Abstract

The goal of this research is to clarify the development of metaphor by using structure-mapping theory to make distinctions among kinds of metaphors. In particular, it is proposed that children can understand metaphors based on shared object attributes before those based on shared relational structure. This predicts (a) early ability to interpret metaphors based on shared attributes, (b) a developmental increase in ability to interpret metaphors based on shared relational structure, and (c) a shift from primarily attributional to primarily relational interpretations for metaphors that can be understood in either way. Two experiments were performed to test these claims. There were three kinds of metaphors, varying in whether the shared information forming the basis for the interpretation was attributional, relational, or both. In Experiment 1, children aged 5-6 and 9-10 and adults produced interpretations of the three types of metaphors. The attributionality and relationality of their interpretations were scored by independent judges. In Experiment 2, children aged 4-5 and 7-8 and adults chose which of two interpretations--relational or attributional--of a metaphor they preferred. In both experiments, relational responding increased significantly with age, but attributional responding did not. These results indicate a developmental shift toward a focus on relational structure in metaphor interpretation. Metaphor as Structure Mapping 2 METAPHOR AS STRUCTURE MAPPING: THE RELATIONAL SHIFT The study of metaphoric development is fraught with contradictions. Evidence that young children are quite poor at metaphoric interpretation sits side by side with equally compelling evidence that they are uniquely talented at metaphoric language. In experimental studies of metaphor comprehension, young children typically perform quite poorly. A 4-year-old asked "Can a person be sweet?" answers literally (e.g., "Not unless he was made out of chocolate") (Asch & Nerlove, 1960). Young children do badly at matching sentences with metaphorically related pictures (Kogan, 1975) and at choosing appropriate metaphorical completions for sentences (Gardner, Kircher, Winner, & Perkins, 1975). Not until 14 years of age can children explain metaphors such as "the prison guard was a hard rock." These and many other experimental results seem to indicate that metaphorical ability develops very slowly. Indeed, until recently, the dominant position was that metaphor develops only after the child has acquired basic competence at literal language (e.g., Inhelder & Piaget, 1958). This dour assessment contrasts sharply with the impression gained from children's spontaneous speech, which is so full of creative metaphors that some observers have viewed early childhood as a period of linguistic genius (Gardner, 1974). For example, a 15-month-old girl used "moon" to refer not only to the moon but to a half grapefruit and a hangnail (Bowerman, 1982). A 2-year-old boy I observed remarked that a crescent moon was "bent, like a banana," and on another occasion jumped into a pile of pillows and announced "leafs." It is unlikely that all such extensions could be accounted for as errors (Thomson & Chapman, 1975). This seems to leave us with the paradox that young children can produce metaphors on their own but cannot comprehend them in experimental studies. Part of the disparity may be due to the kinds of tasks used to assess comprehension. For example, children can demonstrate metaphoric ability earlier with enactment tasks than with verbal explanation tasks (Gentner, 1977a, 1977b; Pollio & Pickens, 1980; Vosniadou & Ortony, 1986). But even with age-appropriate methodology, young children perform far less well in comprehension tasks than their fluent production would lead us to expect (e.g., Gentner & Toupin, 1986). I suggest that an essential step in sorting out the developmental picture is clarifying what we mean by "metaphor." The term covers a number of different kinds of comparisons, varying in their complexity and in the nature of the commonalities they convey (Lakoff & Johnson, 1980). In this report I use the structure-mapping theory of analogy to derive distinctions among kinds of metaphors (Gentner, 1980, 1983, 1986). To preview the conclusions, I will argue that children can produce and comprehend metaphors based on common object attributes before metaphors based on common relational structure. Structure mapping. The basic intuition of structure-mapping theory is that an analogy is a mapping of knowledge from one domain (the base) into another (the target) which conveys that a system of relations that holds among the base objects also holds among the target objects. Thus an analogy is a way of noticing relational commonalities independently of the objects to which those relations apply. For example, Carnot in 1824 explained heat flow by analogy with a waterfall. The analogy conveys that just as a gradient from a high level to a low level will cause water to flow, given a path, so a gradient from a high temperature to a low temperature will cause heat to flow, given a heat path. This is a typical analogy in that the higher order relational structures are identical in base and target if the proper low-order correspondences among objects and functions are made (i.e., water ----> heat; level ---> temperature; and water path ---> heat path). In interpreting an analogy, people seek to put the objects of the base in one-to-one correspondence with the objects of the target so as to obtain maximum structural match. 