Toward a Dynamical View of Object Perception

Chapter 30Toward a Dynamical View of Object Perception

Mary A. Peterson and Laura Cacciamani

In this chapter, we review our research demonstrating that object perception is adynamical, integrated process in which (a) high-level memory representations areaccessed before objects are perceived; (b) potential objects compete for perceptionand only the winners are perceived; and (c) there is no clear dividing line betweenperception and memory. Our review begins with results that led us to reject the tradi-tional serial hierarchical view of object perception as well as modern versions in theform of feedforward processing models. We then outline the accumulating evidencethat led us to favor a more dynamical, interactive model that involves feedforwardas well as feedback processing between high- and low-levels of the visual hierarchy.In our review, we highlight how our views changed over time.

30.1 Object Perception: What Happens When?

When objects are perceived, they appear clearly separated from regions of the vi-sual field immediately outside their borders; those outside regions appear to sim-ply continue behind the objects as local backgrounds, or as the space surroundingthem. Since the time of the Gestalt psychologists, a commonly held assumption wasthat objects and their grounds are segregated very early in visual processing, andthat memory representations are accessed later in time and only for objects, not forgrounds (see Fig. 30.1). The Gestalt psychologists referred to objects as “figures;”subsequently, the term “figure-ground segregation” has been used by many inves-tigators to refer to temporally early memory-free components of object perception.These scientists have assumed that figure-ground segregation occurs at low levelsin the visual hierarchy, and that object memories are located in higher levels. The

M.A. Peterson (B) · L. CacciamaniDepartment of Psychology, University of Arizona, Tucson, USAe-mail: [email protected]

S.J. Dickinson, Z. Pizlo (eds.), Shape Perception in Human and Computer Vision,Advances in Computer Vision and Pattern Recognition,DOI 10.1007/978-1-4471-5195-1_30, © Springer-Verlag London 2013

443

mailto:[email protected]

http://dx.doi.org/10.1007/978-1-4471-5195-1_30

444 M.A. Peterson and L. Cacciamani

Fig. 30.1 Schematic illustrating the traditional serial hierarchical view of object perception inwhich figure-ground segregation occurs at low levels. For the bipartite black and white display,assume that at this low-level stage the black region is determined to be figure (i.e., the object) andthe white region is determined to be the ground. On the traditional view, information pertaining tothe region perceived as figure is sent via feedforward connections (black arrows) to higher levelswhere object memories and semantics are accessed (i.e., that the perceived object is a face, andassociated semantics), whereas information pertaining to the region perceived as ground is not sentto higher levels (note the white horizontal bar illustrating truncation of high-level processing ofthe ground)

conflation of temporal order and hierarchical level is consistent with a serial hier-archical processing assumption in which visual processes are localized at a certainlevel in the visual hierarchy, processes located in lower levels are completed earlierin time than those located in higher levels, and there is no feedback from higher tolower levels. Indeed, many scientists refer to figure-ground segregation as an early,low-level stage of processing.1

There has never been any evidence that figure-ground segregation is a low-levelstage, however. Certainly, ample evidence indicates that image factors such as con-vexity, symmetry, enclosure, and small area are sufficient for figure-ground percep-tion, and that past experience is not necessary (e.g., [11, 15, 21, 23, 24, 30, 59, 73]).Such results were long taken as evidence that figure-ground perception is based onlow-level assessments of image factors and precedes access to object memories. Butthis conclusion was not warranted: evidence that past experience is not necessary

1A figure-ground segregation stage could be placed at mid- rather than low-levels in the visual hier-archy, as in other models of visual perception that remain serial feedforward models (e.g., [35, 36]).

30 Toward a Dynamical View of Object Perception 445

Fig. 30.2 Figure-ground displays used by Peterson et al. [50]. The white ground regions depict(A) (portions of) standing women, (B) portions of upside-down women, and (C) the parts of thestanding women in (A) spatially-rearranged. Reprinted with permission from Peterson, M.A., Har-vey, E.H., and Weidenbacher, H.L. (1991). Shape recognition inputs to figure-ground organization:Which route counts? Journal of Experimental Psychology: Human Perception and Performance,17, 1075–1089, American Psychological Association

for figure-ground perception does not rule out the possibility that, when present, itinfluences figure-ground perception.2 A systematic investigation of whether past ex-perience (i.e., memory) can affect figure-ground perception (i.e., object perception)was lacking.3

30.2 Systematic Tests of Whether Past Experience InfluencesFigure-Ground Perception

My colleagues and I conducted a series of experiments to investigate whether ob-ject memory influences figure-ground segregation. Using a variety of stimuli andmethods, we found strong evidence that it does (e.g., [9, 45–50, 52]). The set ofdisplays used by Peterson et al. [50] are shown in Fig. 30.2. In each display, a cen-tral black region shares a border with a surrounding white region. The black regionwas closed, symmetric, smaller than, and enclosed by the white region; these im-age properties favored the percept that the center black region was the figure, andthat the surrounding white region simply continued behind the figure as its ground.The critical manipulation concerned the white regions. In Fig. 30.2A, the left andright vertical borders shared by the black and white regions portray portions of well-known objects—women—in their typical upright orientation on the white side. Fig-

