1

A Unified Account of Polysemy within LCCM Theory 1 Vyvyan Evans (v.evans@bangor.ac.uk) 2 Draft, May 2014. To appear in Lingua 3

4 Abstract 5

Within the cognitive linguistics tradition, polysemy has often been viewed as a function of 6 underlying entries in semantic memory: word forms have distinct, albeit related, lexical 7 entries, which thereby give rise to polysemous word senses in language use (e.g., Evans 8 2004; Tyler and Evans 2001, 2003). In this paper, I seek to broaden out the study of 9 polysemy within this tradition by tackling it from two slightly different angles. I argue that 10

polysemy can also arise from the non-linguistic knowledge to which words facilitate access. 11 This phenomenon I refer to as conceptual polysemy. I illustrate this with an analysis of the 12 lexical item book. Moreover, polysemy also arises from different word forms which, at least 13 on first blush, appear to share a common semantic representation. This phenomenon I refer 14 to as inter-lexical polysemy. I illustrate with a detailed case study involving an analysis of 15

the prepositional forms in and on. In presenting my account of these two polysemous 16 phenomena, I introduce a contemporary account of lexical representation: the Theory of 17

Lexical Concepts and Cognitive Models, or LCCM Theory for short (Evans 2006, 2009, 18 2013). This provides a common theoretical architecture which facilitates a joined-up account 19

of these specific phenomena, and of polysemy more generally. 20 21

Keywords: polysemy, conceptual polysemy, inter-lexical polysemy, LCCM Theory, the 22 Theory of Lexical Concepts and Cognitive Models 23 24

1. Introduction 25 The issue I address in this paper relates, on the face of it, to two distinct types of polysemous 26

phenomena. The first type concerns examples of the following kind, where a single lexical 27 itemin this case bookobtains distinct readings in different contexts of use. For reasons to 28 be explained below, I shall refer to this phenomenon as conceptual polysemy: 29

30

Phenomenon 1: 31 (1) a. Thats a heavy book tome 32 b. That antiquarian book is illegible text 33 c. Thats a boring book level of interest 34 d. Thats a long book duration 35 36 In each of these examples, book means something slightly different. In (1a), book refers to 37 the physical artefact, and the fact that as a physical artefact books can be heavythis might 38 be dubbed the tome reading. In (1b), book again obtains a reading that concerns the 39 physical properties of books; but here, specifically, the reading relates to the physical text 40

with which a reader engagesthis might be dubbed the text reading. In (1c) book is 41 understood to refer to the activity of reading, and specifically the level of interest achieved by 42

the readerwhat we might think of as the level of interest reading. And finally, in (1d), 43 which again concerns the activity rather than the physical properties of the book, the reading 44 concerns length of time taken to read the bookwhat might be dubbed a duration reading. 45 Collectively, these examples illustrate how an open-class lexical item can take on different 46 interpretations in different linguistic contexts. 47

I refer to this phenomenon as conceptual polysemy as some analysts, (e.g., Evans 48 2009; Langacker 1987) have suggested that this type of polysemy arises in the following 49

way: linguistic context can serve to differentially highlight different aspects of the non-50

2

linguistic or encyclopaedic knowledge to which a word form facilitates access. The 51

polysemy is made possible precisely because our knowledge representation for a book is not 52 a monolithic structure, but consists of a complex array of interdependent components. For 53 instance, Langacker (1987) accounts for polysemy such as this being due to active zones: 54 distinct facets of our conceptual representation for the physical artefact book become active 55 during language use, triggered by linguistic context. As such, context plays a role in 56 modulating (Cruse 1986) exactly which parts of conceptual representation become activated 57 during language understanding. 58 The second type of polysemous phenomenon relates to distinct lexical items, in this 59 case prepositions, appearing to have, at least on the face of it, broadly similar readings: a 60

phenomenon we might term inter-lexical polysemy. In the following representative 61 examples, I am focusing on the English prepositions in and on: 62 63 Phenomenon 2: 64 (2) a. We are in love/shock/pain state 65 Cf. We are in a room spatial 66 b. We are on alert best behaviour/look-out/the run state 67 Cf. We are on the sand spatial 68 69

In these examples, the prepositions in and on both appear to have a distinct state sense70 which contrasts with a spatial sense associated with each preposition. But while the state 71 senses of in and on are broadly similar, the referentsin the sense of the semantic arguments 72 that can co-occur with each prepositionare markedly distinct, as we see from the following 73 examples: 74

75 (3) a. *We are in alert/best behaviour/look-out/the run (Intended meaning: state) 76 b. *We are on love/shock/pain (Intended meaning: state) 77 78 The examples in (3) demonstrate this: the range of semantic arguments that the state senses 79 of in and on can take are in complementary distribution; the state senses associated with in 80 and on are not, in fact, identical. 81 In earlier work (Evans 2009, 2010), the first work I am aware of concerning inter-82 lexical polysemy, I argued that this distinction arises from distinct sets of parameters 83

associated with the state senses of in and on. While I develop the notion of a parameter in 84 some detail later in this article, briefly, a parameter is a schematic unit, or atom of linguistic 85 knowledge. The claim I defend below is that word senses are constituted of an assortment of 86 parameters: atoms of schematic linguistic meaning that, collectively, give rise to a words 87 lexical concept. 88

This construct of a lexical concept accounts, I shall argue, for the distinction observed 89 in comparable senses across distinct lexical items: the phenomenon of inter-lexical polysemy. 90

In contrast, the phenomenon of conceptual polysemy arises, I shall argue, from the non-91 linguistic knowledge to which lexical items facilitate access. Non-linguistic knowledge is 92

constituted of what I shall refer to as cognitive models. And lexical concepts facilitate access 93 to a set of cognitive models: a words semantic potential. While conceptual polyemy arises 94 due to the nature and organisation of non-linguistic knowledge, linguistic context 95 nevertheless plays a role in guiding what sorts of non-linguistic knowledge are activated, in 96 order to give rise to the requisite readings obtained, as evidenced by the examples in (1). 97

The two theoretical constructsthe lexical concept and the cognitive modelthat I 98 use to account for these two types of polysemy arise under the aegis of the Theory of Lexical 99

Concepts and Cognitive Models, or LCCM Theory for short. This provides an account of 100

3

lexical representation and compositional semantics: these two fundamental theoretical 101

constructs provide the theory with its name. My overall objective in this paper, and one 102 facilitated by LCCM Theory, is to provide a unified account of contextual and inter-lexical 103 polysemy, deploying the theoretical insights provided by LCCM Theory. The approach I 104 develop is not unified in the sense that these distinct phenomena can receive a single 105 account. Rather, with a psychologically-appropriate account of language and its relationship 106 with conceptual structure, these two distinct types of polysemy can be viewed as distinct, 107 albeit predictable, symptoms of our universal meaning-making capacity. 108 Accordingly, my argument amounts to this: meaning construction involves linguistic 109 and non-linguistic representations; and these representations inhere in the linguistic and 110

conceptual systems respectively. The representations that are purely linguistic in nature111 lexical conceptsaccount for inter-lexical polysemy. In contrast, the representations that are 112 non-linguistic in naturecognitive modelsaccount for conceptual polysemy. As LCCM 113 Theory provides a joined-up account of the role of both linguistic and conceptual knowledge 114 in meaning construction, this affords a unified account of these two distinct types of 115

polysemy. Moreover, and more generally, the paper makes the case for the distinct and 116 differential role of linguistic and non-linguistic knowledge in accounting for (distinct types 117

of) polysemy. 118 The paper is structured as follows. In Sections 2 and 3, I present the theoretical basis 119

for the account of polysemy I provide later in the paper. Section 2 is concerned with the 120 distinction in the representational format of linguistic versus non-linguistic knowledge. And 121

in Section 3, I situate that distinction in the framework of LCCM Theory. Sections 4 and 5 122 then deploy the theoretical constructs developed, in order to provide an analysis of conceptual 123 polysemy (Section 4), and inter-lexical polysemy (Section 5). In Section 6 I draw 124

conclusions, making the point that LCCM Theory provides a framework that enables a 125 joined-up account of polysemy. 126

127 2. Parametric versus analogue concepts 128 In this section I make the case for a qualitative distinction between conceptual structure on 129

the one hand, and semantic structure on the other. Semantic structure is the representational 130

format that inheres in the linguistic system, while conceptual structure is the representational 131 format that sub-serves the conceptual system. I summarise, here, some of the reasons for 132 thinking there is a principled distinction of this sort in human semantic representation, 133

although for detailed discussion see Evans (2009; 2013: chapter 2; in press a). 134 Research in cognitive science has definitively established that humans are not alone in 135

possessing conceptual systems (see Hurford 2007; and Evans 2014 for overviews). A 136 conceptual system, which serves as our repository for concepts, is essential for a wide array 137 of tasks upon which basic mental function is contingent. These include categorisation, 138

learning, choice, wayfinding, and reasoning. Many non-human species have fairly 139 sophisticated reasoning capabilities (see Hurford 2007 for a review). 140

