5.3: Lexical ambiguity

5.3.1 Ambiguity, vagueness, and indeterminacy

In Chapter 2 we discussed cases of lexical ambiguity like those in (2). These sentences are ambiguous because they contain a word-form which has more than one sense, and as a result can be used to refer to very different kinds of things. For example, we can use the word case to refer to a kind of container or to a legal proceeding; lies can be a noun referring to false statements or a verb specifying the posture or location of something. These words have a variety of referents because they have multiple senses, i.e., they are ambiguous. And as we noted in Chapter 2, the truth value of each of these sentences in a particular context will depend on which sense of the ambiguous word is chosen.

(2) a. The farmer allows walkers to cross the field for free, but the bull charges.
b. Headline: Drunk gets nine months in violin case.
c. Headline: Reagan wins on budget, but more lies ahead.

However, there are other kinds of variable reference as well, ways in which a word can be used to refer to different sorts of things even though it may have only a single sense. For example, I can use the word cousin to refer to a child of my parent’s sibling, but the person referred to may be either male or female. Similarly, the word kick means to hit something with one’s foot, but does not specify whether the left or right foot is used.¹ We will say that the word cousin is indeterminate with respect to gender, and that the word kick is indeterminate with respect to which foot is used.² We will argue that such examples are not instances of lexical ambiguity: neither of these cases requires us to posit two distinct senses for a single word form. Our basis for making this claim will be discussed in §5.3.2 below.

Another kind of variable reference is observed with words like tall or bald. How tall does a person have to be to be called “tall”? How much hair can a person lose without being considered “bald”? Context is a factor; a young man who is considered tall among the members of his gymnastics club might not be considered tall if he tries out for a professional basketball team. But even if we restrict our discussion to professional basketball players, there is no specific height (e.g. two meters) above which a player is considered tall and below which he is not considered tall. We say that such words are vague, meaning that the limits of their possible denotations cannot be precisely defined.³

Kennedy (2011) mentions three distinguishing characteristics of vagueness. First, context-dependent truth conditions: we have already seen that a single individual may be truly said to be tall in one context (a gymnastics club) but not tall in another (a professional basketball team). This is not the case with indeterminacy; if a certain person is my cousin in one context, he or she will normally be my cousin in other contexts as well.

Second, vague predicates have borderline cases. Most people would probably agree that a bottle of wine costing two dollars is cheap, while one that costs five hundred dollars is expensive. But what about a bottle that costs fifty dollars? Most people would probably agree that Einstein was a genius, and that certain other individuals are clearly not. But there are extremely bright people about whom we might disagree when asked whether the term genius can be applied to them; or we might simply say “I’m not sure”. Such borderline cases do not typically arise with indeterminacy; we do not usually disagree about whether a certain person is or is not our cousin.

Gillon (1990) provides another example:

Vagueness is well exemplified by such words as city. Though a definite answer does exist as to whether or not it applies to Montreal [1991 population: 1,016,376 within the city limits] or to Kingsville (Ontario) [1991 population: 5,716]; nonetheless, no definite answer exists as to whether or not it applies to Red Deer (Alberta) [1991 population: 58,145] or Moose Jaw (Saskatchewan) [1991 population: 33,593]. Nor is the lack of an answer here due to ignorance (at least if one is familiar with the geography of Western Canada): no amount of knowledge about Red Deer or Moose Jaw will settle whether or not city applies. Any case in which further knowledge will settle whether or not the expression applies is simply not a case evincing the expression’s vagueness; rather it evinces the ignorance of its user… Vagueness is not alleviated by the growth of knowledge, ignorance is.

Third, vague predicates give rise to “little-by-little” paradoxes.⁴ For example, Ringo Starr was clearly not bald in 1964; in fact, the Beatles’ famous haircut was an important part of their image during that era. Now if in 1964 Ringo had allowed you to pluck out one of his hairs as a souvenir, he would still not have been bald. It seems reasonable to assume that a man who is not bald can always lose one hair without becoming bald. But if Ringo had given permission for every person in Europe to pluck out one of his hairs, he would have become bald long before every fan was satisfied. But it would be impossible to say which specific hair it was whose loss caused him to become bald, because bald is a vague predicate.

