Many set-denoting adjectives come in antonym pairs that can be placed on a single scale. Some examples are given in (60). The following subsections will show, however, that the scales implied by these antonym pairs can differ in various ways.

The scales in (61), (64), and (65) have a neutral zone to which neither adjective applies. This zone is often more or less fixed for the speaker in question. For some adjectives, however, the neutral zone is more flexible and can be determined by the entity of which the adjective is predicated or by the context in which the adjectives are used. This is especially true for measure adjectives of the kind in (73).

The examples in (74) show that the placement of the neutral zone, i.e. the interpretation of the measure adjectives, depends on the argument of which the adjective is predicated. We will consider only the examples with the adjective groot, but the discussion is also applicable to klein.

Although groot can be predicated of the two noun phrases deze muisthis mouse and deze walvisthis whale, it is clear that the two entities to which these noun phrases refer cannot be assumed to be of similar size. Thus, the placement of the neutral zone on the implied scales of size will be different. In the case of mice, the scale is expressed in terms of centimeters, as in (75a), whereas in the case of whales, the scale is instead expressed in terms of meters, as in (75b).

75

a.

Scale of size for mice in centimeters

b.

Scale of size for whales in meters

This shows that the position of the neutral zone is at least partly determined by the argument of which the adjective is predicated; it indicates the “normal” or “average” size of mice/whales. In other words, examples such as (74) tacitly introduce a comparison class, namely the class of mice/whales, which determines the precise position of the neutral zone on the implied scale. Often a voor-PP is introduced to make the comparison class explicit and to clarify the intended neutral zone, as in (76).

The comparison class, and hence the neutral zone, is not always determined by the argument of which the adjective is predicated; the context may also play a decisive role. In a discussion of mammals in general, statement (77a) will be true, while the statement in (77b) will be false: since the comparison class consists of mammals, the neutral zone will be determined by the average size of mammals, and Indian elephants are definitely larger than that. However, if we discuss the different subspecies of elephants, statement (77b) can be considered true: the comparison class is formed by elephants, and the Indian elephant is indeed small compared to the African elephant. We can therefore conclude that the truth of (77b) depends on the comparison class evoked by the context.

Although the placement of the neutral zone on the scale implicit in the measure adjective depends on extra-linguistic information, the scale itself is objective once speakers have established the neutral zone, so that they can objectively determine whether a given statement is true or false. The fact that the scale implied by the measure adjectives is objective is also supported by the fact that (in some cases) the exact position on the scale can be indicated by nominal measure phrases like twee dagen and twintig meter in (78).

On the other hand, with adjectives like lelijk/mooiugly/beautiful and saai/boeiendboring/exciting, the exact position of the relevant entities on the implied scale is a more subjective matter. It may well depend entirely on the language user, which can be brought out by embedding the adjective under the verb vindenconsider, as in the (a)-examples in (79). Occasionally, the entity whose evaluation is assumed can be expressed syntactically by a voor-PP; this is illustrated in the (b)-examples.

The pairs of measure adjectives in (73) can be regarded as true antonyms: the two (a)-examples in (80) are fully equivalent. However, this equivalence does not seem to hold for the subjective adjectives in the (b)-examples, which suggests that the comparative forms mooiermore beautiful and lelijkeruglier are not true antonyms but quasi-antonyms.

This difference may be related to the following observation. The use of the comparative form of objective adjectives like kleinsmall and grootbig in the (a)-examples of (80) does not necessarily imply that the argument of which the adjective is predicated is in fact small or big. The use of the comparative form of the subjective adjectives mooibeautiful and lelijkugly, on the other hand, suggests that the argument of which the adjective is predicated is in fact beautiful or ugly. Consequently, the true antonym of mooier is the comparative form minder mooiless beautiful, as is clear from the equivalence between (81a) and (81b).

Of course, the true antonymy relationship also holds between groterbigger and minder grootless big. The crucial point, however, is that minder groot is virtually synonymous with kleinersmaller, while minder mooiless beautiful and lelijkeruglier have distinctly different meanings. The difference between objective and subjective adjectives discussed in this subsection may be lexically encoded; reasons for assuming this will be given in Subsection F below.

The examples in (78) have already shown that the measure adjectives can be modified by nominal measure phrases; cf. Klooster (1972) and Corver (1990:ch.8). For each pair of antonyms in (73), however, this holds only for the adjective in the primeless example; some illustrations are given in (82). Of course, the use of two days old in (82a) does not express the fact that the kitten is old; it is in fact quite young, which can be emphasized by using the evaluative particle pasonly. Therefore, it is clear that the adjective oud has lost the antonymous part of its meaning; the same is true for the adjective lang in (82b). Since oud and lang do not need to express the antonymous part of their meaning, they can be seen as the neutral forms of the respective pairs; the adjectives jong and kort cannot be used in this neutral way, hence the use of the percentage sign.

Similar conclusions can be drawn from the interrogative sentences in (83): the neutral form oud/lang gives rise to a perfectly natural question and does not imply that the subject of the clause should be characterized as old/long, whereas the non-neutral form jong/kort gives rise to a marked result and, to the extent that it is usable at all, seems to imply that the kitten is young and the road is short.

