Idioms, polysemy, context

Idioms, Polysemy, and Context: A Model Based on Nigerian Arabic

JONATHAN OWENS Bayreuth University Abstract. This article contributes to the debate between monosemic and polysemic approaches to linguistic semantics by a close examination of Nigerian Arabic idiomatic expressions involving the keyword ‘head’. Three broad categories of constraints on polysemy can be identified, which limit polysemy, but not to the extent that a fully monosemic account can be motivated. An inherent, stipulative polysemy resides in idioms and in their constitutive lexemes. The attempt to motivate a monosemic account highlights factors constraining polysemy, sets limits to their effects, suggests a taxonomy which brings together essential structural and semantic aspects of idioms, and simultaneously elucidates the rich, multifaceted world of a simple lexeme.

1. Monosemy, polysemy, context. Undoubtedly one of the central problems of semantics is polysemy. Anticipating the pivot of discussion in this article, does Nigerian Arabic \aas ‘head’ have a single meaning, or does it have many meanings? The translational equivalents in (1a)—(1g) suggest that the answer is “many.”1 (1a) al¤hebíl kula bu¤rub:¤ú fœ \aas¤a DEF-rope also 3-tie-they.it on head-his ‘The rope as well, they tie it on its head.’ (1b) \aas¤a bi¤šiil al¤gœ\á ajala head-his 3-carry DEF-study quickly ‘He learns quickly.’ (1c) al¤kalaam šaal \uusse katiir¤aat DEF-word carried heads many-PL ‘The issue had a number of ramifications.’ / ‘The issue became complicated.’ (1d) šaal ar¤\aas carried DEF-head ‘He took the lead.’ / ‘He headed the column.’ (1e) raas¤a šaayil head-his carrying ‘He is troubled.’ (1f) šuqul šaal \aas¤í Something carried head-my ‘Something distracted me.’ 1

2

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

(1g) kalaam¤ak šaal \aas¤í word-your carried head-my ‘What you said convinced me.’

Linguistic scrutiny, however, reveals the matter to be more complicated. As is well known, there are two opposed positions on this issue. What appears to be a minority position is the reductionist, monosemic interpretation, the one-\aas approach. The opposed position countenances a high degree of polysemy. Indeed, in the so-called classical structural approach to polysemy, a simple test would demand that a high degree of polysemy be recognized in the word \aas. 1.1. Monosemy and polysemy. The classic characterization (Quine 1960) recognizes polysemy when different senses of a lexeme can be meaningfully conjoined. In (2), different senses of lamma raas, literally, ‘gather head’ (where raas is a phonetic variant of \aas, as noted above), are represented in the same conjoined phrase (the first line of the example); one sense is asserted in the first conjunct (‘we joined together’ in the literal sense of assembling in one place), and a different sense is negated in the second conjunct (‘we did not unite’ in the figurative sense of agreeing). The conjoined phrase is not felt to be contradictory; this implies that the two senses are distinct. (The additional contextualization provided by the rest of the example shows how consultants understood the different senses.) (2) A: lammeena raasna haw ma lammeená da ‘We united and at the same time, didn’t unite.’ B: keef da ‘What do you mean?’ A: lammeená da, geýedna bakaan waahid; ma lammeená da, ma qassadna foog al¤kalaam ‘We got together, in the sense that we all gathered (to discuss) in one spot. At the same time, we didn’t unite because we couldn’t agree on the matter.’

The logical test is useful in that it shows that one is not dealing with vagueness (Tuggy 1993).2 A preliminary demonstration of polysemy, or “ostensible polysemy” as it can be called, does not appear particularly rewarding, however, when confronted with the reality that in the database used here, for a single keyword that serves as the basis of the current article, there are something like sixty or more idioms whose senses can be demonstrated to be distinct from the literal meaning of \aas, in a way analogous to (2). As noted, there have been two broad responses to ostensible polysemy. One is, as it were, to deny it, while the other has been to embrace it, albeit within the framework of a theory of semantics or cognitive linguistics that structures polysemy in particular ways.

2015

JONATHAN OWENS

3

Those who deny it are the monosemists. An early monosemist position, and one still relevant as an analytic standard to conceptualizations of figurative language (see discussion around (20) and (23) in section 4.2.2), is that of Searle (1980; see Taylor [1995] for further summary of earlier monosemic approaches). Searle assumes as a primitive a difference between literal and figurative (1980: 122)3 and given it, once an utterance is recognized as figurative, derives the meanings in the following illustrative way (1980:115, 120). The first step is (3). (3) An utterance is recognized as not literal.

Given (3), certain principles (Searle outlines seven, though he says these are nonexhaustive) enable interlocutors to associate the predicate with metaphoric meaning. These principles are possible because of shared strategies that restrict the possible interpretations of the metaphoric meaning. Importantly, Searle does not hold that the relevant, metaphorically interpreted attributes of a predicate are inherent parts of the predicate. Rather, they are derived online via the set of principles. The utterance (4) will, via his Principle 2–“Things that are P are contingently R” (1980:116) (where R = metaphoric meaning)–allow the predicate, be a pig be contingently understood as in (5). (4) Sam is a pig. (5) ‘Sam is filthy, gluttonous, sloppy . . .’

‘Filthy’, ‘gluttonous’, and ‘sloppy’ are literal, inherent attributes of the noun pig, which interlocutors will pick up on to interpret (5). However, pig does not have the sense of a ‘gluttonous person’. Those who embrace the polysemic position probably represent the dominant perspective today. Among cognitive linguists in particular there is an underlying assumption that lexical structure is biased towards polysemy. LewandowskaTomaszczyk expresses this position somewhat programmatically: “Whereas monosemy assumes a minimal, narrow semantic representation, Cognitive Linguistics tends to favor a rich form of representation in which each lexical meaning is an access point to a network of related categories” (2010:153). Also within the cognitive linguistic perspective, Tuggy simply states that “polysemy is rampant” (2003:324, see also Tuggy 1999:357), and he then proceeds to describe the Nahuatl (Nawatl) verb kiisa ‘emerge’ in terms of schemas of different levels of generality. Brugman and Lakoff again assume, rather than argue for, the polysemy of ‘over’ (1988:290), and describe a network of related, polysemous senses in terms of image schemas, connected by links of various types.4 Taylor alludes to basic problems in “rampant polysemy” (2006:52—53), namely, that it implies a very large set of fixed meanings in the storage lexicon. Moreover, there is perhaps no upper limit to the number of polysemous senses that can be discerned for a lexeme.

4

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

Some recent approaches build a bridge of sorts between the monosemist and polysemist approach, endorsing polysemy, but not without conditions, as it were. Evans writes that polysemy develops when “situated implicatures associated with a particular context can become reanalysed as distinct sense units” (2009: 166). However, she does not expand on how it is discerned when a distinct sense unit has indeed become established. Looking at the issue in Searlean terms, how does one recognize that a new sense unit has a cognitive reality, and is not merely a token contingently derived via a set of principles applied online? Riemer (2005) recognizes a core meaning of a lexeme and develops further meanings via metonymic or metaphoric interpretation. He begins with what he terms “prototypical centres” (2005:327), also termed “core meaning” (2005:345). Analyzing the Warlpiri verb pakarni, he begins with a meaning of ‘hit’ or ‘hit with an object such as a hand’. This verb has a number of further meanings, including ‘kill’, ‘pierce’, ‘paint’, and ‘perform dance ceremony’. Each of these meanings is derived via a metaphoric or metonymic application. The meanings ‘kill’ and ‘perform a dance’, for instance, are both seen as effect metonymies: the meaning ‘performing a dance ceremony’ reflects the fact that dances involve hitting feet or instruments against the ground, while ‘killing’ is a causal metonymy from hitting. Both of these approaches begin with a basic meaning and a set of principles or procedures by which polysemy develops. The link to monosemy resides in identifying a core meaning as basic. 1.2. Contextualized approaches. The perspectives outlined above concentrate the question of polysemy in the single lexical item. An alternative, which is sometimes loosely associated with a monosemic approach, is to highlight the role of context in defining polysemy. One explicit expression of this position is that of Allwood (2003), who offers a different approach to monosemy, searching for a way to bring all meanings of a lexeme within a single concept.5 He first outlines two approaches. One is what he terms the “Gesamtbedeutung” (‘holistic meaning’) approach, which looks for the “largest common denominator” (2003: 39) into which all token meanings of a lexeme can be fit. On the other hand, the “Grundbedeutung” (‘basic meaning’) approach searches for an essential, core meaning from which single, differentiated meanings can be derived. Allwood is skeptical of both approaches, in particular because of the difficulty of specifying how, linguistically, one gets to individual, conventionalized meanings. Instead, Allwood advocates what he terms “meaning potentials” (2003:41—45), whose basic idea rests on identifying a simple monosemic meaning whose apparent polysemy is constrained “by other words and by extralinguistic context” (2003: 44).6 Pragmatic and collocational environment together form what Allwood terms “meaning potentials.” He is, however, unfortunately not very specific about how the collocational factor can be operationalized in the conceptualization of monosemy.

2015

JONATHAN OWENS

5

In this article, I would like to address the polysemy-monosemy debate from a different, methodological angle. Whereas it is possible to offer descriptive systems supporting a monosemic (e.g., Searle) or polysemic account of meaning (e.g., Tuggy 2003, 2010; Lewandowska-Tomaszczyk 2010), it is equally necessary to ask what factors and criteria are relevant to supporting the one approach or the other. A crucial aspect of this inquiry pertains to the multifaceted role of context in either limiting, or supporting polysemy as opposed to monosemy. A close examination of contextual factors suggests that neither an unambiguous “monosemy only” or “unlimited polysemy” approach does justice to the nuanced meanings that emerge from a systematic examination of individual factors which bear on semantic interpretation.7 1.3. Criteria. Whereas much of the recent discussion on monosemy-polysemy has concentrated on the best representation of the phenomena, often using diagrams, there has been far less discussion of what criteria should be invoked to decide where a monosemic representation should end in favor of polysemy, or, alternatively, how or whether polysemy should be constrained. For instance, in Tuggy’s (2003) article, different senses of Nahuatl kiisa (okis and okiski in the examples below) are found, such as those in (6a) and (6b). (6a) kittown¤n tomasa tlenoh imak okis she.says-the Thomasina what.is on.her.hand it.emerged ‘Thomasina wonders what it is that has broken out on her hand.’ (Tuggy 2003:330) (6b) n rufino no okimaka totonki yekin yalla okiski the Rufino also it.had.hit.him hot barely yesterday it.emerged ‘Rufino had also gotten a fever; it did not stop (break) till yesterday.’ (Tuggy 2003: 331)

In Nahuatl, kiisa is part of expressions that convey the meanings ‘break out (of sickness)’ and ‘break (of fever)’ (i.e., sickness emerges and leaves the body). That the two sentences (6a) and (6b) have different meanings is beyond doubt. What Tuggy does not explore, however, is the alternative to attributing the different sentence meanings to the different polysemic meanings of kiisa. In this case in particular, he does not discuss the degree to which other parts of the sentence, above all the subject, contribute to–i.e., constrain–different interpretations of kiisa. The translational equivalents in this case, ‘break out’ and ‘break (fever)’, might be interpreted as being wholly dependent on the nature of the subject (this point was already made by Searle [1980:115]). Instead, like Lakoff and Brugman, Tuggy assumes that the polysemy resides in a single lexeme, kiisa (see (25c) and (25d) below). The fact that this polysemy is represented on a hierarchicalized schematic diagram does not hide the fact that interesting linguistic dependencies might exist explaining at least some of the polysemy.

6

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

To consider the factor of context systematically, it is heuristically useful to take as a baseline approach one of the two approaches to meaning. In such a debate one can begin with the simpler of the two positions and see how it fares against a detailed set of data, such as the one from Nigerian Arabic that I present below. The simpler of the two positions is the monosemic one, since it would hold that despite the seven translational contrasts in (1a)—(1g), there is still a single lexeme, \aas, with a unitary meaning. This article thus tries to motivate this position, moving to a polysemous approach only if it fails. Indeed, it turns out that the monosemist position fails in different ways. At the same time, pursuing the monosemic position brings out interesting ways in which polysemy is both defined and constrained. The data itself comes from a detailed examination of idioms in Nigerian Arabic (NA), concentrating on the rich idiomaticity of a single lexeme, \aas ‘head’. I examine this question from four different perspectives, all of them embedded in a discussion and description of idioms in Nigerian Arabic as introduced in section 2. Within each perspective, the background assumption is that, other things being equal, a monosemic representation is preferable. This serves as a null hypothesis. A monosemic approach assumes that elements of context– collocational and pragmatic, for instance (discussed above)–are operative in producing the observed, differentiated meaning of single lexical items. To this end, in section 4 the roles of grammatical, functional, lexical, and pragmatic context in limiting polysemy are considered, and in section 5 these factors are incorporated in a contextualized attribute extension sense taxonomy of one idiom keyword which brings the factors together in a single representation. In section 3, I add a further methodological perspective, namely the behavior of \aas in actual usage, as defined by a large corpus of spoken Nigerian Arabic. As becomes evident, reference to the corpus helps to decide certain issues while contributing a greater degree of precision to others, even if important questions are still left unanswered. 2. Idioms and the data. Idioms would appear to be very good candidates to contribute to the debate on polysemy because idioms inherently imply contextualization of individual lexical meanings. Indeed, this exceptional property of idioms inherently contradicts the summary of Ravin and Leacock, who write that “the common model . . . is to define the meaning of words independent of context . . . and then establish principles according to which word meanings interact when found together in a particular context” (2000:5). As is seen below, Nigerian Arabic \aas ‘head’ has a rich semantics, yet in a sense, no semantics of interest outside of the severely fixed idiomatic collocations it occurs in. One needs to know the context before one can ascertain its meaning. This makes idioms interesting for present purposes, since the monosemic approach gives particular weight to context. As Nunberg, Sag, and Wasow (1994) note (see

2015

JONATHAN OWENS

7

discussion below), idioms by definition imply a disjunction between the meaning of an individual word in isolation, and the meaning of the idiom as a whole. If idioms are not “sense units,” they are not idioms. The idioms used in this study consist of fixed collocations. Lexical constituents of an idiom are represented in curly brackets, e.g., {lamma \aas} ‘unite’ represents an idiom or set of idioms whose lexical constituents are lamma and \aas; if a particular idiom involves some distinctive grammatical or other feature, that distinctive feature is also included within the brackets, e.g., {šaal a\¤\aas} ‘take the lead’, where the definite article ar¤ uniquely marks this idiom (see (25a)). All the idioms discussed in this article consist of two lexical items.8 I concentrate on idioms containing one keyword, \aas (with variants raas and \aab) whose literal meaning is ‘head’, but whose idiomatic usage, as is seen below, encompasses a considerable range of ostensible meanings. Other idiomatic keywords are adduced as necessary. Lexical components other than the headword–the collocates–in the idioms treated here are of two types. One is a single lexical item that itself is constitutive of the idiom in question. In (1e) above, for instance, the meaning ‘convince’ requires both the keyword \aas and the collocate šaal ‘carry, pick up and keep’. A semantically related lexeme such as axad ‘take’ would force a literal meaning, axad \aas¤í ‘he took my head’ (somewhere, for some purpose). I term these as “fixed collocate idioms.” The other type of collocate is an open set of items delimitable by a simple parameter, as in (7), discussed further in (43) below. (7) \aas al¤qalla head DEF-grain ‘tassel of grain’

For \aas to take the sense ‘tassel’, it requires a collocate that denotes some type of grain that sends out flowers (its tassel)–this could be any of many types of millet or guinea corn typical of northeastern Nigerian agriculture. The class of collocates is open ended, but its identifying parameters are fixed. These are termed “open-ended idioms.” The division into keyword vs. collocate engenders two questions, one of major significance. The first, lesser issue, is the basis on which the individual lexemes are identified as keyword or collocate. This has two answers. The first is severely practical: this article is interested in exploring the semantics of \aas, and therefore it is designated the keyword. A second answer is that usually in an idiomatic collocation one of the two collocates occurs in a wider range of idiomatic usages than the other, and this enables it to be identified as the keyword. For instance, thus far no other idioms have been found in which qalla ‘grain’ is combined with some lexical item other than \aas. However, exploring the range of issues associated with the idiomaticity of individual words is beyond the scope of this article. Some methodological aspects of the problem are touched on section 3, when the corpus is examined.