Another property which may distinguish vagueness from indeterminacy is the degree to which these properties are preserved in translation. Indeterminacy tends to be language-specific. There are many interesting and well-known cases where pairs of translation equivalents differ with respect to their degree of specificity. For example, Malay has no exact equivalent for the English words brother and sister. The language uses three terms for siblings: abang ‘older brother’, kakak ‘older sister’, and adek ‘younger sibling’. The term adek is indeterminate with respect to gender, while the English words brother and sister are indeterminate with respect to relative age.

Mandarin has several different and more specific words which would all be translated by the English word uncle: 伯伯 (bóbo) ‘father’s elder brother’; 叔叔 (shūshu) ‘father’s younger brother’; 姑丈 (gūzhàng) ‘father’s sister’s husband’; 舅舅 (jiùjiu) ‘mother’s brother’; 姨丈 (yízhàng) ‘mother’s sister’s husband’.⁵ Thus the English word uncle is indeterminate with respect to various factors that are lexically distinguished in Mandarin.

The English word carry is indeterminate with respect to manner, but many other languages use different words for specific ways of carrying. Tzeltal, a Mayan language spoken in the State of Chiapas (Mexico), is reported to have twenty-five words for ‘carry’:⁶

(3) 1. cuch ‘carry on one’s back’
2. q’uech ‘carry on one’s shoulder’
3. pach ‘carry on one’s head’
4. cajnuc’tay ‘carry over one’s shoulder’
5. lats’ ‘carry under one’s arm’
6. chup ‘carry in one’s pocket’
7. tom ‘carry in a bundle’
8. pet ‘carry in one’s arms’
9. nol ‘carry in one’s palm’
10. jelup’in ‘carry across one’s shoulder’
11. nop’ ‘carry in one’s fist’
12. lat’ ‘carry on a plate’
13. lip’ ‘carry by the corner’

14. chuy ‘carry in a bag’
15. lup ‘carry in a spoon’
16. cats’ ‘carry between one’s teeth’
17. tuch ‘carry upright’
18. toy ‘carry holding up high’
19. lic ‘carry dangling from the hand’
20. bal ‘carry rolled up (like a map)’
21. ch’et ‘carry coiled up (like a rope)’
22. chech ‘carry by both sides’
23. lut’ ‘carry with tongs’
24. yom ‘carry several things together’
25. pich’ ‘carry by the neck’

In contrast, words which are vague in English tend to have translation equivalents in other languages which are also vague. This is because vagueness is associated with certain semantic classes of words, notably with scalar adjectives like big, tall, expensive, etc. Vagueness is a particularly interesting and challenging problem for semantic analysis, and we will discuss it again in later chapters.

5.3.2 Distinguishing ambiguity from vagueness and indeterminacy

The Spanish word llave can be used to refer to things which would be called key, faucet or wrench/spanner in English.⁷ How do we figure out whether llave has multiple senses (i.e. is ambiguous), or whether it has a single sense that is vague or indeterminate? A number of linguistic tests have been proposed which can help us to make this decision.

The most common tests are based on the principle that distinct senses of an ambiguous word are antagonistic.⁸ This means that two senses of the word cannot both apply simultaneously. Sentences which seem to require two senses for a single use of a particular word, like those in (4), are called puns.

(4) a. The hunter went home with five bucks in his pocket.
b. The batteries were given out free of charge.
c. I didn’t like my beard at first. Then it grew on me.
d. When she saw her first strands of gray hair, she thought she’d dye.
e. When the chair in the Philosophy Department became vacant, the Appointment Committee sat on it for six months.⁹

A clash or incompatibility of senses for a single word in sentences containing a co-ordinate structure, like those in (5), is often referred to using the Greek term zeugma (pronounced [ˈzuɡmə]).

(5) a. Mary and her visa expired on the same day.¹⁰
b. He carried a strobe light and the responsibility for the lives of his men.¹¹
c. On his fishing trip, he caught three trout and a cold.¹²

The odd or humorous nature of sentences like those in (4) and (5) provides evidence that two distinct senses are involved; that is, evidence for a real lexical ambiguity. Another widely used test for antagonism between two senses is the identity test.¹³ This test makes use of the fact that certain kinds of ellipsis require parallel interpretations for the deleted material and its antecedent. We will illustrate the test first with an instance of structural ambiguity:¹⁴

(6) a. The fish is ready to eat.
b. The fish is ready to eat, and so is the chicken.
c. The fish is ready to eat, but the chicken is not.
d. #The potatoes are ready to eat, but the children are not.

