Invariance of long-term visual priming to scale, reflection, translation, and hemisphere

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements

Visual translation tolerance refers to our capacity to recognize objects over a wide range of different retinal locations. Although translation is perhaps the simplest spatial transform that the visual system needs to cope with, the extent to which the human visual system can identify objects at previously unseen locations is unclear, with some studies reporting near complete invariance over 10 degrees and other reporting zero invariance at 4 degrees of visual angle. Similarly, there is confusion regarding the extent of translation tolerance in computational models of vision, as well as the degree of match between human and model performance. Here, we report a series of eye-tracking studies (total N = 70) demonstrating that novel objects trained at one retinal location can be recognized at high accuracy rates following translations up to 18 degrees. We also show that standard deep convolutional neural networks (DCNNs) support our findings when pretrained to classify another set of stimuli across a range of locations, or when a global average pooling (GAP) layer is added to produce larger receptive fields. Our findings provide a strong constraint for theories of human vision and help explain inconsistent findings previously reported with convolutional neural networks (CNNs).

The nature of shape constancy mechanisms as revealed by shape priming

Five shape priming experiments are reported in which the target was either a five- or six-sided line-drawn figure and participants made a speeded two-alternative forced-choice judgment about the target's number of sides. On priming trials, the target was preceded by a briefly presented smaller line figure (the prime) and performance on these trials was gauged relative to a no-prime condition. In the first two experiments, primes were rendered invisible by the presentation of a backwards visual noise mask, respectively for a short (∼40 ms) or long duration (∼93 ms). No reliable priming effects arose under masked conditions. When these experiments were repeated without the mask, participants were speeded when the prime and target were related by a rigid through-the-plane rotation but not when the prime was a nonrigid, stretched version of the target. The same pattern of priming effects arose when, in a final experiment, novel irregular shapes were used. Collectively, the data reveal the operation of shape constancy mechanisms that are particularly sensitive to shape rigidity. The findings suggest that the visual system attempts to secure a correspondence between the rapid and successive presentations of the prime and the target by matching shapes according to a rigidity constraint.

A new look on S-R associations: How S and R link

Humans can learn associations between stimuli and responses which allow for faster, more efficient behavior when the same response is required to the same stimulus in the future. This is called stimulus-response (S-R) priming. Perceptual representations are known to be modular and hierarchical, i.e. different brain areas represent different perceptual features and higher brain areas represent increasingly abstract properties of the stimulus. In this study we investigated how perceptually specific the stimulus in S-R priming is. In particular we wanted to test whether basic visual features play a role in the S-R associations. We used a novel stimulus: images of objects built from basic visual features. Participants performed a classification task on the objects. We found no significant effect on reaction times of switching vs. repeating perceptual features between presentations of the same object. This suggests that S-R associations involve a perceptually non-specific stimulus representation.

Neocortical Rebound Depolarization Enhances Visual Perception

Animals are constantly exposed to the time-varying visual world. Because visual perception is modulated by immediately prior visual experience, visual cortical neurons may register recent visual history into a specific form of offline activity and link it to later visual input. To examine how preceding visual inputs interact with upcoming information at the single neuron level, we designed a simple stimulation protocol in which a brief, orientated flashing stimulus was subsequently coupled to visual stimuli with identical or different features. Using in vivo whole-cell patch-clamp recording and functional two-photon calcium imaging from the primary visual cortex (V1) of awake mice, we discovered that a flash of sinusoidal grating per se induces an early, transient activation as well as a long-delayed reactivation in V1 neurons. This late response, which started hundreds of milliseconds after the flash and persisted for approximately 2 s, was also observed in human V1 electroencephalogram. When another drifting grating stimulus arrived during the late response, the V1 neurons exhibited a sublinear, but apparently increased response, especially to the same grating orientation. In behavioral tests of mice and humans, the flashing stimulation enhanced the detection power of the identically orientated visual stimulation only when the second stimulation was presented during the time window of the late response. Therefore, V1 late responses likely provide a neural basis for admixing temporally separated stimuli and extracting identical features in time-varying visual environments.

A study of mice and humans shows that prior activity in the visual cortex induces a long-delayed depolarization that enhances perception of subsequent visual stimuli if these are identical to the previous one, thereby extracting invariant visual features from the constantly changing visual world.

