Categorical perception of relative orientation in visual object recognition

Mutual Information and Categorical Perception

Categorical perception refers to the enhancement of perceptual sensitivity near category boundaries, generally along dimensions that are informative about category membership. However, it remains unclear exactly which dimensions are treated as informative and why. This article reports a series of experiments in which subjects were asked to learn statistically defined categories in a novel, unfamiliar 2D perceptual space of shapes. Perceptual discrimination was tested before and after category learning of various features in the space, each defined by its position and orientation relative to the maximally informative dimension. The results support a remarkably simple generalization: The magnitude of improvement in perceptual discrimination of each feature is proportional to the mutual information between the feature and the category variable. This finding suggests a rational basis for categorical perception in which the precision of perceptual discrimination is tuned to the statistical structure of the environment.

Topological Relations Between Objects Are Categorically Coded

How do individuals compare images-for example, two graphs or diagrams-to identify differences between them? We argue that categorical relations between objects play a critical role. These relations divide continuous space into discrete categories, such as "above" and "below," or "containing" and "overlapping," which are remembered and compared more easily than precise metric values. These relations should lead to categorical perception, such that viewers find it easier to notice a change that crosses a category boundary (one object is now above, rather than below, another, or now contains, rather than overlaps with, another) than a change of equal magnitude that does not cross a boundary. We tested the influence of a set of topological categorical relations from the cognitive-modeling literature. In a visual same/different comparison task, viewers more accurately noticed changes that crossed relational category boundaries, compared with changes that did not cross these boundaries. The results highlight the potential of systematic exploration of the boundaries of between-object relational categories.

The activation of alternative response candidates: when do doubts kick in?

In the current study, we investigated at which moment during visual object categorization alternative interpretations are most strongly activated. According to an early activation account, we are uncertain about how to interpret the visual information early in the categorization process. This uncertainty will vanish over time and therefore, the number of possible response candidates decreases over time. According to a late activation account, the visual information is categorized quickly, but after extensive viewing alternative interpretations become more strongly activated. Therefore, the number of possible response candidates increases over time. To increase perceptual uncertainty we used morphed figures composed of a dominant and nondominant object. The similarity rating between morphed figures and their nondominant object was taken as indicator for the activation of the nondominant response candidate: high similarity indicates that the nondominant object is relatively strongly activated as an alternative response candidate. Presentation times were varied in order to distinguish between the early and late activation account. Using a Bayesian model selection approach, we found support for the late activation account, but not for the early activation account. It thus seems that in a late stage of the categorization process the influence of the nondominant response candidate is strongest.

Object naming induces reliance on orientation-independent representations during longer-term, but not short-term, visual remembering

Naming novel objects with novel count nouns changes how the objects are drawn from memory, revealing that object categorisation induces reliance on orientation-independent visual representations during longer-term remembering, but not during short-term remembering. Serial position effects integrate this finding with a more established conceptualisation of short-term and longer-term visual remembering in which the former is identified as keeping an item in mind. Adults were shown a series of four novel objects in orientations in which they would not normally be drawn from memory. When not named ("Look at this object"), the objects were drawn in the orientations in which they had been seen. When named with a novel count noun (e.g., "Look at this dax"), the final object continued to be depicted in the orientation in which it had been seen, but all other objects were depicted in an unseen but preferred (canonical) orientation, even though participants could still remember the orientations in which they had been seen. Although orientation-dependent exemplar representations appear to be more accessible than orientation-independent generic representations during short-term remembering, the reverse is the case during longer-term remembering. How the theoretical framework emerging from these observations accommodates a broader body of evidence is discussed.

Hemispheric asymmetries in categorical perception of orientation in infants and adults

Orientation CP is the faster or more accurate discrimination of two orientations from different categories (e.g., oblique1 and vertical1) compared to two orientations from the same category (e.g., oblique1 and oblique2), even when the degree of difference is equated across conditions. Here, we assess whether there are hemispheric asymmetries in this effect for adults and 5-month-old infants. Experiment 1 identified the location of the vertical-oblique category boundary. Experiment 2, using a visual search task with oriented lines found that adult search was more accurate when the target and distractors were from different orientation categories, compared to targets and distractors of an equivalent physical difference taken from the same category. This effect was stronger for targets lateralized to the left visual field (LVF) than the right visual field (RVF), indicating a right hemisphere (RH) bias in adult orientation CP. Experiment 3, replicated the RH bias using different stimuli and also investigated the impact of visual and verbal interference on the category effect. Experiment 4, using the same visual search task, found that infant search was also faster when the target and distractors were from different orientation categories than the same, yet this category effect was stronger for RVF than LVF lateralized targets, indicating a LH bias in orientation CP at 5 months. These findings are contrasted to equivalent studies on the lateralization of color CP (e.g., Gilbert, Regier, Kay, & Ivry, 2005). The implications for theories on the contribution of the left and right hemispheres of the infant and adult brain to categorical computations are discussed.