Sentence (6a) is structurally ambiguous: the fish can be interpreted as either the agent or the patient of eat. Both of the clauses in example (6b) are ambiguous in the same way. This predicts that there should be four logically possible interpretations of this sentence; but in fact only two are acceptable to most English speakers. If the fish is interpreted as an agent, then the chicken must be interpreted as an agent; if the fish is interpreted as a patient, then the chicken must be interpreted as a patient. The parallelism constraint rules out readings where the fish is the eater while the chicken is eaten, or vice versa. The same holds true for example (6c). Sentence (6d) is odd because the nouns used strongly favor different interpretations for the two clauses: the potatoes must be the patient, while the children must be the agent, violating the parallelism constraint.

Example (7) illustrates the use of the identity test with an apparent case of lexical ambiguity: duck can refer to an action (lowering the head or upper body) or to a water fowl. (In fact, this is a fairly obvious case of lexical ambiguity since the two uses have different parts of speech, which is not normally possible with vagueness or indeterminacy. Our purpose here is to validate the test, showing that it gives the expected results in the clear cases, and thus provides a reasonable source of evidence for deciding the less obvious cases.)

Sentence (7a) is ambiguous, because the two senses of duck generate two different readings, and one of these readings could be true while the other was false in a particular situation. The same potential ambiguity applies to both of the clauses in (7b), so again we would predict that four interpretations should be logically possible; but in fact only two are acceptable. Sentence (7b) can mean either that John and Bill both saw her perform a certain action or that they both saw a water fowl belonging to her. The fact that the parallelism constraint blocks the “crossed” readings provides evidence that these two different interpretations of duck are truly distinct senses, i.e. that duck is in fact lexically ambiguous.

(7) a. John saw her duck.
b. John saw her duck, and so did Bill.

Contrast this with the examples in (8). The word cousin in the first clause of (8a) refers to a male person, while the implicit reference to cousin in the second clause of (8a) refers to a female person. This difference of reference does not violate the parallelism constraint, because the two uses of cousin are not distinct senses, even though they would be translated by different words in a language like Italian. The identity test indicates that cousin is not lexically ambiguous, but merely unspecified for gender.

(8) a. John is my cousin, and so is Mary.
b. John carried a briefcase, and Bill a backpack.
c. That three-year old is quite tall, but then so is his father.

Similarly, the word carry in the first clause of (8b) probably describes a different action from the implicit reference to carry in the second clause. The sentence allows an interpretation under which John carried the briefcase by holding it at his side with one hand, while Bill carried the backpack on his back; in fact, this would be the most likely interpretation in most contexts. The fact that this interpretation is not blocked by the parallelism constraint indicates that carry is not lexically ambiguous, but merely unspecified (i.e., indeterminate) for manner. The two uses of carry would be translated by different words in a language like Tzeltal, but they are not distinct senses.

The actual height described by the word tall in the first clause of (8c) is presumably much less than the height described by the implicit reference to tall in the second clause. The fact that this interpretation is acceptable indicates that tall is not lexically ambiguous, but merely vague.

Example (9) shows how we might use the identity test to investigate the ambiguity of the Spanish word llave mentioned above. These sentences could appropriately be used if both Pedro and Juan bought, broke or found the same kind of thing, whether keys, faucets, or wrenches. But the sentences cannot naturally describe a situation where different objects are involved, e.g. if Pedro bought a key but Juan bought a wrench, etc.¹⁵ This fact provides evidence that llave is truly ambiguous and not merely indeterminate or vague.

(9) a. Pedro compró/rompió una llave y también Juan.

Pedro bought/broke a key/etc. and also Juan

‘Pedro bought/broke a key/faucet/wrench, and so did Juan.’

b. Pedro encontró una llave al igual que Juan.

Pedro found a key/etc. to.the same that Juan

‘Pedro found a key/faucet/wrench, just like Juan did.’

Another test which is sometimes used is the sense relations test: distinct senses will have different sets of synonyms, antonyms, etc. (see discussion of sense relations in Chapter 6). For example, the word light has two distinct senses; one is the opposite of heavy, the other is the opposite of dark. However, Cruse (1986: 56–57) warns that this test is not always reliable, because contextual features may restrict the range of possible synonyms or antonyms for a particular use of a word which is merely vague or indeterminate.

