Theme: polysemy subject: Lexicology Compiled by: Tursunboyev Sardor, group -60 Supervisor: F. f f. d. (PhD) Gavharoy Isroiljon kizi Andizhan 2023 Theme: Polysemy

5b. The painting is over the couch. (‘above’)
5c. The truck ran over the rabbit. (‘across’)
The idea is that over constitutes a radial category composed of a range of distinct but related senses, organized around the prototypical, or central, sense (which Brugman and Lakoff take to be the ‘above and across’ sense in (5a)), in a lexical network structure. (A slightly different, albeit similar manifestation of the network model of polysemy representation is given by Langacker, 1988.) The different senses of over exhibit typicality effects; more typical senses are located “closer” to the prototypical sense in the network, while less typical senses are located in its periphery. Such peripheral senses are derived from more typical senses by a set of cognitive principles for meaning extension (e.g., “conceptual metaphors,” cf. Lakoff & Johnson, 1980), giving rise to meaning chains (e.g., sense A is related to sense B in virtue of some shared attribute(s), sense B is related to sense C, which is related to sense D, and so on). For instance, the “control” sense in (5g) is seen as being derived from the “above” sense in (5b) on the basis of the metaphorical schema CONTROL IS UP, LACK OF CONTROL IS DOWN (Lakoff, 1987). Sense relations, then, concern, in the first instance, adjacent members of the category, while members that are only indirectly connected in the semantic network may be very different in semantic content. Wittgenstein’s (1953) metaphor of “family resemblance” (cf. Rosch & Mervis, 1975) is often used to describe such polysemous categories within the cognitive linguistics framework (Taylor, 2003).
A central aspect of Lakoff and Brugman’s approach is that radial categories are stored in the long-term semantic memory of speakers. In this respect, the radial category account of polysemy is a radical version of a sense enumeration lexicon, in that the full range of senses are taken to be stored as part of a semantic network (hence, it is sometimes referred to as the “full-specification approach,” cf. Evans & Green, 2006). A common criticism of the full-specification approach, and of sense enumerative accounts more generally, is that they seem to entail a (potentially) indefinite proliferation of mentally stored senses in order to cover the range of uses of lexical forms (for instance, Brugman, 1988, identifies nearly a hundred different uses of over), and thereby fails to distinguish those aspects of meaning that are part of the word meaning proper and those that result from its interaction with the context, a problem that is sometimes referred to as the “polysemy fallacy” (Sandra, 1998). It is possible that work in distributional semantics, which makes use of nondiscrete representations, might give a different shape to the notion of a radial category, one which is not committed to the storage of each individual sense that forms part of the network.
More recently, scholars working within the cognitive linguistics paradigm have refined their accounts of polysemy, acknowledging the context-dependence of word meanings (Allwood, 2003; Evans, 2005, 2009; Taylor, 2006; Tyler & Evans, 2001, 2003; Zlatev, 2003). In particular, Tyler and Evans (2001, 2003) have developed an account of polysemy that, while espousing the Lakoff-Brugman idea that polysemous senses are represented in terms of sense networks centered around a prototypical sense, proposes a set of criteria that tries to (i) determine whether a particular sense of a word counts as a distinct sense, and (ii) establish the central sense of a polysemous lexical item. The “Principled Polysemy approach” is also an attempt at avoiding the polysemy fallacy by distinguishing between those senses that are stored in semantic memory and those that are pragmatically constructed during online processing.
One-Representation Approaches
Ruhl (1989) was the first to spell out in some detail what we call the one-representation approach (sometimes also referred to as the core meaning approach). Ruhl’s work was a critique of the idea that words in general have multiple meanings, which was the mainstream view in linguistic theory at the time, and which had led to the postulation of sense enumeration lexicons in various forms. According to Ruhl’s monosemy approach, words should initially be presumed to have only a single meaning, and polysemy (or homonymy) be posited only when extended attempts to describe this meaning fail. Ruhl’s methodology involved collecting a large set of data corresponding to a word’s actual uses and extracting from this set a highly abstract, unitary schema that covered all its uses. For instance, the corpus analyses presented in Ruhl (1989) included several hundred attested uses of the verbs bear, hit, kick and slap, typically taken to be highly polysemous. Ruhl claimed that each of these uses revealed an abstract unified meaning, albeit not one that could that could be comprehensively captured by a single word or phrase (Ruhl, 1989, p. 63). One important insight of Ruhl’s proposal was that a large part of what had traditionally been treated as part of the lexical meaning of a word, and which had led to a proliferation of senses on previous accounts, was more adequately described as senses that were generated from an underspecified meaning as a result of its interaction with the linguistic and/or extralinguistic context.