However, while other species have a conceptual system, humans are the only species 141 with language (Evans 2014). Moreover, language appears to be an evolutionary trait that is 142

specific to the genus Homo, common to humans as well as, on some accounts, extinct species 143 of our genus (e.g., Levinson and Gray 2012). In previous work I have argued that as the 144 human conceptual system must have preceded language in evolutionary terms, it is plausible 145 that language evolved in order to serve as an executive control system on the conceptual 146 system (Evans 2009). Language harnesses representations in the conceptual systemwhich 147 evolved for other purposesin order to facilitate linguistically-mediated representation. For 148 this to have happened, it stands to reason that representations in the linguistic system149 semantic structureis qualitatively distinct from representations in the conceptual system150

4

conceptual structure. While representations, or concepts, in the conceptual system are rich in 151

nature, deriving from sensory-motor and interoceptive experiences in which they are directly 152 grounded, representations in the linguistic system are schematic or much sparser in terms of 153 the perceptual detail that they encode (Evans 2009; 2013). 154 From the present perspective, words are in fact cues that index or point to body-based 155

states, processed and stored by the brain (Evans 2009, 2013; Glenberg and Roberston 1999; 156 Fischer and Zwaan 2008). To illustrate, consider the use of red in the following example 157 sentences (adapted from Zwaan 2004): 158 159 (4) a. The school teacher scrawled in red ink all over the pupils homework book 160 b. The red squirrel is in danger of extinction in the British Isles 161 162 In the first example, the use of red evokes a bright, vivid red. In the second, a dun or browny 163 red is most likely evoked. This illustrates the following: The meaning of red is not, in any 164 sense, there in the wordalthough I nuance this view below. Rather, words cue perceptual 165 and interoceptive states stored in the conceptual system. And these body-based and 166 subjective states are re-activated during language use. Put another way, the word form red 167

gives rise to distinct re-activations, or more technically simulations for different hues of red. 168 These simulations arise as a consequence of reactivating stored experience types in the 169

conceptual system. These reactivated experiences we might refer to as analogue concepts170 concepts that are directly grounded in the experiences that give rise to them. How then does 171

semantic structure (in language) differ from this level of conceptual structurewhich is to 172 say, from analogue concepts? 173 To illustrate, I consider the use of the adjective red, and the noun redness, in the 174

following examples, adapted from a skin care product advert: 175 176

(5) a. Treat redness with Clinique urgent relief cream. 177 b. Treat red skin with Clinique urgent relief cream. 178 179

Both wordsred and rednessrelate to the same perceptual state, presumably. But the 180 words packagewhich is to say, serve to construethe content in a different way, giving 181 rise to distinct simulations. In the example in (5a), redness gives rise to an interpretation 182 relating to a skin condition. In the second, (5b), red refers more straightforwardly to an 183 unwanted property of the skin. 184 The different interpretations arising from these sentences are not due to a different 185

hue being indexed. Rather, the wordsnoun versus adjectivemodulate the perceptual hue 186 in a slightly different way, giving rise to distinct simulations: skin condition versus 187 discolouration of skin interpretations. In other words, the grammatical distinction between 188 red (adjective) and redness (noun) appears to relate to a semantic distinction between the 189 notion of property versus thing. The words red and redness, while indexing the same (or 190

similar) perceptual state, also encode schematic concepts: PROPERTY versus THING (cf. 191 Langacker 2008). 192

But unlike the body-based perceptual statethe hue of redwhich is analogue in 193 nature, PROPERTY and THING are highly schematic notions. They are what I refer to as 194 parametric concepts. Unlike the perceptual experience of redness, with comes to mind when 195 we variously imagine a fire engine, a Royal Mail post box (ubiquitous in the UK), henna, fire, 196 the Chinese flag, or supermans cape, parametric concepts are not like veridical embodied 197 experiences. There is nothing about the (parametric) concepts PROPERTY or THING which is 198 akin to the perceptual experience of redness (an analogue concept). Parameters are abstracted 199

from embodied (= perceptual and interoceptive) states, filtering out all points of difference to 200

5

leave highly image schematic content: the parameter.1 The word form red encodes the 201

parameter PROPERTY, while redness encodes the parameter THING. This is another way of 202 saying that red is an adjectiveit describes a property of a thingwhile redness is a noun203 it describes a property that is objectified in some way, and established as being identifiable, 204 in principle, in its own right, independent of other entities in world. Figure 1 captures the 205

relationship between a word form, and its parametric and analogue concepts. 206 207

208 Analogue concept: 209

210 211 212 213 214 Parametric concepts: 215 216 217 218 Word forms: 219 220

Figure 1. Analogue and parametric concepts 221 222

223 My claim, then, is this. There is a distinction between analogue concepts on the one 224 hand, and parametric concepts on the other. The former relate to non-linguistic concept-types 225

that, in evolutionary terms, had to precede the existence of language. Parametric concepts, on 226 the other hand, constitute a species of concept that arose as a consequence of the emergence 227

of language. They provide a level of schematic representation directly encoded by language: 228 parametric concepts guide how analogue concepts are activated, and, consequently, how 229

simulations are constructed in the service of linguistically-mediated meaning construction. 230 For instance, the forms red and redness both index the same perceptual state(s). But 231 they parcellate the conceptual content in a different way, giving rise to distinct simulations: 232

redness = condition; red = (unwanted) property of skin. The schematic parametric concepts, 233 which is to say, that part of meaning that is native to language, relates to THING versus 234

PROPERTY. Parametric concepts are language-specific affordances, rather than affordances of 235 the conceptual system. 236

Related proposals have been put forward by Bergen (2012), and Taylor and Zwaan (e.g., 237

2008, 2009). Taylor and Zwaan have captured this view in terms of what they dub the 238 Linguistic Focus Hypothesis. They argue that during language understanding, motor 239 representations are activated that are under the governance of linguistic constructions. These 240

serve to differentially direct focus on the referential world. Bergens findings are consonant 241 with this hypothesis. In one set of behavioural experiments, Bergen (2012: 114) found that 242 the grammatical subject, for instance, the use of I versus you, influences the perspective that a 243 language user perceives a scene from. 244 In the light of this discussion, what then is the function of language, and specifically, 245

parametric concepts in meaning construction? My answer is that parametric concepts, 246

1 Cf. the related notion of image schema developed in the work of Johnson (1987)

PROPERTY THING

red redness

6

encoded by language, guide the formation of complex simulations for purposes of 247

linguistically-mediated communication. Parametric concepts guide the parcellation (focal 248 adjustments, in Langackers 2008 terms) of analogue (i.e., perceptual and interoceptive) 249 concepts, in the construction of simulations. Parametric concepts encode schematic, which is 250 to say, digitised content. Content of this sort is abstracted from analogue, which is to say, 251 perceptual and interoceptive representations. Hence, the parameters THING versus PROPERTY 252 are schemas drawn from embodied experience. Table 1 represents a summary of the 253 distinction between parametric and analogue concepts. 254 255

256

Parametric concepts Analogue concepts

Specific to language Parametric (abstracted from

embodied states, filtering out all points of difference to leave a highly schematic properties or parameters)

Underpin all linguistic units (where a linguistic unit is a form/parametric content unit of any complexity)

Specific to the conceptual system Analogue (albeit attenuated)

representations of body-based states

Arise directly from perceptual (conscious experience), and reside in

the same neural system(s) as body-

based states

Re-activated or simulated (by language, imagination, etc., and can

be combined to form complex and

novel simulations

257

Table 1. Parametric versus Analogue Concepts 258 259

260

3. Towards an account of meaning construction 261 Having distinguished between analogue and parametric concepts, we now require an account 262 of their respective contribution to the meaning construction process. This will facilitate my 263 account of polysemy in subsequent sections. 264 265 3.1. LCCM Theory 266 In two book-length treatments (Evans 2009, 2013) I have developed a theoretical account of 267 lexical representation and semantic composition dubbed the Theory of Lexical Concepts and 268 Cognitive Models, or LCCM Theory for short. The claim at its heart is enshrined in the 269 distinction between its two foundational theoretical constructsthe lexical concept and 270 cognitive model: there is a qualitative distinction between the representations captured by 271 these theoretical constructs. This distinction relates, ultimately, to the bifurcation between 272 analogue versus parametric concepts, which respectively structure lexical concepts and 273 cognitive models. 274 In keeping with the thrust of the argument developed in the previous section, LCCM 275 Theory assumes the linguistic system emerged, in evolutionary terms, much later than the 276 earlier conceptual system. The utility of a linguistic system, on my account, is that it provides 277 an executive control mechanism facilitating the deployment of conceptual representations in 278 service of linguistically-mediated meaning construction. Hence, semantic representations in 279 the two systems are of a qualitatively distinct kind. I model semantic structurethe primary 280 representational substrate of the linguistic systemin terms of the theoretical construct of the 281 lexical concept. A lexical concept is a component of linguistic knowledgethe semantic pole 282 of a symbolic unit (in Langackers e.g., 1987 terms)which encodes a bundle of various 283