Another kind of evidence for lexical ambiguity is provided by the test of contradiction.¹⁶ If a sentence of the form X but not X can be true (i.e., not a contradiction), then expression X must be ambiguous. For example, the fact that the statement in (10) is not felt to be a contradiction provides good evidence for the claim that the two uses of child represented here (‘offspring’ vs. ‘preadolescent human’) are truly distinct senses.

(10) (Aged mother discussing her grown sons and daughters)
They are not children any more, but they are still my children.

This is an excellent test in some ways, because the essential property of ambiguity is that the two senses must have different truth conditions, and this test involves asserting one reading while simultaneously denying the other. In many cases, however, it can be difficult to find contexts in which such sentences sound truly natural. A few attempts at creating such examples are presented in (11). The fact that such sentences are even possible provides strong evidence that the relevant words have two distinct senses.

(11) a. Criminal mastermind planning to stage a traffic accident in order to cheat the insurance company: After the crash, you lie down behind the bus and tell the police you were thrown out of the bus through a window.
Unwilling accomplice: I’ll lie there, but I won’t lie.

b. Foreman: I told you to collect a sample of uranium ore from the pit and row it across the river to be tested.
Miner: I have the ore but I don’t have the oar.

c. Rancher (speaking on the telephone): I’ve lost my expensive fountain pen; I think I may have dropped it while we were inspecting the sheep. Can you check the sheep pen to see if it is there?
Hired hand: I am looking at the pen, but I don’t see a pen.

An equivalent way of describing this test is to say that if there exists some state of affairs or context in which a sentence can be both truly affirmed and truly denied, then the sentence must be ambiguous.¹⁷ An example showing how this test might be applied to two uses of the word drink (alcoholic beverage vs. any beverage) is quoted in (12)

(12) a. Ferrell has a drink each night before going to bed.
b. “Imagine… this state of affairs: Ferrell has a medical problem which requires that he consume no alcoholic beverages but that he have a glass of water each night before going to bed. One person knows only that he does not consume alcoholic beverages; another knows only that he has a glass of water each night at bedtime. The latter person can truly affirm the sentence in (12a)… But the former person can truly deny it.” (Gillon 1990: 407)

Gillon points out that this is a very useful test because “generality and indeterminacy do not permit a sentence to be both truly affirmed and truly denied” (1990: 410). Sentences like those in (13) can only be interpreted as contradictions; they require some kind of pragmatic inference in order to make sense.¹⁸

(13) a. # She is my cousin and she is not my cousin.
b. # I am carrying the bag and I am not carrying the bag.
c. # This creature is a vertebrate and it is not a vertebrate.

5.3.3 Polysemy vs. homonymy

Two types of lexical ambiguity are traditionally distinguished: polysemy (one word with multiple senses) vs. homonymy (different words that happen to sound the same). Both cases involve an ambiguous word form; the difference lies in how the information is organized in the speaker’s mental lexicon.

Of course, it is not easy to determine how information is stored in the mental lexicon. This is not something that native speakers are consciously aware of, so asking them directly whether two senses are “the same word” or not is generally not a reliable procedure. The basic criterion for making this distinction is that in cases of polysemy, the two senses are felt to be “related” in some way; there is “an intelligible connection of some sort” between the two senses.¹⁹ In cases of homonymy, the two senses are unrelated; that is, the semantic relationship between the two senses is similar to that between any two words selected at random.

It is difficult to draw a clear boundary between these two types of ambiguity, and some authors reject the distinction entirely. However, many ambiguous words clearly belong to one type or the other, and the distinction is a useful one. We will adopt a prototype approach, suggesting some properties that are prototypical of polysemy vs. homonymy while recognizing there will be cases which are very difficult to classify.