2Note that none of the image factors was shown to be necessary for figure-ground perception either.3An experiment reported by Rubin [61] suggested that past experience can influence figure assign-ment, and a subsequent experiment by Schafer & Murphy [63] made the same claim for motivation,which was based on prior experience. These initial claims were rejected because they were opento alternative interpretations (e.g., [65]; see [40] for review).


ure 30.2B is an upside-down version of Fig. 30.2A; now the well-known objects (thewomen) on the white side of the border are portrayed in an unfamiliar upside-downorientation. In Fig. 30.2C, the parts portrayed on the white side of the border are thesame familiar parts as in Fig. 30.2A, but they have been spatially rearranged so thatthe configurations they form are no longer familiar (in Fig. 30.2C, the parts arrayedfrom top to bottom are, using the part names from Fig. 30.2A: the woman’s skirt;her shoulders and arms; her feet; and her head).

Peterson et al. [50] asked observers to report about reversals of perceived figure-ground status in the displays shown in Fig. 30.2. They found that the white regionswere more likely to be perceived as figures at the borders they shared with the blackregions when they portrayed portions of familiar objects that were upright (in theirtypically-experienced orientation) rather than inverted (Fig. 30.2A vs. Fig. 30.2B).Inverting the stimuli rendered the configurations on the white side of the border un-familiar, but did not change other image features (e.g., convexity) known to affectfigure assignment. Moreover, inverted familiar configurations are not permanentlyunfamiliar; it simply takes longer for inverted versions of familiar stimuli to be rec-ognized (e.g., [16, 69]) because access to memory representations is delayed [37].Thus, the finding that familiar configurations influenced border assignment whenthe stimuli were upright and not when they were inverted showed that object memo-ries can contribute to figure-ground segregation, but only when the object memoriesare accessed rapidly. Finally, the presence of upright, familiar parts arranged in anunfamiliar configuration, as in Fig. 30.2C, did not result in more reports of the whiteregions as figure than in Fig. 30.2B, leading Peterson et al. to conclude that effectsof past experience originated at high levels of processing where familiar configura-tions were represented and not at lower levels where familiar parts are represented.Notably, Peterson et al. [50] found that informing participants of the correspondencebetween the upright and inverted displays, or between the intact and part-rearrangeddisplays, did not alter the pattern of results, indicating that these effects of past ex-perience required stimulus input; knowledge alone without fast access to memoriesof familiar objects was insufficient.

The pattern of results described above was evident both in the duration that sub-jects maintained the white regions as figures once they were perceived as figuresand also in the likelihood that the white regions were obtained as figures by rever-sal out of the black-region-as-figure percept. The latter result in particular led us toconclude that, contrary to the traditional assumption that figure-ground perceptionprecedes access to past experience, past experience exerts an influence on figure-ground perception, at least when it is accessed quickly via upright familiar con-figurations. Peterson and Gibson [9, 48] showed the same pattern of results whensubjects reported their first figure-ground percept for briefly exposed displays.

The experiments reviewed above showed that representations of familiar configu-rations at high levels in the visual hierarchy are accessed sufficiently early in time toinfluence figure-ground perception. High levels were implicated because the recep-tive fields of cells mediating these effects of object memories had to be large enoughto encompass the configuration of parts (up to ∼5° of visual angle). For convergingevidence using different stimuli and/or different methods see [45, 47, 49, 52].


Fig. 30.3 Figure-ground perception is not the only possible outcome when two contiguous re-gions share a border. Shared borders can delimit regions of a flat pattern (A) or the corners of athree-dimensional object (B). Adapted from Goldreich, D., & Peterson, M.A. (2012). A Bayesianobserver replicates convexity context effects in figure-ground perception. Seeing and Perceiving,25, 365–395, with permission from Brill and Martinus Nijhoff

30.3 Where Is Object Perception Accomplished?

The results reviewed so far showed that figure-ground perception takes input fromhigh-level representations, but they were agnostic as to where in the visual hierarchyfigure-ground segregation takes place. At first there seemed to be no reason to rejectthe idea of a low-level figure-ground computation. Seemingly consistent with thatview, Lamme and colleagues (e.g., [26, 75]) had shown that V1 neurons responddifferentially when a figure versus a ground lies in their receptive fields (cf., morerecently [74]). Therefore, we originally maintained the assumption that the compu-tations leading to figure-ground perception took place in a low level in the visualhierarchy (V1 or V2) and that contributions from past experience were mediated byfeedback from the higher-level regions implicated by the past experience effects wehad observed (see also [26, 38, 39, 49, 71, 72]).

