Research ReportThe role of apparent size in building- and object-specific regions of ventral visual cortex
Research Highlights
►Small perceived size amplified LO responses to otherwise non-preferred stimuli (buildings). ►Right hemisphere LO responded preferentially to geometric shapes when they were perceived to have the size of a small object. ►PPA showed independent preferences for semantic category (buildings) and for perceived size (large). ►The perceived size manipulation generally produced statistically larger effects than did semantic category.
Introduction
Buildings and small everyday objects are two categories of images that have been shown to activate distinct regions of the brain in human neuroimaging studies. Among these are the parahippocampal place area (PPA; Epstein and Kanwisher, 1998), which has been shown to respond preferentially to scenes and buildings, and the lateral occipital area (LO; Malach et al., 1995), which has been shown to respond preferentially to everyday objects. It is often assumed that this specificity is based on the unique shape cues that define a particular category.
The PPA not only typically responds preferentially to images of many varieties of scenes (Epstein and Kanwisher, 1998, Epstein et al., 1999, Kohler et al., 2002, Goh et al., 2004, Steeves et al., 2004, Epstein et al., 2007) and also responds more to cropped images of single buildings (Maguire et al., 2001) than it does to images of household objects and faces (Epstein and Kanwisher, 1998, Epstein et al., 1999), even though many buildings and common everyday objects share a very similar global rectilinear form. A similar specificity for buildings has been observed in other ventral-stream regions including the right lingual gyrus (LG) (Gorno-Tempini and Price, 2001, Maguire et al., 2001, Rosenbaum et al., 2004) and the posterior collateral sulcus (CS) (Haxby et al., 1999, Ishai et al., 1999, Levy et al., 2001, Hasson et al., 2002, Malach et al., 2002, Avidan et al., 2003, Grill-Spector, 2003, Lerner et al., 2003, Levy et al., 2004). A related review of ventral stream stimulus specificity (Haxby et al., 2000) revealed that a region of building-specific activation was posterior to and distinct from the more scene-general PPA, even though the two may overlap in some studies.
LO is broadly tuned with respect to shape, with its preferred category consisting of any well-formed image of a concrete object (Grill-Spector et al., 1998a, Grill-Spector et al., 1998b, Grill-Spector et al., 1999, Kourtzi and Kanwisher, 2000, Lerner et al., 2001, Ferber et al., 2005). Thus, LO will respond to some degree to images of buildings, although not as much as to objects (Grill-Spector, 2003).
The reviewed neuroimaging evidence is consistent with neuropsychological studies showing dissociations between visual object agnosia and topographical or landmark agnosia. Lesions of the anterior LG have been implicated in landmark agnosia, a rare disorder that impairs the recognition of large built edifices but not of small objects. In the relatively few documented cases of this disorder, the patients were able to recognize small objects and shapes but not buildings as a whole (Whiteley and Warrington, 1978); Takahashi and Kawamura, 2002. The effects of LG lesions appear to be distinct from disorders caused by rare focal lesions to the parahippocampal gyrus (PHG). The latter lesions impair the ability to learn the navigation of new routes (Aguirre and d'Esposito, 1999, Barrash et al., 2000), although there is some evidence that PHG lesions also impair the ability to recognize scenes (Lee et al., 2005a, Lee et al., 2005b, and landmarks Takahashi and Kawamura, 2002). In contrast, patient DF who developed visual form agnosia following damage to the lateral posterior occipitotemporal cortex largely centered on area LO (Goodale et al., 1991, Milner et al., 1991, James et al., 2003) shows extremely poor perception of visual objects on the basis of their shape. However, DF does retain some ability to discriminate among scenes, particularly if those scenes have diagnostic color and texture cues (Steeves et al., 2004).
Remarkably, although the role of shape cues in determining category-specific responses in the ventral stream has been well studied, little is known about the role of size, perhaps one of the most salient differences between buildings and common objects. In fact, almost any building will be significantly larger than almost any object. Because they are larger, buildings are more likely to project a larger image on the retina than objects, which means that they occupy larger visual field eccentricities (Levy et al., 2001, Hasson et al., 2002, Levy et al., 2004) and observers may perceive them using lower spatial frequencies (Peyrin et al., 2003, Peyrin et al., 2004). But of course, in most fMRI studies, these salient differences are not evident because participants are almost always shown pictures of buildings and objects where differences in retinotopic size and spatial frequencies are minimized.
