Direct Criteria for Ambiguity
Three different types of criteria for ambiguity will be proposed. It may ultimately be possible to show that they all reduce to a single basic criterion, but here they will be presented separately. Generally speaking, unless there are specific reasons why one or other of the criteria should be inapplicable (some of these reasons will be discussed below, we shall expect an ambiguous item to satisfy all the criteria.
The first criterion is frequently difficult to apply in practice, but it is conceptually important. It is that the senses of an ambiguous word form should not in every case be totally conditioned by their contexts, unlike the interpretations which arise as a result of contextual modulation. This means that an ambiguous word form set in a disambiguating context may well carry more information than can be accounted for in terms of interaction between the context-independent meaning of the word form, and the semantic properties of the context. In cases of contextual modulations, on the other hand, all information is derived from these sources.
Consider sentences 16 and 17:
16. Arthur washed and polished the car.
17. John lubricated the car.
The most likely interpretation of 16 is that not every part of the car underwent washing and polishing, but the exterior surface only. What is the basis for this conclusion? It is derived entirely from the general meaning of car, together with the semantic properties of the context (remember that general knowledge concerning cars and operations carried out on them is, on the view of meaning adopted in this book, embedded in the meaning of car, wash, polish, etc). A similar account can be given of the most likely interpretation of car in 17. Or take the case of monarch in 14 (repeated here for convenience):
14. The Ruritanian monarch is expecting her second baby.
We can be virtually certain that the monarch in question is a queen, because of the restricting effect of the context on the general meaning of monarch. Notice that a similar interpretation would arise, and no loss of information would result, if monarch were replaced by a synonym or paraphrase such as sovereign, or crowned head (and automobile would interact in the same way with the context if it were substituted for car in 16 and 17). Contrast these, however, with bank in 18 and 19:
18. Her husband is the manager of a local bank.
19. At this point, the bank was covered with brambles.
Let us try to account for the (most probable) different interpretations of bank in the way that we did for car. It is first necessary to decide on a synonym or paraphrase of the context-invariant meaning of bank. This already poses problems, but let us say, for the sake of argument, that it is equivalent to place. We can then observe the effect of substituting place for bank in 18 and 19:
20. Her husband is the manager of a local place.
21. At this point, the place was covered with the brambles.
There is quite clearly a loss of information, so we have failed to show that the interpretations of bank are the result of contextual modulation of a general meaning. It may be concluded, therefore, that the different contexts are selecting discrete senses of bank. Another instance of incomplete contextual determination is to be observed with dog. Let us for the moment take it as established that dog has a general sense, denoting the whole species, irrespective of sex. In sentences such as 22, however, dog has a more specific meaning, and refer only to males:
22. John prefers bitches to dogs
Now it might be argued that the resultant sense of dog here is caused by contextual modulation of the general sense: dog cannot in this context refer to females if logical consistency is to be preserved, which leaves only males as possible referents. Consider now, however,23:
23. Incredibly, John prefers an aged, half-blind bitch to a dog, as his canine companion.
If the interpretation of dog in this sentence were the result of contextual modulation of the general sense, it ought to include reference to, for instance, young females with good eyesight. But once again, it refers to male dogs only. This reading cannot be explained by contextual modulation, so it must be the result of selection from a set of discrete possibilities. In fact, the same is true of 22. That contextual modulation of the general sense of dog cannot explain the specific interpretation in 22 is shown by the lack of a parallel specific interpretation of canine (in its jocular use as a noun ) when it is substituted for dog:
24. ? John prefers bitches to canines.
(We shall consider below why 24 should be anomalous)
Some understanding of the way the semantic effects of selection may be independent of, and indeed may transcend, those properties of the context which are responsible for the selection can be gleaned from the following analogy. Suppose that is known that a certain event is to occur on a certain day, but may take place at only one of two possible times, namely, 12.00 n00n or 12.00 midnight. If one were subsequently to receive a report that when the event occurred, the sun had set, one would be able to infer that it had taken place at exactly 12.00 midnight. The precision of this inference goes well beyond what is explicitly present in the report, which acts rather like a trigger setting off one of two preexisting possibilities. In a similar manner, the context of dog in 22 and 23 acts like a trigger which activates one of a set of pre-existing bundles of semantic properties, each having a precision and richness not directly sanctioned by the context. In principle all ambiguous items should be capable of manifesting these characteristics.
Our second criterion for ambiguity is that separate senses should be independently maximisable. Under certain conditions, the application of certain terms must be maximised within the current universe of discourse, even at the expense of oddness. Consider 25 (which resembles 24):
25: ? Mary likes mares better than horses.