We later moved beyond the assumption that a figure-ground computation occursat a low level in the visual hierarchy, and argued instead that figure-ground percep-tion is simply one possible outcome of perceptual organizing processes [41], onethat we now understand to span multiple hierarchical levels [1, 51]. We took thisposition because it is only sometimes the case that one of two abutting regions inthe visual field is perceived as an object (figure) and the other region is perceived asa shapeless ground [10, 17]. Other outcomes are possible; for instance, two abuttingregions can be perceived as sections of a two-dimensional pattern (Fig. 30.3A) oras two sides of a three-dimensional object, such as a cube (Fig. 30.3B). The per-ceived outcome depends on a host of factors, among them the context surroundingthe abutting regions (e.g., [10, 53, 56]).

These observations call into question the notion of a figure-ground “computa-tion” (or “process” or “stage”). As an alternative, we proposed that (1) candidateobjects that might be perceived on opposite sides of borders are processed to highlevels (e.g., object memories, semantics) in a first fast feedforward pass of pro-cessing in which properties relevant to object status (figural status) are detected;(2) these candidate objects compete to be perceived as the object (i.e., as the figure)


Fig. 30.4 In our current dynamical view of object perception, candidate objects that might be per-ceived on opposites sides of borders are processed to high levels (e.g., semantics, object memories)in a first feedforward pass of processing. (Upward pointing black and white arrows symbolize thatthis feedforward pass occurs for both the black region and the white region sharing a border in thebipartite stimulus below.) Feedback from high levels then influences processing at multiple lowerlevels (see curved, downward pointing black and white arrows). (For simplicity, the illustrationshows feedback to one low-level region.) The opposing candidate objects compete for represen-tation at many levels; the competition is inhibitory. (See the horizontal gray bars with caps onthe ends.) The candidate object that wins the competition is perceived as the object shaped bythe border (i.e., the figure). The losing candidate object is not perceived; its side of the border issimply perceived as a shapeless ground to the object. The outcome of the competition is repre-sented across high and low levels of processing. Object perception emerges from this dynamicalinteractive system

at many levels in the visual hierarchy; and (3) the competition, and ultimately theperceived outcome, is influenced by context and by dynamical interactions betweenhigh and low levels of visual processing. Object perception—which may entail theassignment of a shared border as the edge of an object on one side but not the other(figure-ground assignment)—emerges from this dynamical interactive system. (SeeFig. 30.4.) In the sections that follow, we review the evidence that led us to ourcurrent view.

30.4 Competition

We were not the first to propose that figure-ground perception entails competi-tion. Sejnowski and his colleagues [19, 64] proposed the first competitive modelof figure-ground perception. Their model included figure units separated by pairs of


edge units facing in opposite directions (e.g., edge-left and edge-right units). Theseopposing edge units inhibited each other but engaged in mutual excitation with fig-ure units lying on their preferred side. Kienker et al. [19] used focused attention asa seed to increase the activity in one set of figure units and their associated edgeunits; these edge units suppressed edge units facing in the opposite direction, whichin turn suppressed the figure units on the opposite side of the edge. The authors con-cluded that this relatively enhanced activity in the figure units on one side of an edgewas the mechanism behind figural assignment (see also [13, 14]; and more recently,[7, 22]). Vecera and O’Reilly [71, 72] extended Kienker et al.’s model to accountfor Peterson et al.’s [46–48, 50] effects of familiar configuration by using feedbackfrom high-level cells to increase the activity of the feature units lying on one side ofan edge. Although many of the competitive models reviewed above implement in-fluences on figure-ground perception from levels somewhat higher in the hierarchy,none implements semantic influences and all assume that inhibitory competitionoccurs only between low-level edge units.

Peterson, de Gelder, Rapcsak, Gerhardstein, and Bachoud-Lévi ([44], see [55],for a review; [54]) proposed instead that object perception entails inhibitory compe-tition between ensembles of object properties (i.e., familiar configuration and tradi-tional image-based factors) on opposite sides of a border rather than, or in additionto, competition between edge units or figure units. In Peterson et al.’s framework,object properties on the same side of the edge cooperate, and those on opposite sidescompete. We use the term “candidate objects” to refer to these ensembles of objectproperties on opposite sides of borders. Ceteris paribus, the more strongly cued sideof the edge wins the competition and the border is perceived as a bounding edge ofthe candidate object on that side of the border. Critically, Peterson et al. proposedthat the candidate object on the relatively weakly cued side of a border (the side thatloses the competition for object (figure) status) is inhibited. They suggested that thisinhibition accounts, in part, for the patently unshaped nature of grounds near theborder they share with perceived objects. It is important to emphasize that none ofthe other models of figure-ground perception predicted the inhibition of a candidateobject on the losing side of a border; they predicted inhibition of edge units facingin the ground direction, and enhanced neural responses to the features comprisingthe object (i.e., the figure).