The caricature effect in drawing: evidence for the use of categorical relations when drawing abstract pictures

Five experiments were conducted to determine how novice and expert drawers represent relative size for the purposes of drawing. Participants were shown images of two-part or three-part geometric figures composed of two spatially separated shapes. In each picture there was a small but noticeable relative-size difference between the constituent shapes (one part of the picture was always 25% larger than another part). Participants later drew the pictures from memory. The results showed that novice and expert drawers consistently exaggerated the relative size relationship between the shapes in the picture when attempting to draw it from memory and when copying (the 'caricature effect'), although the effect was reduced for the experts. The results are consistent with the idea that people represent size in memory using categorical descriptors (e.g., 'smaller than', 'larger than') rather than as precise metrics. Further, the results suggest that the process of becoming a skilled drawer may involve overcoming this categorical bias.

Object categorisation, object naming, and viewpoint independence in visual remembering: evidence from young children's drawings of a novel object

A simple object-drawing task confirms a three-way association between object categorisation, viewpoint independence, and longer-term visual remembering. Young children (5- to 7-year-olds) drew a familiar object or a novel object, immediately after it had been hidden from view or on the following day. Both objects were shown from a full range of viewpoints or from just two viewpoints, from neither of which would either object normally be drawn after unrestricted viewing. When drawing from short-term memory after restricted viewing, both objects were most likely to be depicted from a seen viewpoint. When drawing from longer-term memory after restricted viewing, the novel object continued to be drawn from a seen viewpoint, but the mug was now most likely to be drawn from a preferred viewpoint from which it had not been seen. Naming the novel object with a novel count noun ("Look at this. This is a dax"), to signal that it belonged to an object category, resulted in it being drawn in the same way as the familiar object. The results concur with other evidence indicating that short-term and longer-term visual remembering are differentially associated with viewpoint-dependent representations of individual objects and viewpoint independent representations of object categories, respectively.

Categorical and coordinate relations in faces, or Fechner's law and face space instead?

E. E. Cooper and T. J. Wojan (2000) applied the categorical-coordinate relations distinction to faces on the basis of a finding that two-eyes-up versus one-eye-up distortions had opposite effects in between-class (face normality) and within-class (face identity) tasks. However, Cooper and Wojan failed to match amount of metric change between their 2 deviation types and tested only 1 deviation level. In the present study, eyeheight was shifted (e.g., both eyes up or both eyes down vs. one eye up and one eye down) parametrically (11 levels) and normality and identity ratings obtained. There was no evidence of categorical changes in perception where these would have been predicted by Cooper and Wojan's theory. In all cases, the relationship between physical and perceived distortion followed Fechner's law. Differences across distortion types in Fechner threshold (the minimum deviation altering perceived normality or identity) are explained in terms of the variability associated with different dimensions in face space.

Visual perception of orientation is categorical near vertical and continuous near horizontal

Four experiments were conducted to examine whether visual-orientation information is perceived categorically. In experiments 1 and 3, adult participants sorted oriented line stimuli into broad oblique and narrow vertical or horizontal categories. Experiments 2 and 4 showed that categorical discrimination of orientation occurred only near the vertical-oblique boundary. The data indicate that there is categorical perception near vertical and more continuous perception near horizontal. The results are relevant to the debate over whether categorical perception is derived from perceptual structure, verbal coding, or within-task learning. In addition, the asymmetrical perception of orientation around vertical and horizontal is consistent with the possibility that there may be differences in the functional significance of orientation near the two main axes.

Visual working memory for image statistics

To define the role of statistical features of images in visual working memory, we compared the ability of subjects (N=6) to identify changes in arrays of black and white checks when these changes altered some aspect of their statistical structure, versus when these changes did not. Alteration of luminance statistics or local higher-order statistics improved performance, but alteration of the degree of bilateral symmetry did not. The dependence of performance on the degree of statistical change indicated that statistical information was represented in a graded, rather than categorical, fashion.

Categorical perception of familiar objects

We report three experiments where the categorical perception of familiar, three-dimensional objects was investigated. A continuum of shape change between 15 pairs of objects was created and the images along the continuum were used as stimuli. In Experiment 1 participants were first required to discriminate pairs of images of objects that lay along the shape continuum. Then participants were asked to classify each morph-image into one of two pre-specified classes. We found evidence for categorical perception in some but not all of our object pairs. In Experiment 2 we varied the viewpoint of the objects in the discrimination task and found that effects of categorical perception generalized across changes in view. In Experiment 3 similarity ratings for each object pair were collected. These similarity scores correlated with the degree of perceptual categorization found for the object pairs. Our findings suggest that some familiar objects are perceived categorically and that categorical perception is closely tied to inter-object perceptual similarity.

Attentional coding of categorical relations in scene perception: evidence from the flicker paradigm

The purpose of the present investigation was to determine whether the positions of objects in a scene are coded relative to one another categorically (i.e., above, below, or side of; Experiment 1) and to determine whether spatial position in scene perception is coded preattentively or only under focused attention (Experiment 2). In Experiment 1, participants viewed alternating versions of a scene in which one of the objects in the scene changed its categorical relationship to the closest object in the scene, changed only its metric relationship to the closest object in a scene, or appeared and disappeared. Participants were faster at detecting changes that disrupted categorical relations than at detecting changes that disrupted only metric relations. In Experiment 2, this categorical advantage still occurred even when participants were cued to the location of the change. These results suggest that categorical spatial relations are being coded in scene perception and that attention is required in order to encode spatial relations.