8

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

The more fundamental question in the keyword-collocate division is which element, if any, is polysemous–the keyword \aas, the collocate, both words individually, or the entire unit? My answer to this emerges in part in section 5, when a model of idiom structure is developed. The idioms discussed in this article are all idiomatically combining expressions in the sense of Nunberg, Sag, and Wasow (1994). In fact, they have no grammatical properties that distinguish them from nonidiomatic collocations. For instance, given (1e), repeated as (8a), one can topicalize, as in (8b), or passivize, as in (8c) (for detailed discussion, see Owens 2014:135—38; see also (28b) below, where nominalization of a V + O idiom is given). (8a) kalaam¤ak šaal \aas¤í word-your carried head-my ‘You convinced me.’ (8b) \aas¤í kalaam¤ak šaal¤a head-my word-your carried-it.M ‘As for what you said, it convinced me.’ (8c) \aas¤í an¤šaal head-my PSV-carry ‘I was convinced.’

I do not utilize Nunberg, Sag, and Wasow’s further criteria for characterizing idioms, opacity, compositionality, and conventionality. Opacity and compositionality measure, as it were, the disjunction between the idiomatic meaning and “the meaning we would predict for the collocation if we were to consult only the rules that determine the meanings of the constituents in isolation” (1994: 498). While agreeing that these properties are a relevant way of characterizing how idioms are recognized, they do not form an essential aspect of the issue of polysemy and idioms as discussed here. More fundamentally, when working in a very foreign cultural context, before operationalizing these criteria one would need to form some sort of baseline of expectations for what, for instance, {šaal \aas} ‘carry head’ would mean in a Nigerian Arabic context. Expectations depend to an important degree on what linguistic and cultural background one comes from. Instead, this article begins with a simpler and more austerely Searlean approach: meanings are either literal or nonliteral or abstract, idioms by definition being nonliteral. I do use the terms “compositionality” and “conventionality,” though in ways different from Nunberg, Sag, and Wasow. Compositionality refers to the individual lexemic senses of keyword plus collocate accessed in order to compose a given idiomatic meaning, as is discussed in greater detail in section 5. Note that this usage differs also from that sometimes used in psycholinguistic studies (e.g., Titone and Connine 1999:1656), whereby compositionality refers to the

2015

JONATHAN OWENS

9

accessing of the literal meanings of idiom collocates, in this sense opposed to a “noncompositional” or “lexical representation” (Swinney and Cutler 1979) model. As to conventionality, rather than considering it to be a semantic property of idioms, I use it in the more mundane sense of what is adhered to in the community–community norms (see section 3). Community norms are established in this article by the frequency of a given idiom in a linguistic corpus. The corpus was collected over a period of nearly ten years (1991—2001) in northeast Nigeria in the context of investigating Nigerian Arabic from a number of perspectives, grammatical (Owens 1993), sociolinguistic (Owens 1998), and situational (Owens 2005, 2007a).9 It is representative of Nigerian Arabic in a way to be expected of sociolinguistically based corpora, including texts both from rural areas and the major metropolis of Maiduguri, and having different genres. In all, four hundred thousand words are available for rapid search,10 about two hundred fifty thousand of which are also available in audio format (Owens and Hassan 2012). 3. Conventionality. Conventionality in this article is understood as a property defined against a reasonably large and representative corpus, and taken to be a property of the community of usage. The corpus consists of four hundred thousand words of oral texts. A full description of the data is provided by Owens and Hassan (2012) (see above). About ten hours of the texts have been translated into English and are partly annotated. To my knowledge, this is largest oral corpus available for any single Arabic dialect area, anywhere. All in all, as oral corpora go, its size compares favorably with a number of those for much better-studied languages such as Spanish or French (Cresti and Moneglia 2005), and I assume it to be representative of spoken Nigerian Arabic until more resources become available. In addition, much briefer comparative reference is made to Egyptian Arabic, also in corpus form (as described in n. 25). It is in particular from the corpus that the rich idiomaticity of \aas was identified–upwards of sixty ostensibly distinct idioms involving this word. The appendix lists, according to a categorization developed in section 5, a representative sample of these \aas-based idioms in Nigerian Arabic. 3.1. The keyword \aas: conventionalized as what? The examination of the corpus properties of \aas serves two purposes. A basic one is informational: does idiomaticity in fact play an important role in understanding \aas in Nigerian Arabic? A second one is more ambitious: from the corpus data, can one automatically discern evidence bearing on the question of whether \aas is monosemic or polysemic? This inquiry proceeds in two stages. First, it is shown that frequencies alone cannot be used to differentiate figurative from literal meanings (in the present section and section 3.2); it is then shown (section 3.3) that

10

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

corpus data does indeed reveal important differences between literal and figurative meanings of collocates, though not from a simple count of collocational adjacency (as in sections 3.1 and 3.2), but rather in terms of something that can be termed “discourse embeddedness, which, as will be seen, refers to the availability of a referring entity to coreference in adjacent discourse.” Among words for body parts in the Nigerian Arabic corpus, \aas appears to be the most frequent. There are 308 tokens in all (literal or idiomatic), amounting to 0.08 percent of all words in the corpus. No exhaustive comparison with all other body parts is available at this point, but by way of comparison, gakb ‘heart’, another frequent body part, has 101 tokens (0.025 percent), rijil ‘foot’ has 115 (0.028 percent), ýeen ‘eye’ has 80 (0.02025 percent), and kitif ‘shoulder’ just two. A first, striking observation about \aas is that its literal meaning, a physical ‘head’, represents a rather small minority of its tokens. Examples such as (1a) are somewhat unusual. Basically \aas has three types of meanings: literal (as in (1a)), abstract or idiomatic, and reflexive. The last, not discussed in the present article, is illustrated in (9), where \aas is used as a reflexive pronoun (see Owens 2014:146— 47). (9) bi¤yeerit le Wumt¤a mine kula bi¤yeerit le \aas¤a 3-farm for self-his who all 3-farm for head-his ‘He farms for himself; each one farms for themselves.’ (TV72c)

Counts of tokens of the three meaning types are shown in table 1. One token in which \aas occurs as part of a name of village, \aas al¤fiil ‘head of the elephant’, was not counted, and there are eight tokens still unclassified. These are included in the “other” category in the table. Note that only the 308 tokens of the “abstract” and “literal” categories are included in the following analysis. Table 1. Three Types of \aas ABSTRACT LITERAL REFLEXIVE OTHER

251 57 39 9

70.5% 16.0% 11.0% 2.5%

TOTAL

356

100.0%

It is outside the scope of the present article to pursue the broader comparative issue of how far other body part lexemes behave like \aas. It does appear that \aas is at the higher end of an abstraction scale. Looking at other prominent body parts occurring in idioms, iid ‘hand’ and ýeen ‘eye’ also divide between abstract and literal meanings, but with these two lexical items the literal meanings clearly dominate, as the raw scores show: 65 literal tokens vs. 29 idiomatic ones for iid and 58 literal tokens vs. 22 idiomatic ones for ýeen. On

2015

JONATHAN OWENS

11

the other hand gakb ‘heart’, discussed in section 4.2.3 below, has comparatively a far higher number of idiomatic tokens, all of its 101 tokens being idiomatic. In any case, leaving reflexives and “other” out of the discussion for the rest of this article, in general it can be said that Nigerian Arabic \aas is idiomatic to a far greater degree than it is literal, so that the answer to the first question is that indeed, the idiomaticity of \aas does play a fundamental role in this language. The ratio of idiomatic to literal tokens is close to 3.5:1. Idiomaticity implies a figurative meaning, as is elaborated on in the following sections. An initial interpretation of the figures might be that what is responsible for the great disparity between the frequencies of figurative and literal uses is the polysemy of \aas. As a starting point, table 2 gives the numbers of tokens of every \aas idiom that is represented by five or more tokens in the corpus (the “Token” column), along with the percentage that this represents of the total number (308) of tokens of \aas in the corpus (the “% \aas column); the “% collocate” column of the table shows what percentage of the total number of occurrences of the collocate in the corpus (in parentheses) the tokens in the idiom represent (thus, in the first line, there are 461 total instances of lamma in the corpus, of which 10.2 percent occur in the 47 tokens of the idiom pattern {lamma \aas}). Furthermore, for all fixed collocations where the collocate is a content word (rather than a function words such as a preposition), the sum of the number of tokens of \aas and the collocate that occur in the idiom is divided by the total number of tokens of both \aas and the collocate in the corpus, and this is given as a percentage in the “% key + collocate” column.11 Table 2. \aas Idiomatic Collocations with Five or More Tokens in the Corpus IDIOM {lamma \aas} ‘unite’ (see (30)) {fata \aas} ‘enlighten’ (see (46)) {\aas X = top} (see (41)) {fi \aas} ‘on/at head’, i.e., ‘for self’ {hana \aas} ‘of head’, i.e., ‘for self, own initiative’ {ka\ab \aas} ‘remember’ {ligi \aas} (see (17)) {foog \aas} ‘on head’ ‘responsible for’

TOKEN 47

TOTAL

% \aas 15.2%

34

11.0%

13 12

4.2% 3.8%

11

3.5%

6 5 5

1.9% 1.6% 1.6%

133

41.0%

% COLLOCATE % KEY + COLLOCATE 10.2% (461) 12.2% 54.0% (63)

2.4% (244) 0.4% (1071)

18.3%

2.1% 0.7%

The idioms in table 2 account for 60 percent of all figurative tokens of \aas. Note that there are nearly as many tokens of the single idiom {lamma \aas} as there are of all literal \aas meanings together, and, indeed, the entire idiom, i.e.,

12

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

{keyword + collocate} together can constitute a high percentage of the total tokens of the constituting lexemes. The idiom {fata \aas} claims nearly 20 percent of the respective lexemic tokens.12 It might be claimed that the different meanings of \aas as defined in the following sections are what are responsible for these distributions. For instance, the idiomatic sense of \aas as representing an individual (an idea developed in section 5 below; see (52)) would sanction {lamma \aas}, or \aas in the sense of ‘individual’s cognitive state’ would sanction {fata \aas}. In support of this, one might observe that these two idiomatic meanings are not only unusual in some sense, but together account for 25 percent of all \aas tokens, as well as large percentages of all {keyword collocate} tokens for these two idioms. It would be tempting to conclude that one can discern idiomaticity merely from determining where significant collocational distributions are established. It might then be argued that if the idiomaticity, in turn, depends on the polysemy of \aas, polysemy plays an important role in determining the frequency of usage of \aas in Nigerian Arabic. This perspective can be provisionally discarded for two reasons. First, there are unequivocally idiomatic constructions, such as {ligi \aas}, that constitute a very low percent of all \aas tokens and a tiny proportion of the {keyword collocate} complex in this case. However, {ligi \aas} is no less idiomatic than {lamma \aas} in meaning. Idiomaticity as such, as an abstract property, does not automatically ensure widespread collocational usage. Secondly, if one simply counts tokens without relating the collocations to a wider context, the frequencies of the idiomatic collocations here are no different from those of nonidiomatic collocations, such as sara(h) (be) + bagar ‘take cattle to daily pasture’. An extended comparison elucidates this point. 3.2. Statistics for literally interpreted collocations. Nigerian Arabs were originally, and until today to a large degree still are, a cattle-rearing culture. Whether they keep their cattle in villages, or herd with them in nomadic camps, a daily activity is to take them to pasture, expressed in the verb sara(h) ‘pasture cattle (or goats/sheep)’. This activity is described in many of the texts, so that a total of 106 tokens of sara(h) are recorded. These occur in verb (perfect, imperfect)13 and verbal noun (sára/sarha) form. As might be expected, sara occurs overwhelmingly, though not exclusively, in the context of bagar (or, less commonly, qanam ‘sheep’). There are three common constructions, shown here in (10)—(12). (10) V + be + O, where O = bagar le an¤nahaar ni¤sœra be l¤bagar to DEF-day we-pasture with DEF-cattle ‘Until the daytime we pasture the cattle.’ (IM10)

2015

JONATHAN OWENS

13

(11) S + V, where S = bagar wa hinaak al¤bagar bi¤sœrh¤an fi lubb al¤gašš and there DEF-cattle 3-pasture-F.PL in middle DEF-grass ‘And there the cattle pasture in the midst of the grass.’ (IM20) (12) VN + possessor, where possessor = bagar bala sarh¤it al¤bagar bi¤hart¤u other pasturing DEF-cattle 3-cultivate-PL ‘Other than pasturing cattle, they farm.’ (TV44)

In all of these cases, as with the {lamma \aas} tokens, the frequency of the collocations can be formally measured, as in table 3. Table 3. Frequency of sara(h) COLLOCATION

TOKENS

% sarah %bagar (1015) % KEY + COLLOCATE

sarah + be + bagar = O bagar = S + sara sarha VN + bagar possessor

56 14 3

54.0% 13.0% 3.0%

5.5% 1.3% 0.2%

TOTAL

73

70.0%

7.0%

sarah/sarha without bagar

31

30.0%

13.0%

The lexeme bagar may occur overtly, as in (10)—(12), or it may be is crossreferenced anaphorically, as in (13). (In this example, hin is a feminine plural pronoun, referring to the feminine plural noun bagar.) (13) bi¤sœra bee¤hin 3-pasture with-them.F ‘He pastures them.’ (GR21)

In some cases, sara does occur absolutely, as it were, without collocating with bagar, as in (14). (14) ana ma saree¤t I not pasture-I ‘I never pastured.’ (IM138)

The lexemes sarah and bagar imply each other to a very high degree; in particular, 73 out of 104 tokens of sara(h) occur with bagar. But this is comparable to the fact that 14.5 percent of all tokens of \aas occur in the presence of lamma or 54 percent of all tokens of fata occur in the presence of \aas. Furthermore, when one considers the {keyword + collocate} complex, the 13 percent ranks between {fata + \aas} and {lamma \aas}. Idiomatic collocations will rank higher than some nonidiomatic ones and lower than others.