7

types of highly schematic linguistic content (see Evans 2006, 2009). Hence, lexical concepts 284 are parametric in nature. 285 While lexical concepts encode highly schematic linguistic content, a subsetthose 286 associated with open-class formsare connected, and hence facilitate access, to the 287 conceptual system. Lexical concepts of this type are termed open-class lexical concepts.2 288 Such lexical concepts are typically associated with multiple areas in the conceptual system, 289 referred to as association areas. 290 The range of association areas to which a given lexical concept facilitates access is 291 termed an access site. LCCM Theory assumes that the access site for any given open-class 292 lexical concept is unique. As lexical concepts facilitate access to a potentially large number of 293 association areas in the conceptual system, any given open-class lexical concept, in principle, 294 facilitates access to a large semantic potential. However, only a small subset of this semantic 295 potential is typically activated in interpretation of a given utterance. 296 In LCCM Theory, conceptual structurethe semantic representational substrate of 297 the conceptual systemis modelled by the theoretical construct of the cognitive model. A 298 cognitive model is a coherent body of multimodal knowledge directly grounded in the brains 299 modal systems, and derives from the full range of experience types processed by the brain 300 including sensory-motor experience, proprioception and subjective experience including 301 affect. Hence, cognitive models are analogue in nature, and as such are constituted by 302 analogue concepts. 303 The conceptual content encoded as cognitive models can become re-activated during 304 the simulation process. Simulation, as implied in earlier discussion, is a general purpose 305 computation performed by the brain in order to implement the range of activities that sub-306 serve a fully functional conceptual system.3 307 308 3.2. Cognitive Model Profile 309 An important construct in LCCM Theory, and one that is essential to providing an account of 310 both polysemy and meaning construction, is that of the cognitive model profile. As an open-311 class lexical concepta noun, verb, adjective or adverbfacilitates access to numerous 312 association areas within the conceptual system, it facilitates access to numerous cognitive 313 models. Moreover, the cognitive models to which a lexical concept facilitates access are 314 themselves connected to other cognitive models. The range of cognitive models to which a 315 given lexical concept facilitates direct access, and the range of additional cognitive models to 316 which it therefore facilitates indirect access is termed its cognitive model profile. 317 To illustrate, consider the cognitive model profile for the lexical concept which I gloss 318 as [FRANCE] associated with the form France. A partial cognitive model profile for [FRANCE] 319 is represented in Figure 2. 320 321

322 323

324 325

326 327 328 329

2 See Evans (2009) for the rationale for this position.

3 For discussion and findings relating to the multimodal nature of conceptual representations and the role of

simulation in drawing on such representations in facilitating conceptual function see, for instance, Barsalou

(1999, 2008), Glenberg (1997), Gallese and Lakoff (2005), and references therein.

8

330

331 332 333 334

335 336 337 338 339

340 341 342 343 344

345 346

347 348

349 350

Figure 2. Partial cognitive model profile for [FRANCE] 351 352 353

Figure 2 is an attempt to capture (a subset of) the sort of knowledge that language 354 users must presumably have access to when speaking and thinking about France. The lexical 355 concept is denoted by the term in small caps and in square brackets: [FRANCE]. This is a 356 mnemonic for the semantic content encoded by the lexical concept, which may include a 357 cluster of parametric knowledgeparametersabout which I shall have more to say in the 358 next section. 359 In contrast, cognitive models are denoted by terms in small caps without square 360 brackets. As illustrated by Figure 2, the lexical concept [FRANCE] provides access to a 361 potentially large number of cognitive models.4 As each cognitive model consists of a complex 362 and structured body of knowledge, which, in turn, provides access to other sorts of 363 knowledge, LCCM Theory distinguishes between cognitive models which are directly 364 accessed via the lexical conceptprimary cognitive modelsand those cognitive models 365 which form sub-structures of those which are directly accessedsecondary cognitive models. 366 These secondary cognitive models are indirectly accessed via the lexical concept. 367 The lexical concept [FRANCE] affords access to a number of primary cognitive 368 models, which make up the primary cognitive model profile for [FRANCE]. These are 369 hypothesised to include: GEOGRAPHICAL LANDMASS, NATION STATE and HOLIDAY 370 DESTINATIONand I reiterate, a cognitive model represents a coherent body of complex 371 information: multimodal information, gleaned through sense-perception, interoceptive 372 experience, and through propositional information achieved via cultural learning, language 373 and other channels. Each of these cognitive models provides access to further cognitive 374 models. 375 In Figure 2, a flavour of this is given by virtue of the various secondary cognitive 376 models which are accessed via the nation state cognitive modelthe secondary cognitive 377

4 Note that the abbreviation [FRANCE] represents the linguistic content that is encoded by the vehicle France.

GEOGRAPHICAL LANDMASS

NATION STATE

CONSTITUTIONAL

SYSTEM

POLITICAL

SYSTEM

[FRANCE]

ELECTORATE

HOLIDAY

DESTINATION

HEAD OF STATE

NATIONAL SPORTS CUISINE

9

model profile. These include NATIONAL SPORTS, POLITICAL SYSTEM and CUISINE. For instance, 378 we may know that in France, the French engage in national sports of particular types, for 379 instance, football, rugby, athletics, and so on, rather than others: the French dont typically 380 engage in American football, ice hockey, cricket, and so on. We may also know that as a 381 sporting nation they take part in international sports competitions of various kinds, including 382 the FIFA football world cup, the Six Nations rugby competition, the rugby world cup, the 383 Olympics, and so on. 384 That is, we may have access to a large body of knowledge concerning the sorts 385 of sports French people engage in. We may also have some knowledge of the funding 386 structures and social and economic conditions and constraints that apply to these sports in 387 France, Frances international standing with respect to these particular sports, famous French 388 sportsmen and women, and further knowledge about the sports themselves including the rules 389 that govern their practice, and so forth. This knowledge is derived from a large number of 390 sources including direct experience and through cultural transmission (including language). 391 With respect to the secondary cognitive model of political system, Figure 2 illustrates 392 a sample of further secondary cognitive models which are accessed via this cognitive model. 393 Hence, each secondary cognitive model has further (secondary) cognitive models to which it 394 provides access. For instance, (FRENCH) ELECTORATE is a cognitive model accessed via the 395 cognitive model (FRENCH) POLITICAL SYSTEM. In turn the cognitive model (FRENCH) 396 POLITICAL SYSTEM is accessed via the cognitive model NATION STATE. Accordingly, NATION 397 STATE is a primary cognitive model while ELECTORATE and POLITICAL SYSTEM are secondary 398 cognitive models.5 399 The utility of the LCCM Approach is that it provides a ready means of accounting for 400 meaning construction. To illustrate, consider the following sentences all involving the lexical 401 item France. 402 403

(6) a. France is a country of outstanding natural beauty. 404 b. France is one of the leading nations in the European Union. 405 c. France beat New Zealand in the 2007 Rugby world cup. 406 d. France voted against the EU constitution in the 2005 referendum. 407 408 In each of these examples, the semantic contribution associated with the form France is 409 slightly distinct: the reading for France varies across these distinct utterances. France, in (6a) 410 has a geographical landmass reading; in (6b) it is France as a political entity, a nation state; in 411 (6c), it is the 15 players who make up the French Rugby team; and in (6d) the reading 412 involves that proportion of the French electorate who voted non when presented, in a 413 national referendum, with the proposal to endorse a constitution for the European Union. The 414 key insight of LCCM Theory is that the reason for this variation is due to differential 415 activation of non-linguistic knowledge structures within the cognitive model profile to which 416 the lexical concept associated with France affords access. 417 The differential readings associated with the examples in (6) arise as follows. In (6a) 418 the interpretation associated with the form France, which relates to a particular geographical 419 region, derives from activation of the GEOGRAPHICAL LANDMASS cognitive model. Individual 420 language users have knowledge relating to the physical aspects of France, including its 421 terrain, and its geographical location. In this example, the utterance context serves to activate 422 this part of the cognitive model profile accessed by the lexical concept [FRANCE]. In the 423 second example, the utterance context serves to activate a different part of the cognitive 424

5 The rationale for distinguishing between primary and secondary levels of cognitive models has been laid out in

detail elsewhere (e.g., Evans 2013: Chapter 2).

