Development of perceptual organization in infancy Paul C. Quinn, University of Delaware & Ramesh S. Bhatt, University of Kentucky

Transcription

1 Development of perceptual organization in infancy Paul C. Quinn, University of Delaware & Ramesh S. Bhatt, University of Kentucky To appear in: Oxford Handbook of Perceptual Organization Oxford University Press Edited by Johan Wagemans Preparation of this chapter was supported by Grant HD from the National Institute of Child Health and Human Development. We thank Johan Wagemans and two anonymous reviewers for their comments. Correspondence should be sent to Paul C. Quinn, Department of Psychology, University of Delaware, Newark, DE Introduction Even simple visual displays can have multiple interpretations. Consider the stimulus depicted in Figure 1a. Why is it that most adults report perceiving an overlapping hexagon and cross, despite the fact that other interpretations, such as those in Figures 1b-1d, are equally physically possible? As put by Metzger (1936/2006, p. 43, italics from original text), the stimulus distribution in the eye is always infinitely ambiguous. One could argue that the favored interpretation receives support from language and instruction given that during development we come to learn that the labels hexagon and cross refer to those particular constellations of contours. However, the rapid emergence of visual cognition (with many grouping phenomena evident in the initial months of life), combined with the difficulty of the problem, suggests that the development of perceptual organization results from the imposition of strong constraints (Quinn, Bhatt, & Hayden, 2008a). This chapter will take up the task of identifying those constraints and explicating their developmental determinants. In particular, we will examine how the constraints are a mix of the inherent operational characteristics of the visual system and the learning engendered by a structured environment (Bhatt & Quinn, 2011). First, however, we consider some theoretical accounts of the ontogenesis of perceptual organization. 1

2 C D A B Figure 1. (A) Configuration of contours perceived as hexagon and cross, even though one could just as readily perceive (B), (C), and (D). (After Metzger, 1936/2006.) 2. Historical theoretical positions on the development of perceptual organization Gestalt accounts. For the Gestaltists, holistic percepts are realized even on initial presentation of a visual pattern (Wagemans et al., 2012). As stated by Köhler (1929, p. 163), elementary organization is an original sensory fact and it occurs because our perceptual systems are constrained to follow certain grouping principles that operate on the basis of the proximity, similarity, common movement, and good continuation properties of the elements. The perception of one organization over other organizations that are equally physically possible reflects adherence to such principles (Wertheimer, 1923/1958). Emphasizing the nativist basis for perceptual organization, Koffka (1935) observed that, Whereas to traditional psychology the articulation of our field into things appears as a clear example of experience or learning, our theory considers this articulation as the direct result of the spontaneous organization aroused by the stimulus mosaic. (p. 209). Zuckerman and Rock (1957) sided with Gestalt claims of innate organizing processes on the grounds of logic and parsimony. That is, if one does not posit such processes, then the starting point for infants is an unorganized mosaic of sensory impressions (p. 278), and experience with different shapes and forms must somehow induce the transformation of sensory data into bounded regions. Such transformation is presumably mediated through memory, but according to Zuckerman and Rock, if that memory consists of amorphous sensations rather than cohesive shapes, then it is unclear how it could lead to subsequent organized percepts. Instead, it is simpler to assume that innate organizing processes account for the initial structuring of visual displays into coherent patterns. As summarized by Zuckerman and Rock (p. 291), the organization of the visual field into shaped areas is not an outcome of learning past experience cannot carve visual form out of initially formless perception. Learning accounts. Two other views of the development of perceptual organization have proposed mechanisms that allow one to more readily envision how organization could emerge, even if it is not 2

3 the initial starting point. For Hebb (1949), perception of a whole object is a learned process that is founded in perception of the individual features of the object and the integration of those perceptions as achieved through eye movements. As described by Hebb (p. 83), If line and angle are the bricks from which form perceptions are built, the primitive unity of the figure might be regarded as the mortar, and eye movement as the hand of the builder. For Hebb, the emergence of perceptual organization would take considerable developmental time because of dependence on improvements in eye movements that yield more wholistic perceptions as visual scanning becomes more systematic. Another account of the emergence of perceptual organization relies neither on inherent constraints nor perceptual learning that occurs from development of visual scanning, but rather on learning of probabilistic image statistics derived from regularities in the environment (Brunswik & Kamiya, 1953; Elder & Goldberg, 2002; Elder, Chapter 7, this volume). Consider the organizing principle of proximity, which specifies that close elements will be grouped together. In the Brunswik and Kamiya view, proximity may actually be learned because image elements that correspond to the same object are likely to be closer to each other than elements that correspond to different objects. Likewise, in the case of lightness similarity, discontinuities in luminance cues are correlated with boundaries where one object ends and another begins. The discovery of such correlations by infants can presumably be used as a basis for integrating sequences of elements that project from common structures in a visual scene. With different theorists offering differing accounts of the development of perceptual organization, some stressing innate grouping factors, and others emphasizing ways in which visual order could emerge through maturation of internal mechanisms or experience with a structured environment, we turn to a discussion of the evidence. 3. Initial eye movement evidence in infants: salapatek (1975) At the time that Gestalt theory and reactions to it were being composed, methodologies were not yet available to investigate perceptual abilities in infants. However, such methods did become available in the 1960s and 1970s, and one technique in particular provided some early evidence relevant to the debate. Specifically, Salapatek (1975) recorded infants eye movements while they visually scanned simple outline figures, and reported a developmental trend over the first months of life in which scanning was initially limited to single features and gradually expanded to include multiple features and eventually the whole pattern. These eye movement data are consistent with a Hebbian account of the development of perceptual organization, although one can question how direct a relation there is between fixation and the surrounding expanse of visual attention. That is, if visual attention is distributed broadly about the fixation point, then an infant who fixates a corner of a triangle could actually be processing information across the entire triangle. For this reason, it is unclear what inferences can be drawn from infant visual scanning data, at least as they pertain to the ontogenesis of perceptual organization. 4. Demonstrations of organizational phenomena in infants 3