Some general guidelines for distinguishing polysemy vs. homonymy:

a. Two senses of a polysemous word generally share at least one salient feature or component of meaning, whereas this is not in general true for homonyms.²⁰ For example, the sense of foot that denotes a unit of length (‘12 inches’) shares with the body-part sense the same approximate size. The sense of foot that means ‘base’ (as in foot of a tree/mountain) shares with the body-part sense the same position or location relative to the object of which it is a part. These common features suggest that foot is polysemous. In contrast, the two senses of row (pull the oars vs. things arranged in a line) seem to have nothing in common, suggesting that row is homonymous.

b. If one sense seems to be a figurative extension of the other (see discussion of figurative senses below), the word is probably polysemous. For example, the sense of run in This road runs from Rangoon to Mandalay is arguably based on a metonymy between the act of running and the path traversed by the runner, suggesting that this is a case of polysemy.

c. Beekman & Callow (1974) suggest that, for polysemous words, one sense can often be identified as the primary sense, with other senses being classified as secondary or figurative. The primary sense will typically be the one most likely to be chosen if you ask a native speaker to illustrate how the word X is used in a sentence, or if you ask a bilingual speaker what the word X means (i.e., ask for a translation equivalent). For homonymous words, neither sense is likely to be “primary” in this way.²¹

d. Etymology (historical source) is used as a criterion in most dictionaries, but it is not a reliable basis for synchronic linguistic analysis. (Speakers may or may not know where certain words come from historically, and their ideas about such questions are often mistaken.) However, there is often a correlation between etymology and the criteria listed above, because figurative extension is a common factor in semantic change over time, as discussed in §5.4. English spelling may give a clue about etymology, but again is not directly relevant to synchronic linguistic analysis, which normally focuses on spoken language.

Point (d) is a specific application of a more general principle in the study of lexical meaning: word meanings may change over time, and the historical meaning of a word may be quite different from its modern meaning. It is important to base our analysis of the current meanings of words on synchronic (i.e., contemporaneous) evidence, unless we are specifically studying the diachronic (historical) developments. Lyons (1977: 244) expresses this principle as follows:

A particular manifestation of the failure to respect the distinction of the diachronic and the synchronic in semantics … is what might be called the etymological fallacy: the common belief that the meaning of words can be determined by investigating their origins. The etymology of a lexeme is, in principle, synchronically irrelevant.

As an example, Lyons points out that it would be silly to claim that the “real” meaning of the word curious in Modern English is ‘careful’, even though that was the meaning of the Latin word from which it is derived.

A number of authors have distinguished between regular or systematic polysemy vs. non-systematic polysemy. Systematic polysemy involves senses which are related in recurring or predictable ways. For example, many verbs naming a change of state (break, melt, split, etc.) have two senses, one transitive (V_tr) and the other intransitive (V_intr), with V_tr meaning roughly ‘cause to V_intr’. Similarly, many nouns that refer to things used as instruments (hammer, saw, paddle, whip, brush, comb, rake, shovel, plow, sandpaper, anchor, tape, chain, telephone, etc.) can also be used as verbs meaning roughly ‘to use the instrument to act on an appropriate object.‘ (A single sense can have only a single part of speech, so the verbal and nominal uses of such words must represent distinct senses.)

The kinds of regularities involved in systematic polysemy are similar to patterns which are associated with derivational morphology in some languages.²² This means that the systematic relationships between senses can be stated in the form of rules. Some authors have suggested that only the base or core meaning needs to be included in the lexicon, because the secondary senses can be derived by rule.²³ But even in the case of systematic polysemy, secondary senses need to be listed because not every extended sense which the rules would license actually occurs in the language. For example, there are no verbal uses for some instrumental nouns, e.g. scalpel, yardstick, hatchet, pliers, tweezers, etc. For others, verbal uses are possible only for non-standard uses of the instrument or non-literal senses:

(14) a. Australian Prime Minister Kevin Rudd has axed the carbon tax.
b. Alaska Airlines axed the flights as a precaution.
c. ?*John axed the tree.

Traditionally it has been assumed that all the senses of a polysemous word will be listed within a single lexical entry, while homonyms will occur in separate lexical entries. Most dictionaries adopt a format that reflects this organization of the lexicon. The format is illustrated in the partial dictionary listing for the word form lean presented in (15).²⁴ The verbal and adjectival uses of lean are treated as homonyms, each with its own lexical entry. Each of the homonyms is analyzed as being polysemous, with the various senses listed inside the appropriate entry.

