Tests for polysemy

The idea of ‘conceptual relation’ (or ‘semantic relation’) featuring in the definition of polysemy discussed above is notoriously unconstrained. If polysemy is defined as ‘the possession of conceptually related senses by a single word’, the fact that we can conceive of a conceptual relation between any two meanings means that we need never diagnose homonymy. The meanings ‘pound’ and ‘book’, for instance, might be conceptually related in that books were typically quite heavy objects, weighing several pounds. To give another example, the French noun pic means both ‘woodpecker’ and ‘peak’. These are not, however, historically related: the ‘woodpecker’ sense comes from popular Latin *piccus, the ‘peak’ meaning from Spanish pico, each of which took on the same pronunciation after it entered French (Rey 1998: pic). However, in order to motivate an analysis of pic as polysemous we could posit a conceptual link between the shape of a woodpecker’s beak and a steep mountain-top: the only reason that mod ern French dictionaries do not do this is the separate origin of the two words. Even so, it might be objected, who is to say that contemporary French speakers do not relate the two meanings in just this way? Ordinary speakers have no access to the etymological history of their own language; when acquiring French, native speakers would have heard simply the single form [pi:k], which they would have learnt to associate with two meanings. Might not they think about the two meanings as related by shape?

 In languages whose histories are not well known, cases like this pose a considerable problem, and the uncertainty is aggravated when we have no clear sense of the plausibility of a conceptual relation between the meanings involved. In Warlpiri, for example, the verb parntarni means both ‘hit on the head’ and ‘name, call’. Without a thorough appreciation of the cultural context, it is entirely unclear whether it would be possible to propose a plausible conceptual link between these two ideas. And even if we did have a detailed knowledge of the cultural context, it’s not obvious what would constitute adequate evidence that the two meanings were ‘conceptually related’. Clearly, then, the idea of ‘conceptual relation’ will not allow us to decide conclusively between cases of polysemy, monosemy and homonymy. What is needed is a more precise criterion which will discriminate the three cases unambiguously. Linguists have devised a number of polysemy tests, of which we will discuss the most important.

The oldest type of polysemy test, the definitional test, due originally to Aristotle (Posterior Analytics II.13; Geeraerts 1993: 230), identifies the number of senses of a word with the number of separate definitions needed to convey its meaning accurately. A word has more than one meaning if there is no single definition which can be applied to all of its uses, and it has no more meanings than the number of maximally general definitions necessary to defi ne its complete denotation. This was the criterion we applied in (28) above in order to delimit the five separate senses of French pièce, and it corresponds to the common-sense idea that a word has as many senses as it requires separate semantic descriptions. Thus, the definitional criterion demonstrates the non monosemy of the noun quarry, since there would seem to be no definition which could simultaneously cover the meanings ‘site dedicated to the open air excavation of stone’ and ‘object of a search or hunt’. Similarly, there seems to be no single definition capable of describing the meanings ‘palace’ and ‘palate’ of the French noun palais. (Note that the definitional criterion will not of itself distinguish polysemy and homonymy.)

Definitional tests for polysemy are widely rejected (Geeraerts 1993; Schütze 1997: 69; Fodor 1998; Dunbar 2001). The most significant problem with them is that, contrary to the beliefs of their proponents, they in fact presuppose that the number of meanings to be defined is already known (Geeraerts 1993: 236). Ironically, therefore, far from being a test of polysemy, they actually require that the question of the number of senses held by a lexical item is already resolved. To see this, let us once again take an example from French and consider the adjective drôle, which can be defined in two different ways, shown in (29) a and b.