14

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

The difference, that {lamma \aas} is defined as idiomatic while sarah (be) + bagar is not, is an evaluative one, not a formal one.14 Numbers alone do not identify idiomaticity or polysemy. 3.3. Discourse embeddedness and the literal vs. figurative contrast. Still, the data suggests that Searle’s simple literal-figurative dichotomy, with the figurative domain accounted for by a set of interpretative principles, does not adequately reflect the reality that within the figurative domain, certain fixed collocations are so established that in natural speech they are as numerous as all literal usages combined. One conclusion that can be drawn from the comparison with nonidiomatic collocations is that whatever processing procedures underlie literal collocations are equally the basis of the idiomatic collocations, given that each involve high-frequency collocations. It is intuitively unlikely that, as Searle would have it, a set of special principles intervenes in the computation of idiomaticity that is lacking in the computation of literal collocations.15 While the corpus data in sections 3.1 and 3.2 allows this inference to be made only indirectly, data can in fact be found bearing on the question of contextualized identification of idiomaticity that definitively answers the question against Searle. In Searle’s analysis, literal meanings are all there are to words such as ‘pig’ in expressions such as (4) above. In a separate study (Owens and Dodsworth 2015), we looked specifically at the properties of words with both a literal and idiomatic meaning; \aas was among these. The basic research question followed in that study was the same as that addressed here, namely, what the nature of idioms is and what the contribution of the individual lexical and grammatical elements to the idiomatic meaning is. As has been noted, in phonological, morphological, and syntactic terms there is no difference between \aas in its literal sense and in its idiomatic sense. Both, for instance, generally have an emphatic \, though both can have the variants \aab and raas, and both have a plural form \uýusse or \uusse. They are both syntactically flexible (see (8) above). Furthermore, as seen in sections 3.1 and 3.2, there is no criterion based on immediate collocational environment alone that distinguishes literal from idiomatic meanings. Against these identities, Owens and Dodsworth (2015) show that there is one striking formal contrast between idiomatic and nonidiomatic meanings of one and the same lexeme, such as \aas. This difference is manifested in the discourse referential properties of the idiomatic and literal meaning of \aas. The idiomatic meaning has a much reduced discourse exposure as compared to the literal meanings, or as described by Owens and Dodsworth, it has a reduced degree of discourse embeddedness or discourse visibility. What is meant by this is that a token of a word with a figurative meaning in an idiomatic collocation is less likely to be referred to in preceding or following discourse than is a token with literal meaning. This property of idiomatic nouns is analogous to what happens in processes such as compounding and noun incorporation; recall the

2015

JONATHAN OWENS

15

work of Hopper and Thompson (1980, 1984), in which, for instance, the noun fox in the compound fox-hunting is largely invisible to reference in discourse (note the oddity of an utterance like !Our fox-hunting adventure wasn’t much fun because we didn’t catch it). In this section, I briefly summarize the results of Owen and Dodsworth (2015), concentrating on the contrast between two key words already discussed in this section, \aas and bagar. As is seen already in section 3.1 above, \aas has both literal and idiomatic meaning; on the other hand, in our relatively large corpus bagar has only literal meanings. Discourse visibility is measured by how often the referent of a keyword–here, \aas or bagar–is referred to in preceding or following discourse (see Owens, Dodsworth, and Kohn [2013] for a detailed description of the basic measures used). As predicted, a literal meaning, as in (1a), will have a higher degree of discourse embeddedness or visibility than will a figurative or idiomatic meaning. This contrast holds irrespective of whether a lexeme is only literal, as with bagar, or either literal or figurative, as with \aas. Thus, in (1a) the pronoun ¤a on \aas¤a refers to the literal keyword, bagara. On the other hand, it is expected that \aas in (15) would not have a referent in following discourse, since it is part of the idiom {lamma \aas} ‘unite’, and this expectation is borne out; in this sentence, only the masculine second person plural pronoun is referentially continued in the following clause. (15) tawwa lammee¤tu \aas¤ku ke haw famfam¤ku addal¤tú mi formerly join-you.PL head-you.PL so and pump-your.PL repair-2PL.it not axeer wa better Q ‘Wouldn’t it have been better if you had united and repaired your pump?’ (TV45)

By contrast, in (16), in which \aas has a literal interpretation, it is referred to by the pronoun object in the following clause. (16) baal foog \aas waahid baal ley¤a urinated on head one urinated on-it ‘He just urinated on a head, he urinated on it.’ (GR47)

Owens and Dodsworth (2015) compares a total of five keywords with idiomatic and nonidiomatic meanings (\aas ‘head’, gakb ‘heart’ [see section 4.2.3 below], iid ‘hand’, ýeen ‘eye’, and ba:un ‘stomach’) with four words that only have literal meanings (bagar ‘cattle’, qalla ‘grain’, ru:aana ‘foreign language’, nugura ‘hole’) and measures the extent to which each keyword was referred to in either previous or following context, using the same corpus as that described at the end of section 2 (Owens and Hassan 2012). Here I reproduce only a comparison between the two words, \aas and bagar, that reflect the overall contrasts between idiomatic and literal keywords. These two words have the frequencies in the corpus shown in table 4.

16

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

Table 4. Frequency of Two Keywords KEYWORD

TOTAL TOKENS

CORPUS %

\aas ‘head’ 308 (247 idiomatic) 0.07725% bagar ‘cattle’533 (none idiomatic)0.22%0 %64

IDIOMATIC % NUMBER OF TEXTS 80%

65

NOTE: corpus % = percentage of the entire corpus represented by tokens; idiomatic % = percentage of tokens that are idiomatic; number of texts = how many individual texts (out of a total of 94) the tokens occur in.

The key question here is whether idiomatic \aas differs in the degree to which it is embedded in following discourse from nonliteral usages and, if so, in what way. Table 5 gives the raw counts of tokens relevant to this point. Table 5. Discourse Embeddedness, Literal vs. Idiomatic Tokens IDIOMATIC \aas Referent of keyword not referred to in immediately adjacent clause Referent of keyword referred to in immediately adjacent clause

224 27

LITERAL \aas

bagar (LITERAL)

40 17

306 201

By inspection, the most obvious point is that bagar, which has only literal usage in our corpus, has a far higher degree of discourse embeddedness than does \aas in either idiomatic or literal guise. It is equally apparent, however, that relative to the total tokens, literal usages of \aas are referred to far more often in neighboring discourse than are idiomatic instances of \aas.16 The latter are referred to in surrounding discourse in barely one token in ten. The three values can be ranked as follows in terms of likelihood of reference in surrounding discourse: bagar > literal \aas > idiomatic \aas. The contrasts set out in table 5 present strong empirical evidence against a Searle-like literal-only analysis of the keyword. Whereas \aas in its idiomatic and nonidiomatic senses is identical phonologically, morphologically, and in terms of clause-internal syntax, there are measurable differences between the two when the larger discourse context is considered. Here it becomes apparent that there is, as it were, a literal \aas1 and an idiomatic \aas2, distinguished from each other by the degree to which the noun can be referred to in the surrounding discourse. Note that these differences are, in all likelihood, impossible to discern with the introspective methodology used by Searle. They become apparent only via quantitative analysis of natural speech. Comparing these results to those discussed in sections 3.1 and 3.2, where occurrences of specific collocations were counted, equally shows that quantitative data alone does not shed decisive light on the nature of idomaticity; how informative quantification is depends on the specific linguistic questions to which it is addressed. Understanding the cross-clausal behavior of collocations turns out to be crucial for a fuller understanding of their nature.

2015

JONATHAN OWENS

17

The discussion in this section leads in two directions. On the one hand, it shows how important context is for discerning the nature of lexical-semantic differentiation. On the other, it offers one piece of evidence in favor of the reality of polysemy. While polysemy can, logically and methodologically, be explained by pragmatic principles, the consistency with which literal and idiomatic meanings of the same lexeme behave contrastively in discourse can be accounted for if \aas minimally is polysemous as literal \aas1 and figurative or idiomatic \aas2. 4. Contexts. Section 3 concentrates on the relation between lexical collocates in idioms and their role in discourse as it can be ascertained in natural corpora. Lexemes in idioms, however, are sensitive to grammatical, semantic, and pragmatic contextualization as well, as many who deal with semantics and cognitive linguistics have emphasized, even researchers who do not assume that a sufficient consideration of context will decisively tip the analytical balance towards monosemy. Various types of context have been considered, including that provided by derivational affixes (Tuggy 1993:277), syntactic constructions (Dowty 2000), collocational information (e.g., Cruse 1986; Deane 1988:342; Charles 1988; Miller and Leacock 2000; Dunbar 2001), and pragmatic information (Cruse 2000). Riemer (2005:242—49), in the context of his explanation of polysemy as metaphoric or metonymic extension, has an extended discussion of the problems of determining in which parts of a clause the metaphor or metonymy is to be located. In the framework of the current work, as already noted, Allwood is one of the few who has at least suggested the role of context in systematically limiting polysemy. In this section and section 5, I attempt to give concrete expression to the way in which context interacts with lexical meaning in regard to polysemy. In the present section I outline the basic contextual categories. An underlying idea is that in important respects polysemy is adventitious upon familiar grammatical categories, and that the many individual idiom types that define the polysemy are in a significant number of cases grammatically in complementary distribution with one another. 4.1. Basic contextual categories. Three different types of contextualization are exemplified below. The discussion shows that small grammatical, lexical, and pragmatic differences can have vastly different effects on the interpretation of idioms. The contextualization types can be divided into three broad kinds– clause-internal (or frame-internal; see below), interclausal or interframe, and pragmatic. Intraclausal contextualization defines the domain of what I term “distributed polysemy,” which I turn to now. 4.2. Distributed polysemy. Distributed polysemy alludes to the fact that contrasts bearing on the interpretation of polysemy are generally distributed over more than one location in a grammatical clause or frame and that, intuitively, the nature of the clause plays a role in determining the interpretation

18

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

of the idiom. This idea is implicit in some treatments of idioms. In particular, Fellbaum (1993) examined in detail the role of the determiner in defining idiom meaning, arguing that the behavior of a determiner in an idiom tracks its behavior in nonidiomatic usage. For instance, in the idiom he has an axe to grind, the indefinite article a is appropriate because the meaning of the idiom is noncompositional–‘has a grievance or ulterior motive’–and the idiomatic meaning is nonspecific, licensing the use of nonspecific a. However, Fellbaum’s well-documented article emphasizes the isomorphy between idiomatic meanings and literal meanings, not the role of the determiner in establishing contrasts among idioms themselves.17 Indeed, Fellbaum suggests that “The literal equivalent of a noun in most idioms is somewhat vague and admits of severally semantic similar interpretations . . . the exact meaning [of axe] is not important, because the meaning of the idiom as a whole is not dependent on such a fairly subtle difference” (1993:281). This follows a model of keyword interpretation in idioms in which the idiomatic meaning of the keyword noun is either allowed to be derived online, as with Searle, or one in which, as with Glucksberg (1993, 2001:78), literal and abstract meanings coexist in processing time. While agreeing with Fellbaum that idioms are sensitive to more than the meaning of the noun keyword, I show here that once one begins dissecting larger clause-based dependencies in idioms, the meaning of the idiomatic whole is indeed often dependent on very specific nonlexical functional elements and contrasts. For heuristic purposes, distributed polysemy is defined within the clause. It is realized via basic functional categories which can simply be represented as standard categories, subject, predicate, object, and so on, with the nominal arguments bearing thematic roles, Agent, Theme, Source, Goal, Frequency, Manner, and many more, and individual nouns defined by inherent features of number and gender (in Arabic), and various syntactic relations such as definiteness. The clause is also a domain for relations of anaphora. As a terminological note, I sometimes refer below to the clausal domain of distributed polysemy as a “frame.”18 As far as this discussion of polysemy goes, there are two key aspects of frames relevant to delimiting polysemy. One pertains to anaphora and exophora, which below is termed simply “anaphora,”19 the other to what might be termed grammatical and lexicorepresentational elements. 4.2.1. Anaphora and agreement. Anaphora and agreement pertain to the relations of anaphora within the clause itself. In Arabic, the identity of the subject is marked morphologically on the verb itself through the categories person, number, and gender, these agreeing elements being analysed as pronominal elements with anaphoric values (e.g., Owens, Dodsworth, and Rockwood 2009; Owens, Dodsworth, and Kohn 2013). A number of idioms are minimally distinguished by the nature of the anaphoric relations between referring expressions

2015

JONATHAN OWENS

19

in the clause. For instance, the idiomatic structure {ligi \aas¤PSSR} (get headPSSR) requires the verb ligi and the object \aas, and the latter must be in a possessed relation. The possessor of \aas must be human (as a noun token or by reference, if pronominal). This is illustrated in (17a)—(19b). All involve differential interpretations based on what broadly can be thought of as differences of reference and anaphor.20 (17a) lig¤at \aas¤hum get-F head-their ‘She got their support.’ (17b) lig¤at \aas an¤naas get-F head DEF-people ‘She got the people’s support.’ (18) ligi \aas¤a get head-his ‘He escaped.’ (19a) lig¤at \aas¤ha get-F head-her ‘She gave birth.’ (19b) ligii¤na \aas¤na got-we head-our ‘We (female speakers) gave birth.’

In each case a difference of meaning correlates with a difference in the grammatical properties of each idiom. The differences within the common structure can be represented as in (20a)—(20c). (20a) ligi1 \aas-possessor2 (in (17a) and (17b)) (20b) ligi1 \aas-possessor1 (in (18)) (20c) lig¤at1F \aas-possessor1F (in (19a) and (19b))

In structure (20a), the possessor’s reference must be disjoint from that of the subject. The possessor can be either a pronoun, as in (17a), or a noun, as in (17b). In structure (20b), the possessor must be coreferential with the subject, so it must be a pronoun cross-referencing the subject. Herein lies the fundamental difference between (17a) and (17b), on the one hand, and (18)—(19b), on the other. In (19a) and (19b), too, the possessor must cross-reference the subject, but in addition, the subject and the possessor that coreferences it must be feminine. The cross-reference can be speech-situational in the case of a first person subject; (19b) implies that the speakers are female.

20

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

These examples indicate that how raas is bound referentially in the clause fundamentally constrains its interpretation. Two classes of examples can be distinguished here. In the first class, features of the subject itself, and accompanying cross-referentiality in the possessor pronoun, delimit the possible interpretation of the utterance, as in (19a) and (19b). For the meaning ‘give birth’, the subject must be feminine–either overtly marked as such, or so interpretable.21 Choice of gender is not free in this case. In the second class, the choice of subject is free, but whether or not there is cross-referentiality between subject and possessive pronoun of raas induces a basic difference in meaning; this is seen in the contrast between (17a) and (17b), on the one hand, versus (18)— (19b), on the other. Summarizing the cases exemplified here, one and the same basic lexical collocation, {ligi raas¤PSSR}, participates in three distinct idioms, each defined by subtle grammatical and referential properties. These correlations between meaning and structure are summarized in (21). (21) Coreferentiality between subject and possessor of \aas: meaning outcomes 1. Subject and possessor are not coreferential: ‘support’ 2. Subject and possessor are coreferential: a. Subject is not feminine: only ‘escape’ b. Subject is feminine: ‘give birth’

It can be noted in this context that of the three categories listed here, by far the most general one in defining contrastive idioms is that of coreferentiality between subject and noun possessor, i.e., cases 1 and 2a in (21). 4.2.2. Grammatical and lexicorepresentational factors. The grammatical basis of the frame is defined by traditional grammatical categories such as definiteness and number, by grammatical functions such as subject and object, and by what can be termed lexicorepresentational factors such as choice of lexemes and their semantic values, including their thematic values within the clause. As both Fellbaum (1993) and Nunberg, Sag, and Wasow (1994) emphasize, idiomatic structure is adventitious on nonidiomatic and hence the nonidiomatic structure serves as a starting point. The clausal frame as relevant to idioms begins with a basic, literal meaning built around a verbal predicate. As above, the verb šaal ‘carry’ is used for illustration in the following discussion. The basic grammatical and lexicorepresentational elements of the frame for šaal ‘carry’ can be characterized as in (22). (22) basic frame for šaal Nature of event: pick up and keep (for one’s benefit; see (23) below for elaboration) Agent = subject: NP, usually human Patient = object: NP, usually tangible object Source: PP Goal: PP

2015

JONATHAN OWENS

21

The agentive subject is normally human. A look at the first two hundred of the 1,006 tokens of šaal in the corpus shows that 180 (90 percent) have a human subject. The subject can also be a being or object with inherent motive force–for instance, a cow carries individuals, as do cars and trucks. In a few cases it is an abstract entity; for example, hukuuma ‘government’, with which šaal is understood as ‘appointing’ an individual. I discuss these and other types of arguments below. The object is semantically and grammatically unconstrained. It can be either singular or plural, be a single discrete object or a collection (see section 4.3 for the contrasting case of lamma ‘join’). Like all verbs, šaal occurs with various adjuncts (noncore thematic roles)–time, reason, etc. In addition, it occurs with a specific Source-Goal complement, min . . . le ‘from . . . to’. These do not figure in the discussion below, though there are idioms that utilize them, e.g., an idiom that indicates the linear extent of territory by specifying its endpoints, as in (23). (23) šiil¤a min hine wadd¤a le baama carry-it from here send-it to Bama ‘It lies between here and Bama.’ (lit., ‘Carry it [i.e., the hypothetical line of distance] from here and send it to Bama.’)

A key aspect of šaal is its inherent semantics. Basically, it indicates a composite action of picking up an object, carrying it, and either keeping it or disposing of it on one’s own behalf (if the subject is human); the latter is seen in (24). (24) al¤ma\a šaal¤at al¤\aaba le s¤suug DEF-woman carry-F DEF-sour milk to DEF-market ‘The woman carried the sour milk to the market.’

