Jiro Gyoba

Datasets often circulate beyond their original research contexts, becoming benchmarks, being reformulated for new learning paradigms, and acquiring altered identities through repeated reuse. Through the case of the Japanese Female Facial Expression (JAFFE) dataset and its subsequent use in facial expression recognition and label-distribution learning, we develop the concept of dataset use-drift and show how it can obscure provenance, original design, and the meaning of annotations. We then describe how use-drifted dataset identity can lead to backward projected critique of the original resource. We argue that dataset critique should distinguish original documentation from interpretations introduced through later reuse and should meet the same standards of documentation, care, and accuracy expected of dataset construction itself.

Datasets often circulate beyond their original research contexts, becoming benchmarks, being reformulated for new learning paradigms, and acquiring altered identities through repeated reuse. Through the case of the Japanese Female Facial Expression (JAFFE) dataset and its subsequent use in facial expression recognition and label-distribution learning, we develop the concept of dataset use-drift and show how it can obscure provenance, original design, and the meaning of annotations. We then describe how use-drifted dataset identity can lead to backward projected critique of the original resource. We argue that dataset critique should distinguish original documentation from interpretations introduced through later reuse and should meet the same standards of documentation, care, and accuracy expected of dataset construction itself.

This study aimed to construct a cognitive model of facial youthfulness. First, we investigated which higher-order impressions (e.g. lively, clean, etc.) compose youthfulness on the hypothesis that youthfulness is a complex impression. Then, we investigated which lower-order impressions (e.g. wrinkles, translucency, etc.) account for the constituent impressions of youthfulness. Eight trained participants rated 27 impressions of 32 women’s facial images in their 50s. Factorial analysis and multiple regression analysis were conducted on the rating data. Our investigation revealed that youthfulness consisted of lively impression and clean impression in addition to estimated age, while lively impression was the main constituent of youthfulness. Further analyses showed that those three impressions respectively related to strong appearance of eyes, perceived translucency, and wrinkled appearance. Based on these results, we proposed a cognitive model representing that youthfulness is a complex impression composed of three higher-order impressions, which are affected by particular facial features.

We examined the influence of expansion and contraction on visual short-term memory (VSTM) using a change-detection task. In each trial, several expanding/contracting line segments with various orientations were presented in two successive displays. The orientation of objects in the second display was either identical to, or different from, that of the first display. Observers were asked to judge the presence or absence of an orientation change in successive displays. Results showed that memory performance for expanding objects was higher than for contracting objects (expansion benefit: Experiments 1A and 1B). Further experiments focused on VSTM processing (encoding, storage, and retrieval). Regarding the retrieval stage, an expansion benefit was replicated only when the direction of motion was consistent between two successive displays (Experiment 2A). A cueing stimulus enhanced the memory performance for both expanding and contracting motions and eliminated the expansion benefit (Experiment 2B). Regarding the storage stage, we found the expansion benefit occurred only for shorter blank intervals between the two displays (Experiment 3). Regarding the encoding stage, the expansion benefit was observed regardless of presentation times (Experiment 4). These results indicate the possibility that expanding and contracting motions modulate VSTM.

The word order that is easiest to understand in a language generally coincides with the word order most frequently used in that language. In Kaqchikel, however, there is a discrepancy between the two: the syntactically basic VOS incurs the least cognitive load, whereas SVO is most frequently employed. This suggests that processing load is primarily determined by grammatical processes, whereas word order selection is affected by additional conceptual factors. Thus, the agent could be conceptually more salient than other elements even for Kaqchikel speakers. This hypothesis leads us to the following expectations: (1) utterance latency should be shorter for SVO sentences than for VOS sentences; (2) Kaqchikel speakers should pay more attention to agents than to other elements during sentence production; and (3) despite these, the cognitive load during sentence production should be higher for SVO than for VOS. A Kaqchikel sentence production experiment confirmed all three expectations.

We investigated the effect of an observer’s hand postures on visual motion perception using the stream/bounce display. When two identical visual objects move across collinear horizontal trajectories toward each other in a two-dimensional display, observers perceive them as either streaming or bouncing. In our previous study, we found that when observers put their palms together just below the coincidence point of the two objects, the percentage of bouncing responses increased, mainly depending on the proprioceptive information from their own hands. However, it remains unclear if the tactile or haptic (force) information produced by the postures mostly influences the stream/bounce perception. We solved this problem by changing the tactile and haptic information on the palms of the hands. Experiment 1 showed that the promotion of bouncing perception was observed only when the posture of directly putting one’s palms together was used, while there was no effect when a brick was sandwiched between the participant’s palms. Experiment 2 demonstrated that the strength of force used when putting the palms together had no effect on increasing bounce perception. Our findings indicate that the hands-induced bounce effect derives from the tactile information produced by the direct contact between both palms.

We found that a hand posture with the palms together located just below the stream/bounce display could increase the proportion of bouncing perception. This effect, called the hands-induced bounce (HIB) effect, did not occur in the hands-cross condition or in the one-hand condition. By using rubber hands or covering the participants’ hands with a cloth, we demonstrated that the visual information of the hand shapes was not a critical factor in producing the HIB effect, whereas proprioceptive information seemed to be important. We also found that the HIB effect did not occur when the participants’ hands were far from the coincidence point, suggesting that the HIB effect might be produced within a limited spatial area around the hands.

We investigated how auditory space would be represented during linear self-motion and visually-induced self-motion (vection) (Teramoto et al., 2012, 2014). The previous studies indicated that the subjective coronal plane (SCP) was displaced during forward self-motion, while during backward motion in the case of vection. The present study investigated how auditory space perception would be altered when both visual and vestibular information were presented. A random-dot pattern simulating linear self-motion was used. At the beginning of the trial, the random-dot pattern started to move at a velocity of ±0.1 m/s. When an observer perceived vection, the velocity was changed at an acceleration of ±0.2, 0, and 0.4 m/s2. At the same time, the observer was moved forward at an acceleration of 0.2 m/s2 by using a linear-motor-driven chair. A short noise-burst was presented from a loudspeaker when the observer moved 2 m. The observers indicated the direction in which the sound was perceived relative to their coronal plane. The results showed that the direction of the SCP shift was different only when the acceleration of both visual and vestibular information was identical. These results suggest that auditory space distortion effect occurs closely related to the integration of vestibular and visual information.

Previous studies have shown that the perceived duration of visual stimuli can be strongly distorted by auditory stimuli presented simultaneously. In this study, we examine whether sounds presented separately from target visual stimuli alter the perceived duration of the target’s presentation. The participants’ task was to classify the duration of the target visual stimuli as perceived by them into four categories. Our results demonstrate that a sound presented before and after a visual target increases or decreases the perceived visual duration depending on the inter-stimulus interval between the sounds and the visual stimulus. In addition, three tones presented before and after a visual target did not increase or decrease the perceived visual duration. This indicates that auditory perceptual grouping prevents intermodal perceptual grouping, and eliminates crossmodal effects. These findings suggest that the auditory–visual integration, rather than a high arousal state caused by the presentation of the preceding sound, can induce distortions of perceived visual duration, and that inter- and intramodal perceptual grouping plays an important role in crossmodal time perception. These findings are discussed with reference to the Scalar Expectancy Theory.