(15) lean₁ (V): 1. to incline, deviate, or bend from a vertical position; 2. to cast
one’s weight to one side for support; 3. to rely on for support or
inspiration; 4. to incline in opinion, taste, or desire (e.g., leaning toward a
career in chemistry).

lean₂ (Adj): 1. lacking or deficient in flesh; 2. containing little or no fat
(lean meat); 3. lacking richness, sufficiency, or productiveness (lean
profits, the lean years); 4. deficient in an essential or important quality or
ingredient, e.g. (a) of ore: containing little valuable mineral; (b) of fuel
mixtures: low in combustible component.

This is not the only way in which a lexicon could be organized, but we will not explore the various alternatives here. The crucial point is that polysemous senses are “related” while homonymous senses are not.

5.3.4 One sense at a time

When a lexically ambiguous word is used, the context normally makes it clear which of the senses is intended. As Cruse (1986: 53) points out, a speaker generally intends the hearer to be able to identify the single intended sense based on context:

[A] context normally also acts in such a way as to cause a single sense, from among those associated with any ambiguous word form, to become operative. When a sentence is uttered, it is rarely the utterer’s intention that it should be interpreted in two (or more) different ways simultaneously… This means that, for the vast majority of utterances, hearers are expected to identify specific intended senses for every ambiguous word form that they contain.

Cruse (1986: 54) cites the sentence in (16), which contains five lexically ambiguous words. (Note that the intended sense of burn in this sentence, ‘a small stream’, is characteristic of Scottish English.)

(16) Several rare ferns grow on the steep banks of the burn where it runs into the lake.

Cruse writes,

In such cases, there will occur a kind of mutual negotiation between the various options [so as to determine which sense for each word produces a coherent meaning for the sentence as a whole]… It is highly unlikely that any reader of this sentence will interpret rare in the sense of ‘undercooked’ (as in rare steak), or steep in the sense of ‘unjustifiably high’ (as in steep charges)… or run in the sense of ‘progress by advancing each foot alternately never having both feet on the ground simultaneously’, etc.

A very interesting use of this principle occurs in the short story “Xingu”, by Edith Wharton (1916). In the following passage, Mrs. Roby is describing something to the members of her ladies’ club, which they believe (and which she allows them to believe) to be a deep, philosophical book. After the discussion is over, however, the other members discover that she was actually describing a river in Brazil. The words which are italicized below are ambiguous; all of them must be interpreted with one sense in a discussion of a philosophical work, but another sense in a discussion of a river.

(17) “Of course,” Mrs. Roby admitted, “the difficulty is that one must give up so much time to it. It’s very long.”
“I can’t imagine,” said Miss Van Vluyck tartly, “grudging the time given to such a subject.”
“And deep in places,” Mrs. Roby pursued; (so then it was a book!) “And it isn’t easy to skip.”
“I never skip,” said Mrs. Plinth dogmatically.
“Ah, it’s dangerous to, in Xingu. Even at the start there are places where one can’t. One must just wade through.”

“I should hardly call it wading ,” said Mrs. Ballinger sarcastically.
Mrs. Roby sent her a look of interest. “Ah — you always found it went swimmingly?”
Mrs. Ballinger hesitated. “Of course there are difficult passages,” she conceded modestly.
“Yes; some are not at all clear — even,” Mrs. Roby added, “if one is familiar with the original.²⁵”
“As I suppose you are?” Osric Dane interposed, suddenly fixing her with a look of challenge.
Mrs. Roby met it by a deprecating smile. “Oh, it’s really not difficult up to a certain point; though some of the branches are very little known, and it’s almost impossible to get at the source.”

Mrs. Roby’s motives seem to be noble — she is rescuing the ladies of the club from further humiliation by an arrogant visiting celebrity, Mrs. Osric Dane (a popular author). But when the other members discover the deception, they are so provoked that they demand Mrs. Roby’s resignation.

Cotterell & Turner (1989: 175) point out the implications of the “one sense at a time” principle for exegetical work:

The context of the utterance usually singles out … the one sense, which is intended, from amongst the various senses of which the word is potentially capable… When an interpreter tells us his author could be using such-andsuch a word with sense a, or he could be using it with sense b, and then sits on the fence claiming perhaps the author means both, we should not too easily be discouraged from the suspicion that the interpreter is simply fudging the exegesis.