Animals are constantly exposed to a visual world that varies over time. To examine how the visual cortex integrates visual information that is temporally spaced, we monitored neuronal activity of the primary visual cortex (V1) using single- and multicell recording techniques. We discovered that a brief visual stimulus induced an early, transient activation as well as a delayed reactivation of V1 neurons in mice and humans. Notably, this reactivation of visual cortex conveyed information about stimulus orientation: presentation of a second visual stimulus during this reactivation enhanced the V1 response specifically when the orientations of the two stimuli were identical. Behavioral tests in mice and humans revealed that the ability to detect visual stimuli was also enhanced when the second stimulus was presented during the time window of V1 reactivation. Because animals extract visual information from an environment in constant change, the modulation of visual responses through cortical reactivation might be a strategy commonly used in the visual system.

Repetition priming in picture naming: sustained learning through the speeding of multiple processes

Picture naming has been used by vision researchers to study object identification, by language researchers to study word production, and by memory researchers to study implicit memory. Response times for naming repeated pictures decrease with successive repetitions. Repetition priming in picture naming involves an implicit, nonhippocampal form of memory. In this review, the processes speeded with repetition are decomposed, the time course of the effect is characterized, the factors affecting the magnitude of priming are enumerated, and possible mechanisms of priming are evaluated. Both behavioral response time and neuroimaging studies are considered. The processes that are speeded with repetition include high-level object identification and word production processes, but not low-level visual processes or articulation. Repetition priming lasts for at least several weeks and follows a typical forgetting function. The mechanism of priming is concluded to be speeded completion of the component processes of picture naming.

A cultural side effect: learning to read interferes with identity processing of familiar objects

Based on the neuronal recycling hypothesis (Dehaene and Cohen, 2007), we examined whether reading acquisition has a cost for the recognition of non-linguistic visual materials. More specifically, we checked whether the ability to discriminate between mirror images, which develops through literacy acquisition, interferes with object identity judgments, and whether interference strength varies as a function of the nature of the non-linguistic material. To these aims we presented illiterate, late literate (who learned to read at adult age), and early literate adults with an orientation-independent, identity-based same-different comparison task in which they had to respond “same” to both physically identical and mirrored or plane-rotated images of pictures of familiar objects (Experiment 1) or of geometric shapes (Experiment 2). Interference from irrelevant orientation variations was stronger with plane rotations than with mirror images, and stronger with geometric shapes than with objects. Illiterates were the only participants almost immune to mirror variations, but only for familiar objects. Thus, the process of unlearning mirror-image generalization, necessary to acquire literacy in the Latin alphabet, has a cost for a basic function of the visual ventral object recognition stream, i.e., identification of familiar objects. This demonstrates that neural recycling is not just an adaptation to multi-use but a process of at least partial exaptation.

The transfer of object learning across exemplars and their orientation is related to perceptual similarity

Recognition of objects improves after training. The exact characteristics of this visual learning process remain unclear. We examined to which extent object learning depends on the exact exemplar and orientation used during training. Participants were trained to name object pictures at as short a picture presentation time as possible. The required presentation time diminished over training. After training participants were tested with a completely new set of objects as well as with two variants of the trained object set, namely an orientation change and a change of the exact exemplar shown. Both manipulations led to a decrease in performance compared to the original picture set. Nevertheless, performance with the manipulated versions of the trained stimuli was better than performance with the completely new set, at least when only one manipulation was performed. Amount of transfer to new images of an object was related to perceptual similarity, but not to pixel overlap or to measurements of similarity in the different layers of a popular hierarchical object recognition model (HMAX). Thus, object learning generalizes only partially over changes in exemplars and orientation, which is consistent with the tuning properties of neurons in object-selective cortical regions and the role of perceptual similarity in these representations.

Including cognitive biases and distance-based rewards in a connectionist model of complex problem solving

We present a cognitive, connectionist-based model of complex problem solving that integrates cognitive biases and distance-based and environmental rewards under a temporal-difference learning mechanism. The model is tested against experimental data obtained in a well-defined and planning-intensive problem. We show that incorporating cognitive biases (symmetry and simplicity) in a temporal-difference learning rule (SARSA) increases model adequacy-the solution space explored by biased models better fits observed human solutions. While learning from explicit rewards alone is intrinsically slow, adding distance-based rewards, a measure of closeness to goal, to the learning rule significantly accelerates learning. Finally, the model correctly predicts that explicit rewards have little impact on problem solvers' ability to discover optimal solutions.

