In reviewing evidence on this question, several kinds of ambiguity emerge. What I will call boundary ambiguity concerns how clearly entities are distinguished from their parts and their superordinate categories. Physical objects (such as apples or cars) are more distinct or bounded than activities (such as going to a movie or to a restaurant) or mental events (such as thinking or dreaming). Rips and Estin (1998) compared objects, scripted activities, and mental events in several ways. All of the comparisons suggested that objects are the most distinctive or bounded.
In their first study, participants generated parts of and superordinate categories for entities (specifically objects, activities, and mental events). Thus, for objects, an apple core is a part of an apple, and an apple is a kind of fruit. Similarly, for mental events, using logic is part of reasoning, and reasoning is a kind of thinking. For each entity (e.g., apple or reasoning), participants then judged whether the part (e.g., apple core or using logic, respectively) "is a kind of" the superordinate (e.g., fruit or thinking, respectively), and whether the entity (apple, reasoning) "is a part of" the superordinate (fruit, thinking). Although parts of objects were rarely seen as parts of their superordinate categories, parts of scripts and mental events often were. An apple core is not a kind of fruit, but using logic is a kind of thinking. That is, there was more homogeneity among the parts and wholes of scripts and mental events than objects. Furthermore, participants could name more distinctive properties for their objects than their scripts, and more distinctive properties for their scripts than their mental events. In addition, ratings of "bound-edness" were highest for objects and lowest for mental events, and this boundedness predicted the other results better than other ratings. Thus metonymy (i.e., naming parts and wholes with the same term) is more common for scripted activities and mental events than for objects. Such terms are polysemous in terms of their referents' boundaries. Hampson, John, and Goldberg (1986) made a similar point about the way trait terms are structured, without providing explicit comparisons with other kinds of concepts.
Another way to demonstrate words' ambiguity is to compare their meanings in contexts with their meanings in isolation. The meanings of adjectives (such as traits) usually change when they are combined with nouns. Murphy and Andrew (1993) asked participants to list the oppo-sites of 14 isolated adjectives and then list the opposites of 63 adjectives as used in adjective-noun phrases. Some phrases (e.g., cold water, bright light) were designed to elicit the same opposites as their adjectives alone, whereas others were not (e.g., cold facts, bright child). Indeed, participants' opposites of the former matched their opposites of the corresponding adjectives alone 66% of the time, whereas opposites of the latter matched them only 25% of the time. Some combinations appeared likely to elicit distinctly different meanings of the same adjectives (i.e., to represent homonyms; see note 1), whereas others (e.g., healthy appetite, healthy grip; fresh shirt, fresh water) appeared to be more polysemous. But the analyses do not distinguish between these two. In a second study, where adjective-noun phrases were selected at random from a large corpus of text, the mean match rate was only 51%. Two other studies asked for synonyms rather than antonyms, with similar results, including a mean match rate of only 44% for random adjective-noun phrases. Thus, adjectives' meanings show a high degree of what I will call combinatorial ambiguity.
Murphy and Andrew (1993) make a strong case that the meanings of many or most adjective-noun phrases are computed online rather than prestored. Traits are adjectives, and all 14 of the adjectives used in two of their studies could be used as traits. Indeed, illustrating the combinatorial ambiguity of trait adjectives, Kunda, Sinclair, and Griffin (1997) found that occupational stereotypes of actors changed the behavioral meanings of trait terms. For example, the combination of "aggressive" and "construction worker" implied physical aggression to their participants, whereas "aggressive lawyer" implied verbal aggression.
Many traits are also ambiguous in the sense that they can refer to, and be implied by, a wide range of very different behaviors. In this sense, creative is a relatively ambiguous trait and punctual is a relatively unambiguous trait. Such referential ambiguity is clearly documented, and its implications are developed in the research by Dunning and his colleagues described above.
Finally (but not exhaustively), behaviors show valence ambiguity in that they often imply traits of opposite valence. "Donald spent a great amount of his time in search of what he liked to call excitement. He had already climbed Mt. McKinley, shot the Colorado rapids in a kayak, driven in a demolition derby, and piloted a jet-powered boat—without knowing very much about boats. He had risked injury, and even death, a number of times" can imply that Donald is either adventurous (positive) or reckless (negative) (Higgins et al., 1977). This same classic study and many subsequent ones have described Donald's other behavior in ways that can be interpreted as independent or aloof, and as persistent or stubborn. Trait inferences from behavior may even require resolution of behaviors' valence ambiguity more often than not, because Cruse (1965) found that evaluations of English trait terms have a bimodal distribution, making neutral traits rare.
In short, behavior descriptions and mental action terms are more ambiguous than terms for objects in at least one way, and trait and behavior descriptions are ambiguous in other ways too. Yet, the meanings of all of these words are readily and adequately disambiguated in most daily discourse. There appears to be no "best" or "true" interpretation of trait terms independent of the (syntactic, semantic, and pragmatic) context in which they are used.
Trait terms can be used to describe act frequencies for individuals (Buss & Craik, 1981) or the goals of those acts (Read, Jones, & Miller, 1990). Borkenau (1990) provided evidence that traits represent ideals rather than central tendencies, "conveying information on the aptitude of persons for [attaining] several [of their own] goals" (p. 394) rather than their typical behaviors. Thus, people may be "helpful" in the sense that they just gave assistance, or frequently give assistance, or chronically want to give assistance, or are capable of giving assistance. Of course, particular forms of assistance depend on the circumstances and the person's other characteristics, making the multiple meanings of helpful innumerable if not infinite.
People prefer trait descriptions over goal descriptions when they are forming impressions or predicting future behavior, but goals are preferred when memorizing someone's acts or empathizing with them (Hoffman, Mischel, & Baer, 1984; Hoffman, Mischel, & Mazze, 1981). Traits can describe people or merely their behaviors (Todorov & Uleman, 2002), or even become incidentally associated with other people and objects (Brown & Bassili, 2001; Skowonski, Carlston, Mae, & Crawford, 1998), depending on the conditions under which traits are initially inferred from behaviors. Traits can provide interpretations of ambiguous acts, so that traits and act descriptions are assimilated, as when Donald's acts are interpreted as reckless after the concept of reckless has been activated in an apparently unrelated context (Higgins et al., 1977). Or traits can provide standards or ideals against which ambiguous acts are judged, so that traits and acts are contrasted, as when first explicitly describing reckless Erik makes Donald seem not so reckless but adventurous instead (Stapel & Koomen, 2001), again depending on the other conditions under which Donald's traits are initially inferred. Traits can describe temporary states or enduring characteristics, depending on one's beliefs about traits' malleability (Dweck, 1999).
Of course, this does not mean that trait terms can mean anything. But it does mean that traits are polysemous and ambiguous in other multiple ways. As Murphy and Andrew (1993) suggest, most of these meanings are probably computed online. So, rather than trying to discover traits' "true meaning," it seems more productive to ask what classes of knowledge give traits their meanings in particular situations.
Was this article helpful?