Sometimes, of course, the speaker does intend both senses to be available to the hearer; but this is normally intended as some kind of play on words, e.g. a pun. The humor in a pun (for those people who enjoy them) lies precisely in the fact that this is not the way language is normally used.

5.3.5 Disambiguation in context

Word meanings are clarified or restricted by their context of use in several different ways. If a word is indeterminate with respect to a certain feature, the feature can be specified by linguistic or pragmatic context. For example, the word nurse is indeterminate with respect to gender; but if I say The nurse who checked my blood pressure was pregnant, the context makes it clear that the nurse I am referring to is female.

We noted in the preceding section that the context of use generally makes it clear which sense of a lexically ambiguous word is intended. This is not to say that misunderstandings never arise, but in a large majority of cases hearers filter out unintended senses automatically and unconsciously. It is important to recognize that knowledge about the world plays an important role in making this disambiguation possible. For example, a slogan on the package of Wasa crispbread proudly announces, Baked since 1919. There is a potential ambiguity in the aspect of the past participle here. It is our knowledge about the world (and specifically about how long breads and crackers can safely be left in the oven), rather than any feature of the linguistic context, which enables us to correctly select the habitual, rather than the durative, reading. The process is automatic; most people who see the slogan are probably not even aware of the ambiguity.

Because knowledge about the world plays such an important role, disambiguation will be more difficult with translated material, or in other situations where the content is culturally unfamiliar to the reader/hearer. But in most monocultural settings, Ravin & Leacock’s (2000) assessment seems fair:

Polysemy is rarely a problem for communication among people. We are so adept at using contextual cues that we select the appropriate senses of words effortlessly and unconsciously… Although rarely a problem in language use, except as a source of humour and puns, polysemy poses a problem for semantic theory and in semantic applications, such as translation or lexicography.

If lexical ambiguity is not (usually) a problem for human speakers, it is a significant problem for computers. Much of the recent work on polysemy has been carried out within the field of computational linguistics. Because computational work typically deals with written language, more attention has been paid to homographs (words which are spelled the same) than to homophones (words which are pronounced the same), in contrast to traditional linguistics which has been more concerned with spoken language. Because of English spelling inconsistencies, the two cases do not always coincide; Ravin & Leacock cite the example of bass [bæs] ‘fish species’ vs. bass [bej s] ‘voice or instrument with lowest range’, homographs which are not homophones.

As Ravin & Leacock note, lexical ambiguity poses a problem for translation. The problem arises because distinct senses of a given word-form are unlikely to have the same translation equivalent in another language. Lexical ambiguity can cause problems for translation in at least two ways: either the wrong sense may be chosen for a word which is ambiguous in the source language, or the nearest translation equivalent for some word in the source language may be ambiguous in the target language. In the latter case, the translated version may be ambiguous in a way that the original version was not.

A striking example of the former type occurred in the English text of a bilingual menu in a Chinese restaurant, which offered ‘deep-fried enema’ rather than ‘deep-fried sausage’. The Chinese name of the dish is zhá guànchang (炸灌腸). The last two characters in the name refer to a kind of sausage made of wheat flour stuffed into hog casings; but they also have another sense, namely ‘enema’. The translator (whether human or machine) chose the wrong sense for this context.²⁶

Much medieval and renaissance art, most famously the sculptural masterpiece by Michelangelo, depicts Moses with horns coming out of his forehead. This practice was based on the Latin Vulgate translation of a passage in Exodus which describes Moses’ appearance when he came down from Mt. Sinai.²⁷ The Hebrew text uses the verb qaran to describe his face. This verb is derived from the noun qeren meaning ‘horn’, and in some contexts it can mean ‘having horns’;²⁸ but most translators, both ancient and modern, have agreed that in this context it has another sense, namely ‘shining, radiant’ or ‘emitting rays’. St. Jerome, however, translated qaran with the Latin adjective cornuta ‘horned’.²⁹

As noted above, a translation equivalent which is ambiguous in the target language can create ambiguity in the translated version that is not present in the original. For example, the French word apprivoiser ‘to tame’ plays a major role in the book Le Petit Prince ‘The Little Prince’ by Antoine de Saint-Exupéry. In most (if not all) Portuguese versions this word is translated as cativar, which can mean ‘tame’ but can also mean ‘catch’, ‘capture’, ‘enslave’, ‘captivate’, ‘enthrall’, ‘charm’, etc. This means that the translation is potentially ambiguous in a way that the original is not. The first occurrence of the word is spoken by a fox, who explains to the little prince what the word means; so in that context the intended sense is clear. However, the word occurs frequently in the book, and many of the later occurrences might be difficult for readers to disambiguate on the basis of the immediate context alone.