4 Looking-time procedures used to study perceptual organization in infants are based on the infant s visual preference for novel stimuli (Fantz, 1964). To determine whether two stimuli are discriminable, for example, infants can be familiarized with one of the stimuli and subsequently presented with the familiar stimulus paired with the novel stimulus. A preference for the novel stimulus that cannot be attributed to a priori preference implies that infants have recognized the familiar stimulus and can discriminate between it and the novel stimulus Configural superiority A strategy for researchers interested in the start-up of visual cognition has been to take empirical phenomena supportive of a particular mental faculty in adults and adapt looking time procedures to study those same phenomena in infants. One such occurrence relevant to perceptual organization is the configural-superiority effect (Pomerantz, 1981, Chapter 26, current volume). In adults, configural superiority is in evidence when the mirror image line elements shown in Figure 2A are found easier to discriminate when embedded in the non-informative contextual frame shown in Figure 2B (Pomerantz, Sager, & Stoever, 1977). This result poses difficulty for feature analytic models of visual processing, because if one were processing only the features of the visual forms (i.e., the individual line segments), then the stimuli in Figure 2B should be more confusable than those in Figure 2A given the featural overlap in the horizontal and vertical line segments. Instead, the finding suggests that emergent relations between features (i.e., angles, corners, whole forms) are represented when processing visual patterns. It could be argued that the configural-superiority effect shown in Figure 2A and 2B is linguisticallybased given that labels such as arrow versus triangle may generate an acquired distinctiveness of the patterns. However, that interpretation is defeated by demonstrations of configural superiority in young infants (Bomba, Eimas, Siqueland, & Miller, 1984; Colombo, Laurie, Martelli, & Hartig, 1984; Quinn & Eimas, 1986). In Quinn and Eimas, for example, 3- to 4-month-olds familiarized with a single line element showed no preference when tested with the familiar element paired with its mirror image line element (Figure 2A). By contrast, when these elements were embedded in the right angle contextual frame (Figure 2B), the infants reliably preferred the novel stimulus. These results suggest that the configural-superiority effect is perceptually-based, and that young infants represent more global visual processing units that emerge when simple components are grouped together Global precedence Another perceptual effect that has been considered as evidence of organization in adults and that has been of interest to developmentalists is the global-precedence effect (Kimchi, Chapter 18, current volume; Navon, 1977). In the procedure used to generate this effect, adult observers are presented with a multilevel stimulus consisting of a large letter made from small letters. The global letter matches or does not match the local letter and the observer s task is to identify either the global letter or local letters. The key findings are that (1) response times are faster to the global letter, (2) conflicting local letters do not impact processing of the global level, and (3) a conflicting global letter interferes with processing of the local letters. This pattern of outcomes indicates that global aspects of a stimulus are processed and recognized before local aspects. Ghim and Eimas (1988) investigated whether a global-precedence effect could be demonstrated in 4

5 young infants. In one condition, 3- to 4-month-olds were familiarized with a global square made up of local squares followed by either a local or global preference test. The local test contrasted a pair of global diamond stimuli, one constructed from local squares and the other from local diamonds. By contrast, the global test paired a global square with a global diamond, each composed of novel local diamonds. If global precedence is occurring, then in the local test, the novelty at the global level would lead infants to divide their attention evenly between the two stimuli, even though there is a source of novelty at the local level residing in the local diamonds. However, in the global test, infants should prefer the global diamond, even though there is a competing source of novelty from the local diamonds. These predictions were confirmed: Infants in the local test did not respond differentially, whereas those in the global test preferred the global diamond (even though a control condition showed that infants were sensitive to the change in the local elements). The findings provide evidence that, as is the case with adults, global information has a processing advantage over local information in young infants (see also Frick, Colombo, & Allen, 2000) Subjective contours Yet another manifestation of organization in adult vision is the perception of subjective contours (Kanizsa, 1955; van Lier & Gerbino, Chapter 11, current volume). Consider Figure 2C: adults perceive a white square atop some pacman shapes. The contour appears to continue across the white space between the shapes, thereby suggesting a completion process. Although one can argue for a topdown explanation and suggest that the completion process is facilitated by knowledge of the square form, this explanation is weakened by demonstrations that infants perceive illusory contours (Ghim, 1990; Hayden, Bhatt, & Quinn, 2008; Johnson & Aslin, 1998; Kavsek, 2002). For example, Ghim (1990) reported that 3- to 4-month-olds were more likely to display novelty preferences in tasks involving a pattern that elicited the perception of subjective contours (Figure 2C) versus one that did not (Figure 2D) relative to tasks involving two patterns neither of which produced subjective contours. In addition, after familiarization to an outline square, infants preferred a pattern that did not produce subjective contours (Figure 2D) to one that did produce the illusory square in adults (Figure 2C). This evidence suggests that, like adults, young infants are capable of a completion process that produces the perception of subjective contours. 5