In support of the competitive model proposed by Peterson et al. [44], Petersonand Skow [54] observed suppression of responses to object candidates on the per-ceived ground side of a border. Their participants classified line drawings as de-picting real-world objects or novel objects. The line drawings were preceded bybriefly-exposed silhouettes that were designed so that the inside, bounded regionwould be perceived as the figure/object: they were closed, symmetric, smaller inarea than, and enclosed by a large surrounding white backdrop (see Fig. 30.5 forsample stimuli). A portion of a familiar configuration was suggested on the outsideof half of these novel black silhouettes, but participants were unaware of these fa-miliar configurations; they perceived the outside of these “experimental silhouettes”as a shapeless ground to the novel black silhouette (because the ensemble of objectproperties favoring the inside as the object was stronger than the familiar configura-tion cue on the outside). For the other half of the novel black silhouettes, no portion


Fig. 30.5 Sample stimuli used by Peterson & Skow [54]. Subjects categorized line drawings asportraying real-world or novel objects. (Only real-world objects are shown in this figure; line draw-ings of novel objects came from the Kroll and Potter [25] set.) Line drawings were preceded bysmall, symmetric, enclosed, black silhouettes with a novel shape. For half of these novel blacksilhouettes, a portion of a familiar object was suggested on the outside; these were experimen-tal silhouettes (Exp Silh). Participants were unaware that these familiar objects were potentiallypresent. A portion of a house is suggested on the outside of the experimental silhouettes in thisfigure. For the other half of the novel black silhouettes, no familiar objects were suggested onthe outside; these were control silhouettes (Ctrl Silh). Left panel: the “Same Category” condi-tion, where the familiar object suggested in white on the outside of the experimental silhouette(here, a house) was from the same basic-level category as the line drawing that followed (also ahouse). Right panel: the “Different Category” condition, where the familiar object suggested inwhite on the outside of the experimental silhouette (again, a house) was from a different categoryfrom the subsequent line drawing (here, a duck). The same line drawings also followed control sil-houettes in both conditions. Reprinted with permission from Peterson, M.A., & Skow, E. (2008).Suppression of shape properties on the ground side of an edge: Evidence for a competitive modelof figure assignment. Journal of Experimental Psychology: Human Perception and Performance,34(2), 251–267, American Psychological Association

of a familiar object was suggested on the outside. Participants also perceived theoutside of these “control silhouettes” as a shapeless ground to the novel black sil-houette.

Despite the fact that observers were unaware of the familiar object suggested onthe outside of the experimental silhouettes, they took longer to classify line draw-ings of objects from the same basic level category as the unseen object rather thana different category. After ruling out alternative interpretations (including the inter-pretation that edge units facing toward the ground were suppressed), Peterson andSkow [54] attributed the longer response times to suppression of the candidate ob-ject that lost the competition for figural status. They suggested that the competitionfor figural status was an instance of the competition for representation investigatedby Desimone and Duncan [6] and their colleagues, described next.

Moran and Desimone [31, 32, 60] found that the response of a neuron is reducedwhen two stimuli are present in its receptive field, even when one of the stimulielicits a vigorous response when presented alone and the other elicits little or no


response when presented alone. These effects are observed only when the stim-uli lie close enough to each other to fall within the same receptive field [29, 32].Desimone and his colleagues concluded that the response reduction they observedoccurred because the two stimuli were competing for representation by the neuron.They found that the competition is resolved in favor of the stimulus that is higherin contrast or that is attended ([6, 8, 58]; see [57], for a review). For instance, if ananimal attends to one of two stimuli within a neuron’s receptive field, the neuron’sresponse pattern changes to resemble the pattern obtained when only the attendedstimulus is present (regardless of whether the attended stimulus elicits a strong ora weak response). Duncan and Desimone’s model has come to be called the BiasedCompetition Model of Attention (although note that in addition to attention, contrastcan bias the competition).

Peterson and Skow [54] considered it natural to extend the Desimone and Duncanmodel to figure-ground perception because the two objects potentially representedon opposite sides of a border necessarily lie in the same receptive field.4 Moreover,like Desimone and colleagues, they found that responses to object properties thatlose the competition are suppressed. Biased competition has been shown to occur atmany levels of the visual system (i.e., V2, V4, TE, IT) via a variety of methods (i.e.,single cell recording, event related potentials, and functional magnetic resonanceimaging [fMRI]).

30.5 High-Level Object Memories and Dynamical InteractionsBetween High and Low Levels

Peterson and Skow’s [54] results are consistent with the hypothesis that competitionfor object (i.e., figure) status can occur at high levels: In their experiment, the famil-iar configurations that lost the competition were approximately 3° in vertical extent;hence they are represented in high-level brain regions with large receptive fields.It is likely that competition for object (figure) status can occur in lower-level brainregions as well, given that simple image properties like convex parts, small area,and closure can influence figure assignment. We originally thought that familiarparts, represented in lower levels than familiar configurations, could not influencethe competition for figural status, however. This was because in tests with neuro-logically normal individuals, no effects of past experience were evident when the(familiar) parts of familiar configurations were rearranged to form a novel configura-tion: Participants were no more likely to see regions portraying such part-rearrangednovel configurations as figure than regions portraying inverted versions of familiarconfigurations (see Sect. 30.2 and Fig. 30.2C). Therefore, Peterson [42, 54] hy-pothesized that only familiar configurations and not familiar parts can affect figure

4We were not the first investigators to see a relationship between the biased competition model andfigure-ground perception (cf. [18, 70]), but previous authors neither elaborated on nor exploredtheir suggestion.


assignment. They further hypothesized that familiar configuration effects on objectperception were mediated by brain regions such as V4 with receptive fields largeenough to encompass their configurations.