10

model profile to which the lexical concept [FRANCE] affords access. In this example, the 425 reading derives from activation of the NATION STATE cognitive model. The use of France in 426 the example in (6c) relates to the group of 15 French individuals who play as a team and 427 thereby represent the French nation on the rugby field. In the example in (6d) the form 428 France relates not to a geographical landmass, nor a political entitya nation-statenor to a 429 group of 15 rugby players who happen to be representing the entire population of France. 430 Rather, it relates to that portion of the French electorate that voted against ratification of the 431 EU constitution in a referendum held in 2005. Accordingly, what is activated here is the 432 ELECTORATE cognitive model. 433 This last example provides an elegant illustration of the way in which activation of a 434 cognitive model serves to provide a situated interpretation of a lexical concept by giving rise 435 to an access route through the semantic potential. In this example, interpretation requires that 436 an access route is established through the cognitive model profile accessed via the lexical 437 concept [FRANCE] in a way that is consistent with the lexical concepts associated with the 438 other linguistic forms and units in the utterance. The interpretation associated with France in 439 this example has to do with the French electorate, and specifically that part of the French 440 electorate which voted against ratification of the EU constitution. In other words, [FRANCE] 441 in this example achieves an interpretation which is facilitated by activating the cognitive 442 models which are shown in bold in Figure 3. 443 444 445

446 447 448

449 450

451 452 453

454

455 456 457

458 459

460 461 462

463 464

465 Figure 3. Access route established by the interpretation of [FRANCE] in the utterance France 466

voted against the EU constitution 467 468 An important consequence of assuming a distinction between primary and secondary 469 cognitive models relates to figurative language. Specifically, some aspects of the distinction 470 between literal and figurative language can be elegantly accounted for. For instance, consider 471 knowledge representation for the celebrated French novelist, critic and essayist Marcel 472 Proust. Many native speakers of English may only be dimly aware of Prousts literary 473 contribution. Speakers in this category may simply know that Proust was a French literary 474

GEOGRAPHICAL LANDMASS

NATION STATE

CONSTITUTIONAL

SYSTEM

POLITICAL

SYSTEM

[FRANCE]

ELECTORATE

HOLIDAY

DESTINATION

HEAD OF STATE

NATIONAL SPORTS CUISINE

11

figure. They may be unaware precisely when he lived, what his literary output related to, and 475 indeed any other information about him. Cognitive model profiles relating to Proust, for 476 these speakers, will involve knowledge inherited from type cognitive models. Such cognitive 477 models facilitate inheritance of content in order to populate a cognitive model profile for an 478 individual. In this case, a schematic cognitive model profile will be derived. Such a cognitive 479 model profile is presented in Figure 4. 480 481

482 483

Figure 4. Schematic cognitive model profile for [PROUST] 484 485 486

In the schematic cognitive model profile in Figure 4, there are at least two primary 487 cognitive models, for MAN and AUTHOR respectively. Each will consist of a range of 488 attributes, inherited from type cognitive models for man and author. For instance, the type 489 cognitive model for MAN will include generic information relating to aspects of physiology, 490 appearance, personality, socio-cultural role, dress, behavioural traits, and so on. The 491 schematic cognitive model for AUTHOR will include generic information relating to the 492 generic habits and qualities associated with being an author, the nature of the activities 493 engaged in, potential for success, and so on. A salient secondary type cognitive model also 494 inherited by the schematic cognitive model profile is likely to relate to BODY OF WORK. This 495 might include generic knowledge about the type and nature of the output associated with 496 being an author, some information about the publishing process, the requirement to have a 497 literary agent, the role of booksellers, and so on. 498 Now consider the following sentences: 499 500

(7) a. Proust had a moustache. 501 b. Proust is difficult to read. 502 503 The sentence in (7a) gives rise to a reading in which the man identified as Proust had a 504 moustache. In contrast, the example in (7b) relates not to the man per se, but rather to his 505 literary output. That is, in an example such as this Proust would normally be taken as 506 referring not to the man, but rather to the literary works produced by Proust the man. 507 Moreover, the interpretation of Proust in (7a) would normally be judged to be literal, while 508

[PROUST]

MAN AUTHOR

BODY OF WORK

12

the interpretation in (4b) would be judged as figurative, and more specifically an instance of 509 metonymy: Proust stands for the works created by the manPRODUCER FOR PRODUCT. 510 A central claim of LCCM Theory is that one reason for the distinction in literal versus 511 figurative interpretations is a consequence of the cognitive model profile, and a distinction, 512 therefore, in terms of the range of analogue concepts directly and indirectly accessed by the 513 lexical concept. Literal interpretations involve activation of a primary cognitive modelin 514 this case MANwhile figurative interpretations involve activation of secondary cognitive 515 modelsin this case BODY OF WORK. And intuitively, it does seem as if there is some sense 516 in which the body of literary output is more peripherally accessed by the lexical concept 517 [PROUST], than that of being a human male, a man, and having a particular profession, namely 518 being an author. In other words, the explicit claim made by LCCM Theory is that cognitive 519 model profiles accessed by open-class lexical concepts exhibit a qualitative distinction 520 between cognitive models that are in some sense more central to the knowledge associated 521 with, for instance, Proust, and knowledge that is less central. While there is unlikely to be a 522 neat distinction between primary and secondary cognitive models, and while the distinction is 523 likely to vary from individual to individual, and indeed across discourse communities, there 524 appears to be grounds for making a qualitative distinction of this sort. 525 526 4. Conceptual polysemy 527 Now lets return to the first of the phenomena under the microscope in this paper: conceptual 528 polysemy. I reproduce the relevant data below. 529 530 Phenomenon 1: 531 (8) a. Thats a heavy book tome 532 b. That antiquarian book is illegible text 533 c. Thats a boring book level of interest 534 d. Thats a long book duration 535 536 We now begin to see that polysemy of this sort is a consequence of differential activation of 537 regions of the cognitive model profilethe vast semantic potentialto which the lexical 538 concept [BOOK] facilitates access. To see how this works, lets examine the sorts of 539 knowledge that the lexical concept [BOOK] must afford access to. 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555

Figure 5. The relationship between lexical concepts, cognitive models, and attributes and 556 values 557

558

PHYSICAL

STRUCTURE

READING ACTIVITY

TOME TEXT DURATION

[BOOK]

Attributes

Cognitive models

Lexical concept

READER

LEVEL OF

INTEREST

13

559

Lets consider the cognitive models accessed via [BOOK]. As illustrated in the partial 560 cognitive model profile given in figure 5, the knowledge accessed by [BOOK] includes, at the 561 very least, that a book is a physical entity and is interacted with via a process of reading. 562 These two distinct sorts of knowledgeknowledge relating to an artefact, on one hand, and 563 the process of reading, on the otherare captured in Figure 5 by the two cognitive models: 564 PHYSICAL STRUCTURE and READING ACTIVITY respectively. The two cognitive models are 565 related by virtue of a readerthe entity that interacts with the physical artefact by handling 566 the tome and reading the printed text. Relationships of this sort holding between cognitive 567 models I refer to as a structural invariant: a stable knowledge structure that is relational in 568 nature. I capture the structural invariant in Figure 5 by a double-headed arrow, and the 569 specific relation involved is signalled by the mnemonic READER. In addition, cognitive 570 models consist of a large, detailed, but structured, body of knowledge: what I refer to as 571 attributes. Figure 5 provides two attributes for each of the cognitive models which [BOOK] 572 provides access to. 573 The cognitive model PHYSICAL STRUCTURE relates to the physical artefact, consisting 574 of, at the very least, knowledge as to the physical structure and organisation of a given book. 575 This includes detailed knowledge concerning the material aspects of the artefact, including its 576 dimensions, weight, binding (paper or cloth), and so forth. This aspect of our knowledge 577 about books I refer to as the TOME attribute. In addition to the physical organisation and 578 construction of a book, books consist of text which is interacted with through the process of 579 reading. This I refer to as the TEXT attribute. 580 The READING ACTIVITY cognitive model relates to the process involved in interacting 581 with books, especially the nature of the interaction with the text itself. One consequence of 582 this interaction is that reading takes up a period of time, which I refer to as the DURATION 583 attribute. That is, depending on the amount of text involved, reading can take lesser or 584 greater amounts of time. Another consequence of interaction with books is the level of 585 interest that a given book holds for the reader. This I refer to as the LEVEL OF INTEREST 586 attribute. While the reader might judge the book to be interesting another might be judged to 587 be boring, and so on. 588 Now lets return to the specific type of polysemy under the spotlight in this section. 589 Each of the utterances in (8) involves a distinct interpretation of the [BOOK] lexical concept. 590 This is achieved by virtue of each instance of book being interpreted in a way consistent with 591 the utterance context: consequently, a slightly different access route is established through the 592 cognitive model profile accessed via the lexical concept [BOOK]. 593 For instance, the readings that result from (8a) and (8b) have to do with activation of 594 the PHYSICAL STRUCTURE cognitive model. However, each involves differential activation of 595 attributes associated with this cognitive modela process I term highlighting (Evans 2009). 596 While the reading associated with book in (8a) involves highlighting of the TOME attribute, 597 the reading associated with book in (8b) involves highlighting of the TEXT attribute. 598 In contrast, the readings that result from the utterances in (8c) and (8d) have to do 599 with activation of the READING EVENT cognitive model accessed via [BOOK]. The reading 600 associated with book in (8c) results from highlighting of the DURATION attribute. The reading 601 associated with book in (8d) results from highlighting the LEVEL OF INTEREST attribute. 602 603

5. Lexical concepts in LCCM Theory 604 In order to set the scene for my analysis of the state lexical concepts, in the remainder of 605 this paper, in this section I present the theoretical assumptions that underpin the construct of 606 the lexical concept within LCCM Theory. 607 608