In fMRI research, the image sets used to determine stimulus-specific regions-of-interest (ROIs) in the ventral visual pathway are chosen overwhelmingly on the basis of their membership in a semantic category rather than as groups according to their values on a continuous dimension like size or shape (c.f. Tootell et al., 2008). There are practical reasons for avoiding image size manipulations in neuroimaging experiments, namely that the small bore of an fMRI scanner typically limits the visual angle of stimuli. Also, higher-level ventral stream regions show significantly greater (but not complete, e.g., Op de Beeck and Vogels, 2000) invariance across retinotopic transformations than do early visual areas and so there may have been an expectation that image size would have little effect on their activation. Early studies of region LO reported translation-, scale-, and contrast-invariant BOLD responses, and similar findings have been discussed for the PPA (MacEvoy and Epstein, 2007) and FFA (Kanwisher et al., 1999, Ewbank et al., 2005).
More recently, however, both LO and the PPA have been found to show some sensitivity to retinotopic size and translation, although on a much lower scale than is typically observed in classical occipital retinotopic fields (Grill-Spector and Malach, 2001). In fact, LO appears to respond more strongly to contralateral stimuli (Niemeier et al., 2005, Hemond et al., 2007), and two LO subregions have been shown to be retinotopically mapped with respect to both polar angle and eccentricity (Wandell et al., 2005, Larsson and Heeger, 2006, Sayres and Grill-Spector, 2008). The PPA has been found to be sensitive to the hemifield in which scene images were presented in the fMRI scanner (Schwarzlose et al., 2008) and to encode retinotopic location information more generally (Carlson et al., 2003). Two scene-selective regions of the parahippocampal gyrus have also been found to show a degree of retinotopic organization (Arcaro et al., 2009).
Sensitivity to retinotopic eccentricity has been proposed as a way to account for the apparent category specificity of higher-level visual regions (Levy et al., 2001, Hasson et al., 2002, Levy et al., 2004). Even within the relatively small range of eccentricities feasible in a typical fMRI setup (e.g., within 20°; Levy et al., 2004), it has been found that the activation of building-selective regions in the CS was modulated largely by the eccentricity of visual stimuli, with little regard to what they depicted (Levy et al., 2001, Levy et al., 2004). Specifically, larger eccentricities produced greater activation in the building-selective regions, which the authors attributed to the fact that buildings viewed in the real world extend to larger retinotopic eccentricities than smaller objects. These authors suggested that the CS building-selective region integrates information across broad regions of the visual field, requiring only low acuity and giving rise to the peripheral preference, an interpretation that was bolstered by similar experiments with amblyopic patients (Lerner et al., 2003).
Complementary to findings suggesting that higher-level areas can be sensitive to retinotopic image properties, it has been found that purely perceptual changes in size modulate fMRI activation in retinotopic primary visual cortex. Specifically, simple stimuli positioned in different parts of a pictorial corridor illusion evoked V1 activation that corresponded to appropriately larger and smaller stimulus visual angles (Murray et al., 2006). Importantly, these effects were found using 2D stimuli both with and without rich cues to pictorial depth (e.g., texture gradients); a simple background consisting of two parallel lines receding to a vanishing point yielded significant behavioral and fMRI results. An important remaining question is whether or not manipulations of perceived size would have an effect on the activity of higher-level visual areas as well.
It was this issue that we addressed in the current study; in other words, we asked whether or not the apparent category-specific activity in areas LO, PPA, and posterior collateral sulcus would be influenced by purely perceived changes in size. Specifically, we hypothesized that category-selective ROIs would respond more strongly to an item that appeared to have the size appropriate for the category in question. We predicted that object-selective LO would respond more when subjects viewed items that appeared to be the size of a small object and that building-selective PPA and posterior CS would respond more strongly when subjects viewed items that appeared to be building-sized. In this sense, our reasoning is analogous to that of previous studies on retinotopic eccentricity bias (Levy et al., 2001, Hasson et al., 2002, Levy et al., 2004). We hypothesized that the introduction of contextual cues to bias interpretation of apparent size of the target items would modulate activation in category-selective ROIs (Cate et al., 2006, Amit et al., 2008) and that such a top-down cognitive mechanism would not rely on truly illusory impressions of size.