6 Configural Superiority A B Subjective Contours C D Figure 2. Configural Superiority: (A) Line segments in isolation. (B) Line segments embedded in a right angle contextual frame. (After Quinn & Eimas, 1986.). Subjective Contours: Configuration of elements produce (C) and do not produce (D) the perception of a square shape. (After Ghim, 1990.) Demonstrations of configural superiority, global precedence, and subjective contours in infants suggest that at least some of the mechanisms that produce perceptual organization in adults are also functional in the initial months of life. However, these demonstrations are less informative about how infants relate individual elements to each other. For example, in the cases of configural superiority and global precedence, was it the Gestalt principles of closure, good continuation, proximity, lightness similarity, form similarity, or some combination, that allowed infants to organize the patterns? Similarly, in the case of subjective contours, any of the above principles with the exception of proximity, could be involved. To better identify which specific grouping factors are functional during early development, some investigators have taken the approach of studying how infants will respond to displays of elements that could be organized by one or another principle. The chapter turns to a discussion of these studies. 5. An influential study: Kellman and Spelke (1983) Kellman and Spelke (1983) presented 4-month-olds with a display of a rod partly occluded by a block. The question was whether infants represented the continuity of the rod behind the block based on the gestalt properties of the visible portions of the rod including their good continuation and similarity of shape. After familiarization with the rod-block display, infants were presented with a complete rod versus a broken rod. If infants represented the continuity of the rod, then they should respond to the complete rod as familiar and the broken rod as novel. However, if the rod fragments were not grouped together, then the broken rod should be perceived as familiar and the complete 6

7 rod as novel. The infants looked equally to the complete and broken rod displays, a null result that was difficult to interpret. A follow-up experiment was conducted that was a replication of the initial experiment, except that the cue of common motion was added to the visible portions of the rod. In this instance, the infants preferred the broken rod. An additional experiment that pitted common motion against good continuation and similarity confirmed that it was common motion alone rather than the combination of common motion, good continuation, and similarity that enabled infants to group the rod. Moreover, using a similar methodology, Spelke (1982) reported that same-aged infants perceived the continuity of two adjacent objects as long as their surfaces were contiguous and even when those surfaces were dissimilar in size, shape, and textural markings. These results led Spelke (1982) to develop a hybrid account of the development of object perception, incorporating innate organizing principles as well as a role for learning based on experience with a structured environment. Specifically, Spelke argued that infants at birth are constrained by two core organizational principles, common movement and connected surface. Adherence to these principles would parse from a visual scene those surfaces that move together and maintain their coherence as they move and grant them the status of objects. The resulting object blobs can then be tracked over real time. Such experience, according to Spelke, allows infants to discover that objects exhibit other properties including proximity of parts, similarity of surface, and good continuation of contour (Brunswik & Kamiya, 1953). In this way, some of the principles that were considered to be innate organizing principles by the Gestaltists were by the Spelke account learned through their natural correlation with the core principles. 7