A recent experiment by Barense et al. [1] caused us to change these views:Barense et al. showed (1) that a brain region higher than V4 is involved in effects ofpast experience on figure assignment and (2) that competition for figural status canbe biased by part familiarity responses. They reached these conclusions after exam-ining effects of past experience on figure assignment in patients with damage to theperirhinal cortex of the medial temporal lobe. The perirhinal cortex lies anterior tothe brain regions traditionally thought to be involved in visual perception; it has longbeen thought to be involved in declarative memory only, and not in perception (e.g.,[3, 20, 66–68]). An alternative view has emerged in the last decade—that the repre-sentations and computations in the perirhinal cortex subserve perception as well asmemory when the task-relevant stimulus set contains many complex conjunctionsconstructed from similar features (e.g., [2, 4, 12, 28, 33, 34]).

To test whether the perirhinal cortex was involved in effects of familiar configu-ration on figure-ground perception, Barense et al. asked perirhinal cortex-damagedindividuals to report the perceived figure-ground organization of displays like thosein Fig. 30.6. Each display had a critical region and a matched complementary re-gion. The critical regions depicted either intact familiar configurations (Fig. 30.6A);novel configurations created by rearranging the parts of the familiar configurations(part-rearranged novel configurations, Fig. 30.6B); or novel configurations com-posed of a novel ensemble of parts created by inverting the part-rearranged novelconfigurations (control novel configurations, Fig. 30.6C).5 In contrast to non-brain-damaged participants, perirhinal cortex-damaged individuals perceived regions de-picting part-rearranged novel configurations as figure approximately equally as of-ten as intact familiar configurations and more often than control novel configurations(see Fig. 30.6D). In other words, participants with perirhinal cortex damage showedeffects of familiar parts on figure assignment—an effect that has not been observedwith non-brain-damaged participants.

To account for their results, Barense et al. [1] hypothesized that when part-rearranged novel configurations are present and the perirhinal cortex is intact (asin the control participants), the perirhinal cortex detects the mismatch between thefamiliarity of the parts and the novelty of the configuration and suppresses part-familiarity responses in lower levels of the visual hierarchy so that the familiarityresponses at low and high levels correspond. As a consequence, part familiarity can-not exert an influence on the competition for object perception in part-rearrangednovel configurations. However, when the perirhinal cortex is damaged (as it is inthe patients), it does not distinguish between novel and familiar configurations;hence it does not inhibit low-level familiarity responses to the familiar parts in part-rearranged novel configurations. Because low-level part familiarity responses are

5Further exploration is necessary to determine whether the individual parts of the novel configu-rations are novel, but we do know that as an ensemble, the parts in the novel configurations arenovel, whereas the ensemble of parts in the part-rearranged novel configuration is familiar.


Fig. 30.6 (A)–(C) Stimuli Barense et al. [1] used to test effects of configuration familiarity onfigure assignment. Each stimulus was divided into two regions by a central border. One of the tworegions was a critical region. Here, the critical regions are shown in black on the left; black/whitecolor and left/right location of critical regions were balanced in the experiment. (A) Familiar Con-figurations. The critical regions depict portions of familiar objects (from left to right: a woman,a lamp, and a guitar). (B) Part-Rearranged Novel Configurations. The critical regions in (B) werecreated by cutting the familiar configurations in (A) into parts at minima of curvature along thecentral border and spatially rearranging them into novel configurations. The parts in (B) are thesame familiar parts as in (A), albeit in a different spatial relationship. (C) Control Novel Config-urations. The critical regions are inverted versions of those shown in (B). Inversion is known toreduce familiarity of configurations; here we use it to reduce the familiarity of the parts of a novelconfiguration. The parts are unfamiliar when inverted, because the familiar configuration fromwhich they were extracted has not often been seen in this orientation. (A, B, and C are reprintedwith permission from Barense et al. [1] Fig. 1A.) (D) The percentage of trials on which subjectsreported perceiving the critical regions as figures in the three types of stimuli. The data from twopatients with perirhinal cortex damage (RFR and JN) are shown on the right. The data from age–matched control subjects are shown on the left. Config = Configuration. Adapted from Barense,M.D., Ngo, J.K.W., Hung, L.H.T., Peterson, M.A. (2012). Interactions of memory and perceptionin amnesia: the figure-ground perspective. Cerebral Cortex, 22, 2680–2691, with permission fromOxford University Press

not suppressed, their effects on the competition for object perception are revealed infigure reports regarding the part-rearranged novel configurations. Thus, familiarityresponses at low levels can affect object perception, but tests of brain-damaged par-ticipants revealed these effects. For non-brain-damaged individuals, the effects offamiliar parts and familiar configurations cannot be distinguished because feedbackfrom the perirhinal cortex of the medial temporal lobe reduced low-level familiarityresponses to familiar parts that are arranged to form a novel configuration.

