number of verbs, including ``buy,'' ``sell,'' and ``pay,'' are related to one another in virtue of how they highlight certain aspects of the same Commercial Event frame. Indeed a number of verbs can be understood as evoking the same frame, but accentuating (or profiling) the perspective, motives, or intentions of particular participants. Examples include ``buy'' versus ``sell'' (Fillmore, 1977), ``give'' versus ``take''
    (Fisher, Hall, Rakowitz, & Gleitman, 1991), and ``substitute'' versus
    ``replace'' (Landau & Gleitman, 1985). Examples such as this accentuate how meaning cues the particular construal of events rather than merely providing speakers with an objective characterization. Moreover, they demonstrate how frames are motivated by human experiences, social institutions, and cultural practices. Similarly, the meaning of ``bachelor,'' classically defined as an unmarried man, can be shown to depend on the existence of background information grounded in social practice. Questioning whether, say, the pope, tarzan, or a gay man in a long-term relationship count as bachelors, Fillmore argues that the definition of ``bachelor'' as an unmarried man relies on the existence of a frame, or set of propositions which represent common assumptions about the normal course of a man's life in western society. Much as our understanding of ``on'' in (6) and (7) involves a tacit assumption of a gravitational field, talk about bachelors involves implicit acceptance of its background assumptions. When these assumptions do not obtain for a particular man, say Pope John Paul II, we are hesitant to apply the term
    ``bachelor.'' Lakoff (1987) emphasizes the idealized character of the background assumptions represented in frames. Because these assumptions involve a large degree of oversimplification, they apply more easily to certain segments of society than others. Thus grounding lexical semantics in these idealized cognitive models has the virtue of providing an account of prototype effects in categorization. The frames invoked by linguists to understand lexical semantics can also be used to explain other cognitive tasks, including reasoning, problem solving, and making judgments about the behavior of others.

    Frames as data structures. Fillmore's linguistically motivated account of a frame is paralleled by similar suggestions from researchers in other branches of cognitive science. In the field of artificial intelligence, Minsky (1975) proposed the term frame for a data structure used to represent commonly encountered, stereotyped situations. Minsky offered a child's birthday party as an example of the sort of thing a frame might be used to represent. A birthday party frame includes slots, such as food, games, and presents, which specify general features of the event. Slots are bound to fillers, or representations of the particulars of a situation. In a process called slot-filling, slots such as food are bound to fillers such as cake and ice-cream. The efficiency of frames as data structures derives from the organization of general slots which can be bound to particular fillers. This provides a means of organizing the similarities as well as the differences which exist between our various experiences of children's birthday parties. Activating a frame creates expectations about important aspects of the context by directing the agent to fill the slots with available information. Moreover, the real power of frames derives from the use of default values which consist of the most typical and/or frequent filler for each slot. If information about the actual slot-filler is unavailable, a slot is assumed to be filled by the default. The beauty of Minsky's suggestion was the notion that frames contain the sorts of information needed to understand a particular sort of event or scenario, as well as default information about the most probable fillers for any given slot. Moreover, the representational structure in frames proved to be valuable in building computer systems to understand natural language. In the course of developing a system that could understand simple stories, Schank & Abelson (1977) postulated scripts as analogous to Minsky's frames. Scripts represent stereotyped sequences of events such as going to a restaurant, and contain slots which are either filled by binding the particular fillers manifest in the situation at hand, or by instantiating the default value for any particular slot. While the frame type data structure was largely motivated by considerations of representational utility, cognitive psychologists have found considerable evidence that people utilize frames, or schemata (schemas) as they are called in the psychological literature, in a variety of cognitive tasks. People use frames in perception, planning, and memory for events (Barsalou, 1992). Moreover, frames have been used to explain human ability to make inferences in complex situations, to make default assumptions about unmentioned aspects of situations, and to make predictions about the consequences of actions. <footnode.html> <footnode.html> In cognitive semantics, meaning does not involve mapping from terms to objects, actions, and events in the world. Rather, words designate elements and relations in frames which may represent objective aspects of reality, but need not (Fauconnier, 1997). So, instead of positing one set of processes to track correspondences between terms and objects, and another for terms and various abstract, relational properties, we can see the former as following trivially from the latter. That is, words are always understood as setting up frames, regardless of whether those frames apply to actual, representational, or hypothetical referents. Apparent core cases - where frames apply directly to real-world referents - are merely a subset of a more inclusive phenomenon. Consequently, there is nothing to be gained from treating these so-called core meanings as more fundamental than more exotic looking utterance meanings. For one thing, doing away with the notion of core meanings dissolves the problem of how to circumscribe the core and parameterize its extensions. Moreover, addressing more exotic cases of meaning construction has led to an important locus of generativity in language production, namely the human capacity to map within and between frames and scenarios.


    _________________________________________________________________ MSN 8 with e-mail virus protection service: 2 months FREE* http://join.msn.com/?page=features/virus

    =============================================================== This was distributed via the memetics list associated with the Journal of Memetics - Evolutionary Models of Information Transmission For information about the journal and the list (e.g. unsubscribing) see: http://www.cpm.mmu.ac.uk/jom-emit

    This archive was generated by hypermail 2.1.5 : Sat 22 Feb 2003 - 16:30:37 GMT