8 6. Further work on perceptual grouping of visual displays by infants via classic organizational principles Questions that arise from the initial Spelke (1982) account concern the status of grouping principles such as similarity, good continuation, and proximity. While common motion was found to be a more potent determinant of grouping relative to similarity and good continuation in a stimulus display in which different sources of information were in competition, one can ask whether similarity and good continuation might be functional when not in competition with another principle. Moreover, although similarity and good continuation were not sufficiently strong to provide a basis for grouping when they were the sole sources for organization in the case of partial occlusion with the rod-box display, these principles might be functional for displays in which the elements are completely visible. In addition, Kellman and Spelke (1983) investigated a particular type of similarity (form), thus leaving unresolved the issue of whether infants might be able to utilize other forms of similarity (e.g., lightness). Furthermore, one other Gestalt principle, proximity, was not investigated, thereby keeping open the question of its functionality Lightness similarity Quinn, Burke and Rush (1993) asked whether 3-month-olds could utilize lightness similarity to organize columns or rows of elements that could be grouped only on the basis of their lightness versus darkness (see also Quinn & Bhatt, 2006). The test stimuli were horizontal versus vertical bars (see Figure 3, top panel). If the organization in the row and column arrays is apprehended, then infants familiarized with columns should prefer horizontal bars and infants familiarized with rows should prefer vertical bars. The findings provided positive evidence for use of lightness similarity: infants preferred the novel organization of bars. An additional control experiment showed that infants could discriminate between arrays differing in the shape (square vs. diamond) of the dark or light elements. This latter finding mitigates explanations of the preference for the novel organization based on immature resolution acuity and indicates that infants were able to perceive the individual elements of the displays and organize them into larger perceptual units (i.e., rows vs. columns) based on lightness similarity. Of note is that Farroni, Simion, Umilta, and Valenza (2000) used a similar methodology to argue that even newborns adhere to lightness similarity when organizing visual patterns; however, because that study did not determine if the individual light elements could be resolved, it left open the question of whether the displays were organized via the proximity of the dark elements Proximity Another classic grouping principle investigated was proximity (Quinn, Bhatt, & Hayden, 2008b). As shown in Figure 3 (bottom panel), using the same methodology of Quinn et al. (1993), 3- to 4-montholds were presented with arrays of elements that could be organized into rows or columns via proximity, and then tested with horizontal versus vertical bars. Infants preferred the test stimuli with the novel organization, and subsequent control experiments indicated that the preferences were not attributable to an a priori preference or to an inability to resolve elements within the rows and columns. The results indicate that proximity joins lightness similarity as a grouping principle that can be used to organize visual patterns by young infants. 8

9 LUMINANCE FAMILIAR TEST vs. PROXIMITY FAMILIAR TEST Columns vs. Rows vs. Figure 3. Luminance (top panel): Familiarization and test stimuli used in the Quinn, Burke and Rush (1993) study investigating whether 3-month-olds can organize visual patterns in accord with lightness similarity. Proximity (bottom panel): Familiarization and test stimuli used to determine whether infants adhere to proximity when organizing visual patterns. From Quinn, Bhatt, and Hayden (2008b) Good continuation A third classic static principle investigated was good continuation (Quinn & Bhatt, 2005a). In contrast to the column versus row methodology used to study lightness similarity and proximity, a methodology was adopted that had been used to investigate good continuation grouping by adults (Prinzmetal & Banks, 1977). The displays (shown in Figure 4) consisted of a line of circular distractors and a square or diamond target. Infants were presented with one pattern and then tested for discrimination between the familiar pattern and a novel one. In the top panel (a), the target appeared in line, embedded, or aligned (and thus in good continuation) with the distractors, whereas in the bottom panel (b), the target was off line with the distractors. The expectation is that if infants perceived the patterns in accord with good continuation, then the change in the target should be more difficult to detect when the target is in a good continuation relation with the distractors as in the in-line condition. By contrast, in the off-line condition, the target would not group with the distractors and would retain its status as an independently processed unit of information, thereby increasing the likelihood that a change in its form would be detected. Three- to 4-month-olds preferred the novel test stimulus in the off-line condition, but not in the in-line condition. This evidence suggests that good continuation is a third organizational principle available to young 9

10 infants. Figure 4. Examples of the familiarization and test stimuli used in Quinn and Bhatt (2005a). The in-line condition is depicted in panel (a); the off-line condition is shown in Panel (b) Form similarity The functionality of form similarity in young infants was examined by Quinn, Bhatt, Brush, Grimes, and Sharpnack (2002), who drew upon the methodology that was used to investigate lightness similarity and proximity. As shown in Figure 5, 3- to 4-month-olds were familiarized with rows or columns of Xs versus Os, and then tested with horizontal versus vertical bars. If infants group the familiarization stimulus into rows or columns via form similarity, then they should prefer the novel organization of bars. However, the infants did not display such a preference, even when familiarization time was doubled; instead, attention was divided between the test stimuli. A control study showed that infants were capable of discriminating between the familiarization arrays and arrays that consisted entirely of Xs or Os. This latter result indicates that failure of the infants to use form similarity was not due simply to an inability to discriminate between the constituent X and O shapes. 10