Peterson, Cacciamani, Barense, and Scalf [43] tested Barense et al.’s [1] pro-posal using fMRI in non-brain-damaged participants. They found that, for stimulipresented in the right visual field, the perirhinal cortex does distinguish between


intact familiar configurations and part-rearranged novel configurations, respondingmost strongly to the former and least strongly to the latter, with responses to controlnovel configurations in between. They also found a stronger response to the familiarparts in the low-level cortical region V2 (left hemisphere) when the familiar partswere arranged in a familiar configuration rather than a novel configuration. Petersonet al. took the differential responses to the same familiar parts in V2 as evidence offeedback from higher levels where the configurations were represented; the perirhi-nal cortex was a candidate given the similarity between its response pattern andthe pattern observed in V2. Thus, the fMRI results confirmed the feedback modelproposed by Barense et al. [1].

The results of the experiments by Barense et al. [1] and Peterson et al. [43]showed that memory representations at higher levels than previously supposed—theperirhinal cortex—are involved in effects of past experience on figure assignment.They also confirmed the hypothesis that object perception in general, and figure-ground perception in particular, is a dynamical, interactive process—one in whichthere is no clear dividing line between perception and memory.

30.6 Conclusion

Object perception seems immediate and unambiguous to human perceivers, yet it isneither. At first blush, it seems that serial processing assumptions and feedforwardmodels can account for perception, yet they cannot. Our investigation of whetherpast experience can affect object perception revealed that object perception is theresult of dynamical feedforward and feedback interactions between low- and high-level brain regions—both regions traditionally thought to be involved in perceptionand those traditionally thought to be involved in declarative memory only. A chal-lenge for the future is to unpack the entire dynamical system, to identify (1) thelevels where competition for object perception occurs (our recent results suggest itcan occur both where part familiarity and where configuration familiarity are rep-resented), (2) which levels only influence the competition (e.g., does the perirhinalcortex only influence the competition?), and (3) which levels reflect the outcomeof the competition [a candidate for the latter is V1; see [27, 62], although criticalcomputations may occur there as well (e.g., [5, 74])]. This is a fertile program ofresearch, the results of which promise to elucidate object perception.

Acknowledgements MAP acknowledges the support of NSF BCS 0960529 while writing thischapter.

References

1. Barense MD, Ngo JKW, Hung LHT, Peterson MA (2012) Interactions of memory and per-ception in amnesia: the figure-ground perspective. Cereb Cortex 22:2680–2691


2. Baxter MG (2009) Involvement of medial temporal lobe structures in memory and perception.Neuron 61:667–677

3. Clark RE, Reinagel P, Broadbent NJ, Flister ED, Squire LR (2011) Intact performanceon feature-ambiguous discriminations in rats with lesions of the perirhinal cortex. Neuron70:132–140

4. Cowell RA, Bussey TJ, Saksida LM (2010) Components of recognition memory: disso-ciable cognitive processes or just differences in representational complexity? Hippocampus20:1245–1262

5. Craft E, Schütze H, Niebur E, von der Heydt R (2007) A neural model of figure-ground orga-nization. J Neurophysiol 97:4310–4326. doi:10.1152/jn.00203.2007

6. Desimone R, Duncan J (1995) Neural mechanisms of selective visual attention. Annu RevNeurosci 18:193–222

7. Domijan D, Setic M (2008) A feedback model of figure-ground assignment. J Vis 8(10):11–27

8. Duncan J, Humphreys G, Ward R (1997) Competitive brain activity in visual attention. CurrOpin Neurobiol 7:255–261

9. Gibson BS, Peterson MA (1994) Does orientation-independent object recognition precedeorientation-dependent recognition? Evidence from a cueing paradigm. J Exp Psychol HumPercept Perform 20:299–316

10. Goldreich D, Peterson MA (2012) A Bayesian observer replicates convexity context effects infigure-ground perception. Seeing Perceiving 25:365–395

11. Göttschaldt K (1938) Gestalt factors and repetition (continued). In: Ellis WD (ed) A source-book of Gestalt psychology. Kegan Paul, London

12. Graham KS, Barense MD, Lee AC (2010) Going beyond LTM in the MTL: a synthesis ofneuropsychological and neuroimaging findings on the role of the medial temporal lobe inmemory and perception. Neuropsychologia 48:831–853

13. Grossberg S (1994) 3-d vision in figure-ground separation by visual cortex. Percept Psy-chophys 55:48–120

14. Grossberg S, Mingolla E (1985) Neural dynamics of form perception—boundary completion,illusory figures, and neon color spreading. Psychol Rev 92(2):173–211