Is drôle polysemous or not? The definitional criterion will not help us to decide, since two definitional strategies, each of which gives a different answer, are equally possible and there is not any obvious way to distinguish between them. On strategy (a), drôle has two distinct meanings and is therefore polysemous; on strategy (b) it is monosemous. It might be thought that (29b) is a rather unsatisfactory definition, only possible because of the convenient presence in English of a word which covers the same semantic territory as drôle in French; funny in English clearly covers two distinct notions which a definition should distinguish. We can, however, easily answer this objection by rephrasing (29b) as (30):

This definition combines the two cases in (29a) into a single disjunctive definition (one that contains two clauses linked by ‘or’), thereby preserving the semantic analysis of drôle while abandoning its distinction into two meaning components. On purely formal grounds, there is nothing to distinguish these definitions of drôle: they are all equally accurate, in the sense that they may all be truthfully substituted for the definiendum (see 2.4). Yet they do not resolve the question of the monosemy or polysemy of the adjective.

Another serious problem with the definitional test is that the number of senses it diagnoses for the definiendum will vary according to the meta language in which the definitions are couched. The Kukatja (Pama Nyungan, Australia; Valiquette 1993) verb yungkala is defined in English as meaning either ‘throw and pelt’ or ‘grind’ (it also has other senses which do not concern us here). On the definitional criterion, therefore, it is shown to be polysemous. But if we change the defining metalanguage to Walmajarri (Pama-Nyungan, Australia), a related Australian language, we could simply propose the single definition luwarnu, a verb which is also defined in English as ‘pelt, grind’. With Walmajarri as the defining meta language, then, yungkala turns out to be monosemous. On the basis of this type of example, we can conclude that definitions should not be appealed to as evidence for the polysemy or monosemy of a lexical item.

Another frequently suggested test for polysemy is the logical test (first advanced by Quine 1960). A word (or phrase) is polysemous on this test if it can be simultaneously true and false of the same referent. The reasoning behind this test is that a word could only be simultaneously affirmed and denied if the affirmation and the denial applied to different meanings; otherwise, language would be self-contradictory. Examples of simultaneous affirmation and denial of the same word are given in (31), with the particular sense in question mentioned in brackets:

The adoption of this test, however, would require us to diagnose polysemy (ambiguity) in many cases where we would not, in fact, want to recognize any more than a single meaning for the word in question:

Instead of demonstrating polysemy, what seems to be happening in these utterances is that the speaker is simultaneously entertaining two different points of view, under only one of which the description applies. From one point of view, for example – that according to which windows can be opened – the referent of (32a) qualifies as a window; from the opposite point of view, it does not. To diagnose polysemy, however, in every lexical item that was amenable to this sort of perspectivization would leave virtu ally no monosemous words in the lexicon.

QUESTION Devise some other examples like those in (32) involving the simultaneous affirmation and denial of different aspects of a word’s meaning. In which cases would you want to say that the word was polysemous? What are your motivations?

 A particularly common variety of test used to distinguish between polysemy (ambiguity) and monosemy (vagueness) are the so-called linguistic tests, which involve constructions which predicate the same information of two different subjects. In order not to sound bizarre, punning or just awkward, these constructions require that the same information be predicated of both subjects. For example, the and so construction in (33a) would not be appropriate if the quartet are playing a Schoenberg string quartet and Real Madrid (a football team) are playing sport; rather, it is only appropriate if the two types of playing are the same, as in (33b):

Examples of constructions which, like (33a), are bizarre, punning or awk ward, are referred to as crossed or zeugmatic (Greek zeugma ‘yoke’), since they cross or ‘yoke together’ notions which do not belong together. As a result of the contrast between (33a) and (33b), some linguists (Lakoff 1970, Zwicky and Sadock 1975) have suggested that constructions like and so can be used to differentiate between polysemous and monosemous expressions. Thus, (33a) demonstrates that play is polysemous between the sense ‘perform a musical piece’ and ‘engage in a sporting activity’. Similarly, the fact that (34) is not appropriate when intended with the bracketed senses testifies to the polysemy of mad:

Constructions using and so are far from being the only ones to require this sort of identity between the two parts of the predication. Thus, the pronoun it in (35) has to be understood as coreferential (anaphoric) with its antecedent, time. But since two different senses of time are intended in (35), the resulting sentence takes on a ‘punning’ quality, which has been taken as evidence of the polysemy of time with respect to the bracketed senses:

One major problem with the linguistic test is that whether or not a sentence seems punning, bizarre or awkward is open to significant variation between subjects. Indeed, even the reactions of a single subject to the same sentence may differ at different times. For the present author, for example, the following sentences (Riemer 2005: 141) have in the past seemed both awkward and normal:

Because of this shifting status, the linguistic test would not seem to offer the stable results required for judgements of semantic structure.