11 Figure 5. Examples of the familiarization and test stimuli used to test for perceptual organization by form similarity in Quinn, Bhatt, Brush, Grimes, and Sharpnack (2002). With the data thus far described not demonstrating young infants use of form similarity, Quinn et al. (2002) tested older infants, 6- to 7-month-olds, on the form similarity task. This age group preferred the novel organization. Thus, 6- to 7-month-olds, but not 3- to 4-month-olds, can organize visual patterns in accord with form similarity. In combination with outcomes indicating that 3- to 4-montholds can utilize lightness similarity, proximity, and good continuation to organize visual patterns under similar testing conditions (Quinn & Bhatt, 2005; Quinn et al., 1993, 2008b), the results indicating that only 6- to 7-month-olds can use form similarity suggest that different Gestalt principles may become functional over different time courses of development and that not all principles are readily deployed. The findings are inconsistent with a strict Gestalt view that all organizing principles are automatically activated upon first encounter with a visual pattern (e.g., Köhler, 1929). The data are additionally consistent with evidence indicating that adults have independent luminance- and edge-based grouping mechanisms (Gilchrist, Humphreys, Riddoch, & Neumann, 1997). They are further in accord with the finding that some visual agnosics show intact lightness similarity and proximity grouping, but impaired shape configuring and form-based grouping ability (Behrmann & Kimchi, 2003; Humphreys, 2003), and the result that individuals with Williams Syndrome show superior lightness similarity and good continuation grouping abilities relative to those for form similarity (Farran, 2005). The developmental evidence contrasting the time course of emergence of the principles of proximity and form similarity in infants is moreover consistent with microgenetic evidence in adults indicating that proximity grouping occurs more rapidly than form-based grouping in the time course of processing (Ben-Av & Sagi, 1995; Han, Humphreys, & Chen, 1999). However, we now consider evidence indicating that young infants inability to use form to organize visual images is not absolute Contribution of learning to the development of form-based grouping Because Quinn et al. (2002) reported a later development of form similarity, Quinn and Bhatt (2005b) sought to determine whether this development was driven by maturation or learning. They reasoned that if form similarity is under experiential control, then it might be possible to find a stimulus display or procedural manipulation that would allow 3- to 4-month-olds to organize visual patterns in accord with form similarity. Alternatively, if infants use of form similarity is maturationally determined, then 11

12 methodological variants would not be expected to bring about positive evidence that form similarity is functional in the younger age group. Given that Quinn et al. found that the X-O form contrast yielded null results with 3- to 4-month-olds, Quinn and Bhatt tested this age group with two other form contrasts, square vs. diamond geometric shapes and H vs. I letter shapes, on the form similarity task. Neither contrast was successful in producing a preference for the novel organization; infants in both cases divided attention between the horizontal and vertical bars. This result suggests that young infants inability to organize by form similarity is not a specific deficit with Xs versus Os, but rather a more general phenomenon. A second attempt to determine if 3- to 4-month-olds could be induced to use form similarity employed a training regime. Specifically, Quinn and Bhatt (2005b) asked whether variations in the patterns used to depict rows or columns during familiarization would enhance infants performance in the form similarity task. One may reason that pattern variation will facilitate performance because the invariant organization of the stimuli will be more easily detected against a changing background. In other words, variation might provide infants with the opportunity to form concepts of rows or columns. To investigate this possibility, the form similarity task that had previously produced null results (when each of the three different form contrasts was presented individually) was administered, but in this instance, with each of the three form contrasts presented during a single familiarization session (see Figure 6). Younger infants now preferred the novel organization of bars. This striking result suggests that 3- to 4-month-olds can use form similarity to organize elements if they are provided with varied examples with which to abstract the invariant arrangement of the pattern. The outcome is theoretically significant because it demonstrates that perceptual learning may play a role in acquiring some aspects of visual organization. Moreover, following Goldstone s (2003) proposal that one mechanism by which perceptual learning occurs is by increasing attention to relevant information and decreasing attention to irrelevant information, Bhatt and Quinn (2011) have suggested that variability led to grouping based on shape similarity because it enhanced infant attention to global structures and diminished attention to local elements. A B C FAMILIAR TEST Figure 6. Familiarization and test stimuli used in Quinn and Bhatt (2005b). 12

13 7. Perceptual grouping of visual displays by infants via modern organizational principles While the classic grouping principles were described by Wertheimer (1923/1958), the grouping principles that will be examined in this section, connectedness and common region, were introduced by Palmer and Rock in the 1990s (Palmer, 1992; Palmer & Rock, 1994; Rock & Palmer, 1990; see also Brooks, Chapter 4, this volume) Connectedness Rock and Palmer (1990) described the principle of connectedness as the visual system s tendency to group together connected entities, and remarked that connectedness may be the most fundamental principle of grouping yet uncovered. (p. 86). To determine whether sensitivity to connectedness is operational in early infancy, as shown in Figure 7, infants as young as 3 months of age were habituated to the connected patterns shown in Panels A or B, and then administered a preference test pairing connected elements (Panel C) with disconnected elements (Panel D) (Hayden, Bhatt, & Quinn, 2006). The expectation was that if the infants organize the habituation patterns on the basis of connectedness, then they should display a novelty preference for the disconnected-element test stimulus. This outcome was observed, and a control condition showed that it could not be attributed to a spontaneous preference. The results indicate that young infants are sensitive to the connectedness principle. Figure 7. The stimuli used in Hayden, Bhatt, and Quinn (2006). Infants in the habituation conditions were habituated to the connected patterns in panel A or B and tested with the patterns in panels C and D. Infants in the no-habituation condition were tested with the patterns in panels C and D without prior exposure to the patterns in panels A and B. 13