15. Hebb DO (1949) The organization of behavior. Wiley, New York16. Jolicoeur P (1985) The time to name disoriented objects. Mem Cogn 13:289–30317. Kennedy JM (1974) A psychology of picture perception. Jossey-Bass, San Francisco18. Keysers C, Perrett DI (2002) Visual masking and RSVP reveal neural competition. Trends

Cogn Sci 6:120–12519. Kienker PK, Sejnowski TJ, Hinton GE, Schumacher LE (1986) Separating figure from ground

with a parallel network. Perception 15:197–21620. Kim S, Jeneson A, van der Horst AS, Frascino JC, Hopkins RO, Squire LR (2011) Memory,

visual discrimination performance, and the human hippocampus. J Neurosci 31:2624–262921. Koffka K (1935) Principles of Gestalt psychology. Harcourt Brace, New York22. Kogo N, Strecha C, Van Gool L, Wagemans J (2010) Surface construction by a 2-d

differentiation-integration process: a neurocomputational model for perceived border own-ership, depth, and lightness in Kanizsa figures. Psychol Rev 117:406–439

23. Köhler W (1947) Gestalt psychology. New American Library, New York. (Original work pub-lished 1929)

24. Kosslyn SM (1987) Seeing and imagining in the cerebral hemispheres: a computational ap-proach. Psychol Rev 94:148–175

25. Kroll JF, Potter MC (1984) Recognizing words, pictures, and concepts: a comparison of lexi-cal, object, and reality decisions. J Verbal Learn Verbal Behav 23:39–66

26. Lamme VAF, Rodriguez V, Spekreijse H (1999) Separate processing dynamics for textureelements, boundaries, and surfaces in primary visual cortex of the macaque monkey. CerebCortex 9:406–413

27. Likova LT, Tyler CW (2008) Occipital network for figure/ground organization. Exp Brain Res189:257–267. doi:10.1007/s00221-008-1417-6

http://dx.doi.org/10.1152/jn.00203.2007

http://dx.doi.org/10.1007/s00221-008-1417-6


28. Lee AC, Yeung LK, Barense MD (2012) The hippocampus and visual perception. Front Hu-man Neurosci 6:91

29. Luck SJ, Chelazzi L, Hillyard SA, Desimone R (1997) Neural mechanisms of spatial selectiveattention in areas V1, V2, and V4 of macaque visual cortex. J Neurophysiol 77:24–42

30. Marr D (1982) Vision. Freeman, San Francisco31. Miller EK, Gochin PM, Gross CG (1993) Suppression of visual responses of neurons in infe-

rior temporal cortex of the awake macaque by addition of a 2nd stimulus. Brain Res 616:25–29

32. Moran L, Desimone R (1985) Selective attention gates visual processing in the extrastriatecortex. Science 229:782–784

33. Murray EA, Bussey TJ, Saksida LM (2007) Visual perception and memory: a new view ofmedial temporal lobe function in primates and rodents. Annu Rev Neurosci 30:99–122

34. Murray EA, Wise SP (2012) Why is there a special issue on perirhinal cortex in a journal calledHippocampus?: The perirhinal cortex in historical perspective. Hippocampus 22(10):1941–1951

35. Nakayama K (1999) Mid-level vision. In: The MIT encyclopedia of the cognitive sciences, pp545–546

36. Nakayama K, He ZJ, Shimojo S (1995) Visual surface representation: A critical link betweenlower-level and higher-level vision. Visual cognition: An invitation to cognitive science, vol2, pp 1–70

37. Oram MW, Perrett DI (1992) Time course of neural responses discriminating different viewsof the face and head. J Neurophysiol 68:70–84

38. Peterson MA (1994) Object recognition processes can and do operate before figure-groundorganization. Curr Dir Psychol Sci 3:105–111

39. Peterson MA (1999) What’s in a stage name? J Exp Psychol Hum Percept Perform 25:276–286

40. Peterson MA (1999) Organization, segregation and object recognition. Intellectica 28:37–5141. Peterson MA (2003) On figures, grounds, and varieties of amodal surface completion. In:

Kimchi R, Behrmann M, Olson C (eds) Perceptual organization in vision: behavioral andneural perspectives. LEA, Mahwah, pp 87–116

42. Peterson MA (2003) Overlapping partial configurations in object memory: an alternative so-lution to classic problems in perception and recognition. In: Peterson MA, Rhodes G (eds)Perception of faces, objects, and scenes: analytic and holistic processes. Oxford UniversityPress, New York, pp 269–294

43. Peterson MA, Cacciamani L, Barense MD, Scalf PE (2012) The perirhinal cortex modulatesV2 activity in response to the agreement between part familiarity and configuration familiarity.Hippocampus 22(10):1965–1977

44. Peterson MA, de Gelder B, Rapcsak SZ, Gerhardstein PC, Bachoud-Lévi A (2000) Objectmemory effects on figure assignment: conscious object recognition is not necessary or suffi-cient. Vis Res 40:1549–1567

45. Peterson MA, Enns JT (2005) The edge complex: implicit perceptual memory for cross-edgecompetition leading to figure assignment. Percept Psychophys 14:727–740