Furthermore, as pointed out by Geeraerts (1993: 238), the linguistic test cannot be relied on to give correct results where the polysemy of the word in question is not in doubt. Consider for example (37):

There is nothing awkward or peculiar about this sentence used in the context of a paper deciding to change its format from a broadsheet to a tabloid. Yet newspaper initially refers to the management in charge of publishing the physical newspaper, whereas it, which should be coreferential with this, refers to the physical object itself. Pretheoretically, we clearly recognize two distinct meanings of newspaper, the ‘management/board of directors’ sense and the ‘material object’ sense. Yet these different meanings do not show up on the linguistic test.

QUESTION A possible response to this objection would be that our pre theoretical ideas about the polysemy of newspaper are simply wrong. Is this reasonable?

Another problem with the linguistic test is that it ignores the difference between the sense and reference of the lexemes in question. As pointed out by Tuggy (1993), the linguistic test is sensitive to the referents of the terms involved. For example, sentences on the pattern of (38) have been used to demonstrate polysemy, in this case polysemy of the verb court:

The zeugmatic character of this sentence justifies the postulation of two separate meanings of the verb court: ‘woo’, which is associated with the object Tina, and ‘knowingly risk’, associated with the object disaster. It is the fact that each object corresponds to a different sense of court that gives (38) its zeugmatic quality. We can, however, imagine two different contexts in which (38) might be uttered (Riemer 2005: 141). In the first, the speaker means that in courting Tina, Hank is courting a disaster. In this case, Tina and disaster ultimately both refer to the individual Tina. The second con text is one in which Tina and disaster are in no way coreferential: where, for example, at the same time as ‘courting’ Tina, Hank is also, unrelatedly, contemplating a disastrous career-change. This suggests that it is the referent, not the sense, of the lexeme to which the linguistic test is sensitive. Questions of polysemy and monosemy, which concern sense, not reference, cannot therefore be illuminated by these phenomena.

اخر الاخبار

اخبار العتبة العباسية المقدسة

قسم الصيانة يقيم دورة تدريبية عن آلية تنصيب مطافئ الحريق اليدوية وإدامتها لملاكاته

قسم الشؤون الخدمية ينفذ حملة لغسل محيط العتبة العباسية المقدسة وتنظيفه

ضمن برنامج تأهيل المنتسبين الجدد.. قسمُ الشؤون الدينية يطلق دورته الفقهية الـ59 لمنتسبي العتبة العباسية المقدسة

ضمن مقتنياته التراثية.. متحف الكفيل يوثّق تاجًا ذهبيًّا مرصَّعًا بالأحجار الكريمة

المزيد

الآخبار الصحية

المركز الاوروبي : كل ركاب سفينة "هانتا" مخالطون معرضون لخطر كبير

ألمانيا.. رئيس الأطباء يدعو للإسراع بتطبيق "ضريبة السكر"

بين الفائدة والضرر.. جدل حول تأثير الشاي والقهوة على العظام

أمراض خطيرة.. أبحاث تكشف مخاطر الجلوس لفترة طويلة

المزيد

الآخبار التكنلوجية

سبع حقائق غير متوقعة عن موجات الراديو

الأجسام الغريبة.. ماذا رأى رواد "أبولو 17" على القمر؟

حرارة المحيطات عند مستويات مقلقة.. وعودة "إل نينيو" تقترب

باحثون يحددون تأثير القهوة على الميكروبيوم والمزاج

المزيد

مواضيع ذات صلة

Polysemy, monosemy and homonymy

Polysemy and meaning division

Componential analysis

Connective Adjuncts

Synonymy

Taxonomy

Hyponymy

Meronymy

Antonymy

Lexical relations

Semantics and pragmatics

Relevance theory