Another version of the one-representation approach is the generative lexicon theory (Pustejovsky, 1991, 1995). This theory stands out from the other accounts discussed so far by being developed with the sole purpose of explaining polysemy. Pustejovsky sought to provide a more explanatory account of polysemy than that given by sense enumeration lexicons. In his view, such accounts are inadequate, primarily because they are unable to explain how words may take on an infinite number of meanings in novel contexts. Not only is it impossible for such accounts to list all the possible meanings of a lexical item, they also miss the generalizations that can be made on the basis of what appear to be regular patterns of sense alternations, and fail to capture how polysemous senses may partially overlap and be logically related to one another. A more promising approach, he argues, which is able to meet these explanatory requirements, is a lexicon where items are represented as templates combined with a generative framework for the composition of lexical meanings. The generative lexicon theory is thus an example of a one-representation approach.
On Pustejovsky’s account, the semantics of a lexical item is viewed as a structure consisting of several components. A key component is the qualia structure, which consists of a specification of four different roles: The constitutive role captures the relation between an object and its constituents, or proper parts; the formal role specifies what distinguishes the object within a larger domain; the telic role defines the purpose and function of the object (if there is one); and the agentive role describes the factors involved in the origin or coming into existence of the object. For instance, the qualia structure of the noun novel will specify that a novel is a narrative (constitutive role), it is a book (formal role), its purpose is to be read (telic role), and it comes into being by a process of writing (agentive role). Complementing these underspecified (yet informationally rich) lexical entries is a set of generative mechanisms, which operate to yield compositional interpretations. For instance, selective binding is exemplified by the process by which an adjective takes one event expression contained in the qualia structure of the head noun as input to the interpretation process, as in the uses of good in (6):
6a. American Pastoral is a good novel.
6b. A chef needs a good knife.
Here, good selectively modifies the event description given by the telic roles of the nouns (novels are for reading; knifes are for cutting), giving rise to the interpretations ‘good read’ and ‘a knife that cuts well.’
Pustejovsky’s generative lexicon provides a considerably more explanatory account of polysemy than sense enumeration lexicons, and it has had a profound impact on research in lexical semantics. However, it also has to confront several problems related mainly to inflexibility (Asher, 2011; Blutner, 2002; de Almeida & Dwivedi, 2008; Falkum, 2007; Fodor & Lepore, 1998; Willems, 2006). Other formal approaches build on Pustejovsky’s original insights but seek to avoid some of its problems (Asher, 2011; Babonnaud, Kallmeyer, & Osswald, 2016).
Contemporary Debate
The contemporary debate has been reshaped by psycholinguistics. Experimental work on polysemy started in the 1990s and is gaining influence also within theoretical approaches. Studies have been conducted using a variety of methodologies, including lexical decision and sensicality judgment tasks, eye tracking, magnetoencephalography (MEG) and electroencephalography (EEG) recordings. The main topic in the psycholinguistics literature is how polysemous senses are represented vis-a-vis homonymous meanings in the mental lexicon. Studies conducted during the 1990s had suggested that polysemous senses were stored differently from homonymous meanings (Azuma & van Orden, 1997; Frazier & Rayner, 1990; Williams, 1992). These studies discovered facilitation effects for different senses of polysemous terms and competition between homonymous meanings. Using lexical decision tasks, Azuma and van Orden found that words with many related senses were easier to recognize than words with unrelated meanings, which suggested that, unlike the meanings of homonyms, the senses of polysemous terms do not compete for activation. More recent studies have given consistent results (Beretta, Fiorentino, & Poeppel, 2005; Frisson, 2015; Klepousniotou & Baum, 2007; Klepousniotou et al., 2008; Pickering & Frisson, 2001; Pylkkänen et al., 2006, but cf. Klein & Murphy, 2001).
Let us illustrate the kind of work done in psycholinguistics by explaining in some detail Frisson’s (2015) study of how we process so-called ‘book’ polysemy (e.g., book, manuscript, notice, journal, etc.), with results that can be plausibly extended to other kinds of inherent polysemy. The study consisted of two experiments: one sensicality judgment task and one eye-tracking study. In the sensicality judgment task, subjects were presented with a prime NP in which the adjective focused on either the physical object sense (e.g., bound book) or the information sense (e.g., scary book). Then they were asked to make a sensicality judgment about a target NP in which the adjective focused on either the consistent (e.g., [well-plotted book], scary BOOK) or the inconsistent (e.g., [bound book], scary BOOK] sense. The results showed a clear consistency effect, with increased processing time in the inconsistent condition compared with the consistent condition, but no effect of either sense dominance or direction of sense switch (physical object to information or information to physical object) in the inconsistent condition. In the eye-tracking study, there were three conditions: The neutral condition aimed at testing how quickly a specific sense is assigned to a polysemous word without prior contextual indication. The repeat condition aimed at testing the effect of sense repetition on ease of processing. Finally, the switch condition tested whether switching from one sense involves an extra processing cost. In the neutral condition, subjects did not have more difficulty disambiguating towards the subordinate sense than toward the dominant sense of the polysemous noun. In the repeat condition, subjects spent more time reading the polysemous noun than in the neutral condition, but the time to select a particular sense was not affected by sense frequency. In the switch condition, processing was more difficult than in the neutral context, and switching from a subordinate to a dominant sense induced a greater cost than vice versa.