One might have thought that the context makes it clear that the context makes it clear that horse in to be interpreted as “stallions”; however, such an interpretation is not available for this type of sentence. The reason is that since mares have been mentioned, they fall within the current universe of discourse, and by the rule of maximization (the details of which are not entirely clear) must be included in the reference of horses. This, of course, leads to logical inconsistency, and hence oddness. (Notice, however, that there is no anomaly if the reference of horses is EXPLICITLY restricted: Mary prefers mares to horses which can sire foals or Mary prefers mares to these horses uttered in a situation where only stallions are present.) On the other hand; 26, unlike 25, is perfectly normal:
26. John prefers bitches to dogs.
The general sense of dog would of course give rise to anomaly in 26, because of the rule of maximisation. The reason 26 is not odd is that dog has another sense, which even when maximised excludes bitches, and this is automatically selected by the context. By contrast, 27 selects the general reading of dog (the specific reading would be odd here, but not for reasons connected with maximisation):
27. Arthur breeds dogs.
Thus 26 and 27 taken together constitute strong evidence that dog is ambiguous.
The existence of two independent senses of dog, each independently maximisable, is responsible for the fact that A’s question in 28, if the dog in question is female, can be truthfully answered either ‘Yes’ or ‘No’ (depending on which sense the respondent believes the questioner to be intending):
28. A: Is that a dog?
B: (i) Yes, it’s a spaniel.
(ii) No, it’s a bitch.
There is no parallel set of circumstances in which the question in 29 can be truthfully answered ‘Yes’ or ‘No’:
29: A: Is the subject of this poem a monarch?
B: (i) Yes, it is a queen.
(ii) ? No, it is a king.
Because there is only one sense of monarch, namely, the general one, and because its reference must be maximised, if the subject of the poem was a king or a queen, then ‘Yes’ is the only truthful answer. As with 28, situations can be imagined in which the questions in 30 and 31 can be truthfully answered either negatively or positively:
30. A: Has Charles changed his position?
B: (i) Yes, he’s now sitting next to the chairman.
(ii) No, he still supports corporal punishment.
31. A: Did Arthur make it to the bank?
B: (i) Yes, he’s a strong swimmer.
(ii) No, he was arrested as soon as he came out of the water.
The same should be true, in principle, of any truly ambiguous expression.12
Ambiguity tests of the third kind utilities the fact that independent senses of a lexical form are antagonistic to one another; that is to say, they cannot be brought into play simultaneously without oddness. Contexts which do activate more than one sense at a time give rise to the variety of oddness we have labeled zeugma:
32.? John and his driving license expired last Thursday.
The simultaneous bringing into play of two senses can be effected either by coordination, as in 32, where John and his driving license select different senses of the verb expire, or by anaphora, as in 33:
33.? John’s and his driving license expired last Thursday; so did John.
So did is an anaphoric verb phrase; that is to say, its referential properties operate not directly, but indirectly, through a previously mentioned verb phrase, in this case expired last Thursday, which must be re-applied, this time with John as subject. But since this demands a different sense from the one appropriate to its first occurrence, the result is zeugma.
A general term cannot give rise to zeugma in this way:
34. My cousin, who is pregnant, was born on the same day as Arthur’s, who is the father.
Arthur’s refers anaphorically through cousin. The context makes it clear that the two cousins are of different sexes; however, the sentence is not zeugmatic, so we may conclude that cousin does not have two senses “make cousin” and “ female cousin”.
Antagonism of senses also lies behind the so-called identity test for ambiguity. 13 In 35, each part of the sentence contains an occurrence, either direct, or indirect via anaphora, of the ambiguous adjective light, and can therefore in theory be interpreted in two ways:
35. Mary is wearing a light coat; so is Sue.
However, the whole sentence does not have four (i.e.2 X2) interpretations, but two only. This is because the same reading of light must be selected in each part: either both ladies are wearing “undark” coats, or both are wearing ‘unheavy” coats. What is termed the crossed interpretation, with each part of the sentence manifesting a different sense, is prohibited. This prohibition is not a mysterious property of the grammatical process of anaphora; it is simply a consequence of the fact that light resists, so it were, the simultaneous activation of more than one of its senses. General terms allow crossed interpretations:
36. Mary has adopted a child; so has Sue.
There are four possible distributions of sexes compatible with this sentence, since there is no requirement that the two children should be of the same sex.
Some difficult cases
In this section the operation of ambiguity tests will be illustrated by applying them to a selection of difficult cases. The difficulties mostly concern tests based on the antagonism of sister-senses (i.e. senses associated with a single lexical form). It is not possible simply to dispense with such tests, because there are occasions, especially when dealing with highly context-bound readings which do not appear in ambiguous sentences, when they are the only practicable way of diagnosing ambiguity.
I. The first example involves the unit-type ambiguity. This is quite easy to demonstrate by means of the Yes/No-test:
37. A: Is this the jacket you want?
B: (i) Yes. (it’s the type I want)
(ii) No. (this particular one is shop-soiled)
But it is much more difficult to show antagonism; many contexts which might be expected to manifest it do not:
38. this is our best-selling jacket: do try it on.