14

609 5.1. The nature of the lexical concept in LCCM Theory 610 As intimated earlier, lexical conceptsunlike cognitive modelsare constituted of linguistic 611 knowledge. My premise is that linguistic knowledge is usage-based. That is, I assume that 612 the organisation of our language system is intimately related to, and derives directly from, 613

how language is actually used (Croft 2000; Langacker 2000; Tomasello 2003). Through 614 processes of abstraction and schematisation (Langacker 2000), based on pattern-recognition 615 and intention-reading abilities (Tomasello 2003), language users derive linguistic units. 616 These are relatively well-entrenched mental routines consisting of conventional pairings of 617 form and semantic representation. The semantic representations conventionally associated 618

with a given unit of form, I refer to, as already noted, as a lexical concept. 619 While lexical concepts are mental representations, they underspecify the range of 620

situated meanings associated with a given form in an individual utterance. Thus, I make a 621 fundamental distinction between lexical concept as a mental unit, and its context-dependent 622 realisation in an utterance. This is akin to the distinction in Phonological Theory between the 623

abstract notion of a phoneme and the actual unit of realised context-dependent sound, the 624 allophone. My claim is that there is an essential distinction between lexical representation 625

and meaning. While meaning is a property of the utterance, lexical representations are the 626 mental abstractions which we infer must be stored as part of the language users knowledge 627 of language, in order to produce the range of novel uses associated with situated instances of 628 a particular word such as a preposition. The meaning associated with an utterance I refer to 629

as a conception. Thus, conceptions are a function of language use. 630 There are a number of important properties associated with lexical concepts. I briefly 631

review some of the most relevant here (for detailed discussion see Evans 2009). Firstly, and 632

as noted above, linguistic units, as I use the term, are conventional pairings of form and 633 meaning. From this it follows that lexical concepts are form-specific. Secondly, as 634

mentioned above, although lexical concepts are form-specific, a single form can be 635 conventionally associated with a potentially large number of distinct lexical concepts which 636 are related to degrees as attested by the phenomenon of polysemy.

6 That is, forms are not 637

lexical concept-specific. A consequence of this is that the lexical concepts which share the 638

same form can be modelled in terms of a semantic network (see Evans and Green 2006: 639 chapter 10 for discussion). 640

Thirdly, the definitional property of any given lexical concept is that it has a lexical 641

profile, its unique biometric identifier. A lexical profile is an extension of criteria presented 642 in Evans (2004), and akin to the notion of an ID tag (Atkins 1987) and behavioural profile 643 (Gries 2005). While a lexical concept associated with a particular form can be provided with 644 a semantic gloss, as in the case of lexical concepts associated with over, an example of which 645 I glossed as [ABOVE] or the lexical concepts associated with in, at and on to be examined later 646

which I preliminarily gloss as [STATE], whether a particular usage of a form relates to one 647 lexical concept rather than another is a matter of examining the selectional tendencies (the 648 lexical profile) associated with a given usage. While any given usage of a lexical concept 649 will have its own unique collocational pattern, general patterns can be established, and form 650

part of the conventional knowledge associated with a particular lexical concept. 651 Two sorts of information form a lexical concepts lexical profile. The first relates to 652

semantic selectional tendencies (Evans 2009). The second relates to formal or grammatical 653 selectional tendencies (Evans 2009). Gries (2005) has advocated the way in which corpus 654 methodologies can be used to examine the lexical profile associated with a specific lexical 655

6 See Evans 2005 and Tyler and Evans 2001, 2003 for detailed discussion of polysemy.

15

concept. For instance, each of the state lexical concepts associated with in and on have 656 distinct lexical profiles. 657

To provide a preliminary illustration of the construct of the lexical profile, I briefly 658 consider two lexical concepts, both of which I provisionally gloss as [STATE]although I 659 revise this gloss later in the chapterand which are conventionally encoded by the English 660 prepositional forms in and on. These are evidenced by the following examples: 661 662 (10) a. John is in trouble/danger 663 b. Jane is in love/awe 664 c. Fred is in shock 665

d. Jake is in a critical condition 666 667 (11) a. The guard is on duty 668 b. The blouse is on sale 669 c. The security forces are on red alert 670

671 While both in and on have state lexical concepts conventionally associated with them, the 672 lexical profile for each is distinct. For instance, the [STATE] lexical concept associated with 673 on selects semantic arguments which relate to states which normally hold for a limited period 674

of time, and which contrast with salient (normative) states in which the reverse holds. For 675 instance, being on duty contrasts with being off-duty, the normal state of affairs. Equally, 676 being on sale is, in temporal terms, limited. Sales only occur for limited periods of time at 677 specific seasonal periods during the year (e.g., a winter sale). Similarly, being on red alert 678 contrasts with the normal state of affairs in which a lesser security status holds. Further, the 679

states in question can be construed as volitional, in the sense that to be on duty/sale/red alert 680 requires a volitional agent who decides that a particular state will hold and takes the requisite 681

steps in order to bring such a state of affairs about. 682 In contrast, the semantic arguments selected for by the [STATE] lexical concept for in 683 relates to states which do not necessarily hold for a limited period of time, and do not 684

obviously contrast with a normal state of affairs. Moreover, while states encoded by on are 685 in some sense volitional, states associated with in are, in some sense, non-volitional. That is, 686 we do not usually actively choose to be in love, shock or a critical condition, nor can we, by a 687 conscious act of will, normally bring such states about. That is, these states are those we are 688

affected, constrained and influenced by, rather than those which are actively (in the sense of 689 consciously) chosen. 690

The fourth and final property of lexical concepts that I review here concerns the 691 position that they have bipartite organisation. That is, lexical concepts encode linguistic 692 content and facilitate access to conceptual content. Linguistic content represents the form 693

that conceptual structure takes for direct encoding in language, and constitutes what might be 694 thought of as a bundle of distinct knowledge types. There are a large number of different 695 properties encoded by linguistic content which serve to provide a schematic or skeletal 696 representation, which can be encoded in language (see Evans 2009: chapter 4). The one 697

which is relevant for the present study relates to the notion of parameterisation. 698 One way in which knowledge, in general terms, can be represented is in terms of 699 richly inflected nuances that serve to reflect the complexity of experience. An alternative 700 way is to compress such fine distinctions into two, three or more, much broader, and hence, 701 more general distinctions. These I refer to as parameters. Linguistic content serves to encode 702

content by adopting the latter strategy, which is to say, to employ parameterisation. 703 Parameters are hence part of the bundle of information that a lexical concept serves to 704 encode. 705

16

To illustrate this notion, consider the complex range of expressions that a language 706

user might employ, in English, in order to locate themselves with respect to time, thereby 707 facilitating time-reference. Any one of the following could conceivably be employed, 708 depending upon context: today, May, 2014, the day after yesterday, the day before tomorrow, 709 this moment, now, this second, this minute, this hour, today, this week, this month, this 710

quarter, this year, this half century, this century, this period, the 2nd

day of the month, this 711 era, this millennium, and so on. A potentially unlimited set of finer and finer distinctions can 712 be made (e.g., 1 second ago, 2 seconds ago, 1 hour 4 minutes and 3 second ago, 2 days ago, 713 etc.), reflecting any manner of temporal distinction we might care to make. 714 In contrast, paramaterisation functions by dividing all the possible permutations 715

relating to a given category, such as time-reference, into a small set of divisions: parameters. 716 Such parameters might distinguish between the past, for instance, and the non-past. Indeed, 717 this is the basis for the tense system in English, as illustrated by the following: 718

719 (12) a. He kicked the ball Past 720

b. He kicks the ball Non-past 721 722

English encodes just two parameters that relate to Time-reference: Past versus Non-past, as 723 exhibited by the examples in (12), and thus manifests a binary distinction. Some languages, 724

such as French, have three parameters: Past, Present and Future. Some languages have more 725 than three parameters, distinguishing additionally remote past from recent past, for instance. 726

The language with the most parameters for linguistically encoding time-reference is an 727 African language: Bamileke-Dschang with eleven. Crucially, parameters are encoded by 728 specific lexical concepts, and thus form part of the knowledge bundle that constitutes a 729 lexical concept. For instance, the parameter past is encoded by the lexical concept 730 associated with the ed form in (12a). However, other lexical concepts also include the 731 parameter past such as the lexical concepts associated with the following forms: sang, lost, 732 went, etc. 733

I argue, then, that a key feature of linguistic (as opposed) to conceptual content is that 734

it only encodes knowledge in parametric fashion. This is not to say that conceptual content 735

does not parameterise knowledge. Indeed, parameterisation is simply a highly reductive form 736 of abstraction: it serves to abstract across the complexity exhibited by a particular category. 737 The point, however, is that the parameters encoded by linguistic content serves to strip 738 away most of the differences apparent in the original perceptual experience, thereby 739 reducing it to a highly limited numbers of parameters. 740