Encoding the identity and location of objects in human LOC

We are able to recognize objects independent of their location in the visual field. At the same time, we also keep track of the location of objects to orient ourselves and to interact with the environment. The lateral occipital complex (LOC) has been suggested as the prime cortical region for representation of object identity. However, the extent to which LOC also represents object location has remained debated. In this study we used high-resolution fMRI in combination with multivoxel pattern classification to investigate the cortical encoding of three object exemplars from four different categories presented in two different locations. This approach allowed us to study location-tolerant object information and object-tolerant location information in LOC, both at the level of categories and exemplars. We found evidence for both location-tolerant object information and object-tolerant location information in LOC at the level of categories and exemplars. Our results further highlight the mixing of identity and location information in the ventral visual pathway.

High-level visual object representations are constrained by position

It is widely assumed that high-level visual object representations are position-independent (or invariant). While there is sensitivity to position in high-level object-selective cortex, position and object identity are thought to be encoded independently in the population response such that position information is available across objects and object information is available across positions. Contrary to this view, we show, with both behavior and neuroimaging, that visual object representations are position-dependent (tied to limited portions of the visual field). Behaviorally, we show that the effect of priming an object was greatly reduced with any change in position (within- or between-hemifields), indicating nonoverlapping representations of the same object across different positions. Furthermore, using neuroimaging, we show that object-selective cortex is not only highly sensitive to object position but also the ability to differentiate objects based on its response is greatly reduced across different positions, consistent with the observed behavior and the receptive field properties observed in macaque object-selective neurons. Thus, even at the population level, the object information available in response of object-selective cortex is constrained by position. We conclude that even high-level visual object representations are position-dependent.

Translational and reflectional priming invariance: a retrospective

Biederman and Cooper (1991a) showed that the presentation of a briefly presented image of an object at one position in the visual field facilitated its identification, as assessed by naming speed and accuracy, several minutes later. The facilitation was unaffected by a translation or a reflection of the stimulus. A component of this priming was visual rather than basic-level conceptual or lexical in that there was less facilitation for an object with the same name (and basic-level class) but a different shape. The invariance of priming to view variables has stood up well over the years and appears to be a general phenomenon--as long as the original structural description can be readily resolved--in that it has also been observed for variations in size and orientation in depth. Although priming was unaffected by a change in position, we documented that there was explicit memory for the position (and orientation and size) of the stimulus. The existence of two forms of representation from the identical stimulus presentation--one invariant and the other dependent on view variables--poses a challenge as to what can be concluded about view invariance from single-unit activity.

The remains of the trial: goal-determined inter-trial suppression of selective attention

When an observer is searching through the environment for a target, what are the consequences of not finding a target in a given environment? We examine this issue in detail and propose that the visual system systematically tags environmental information during a search, in an effort to improve performance in future search events. Information that led to search successes is positively tagged, so as to favor future deployments of attention toward that type of information, whereas information that led to search failures is negatively tagged, so as to discourage future deployments of attention toward such failed information. To study this, we use an oddball-search task, where participants search for one item that differs from all others along one feature or belongs to a different visual category, from the other stimuli in the display. We find that when participants perform oddball-search tasks, the absence of a target delays identification of future targets containing the feature or category that was shared by all distractors in the target-absent trial. We interpret this effect as reflecting an implicit assessment of performance: target-absent trials can be viewed as processing "failures" insofar as they do not provide the visual system with the information needed to complete the task. Here, we study the goal-oriented nature of this bias in three ways. First, we show that the direction of the bias is determined by the experimental task. Second, we show that the effect is independent of the mode of presentation of stimuli: it happens with both serial and simultaneous stimuli presentation. Third, we show that, when using categorically defined oddballs as the search stimuli (find the face among houses or vice versa), the bias generalizes to unseen members of the "failed" category. Together, these findings support the idea that this inter-trial attentional biases arise from high-level, task-constrained, implicit assessments of performance, involving categorical associations between classes of stimuli and behavioral outcomes (success/failure), which are independent of attentional modality (temporal vs. spatial attention).