Here šaal implies that the woman prepared the milk, carried it to the market, and sold it on her own behalf. Selling is implied by ‘market’, and that it is for her own benefit is implied by the verb šaal. A contrasting meaning is produced by the verb axad ‘take’, which in the same sentence (al¤ma\a axadat . . .) would suggest that the sour milk which the woman took to the market was not her own. Given the basic frame of šaal, it is clear that there are various ways that it does not account for the idiomatic usages exemplified in (1b)—(1g). In (1f) and (1g), the object is an Experiencer, not a tangible object. In (1b), the subject is ostensibly a physical object, though one of a specific, idiomatic type, as discussed below. One could approach the discrepancy between basic frame and idiomatic meanings in (1b)—(1g) in two ways. On the one hand, one could assume that (22) is in fact the only frame for šaal, and account for the idiomatic meanings via metaphoric extension. This is essentially the approach taken by Goossens (2002: 364) in his analysis of idioms such as ‘catch someone’s ear’. In his analysis, ‘ear’ undergoes metonymic extension, ‘attention’, and the entire phrase is given a metaphoric interpretation ‘get someone’s ear’. On this basis, \aas in (1g) would

22

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

have a metonymic value, such as ‘intelligence, opinion’, and a metaphoric image would be formed of someone’s word appropriating it. This approach can be termed the “metaphor processing approach,” since the meaning, in line with Searle’s description given in the discussion of (3)—(5) in section 1.1 above, is derived via reference to a contextualized, online re-interpretation of the literal meaning. While this treatment has the advantage of keeping a simple frame for šaal, it misses the fact that šaal in its idiomatic usage undergoes a number of systematic transformations (using the term neutrally) of the basic frame, even if it maintains a number of its key elements. These transformations include the following. The key features are that idiomatic šaal often represents a change of state, as in (1b)—(1c), rather than a change of location as in (23) and (24). The subject is frequently a motive force of some sort, as in (1b), and frequently is inanimate (e.g., a government), while the object tends to be an Experiencer, as in (1f)—(1g). In some cases, it is the subject that assumes the role of Experiencer, as in (1e). On the other hand, the durativity of the action and the subject ending up in control of the event described by šaal are maintained across literal and idiomatic meanings. The relation between literal and idiomatic šaal can be represented as in table 6. Table 6. Transformation of Idiomatic šaal BASIC šaal

IDIOMATIC šaal

durative subject in control of event change of location subject = Agent Agent = human object = tangible object

(no change) (no change) change of state subject = Experiencer Agent/Experiencer = motive force object = Experiencer/Patient

® ® ® ®

NOTE:

The arrow symbol (®) means “transformed into”; all idioms do not undergo each of these transformations.

These transformations are pervasive in idiomatic uses of šaal. I assume that idiomatic šaal, the šaal with “transformed” meaning, is a different šaal from literal šaal. That is, rather than adopt Goossens’s “metaphorical processing approach,” I assume that šaal itself is polysemous. I take up these this point in greater detail in section 5. Here, I briefly analyze the six {šaal \aas} idioms of (1b)—(1g) (the examples are repeated below for convenience as (25a—(25d), (27a), and (28b)). The four idioms in (25a)—(25d) have \aas as Patient. (25a) šaal ar¤\aas carried DEF-head ‘He took the lead.’ / ‘He headed the column.’

2015

JONATHAN OWENS

23

(25b) al¤kalaam šaal \uusse katiir¤aat DEF-word carried heads many-PL ‘The issue had a number of ramifications.’ / ‘The issue became complicated.’ (25c) šuqul šaal \aas¤í something carried head-my ‘Something distracted me.’ (25d) kalaam¤ak šaal \aas¤í word-your carried head-my ‘What you said convinced me.’

The distinctive characteristics of each of these idioms are as follows. In (25a), the Patient-object \aas is marked by the definite article. In (25b) and (25d), the Agent-subject is propositional, rather than human. Example (25d) could be paraphrased as (26). (26) al¤inta gul¤t¤a šaal \aas¤i REL-you said-you-it carried head-my ‘What you said convinced me.’

In (25c) and (25d), the object \aas must be possessed, usually by a possessive pronoun. Here \aas is an Experiencer. In (25b), \aas must be an indefinite noun, can be plural, and is typically modified by an adjective, as in the example. The Agent in this case is propositional, similar to (25d). One could paraphrase (25b) in a way parallel to (26), with ruusse katiiraat as object. Note that (25c) is distinguished from (25d) merely by the nature of the subject. If the subject is propositional, the appropriate translation in English is ‘convince’. If it is not, and if it is an arbitrary sound, an unspecified matter, it is ‘distract’. The two {šaal \aas} idioms in (27a) and (27b) have \aas as subject. (27a) \aas¤a bi¤šiil al¤gœ\á ajala head-his 3-carry DEF-study quickly ‘He learns quickly.’ (27b) \aas¤a šaayil kalaam head-his carrying word ‘He has problems.’

In (27a), the patient must be, broadly speaking, propositional. In this pattern, \aas represents an Experiencer. Example (27b) may be expressed without an object–the head is carrying something, but this thing need not be stated. This example is doubly durative. The verb šaal itself is durative, as seen above, and the active participle form of the verb also has a durative function in Nigerian Arabic, as indeed in virtually all varieties of Arabic (Yavrumyan and Owens

24

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

2007). In both of these, again, the idiomatic meaning mirrors the durative meaning of šaal itself. The six idioms obviously are not distinctive in their basic lexical makeup; the {šaal \aas} collocation is at the core of each. Instead, differentiation is achieved via various grammatical and lexical properties. Word order and, attendant upon this, grammatical relation are factors; idioms are differentiated according to whether \aas occurs as subject, as in (27a) and (27b), or as object, as in (25c) and (25d) (in either case it is an Experiencer). The lexical identity of the subject may be distinctive; idioms are differentiated according to whether the item that occurs as subject is propositional or not ((25c) vs. (25d)), and whether it is human or not. Morphosyntactic form can differentiate idioms; whether \aas is definite (as in (25a)), indefinite (as in (25b)), or possessed (as in (25c) and (25d)) are all factors defining particular idioms. It is clear, then, that the different meanings that {šaal \aas} is associated with are distributed across different grammatical parts of the transformed frame and pertain to various aspects of them. It is seen that the polysemy of this collocation, too, is distributed. Here it can be asked in what sense the precise definition of the frame specificities of each idiom helps to constrain the polysemy of \aas. In many cases, contrasting individual examples makes it clear that this does occur. For instance, apparent differences in meaning may be due solely to the nature of the subject, as with (25c) vs. (25d). In both cases, the basic cognitive capacity attributed to the Experiencer is determined by the subject. If the subject is propositional subject, the Experiencer is ‘convinced’; if the subject denotes a random sound, the Experiencer is ‘distracted’. In contrast to the analogous treatment of Tuggy (see (6)), I would not consider (25c) and (25d) to represent distinct idioms; or, to put the matter in lexical terms, I would not consider that the contrast between (25c) and (25d) involves different meanings of \aas and šaal. The larger clausal frame defines a lexical complementarity wherein the different meanings are defined. Such a precise characterization of the frame does not generally lead to a complete elimination of polysemy, however. Knowing that \aas as Patient in (25a) must have a definite article does not obviously lead to the meaning that (25a) has, even if it probably does in some way limit the interpretations of the sentence. The fact that (25a) is not possessed, for instance, will lead one to look for a referent outside of a person’s cognitive state (see section 5.1.1). Furthermore, the definite article points in the direction, potentially, of a unique entity. Distributed polysemy, therefore, covers a range of contextualization factors, some of them very specific, others constraining but not determinative. 4.2.3. \aas vs. gakb, ‘head’ vs. ‘heart’, object vs. subject. An interesting variant on the correlation between grammatical function and idiomatic keyword requires a separate subsection. As noted in section 3, while \aas is probably the

2015

JONATHAN OWENS

25

idiomatic keyword with the largest number of tokens in the corpus, others are important as well, including gakb ‘heart’. One striking difference between \aas and gakb pertains simply to their syntactic distribution. Whereas gakb strongly tends to occur in subject position, \aas is, as it were, an object idiom. Table 7 shows the basic frequencies with the expected frequencies of each syntactic function in order to highlight where the significant differences lie. It is clear that \aas is overrepresented as an object (37 percent), whereas gakb is overrepresented as a subject (52.4 percent). In fact, the relative frequency of \aas as object is, in a sense, higher than it appears in table 5. Under the category “possessor” appear nineteen tokens where \aas is the object in a verb-plus-object construction that is nominalized (see n. 12), and therefore appears as the possessor of the verbal noun. For example, compare, to the ordinary clause in (28a), the nominalized expression in (28b), where the object \aas is the possessor of the verbal noun lamamaan. Table 7. Syntactic Position: \aas vs. gakb \aas BASIC FREQUENCY

subject topic predicate object object of preposition possessor fragment abu

65 6 3 116 78 35 1 4

Total

308

gakb

EXPECTED FREQUENCY

88.9 4.5 2.3 97.1 76.1 35.4 0.8 3.0

BASIC FREQUENCY

53 0 0 13 23 12 0 0

EXPECTED FREQUENCY

29.1 1.5 0.7 31.9 24.9 11.6 0.2 1.0

101

(28a) lamma = verb, \aas = object tawwa lammee¤tu \aas¤ku formerly join-you.PL head-you.PL ‘Formerly you united.’ (TV45) (28b) ýaarf¤iin lamam¤aan a\¤\aas know-PL joining-VN DEF-head ‘They know about unifying.’ (IM20)

There are nineteen such nominalizations, and adding these to the 116 object tokens would bring the total of \aas as object to 135 (44 percent)–nearly half of all tokens of \aas, both literal and idiomatic. There is something in the idiomatic nature of \aas and gakb that predisposes them to occur in these two positions, or to phrase the matter in contextual terms, the context provided by the subject position favors gakb, whereas object position favors \aas. In particular, two main meanings of idiomatic \aas relevant to this discussion are proposed in section

26

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

5.1.1 below: representing the individual as an agentive force to the outside world, and representing the cognitive center where the world is processed. These two meanings are predominantly associated with occurrence in the object position. For \aas to represent the individual as an agentive force to the outside world generally requires two entities in a clause. The first, identified by the subject, is the entity to whom the individual is projecting himself or herself (in (28a), this is ¤tu). The second is an entity representing what is projected, the image of self, represented by \aas. This is a Patient, prototypically in object position. In the case of \aas as cognitive center, \aas is an Experiencer, and here as well the Experiencer is typically an object. Shifting an object to the role of Experiencer, in fact, is one of the transformations noted in table 6 that idiomatic šaal implies. Discussing the idiomaticity of gakb in detail is outside the scope of the current article. For present purposes, gakb represents the initiator of an action, and the initiator, in Arabic, typically occurs in subject position.22 Alternatively, gakb represents the state an individual is in, and here as well it occurs in the subject position of a stative clause. Examples (29a)—(29c) exemplify the point. (29a) ana gakb¤i xaram min il¤xidime hiil naas¤na me heart-my leave from DEF-work of people-our ‘I lost interest in working for our people.’ (lit., ‘my heart left from the work’) (J 5) (29b) šuqul al¤gakb¤ak raad¤a thing RC-heart-your wanted-it ‘what you wanted’ (lit., ‘the thing that your heart wants’) (IM08) (29c) qaadi baýiid da gakb¤ak raagid there far DT heart-your resting ‘There far away you are at ease.’ (GR 139)

These observations do not, of course, rule out the possibility that \aas (or gakb) in these constructions can be moved out of the canonical object (subject) position. Table 5, however, shows that object is the unmarked position for idiomatic \aas. Subject and object each provide a different context for the idiomatic usages of \aas and gakb. The level of contexualizing generality in this case is extremely broad. Very programmatically it can be represented as in (30). (30) Typical functions, ‘head’ vs. ‘heart’ object: Patient/Experiencer (= \aas) subject: Initiator/State-In (= gakb)

The point here is not only that idiomatic \aas is contextualized into object position because of its typical Patient or Experiencer role, while idiomatic gakb is contextualized as subject because of its Initiator or State-In role. It is equally

2015

JONATHAN OWENS

27

that the functions object and subject are part of broad contrastive frames which support the general semantic properties of \aas and gakb and highlight their contrastive meanings. 4.3. Interframe factors. The previous section shows that the overall structure of a frame affects the way in which idioms are conventionally determined. While the idiomatic meaning utilizes the basic grammatical structure of a frame (pace Fellbaum), individual, nonlexical elements can play a crucial, contrastive role. In this section, I expand this latter perspective by comparing different frames, with the view towards determining the extent to which the basic frames themselves can be said to determine the meaning of the idioms. First, it is shown that the nature of the lexemes in a frame give important clues as to idiomatic meaning (section 4.3.1); second, it is seen that there are general idiomatic markers that give important clues as to idiomatic meaning. 4.3.1. Contexts and lexical frame: frame determinism? To a naïve outsider it might appear that which collocation produces which idiomatic meaning is arbitrary. Why should (31) not mean ‘he convinced them’ (e.g., ‘he gathered up their ideas’), or alternatively, why should {šaal \aas} in (32) not mean ‘unite’? (31) lamma \aas¤hum gather head-their ‘He united them.’ (32) šaal \aas¤í carry head-my ‘It convinced me.’ / ‘It distracted me.’

In this case, a closer look at the basic frames of the two verbs allows one to discern why (32) is not a good fit for the meaning of (31); whereas literal šaal places no restrictions on the type of Patient (object) it takes, lamma requires a mass or a plural NP, as in (33a)—(33c). A singular object is allowed only if the object is represented as accumulated, as in (33d); a simple singular object, as in (33e), is odd. (33a) lamm¤o al¤kaare gather-they DEF-load ‘They gathered up the load.’ (33b) lamm¤o al¤bagar gather-they DEF-cattle ‘They rounded up the cattle.’ (33c) lamm¤o al¤hibaan gather-they DEF-ropes ‘They gathered up the ropes.’

28

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

(33d) lamm¤o hebíl wara hebíl gather-they rope after rope ‘They gathered up one rope after another.’ (33e) !lamm¤o hebil¤í gather-they rope-my (!‘They gathered my rope’)

For {lamma \aas} to acquire the meaning of ‘convince’ it would require an extension of its basic lexical frame, since ‘convince’ freely allows a singular object, as in (32). Such an extension is certainly not an impossibility. As seen in the previous section, unusually, {šaal \aas} idioms often have inanimate agentive subjects or even Experiencer subjects. However, šaal is, in this sense, better suited to the task, since it lacks this particular constraint, as is seen in (34a) and (34b). (34a) šaal¤o al¤hebíl carry-they DEF-rope ‘They carried the rope.’ (34b) šaal¤o al¤hibaan carry-they DEF-ropes ‘They carried the ropes.’

While it might appear that a “frame determinism” is at work in steering the {lamma \aas} idioms away from the meaning ‘convince’, frame determinism cannot be applied in an automatic manner.23 When one considers idioms with the structure {ligi \aas}, encountered in (17a)—(19b) above, one finds that ligi, like šaal, places no obvious restrictions on its object. In this case, however, the properties of the subject are contrastive. Whereas šaal can take a propositional subject, ligi cannot, as is seen in the contrast of (35a) with (35b), even though one way a politician can get someone’s support is through his or her words. On the other hand, a human Agent with šaal would only be fully acceptable if the literal meaning were intended, as is seen in the contrast of (36a) with (36b). (35a) !al¤kalaam ligi \aas¤i DEF-word got head-my (Cannot be used for ‘What he said convinced me.’) (35b) al¤kalaam šaal \aas¤í DEF-word carried head-my ‘The matter convinced me.’ (36a) al¤politišan ligi \aas¤i DEF-politician got head-my ‘The politician got my support.’

2015

JONATHAN OWENS

29

(36b) !al¤politišan šaal \aas¤i DEF-politician carried head-my (Cannot be used for ‘The politician convinced me.’)