Jacket in the first clause clearly must have a type reading-one cannot repeatedly sell the same individual jacket. One might have thought that only a particular unit of the type could be ‘tried on’, but that seems not to be the case. One must beware of drawing hasty conclusions in this area. As it happens, it is possible to find contexts which isolate the two readings, and when these are yoked together, zeugma results. Sentence 39 allows only the ‘unit’ reading for skirt 39 (this seems to be a property of belong):
39. That skirt belongs to Mary.
Sentence 40 can only bear a type reading:
40. My sister has the skirt Sue is wearing now.
Try to link these two readings together anaphorically, and the antagonism becomes plain:
41. ? the skirt sue is wearing belongs to Mary; my sister has it, too.
II. It not infrequently happens that ambiguous readings are related in such a way that in certain contexts one reading entails the other. Such cases are a common cause of apparent failure of the zeugma-test 9often called the ‘pun-test’) or the identity test. The two readings of dog are a case in point.14In 42, for example, it appears that a crossed interpretation is possible, in that Mary’s dog could well be male, and Bill’s female:
42. Mary bought a dog; so did Bill.
Does this contradict the evidence presented above that dog is ambiguous? The answer is that it does not. When dog occurs in a sentential context in which the specific interpretation entails the general interpretation, we cannot be sure which sense is operative when reference is made to a male dog: the two senses under these circumstances are effectively inseparable. Hence the normality of 42 when the dogs referred to are of opposite sexes cannot be used as evidence against the existence of two senses of dog, since it can be fully accounted for by claiming that only the general sense is operative. However, the situation is much clearer when dog occurs in a context where neither sense entails the other, as in 43:
43. Arthur wants to know if that is a dog; so does Mike.
A moment’s thought will convince the reader that the crossed reading is prohibited here; this sentence cannot be used to describe a situation where Arthur knows that the animal in question is an Alsatian, but in unsure of its sex, while Mike knows that it is female, but thinks it might be a wolf. The pun-test, too, demands non-entailing contexts:
44a. Dogs can become pregnant at 12 months. (General sense only)
b. Dogs mature later than bitches. (specific sense only).
c. ? Dogs can become pregnant at 12 months, but mature later than bitches.
III. Entailment between readings also bedevils attempts to demonstrate antagonism between the “exactly” and “at least” interpretations of numerals and other expressions of quantity.15The Yes/No-test suggests that this is a genuine ambiguity:
45. A: Have you got £10 in your wallet?
B: (i) Yes. In fact, I’ve got £12.
(ii) No. I’ve got £12
However, John has (exactly) £10 entails John has (at least) £10, which perhaps explains why 46 is not zeugmatic:
46. You need £100 in your account to qualify for free banking.
Arthur has it, now that he has added £50 to the £50 that was already there.
The first mention of £100 clearly demands an “at least” interpretation what Arthur has is “exactly” £100; one might therefore not expect the it of the second sentence to be able to refer anaphorically to £100 in the first sentence without antagonism. However, because of the entailment referred to above, the original and anaphoric occurrences of £100 can both be given the “at least”interpretation, thus avoiding antagonism. It is possible to construct isolating contexts which reveal antagonism, but they are extremely cumbersome:
47a. John, with £11, and Bill, with £12, both have the £10 necessary to open a savings account. (“at least”)
b. Tom, too, now has £10, having spent £2 out of his original £12. (“exactly” reading forced by now) necessary to open a savings account. (“at least”)
c. ? John, with £11, has the £10 necessary to open a savings account, Tom, too, now has it, having spent £2 out of his original £12.
IV. The case of door is interesting (a group of related words such as window, hatch, sky-light, etc. behave similarly). Two senses of door may be observed in 48, which can be truthfully answered either ‘Yes’ or ‘No’ in the following situation: the door in question has a ‘cat-flap’, and is standing open: the cat goes through the cat-flap, but not through the doorway:
48. Did the cat go through the door?
Once again, difficulties arise with the antagonism criteria. It might be predicted, for instance, that 49 would be zeugmatic, since what is smashed (the door-panel) is different from what is bricked up (the doorway):16
49. The door was smashed in so often that it had to be bricked up.
But there is no anomaly of any kind. Again, it appears that contexts of a particular kind must be avoided if the test is to succeed. In this case it is the part-whole relationship which is to blame. For certain predicates, applicability to parts entails applicability to wholes corresponding to the parts. Thus, if I touch the table-leg, by doing so I necessarily touch the table; if the tea-pot handle is broken, so is the tea-pot, and so on. It seems likely that this entailment is interfering with antagonism in49- both events are interpreted as happening to the ‘global door’, of which the door-panel is a part. The remedy, as before, is to avoid such contexts, and to use, to isolate the senses, only those contexts in which part does not entail whole (or, better still, contexts where part entails not-whole). When this done, the antagonism of the senses is easily seen:
50. ? We took the door off its hinges and then walked through it.
The moral to be drawn from these examples is that apparent compatibility of readings must not be hastily accepted as proof of generality: each case must be examined carefully to determine whether there are special factors preventing the appearance of zeugma. It may be reasonably confidently assumed that the different criteria for ambiguity which have been described in fact are sensitive to the same underlying semantic properties, and that in the absence of ‘special factors’ will provide identical diagnoses.
RINA LESTARI 