741 5.2. Two factors in accounting for state lexical concepts: Lexical profiles and parameters 742 In this sub-section, I further develop the notion of lexical profiles and parameters, and explain 743 how LCCM Theory views these constructs as giving rise to distinct state lexical concepts. 744 There are two aspects of linguistic content that are relevant for the discussion of the 745 inter-lexical polysemy exhibited by the range of state lexical concepts in this study. The 746 first concerns the lexical profile exhibited by lexical concepts, as manifested by distinct 747 collocational patterns in language use. As we saw earlier, two sorts of information form a 748 lexical concepts lexical profile: semantic selectional tendencies, and formal or grammatical 749 selectional tendencies. In this remainder of this paper, I employ distinctions in the semantic 750 arguments which, I hypothesise, collocate with distinct state lexical concepts to uncover 751 distinctions in lexical concepts both within and between prepositions. 752 The second aspect of linguistic content that will be relevant relates to 753 parameterisation. One characteristic that serves to distinguish between lexical concepts, 754 both across prepositions and within a single preposition, relates to the parameters encoded. 755

17

For instance, the prototypical spatial lexical concept associated with in, which I gloss as 756 [ENCLOSURE], encodes the parameter Containment, as evidenced by the example in (13). In 757 contrast, the [PSYCHO-SOMATIC] lexical conceptone of the state lexical concepts 758 associated with inencodes the parameter Psycho-somatic State, as evidenced in (14), but 759 not the Containment parameter. 760 761 (13) The kitten is in the box Parameter: Containment 762 (14) John is in love Parameter: Psycho-somatic state 763 764 That is, the [ENCLOSURE] lexical concept in (13) encodes a schematic dimension abstracted 765 from sensory-motor experience in which a trajectorythe motile entity (TR)is contained 766 by the landmarkthe entity with respect to which the TR is foregrounded (LM). Notice that 767 the relation encoded is highly schematic in nature; it says nothing about whether there is 768 contact or not between the TR and LM as in (15), nor as to whether the TR represents part of 769 the LM or not as in (16): 770 771 (15) a. The fly is in the jar (i.e., flying around) 772 b. The fly is in the jar (i.e., stationary on one interior surface) 773 (16) Theres a crack in the vase 774 775 Indeed, the precise spatio-geometric nature of the TR, LM and their relationship is a function 776 of the TR and LM and their possible forms of interaction, rather than the abstract parameter 777 encoded by the [ENCLOSURE] lexical concept associated with the prepositional form in. This 778 information derives from the semantic potential accessed via the open-class lexical concepts, 779 as mediated by compositional processes (see Evans, 2009, for details). 780 In contrast, the [PSYCHO-SOMATIC] lexical concept associated with in encodes the 781 parameter Psycho-somatic state. This information is highly schematic in nature. The 782 parameter encoded does not determine which sorts of psycho-somatic states can collocate 783 with this lexical concept. This is a function of the lexical profile: knowledge relating to the 784 semantic selectional tendencies associated with this lexical concept, and hence the range of 785 psycho-somatic states which can co-occur with the [PSYCHO-SOMATIC] lexical concept. 786 Hence, while the parameters encoded by a lexical concept determine the possible range of 787 semantic arguments that can co-occur, the lexical profile, which relates to stored knowledge 788 based on usage-patterns, provides information relating to the range of permissible states 789 which can co-occur with this lexical concept. 790 In previous work, I introduced a methodology, exploiting the constructs of 791 parameterisation and lexical profiles, in order to provide a falsifiable basis for adducing 792 distinct lexical concepts associated with a linguistic form (see Evans 2009: pp. 141-148). 793 While space precludes a detailed illustration of this methodology, my analysis of distinct 794 state lexical concepts, in the remainder of this paper, is underwritten by it. Of course, future 795 empirical researchdeploying methodologies from psycholinguistics and corpus 796 linguisticswill provide the requisite empirical evidence. And this converging evidence may 797 require revision of the linguistic methodology I have developed. Nevertheless, the essential 798 insight that I adduce, below, relating to the phenomenon of inter-lexical polysemy, is stands 799 regardless of the detail of the distinct lexical concepts for in and on that I argue for. 800 801 5.3. Functional consequences of parameters 802 I now conclude this section of the paper by considering how the state lexical concepts arise 803 from historically earlier spatial lexical concepts, giving rise to the phenomenon of polysemy. 804 Put another way, polysemy is a consequence of new, or derived lexical concepts emerging, 805

18

thereby exhibiting a semantic relationship with a synchronically presentalbeit 806 diachronically antecedentlexical concept. 807 Based on arguments developed in Tyler and Evans (2001, 2003) I argue that the 808 spatio-geometric knowledge, encoded, in present terms, as abstract parameters by the spatial 809 lexical concepts associated with prepositional forms gives rise to the development of non-810 spatial lexical concepts. In other words, state lexical concepts emerge by virtue of 811 parameters such as that of Psycho-somatic state arising as a functional consequence of spatio-812 geometric properties, in particular usage contexts. Hence, the emergence of derived lexical 813 concepts is a consequence of the functional consequences of spatio-geometric parameters in a 814 specific context of use. Such contexts of use Tyler and I (2001, 2003) referred to as spatial 815 scenes. 816 For instance, there are a large number of distinct sorts of spatial scenes that involve 817 the prototypical spatial lexical concept: [ENCLOSURE], associated with in, and which hence 818 encode the parameter Containment. This follows as different bounded landmarks: 819 landmarks which exhibit the structural properties interior, boundary and exterior, have 820 different functions, are employed for different ends and are viewed from different vantage 821 points. For instance, while a playpen, prison cell and a coffee cup all restrict the containee to 822 a specific location, they do so in service of different objectives, respectively: safety, 823 punishment and consumption. Hence, without understanding the functional consequence of 824 being located in a bounded landmark such as a prison (cell), the question in (17) would be 825 uninterpretable: 826 827 (17) What are you in for? 828 829

After all, in, here, does not relate directly to a given spatial relation, but rather to the specific 830 sets of knowledge systems relating to the containment function of prison in a particular 831 society. Thus, in (17), being in relates not purely to containment, a functional consequence 832 of the [ENCLOSURE] lexical concept, but rather, and in addition, to punishment, a functional 833 consequence of being contained in enclosures (i.e., bounded landmarks) of a certain kind, i.e., 834 prisons, which occupy a certain position, and fulfil a specified role in the socio-cultural and 835 legal institutions of a particular society. 836 Now consider a different sort of functional consequence associated with the 837 [ENCLOSURE] lexical concept for in. One consequence of certain sorts of bounded landmarks 838 is their utility in providing security. This is evident in the scenario involving a very small 839 child in a playpen for instance. But it is also true of bounded landmarks such as safes used to 840 safeguard valuable commodities such as money or jewels. Indeed, a functional consequence 841 of bounded landmarks of this sort is that the contents are occluded. This of course assumes 842 that the vantage point from which the bounded landmark is viewed is exterior with respect to 843 the volumetric interior of the bounded landmark in question, here the safe. Thus, 844 containment or location with surety is a functional consequence of the spatial relation (i.e., 845 the lexical concept) conventionally associated with in, i.e., of [ENCLOSURE]. 846 The point is, then, that when in is employed in any given utterance, the conception 847 which derives will almost certainly always relate to a functional consequence attendant on a 848 specific sort of spatial scene, involving a containment relation, but doing so in service of 849 objectives and consequences specific to the sort of spatial scene in question. Put another 850 way, bounded landmarks are of many different kinds, a consequence of the many different 851 ways in which we interact with, and the complex range of functions to which we put, 852 bounded landmarks. 853 In terms of the phenomenon of polysemy, which is to say the emergence of derived 854 lexical concepts, it is precisely functional consequences of this sort which give rise to new 855

19

parameters. Such new parameters become conventionally associated with a lexical form, and 856 hence contribute to the formation of a new lexical concept. The occlusion afforded by certain 857 kinds of bounded landmarks, such as a jewellers safe, is a consequence of placing valuables 858 in a landmark that serves to protect the commodity in question. Typically, such landmarks 859 are made of materials that serve to occlude the contents, a consequencerather than the 860 objectiveof employing the types of materials used for constructing the safe. This 861 functional consequence has become abstracted from such spatial scenes to give rise to a 862 distinct parameter. This forms part of the linguistic content encoded by a distinct lexical 863 concept. Evidence for this comes from examples of the following sort: 864 865

(18) The sun is in 866 867 This utterance relates to lack of visibility of the sun, rather than the sun, the TR, being 868 enclosed by a bounded LM of some sort. That is, the functional consequence of certain sorts 869 of containment relations has given rise to a distinct lexical concept which has a Lack of 870 Visibility parameter encoded as part of its linguistic content. 871 872

6. Inter-lexical polysemy 873 Now lets consider inter-lexical polysemy. Perhaps inevitably, this is a somewhat more 874 complex phenomenon to account for, than conceptual polysemy, involving, as it does, distinct 875 lexical items. I reproduce the relevant examples below: 876 877 Phenomenon 2: 878 (19) a. We are in love/shock/pain state 879 Cf. We are in a room spatial 880 b. We are on alert best behaviour/look-out/the run state 881 Cf. We are on the sand spatial 882 883 The challenge is to account for why it is that prepositions such as in and on have seemingly 884 related state senses. 885 In this section, I will argue that the state lexical concept associated with in selects 886 for co-occurring open-class lexical concepts which access conceptual structure concerning 887 emotional or psychological force such as being in love, in pain and so on. In contrast, 888 the semantic arguments that co-occur with on relate to content that has to do with time-889 restricted activities, as well as actions which involve being currently active. These include 890 being on alert, on duty, and so forth. In short, the types of co-occurring lexical concepts 891 selected by each of the state lexical concepts for these prepositions is of a quite different 892 kind. This suggests that each of the prepositions is associated with a distinct state lexical 893 concept. 894 895

6.1. State lexical concepts for in 896 In this section I present an LCCM analysis of the state lexical concepts associated with in. 897 In particular, I argue that there is more than one distinct state lexical concept conventionally 898 associated with the preposition in. I also show how these state lexical concepts relate to, 899 and are motivated by the prototypical spatial lexical concept, which I gloss as [ENCLOSURE].