The role of perceptual load in object recognition

Predictions from perceptual load theory (Lavie, 1995, 2005) regarding object recognition across the same or different viewpoints were tested. Results showed that high perceptual load reduces distracter recognition levels despite always presenting distracter objects from the same view. They also showed that the levels of distracter recognition were unaffected by a change in the distracter object view under conditions of low perceptual load. These results were found both with repetition priming measures of distracter recognition and with performance on a surprise recognition memory test. The results support load theory proposals that distracter recognition critically depends on the level of perceptual load. The implications for the role of attention in object recognition theories are discussed.

Adaptation to objects in the lateral occipital complex (LOC): shape or semantics?

A change in the basic-level class when viewing a sequence of two objects produces a large release from adaptation in LOC compared to when the images are identical. Is this due to a change in semantics or shape? In an fMRI-adaptation experiment, subjects viewed a sequence of two objects and judged whether the stimuli were identical in shape. Different-shaped stimuli could be from the same or different basic-level classes, where the physical similarities of the pairs in the two conditions were equated by a model of simple cell similarity. BOLD responses in LOC for the two conditions were equivalent, and higher than that of the identical condition, indicating that LOC is sensitive to shape rather than to basic-level semantics.

Memory-prediction errors and their consequences in schizophrenia

Cognitive deficits play a central role in the onset of schizophrenia. Cognitive impairment precedes the onset of psychosis in at least a subgroup of patients, and accounts for considerable dysfunction. Yet cognitive deficits as currently measured are not significantly related to hallucinations and delusions. Part of this counterintuitive absence of a relationship may be caused by the lack of an organizing principle of cognitive impairment in schizophrenia research. We review literature suggesting that a system of memory-based prediction is central to human perception, thought and action , and forward the notion that many of the symptoms of schizophrenia are a result of a failure of this system.

Electrophysiological evidence for size invariance in masked picture repetition priming

This experiment examined invariance in object representations through measuring event-related potentials (ERPs) to pictures in a masked repetition priming paradigm. Pairs of pictures were presented where the prime was either the same size or half the size of the target object and the target was either presented in a normal orientation or was a normal sized mirror reflection of the prime object. Previous masked repetition priming studies have found a cascade of priming effect sensitive to perceptual (N190/P190) and semantic (N400) properties of the stimulus. This experiment found that both early (N190/P190 effects) and later effects (N400) were invariant to size, whereas only the N190/P190 effect was invariant to mirror reflection. The combination of a small prime and a mirror reflected target led to no significant priming effects. Taken together, the results of this set of experiments suggests that object recognition, more specifically, activating an object representation, occurs in a hierarchical fashion where overlapping perceptual information between the prime and target is necessary, although not always sufficient, to activate a higher level semantic representation.

The deleterious effect of contrast reversal on recognition is unique to faces, not objects

Faces reversed in contrast cannot be readily recognized, an effect absent in object recognition. Why? Four factors: expertise, reflectance (pigmentation), high similarity, and the need to discriminate metrically varying smooth surfaces have been offered as explanations. Observers achieved expertise on discriminating smoothly shaped, pigmented, non-face blobs with positive contrast, where distractor similarity matched that of a set of faces in shape and reflectance. On a match-to-sample task, reversal of contrast between sample and matching images had no effect when matching such blobs, but markedly degraded performance when matching faces suggesting that this effect is unique to faces.

"What" and "where" in word reading: ventral coding of written words revealed by parietal atrophy

The visual system of literate adults develops a remarkable perceptual expertise for printed words. To delineate the aspects of this competence intrinsic to the occipitotemporal "what" pathway, we studied a patient with bilateral lesions of the occipitoparietal "where" pathway. Depending on critical geometric features of the display (rotation angle, letter spacing, mirror reversal, etc.), she switched from a good performance, when her intact ventral pathway was sufficient to encode words, to severely impaired reading, when her parietal lesions prevented the use of alternative reading strategies as a result of spatial and attentional impairments. In particular, reading was disrupted (a) by rotating word by more than 50 degrees , providing an approximation of the invariance range for words encoding in the ventral pathway; (b) by separating letters with double spaces, revealing the limits of letter grouping into perceptual wholes; (c) by mirror-reversing words, showing that words escape the default mirror-invariant representation of visual objects in the ventral pathway. Moreover, because of her parietal lesions, she was unable to discriminate mirror images of common objects, although she was excellent with reversible pseudowords, confirming that the breaking of mirror symmetry was intrinsic to the occipitotemporal cortex. Thus, charting the display conditions associated with preserved or impaired performance allowed us to infer properties of word coding in the normal ventral pathway and to delineate the roles of the parietal lobes in single-word recognition.

