Visual Psychophysics


Flash lag


  • Perceived size of flashed objects influenced by flash-lag effect
    E. Bush, K. Watanabe, R. Nijhawan

Purpose: It is now well established that shape and motion are processed in parallel pathways. It is, however, still unclear to what extent these pathways interact. We have identified an illusion in which moving dots bias the perceived length of a line.

Methods: Observers fixated the center of a computer screen while two horizontal lines were flashed .8 degrees above and below the fixation point. The bottom line was 3.5 degrees long. The length of the top line could be varied, and observers were instructed to adjust its length to match that of the bottom line. On each trial, two dots moved away from each other on horizontal trajectories. The lower line was flashed on top of these trajectories. In the frame of the flash the dots were aligned with the ends of the line such that they were completely covered. There were three types of trials. In the complete cycle (CC), the moving dots appeared before the flash, and remained visible after the flash. In the flash-terminated cycle (FTC), the dots appeared before the flash and disappeared along with it. In the flash-initiated cycle (FIC), the dots first appeared in the same frame as the flash. For comparison, we measured the flash-lag effect with the top line replaced by dots at its end-points, and the bottom line eliminated.

In the dots only condition, we observed a flash-lag effect in which the bottom dots appeared farther apart than the top dots, when the physical distances were equal. Interestingly, when the lines were flashed in the presence of the outwardly moving dots, the bottom line appeared on average (n=4) 14% longer. Observers reported little or no expansion on the FTC trials, but significant, and roughly equivalent expansion on the CC and FIC trials.

The flash-lag effect is very likely one contributor to this expansion illusion. The magnitude of the line expansion effect and the flash-lag effect are comparable, and as with the flash-lag effect (Nijhawan, Nature 370,256, 1994), flash-initiated trials produce an effect, while flash-terminated trials do not. The expansion of the line suggests that the flash-lag effect can influence the final computation of shape. Thus, at some level in the processing stream, location information based on motion can interact with shape processing.

Visual aftereffects


  • Afterimages reveal multiple surface representations
    in Neon color Filling-in
    S. Shimojo, Y. Kamitani

When the missing wedge portions of "pacmen" in a Kanizsa-type illusory rectangle are filled with color, it spreads into the central portion of the array, forming a global impression of a semi-transparent colored surface (neon color filling-in). The central issue has been whether a neural representation of such filled-in surface exists, independently of the representation of the inducing stimulus.

We have found that after prolonged adaptation (>10 sec.) to this configuration, an afterimage of the filled-in surface with the opponent color is observed. Two hypotheses were considered about the underlying mechanism. (a)The Local hypothesis: the "local" afterimages of inducers are formed, which are in turn fed into the ordinary neon-color mechanism to yield filling-in and transparency. (b)The Global hypothesis: adaptation to the filled-in surface directly induces its "global" afterimage. Note that a specific and independent reprensetation of the filled-in surface is necessary in this case.

A series of experiments (N=5) provided evidence against (a), and for (b);. 1)The afterimage of the filled-in surface and those of inducers occurred independently during the time course that followed the termination of the display. There were periods when the global afterimage was visible without local afterimages of inducers. 2) Even when both the global and the local afterimages were perceived, the global one appeared opaque, rather than transparent, the latter being the case in the neon color produced by real stimulus. 3)When the luminance contrasts of inducers were manipulated, the perceived strength of the global afterimage was correlated with that of color filling-in during adaptation. 4)When the contrasts of the afterimages of inducers were simulated in real stimulus, little color filling-in was observed.

The results were confirmed with another configuration of adapting stimulus, in which vertical stripes were perceptually completed on top of horizontal inducers, owing to the neon color filling-in. We conclude, that there are separate neural representations for multiple surfaces that are perceived in displays producing neon color.


  • Adaptation to color filling-in leads to a global afterimage
    Shimojo, S., Kamitani, Y. Nishida N.

When four white pacmen with inner colored wedge portions are presented on black (Varin configuration), the color fills into the central blank area to form an impression of a semi-transparent surface on top of the white disks. We reported (SFN ‘99) that prolonged exposure to this neon-color filling-in configuration led to a global, negative afterimage of the color-filled surface.

Purpose: When the portions of inducers are presented in different temporal phases, the color filling-in during adaptation depends on the phase, while the local afterimages of the inducers are equally formed. We examined whether the global afterimage is due to direct cortical adaptation to the perceptually filled surface (Global Hypothesis) or due to filling-in caused by the local afterimages of the inducers (Local Hypothesis), by varying the temporal phase of dynamic adapting stimuli.