7 900

In LCCM Theory, the semantic content that makes up a lexical concept consists of a 901 number of parameters: schematic units or atoms of linguistic contentas discussed in the 902

7 For detailed argumentation for the claim that [ENCLOSURE] is the prototypical lexical concept for in see Evans

(2009, 2010).

20

previous section. The prototypical spatial lexical concept for inthe [ENCLOSURE] lexical 903 conceptencodes the Enclosure parameter, as evidenced by the example in (13), above, 904 reproduced below. In contrast, the [PSCYHO-SOMATIC STATE] lexical conceptone of the 905 state lexical concepts associated with inencodes the parameter Psycho-somatic State, but 906 not the Enclosure parameter, as evidenced in (14) above, also reproduced below. 907

908 (13) The kitten is in the box Salient parameter: Enclosure 909 (14) John is in love Salient parameter: Psycho-somatic state 910 911 [ENCLOSURE] AND ITS PARAMETERS 912

I now turn to a more detailed analysis of the [ENCLOSURE] lexical concept. This is necessary 913 as the state lexical concepts for in have derived from this more prototypical lexical concept 914 for in. As noted above, the [ENCLOSURE] lexical concept encodes a spatio-topological 915 relation holding between a schematic F, the entity enclosed, and a bounded landmark. 916 Bounded landmarks themselves consist of many types even in everyday experience. A 917

bounded landmark includes an interior, which further subsumes an interior surface, and the 918 volumetric interior bounded by the interior surface. It also subsumes a boundary, which can 919

be rigid, as in a metal safe, or non-rigid, as in a plastic carrier bag. The boundary also has 920 other physical characteristics such as permeability and degrees of opacity. Finally, the 921

bounded landmark has, by definition, an exterior: that region which constitutes the inverse of 922 the volumetric interior. Accordingly, part of the exterior includes the exterior surface. The 923

spatio-topological attributes just described relate to enclosure. 924 Due to human interaction with enclosures, the [ENCLOSURE] lexical concept, as 925 manifested in usage events, is associated with a number of functional consequences. There 926

are a number of identifiably distinct sorts of functional categories associated with spatial 927 scenes involving enclosure. In addition to the functional category of enclosure itself, these 928

additionally include Location with Surety (aka containment), Occlusion and Affecting 929 Conditions, summarised in Figure 6. 930 931

932 933

934

935

936

937

938

939

940

Figure 6. Parameters deriving from spatial scenes involving the spatio-topological relation: 941 Enclosure 942

943

Enclosure

Spatial scenes involving Enclosure

Location with Surety

Occlusion

Affecting Conditions

21

Bounded landmarks that are specialised for providing a Location with Surety function 944

are known as containers. This functional category is encoded in linguistic content in terms 945 of what I refer to as the Location with Surety parameter. Containers can provide a support 946 function by virtue of locating by fixingi.e., holding and restrictingthe location of the F, 947 as in the case of coffee in a coffee cup. Alternatively, containers can restrict access (and 948

escape), as in the case of prisons, and safes. 949 The second functional category mentioned relates to Occlusion. A consequence of 950 certain bounded landmarks, due to the opacity of the material which forms the boundary, is 951 that the figure located on the volumetric interior is occluded, and hence hidden from view. 952 This functional category gives rise to the Occlusion parameter. 953

The third functional category, that of Affecting conditions, relates to the fact that an 954 enclosure provides a delimited environment which thereby affects the F located on the 955 volumetric interior. For instance, a prisoner held in solitary confinement in a windowless 956 sound-proofed room is thereby subjected to a particular sensory environment that is a direct 957 consequence of the nature of the bounded landmark in which s/he is located. 958

My claim is, then, that by virtue of interacting in humanly relevant ways with the 959 spatio-topological relation Enclosure, a number of distinct functional consequences arise, 960

which I formalise as distinct and identifiable categories. These functional categories amount 961 to schematic parameters which come to be encoded as semantic atoms, forming part of the 962 bundle of linguistic content encoded by the [ENCLOSURE] lexical concept. In essence, the 963 lexical concept [ENCLOSURE] encodes the spatio-topological relation Enclosure, a schematic 964

unit of knowledge akin to a parameter, and the parametersarising from the encoding of 965 distinct functional categoriesLocation with surety, Occlusion and Affecting Conditions. 966 How then does polysemy arise? Due to the multiplicity of parameters encoded by a 967

single lexical concept, under certain conditions, a parameter (or parameters) that is (or are) 968 particularly salient in a given context of use can become reanalysed as a distinct sense-unit, 969

giving rise to a new lexical concept in its own right. This doesnt mean, for instance, that the 970 [ENCLOSURE] lexical concept loses the Affecting Conditions parameter from its linguistic 971 content. Rather, the Affecting Conditions parameter can become established as the core 972

parameter of a new lexical concept. 973

974 STATE LEXICAL CONCEPTS FOR IN 975 I now consider the state lexical concepts for in, in order to see how these arise from the core 976 spatial lexical concept: [ENCLOSURE]. Consider the following examples, all involving in. 977 978

(20) a. He is in good health. 979 b. The girl is in love. 980 c. John is in trouble/debt. 981

d. Hes in banking [i.e., works in the banking industry]. 982 983

While each relates to a state of some kind, these examples in fact relate to slightly different 984 states: those that have a physical cause, as in (20a)the state of being in good health, 985 which is a consequence of the physical condition of an organisms bodythose that have a 986 psychological or emotional cause, as in (20b)the state is a consequence of a subjective 987 state, which may (or may not) have physical, i.e., observable, manifestationsthose that have 988 a social/inter-personal cause, as in (20c)resulting from social/interpersonal interactions 989 which result in an externally-maintained stateand those that are a result of a habitual 990 professional activity, as in (20d). Put another way, each of these states co-occurs with 991 distinct lexical conceptsthey take distinct semantic argumentswhich relate a particular 992 entity to quite different sorts of states. Hence, there are four distinct sorts of semantic 993

22

tendencies in evidence, supporting the view that we are dealing with four distinct lexical 994

concepts for in. This is illustrated more clearly in the examples below: 995 996 [PHYSIOLOGICAL STATE] (i.e., bodily state) 997 (21) a. Hes in poor/good health 998 b. The woman is in labour 999 1000

[PSYCHO-SOMATIC STATE] (i.e., subjective/internal state) 1001 (22) a. John is in shock/pain (over the break-up of the relationship) 1002 b. John is in love (with himself/the girl) 1003

1004 [SOCIO-INTERPERSONAL STATE] (i.e., externally-maintained state) 1005 (23) a. The girl is in trouble (with the authorities) 1006 b. John is in debt (to the tune of 1000) 1007 1008

[PROFESSIONAL STATE] (i.e., professional activity habitually engaged in) 1009 (24) a. He is in banking 1010

b. She is in insurance 1011 1012

In addition to evidence based on distinct patterns in terms of semantic arguments, 1013 there is an additional line of evidence to support the position that there must be a number of 1014

distinct state lexical concepts associated with in. This is demonstrated by virtue of 1015 ambiguities associated with an utterance of the following kind: 1016

1017

(25) Shes in milk 1018 1019

The utterance in (25) could potentially be interpreted as relating to a woman who is nursing a 1020 baby, and thus lactating. Alternatively, it could relate to a woman who works in the dairy 1021 industry. Given an appropriate extra-linguistic context, an example such as this can be 1022

interpreted in at least two ways. The potential for divergent interpretations is a consequence, 1023

in part, of our knowledge that in has a number of distinct lexical concepts associated with it: 1024 what is relevant for this example is the distinction between a [PHYSIOLOGICAL STATE] lexical 1025 concept and a [PROFESSIONAL STATE] lexical concept. Moreover, ambiguities can be 1026

generated even when a relatively well entrenched example is employed. For instance, even 1027 examples of the following kind: 1028