Adaptation to asymmetrically distorted faces and its lack of effect on mirror images

Previous research showed that viewing symmetrically distorted faces for a few minutes causes undistorted faces to appear distorted in the opposite manner (face-distortion aftereffect, FDAE). Three experiments with 90 observers demonstrated that adaptation to an asymmetrically distorted face also causes FDAE, but does not affect perception of its mirror image. The results suggested the FDAE occurs at the level of visual processing where distinct neural populations respond to a non-frontal facial image and its mirror image. Unlike most aftereffects, this FDAE lasts at least 30 min. Spatial and temporal characteristics of the FDAE and its relevance to portrait drawing and painting are discussed.

Selectivity of inferior temporal neurons for realistic pictures predicted by algorithms for image database navigation

Primates have a remarkable ability to perceive, recognize, and discriminate among the plethora of people, places, and things that they see, and neural selectivity in the primate inferotemporal (IT) cortex is thought to underlie this ability. Here we investigated the relationship between neural response and perception by recording from IT neurons in monkeys while they viewed realistic images. We then compared the similarity of neural responses elicited by images to the quantitative similarity of the images. Image similarity was approximated using several algorithms, two of which were designed to search image databases for perceptually similar images. Some algorithms for image similarity correlated well with human perception, and these algorithms explained part of the stimulus selectivity of IT neurons. Images that elicited similar neural responses were ranked as more similar by these algorithms than images that elicited different neural responses, and images ranked as similar by the algorithms elicited similar responses from neurons. Neural selectivity was predicted more accurately when the reference images for algorithm similarity elicited either very strong or very weak responses from the neuron. The degree to which algorithms for image similarity were correlated with human perception was related to the degree to which algorithms explained the selectivity of IT neurons, providing support for the proposal that the selectivity of IT neurons is related to perceptual similarity of images.

Depth rotation and mirror-image reflection reduce affective preference as well as recognition memory for pictures of novel objects

In two experiments, the identification of novel 3-D objects was worse for depth-rotated and mirror-reflected views, compared with the study view in an implicit affective preference memory task, as well as in an explicit recognition memory task. In Experiment 1, recognition was worse and preference was lower when depth-rotated views of an object were paired with an unstudied object relative to trials when the study view of that object was shown. There was a similar trend for mirror-reflected views. In Experiment 2, the study view of an object was both recognized and preferred above chance when it was paired with either depth-rotated or mirror-reflected views of that object. These results suggest that view-sensitive representations of objects mediate performance in implicit, as well as explicit, memory tasks. The findings do not support the claim that separate episodic and structural description representations underlie performance in implicit and explicit memory tasks, respectively.

The interaction of shape- and location-based priming in object categorisation: Evidence for a hybrid “what+where” representation stage

The relationship between part shape and location is not well elucidated in current theories of object recognition. Here we investigated the role of shape and location of object parts on recognition, using a classification priming paradigm with novel 3D objects. In Experiment 1, the relative displacement of two parts comprising the prime gradually reduced the priming effect. In Experiment 2, presenting single-part primes in locations progressively different from those in the composite target had no effect on priming. In Experiment 3, manipulating the relative position of composite prime and target strongly affected priming. Finally, in Experiment 4 the relative displacement of single-part primes and composite targets did influence response time. Together, these findings are best interpreted in terms of a hybrid theory, according to which conjunctions of shape and location are explicitly represented at some stage of visual object processing.

Negative compatibility or object updating? A cautionary tale of mask-dependent priming

The negative compatibility effect (NCE) is the surprising result that visual targets that follow a brief prime stimulus and a mask can be identified more rapidly when they are opposite rather than identical to the prime. In a recent article in this journal, S. T. Klapp and L. B. Hinkley (2002) proposed that this reflected a competition between inhibitory unconscious processes and excitatory conscious processes. The authors of the current article report 7 experiments with results countering this theory and propose an alternative account within the framework of object substitution masking. In this account, the NCE reflects the updating of perceptual objects, including their links to responses closely associated with those objects.

