Global and local processing by 3- and 4-month-old infants

HEI-RHEE GHIM 0 1 PETER D. EIMAS 0 1 0 This research was supported by Grant HD 05331-17 from the Na tional Institute of Child Health and Human Development. We thank Joanne L. Miller, Gregory L. Murphy, and Bryan E. Shepp for their critical comments on earlier versions of this article and we thank June Shepp and Bonnie Tracy for assistance in testing subjects. Correspon dence and requests for reprints should be sent to Peter D. Eimas at the Department of Psychology, Brown University , Providence, RI 02912 1 Brown University , Providence, Rhode Island In two experiments with 3- and 4-month-old infants, we used a familiarization/novelty preference procedure to assess the ability of infants to acquire information about the global and local information in a complex visual pattern. The initial experiment established that individual infants were able to acquire and remember information about both the global forms and the local forms from which the global patterns were constructed. In addition, we found that the global and local forms were of nearly equal discriminability. Using these patterns and a Stroop-like interference paradigm, in the second experiment we obtained evidence for a global precedence effect that could not, we argue, be attributed solely to a difference in discriminability favoring the global stimuli. - In recent years, the perception of visual patterns has been shown to involve the processing of holistic proper ties as well as more elementary featural properties. Moreover, the presence of holistic information has often influenced the perception of the components. For exam ple, Weisstein and Harris (1974) demonstrated that the identification of simple line targets was better when the targets were embedded in a coherent contextual frame that was perceived as three-dimensional than when they were embedded in less coherent arrays that appeared two dimensional. Even more compelling evidence for an ef fect of holistic information on the perception of compo nent parts was provided by Williams and Weisstein (1978). They showed a facilitating effect of a contextual frame on the identification of line targets even when the comparison condition involved presenting the line targets in isolation. This pattern-line effect has also been obtained with measures of discriminability (e.g., Pomerantz, Sager, & Stoever, 1977), and, interestingly, it has been repli cated with infants as young as 3 months of age (Bomba, Eimas, Siqueland, & Miller, 1984; Colombo, Laurie, Martelli, & Hartig, 1984; Quinn & Eimas, 1986). The existence of pattern-line effects is taken as strong evidence that the perception of visual patterns does not proceed solely by first processing simple featural information, but rather involves the use of more complex configural in formation during the early stages of perception (e.g., Enos & Printzmetal, 1984; McClelland & Miller, 1979; Navon, 1977). Moreover, Navon (1977,1981, 1983) has argued strongly for what he has called a global precedence effect-the hypothesis that "perceptual processing proceeds from the global structure to the more local details" (Navon, 1983, p. 239). This strong view of processing has generated both controversy and considerable research. A common method of testing for a global (or local) precedence presents observers with large letters made up of small letters. The global and local letters may either match or not-that is, they may be consistent or inconsistent-and the observer's task is to identify the global letters on some trials and the local letters on other trials. The patterns are presented briefly and the observers are urged to respond rapidly. What has been found often by Navon (1977) and others is that response times to the global letters are unaffected by conflicting local charac ters, whereas the response times to the local letters are reliably slowed by the presence of inconsistent global letters-a Stroop-like form of interference (see Pomer antz, 1983). However, a global precedence is not always evidenced. For example, Martin (1979) found a local precedence when the global letters were constructed from only a few local letters and Hoffman (1980) found either a global or local precedence by reducing the quality (good ness) of the local or global letters, respectively. Not sur prisingly, the size of the forms appears to determine which precedence effect will be obtained (Kinchla & Wolfe, 1979), as does the relative discriminability of the global and local information (Pomerantz, 1983). As Pomerantz noted, there is reasonable evidence for both local and global precedence effects, but there is very little in the way of theory to explain how or when these effects will appear or what level or levels of processing they involve (see Miller, 1981). Moreover, in light of these results, we would argue that Pomerantz's (1983) definition of precedence effects in terms of one level's dominating the other and thereby capturing attention is more fitting than definitions that involve assumptions about an obligatory order of processing. Although there have been direct tests of pattern-line effects in infants and even an attempt to determine the manner in which contextual frames alter the discriminability of line targets in these subjects (Quinn & Eimas, 1986), there has been only one study to our knowledge that is directly related to the study of precedence effects in infants, that by Vurpillot, Ruel, and Castrec (1977).1 They found that 2- and 4-month-old infants perceived the differences in global forms that were made of local ele ments. The local elements, which were small, were not discriminated. When the local elements were larger, how ever, the infants perceived a change in the local elements but not a change in the global pattern. It is apparent that inferences about the existence of precedence effects must be drawn with caution from this study, inasmuch as Vur pillot et al. never demonstrated that their infants could in fact process both global and local information in a sin gle experimental setting. Furthermore, as Van Giffen and Haith (1984) noted, the perception of a change in the global pattern may actually have been based on a change in area of the global patterns or positioning of the local elements. Be that as it may, Vurpillot et al. 's study makes it quite clear that it is necessary to determine that infants are able to process the information at both the global and local levels and that both sources of information are ap proximately equally discriminable before statements about precedence effects are possible. It is also necessary, as Navon (1977) and Miller (1981) noted, to construct the patterns with identical global and local forms. This avoids confounding factors that might arise from differences in the importance or salience of the global and local proper ties or from differences in experience with them. In the present experiments we tested for global and local process ing with global and local forms that were i (...truncated)