Vision Sciences Society
Fourth Annual Meeting, Sarasota, Florida. May 2004
|
|
Vision Sciences SocietyFourth Annual Meeting, Sarasota, Florida. May 2004 |
Color and motion are registered in distinct brain regions. Functional separation could impose a constraint on how fast these attributes are combined into a unified percept. We measured the temporal resolution of color-motion binding by presenting two alternating random dot patterns (red moving left vs. green moving right). Observers were asked to pair color and direction of motion. Pairing was impaired as the alternation rate increased from 1.4 Hz to 4.2 Hz, suggesting that perceptual-binding is a slow process that requires temporal integration, and that consecutive attributes interfere with each other. Performance recovered after adding either of the following surface transparency cues: 1) presenting red and green pattern concurrently, 2) superimposing irrelevant patterns (blue/gray, with vertical motion), 3) superimposing stationary patterns, or 4) increasing the alternation frequency to 8.5 Hz (which results in perceptual transparency). Transparency provides a segregation cue which prevents interference between surfaces, thus only features that belong to the same visual entity can be bound together. Next we examined if segregation precedes binding. Two stationary patterns (red, green) were superimposed. One pattern moved briefly (24 – 120 ms). Both patterns were temporarily turned gray during motion, and either returned to their original color or switched colors. Observers were asked to report the color of the moving dots. Interestingly, gray was reported in less than 10% of the trials. In 70% of the trials, observed reported the color that followed motion, regardless of the original color. Our findings imply that asynchrony in binding does not reflect different latencies, temporal references, or dynamics for color and motion. Instead, we argue that color is registered over a temporal window defined by perceptual (motion-based) segregation, which lags behind the physical onset of the motion.
Handout ( PDF 277KB), demoes , paper
Recently, fMRI studies have reported that activity in the Fusiform Face Area (FFA) is reduced or eliminated in the absence of attention. In these studies, an “attended” condition (where subjects make a behavioral report on face images) is compared to an “unattended” condition where another task is performed and the faces are behaviorally irrelevant. It is unclear, however, how activity correlates with behavioral performance (since there is no required performance for unattended faces), and thus if the observed decrease in activity is purely due to the lack of top-down attention or if it is explained by the behavioral irrelevance of the faces. Recently, using a dual-task paradigm, we showed that the attentional cost associated with face-gender discrimination is minimal. When subjects reported the gender of a peripherally presented face while performing an attentionally engaging task at fixation, their performance suffered but little compared to when attention was available. In this condition, minimal attentional resources are available to the peripheral face, but it remains behaviorally relevant. Thus we are in a position to determine how neural activity varies with behavioral performance and attentional modulation. We use fMRI to examine the effects of attention on brain activation using the dual-task paradigm. Preliminary results show that FFA activity is mostly unaffected by whether subjects perform the face task alone (peripheral-task condition) or in conjunction with the central task (dual-task condition). Interestingly, the activity is decreased in the “central-task” condition when the faces are still present but are behaviorally irrelevant. This latter result is consistent with previous reports: neural activity is affected by the behavioral relevance of stimuli. In addition, however, the fact that FFA activity is maintained in the dual-task condition indicates that attentional manipulation per se leaves both performance and neural activity unaffected.
Handout ( PDF 1,143KB, jpeg 444KB)
http://www.klab.caltech.edu/~farshadm/vss2004/