1 The corresponding objects in the base and target do not have to resemble each other at all; object correspondences are determined by roles in the matching relational structures. Central to the mapping process is the principle of systematicity: People prefer to map systems of predicates linked by higher order relations with inferential import rather than to map isolated predicates. The systematicity principle is a structural expression of our tacit preference for coherence and deductive power in interpreting analogy. Besides analogy, other kinds of similarity matches can be distinguished in this framework, according to whether the match is one of relational structure, object descriptions, or both. As noted above, analogies discard object descriptions and map relational structure. Mere appearance matches are the opposite: They map Gentner Metaphor as Structure Mapping 3 aspects of object descriptions and discard relational structure. Literal similarity matches map both relational structure and object descriptions. Metaphor. Now let us apply this framework to metaphor. Metaphors can be divided into four partially overlapping categories: attributional metaphors, relational metaphors, double metaphors, and a category of complex metaphors that resist analysis as one-one mappings and that I will not consider here (Gentner, Falkenhainer, & Skorstad, 1987). Attributional metaphors2 (e.g., "Her arms were like twin swans") are mere appearance matches: they convey common object attributes. Here, the attributes "long;" "thin," and "graceful" can be mapped from the base domain of swans to the target domain of her arms. Relational metaphors (e.g., Shakespeare's "Look, he's winding up the watch of his wit; by and by it will strike") can be analyzed as analogies: they convey that the base and target share common relational structure. Here, the intended commonalities have nothing to do with the object attributes of a watch--a glass face, metal cogs, and so on; instead, the metaphor conveys the common relational structure of a person setting a mechanism that will later produce seemingly spontaneous external effects. Finally, besides pure relational and attributional metaphors, there are also double metaphors that are mixtures of the two, such as Verbrugge and McCarrell's (1977) example, "Plant stems are drinking straws for thirsty trees." This metaphor conveys both the common attributes "long, thin, tubular" and the common relational structure "raises fluids from a lower to a higher place in order to benefit some creature." Adults have been found to prefer relational metaphors over attributional metaphors and to focus primarily on relational commonalities in interpreting metaphors (Gentner, 1986; Gentner & Landers, 1985). In a prior study on which the present research is based, subjects produced interpretations of metaphors and rated their aptness, having first written out descriptions of the objects contained in the metaphors (Gentner, 1986). The results indicated a relational focus in metaphor interpretation in two ways. First, although the object descriptions contained both object attributes and relational information, only the relational information survived into the metaphor interpretations. For example, one subject described "cigarette" as follows: "chopped cured tobacco in a paper roll/with or without a filter at the end/held in the mouth/lit with a match and breathed through to draw smoke into the lungs/found widely among humans/known by some cultures to be damaging to the lungs/once considered beneficial to health." This description contains both relational and attributional information. But the same person, when interpreting the metaphor "Cigarettes are like time bombs," included only relational information: "They do their damage after some period of time during which no damage may be evident." The second indication of relational focus is that subjects considered metaphors more apt when they could find relational interpretations: Subjects' aptness ratings for metaphors correlated positively with the judged relationality of their interpretations. In contrast, attributional commonalities did not contribute to aptness: The correlation between aptness ratings and attributionality was nonsignificant and negatively trending. Adults thus demonstrate a relational focus both in interpreting and in judging metaphor. Do children show this same relational focus? Altho