Methods: Various combinations of the inducer portions, i.e. the pacmen (P1,2,3,4, clockwise) and the colored wedges(W1,2,3,4), were alternated (667 ms duration each, 15 cycles) in five conditions: (1) [P all]<->[W all], (2) [P1,3,W1,3]<->[P2,4,W2,4], (3) [W all]<->[blank], (4) [P all, W all]<->[blank], (5) [P1,3,W2,4]<->[P2,4,W1,3]. Note that the total duration of each portion was identical, and that only (2) and (4) included simultaneous presence of adjacent Ps & Ws with their borders, thus led to color filling-in during adaptation. After adaptation, observers (N=6) fixated further at the FP for 20 sec and reported the strength of the global, rectangular afterimage by magnitude estimation. In some conditions, the observers monitored appearance/disappearance of local and global afterimages with buttons.

Results: The magnitude estimates were qualitatively consistent across observers: (4)>=(2)>(1)>=(5)>=(3). The monitoring results indicated that the global afterimage was visible even when the local ones were invisible.

Conclusions: Cortical processes, not just peripheral photo receptors, are the critical source of the global afterimage. During adaptation, the filled surface and the white disks are perceptually (and neurally) segregated in depth. Their cortical representations may then undergo adaptation separately, thus lead to global and local afterimages segregated in time.


  • Motion of the Surround drags objects in
    spatial memory: A motion after-aftereffect 
    B.R.Sheth, K.Watanabe, S. Shimojo

Purpose: To understand how a dynamic visual environment can systematically distort spatial memory.

Methods: I) The display consisted of a horizontally moving target (1s duration), in a field of vertically moving random dots. The target was perceived to have a vertical component of motion opposite in direction from that of the background dots (induced motion). Following target offset, the surround dots either a) continued moving in the same direction, b) reversed direction or, c) stopped moving, but remained present. As soon as the dots disappeared (2.9s later), observers (n=6) had to localize the final position of the target from memory. II) Only after the offset of a stationary target, a random dots display either began coherently drifting in one of two vertical directions, or remained motionless. When motion stopped 2.9s later, subjects had to localize the target position. III) The order of target presentation and movement of the background dots was reversed. At trial onset, a random dots display began drifting either upward or downward, or remained stationary. 2.9s later, the movement terminated, and a stationary target appeared for 1s. Upon concurrent target and background offset, observers then localized the target position. In II) and III), the periods of target presentation and background motion did not overlap at all.

Results: I) Observers’ remembered estimates of target position were influenced by the direction of motion of the surround dots, both during target presentation, and following target offset–but in opposite ways. If the background drifted upward (downward) during the time the target was present, estimates of target position were significantly below (above) the true target position. On the other hand, upward (downward) background flow following target offset, caused estimates of target position to be significantly above (below) the true target position. II) As in I, observers judged the target to be significantly above (below) its true location when the random dots display drifted up (down) following target offset. III) The error biases were opposite from those in II. Estimates of target position were significantly below (above) those in the post-target stationary background condition when the background underwent upward (downward) motion.

Conclusions: The following hypothesis can explain the seemingly contradictory data: Compared to sensory cues, internal, memory generated activity corresponding to an object is inherently less robust. Because of this susceptibility to external stimuli, induced motion switches to motion capture, just as in the case of perception (Murakami and Shimojo,1993). The surround (true motion in II or motion aftereffect in III) drags the memory of the object location with it. This hypothesis provides a general and unifying framework for understanding the relationship between motion and position in the context of spatial memory. BS was supported by a Caltech fellowship.


  • Filling-in induced by high-contrast edge adaptation
    Shinsuke Shimojo*# & Yukiyasu Kamitani*
    *Biology/CNS, Caltech, Pasadena, CA.
    #NTT Com. Sci. Lab., Atsugi, Japan.

Purposes. We found that adaptation to high-contrast edges dramatically delays detection of a test object defined by low-contrast edges at the same location. During the delay, the observer sees only the background which filled into the test region. We aimed to measure the effect as a function of adapting duration, and to determine if it should be attributed to the afterimage interfering/canceling the test, or rather to failure of edge detection due to adaptation leading to filling-in.