In these cases, there are no frame-contrastive elements that explain these differences other than the contrastive semantic features of the subject, which combines with the collocations to form the given meanings: {ligi \aas} ‘support’ requires that the subject be human; {šaal \aas} ‘convince’ requires that the subject be propositional. Moreover, as noted above, having a propositional subject is not the normal statistical tendency of šaal. As it were, not only is the collocation {šaal \aas} idiomatic, but the type of subject that the idiomatic collocation allows is itself grammatically idiomatic. In the final analysis, then, one is left in a by now familiar situation: context is contrastive, but generalizing across different frame-defined contexts problematic. 4.3.2. Interframe collocational paradigms. A second type of frame determination is seen when the collocates of a given keyword (for present purposes, of \aas) themselves occur in a series of idioms, contributing a generalizable meaning to the idioms they occur in. This establishes what can be called an interframe regularity, where the frames are linked by the collocate. Collocates of this sort come particularly from polar adjectives, the best exemplification being the pair xafiif ‘light’ and tagiil ‘heavy’. Generally speaking, xafiif indicates a tendency to overstep an implicit boundary, often socially defined, whereas tagiil indicates the opposite; some examples are shown in (37a)—(41b). (37a) iid¤a xafiif¤e hand-his light-F ‘He steals.’ (37b) iid¤a tagiil¤e hand-his heavy-F ‘He doesn’t steal.’ (38a) qašim¤a xafiif mouth-his light ‘He is talkative.’ (38b) qašim¤a tagiil mouth-his heavy ‘He is tight-lipped, does not give away secrets.’ (39) \aas¤a xafiif head-his light ‘He is light-headed.’ / ‘He is inebriated.’ (See (42) below.)

30

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

(40a) gakb¤a xafiif heart-his light ‘He is a copycat.’ (40b) gakb¤a tagiil heart-his heavy ‘He is stubborn.’ / ‘He is not influenced by others.’ (41a) jild¤a xafiif body-his light ‘He is active.’ (41b) jild¤a tagiil body-his heavy ‘He is lethargic.’

To the extent that xafiif and tagiil consistently introduce the idea of staying outside of or within a boundary, respectively, their contribution to the meaning of the idiom is contextually regular, a semantic classifier as it were. The proviso “to the extent that” is necessary, since in fact xafiif and tagiil are not always straightforwardly associated with these boundary-related meanings. In particular, in the case of the main keyword of this article, tagiil gives the meaning seen in (42), reasonable in view of the basic meaning of tagiil, but not necessarily an obvious opposite of (39). (42) \aas¤a tagiil head-his heavy ‘He is dumb.’

4.4. Pragmatics of culture. A final type of contextualization is pragmatic or speech-situational. The relevance of the speech situation was already discussed briefly for (19b) above. It is also relevant for (19a). Normally, (19a) would be interpreted as glossed in the example (‘she gave birth’). However, if it is known that the referent is either an underaged girl or a postmenopausal woman, then it would be interpreted in the same way as (18) (i.e., as ‘she escaped’). Note that this pragmatically effect on interpretation of this example is limited to one of the two interpretations ‘give birth’ and ‘escape’. 4.5. Summary. Above, three broad ways are described in which different idiomatic meanings of \aas correlate with differences in grammatical, lexical, and pragmatic properties of the idiomatic collocations and their larger syntactic and pragmatic environment. These systematic differences for a single keyword potentially provide support for Allwood’s (2013) emphasis on the predictive role of context, and therefore, potentially, for a monosemist perspective. Two contrasting idealized views can be taken as to how predictive the contextual factors discussed here are. On the one hand, a given factor may actually

2015

JONATHAN OWENS

31

reduce what are ostensibly different idioms to a single type. A prime example here concerns (18) versus (19a) and (19b). On encountering the contrast, I assumed that these represented separate idioms, until, during a talk at the University of Maiduguri, it was pointed out that in northeastern Nigeria, as indeed in neighboring regions, to give birth is to escape a dangerous situation. Childbirth is the prototypical escape from danger of a female. Only women can give birth, and Arabic distinguishes grammatically between masculine and feminine. Thus, despite the quite different equivalents of these two expressions in English translation, it is reasonable to assume that only a single idiom is involved, whose apparently different interpretations is controlled grammatically (by gender agreement) and pragmatically. A similar issue arises in the discussion relating to (25c) and (25d); the difference between ‘convince’ and ‘distract’, deriving from šaal raas, depends on the lexical nature of the subject. On the other hand, only in a minority of the cases discussed can two apparently distinct idioms plausibly be reduced to one. Moreover, such reductionism very quickly runs up against a basic fact: the collocations themselves are arbitrarily given. That idioms generally speaking are explicable on a post hoc basis, but are not predictable in advance, is a truism in research on idioms (Nunberg, Sag, and Wasow 1994:497). There is no a priori reason why {ligi \aas} should mean ‘gain support of, escape, give birth’, rather than, say ‘capture, hire’ (cf. ‘headhunters’ in both cases), ‘brainwash’, or ‘get orientated, come to, get one’s head back’.24 Indeed, there is no nonhistorical a priori reason (in contrast to historical reasons; see n. 9) why the individual lexemes ligi and raas should be used at all to induce the meanings they have in Nigerian Arabic–why, for instance, ‘get support of’ should not be expressed by šaal raasa (cf. (25c)).25 A reading of šaal raas as ‘gain support’ would be very much in keeping with the use of šaal for cases of acquiring and maintaining an abstract attribute. Ultimately, idioms have the same arbitrariness as do individual words, with the important proviso that a further element of arbitrariness intrudes in the need to specify which collocates are necessary to produce the arbitrarily given meaning. However, there is also a large in-between area where it is intuitively appealing to see context at work in circumscribing idiom meaning, even if one cannot generalize the constraining factors linguistically or determine the idiom meaning on the basis of these factors alone. In one of the relatively few works that addresses the linguistic peculiarities of idioms, Glucksberg (2001:83—86) asks what factors constrain their flexibility, emphasizing semantic and pragmatic factors. He silently and swiftly kicked the bucket (where kicked the bucket is interpreted as ‘died’) is acceptable, but he sharply kicked the bucket is not because die sharply is uninterpretable. The present section argues for a broader perspective. When the internal grammar of idiomatic clauses is contrasted with that of related idioms or with different idiomatic keywords, in many cases the meanings of idioms are either constrained or determined by specific grammatical, referential, and contrastively paradigmatic elements. If section 3.3 argued

32

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

for the polysemous reality of idiomatic keywords, this section cautions that ostensible lexical polysemy is far from being boundless. 5. A contextualized attribute extension sense taxonomy for idioms. Section 4 shows that in some cases contextual factors can be invoked that plausibly explain away seemingly polysemous meanings of \aas. At the same time, that section suggests that only a minority of ostensible semantic contrasts can be accounted for without polysemy. Context in the majority of cases is of limiting or constraining character. It is the task of this section to introduce a representational model of polysemy that gives due recognition to factors that might place constraints on it. I continue to use \aas for exemplification below, incorporating the analytical concepts discussed thus far, as well as further ones discussed in the present section. To this end, sections 5.1 and 5.2 develop what can be termed a contextualized attribute extension sense taxonomy of \aas. The taxonomy has two aspects. On the one hand, main semantic components of \aas are proposed, which derive from inherent properties of \aas itself. The nature of these properties is discussed briefly in section 5.1.1. On the other hand, the polysemy of idiomatic \aas does not exist independently of the collocates that are constitutive of the idioms themselves nor of the contexts in which it occurs (sections 5.1.2 and 5.1.3). I should emphasize that the task in the rest of this section is not to solve the intractable issue of how many lexemes of the shape \aas there are, but rather to offer a systematic framework for addressing the issue.26 5.1. An account of idiomatic \aas. The taxonomy has three basic components: a contextualization component, discussed in section 4, the stipulation of collocations, and a hierarchicalized taxonomy based on the idea of attribute extension. Attribute extension is discussed in section 5.1.1; I then turn to collocations and related issues in 5.1.2, and then in 5.1.3 bring all of the elements together in a single representation of idiomaticity. 5.1.1. An attribute extension taxonomy for idiomatic \aas. The idea of attribute extension is based on the work of two different traditions. One is the work of Riemer (2005), introduced briefly in section 1.1, and the other, that of Glucksberg and his associates. I introduce the issue with Riemer, but develop this section mainly on the basis of Glucksberg, returning to Riemer in section 5.2. Riemer develops a theory of lexical polysemy based on the idea of metaphorical and metonymic extensions. Analyzing the Warlpiri verb pakarni, Riemer begins with a basic meaning of ‘hit’ or ‘hit with an object such as a hand’. He terms this a “prototypical centre” (2005:327) or a “core meaning” (2005:345). Warlpiri pakarni has a number of further meanings, including ‘kill’, ‘pierce’, ‘paint’, and ‘perform dance ceremony’. Each of these meanings is derived via

2015

JONATHAN OWENS

33

metaphor or metonymy. The meanings ‘kill’ and ‘perform a dance’, for instance, are both seen as effect metonymies–‘kill’ being a causal metonymy from hitting, and ‘perform a dance ceremony’ based on the action of hitting feet or instruments against the ground. Although these further meanings are derived via metaphoric or metonymic extension, Riemer considers them conventionalized in the pakarni lexeme, i.e., they are no less a part of its meaning than is ‘hit’.27 Given the idea that figurative meanings are derived from properties of lexical items themselves, it is but a short step to the work of Glucksberg and his colleagues. In the course of a long debate with proponents of conceptual metaphor, Glucksberg developed the idea of property attribution to explain how metaphors work. The concept of attribute extension as used here borrows loosely from the idea of property attribution as developed in psycholinguistic work (Cacciari and Tabossi 1988; Glucksberg, Brown, and McGlone 1993; Glucksberg and McGlone 1997, 1999; McGlone 2007; Haser 2005:42). As McGlone (1996:457) explains, the metaphoricity of Our marriage was a rollercoaster ride is interpreted as matching the marriage with the properties that typify a rollercoaster ride–exciting, full of ups and downs, or scary, for instance. One of these properties is attributed to marriage. In the attribute extension taxonomy, it is experiential properties of \aas on which the main figurative senses proposed are based. The perception of where it sits on the body, and its physical shape, for instance, is the basis of examples (43a)—(46) below, whereas its cognitive control function is paramount in (47)— (50). In each case, an inherent, figurative attribute of \aas is singled out and forms the basis of attracting a large range of the collocates with which \aas forms the idioms. Properties of \aas are not “matched with” and “attributed to,” but rather “extended out of.” This manner of understanding the abstract extensions of \aas tallies well with Riemer’s (2005) treatment of lexical polysemy, as discussed above. Given a “core meaning” or “prototypical center” for a lexeme, figurative meanings can be derived based on metonymic and metaphorical attributes which inhere in, or are interpreted experientially to accompany, the core meaning. The contextualized attribute extension taxonomy of \aas begins with the basic distinction between literal and abstract (idiomatic) and proceeds to group the sixty-odd meanings under one of four general categories. The grouping is intuitive, but is implicitly based on metonymic-metaphorical extensions that inhere in the properties of \aas itself. As is briefly discussed in 5.2 below, it is probably impossible to decide for each proposed taxonomic parameter whether its basis is metaphoric or metonymic or a mixture of the two. Some extensions are based on \aas as denoting a prominent (physical) structure. In idioms of this sort, \aas combines with various open-ended collocate classes to represent the top, headlike part of an object. Examples are seen in (43a) and (43b).

34

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

(43a) \aas al¤qalla/duxun head DEF-grain/millet ‘tassel of grain/millet’ (43b) \aas al¤beet/kuuzi/watiir head DEF-house/hut/car ‘roof of the house/hut/car’

In these examples, the prominence is in a class of physical objects–in (43a), any grain that sends out a headlike tassel, and in (43b), a relatively flat object that covers an extensive area attached underneath it. The idea of prominence has an abstract interpretation as well, where it combines with propositional and “conjunctive metonymic” (Dirven 2002) collocates, as in (44). (44) \aas al¤madrasa/xidime/hiraata head DEF-school/word/farming ‘the topic/essence of school/work/farming’

In this first set of collocations, \aas stands in a part-whole relationship to the collocate. It can have the sense of prominent part in an absolute sense as well, as in (45) and (46). (45) mine akal \aas who ate head ‘Who won?’ (46) šiil ar¤\aas (= (25a)) carry DEF-head ‘Take the lead.’

Here \aas represents implicitly what is the first or front of something, without a possessor noun making explicit what this entity is. Before moving on, it should be emphasized that \aas in the senses described thus far does not merely translate as ‘top’ or ‘on top of’ (equivalent to foog ‘on’), nor can its idiomatic senses can be understood independently of its open-ended collocates. This point can be underscored with the following two examples. First, \aas takes the sense of ‘tassel’, as in (43a), only if the tassel itself has a headlike appearance. Millet (duxun) and guinea corn (tambuuna, ajaqama– various types), the two staples of farming in northeast Nigeria, fulfill this description. Corn (maize = mabir), on the other hand, does not. The tassels of corn hang down on the sides, and are designated ganduul (cf. Classical Arabic qanduul ‘aspalathus, type of gorselike flower’).28 Corn does not have a \aas. For the second, a \aas at¤tœraab ‘head of the soil’ is a headman over a relatively large area of land. He might be a traditional district (lawaan) or province

2015

JONATHAN OWENS

35

(ajá) head, for instance. In addition, the term can be applied if someone is the owner, and therefore has control over, farms or other extensive holdings. One can speak of a \aas az¤zerýaat ‘head of the farms’ in this case. It would not be appropriate to apply this term to the owner of a single house, however, no matter how big it is, nor to the owner of single farm, hence it is not possible to say *\aas al¤beet29 or *\aas az¤zerý. In short, the meaning of \aas is finely calibrated with the nature of its collocates. As seen in many examples discussed thus far, \aas also collocates with various lexemes, mostly verbs and adjectives, to induce various abstract meanings. In these idioms, \aas derives from the basic function of a head as housing the brain, which is the psychological, intellectual, and sensory center of an individual. Without the head the individual has no cognitive capacity, and without the basic functions of the head the individual has no ability to project himself or herself or to interact with the outside world. This, then, gives two basic perspectives about Nigerian Arabic \aas. It can represent the cognitive state of an individual, as in (47)—(50), or it can represent an agentive individual to the outside world, as in (51)—(54). Further examples are found in the appendix. (47) ka\ab¤ni fi \aas¤a hold-me in head-his ‘He remembered me.’ (48) \aas¤a xafiif (= 36) head-his light ‘He is light-headed.’ / ‘He is careless.’ (49) fatee¤na lo¤hum \aas¤hum open-we for-them head-their ‘We enlightened them.’ (50) kalaam¤a šaal raas¤í (= 24d) word-your carried head-my ‘He convinced me.’ (51) na¤lumm \aas¤na we-gather head-our ‘We unite.’ (52) ka\ab¤na \aas¤hum min ad¤duwaas hold-we head-their from DEF-fight ‘We kept them out of the conflict.’ (53) ligii¤na raas¤hum got-we head-their ‘We got their support.’

36

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

(54) bi¤tallif \aas¤a 3-spoil head-his ‘He makes a fool of himself.’

The senses discussed here can be represented by the hierarchical taxonomy in (55). (55) attribute extension sense taxonomy for \aas Literal ‘head’ : (1a) Figurative 1. Prominent part (a) Part-whole (i) Physical: (1e)—(1f), (43a)—(43b) (ii) Abstract: (44) (b) Whole: (45), (46) 2. Psychological/intellectual/sensory capacity of the individual (abbreviated to “cognitive and individual center”) (a) represents the cognitive state of an individual: (47)—(50) (b) represents an agentive individual: (51)—(54)

The taxonomy represents a basic split at the literal-abstract interface. Essentially this makes the claim that literal and abstract meanings are different submeanings, that polysemy begins, as it were, at the literal-abstract divide. A different representation would be that shown in (56). (56) literal \aas > abstract \aas Prominent part Cognitive center etc.