In our study, participants viewed images of buildings and objects in sparse pictorial contexts that indicated different real-world sizes. We used 2D displays in which an item's perceived size was altered by varying its apparent relation with other items in the picture, as well as by a very simple perspective cue (two converging floor lines). These size cues were introduced to emphasize conceptual rather than low-level perceptual information about item scale. For similar reasons, and because the activation of regions with scene-specific responses is known to be influenced by the perception of material surface properties (Cant and Goodale, 2007), we avoided rendering the background with any surfaces or textures. Perceived size in the resulting displays was always linked with perceived distance; after all, the apparent size of a stimulus is inversely related to its apparent distance from the observer.
In real-world buildings and objects, category membership and size are inextricably entwined. The current study employed a 2 × 2 experimental design which included category and perceived size as crossed factors. We selected experimental stimuli from two subcategories of buildings and objects, namely garages and cameras; these stimulus classes were chosen in order to minimize differences in visual form across categories and again to emphasize conceptual over perceptual category differences. The resulting design allowed us to compare the size of the two main effects directly and to assess whether or not the effects of category and size are independent. We also included additional control conditions to see if the effects of size and category could be established in isolation and to assess whether category cues have different effects when items are presented with or without scene contexts. In all analyses, we specifically focused on functionally defined ROIs that have been linked to category-specific stimuli in past research. Thus, we used localizers to identify areas LO, PPA, and the posterior collateral sulcus.
Section snippets
Behavioral results
Each subject's mean hit rates for oddball detection in the four conditions of the 2 × 2 factorial design were calculated and submitted to a repeated-measures ANOVA with image category and apparent size/distance as independent factors. The data for the two pairs of control conditions (blank background and rectangle) were compared using paired-sample t-tests.
The ANOVA of RTs showed no significant main effect of image category (F(1,8) = 1.69, p = 0.23) nor of apparent size (F(1,8) = 2.12, p = 0.18) nor an
Discussion
This purpose of the present study was to investigate whether perceived size differences modulate category-specific responses in the ventral visual pathway. Towards this end, we manipulated the perceived size of two subcategories of objects and buildings, namely cameras and garages. We found that the BOLD signal in ventral visual pathway regions with typical category-specific responses—as revealed with our functional localizers—is indeed modulated by changes in the apparent size of
Participants
Nine healthy individuals (two females) participated in the experiment. All participants (mean age, 23.9 years; SD, 5.0; range, 18–34) were right handed, reported normal or corrected-to-normal visual acuity, gave their informed consent to participate in the study, and had no known history of neurological disorder. The participants were members of the University of Western Ontario community. The procedures and protocols for this study were approved by the Research Ethics Board for Health Sciences
Acknowledgments
This research was funded by the Canadian Institutes of Health Research (CIHR) and the Natural Sciences and Engineering Research Council of Canada (NSERC).
References (96)
- et al.
An area within human ventral cortex sensitive to “building” stimuli: evidence and implications
Neuron
(1998) - et al.
Spatial vs. object specific attention in high-order visual areas
Neuroimage
(2003) - et al.
Cortical analysis of visual context
Neuron
(2003) - et al.
Cortical mechanisms specific to explicit visual object recognition
Neuron
(2001) - et al.
Perception of the corridor illusion by baboons (Papio papio)
Behav. Brain Res.
(2002) - et al.
The neuroanatomical correlates of route learning impairment
Neuropsychologia
(2000) - et al.
Object memory and perception in the medial temporal lobe: an alternative approach
Curr. Opin. Neurobiol.
(2005) - et al.
Repetition suppression in occipital–temporal visual areas is modulated by physical rather than semantic features of objects
Neuroimage
(2008) - et al.
The parahippocampal place area: recognition, navigation, or encoding?
Neuron
(1999) - et al.
fMR-adaptation reveals a distributed representation of inanimate objects and places in human visual cortex
Neuroimage
(2005)