Frisson’s results suggest that “book” polysemy is processed differently from both homonyms and cases of polysemy where the senses are semantically distant. They also suggest that the different senses of inherently polysemous expressions might be stored together as part of a single representation. However, similar results have been found for other, non-inherent types of polysemy. For instance, MacGregor et al. (2015) investigated the processing of the similarity-based polysemy of mouth as in “mouth of a person”, “mouth of a river”, and “mouth of a cave” and observed the same overall pattern of co-priming and facilitation effects observed for “book” polysemy.
Overall, researchers agree that this pattern of results indicates that the senses of a polysemous expression relate to a single representation. What is not clear at this stage, as MacGregor and colleagues (2015, p. 137) point out, is the nature of such a representation (see also Frisson, 2009, p. 122).
The alternatives psycholinguists consider are:
An underspecification (thin or core meaning) account: The meaning of a polysemous expression is an underspecified, abstract, and summary representation that encompasses and gives access to its different senses in context.
An overspecification (rich) account: The meaning of a polysemous expression includes all its different senses, which are stored as part of a single representation. Senses are component parts of the meaning of the word, which are selected in context.
Literalism: Each polysemous expression has a literal, denotational meaning. Its other senses are generated via linguistic rules, coercion mechanisms, or pragmatic inferences.
The details of the underspecified and overspecified meaning representations remain unclear. Underspecified meaning representations are usually understood as including features that are shared by all the different senses of the polyseme (this is the “core meaning” approach), while overspecified representations are supposed to include all the senses that the polysemous expression gives access to, which can include a sheer collection, correspond to some kind of structured representation (such as, for instance, one of Pustejovsky’s qualia structures, cf. Pustejovsky, 1995; Frisson, 2009; Vicente, 2015; Del Pinal, 2015), or to distributional semantic representations (Lenci, 2008; Baroni, Bernardi, & Zamparelli, 2014). Literalism has been scarcely considered and tested by psycholinguists (but see Frisson & Frazier, 2005; Frisson, 2015), but has a strong position in many other fields, ranging from computational semantics to cognitive pragmatics. It is endorsed by computational linguists such as Asher (2015), who aim to explain most meaning variations in terms of coercion mechanisms, Copestake and Briscoe (1995), who account for regular polysemy in terms of lexical rules, and some version of it is also assumed by representatives of Relevance Theory’s pragmatic approach (Carston, 2002; Falkum, 2011, 2015; Wilson & Carston, 2007). Although underexplored in the psycholinguistic literature, the results seem, on the face of it, to speak against literalism, given that there is no record of a preference for alleged literal meanings except in some cases such as mass–count polysemy (Frisson & Frazier, 2005), or of the operation of processes—such as coercion—working on them. However, the psychological plausibility of literalism would be a fruitful topic to explore in future experimental studies.
The issue of how polysemous expressions are semantically represented is of course strongly relevant to the broader issue of lexical meaning representation. Given that most, if not all, words in a language are polysemous (Zipf, 1945), it seems safe to transfer the above views about the meaning of polysemous expressions to views about word meaning simpliciter (Falkum & Vicente, 2015; Vicente, 2017). Mainstream semantics has been operating with an idealized literalist position on standing word meaning, and this position has been questioned on various fronts. There is a number of authors from different schools who defend an underspecification approach (e.g., Bierwisch & Schreuder, 1992; Carston, 2013; Evans, 2009; Yalcin, 2014) or an overspecification account (e.g., Elman, 2011; Pustejovsky, 1995; Rayo, 2013; Vicente & Martínez Manrique, 2016; Zwarts, 2004). Some interesting recent work also points in the direction of a renewed interest in the sense enumeration hypothesis, though from a more cognitively plausible viewpoint in which the interaction between stable senses and contextual factors plays a key role (Carston, 2016, forthcoming; Recanati, 2016). In short, there is an increasing interest in lexical meaning, and a deep but still scattered debate across the different traditions regarding the nature of the meaning of lexical content words. Future research on polysemy is bound to have a decisive impact on this question.