The representation in (56) essentially follows Searle’s extralexemic postulate of a literal vs. a metaphoric or figurative meaning, with the figurative meaning always derived online from the literal. In section 3.3, empirical evidence is presented arguing for the categorical difference of literal and figurative meanings, hence of (55) over (56). 5.1.2. Collocational stipulation and semantic mapping. A basic property of the polysemy of \aas is that a differentiated meaning can be discerned only in the context of a further collocate. What is termed a “collocational stipulation” is a part of the representation of the collocation. This has a distributional and a semantic aspect. Distributionally, the collocates are defined according to a fixed parameter. The open-ended collocates, as seen above, constitute a set defined by a certain characteristic–tassels of a certain shape, for instance. The fixed collocates are unique lexemes. These latter collocates themselves typically have a sense hierarchy with literal and figurative senses, with the figurative sense again defined only against its collocation with certain lexemes. As a provisional

2015

JONATHAN OWENS

37

illustration the familiar verb šaal ‘carry’ can be used. This verb is itself registered in the database with twenty-six separate ostensible idioms, including those in (1b)—(1g).30 It may well be that closer consideration of the meanings of these idioms will justify a finer hierarchicalized sense taxonomy for šaal, similar to that for \aas in (55). However, since motivating such a taxonomy is outside the scope of this article, I simply represent the figurative meanings of šaal by copying over the transformations from literal to abstract as discussed in table 6 above; this is shown in (57). (57) (abbreviated) attribute extension sense taxonomy, šaal Literal: ‘carry’, durative, subject control Figurative change of location ® change of state subject = Agent ® subject = Experiencer agent = human ® Agent/Experiencer = motive force object = tangible object ® object = Experiencer/Patient

The collocational stipulation is relevant only to produce a meaningful idiom. For instance, axad \aashum ‘he took their head’ is a possible collocation, but it does not need to be stipulated since the collocation of verb plus object that underlies it would be created by general rule, and the literal meaning that results would be computed by whatever process underlies the interpretation of verbs and objects. The sense ‘convince’, as in šaal \aashum ‘he carried their head, i.e., he convinced them’, on the other hand, requires a stipulation of the particular collocation. In the framework developed here, the lexemes that enter into the idiom can be said to be noncompositional vis-à-vis a literal reading of their constituents, but they are semicompositional against the figurative meanings of their stipulated collocates, as defined by contextualized attribute extension sense taxonomies like those of (55) and (57). The idioms access a specific part of the taxonomy (e.g., \aas in (1b) has specifically the meaning ‘cognitive state’), but this only limits the meaning of the idiom without predicting the precise meaning (hence “semicompositional” above). The collocational stipulation is associated with what I term a “semantic mapping” (Owens and Dodsworth 2015). Typically the idioms are paraphrasable with a single word. Example (1b), for instance {\aas šaal}, repeated as (58a), can be paraphrased by (58b). (58a) \aas¤a bi¤šiil al¤gœ\á ajala head-his 3-carry DEF-lesson quickly ‘He learns quickly.’ (58b) bi¤lýallam al¤gœ\á ajala 3-learn DEF-lesson quickly ‘He learns quickly.’

38

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

In the paraphrase, bilýallam is equivalent to idiomatic \aas bišiil. For present purposes, the term “semantic mapping” can be thought of as an equivalence between the idiomatic collocation and a single lexical item. Example (59) represents the mapping for (58a) and (58b), where the dotted line represents a semantic relationship. (59)

Note that semantic mapping as conceived of here provides a representation for the phenomenon described in section 3.3, whereby idiomatic nouns (such as \aas in (59)) are invisible to discourse in the way literal nouns are not. They are so because, as parts of idioms, they are incorporated into a semantic whole, which preempts the visibility of the single lexeme \aas to discourse.31 5.1.3. Contextualized attribute extension sense taxonomy. The final component of the representation of the \aas-based idioms are the clausally based frame properties discussed in section 4.2 as distributed polysemy. As seen there, it is not only the choice of the key collocational lexemes that determines the interpretation of idioms, but also various clause-internal anaphoric and grammatical relations, as well as the semantic identity of different arguments in the clause. A consideration of all the idiom factors described here leads to the following three basic elements: • the postulated attribute extension sense taxonomies; • an arbitrary, though essential link to a collocate or an open-ended class of collocates with which the idiom is constituted, the collocation stipulation, and related semantic mapping; • the specifics of the grammatical frame that characterizes the collocation, “contextualization.” The meaning of the idiom is located in the interaction of these three attributes. A skeletal fragment of what information a more complete contextualized attribute extension sense taxonomy might look like is given in (60), incorporating all three basic attributes. The scheme in (60) should be read as an illustration of what information will be found in the contextualized attribute extension sense taxonomy, not as an optimal formal representation of the information that is found in it. In line with the exposition in sections 4.1 and 4.2, frame elements, given within slashes, are associated with individual idioms. Clearly, generalizations of various kinds would need to be built into the representation–generalizations that potentially would filter some of the frame information given into higher order components of an as yet unspecified nature.32 It should also be clear that only frame-specific elements that are essential for

2015

JONATHAN OWENS

39

uniquely defining an idiom are incorporated in the contextualized attribute extension sense taxonomy.33 (60) contextualized attribute extension sense taxonomy for \aas (based on (55)) … Figurative meaning 1. Prominent part [collocational stipulation: line 1(b) in the Figurative part of (55), Figurative part of (57); \aas: collocates with figurative šaal ‘take the lead’] /Agent: human, \aas = Patient, \aas = definite/ … 2. Psychological/intellectual/sensory capacity of the individual: a. represents the cognitive state of an individual [collocational stipulation: line 2(b) in the Figurative part of (55), Figurative part of (57); \aas: collocates with figurative šaal ‘convince’] /Agent: propositional, \aas = Patient, \aas¤PSSR and disjoint with PSSR of Agent …/ … b. represents an agentive individual to the world [collocational stipulation: \aas collocates with lamma ‘unite’ ] /\aas ¤PSSR; PSSR = plural/ …

The question can be posed whether one needs a sense taxonomy at all. Instead it might be claimed that all one needs to do is to specify cooccurrence restrictions on the collocates–that \aas, for instance, can collocate with types of grain that have flowers that form headlike tassels, that it collocates with lamma, and so on. This information needs to be stated anyway, since the basis of the differentiated sense taxonomy as given here is the differing collocates implied in each idiom. To consider this perspective in adequate detail would again take one into the large domain of the motivation of idioms in terms of metonymy and metaphor, a domain too large for discussion here (though see 5.2 below). Instead, two basic points can be made. First, this suggestion further implies that collocations can be precisely enough characterized without regard to meaning, i.e., in terms of cooccurrences as ascertained in a corpus–that meaning can, as it were, be dispensed with. In response to this, section 3 above shows that at this stage in the study of Nigerian Arabic, at least, not enough data is available to seriously motivate this approach. Second, reducing the issue to mere cooccurrence patterns devoid of meaning determination implies doing away with the literal-figurative contrast that is the basis of the sense taxonomy. Innocuous though this second consideration may seem, it has implications for the contextualized attribute extension sense taxonomy in general. In the spirit of Searle, the assumption that \aas is basically polysemous, that it has a literal and a figurative sense, entails exploring the many facets of figurativeness. This involves, inter alia, layers intermediate between individual idioms and the most general “figurative” level, ‘prominent part’ and ‘cognitive and individual center’ being two such. The current representation should be understood as a classifying heuristic, commensurate with

40

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

the data, and in principle verifiable via psycholinguistic testing, further work with corpora, and more input from native speakers. 5.2. Nature of contextualized attribute extension sense taxonomy. It is relevant to briefly consider what the basis of the divisions in the contextualized attribute extension sense taxonomy are–what is the basis of generalizing major abstract categories? With Riemer (2005), one can start with the familiar categories of metonymy and metaphor. As noted in section 1, Riemer’s analysis recalls that of Allwood (2003), except that it develops a precise mechanism of polysemic extension. Riemer recognizes a basic meaning of lexemes and, as already explained above, accounts for polysemy via metaphorical or metonymic extensions. From Riemer’s perspective, as a straightforward taxonomic exercise, one could probably say that the ‘prominent part’ set of meanings for \aas (line 1 in the “Figurative” part of the taxonomy in (55)) are metaphoric. The tassel of grain is similar to the head of a person or animal. The ‘cognitive center’ meaning (line 2(a) in the “Figurative” part of (55)) is metonymic, representing what might be termed a “category-and-property metonymy” (Radden and Kövecses 2007: 344)–the head is the part of the body that controls attitudes, actions, and cognitive states in general, and \aas is representative of these properties. “Representation of” in the sense of “substitution for” is a typical sort of metonymy (Warren 1999, 2002). In its uses representing an agentive individual (2(b) in the “Figurative” part of the taxonomy in (55)), \aas appears at first sight to be metonymic. It is considered here to represent an individual to the outside world; that implies a partwhole relation, whereby the head stands for the entire individual–a typical form of metonymy. However, matters are complicated by the fact that the prototypical “representational” meaning of \aas in Nigerian Arabic is that where \aas is used as a reflexive pronoun (Owens 2014:146), as in (61) (see also (9) above). (61) šif \aas¤í saw.I head-my ‘I saw myself.’

While reflexive \aas and \aas in the sense of representing the individual are probably related, the latter, idiomatic sense as described here must be something more than a part of the body representing a whole individual. Reflexive \aas already does this. If the two cases are distinguished by taking idiomatic \aas to represent the individual to the world, as an active agent, how does the addendum “as an active agent” fit into the simple classification of a meaning as metonymic or metaphoric? “Active agent” is here understood more exactly as “in the capacity of an active agent; in active-agentive terms,” even “like an active agent.”34 These interpretations, however, slide into a metaphorical

2015

JONATHAN OWENS

41

understanding of this third major sense category of \aas, so that a conclusion that \aas in this sense is both metonymic and metaphoric is hard to escape. While the difficulty of drawing a distinction between metonymy and metaphor has been discussed in many places, it is sometimes thought that a distinction for any given figure can be made, even if one invokes the idea of scalar values, some figures being more metonymic and less metaphorical than others, or vice versa (e.g., Barcelona 2002a, 2002b; Dirven 2002; Riemer 2005:186). Other scholars, however, either question the possibility or the efficacy of consistently drawing a distinction at all (Barnden 2010), or argue that definitional criteria for the distinction that have been offered yield ambiguous classifications at best (Haser 2005: chapter 2), or simply suggest that precise definitions are not crucial to a discussion of figurative language (Ritchie 2006:11). Given both the challenge of defining a given Nigerian Arabic idiom as metonymic or metaphoric, and the general issue of distinguishing between the two, the current approach, which accounts for figurative meanings in terms of property extension, avoids the necessity of developing an extensive taxonomy of figurative language and applying it to the large database examined here.35 6. Conclusion. In this section, I highlight four points, summarizing what is said above, and pointing to avenues of future research. To begin with, it may be asked whether idioms in fact deserve special attention. Sinclair (2004) is one who considers all language idiomatic. From this perspective, presumably different collocational possibilities of idioms differ only in degree, not in kind, from collocations in general.36 Section 3 of the present article provides evidence that idioms should indeed be distinguished from other collocations. Although, when one looks simply at collocational frequencies, idioms are indeed no different from literal collocations (sections 3.1—3.2), the behavior of idiomatic nouns in discourse reveals them to be different from literal nouns, even those with which they share the same phonological, morphological, and syntactic entry. More generally, along with a collocational stipulation–idioms exist only in specific collocations (section 5.1.2)–idioms are associated with a specific semantic mapping. It can be suggested here that idioms do deserve special linguistic study, and that they merit recognition as a discrete category, less constrained than derivation, compounding, and noun incorporation, but more constrained than literal collocations. Introducing the idea of constraints leads to the second point and to the core issues discussed in this article, including the role of polysemy in the constituent idiomatic lexemes. The perspective argued for here is that the limits of polysemy should be seen not as absolute boundaries, but rather in terms of the multiple factors that work to constrain an ostensibly limitless polysemy. A major task of the article is to define grammatical parameters that are relevant in this regard. This leads into a third point, namely, how these factors are best represented. Certainly there is much room for discussion. I limit myself here to one point. It

42

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

could be argued that the hierarchicalized attribute extension sense taxonomy is inadequate in two directions. On the one hand, it might be claimed that it is not finely enough differentiated to account for all of the sixty-some ostensible \aas idioms (nor even for the thirty or forty that might result if they could be reduced further, as argued in the discussion of (25c) and (25d) towards the end of section 4.2.2). This problem is implicit in the term “semicompositional” introduced in section 5.1.2. On the other, it could be said that such taxonomies force binary cuts to be made even in cases where items in different branches might have common properties (an inherent problem in a branching taxonomy). As far as the first problem goes, what can be said here is that further experimentation is needed to test the degree to which finer taxonomic distinctions are rewarded with a higher degree of figurative compositionality. The objection cannot be answered in the abstract. As to the second problem, it is an easy enough matter to allow for cross-cutting connections. However, unless these are well motivated, they potentially add a high degree of arbitrariness to the hierarchy itself. In both cases, the major goal of the contextualized hierarchicalized attribute extension sense taxonomy outlined in (60) is to define three differentiated domains that need to be recognized for a closer understanding of idioms: inherent senses of the collocates, clausally based constraining factors, and arbitrary idiomatic stipulation. How to do this is an ongoing issue. Finally it is interesting simply to reflect on what studies such as the current one mean for understanding figurative language in general. There have, of course, been a number of studies concentrating on comparative metaphorical language. However, in recent years, these generally assume a Cognitive Linguistic perspective in which figurative language is not treated in terms of idiomaticity as such, but rather as a reflex of conceptual metaphors (e.g., the collections in Sharifian et al. [2008] and Maalej and Yu [2011]).37 The perspective pursued here, however, is that idiomaticity can itself be treated as an independent domain. This allows comparative questions to be asked that can probably only be answered using comparable detailed, corpus-based databases such as the one that serves the current study. It is striking, for instance, that so much idiomaticity should be loaded into a limited number of keywords. Nigerian Arabic might be termed a šaal ‘carry’ language, not an axad ‘take’ language (see the discussion of šaal vs. axad for (24) above). Looked at merely in terms of semantic mapping, in terms of the meanings of the individual idioms, this might appear to risk semantic overload on šaal or \aas or ligi ‘get’. Idioms, however, as explained in detail in section 4, are more than lexical collocations. A single idiomatic lexical collocation can be remarkably efficient in, to speak metaphorically again, building upon basic grammatical structures to define a range of conventionalized meanings. A comparative perspective that looks beyond the lexical idiomatic collocates will do justice to the intricacy of the issue.