The examples in (84) show that the two forms also seem to differ when it comes to the derivation of nouns from adjectives by suffixation with –te; the neutral forms of the measure adjectives seem to be preferred as input for this morphological rule, although it must be pointed out that the formations marked with a number sign can also be found on the internet, but with a much lower frequency. Note that the formation of *oudte in (84c) is probably blocked by the existing noun leeftijdage. We refer the reader to Section 25.1.2, sub II, for further discussion of measure adjectives.

The adjectives in (85) seem to be similar to the measure adjectives in (83): the primeless examples are unmarked and do not presuppose that the property denoted by the adjective is applicable; the primed examples, on the other hand, are marked and strongly suggest that the property denoted by the adjective is applicable.

Often the subjective adjectives not only imply a subjective scale, but also express a negative or positive evaluation. In the pair slechtbad/evil and goedgood, the first adjective denotes a negatively valued property, while the second denotes a positively valued property. The examples in (86) show that this distinction is also reflected in their modification possibilities: the primeless examples contain negatively valued adjectives, and modification by elements like knapquite and flinkquite is possible; the primed examples, on the other hand, contain positively valued adjectives, and modification by knap and flink is excluded.

The examples in (87) show that litotes (the trope in literary and formal language by which one emphasizes a property by negating its antonym) prefers an adjective denoting a negatively valued property; when the adjective denotes a positively valued property, as in the primed examples, the desired interpretation usually does not arise; a notable exception is the colloquial expression Da’s niet goed!, in which the deictic power of the demonstrative has faded; cf. English That’s not good! (Carole Boster, p.c.).

The modifier welrather in (88), on the other hand, requires an adjective denoting a positively valued property; cf. van Riemsdijk (2005). Note that the primed examples are fully acceptable if wel is interpreted as the affirmative marker wel; the two forms differ in that the affirmative marker is accented while the modifier is not. The uses of niet and wel in (87) and (88) are discussed in more detail in Section 25.3, sub II.

Occasionally, the modifier is sensitive to both the positive/negative value of the adjective and the syntactic environment. In declarative clauses, for example, the modifier een beetjea bit may prefer a negatively valued adjective (or an adjective that has no antonym, such as verliefdin love). In questions and imperatives, on the other hand, this modifier may prefer an adjective that denotes a positively valued property. The use of the percentage sign indicates that judgments on the degree of markedness of these examples vary from speaker to speaker.

This subsection discusses the fact that negative polarity items can be licensed by the subset of antonymous adjectives that were called true antonyms in subsection C above. To do so, we will first discuss some logical properties of these adjectives. Subsection C has already shown that true antonyms have the defining property that they allow the inference in (90a), where A and A' are antonymous adjectives. Consider again example (80a), repeated here as (90b): if Jan is taller than Marie, we can conclude that Marie is shorter than Jan, and similarly, if Marie is shorter than Jan, we can conclude that Jan is taller than Marie. This equivalence does not hold for quasi-antonymous adjectives such as mooibeautiful and lelijkugly; cf. the earlier discussion of example (80b).

True and quasi-antonymous adjectives are similar in that the implications in (91a) do not hold for either of them; in both cases the implied scale may have a neutral zone. This has already been discussed for the quasi-antonymous adjectives slechtbad and goedgood in Subsection A, so we limit ourselves here to giving similar examples for the true antonymous adjectives grootbig and kleinsmall.

Despite the fact that the implications in (91a) do not hold, we will show in this subsection that for the true antonyms above, the pair not A and A' as well as the pair not A' and A show certain similarities in semantic behavior that may be relevant when it comes to the licensing of negative polarity items like ook maar ietsanything. To demonstrate this, we have chosen the adjectives in (92) because they can take a clausal complement, which is crucial for our purposes below, since negative polarity items like ook maar iets are normally only possible in embedded clauses; cf. Section V13.3.1, sub III.

Consider the examples in (93). In the primeless examples, the complement clause refers to a broader set of eventualities than does the complement in the primed examples; the addition of an adverb in the latter cases makes the state of affairs to which the complement clause refers more specific, and thus applicable to a smaller number of situations. For example, there are many cases in which a problem is solved, but only in a subset of those cases is the problem solved quickly.

Now, it is important to note that one cannot conclude from the truth of the primeless examples that the primed examples are also true. However, one could conclude from the truth of the primed examples that the primeless ones are also true. The environments in (93), in which an expression like snel oplossento solve quickly can be replaced by a more general one like oplossento solve without changing the truth value of the expression, are known as upward entailing.

The inferences change radically if we replace the adjectives in (93) by their antonyms, as in (94). Now we can conclude from the truth of the primeless examples that the primed examples are also true, and not vice versa. The environments in (94), where an expression like oplossento solve can be replaced by a more specific one like snel oplossento solve quickly without changing the truth value of the expression, are called downward entailing.

From the examples in (93) and (94) we can conclude that the adjectives in the primeless examples of (92) create upward entailing environments, whereas the adjectives in the primed examples of (92) create downward entailing environments. Negation can reverse this property. If we add the adverb nietnot to the examples in (93), the environments become downward entailing, and if we add niet to the examples in (94), the environments become upward entailing. For example, niet gemakkelijknot easy in (95a&a') behaves like moeilijkdifficult in (94a&a'), and niet moeilijknot difficult in (95b&b') behaves like gemakkelijkeasy in (93a&a') in this respect.