Repetition priming from moving faces

Recent experiments have suggested that seeing a familiar face move provides additional dynamic information to the viewer, useful in the recognition of identity. In four experiments, repetition priming was used to investigate whether dynamic information is intrinsic to the underlying face representations. The results suggest that a moving image primes more effectively than a static image, even when the same static image is shown in the prime and the test phases (Experiment 1). Furthermore, when moving images are presented in the test phase (Experiment 2), there is an advantage for moving prime images. The most priming advantage is found with naturally moving faces, rather than with those shown in slow motion (Experiment 3). Finally, showing the same moving sequence at prime and test produced more priming than that found when different moving sequences were shown (Experiment 4). The results suggest that dynamic information is intrinsic to the face representations and that there is an advantage to viewing the same moving sequence at prime and test.

A New Standardized Set of Ecological Pictures for Experimental and Clinical Research on Visual Object Processing

A new set of 174 pictures in black-and-white, coloured and spatially filtered versions, taken from photographs of real objects belonging to different semantic categories, was realised for experimental and clinical research on visual object processing. Two samples, one of English speakers and one of Italian speakers, were tested in order to provide the normative data for each picture, in both black-and-white and coloured versions, in relation to familiarity, visual complexity and name agreement.

Perception of temporally interleaved ambiguous patterns

BACKGROUND

Continuous viewing of ambiguous patterns is characterized by wavering perception that alternates between two or more equally valid visual solutions. However, when such patterns are viewed intermittently, either by repetitive presentation or by periodic closing of the eyes, perception can become locked or "frozen" in one configuration for several minutes at a time. One aspect of this stabilization is the possible existence of a perceptual memory that persists during periods in which the ambiguous stimulus is absent. Here, we use a novel paradigm of temporally interleaved ambiguous stimuli to explore the nature of this memory, with particular regard to its potential impact on perceptual organization.

RESULTS

We found that the persistence of a perceptual configuration was robust to interposed visual patterns, and, further, that at least three ambiguous patterns, when interleaved in time, could undergo parallel, stable time courses. Then, using an interleaved presentation paradigm, we established that the occasional reversal in one pattern could be coupled with that of its interleaved counterpart, and that this coupling was a function of the structural similarity between the patterns.

CONCLUSIONS

We postulate that the stabilization observed with repetitive presentation of ambiguous patterns can be at least partially accounted for by processes that retain a recent perceptual interpretation, and we speculate that such memory may be important in natural vision. We further propose that the interleaved paradigm introduced here may be of great value to gauge aspects of stimulus similarity that appeal to particular mechanisms of perceptual organization.

Using ERPs to dissociate recollection from familiarity in picture recognition

Dual process theories posit that separate recollection and familiarity processes contribute to recognition memory. Previous research, testing recognition memory for words, indicates that event-related brain potentials (ERPs) can be used to dissociate recollection from familiarity. It has been hypothesized that the FN400 ERP old/new effect (300-500 ms) varies with stimulus familiarity, but the parietal ERP old/new effect (400-800 ms) varies with recollection. The results reported here are consistent with this hypothesis, extending it to the recognition of pictures when subjects had to discriminate between studied pictures, highly familiar lures (mirror-reversals of studied pictures), and new pictures. Furthermore, the parietal old/new effect showed significant recollection-related differences only for subjects with good behavioral discrimination between studied items and similar lures.

Multiple levels of visual object constancy revealed by event-related fMRI of repetition priming

We conducted two event-related functional magnetic resonance imaging (fMRI) experiments to investigate the neural substrates of visual object recognition in humans. We used a repetition-priming method with visual stimuli recurring at unpredictable intervals, either with the same appearance or with changes in size, viewpoint or exemplar. Lateral occipital and posterior inferior temporal cortex showed lower activity for repetitions of both real and non-sense objects; fusiform and left inferior frontal regions showed decreases for repetitions of only real objects. Repetition of different exemplars with the same name affected only the left inferior frontal cortex. Crucially, priming-induced decreases in activity of the right fusiform cortex depended on whether the three-dimensional objects were repeated with the same viewpoint, regardless of whether retinal image size changed; left fusiform decreases were independent of both viewpoint and size. These data show that dissociable subsystems in ventral visual cortex maintain distinct view-dependent and view-invariant object representations.