14 7.2. Common region Another newer grouping principle is common region, which states that elements within a region are grouped together and separated from those in other regions (Palmer, 1992). Palmer has also proposed that common region is driven by a characteristic that is external to the elements themselves. In other words, the common region quality that engenders grouping of elements is not inherent in the elements themselves. By contrast, other grouping principles such as similarity are based on intrinsic characteristics of the to-begrouped elements. Palmer thus distinguished between extrinsic versus intrinsic organizational cues and suggested that common region is an extrinsic cue. This distinction raises the possibility that common region could be a different kind of organizational cue than many others, thereby adding to the importance of understanding its emergence in infants. To examine whether young infants use common region to organize visual patterns, 3- to 4-month-olds were familiarized with a display consisting of two pairs of shapes, with one pair (e.g., A and B) located together in a region and the other pair (e.g., C and D) located together in another region, see Figure 8 (Bhatt, Hayden, & Quinn, 2007). The locations of the individual shapes changed from one trial to the next, but the shapes A and B always shared a region while the shapes C and D shared another region. Infants were then tested with a within-region grouping (e.g., AB) versus a between-region grouping (e.g., BC; see Figure 8). Importantly, because the physical distance between A and B versus B and C was equivalent, the only difference between the A and B versus B and C pairs was that the former pair shared the same region, whereas the latter pair was from different regions. If common region is functional in infancy, then the A and B elements should be grouped together because they always shared the same region. That is, infants should find the within-region grouping to be familiar and the between-region grouping to be novel, and respond differentially to these patterns during test. 14

15 Figure 8. Examples of the stimuli used in Bhatt, Hayden, and Quinn (2007). Infants were habituated to two pairs of shapes, with one pair sharing a vertical region and the other pair a different vertical region. Infants were then tested for their preference between a pair of shapes that had shared a common region during habitation (within-region pair) versus a pair of shapes that had been in different regions during habituation (between-region pair), both presented in novel horizontal regions. 15

16 Another aspect of Bhatt et al. (2007) is that it asks whether grouping will carry over to novel regions, given that infants were habituated to vertical regions and tested with horizontal regions. This manipulation allows one to determine whether the perceptual system expects grouping to remain intact when presented with elements that were previously grouped based on one set of regions are subsequently encountered in novel regional configurations. Presumably, if grouping and perceptual organization are to be functionally advantageous, they need to allow the world to be structured into meaningful entities that transcend particular situations. The major result from Bhatt et al. (2007) was that the infants discriminated the grouping of elements from different regions from the grouping of elements that had shared a common region during habituation. Moreover, Hayden et al. (2008) extended these results to regions formed by illusory contours. The findings that infants are sensitive to common region suggest that the extrinsic nature of this cue did not preclude its role as an organizing factor. In other words, infants, like adults, are not solely dependent upon the intrinsic nature of elements to organize them; they are able to use extrinsic factors such as common region to organize. Additionally, the result that performance transferred from differently shaped regions from familiarization to test provides evidence that the perceptual organization abilities of infants can produce processing units of an abstract nature. This latter result actually points toward a unitization process by which previously disparate elements become grouped and begin to function as coherent units in new contexts (Bhatt & Quinn, 2011; Goldstone, 2003). 8. Relations among the principles 8.1. Transfer of organization across principles Although the research reviewed thus far in the chapter suggests that there is flexibility in early grouping in that perceptual units formed by applying a particular grouping principle can be generalized to novel patterns organized by the same principle, one may also ask whether perceptual units formed from application of one principle can be transferred to process a visual pattern organized by a different principle. To this end, Quinn and Bhatt (2009) investigated the possibility of transfer of organization between two principles, lightness similarity and form similarity, both of which were previously shown to be functional in 6- to 7-month-olds. Six- to 7-month-olds were familiarized with arrays that could be organized into columns or rows based on lightness similarity. The infants were then given a novelty-preference test that paired arrays that could be organized into columns or rows based on form similarity (see Figure 9, top panel). If infants can organize the familiarization patterns by lightness similarity and use the represented organization as a basis for processing test patterns organized by form similarity, then they should prefer the novel organization. It should be noted that this transfer task is more demanding than the lightness similarity task of Quinn et al. (1993) because it calls upon infants to group the elements in the test displays based on their form similarity. The infants performed in accord with this expectation, indicating that perceptual units formed from application of one grouping principle (lightness similarity) can be transferred to apprehend an organization defined by a different grouping principle (form similarity). 16

17 Luminance Shape Shape Shape Figure 9. Illustrations of the luminance shape (top panel) and shape shape tasks (bottom panel) presented to infants in Quinn and Bhatt (2009) to examine whether infants will learn to use shape cues to organize if presented in the context of organization based on luminance cues. 17