2015

JONATHAN OWENS

43

Appendix The lexeme \aas is a rich source of idiom in Nigerian Arabic. Over sixty ostensible idioms have been recorded. A large sample of these is presented below, arranged according to the contextualized attribute extension sense taxonomy in (60). The identifying codes from the database have been left intact. Prominent Part 4.46 ka\abat \aasha ‘she plaited her hair’, lit., ‘she held her hair’ 4.47 boko haraam ka\ab \aas maiduguri ‘Boko Haram has occupied Maiduguri’, lit., ‘Boko Haram held the head of Maiduguri’ 4.48 mine Warab \aas ‘who won’, lit., ‘who hit head’ 4.51 šiil ar¤\aas ‘take the lead’, lit., ‘carry the head’ 4.52 mine akal \aas ‘who won’, lit., ‘who ate head’ Psychological/Intellectual/Sensory State of the Individual (Or Cognitive and Individual Center) REPRESENTS A COGNITIVE STATE OF AN INDIVIDUAL 4.05 karabni fi \aasa ‘he remembered me’, lit., ‘he held me in his head’ 4.06 \aasa heelu ‘he is smart’, lit., ‘his head is sweet’ 4.10 \aasa xafiif ‘he is light-headed/careless’, lit., ‘his head is light’ 4.12 \aasa tagiil ‘he is stubborn’, lit., ‘his head is heavy’ 4.15 fateena lohum \aashum ‘we enlightened them’, lit., ‘we opened them their heads’ 4.18 Warabo leena \aasna ‘they disturbed us’, lit., ‘they hit our heads for us’ 4.24 \aasa gawi ‘he is stubborn’, lit., ‘his head is hard’ 4.25 \aashum ja ‘they came to their senses’, lit., ‘their head came’ 4.54 \aasa bišiil al gœrá ajala ‘he learns quickly’, lit., ‘his head carries studies quickly’ 4.53 \aasa yaabis ‘he is stubborn’, lit., ‘his head is dry’ 4.59 katiiriin millabbidiin \aashum ‘many do not reveal themselves’, lit., ‘many conceal their heads’ 4.69 \aasa šaayil ‘he has problems’, lit., ‘his head is carrying’ REPRESENTS AGENTIVE INDIVIDUAL 4.01 nalumm \aasna ‘we unite’, lit., ‘we gather our head’ 4.03 ka\abna \aasna min ad-duwaas ‘we kept out of the conflict’, lit., ‘we held our head from the conflict’ 4.04 nuku\ub \aasna ‘we are proud’, lit., ‘we hold our head’ 4.08 bitallif \aasa ‘he makes a fool of himself’, lit., ‘he spoils his head’ 4.11 \aasna waahid ‘we are united’, lit., ‘our head is one’ 4.19. ma bidalli \aasa le naadum ‘he doesn’t submit to anyone’, lit., ‘he doesn’t lower his head to a person’ 4.21 bixadim foog \aasa ‘he works for himself, independently’, lit., ‘he works on his head’ 4.26 ligi \aashum ‘he got their support’, lit., ‘he got their head’ 4.32 kalaama bidangir \aasak ‘you will submit to them’, lit., ‘his word bends down your head’ 4.37 hu naadum \aasa ‘he is independent’, lit., ‘he is a person of his head’ 4.39 ligiina \aasna ‘we escaped’, lit., ‘we got our head’ 4.43 xarrabo \aasna ‘they made us flee’, lit., ‘they destroyed our head’ 4.62 al¤bitt sanadat \aasa leehum ‘the girl competes with them’, lit., ‘the girl supports her head towards them’

44

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

Notes Acknowledgments. I would like to thank Robin Dodsworth for comments on an earlier draft of this article as well as Jidda Hassan, Sherif Abdulahi, Ibrahim Adamu, and Kellu Ibrahim for their persistent support in the research on Nigerian Arabic. Research has been supported by the Deutsche Forschungsgemeinschaft (DFG) under grant OW 5/5-1/3, “Idiomaticity and Lexical Realignment in Nigerian Arabic.” Transcription. IPA symbols are used, except that : represents a voiced emphatic implosive, j a voiced alveopalatal affricate, š a voiceless alveopalatal fricative, and y a palatal glide. The source of examples drawn from texts is placed in parentheses after the translation. Abbreviations. The following grammatical abbreviations are used in glosses and elsewhere: 3 = third person; DEF = definite; DT =_______; F = feminine; LIN = linker; M = masculine; NP = noun phrase; O = object; PL, pl. = plural; PP = prepositional phrase; PSSR = possessor; PSV = passive; Q = question marker; RC = relative clause; REL = relative; S = subject; SG, sg. = singular; V = verb; VN = verbal noun. Other abbreviations: ! = semantic or pragmatic anomaly; NA = Nigerian Arabic. 1. The phonetic realization of the lexeme varies between raas, \aas, or \aab, as seen in some of the examples below; in (1c), it appears in its plural form. Low-level phonetic variation such as this is speaker- and context-dependent. A close linguistic definition of it would require a variationist treatment. 2. The logical test can be used to demonstrate a systematic contrast between literal and abstract or idiomatic meanings, and hence an ambiguity in the lexeme \aas. The borderline between polysemy and vagueness discussed in Tuggy’s (1993) article, on the other hand, is subject to various factors, including those discussed in detail in section 4. Arguably, one might want to assimilate what I interpret below as context-sensitive, conditioned contrasts to cases of vagueness. The issue, however, requires an explicit treatment that is not developed in the present article. 3. While Searle dealt ostensibly with metaphor, he does suggest (1980:108) that his stock example, Sally is a block of ice, approaches a fixed expression–in his words, a “dead metaphor,” i.e., an idiom. A glance at the British National Corpus yields eleven tokens of the phrase block of ice. Of these, seven are literal meanings, three are figurative, plus one that I am unable to classify. Of the figurative meanings, one is simply ‘(a body) become cold’ (turn into a block of ice), and two are close to the meaning of Searle’s example, e.g., “the cold effectiveness of a block of ice applied to her nervous system,” i.e., made her become unfeeling. Note that in general the problems that Searle raises in interpreting Sally is a block of ice could be equally shown to be problematic for the idioms discussed in this article. For instance, the difficulty of interpreting raas al¤qalla, lit., ‘head of the grain’ as ‘tassel’ (see (7), (43a)), as opposed to simply ‘top of the grain’, is comparable to the difficulty of discerning which characteristic of a block of ice is relevant for interpreting Sally is a block of ice. 4. In their discussion of over, Brugmann and Lakoff (1988:300) do distinguish between a “minimal specification,” which they do not develop, and a “full specification” approach. In the full specification approach, a high degree of polysemy is represented by image schemas for over. In Lakoff (1987:378) there are said to be over one hundred different senses for over. A minimal specification approach would presumably give greater weight to contextual factors. 5. It is important not to paint all monosemist approaches with the same brush. Lewandowska-Tomaszczyk, for instance, rather contentiously writes, “However, whereas the radical monosemy position defends an abstract, minimal semantic representation for a decontextualized general sense from which polysemic instances are derived by

2015

JONATHAN OWENS

45

contextual (pragmatic) constraints, Cognitive Linguistics tends to defend the view that polysemic senses of one lexical item form interrelated sets” (2010:153). As seen, Searle accounts for polysemy via interpretive principles that are only partly pragmatic in nature, Allwood places as much emphasis on linguistic as on pragmatic context, and Riemer’s core meaning, discussed in this section, is expanded by rules of metonymy and metaphor, not pragmatics. Riemer, it can be added, is neutral as to whether these derivations support a monosemic or polysemic approach. 6. The appeal to extralinguistic context is developed elsewhere among monosemists. Jannsen (2003:101), working within a cognitive linguistics framework, argues, for instance, that translationally distinct utterances in Dutch reflect a single meaning when looked at in terms of inference from our worldly knowledge about walking sticks. 7. I leave out of the discussion Taylor’s (2006) somewhat radical proposal of devolving the monosemy-polysemy debate to what can be termed a “radical usage” approach: “speakers understand these expressions [such as open a book and open your eyes] because they have learned them as such” (2006:75). Taylor here, it would appear, is arguing against interpretation via lexical compositionality. A basic problem in this approach is to ascertain what it is that speakers have learned in these expressions. In particular, how have they learned two different meanings for open your eyes, one (in the traditional terminology) the compositional sense (doctor to patient: you can open your eyes) and the other the idiomatic ‘look reality in the face’? This is the question addressed in section 5 of this article. 8. Whether there are three-member fixed idioms in Nigerian Arabic is a question outside the scope of this article. 9. Arabs reached the Lake Chad area by 1400, having come originally from Upper Egypt. Hence they are a well-established part of the social and cultural environment in the region (see articles in Owens 1994). They constitute one of the largest minority groups in the region which, since before their arrival in the region, has been dominated by a Kanuri-speaking majority. Nigerian Arabic idiomaticity is to a very large degree based on borrowing from Kanuri (Owens 2014). 10. I would like to thank Bernhard Volz, Lars Ackermann, and Lutz Lukas of the Applied Computer Science Faculty, Bayreuth University, for building a morphological segmenter (Ackermann et al. forthcoming) to expedite searching and creating a concordance for the corpus. 11. For instance, for {lamma + \aas}, 47 out of the 308 total tokens of \aas in the corpus, and 47 out of the 461 total tokens of lamma, appear in this idiom; (47+47)/ (308+461) = 94/769 = 0.122 (or 12.2%), the denominator 769 being the sum of the total number of tokens of both lamma and \aas in the corpus. 12. These numbers include variant forms of elements of the idioms. For instance, lamma in the idiom {lamma raas} ‘unite’ has a number of different morphological forms; those that appear in this database include lamma (perfect and imperfect verb), allamma ‘be joined’ (form VII in Arabic derivational terms), lamlam, reduplicated verb form ‘unite repeatedly’, laamm (active participle), malamm ‘a gathering’ (note, not ‘place of gathering’ as in Egyptian Arabic), and lammiin ‘uniting’ (verbal noun). Furthermore, I have included among the figures for {fata raas} its dialectal or idiolectal variant {kaša(h) \aas} (same meaning), which occurs in five of the thirty-four tokens. 13. No tokens of the active participle saari(h) happen to occur in the texts. See Owens (2007b) on the variable value of the final laryngeal in sara(h) and fata(h). 14. This needs to be qualified slightly, however. There are two tokens of the imperfect isœra in the sense ‘climb up, wind about’, applied only to climbing plants. Beans climb up sticks, poles, etc., as in al¤libiya bisra fœ \aasa ‘the beans climb up its top’ (GR 132). The etymology needs interpretation. Probably it is a metaphoric extension of sarah ‘pasture’. Pasturing cattle involves a circuitous search for water and grass, like

46

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

the circuitous climbing of beans and other such plants. These two tokens have been left out of the count. 15. Moreover, anticipating the discussion in section 4.2 below, the comparison indicates that in terms of collocational propensities, literal and idiomatic expressions begin on the same basis. It underscores the fact that what makes idiomaticity is idiosyncratic. 16. The chi-square value of 66.6 (df 2) is significant (p > .000). More relevant observationally is the cline of differences between observed and expected values. Idiomatic \aas expects a far higher degree of continued reference in a following clause, and bagar expects a lower value, with literal \aas falling within its expected range. A more detailed and comprehensive multivariate analysis is provided by Owens and Dodsworth (2015). 17. A further perspective that gives prominence to the grammatical clausal properties of idioms is Nunberg, Sag, and Wasow (1994). They use their observations to argue for the compositionality of idioms, not to define the role of grammatical elements in defining different idioms. 18. Thematic roles are an integral part of the frame construct in FrameNet (http:// framenet.icsi.berkeley.edu/fndrupal/home), based on the work of Fillmore and others on Frame Semantics (e.g., Fillmore 1982). In this approach, frames are defined at different levels of generality, with lower-order frames inheriting frame attributes from higherlevel ones. Frames are realized in individual lexical units, which are said to “evoke” the frame they instantiate. Each lexical unit that instantiates a frame has the attributes of that frame. The frame for “Bringing” describes a certain type of event and is realized by such verb as carry, bring, and fetch. While FrameNet is in principle language-neutral, it effectively is applicable only if there are adequate resources (individuals, time) able to define the various levels and inheritance relations between them, which is not the case with Nigerian Arabic; hence no attempt is made to adapt it to the current data. 19. Anaphora in a narrower sense covers reference relations of referring expressions to other referring expressions within a clause and a larger discourse context, while exophora is situational-based, language-external reference. Exophora pertains in particular to first and second person reference, only rarely to third person (Prince 1981:232). I am aware of no studies on spoken Arabic that highlight the properties of anaphoric vs. exophoric elements, though exophora is recognized as an independent factor group in the discourse studies by Owens and collaborators (e.g., Owens, Dodsworth, and Kohn 2013). In these studies, situationally given pronouns show no special discourse properties against other referring expressions, relative to the parameters used to define reference within discourse. 20. Subject-verb agreement is counted here as anaphoric. Nigerian Arabic being a pro-drop language, (Owens, Dodsworth, and Kohn 2013) a verbal clause more often than not has no overt subject, the person-number-gender inflection on the verb being the main indicator of what the subject is. This marker is therefore referential-anaphoric. 21. That is, not ligo raashum ‘they (masculine) escaped’ (they.M got head-their.M). 22. Cf., for instance, the metaphor for “subject” in the Classical Arabic linguistic tradition, faaýil ‘agent’. 23. A further perspective pointed out by an anonymous reader is that a literal meaning of lammo \aas¤hum might be impossible due to the fact that idiomatic \aas is in the singular. This requires further investigation with consultants. There are contexts where morphologically singular, literal nouns can have a plural reading, e.g., waas¤o wataayir¤hum al¤katiir¤aat ‘they repaired their many cars’ (wataayir ‘cars (pl.)’) vs. waaso watiir¤hum al¤katiiraat ‘they repaired their many cars one by one’ (watiir ‘car (sg.)’). Still, the overwhelming predominance of singular nouns in most idioms considered here is potentially a further aspect of distributed polysemy. 24. Cf. raasa ja ‘his head came’, i.e., ‘he got orientated’.

2015

JONATHAN OWENS

47

25. Lest one suppose that the difficulty of construing ‘get head’ as ‘escape’ or ‘give birth’ is merely due to the attempt to understand Nigerian Arabic from a Western perspective, a large sample of these idioms was presented to speakers of Egyptian Arabic, who were equally perplexed by most of them (Owens 1996, 2014). These native Egyptian intuitions are confirmed in a corpus of about 480,000 words which derives from two sources. One is a part of the LDC (Linguistic Data Consortium Call Home series). A second consists of the considerable corpus collected over many years by the doyens of Arabic dialectology, Manfred Woidich and Peter Behnstedt. They were so kind as to make available all of Behnstedt and Woidich (1987, 1988) in electronic format, as well as Woidich’s extensive material from the data on Farafira, Bahariyya, and Dakhla oases, in addition to material from Cairene Arabic. This corpus represents varieties from all over Egypt. The corpus reveals no tokens of liýi/ligi + \aas (dimaag, etc.) collocations among the 320 (including 128 idiomatic) tokens of \aas (or dimaag). 26. One basic point missing in a foundational discussion of the monosemy-polysemy divide and the ability to systematically represent semantic description at all (Croft 1998; Sandra 1998; Tuggy 1999) is the practical, behavioral reality of language contact, bilingualism, and second language learning. If one cannot collocate \aas correctly, as described here, and, very probably, if one is not aware of the underlying collocational frequencies attendant upon \aas, one cannot know Nigerian Arabic. This has behavioral consequences. If one does not know {lamma \aas} ‘unite’ one cannot effectively express this concept to a monolingual Nigerian Arab (of which there are not a few). Admittedly, this observation does not logically entail that the monosemy-polysemy issue must be confronted, but it does effectively force that issue on a practical level, in that the learning of which collocations with \aas are conventional correlates with learning the idiom itself, i.e., its meaning. What does one learn: collocations, idioms as a whole, abstract meanings? And if abstract meanings, which ones, and how are they structured? 27. See Geeraerts (1997:68—79) for related discussion of the historical development of lexical senses in terms of subsets. Robert (2008:74—79) similarly describes the development of opaque subsenses from originally motivated metonymic or metaphoric associations. For the psycholinguistic tradition, see Glucksberg and McGlone (1999:1543). 28. In Ibn Man£ —ur’s compendious Lisaan al-ýArab dictionary from the thirteenth century (11:570), a qandal, with variants qundaal and qandawiil (where q corresponds to NA g) is described as a camel, or a person with a large (Waxm) head, or a long-necked person. Taking this meaning at its face value, this would imply that NA ganduul (assuming that it has the same etymological root) in the sense of ‘tassel of corn’ (i.e., long and flowing) would derive from ‘having an appearance like that of a long neck’, i.e., it would owe its historical etymology to the same metaphoric association as \aas al¤qalla, a comparison to the physical part of a person or animal. 29. Except, of course, in the different sense of ‘roof’ in this case. 30. With (25c) and (25d) registered as one idiom. 31. In fact, there are idioms such as {lamma \aas} ‘unite’ that do not map onto discrete lexemes in Nigerian Arabic, i.e., there is no single verb meaning ‘unite’ in the dialect. The mapping, rather, is to a postulated semantic space that may, as with {bišiil \aas} = bilýallam, also be accessed by a single lexeme, but need not be. 32. For instance, the parameter of coreference of a possessive pronoun with the subject (conjoint or disjoint) is crucial for many idioms, and conceivably is subject to cross-idiom generalization. 33. The meaning given is the meaning of the idiom. The complete taxonomy has three parts. It begins with the hierarchicalized attribute extension sense taxonomy, as described in 5.1.1. For šaal reference is to the provisional Figurative part of (57). The “collocational stipulations” in square brackets “[. . .]” specify what lexemes need to be