It is not surprising that homonymy should pose a problem for translation, because homonymy is an accidental similarity of form; there is no reason to expect the two senses to be associated with a single form in another language. If we do happen to find a pair of homonyms in some other language which are good translation equivalents for a pair of English homonyms, we regard it as a remarkable coincidence. But even with polysemy, where the senses are related in some way, we cannot in general expect that the different senses can be translated using the same word in the target language. Beekman & Callow (1974: 103) state:

Whether multiple senses of a word arise from a shared [component] of meaning or from relations which associate the senses [i.e. figurative extensions—PK], the cluster of senses symbolized by a single word is always specific to the language under study.

Perhaps Beekman & Callow overstate the unlikelihood that a single word in the target language can carry some or all of the senses of a polysemous word in the source language. Since there is an intelligible relationship between polysemous senses, it is certainly possible for the same relationship to be found in more than one language; but often this turns out not to be the case, which is why polysemy can be a source of problems for translators.

---

¹ Lakoff (1970).

² We follow Kennedy (2011) in using the term indeterminacy; as he points out, some other authors have used the term generality instead. Gillon (1990) makes a distinction between the two terms, using generality for superordinate terms.

³ A number of authors (Ruth M. Kempson 1977, Lakoff 1970, Tuggy 1993) have used the term vagueness as a cover term which includes generality or indeterminacy as a sub-type.

⁴ The technical term is the sorites paradox, also known as the paradox of the heap, the fallacy of the beard, the continuum fallacy, etc.

⁵ http://www.omniglot.com/language/kinship/chinese.htm

⁶ www-01.sil.org/mexico/museo/3di-Carry.htm

⁷ Jonatan Cordova (p.c.) informs me that the word can also be used to mean ‘lock’ in wrestling.

⁸ Cruse (1986: 61).

⁹ Cruse (2000: 108).

¹⁰ Adapted from Cruse (1986: 61).

¹¹ Tim O’Brien, The Things They Carried, via grammar.about.com.

¹² http://dictionary.reference.com/browse/zeugma

¹³ Lakoff (1970); Zwicky & Sadock (1975).

¹⁴ Examples adapted from Kennedy (2011: 512).

¹⁵ Jonatan Cordova, Steve and Monica Parker (p.c.).

¹⁶ Quine (1960); Zwicky & Sadock (1975); Kennedy (2011)

¹⁷ Adapted from Gillon (1990: 407).

¹⁸ The word vertebrate is more “general”, in Gillon’s terms, than words like fish or dog. We will discuss this kind of sense relation in the next chapter.

¹⁹ Cruse (2000: 109).

²⁰ Beekman & Callow (1974) suggest that all the senses of a polysemous word will share at least one component of meaning, but this claim is certainly too strong.

²¹ A similar point is made by Fillmore & Atkins (2000: 100).

²² See Apresjan (1974), Aronoff & Fudeman (2011: ch. 5).

²³ For example, Pustejovsky (1995).

²⁴ Adapted from the Merriam-Webster Online Dictionary (http://www.merriam-webster.com/ dictionary/lean).

²⁵ Apparently a play upon an archaic sense of original meaning ‘source’ or ‘origin’.

²⁶ http://languagelog.ldc.upenn.edu/nll/?p=2236

²⁷ Exodus 34:29–35.

²⁸ Psalm 69:31.

²⁹ There is some disagreement as to whether St. Jerome simply made a mistake, or whether he viewed the reference to horns as a live metaphor and chose to preserve the image in his translation. The latter view seems more likely since he was very familiar with the rendering of the Septuagint, which uses the word ’glorified’. The first artistic depiction of a horned Moses appeared roughly 700 years after Jerome’s translation, which might be taken as an indication that the metaphorical sense was in fact understood by readers of the Vulgate at first, but was lost over time. (see Ruth Mellinkoff. 1970. The Horned Moses in Medieval Art and Thought (California Studies in the History of Art, 14). University of California Press.)