18 Kangas, Zieber, Hayden, Quinn, and Bhatt (2011) also reported transfer of organization from common region to proximity in 6- to 7-month-olds, but not in 3- to 4-month-olds; however, Kangas et al. did demonstrate transfer of organization from connectedness to proximity at the younger age. These latter results indicate that transfer of organization across principles is evident early in life, although it continues to undergo quantitative change during infancy Perceptual scaffolding Given transfer between lightness and form similarity, one can inquire as to whether evidence might be found for perceptual scaffolding, a process by which learning based on an already functional organizational principle enables an organizational process that is not yet functional. That is, might infants who are otherwise not able to group based on an organizational principle be induced to do so if they are previously allowed to group elements based on an already functional organizational process? To answer this question, Quinn and Bhatt (2009) capitalized on the prior evidence showing that 3- to 4-month-old infants readily organize via lightness similarity (Quinn et al., 1993), whereas organization by form similarity is not readily exhibited until 6 to 7 months of age (Quinn et al., 2002), and administered the procedure depicted in Figure 9 (top panel) to a group of 3- to 4-month-olds. The younger infants succeeded in the task, thereby showing that the already-developed luminance-based organizational system facilitated grouping based on form similarity. This conclusion was upheld by the null performance of a control group of 3- to 4-month-olds who were familiarized and tested with the form elements shown in Figure 9 (bottom panel). Taken together, the results highlight a scaffolding process that may engender learning by enabling infants to group based on a new cue using an already functioning organizational process. Importantly, this work demonstrates that new organizational principles can be learned via bootstrapping onto already functioning organizational principles as Spelke (1982) had suggested A salience hierarchy? Although the chapter has thus far documented that a variety of organizational principles are operational in infants, what has not yet been discussed is whether there is differential salience among the cues. That is, are there differences in cue salience when multiple cues are concurrently available in a stimulus display presented to infants? This question derives significance because of the previously discussed differences in how readily principles such as lightness similarity and form similarity are deployed, and because of arguments that connectedness may be the most fundamental of all the principles (Rock & Palmer, 1990). In an initial experiment that tested the salience of connectedness versus form similarity, 6- to 7-month-olds were habituated to a pattern that could be organized on the basis of both connectedness and shape similarity (Hayden, Bhatt, & Quinn, 2009). The stimuli contained alternating rows or columns of two different shapes (Xs and Os). The shapes were connected by a black bar in the same configuration (rows or columns) in which the shapes were organized (see Figure 10). Following habituation, infants were tested with a pair of new stimuli: one in which connectedness was altered (by breaking the connectedness among the shapes), and the other in which shape organization was altered (a change from rows to columns or vice versa). The connectedness manipulation was accomplished by positioning the previously connecting lines higher, rather than using shorter lines in their original familiarization location, to keep the total amount of contour constant across the displays. X-O stimuli were used to depict the shape contrast; while one could have used alternative displays to depict the shape contrast (e.g., square vs. diamond), several different shape contrasts presented to infants have yielded equivalent grouping results (Quinn & Bhatt, 2005b). If 18

19 one of the perceptual organizational cues (connectedness vs. shape similarity) was more salient than the other, the change induced by the manipulation of this cue should be more novel and the infants should look longer at this pattern than at the pattern in which the less salient cue was altered. The key finding was that infants preferred the pattern displaying the change in connectedness, a result suggesting that connectedness is more salient than shape similarity. A B C Habituation Stimulus Change in UC Change in Shape Figure 10. Examples of the stimuli used by Hayden, Bhatt, and Quinn (2009). Infants were habituated to patterns of the kind shown in panel A. These patterns could be organized based on both connectedness cues as well as shape similarity cues. The infants were tested with a pattern in which the connectedness was altered (panel B) paired with a pattern in which the shape similarity was altered (panel C). Hayden et al. (2009) next examined the salience relations of connectedness and lightness similarity by repeating their experimental procedure, except that the patterns previously organized by shape (i.e., x vs. o) were now organized by lightness (i.e., dark vs. light squares). In this case, infants preferred to look at the pattern displaying a luminance change to a significantly greater degree than the pattern displaying a connectedness change, a result suggesting that luminance similarity was more salient than connectedness. The pattern of results of Hayden et al. provide evidence that there is a luminance connectedness shape salience hierarchy operating among the organizational cues to which 6- to 7-month-olds have been shown to be sensitive. 9. Further evidence on the flexibility of the principles While having a set of organizing principles functioning in the initial months establishes the coherence of visual patterns, it is also the case that such principles need to work in conjunction with other cognitive processes such as concept formation. This observation suggests that some flexibility may be needed in the deployment of the principles given that visual features that are diagnostic of a category can in certain instances be features that would not be selected by Gestalt organizing principles. Schyns, Goldstone, and Thibaut (1998) have therefore argued for a flexible system of perceptual unit formation, one in which some of the features that come to define objects are extracted during concept learning. Moreover, concepts possessed by an individual at a specific point in time should affect subsequent perceptual organization processes. 19