1029 (26) She is in labour 1030 (27) He is in love 1031

1032 can be interpreted in alternate ways. For instance, (26) could be interpreted as relating to 1033

childbirth or to a professional activity, e.g., the trade union movement. Similarly, (27) could 1034 be interpreted as relating to an emotional state or a professional activity, e.g., marriage 1035

guidance counselling. The former reading is only possible by virtue of assuming something 1036 akin to an [PSYCHO-SOMATIC STATE] lexical concept which is distinct from a [PROFESSIONAL 1037 STATE] lexical concept. And so, both lexical concepts must inhere in long-term semantic 1038 memory if love can be interpreted in these ways in this example. 1039 1040

DERIVATION OF THE STATE LEXICAL CONCEPTS 1041

23

I now examine how the distinct state lexical concepts may have arisen. My claim is that 1042 these lexical concepts developed from the Affecting Conditions parameter associated with the 1043 prototypical [ENCLOSURE] lexical concept for in. 1044 In terms of an LCCM account of the emergence of the state lexical concepts for in, 1045 the trajectory is as follows. Situated implicatures arise in bridging contexts (Evans 2009). 1046

These are contexts in which a usage sanctioned by the relevant spatial lexical concept, such 1047 as the [ENCLOSURE] lexical concept, also gives rise to a situated implicature, such as an 1048 affecting condition. If the prepositional form is repeatedly used in such bridging contexts, the 1049 situated implicature may give rise to the formation of a new parameter, or the detachment of 1050 an existing parameter, as the core parameter of a new lexical concept. I argue below that 1051

bridging contexts involving the functional category of Affecting Conditions may have given 1052 rise to the formation of a number of related but distinct state parameters, and hence lexical 1053 concepts. 1054 In order to trace the development of the parameter Affecting Conditions, we need to 1055 consider spatial scenes that might provide appropriate bridging contexts. To illustrate, 1056

consider the following expressions: 1057 1058

(28) a. in the dust 1059 b. in the sand 1060

c. in the snow 1061 1062

While dust, sand and snow are physical entities which can enclose they cannot, normally, 1063 fulfil the functions provided by, for instance, containers: they do not typically serve to locate 1064 with surety, exceptional circumstances such as quicksand and avalanches excepted. For 1065

instance, dust, sand and snow, by virtue of enclosing, do not normally have the structural 1066 attributes that allow an entity to be supported and thus transported (cf. a bucket, or a coffee 1067

cup), nor do they normally restrict access in the way a prison cell does, for instance. 1068 Nevertheless, these examples exhibit some of the spatio-topological properties 1069

associated with the [ENCLOSURE] lexical concept. This is a consequence of the properties 1070

associated with these bounded landmarks: they provide an affecting condition, an 1071 environmental influence which affects our behaviour. For instance, they determine the kinds 1072 of apparel we wear, and how we behave when we are exposed to the dust/sand/snow, and so 1073 on. As such, these contexts of use provide bridging contexts: both enclosure and affecting 1074

conditions are implicated, and either (or both) may be understood. While examples such as 1075 sand, snow and dust can be construed as enclosures with boundaries, there are other related 1076

examples of what we might refer to as Prevailing Conditions which are much less clear-cut in 1077 terms of the nature of the boundaries involved: 1078 1079

(29) a. the flag in the storm 1080 b. the flag in the wind 1081

1082 I suggest that these instances of in in (29) are sanctioned by virtue of there existing a 1083

distinct parameter Affecting Conditions, which forms part of the linguistic content encoded 1084 by a distinct [PREVAILING CONDITIONS] lexical concept. Hence, the next stage in the 1085 development of a new lexical concept is for the parameter: Affecting Conditions, to be re-1086 analysed as a core component of an independent lexical concept. Clearly a storm and wind 1087 are much less prototypically enclosures, and more saliently provide prevailing conditions 1088

which thereby constitute an environment which affects us. As such, spatial scenes involving 1089 more prototypical enclosures have given rise to the functional category Affecting Conditions, 1090

which has led to the formation of a distinct Affecting Conditions parameter in semantic 1091

24

memory. The existence of a distinct [PREVAILING CONDITIONS] lexical concept, as evidenced 1092

by examples in (29), provides suggestive evidence that such a distinct Affecting Conditions 1093 parameter exists. 1094

I argue that the distinct state lexical concepts associated with in evidenced in (21) to 1095 (24) encode the parameter Affecting Conditions, but not Enclosure. Indeed, these lexical 1096

concepts are what I have referred to as state lexical concepts, as the states invoked all 1097 provide, in some sense, affecting conditions. Moreover, all these state lexical concepts are 1098 relatively, and to degrees, far removed from the physical notion of enclosure from which they 1099 developed. In essence, once an Affecting Conditions parameter becomes conventionalised, it 1100 can be applied to distinct kinds of affecting conditions, even those that are non-spatial in 1101

nature, such as states. This leads to the development of new lexical concepts. 1102 The first such state lexical concept relates to the physical condition of an organism 1103

which thus provides an affecting condition. Such physical conditions include good/ill health, 1104 pregnancy, and any salient physical aspect of the organisms condition which affects and thus 1105 impacts on the organisms functioning. This lexical concept I gloss as [PHYSIOLOGICAL 1106 STATE]. In addition to environmental and physical conditions, affecting conditions can be 1107 caused by psycho-somatic states, such as grief, happiness and sadness which are internal in 1108

nature. This state gives rise to a [PSYCHO-SOMATIC STATE] lexical concept associated with 1109 in. In addition, social interactions which give rise to social or interpersonal relationships lead 1110

to conditions which may affect the individual. Such extrinsic or socially-induced affecting 1111 conditions might include debts, or other sorts of difficult situations which impose conditions 1112

on the behaviour of an individual. This set of affecting conditions gives rise, I suggest, to 1113 what I gloss as the [SOCIO-INTERPERSONAL STATE] lexical concept associated with in. 1114 Finally, ones habitual professional activity provides an affecting condition by virtue of the 1115 physical and social interactions that are attendant upon such activities. This provides an 1116 affecting condition giving rise to a lexical concept glossed as [PROFESSIONAL STATE] 1117

associated with in. The relationship between the Affecting Conditions parameter and the 1118 range of non-spatial lexical concepts for in discussed is summarised in Figure 7. 1119 1120

1121

1122 1123 1124

1125 1126

1127 1128 1129

1130 1131

1132 1133

1134 1135 1136 1137 1138

1139 1140

1141

25

1142

1143 1144 1145 1146

1147 1148 1149 1150 1151

1152 1153 1154 1155 1156

1157 1158

1159 [prevailing [physiological [psycho-somatic [socio-interpersonal [professional 1160

conditions] state] state] state] state] 1161 1162

Figure 7. Parameters and their relationship with the state lexical concepts for in 1163 1164 1165

6.2. State lexical concepts for on 1166 In this section I deal, somewhat more briefly, with lexical concepts associated with the 1167

prepositional vehicle on. The spatial relation designated by on involves the relation of 1168 contact or proximity to the surface of a LM, and so the functional consequence of being 1169 supported or upheld by it. I gloss the prototypical spatial lexical concept conventionally 1170

associated with on as [CONTACT]. This serves to encode the spatio-topological relation 1171

Contact, giving rise to a Contact parameter encoded by the preposition. A functional 1172 consequence of contact is that the TR is thereby supported by the LM. This gives rise to the 1173 parameter Support, which is also encoded by on. The [CONTACT] lexical concept sanctions 1174

an example of the following sort: 1175 1176

(30) the apple on the table 1177 1178 Notice that evidence that the parameters Contact and Support are both encoded by the 1179

lexical concept [CONTACT] comes from the fact that on can only felicitously be employed to 1180 describe spatial scenes in which both parameters are apparent. For instance, if an apple is 1181

held against a wall by someone, the utterance in (31) is semantically anomalous. However, if 1182 the apple is affixed to the wall, for instance by glue, then (31) is entirely appropriate. 1183

1184 (31) the apple on the wall 1185 1186 In short, while the apple is in contact with the wall in both scenarios, in the first scenario it is 1187 the person, rather than the wall, that affords support, while it is the wall, and the glue, which 1188

employs the wall as a means of affixing the apple, in the second. Hence, the example in (31) 1189 applies only when there is both physical contact between the TR and the LM, and when the 1190

latter has a role in supporting the former. 1191

Enclosure

Spatial scenes involving enclosure

Location with surety Occlusion

Affecting Conditions

26

Indeed, there are a number of distinct support lexical concepts associated with on 1192 which privilege the Support parameter at the expense of the Contact parameter, as illustrated 1193 by the following examples: 1194 1195 [SUPPORTING BODY PART] 1196

(32) a. on ones feet/knees/legs/back 1197 b. on tiptoe 1198 c. on all fours 1199

1200 In the examples in (32), the use of on relates to that part of the body which provides support, 1201

rather than being concerned with contact. That is, on all fours, for instance, does not mean 1202 that something is in contact with all fours. Rather, the conventional interpretation is that all 1203 fours provides the means of support. 1204 1205 [MEANS OF CONVEYANCE] 1206

(33) a. on foot/horseback 1207 b. on the bus 1208

1209 With respect to the example in (33b), it is worth pointing out, as Herskovits (1988) does, that 1210

if children were playing on a stationary bus, for instance, that had been abandoned,