Perception

Young and older adults are more sensitive to deviations from normality in young than older adult faces, suggesting that the dimensions of face space are optimized for young adult faces. Here, we extend these findings to own-race faces and provide converging evidence using an attractiveness rating task. In Experiment 1, Caucasian and Chinese adults were shown own- and other-race face pairs; one member was undistorted and the other had compressed or expanded features. Participants indicated which member of each pair was more normal (a task that requires referencing a norm) and which was more expanded (a task that simply requires discrimination). Participants showed an own-race advantage in the normality task but not the discrimination task. In Experiment 2, participants rated the facial attractiveness of own- and other-race faces (Experiment 2a) or young and older adult faces (Experiment 2b). Between-rater variability in ratings of individual faces was higher for other-race and older adult faces; reduced consensus in attractiveness judgments reflects a less refined face space. Collectively, these results provide direct evidence that the dimensions of face space are optimized for own-race and young adult faces, which may underlie face race- and age-based deficits in recognition.

The visual system uses an extraretinal pursuit eye movement signal to disambiguate the perception of depth from motion parallax. Visual motion in the same direction as the pursuit is perceived nearer in depth while visual motion in the opposite direction as pursuit is perceived farther in depth. This explanation of depth sign applies to either an allocentric frame of reference centered on the fixation point or an egocentric frame of reference centered on the observer. A related problem is that of depth order when two stimuli have a common direction of motion. The first psychophysical study determined whether perception of egocentric depth order is adequately explained by a model employing an allocentric framework, especially when the motion parallax stimuli have common rather than divergent motion. A second study determined whether a reversal in perceived depth order, produced by a reduction in pursuit velocity, is also explained by this model employing this allocentric framework. The results show than an allocentric model can explain both the egocentric perception of depth order with common motion and the perceptual depth order reversal created by a reduction in pursuit velocity. We conclude that an egocentric model is not the only explanation for perceived depth order in these common motion conditions.

Reading faces for identity, character, and expression is as old as humanity but representing these states is relatively recent. From the 16th century, physiognomists classified character in terms of both facial form and represented the types graphically. Darwin distinguished between physiognomy (which concerned static features reflecting character) and expression (which was dynamic and reflected emotions). Artists represented personality, pleasure, and pain in their paintings and drawings, but the scientific study of faces was revolutionized by photography in the 19th century. Rather than relying on artistic abstractions of fleeting facial expressions, scientists photographed what the eye could not discriminate. Photography was applied first to stereoscopic portraiture (by Wheatstone) then to the study of facial expressions (by Duchenne) and to identity (by Galton and Bertillon). Photography opened new methods for investigating face perception, most markedly with Galton’s composites derived from combining aligned photographs of many sitters. In the same decade (1870s), Kühne took the process of photography as a model for the chemical action of light in the retina. These developments and their developers are described and fixed in time, but the ideas they initiated have proved impossible to stop.

In this paper, I investigate conceptual categories derived from developmental processing in a deep neural network. The similarity matrices of deep representation at each layer of neural network are computed and compared with their raw representation. While the clusters generated by raw representation stand at the basic level of abstraction, conceptual categories obtained from deep representation shows a bottom-up transition procedure. Results demonstrate a developmental course of learning from specific to general level of abstraction through learned layers of representations in a deep belief network.

Identification and discrimination of peripheral stimuli are often difficult when a few stimuli adjacent to the target are present (crowding). Our previous study showed that crowding occurs for walking direction discrimination of a biological motion stimulus. In the present study, we attempted to examine whether action congruency between the target and flankers would influence the crowding effect on biological motion stimuli. Each biological motion stimulus comprised one action (e.g., walking, throwing wastepaper, etc.) and was rotated in one of five directions around the vertical axis. In Experiment 1, observers discriminated between the directions of the target stimulus actions, which were surrounded by two flankers in the peripheral visual field. The crowding effect was stronger when the flankers performed the same action as the target and the directions differed. The congruency of action type enhanced the crowding effect in the direction-discrimination task. In Experiment 2, observers discriminated between action types of target stimuli. The crowding effect for the action-discrimination task was not modulated by the congruency of action direction. Thus, identical actions induced a larger crowding effect for action-direction discrimination, but congruent directions did not influence crowding for action-type discrimination. These results suggest that the processes involved in direction discrimination of biological motion are partially distinct from action discrimination processes.

Pitch height and pitch class are different, but strictly related, percepts of music tones. To investigate the influence of pitch height in a pitch class identification task, we systematically analyzed the errors—in terms of direction and amount—committed by a group of musicians. The aim of our study was to verify the existence of constant errors in the identification of pitch classes across consecutive octaves. Stimuli were single piano tones from the C major scale executed in two consecutive octaves. Participants showed different response patterns in the two octaves. The direction of errors revealed a constant tendency to underestimate pitch classes in the lowest octave and to overestimate pitch classes in the highest octave. Thus, pitch height showed to influence pitch class identification. We called this bias “pitch class polarization”, since the same pitch class was judged to be respectively lower and higher, depending on relatively low or high pitch height.

Precise timing and presentation of stimuli is critical in vision research, still, the limiting factor in successful recognition is often the monitor itself that is used to present the stimuli. The most widespread method is the use of monitors controlled by personal computers. Traditionally, most experiments used cathode-ray tubes but they are more and more difficult to access, and instead, liquid–crystal displays are getting more and more popular. The two types have fundamentally different working principles and limitations in displaying the stimulus.

In our experiments, the temporal precision of the stimulus presentation was in focus. We investigated whether liquid–crystal displays, which are not considered to be fit to display fast successive stimuli, can represent an alternative choice for cathode-ray tubes. We used the double flash and the flicker illusion to compare the technical capabilities of the two monitor types. These illusions not only do require a precise timing but also a very short exposure to the stimuli. At the same time, the interstimulus interval is also of extreme importance. In addition, these illusions require peripheral stimulation of the retina, which is more sensitive to the temporal aspects of the visual stimulus. On the basis of previous studies and our own psychophysical results, we suggest that liquid–crystal displays might be a good alternative for precise, frame-to-frame stimulus presentation even if parts of the stimuli are projected on the peripheral retina.