20 Quinn and Schyns (2003) undertook a set of experiments to determine whether features that are specified as coherent by Gestalt principles would be overlooked by young infants if alternative means of perceptual organization are suggested by presenting a category of objects in which the features uniting the objects are non-natural in the Gestalt sense. In Experiment 1, 3- to 4-month-olds were familiarized with a number of complex figures, examples of which are shown in the top portion of Figure 11. Subsequently, during a novelty preference test, the infants were presented with the pacman shape paired with the circle shown in the bottom portion of Figure 11. The infants preferred the pacman shape, a finding which suggests that they had parsed the circle from the complex figures via good continuation (Quinn, Brown, & Streppa, 1997). Figure 11. Examples of the familiarization and test stimuli used in Quinn and Schyns (2003) and Quinn, Schyns, and Goldstone (2006). If the infants can parse the circle from the familiar patterns in accord with good continuation, then they should prefer the pacman shape over the circle during the test trials. 20

21 In Experiment 2, Quinn and Schyns (2003; see also Quinn, Schyns, & Goldstone, 2006) asked whether an invariant part abstracted during category learning would interfere with perceptual organization achieved by good continuation. Experiment 2 consisted of two parts. In Part 1, infants were familiarized with multiple exemplars consistent with category learning, with each exemplar marked by an invariant pacman shape, and subsequently administered a novelty preference test that paired the pacman and circle shapes. Examples of the stimuli are shown in Figure 12. The pacman shape was recognized as familiar, as evidenced by a preference for the circle shape. Part 2 of the procedure was then administered and it consisted of a replication of the procedure from Experiment 1. If the category learning from Part 1 of Experiment 2, in particular, the representation of the invariant pacman shape, could interfere with the Gestalt-based perceptual organization that was observed in Experiment 1, then the preference for the pacman shape observed in Experiment 1 should no longer be observed in Part 2 of Experiment 2. In fact, if representation of the pacman shape carried over from Part 1 to Part 2 of Experiment 2, one would expect the opposite result in which infants continue to prefer the circle in the test phase. The latter result was observed and it suggests that perceptual units formed during category learning can interfere with formation of perceptual units organized by good continuation. The bias set by good continuation can thus be thought of as softwired. More generally, an individual s history of categorization can affect their subsequent organizational processes. Figure 12. Examples of the familiarization and test stimuli used in Quinn and Schyns (2003) and Quinn, Schyns, and Goldstone (2006). If the infants can extract the invariant pacman from the familiar patterns, then they should prefer the circle shape over the pacman shape during the test trials. 21

22 10. Conclusions This chapter has reviewed evidence on the development of perceptual organization, described against the backdrop of different theoretical views, including those that emphasize innate organizing principles and others that highlight perceptual learning. The studies clearly show that several phenomena that have been taken as evidence of perceptual organization in adults, such as configural superiority, global precedence, and subjective contours, can be demonstrated in infants. The data also suggest that different organizational principles may become functional over different time courses of development, may be governed by different developmental determinants, i.e., maturation versus experience, have differential salience, and that not all principles are readily deployed in the manner originally proposed by Gestalt theorists. The principles were additionally shown to be flexible in their operation in terms of producing units of processing that would transfer across different displays organized by the same principle and also across different principles. In this sense, the units produced by the infant s organizational processes may be regarded as conceptual-like in their generalizability. To comment further on the differences among grouping principles, there is evidence for early functionality of classic organizational principles that include common motion, good continuation, lightness similarity, and proximity, as well as for the modern organizational principles of common region and connectedness. By contrast, form similarity was shown to be later developing and not as readily deployed. However, form similarity was shown to be activated when young infants were provided with multiple element contrasts, thereby suggesting a role for perceptual learning in its emergence. Form similarity was also activated when pulled along by the already functional principle of lightness similarity, thus demonstrating a perceptual scaffolding process by which new organizational principles can be learned. Overall, the evidence points to a hybrid model to explain the development of perceptual organization. As contended by the Gestaltists (Koffka, 1935; Kohler, 1929; Metzger, 1936/2006; Wertheimer, 1923/1958), as well as Zuckerman and Rock (1957), a number of grouping principles are operational in the early months. However, as contended by Hebb (1949) and Brunswik and Kamiya (1953), other principles may be learned through perceptual experience (Bhatt & Quinn, 2011). The data actually lend support to the type of model proposed by Spelke (1982) in which some start-up principles enable other principles to be bootstrapped onto them. As we look to the future, there are a number of aspects of the development of perceptual organization that are likely to be subject to further empirical inquiry. First, there are few studies of perceptual organization in newborns, with the majority of studies being conducted with infants 3 months of age and older. Additional work on the functionality of the principles from birth to 3 months of age has the potential to change our understanding of what competencies are part of the infant s initial endowment. Second, given evidence that the development of perceptual organization continues into adolescence (e.g., Hadad & Kimchi, 2006; Hadad, Maurer, & Lewis, 2010; Kimchi, Hadad, Behrmann, & Palmer, 2005; Kovacs, 2000; Scherf, Behrmann, Kimchi, & Luna, 2009), we need to know more about how the perceptual organizing abilities of infants are both continuous and discontinuous with those of children and young adults. A third issue centers on the mechanisms by which infants learn perceptual organization. In sections 6 and 8 of this chapter, we reviewed studies showing that variability exposure and scaffolding based on already 22