The Paradigmatic Rules

The minimal-contrast rule. If a stimulus has a common opposite (an antonym), it will always elicit that opposite more often than anything else. These responses are the most frequent found anywhere in word associations. [...]

The most compelling evidence for this rule comes from the so-called popular adjectives (long v. short, good v. bad, etc.).Deese found that the most frequent association to each of 80 such adjectives was its antonym. At the first stage, the feature list for long, for example, would end with [+polar]. The second-stage associating rule would change [+polar] to [–polar], and the third stage result would be short. Nouns, too, often show alterations of only one feature. Among animate nouns, the sign of the feature [±male] is reversed, giving male — female, man — woman, boy — girl, he — she, him — her, aunt — uncle, etc. (and vice versa) as most frequent responses. Antonymous prepositions, e. g. up — down, above — below, and to — from, strongly elicit each other with a change of feature [±polar] and so do verb ‘converses’, e. g. give — take, sell — buy, go — come, and so on. Other frequent single-feature contrasts include [±polar] in verbs (is — are, was — were, has — have, etc.), [±past] among strong verbs (is — was, are — were, has — had, take — took, etc.). [±nominative] among pronouns (he — him, she — her, they — them, etc.), and [±proximal] among the deictic words (here — there, this — that, now — then, etc.). Obviously, the minimal-contrast rule accounts for a large number of the commonest responses in word associations. [... ]

The marking rule. This rule, a particularization of the minimal-contrast rule, was suggested by some remarks of Greenberg. He pointed out that there was a greater tendency to change a feature from, rather than to, its markedvalue in word association data. [...] Consider the feature [±plural] for nouns. A plus signals the addition of the morpheme pl, usually -z; a minus signals the morpheme sg, usually zero. [...]

[+plural] is therefore the marked value, and [–plural] the unmarked. In word-association data, then, it should be commoner to find, say, dogs — dog, than dog — dogs, and it is. Comparative adjectives also elicit their positive forms (better — good) more often that the reverse (good — better), and past participle verbs their infinitive forms (brought — bring) more often than reverse (bring — brought). Marshall extended this rule to unmarked and marked adjectives (e. g. long and short, respectively). An examination of 16 pairs of adjective stimuli that have only the antonym generally supports this extension, with 14 of the 16pairs consistent with the rule. Also, if we take the accusative case to be unmarked with respect to the nominative case, the rule holds, with the stimulus response pairs I — me, he — him, she — her, they — them, etc., occurring more often than the reverse pairs. Again, if the suffix -less is marked with respect to -ful, the rule holds once more, as in pairs like careless — careful, thoughtless — thoughtful, and useless — useful. And there are other cases which confirm this.

On the other hand, man is generally considered to be unmarked with regard to woman, and he with regard to she, him with regard to her, and so on. Yet man elicits woman more consistently than woman does man, the same is true for he and she, her and him, and certain others. To save the marking rule, man would have to be shown to be marked, and woman unmarked. This, however, would go against the very foundations of marking found in Greenberg and elsewhere. The marking rule therefore cannot be retained as a general rule.

Some results which seem to agree with marking hypotheses can furthermore be explained in an alternative way by considering the surface ambiguity of the stimuli. The unmarked adjective deep, for example, could be assigned either of two senses at stage one: (1) in depth as in three feet deep, or (2) opposite to shallow as in The river is deep. But shallow has only one sense, opposite to deep. If the minimal-contrast rule is invoked at stage two, deep will at times produce words like high, far, etc., from sense (1) and at other times shallow fromsense (2); shallow on the other hand, will always produce deep. The consequence is that shallow should elicit deep more that the reverse, which agrees with the data. Thus the minimal-contrast rule, taking together with the surface ambiguity of unmarked adjectives, might account quite simply for the asymmetry in associations between unmarked and marked words.

The feature deletion and addition rules. There also appear to be rules that either delete features from, or add features to the end of the feature list. As Marshall points out, the deletion rule should have a precedence over theaddition rule, since there are many possible features that might be added, but those to be deleted are exactly specified. Deletion of features generally produces superordinates, like fruit from apple, while addition of features produces subordinates, like apple from fruit. Both superordinates and subordinates occur often in word associations, but subjects generally offer superordinates more quickly than subordinates. [...]

Another example of feature deletion is the dropping of [+cause] from such verbs as kill yielding die. Again, this feature is more often dropped than added, as in pairs like kill — die, teach — learn, feed — eat, show — see, etc. If we assume that for listen — hear, listen is identical to hear except for an additional [+volitive] and let the same is true for look — see, then the precedence of deletion over addition is again confirmed. The feature-deletion and addition rules, like the minimal-contrast rule, actually consist of hierarchy of rules, with single deletions and additions preferred to multiple operations.

Word associations often include near synonyms, like house — home, odour — smell, seem — appear, thing — object, etc. One sense of home appears to contain all the features of house plus some extras indicating that it is someone’s usual residence, but other synonyms differ in different ways. Although partial synonyms have not been characterized in any consistent form in terms of feature theory, it is clear that they usually have feature lists differing on only a few, possibly optional, features. The feature-deletion and addition rules, then, also produce synonyms. But the minimal-contrast rule has priority over these rules, for if the stimulus has a full antonym, it is always more frequently given as a response than is a partial synonym. [...]

The category-preservation rule. A long-standing observation in word-association literature is that stimuli tend to elicit paradigmatic responses. This is not too surprising given the previous rules, since the responses produced by those rules are always paradigmatic. But there are seeming exceptions. Common adjectives elicit other adjectives almost invariably, but uncommon adjectives do so less consistently. As Deese pointed out, this happens because the common adjectives mostly belong to antonym pairs and have minimal contrasts whereas the uncommon ones do not. The category-preservation rule is therefore a negative one: “Do not change features high on the list”, such as the feature [+noun] or [+adjective]. The rule is, in fact, only another aspect of the rates stressing that features at the bottom of the list should be altered first. [...]