48

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

chosen to make the successful idiom. In a fully-fledged sense taxonomy, the stipulation could be abbreviated to “pair sense X with sense Y,” where X and Y are numbers representing individual senses. The third piece of information, enclosed within slashes, is the clausal contextual elements discussed in section 4; for instance, in the idiom šaal a\¤\aas ‘take the lead’, \aas must be definite. As presented here, the contextual properties are unordered, and include thematic information (\aas = Patient), morphological information (e.g., \aas must be definite), and referential information (disjoint reference). This is for expository purposes; the systematization of frame elements is outside the scope of this article. The pattern repeats itself for the other main components; for example, {lamma \aas} = ‘unite’ requires \aas to have a plural possessor. 34. One can ask if the semantics of the expression can even be expressed without both metaphoric and metonymic metalanguage. In a characterization such as “it is as if my head represents me,” “as if” is a prototypical representation of metaphor and “represent” is a prototypical metonymic predicate. 35. Riemer (2005:157) writes that “linguistic analysis cannot aim to supply a unique analysis of any object language meaning, and should rather adopt an interpretive, instrumental role” (2005:157). He elsewhere speaks of “an infinite number of interpretive processes by which the matching between a putative conceptual representation and its denotation is implemented” (2005:158). This idea is echoed in a slightly different context by Haser (2005:181¤93), who emphasizes that source lexemes may be mapped onto different conceptual metaphors in a multiplicity of ways (e.g. ‘defend’ might map on to ARGUMENT IS WAR, ARGUMENT IS PRESERVATION, or other conceptual metaphors), that a multiplicity of conceptual metaphors can be attributed to one concept (e.g., ARGUMENT IS: WAR/PRESERVATION/PHYSICAL SUPPORT/. . .), and that one and the same metaphorical expression may map on to multiple conceptual metaphors (2005:228). 36. Moreover, within studies of idioms different categories of idioms are usually recognized. Glucksberg (2001:73—76), for instance, distinguishes four classes: noncompositional idioms (e.g., by and large), nonanalyzable idioms (e.g., kick the bucket), compositional and transparent idioms, and metaphorically combining expressions. 37. Or, similarly, as Riemer (2005:167) asks, given a theory of lexical organization based on the idea of a core meaning and extensions derived via metonymic and metaphoric processes, are there universal tendencies among languages in the world as to how metonymy and metaphor are used to expand meanings? However, as Riemer would appear to imply, even if a theoretical framework can be agreed upon, we are nowhere near even having adequate databases to undertake such research.

References Ackermann, Lars, Lutz Lukas, J. Owens, and B. Volz forthcoming A Morphological Segmenter for Arabic. Proceedings of the Tenth International Conference on Arabic Dialectology, Doha, November 2013. Allwood, Jens 2003 Meaning Potentials and Context: Some Consequences for the Analysis of Variation in Meaning. In Cognitive Approaches to Lexical Semantics, edited by Hubert Cuyckens, René Dirven, and John Taylor, 29—65. Berlin: Mouton de Gruyter. Barcelona, Antonio 2002a On the Plausibility of Claiming a Metonymic Motivation for Conceptual Metaphor. In Metaphor and Metonymy at the Crossroads: A Cognitive Perspective, edited by Antonio Barcelona, 31—58. Berlin: Mouton de Gruyter.

2015 2002b

JONATHAN OWENS

49

Clarifying and Applying the Notions of Metaphor and Metonymy within Cognitive Linguistics: An Update. In Metaphor and Metonymy in Comparison and Contrast, edited by René Dirven and Ralf Pörings, 207—77. Berlin: Mouton de Gruyter. Barnden, John 2010 Metaphor and Metonymy: Making Their Connections More Slippery. Cognitive Linguistics 21:1—34. Behnstedt, Peter, and Manfred Woidich 1987 Die ägyptisch-arabischen Dialekte. Band 3: Delta-Dialekte. Wiesbaden: Reichert. 1988 Die ägyptisch-arabischen Dialekte. Band 3: Niltaldialekte, Oasendialekte. Wiesbaden: Reichert. Brugman, Claudia, and George Lakoff 1988 Cognitive Topology and Lexical Networks. In Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence, edited by Geert Adriaens, Steven L. Small, Garrison W. Cotrell, and Michael K. Tannenhaus, 477—507. San Mateo, Calif.: Morgan Kaufman. Cacciari, Christina, and Patrizia Tabossi 1988 The Comprehension of Idioms. Journal of Memory and Language 27:668— 83. Charles, Walter 1988 The Categorization of Sentential Contexts. Journal of Psycholinguistic Research 17:403—11. Cresti, E., and Massimo Moneglia, eds. 2005 C-ORAL-ROM: Integrated Reference Corpora for Spoken Romance Languages. Amsterdam: John Benjamins. Croft, William 1998 Linguistic Evidence and Mental Representations. Cognitive Linguistics 9: 151—73. Cruse, Alan 1986 Lexical Semantics. Cambridge: Cambridge University Press. 2000 Aspects of the Micro-structure of Word Meanings. In Polysemy: Theoretical and Computational Approaches, edited by Yael Ravin and Claudia Leacock, 30—51. Oxford: Oxford University Press. Deane, Paul 1988 Polysemy and Cognition. Lingua 75:325—61. Dirven, René 2002 Metonymy and Metaphor: Different Mental Strategies of Conceptualization. In Metaphor and Metonymy in Comparison and Contrast, edited by René Dirven and Ralf Pörings, 75—112. Berlin: Mouton de Gruyter. Dowty, David 2000 “The Garden Swarms with Bees” and the Fallacy of “Argument Alternation.” In Polysemy: Theoretical and Computational Approaches, edited by Yael Ravin and Claudia Leacock, 111—28. Oxford: Oxford University Press. Dunbar, George 2001 Towards a Cognitive Analysis of Polysemy, Ambiguity, and Vagueness. Cognitive Linguistics 12:1—14. Evans, Vyvyan 2009 How Words Mean: Lexical Concepts, Cognitive Models, and Meaning Construction. Oxford: Oxford University Press.

50

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

Fellbaum, Christina 1993 The Determiner in English Idioms. In Idioms: Processing, Structure, and Interpretation, edited by Christina Cacciari and Patrizia Tabossi, 271—95. Hillsdale, N.J.: Lawrence Erlbaum. Fillmore, Charles J. 1982 Frame Semantics. In Linguistics in the Morning Calm: Selected Papers from SICOL-1981, edited by Linguistic Society of Korea, 111—37. Seoul: Hanshin. Geeraerts, Dirk 1997 Diachronic Prototype Semantics: A Contribution to Historical Lexicology. Oxford: Clarendon Press. Glucksberg, Sam 1993 Idiom Meanings and Allusional Content. In Idioms: Processing, Structure, and Interpretation, edited by Christina Cacciari and Patrizia Tabossi, 3—26. Hillsdale, N.J.: Lawrence Erlbaum. 2001 Understanding Figurative Language: From Metaphors to Idioms. Oxford: Oxford University Press. Glucksberg, Sam, Mary Brown, and Matthew McGlone 1993 Conceptual Metaphors Are Not Automatically Accessed during Idiom Comprehension. Memory and Cognition 21:711—19. Glucksberg, Sam, and Matthew McGlone 1997 Property Attribution in Metaphor Comprehension. Journal of Memory and Language 36:50—67. 1999 When Love Is Not a Journey: What Metaphors Mean. Journal of Pragmatics 31:1541—58. Goosens, Louis 2002 Metaphtonymy: The Interaction of Metaphor and Metonymy in Expressions for Linguistic Action. In Metaphor and Metonymy in Comparison and Contrast, edited by René Dirven and Ralf Pörings, 349—77. Berlin: Mouton de Gruyter. Haser, Verena 2005 Metaphor, Metonymy, and Experientialist Philosophy: Challenging Cognitive Semantics. Berlin: Mouton de Gruyter. Hopper, Paul, and Sandra Thompson 1980 Transitivity in Grammar and Discourse. Language 56:251—99. 1984 The Discourse Basis for Lexical Categories in Universal Grammar. Language 60:703—52. Janssen, Theo 2003 Monosemy versus Polysemy. In Cognitive Approaches to Lexical Semantics, edited by Hubert Cuyckens, René Dirven, and John Taylor, 93—122. Berlin: Mouton de Gruyter. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Lewandowska-Tomaszczyk, Barbara 2010 Polysemy, Prototypes, and Radial Categories. In The Oxford Handbook of Cognitive Linguistics, edited by Dirk Geeraerts and Hubert Cuyckens, 139—169. Oxford: Oxford University Press. Maalej, Zouheir A., and Ning Yu, eds. 2011 Embodiment via Body Parts: Studies from Various Languages and Cultures. Amsterdam: John Benjamins.

2015

JONATHAN OWENS

51

Man£ —ur, Ibn Abu FaWl. n.d.Lisaan al¤öArab. Beirut: Dar Al-Fikr. McGlone, Matthew 1996 Conceptual Metaphors and Figurative Language Interpretation: Food for Thought? Journal of Memory and Language 35:544—65. 2007 What Is the Explanatory Value of a Conceptual Metaphor? Language and Communication 27:109—26. Miller, George, and Claudia Leacock 2000 Lexical Representations for Sentence Processing. In Polysemy: Theoretical and Computational Approaches, edited by Yael Ravin and Claudia Leacock, 152—60. Oxford: Oxford University Press. Nunberg, Geoffrey, Ivan Sag, and Thomas Wasow 1994 Idioms. Language. 70:491—538. Owens, Jonathan 1993 A Grammar of Nigerian Arabic. Wiesbaden: Harrassowitz. 1996 Idiomatic Structure and the Theory of Genetic Relationship. Diachronica 13:283—318. 1998 Neighborhood and Ancestry: Variation in the Spoken Arabic of Maiduguri, Nigeria. Amsterdam: John Benjamins. 2005 Ø Forms and Lexical Insertion in Codeswitching: A Processing-Based Account. Bilingualism: Language and Cognition 8:23—38. 2007a Close Encounters of a Different Kind: Two Types of Insertion in Nigerian Arabic Codeswitching. In Arabic in the City: Issues in Dialect Contact and Language Variation, edited by Catherine Miller, Dominique Caubet, Janet Watson, and Enam Al-Wer, 249—74. London: Routledge. 2007b Paradigmatic Stability and Final Laryngeals in Nigerian Arabic, or Why Arabic Language History Repeats Itself, Without Actually Doing So. In Approaches to Arabic Linguistics: Presented to Kees Versteegh on the Occasion of His Sixtieth Birthday, edited by Everhard Ditters and Harald Motzki, 639—52. Leiden: Brill. 2014 Many Heads Are Better than One: The Spread of Motivated Opacity via Contact. Linguistics 52:125—65. Owens, Jonathan, ed. 1994 Arabs and Arabic in the Lake Chad Region. Special edition. Sprache und Geschichte in Afrika 14. Owens, Jonathan, and Robin Dodsworth 2015 Semantic Mapping: What Happens to Idioms in Discourse. MS. Owens, Jonathan, Robin Dodsworth, and Mary Kohn 2013 Subject Expression and Discourse Embeddedness in Emirati Arabic. Language Variation and Change 25:255—85. Owens, Jonathan, Robin Dodsworth, and Trent Rockwood 2009 Subject-Verb Order in Spoken Arabic: Morpholexical and Event-based Factors. Language Variation and Change 21:39—67. Owens, Jonathan, and Jidda Hassan 2012 In Their Own Voices, in Their Own Words: A Corpus of Spoken Nigerian Arabic. http://www.neu.uni-bayreuth.de/de/Uni_Bayreuth/Fakultaeten/ 4_Sprach_und_Literaturwissenschaft/islamwissenschaft/arabistik/en/ Idiomaticity__lexical_realignment__and_semantic_change_in_ spoken_arabic/Nigerian_Arabic/index.html. Prince, Ellen F. 1981 Toward a Taxonomy of Given-New Information. In Radical Pragmatics, edited by Peter Cole, 223—55. New York: Academic Press.

52

ANTHROPOLOGICAL LINGUISTICS

57 NO. 1

Quine, Willard 1960 Word and Object. Cambridge, Mass.: Massachusetts Institute of Technology. Radden, Günter, and Zoltán Kövecses 2007 Towards a Theory of Metonymy. In The Cognitive Linguistics Reader, edited by Vyvyan Evans, Benjamin K. Bergen, and Jörg Zinken, 335—59. London: Equinox. Ravin, Yael, and Claudia Leacock 2000 Polysemy: An Overview. In Polysemy: Theoretical and Computational Approaches, edited by Yael Ravin and Claudia Leacock, 1—29. Oxford: Oxford University Press. Riemer, Nick 2005 The Semantics of Polysemy: Reading Meaning in English and Warlpiri. Berlin: Mouton de Gruyter. Ritchie, L. David 2006 Context and Connection in Metaphor. New York: Palgrave Macmillan. Robert, Stéphane 2008 Words and Their Meanings. In From Polysemy to Semantic Change: Towards a Typology of Lexical Semantic Associations, edited by Martine Vanhove, 55—92. Amsterdam: John Benjamins. Sandra, Dominiek 1998 What Linguists Can and Can’t Tell You about the Human Mind: A Reply to Croft. Cognitive Linguistics 9:361—78. Searle, John 1980 Metaphor. In Metaphor and Thought, edited by Andrew Ortony, 92—125. Cambridge: Cambridge University Press. Sharifian, Farzad, René Dirven, Ning Yu, and Susanne Niemeier, eds. 2008 Culture, Body, and Language: Conceptionalizations of Internal Body Organs, Cultures, and Languages. Berlin: Mouton de Gruyter. Sinclair, John 2004 Trust the Text: Language, Corpus, and Discourse. London: Routledge. Swinney, David, and Anne Cutler 1979 The Access and Processing of Idiomatic Expressions. Journal of Verbal Learning and Behavior 18:523—34. Taylor, John 1995 Approaches to Word Meaning: The Network Model (Langacker) and the Two-Level Model (Bierwisch) in Comparison. In Current Approaches to the Lexicon: A Selection of Papers Presented at the Eighteenth LAUD Symposium, Duisburg, March 1993, edited by René Dirven and Johan Vanparys, 3—26. Frankfurt: Lang. 2006 Polysemy and the Lexicon. In Cognitive Linguistics: Current Applications and Future Perspectives, edited by Gitte Kristiansen, Michel Achard, René Dirven, and Francisco Ruiz de Mendoza Ibanez, 51—80. Berlin: Mouton de Gruyter. Titone, Debra, and Cynthia Connine 1999 On the Compositional and Noncompositional Nature of Idiomatic Expressions. Journal of Pragmatics 31:1655—74. Tuggy, David 1993 Ambiguity, Polysemy, and Vagueness. Cognitive Linguistics 4:273—90. 1999 Linguistic Evidence for Polysemy in the Mind: A Response to William Croft and Dominiek Sandra. Cognitive Linguistics 10:343—68.

2015

JONATHAN OWENS

53

The Nawatl Verb kVsa: A Case Study in Polysemy. In Cognitive Approaches to Lexical Semantics, edited by Hubert Cuyckens, René Dirven, and John Taylor, 232—62. Berlin: Mouton de Gruyter. 2010Schematicity. In The Oxford Handbook of Cognitive Linguistics (online version), edited by Dirk Geeraerts and Hubert Cuyckens. Oxford: Oxford University Press. Warren, Beatrice 1999 Laws of Thought and Lexical Change. In Historical Semantics and Cognition, edited by Andreas Blank and Peter Koch, 215—32. Berlin: Mouton de Gruyter. 2002 An Alternative Account of the Interpretation of Referential Metonymy and Metaphor. In Metaphor and Metonymy in Comparison and Contrast, edited by René Dirven and Ralf Pörings, 113—30. Berlin: Mouton de Gruyter. Yavrumyan, Marat, and Jonathan Owens 2007 The Participle. In Encyclopedia of Arabic Language and Linguistics, edited by Kees Versteegh, 541—46. Leiden: Brill. 2003