46. Peterson MA, Gibson BS (1991) The initial identification of figure-ground relationships: con-tributions from shape recognition routines. Bull Psychon Soc 29:199–202

47. Peterson MA, Gibson BS (1993) Shape recognition contributions to figure-ground organiza-tion in three-dimensional displays. Cogn Psychol 25:383–429

48. Peterson MA, Gibson BS (1994) Must figure-ground organization precede object recognition?An assumption in peril. Psychol Sci 5:253–259

49. Peterson MA, Gibson BS (1994) Object recognition contributions to figure-ground organiza-tion: operations on outlines and subjective contours. Percept Psychophys 56:551–564

50. Peterson MA, Harvey EH, Weidenbacher HL (1991) Shape recognition inputs to figure-groundorganization: which route counts? J Exp Psychol Hum Percept Perform 17:1075–1089

51. Peterson MA, Kimchi R (2013) Perceptual organization. In: Reisberg D (ed) Handbook ofcognitive psychology. Oxford University Press, London, pp 9–31


52. Peterson MA, Lampignano DL (2003) Implicit memory for novel figure-ground displays in-cludes a history of border competition. J Exp Psychol Hum Percept Perform 29:808–822

53. Peterson MA, Salvagio E (2008) Inhibitory competition in figure-ground perception: contextand convexity. J Vis 8(16):4 (13pp)

54. Peterson MA, Skow E (2008) Suppression of shape properties on the ground side of an edge:evidence for a competitive model of figure assignment. J Exp Psychol Hum Percept Perform34(2):251–267

55. Peterson MA, Skow-Grant E (2003) Memory and learning in figure-ground perception. In:Ross B, Irwin D (eds) Cognitive vision: psychology of learning and motivation, vol 42. Aca-demic Press, New York, pp 1–34

56. Rauschenberger R, Peterson MA, Mosca F, Bruno N (2004) Amodal completion in visualsearch: preemption or context effects? Psychol Sci 15:351–355

57. Reynolds JH, Chelazzi L (2004) Attentional modulation of visual processing. Annu Rev Neu-rosci 27:611–647

58. Reynolds JH, Chelazzi L, Desimone R (1999) Competitive mechanisms subserve attention inmacaque areas V2 and V4. J Neurosci 19:1736–1753

59. Rock I (1962) A neglected aspect of the problem of recall: the Hoffding function. In: ScherJM (ed) Theories of the mind. Free Press, New York

60. Rolls ET, Tovee MJ (1995) The responses of single neurons in the temporal visual corticalareas of the macaque when more than one stimulus is present in the receptive-field. Exp BrainRes 103:409–420

61. Rubin E (1958) Figure and ground. In: Beardslee D, Wertheimer M (eds & trans) Readings inperception. Van Nostrand, Princeton, pp 35–101. (Original work published 1915)

62. Salvagio E, Cacciamani L, Peterson MA (2012) Competition-strength-dependent ground sup-pression in figure-ground perception. Atten Percept Psychophys 74:964–978

63. Schafer R, Murphy G (1943) The role of autism in a visual figure-ground relationship. J ExpPsychol 32:335–343

64. Sejnowski TJ, Hinton GE (1987) Separating figure from ground with a Boltzmann machine.In: Arbib MA, Hanson AR (eds) Vision, brain and cooperative computation. MIT Press, Cam-bridge, pp 703–724

65. Smith DEP, Hochberg J (1954) The effect of “punishment” (electric shock) on figure-groundperception. J Psychol 38:83–87

66. Squire LR, Zola-Morgan S (1991) The medial temporal lobe memory system. Science253:1380–1386

67. Squire LR, Wixted JT (2011) The cognitive neuroscience of human memory since H.M. AnnuRev Neurosci 34:259–288

68. Suzuki WA (2009) Perception and the medial temporal lobe: evaluating the current evidence.Neuron 61:657–666

69. Tarr MJ, Pinker S (1990) When does human object recognition use a viewer-centered referenceframe? Psychol Sci 1:253–256

70. Vecera SP (2000) Toward a biased competition account of object-based segregation and atten-tion. Brain Mind 1:353–384

71. Vecera SP, O’Reilly RC (1998) Figure-ground organization and object recognition processes:an interactive account. J Exp Psychol Hum Percept Perform 24:441–462

72. Vecera SP, O’Reilly RC (2000) Graded effects in hierarchical figure-ground organization:reply to Peterson 1999. J Exp Psychol Hum Percept Perform 26:1221–1231

73. Wallach H (1949) Some considerations concerning the relationship between perception andcognition. J Pers 18:6–13

74. Zhou H, Friedman HS, von der Heydt R (2000) Coding of border ownership in monkey visualcortex. J Neurosci 20:6594–6611

75. Zipser K, Lamme VAF, Schiller PH (1996) Contextual modulation in primary visual cortex. JNeurosci 16(22):7376–7389

Toward a Dynamical View of Object Perception

Documents

Transcript of Toward a Dynamical View of Object Perception