Method. A white disk (42.0 cd/m2, 2.8 diameter) was presented at 4.8 deg. distance from a fixation point on a gray background (17.3 cd/m2) for 2, 4, 8, or 16 sec. It was turned off briefly (93 ms) in every 500 ms because this manipulation turned out to minimize afterimage formation. Then, a slightly brighter gray disk  (18.8 cd/m2) of the same size was presented at the same. The subjects (N=5) pressed a button as long as they saw the test disk, an afterimage, or both for 10 sec. In separate control experiments, they monitored visibility of (a) afterimage on a homogeneous gray background, or (b) the test disk without adaptation. Results. (1) The total invisible time increases as a function of adaptation duration, up to 6-7 sec (P<0.02, 0 vs. 8 or 16 sec. adaptation). (2) Afterimage was visible (up to 4-5 sec with longer adaptation), yet shorter than the total invisible time in the main exp, at some or all durations in 4 of 5 observers.

Discussion. The result (1) can not be explained by afterimage (or its interference with the test) since by definition of the task, the button pressing (indicating to see something) should increase (or stay the same) when the afterimage persists. The result (2) is inconsistent with afterimage canceling the visibility of the test, which would require the afterimage longer than, or the same as the invisible time. The effect may rather be related to failure of edge detection due to peripheral and/or cortical adaptation, leading to a surface filling-in process at a higher-level boundary-based representation.


  • Multiplicative and Suppressive Effect of Sustained and Transient Edge Adaptation in Peripheral Target Detection.
    Farshad Moradi and Shinsuke Shimojo

Filling-in can be induced by high-contrast edge adaptation (Shimojo & Kamitani VSS 01), or after prolonged adaptation to a peripheral low-contrast object (Troxler 1904). Adaptation to sustained low-contrast vs. adaptation to transient high-contrast suggests synergy between contrast and edge adaptation, but the possible interactions are not well understood.

We observed that presenting a low-contrast edge for 5-10 seconds and then flashing a high-contrast edge over it could elicit the perceptual disappearance of a subsequent low-contrast edge at the same location. Neither adaptation to the low-contrast edge nor flashing the high-contrast edge alone had any significant effect. We investigated this effect using Gabor signals (2 cpd, 5 deg eccent., sd=1, mean lum. 50cd/m2, background 50cd/m2). Target (contrast=4%) followed either a) a sustained (8 sec) low (4%) contrast stationary or drifting Gabor signal (adaptation only), b) a brief (20ms) high (~100%) contrast Gabor signal (flash only), or c) adaptation followed by flash (combined condition). A random-dot mask followed the target after 1 second. The task was to identify whether the target was present or not. Subjects (n=5) failed in less than 3% of the trials in adaptation only or flash only conditions, but more than 30% in the combined condition (p<.0001).

For combined condition trials, failure of detection was more pronounced after adaptation to a drifting Gabor than a stationary one (p<.05). There was no significant difference between same or opposite contrast polarity (phase insensitivity). In other experiments we found: a) suppression is selective for orientation, and b) disappearance could be transferred to other locations. Results suggest 1) Contrast gain adjustment to transient change is processed separately from adaptation to sustained stimuli; 2) the two mechanisms interact non-linearly. Findings are compatible with non-local orientation selective cortical mechanisms presumably at the level of V1 to V4.


Pupilary response to color flicker (Pokemon)


  • Pupilary Responses to Chromatic Flicker
    R. Sayres, P. J. Drew, K. Watanabe, and S. Shimojo.

Purpose: In photosensitive subjects, flickering stimuli can induce seizures; in most patients both frequency and color seem critical to the likelihood of seizures. As the most direct gain control device for the visual system, the pupil may play a role in the cause of photosensitive epilepsy, although the role of pupilary response has not been investigated. This study sought to determine the effects of color and frequency on the pupilary response to flickering stimuli in normal subjects, while keeping the stimuli equiluminant.

Methods: 12 subjects were presented with 6 seconds of static color, followed by 6 seconds of flicker, followed by 6 seconds of static color. The static color was a fusion between the color components of flicker (e.g., purple for blue-red flicker). The following types of flicker were studied: Blue-Black, Red-Black, Green-Black, Red-Blue, Green-Blue, Red-Green, Yellow-Green, and White-Black, with frequencies of 3, 6, 12.5, 19, and 38 Hz. Stimuli were presented on a computer screen; the subjects’ left pupils were monitored with video pupillometry. Prior to experiments, subjects subjectively matched luminances of the flicker color component s against each other, as well as against a red calibrated to 4 Cd / m^2.

Results: Onset of flicker induced pupilary constrictions for all color-combinations; there was a general loss of effect at higher frequencies (1 9 and 38 Hz). Hue-modulated flicker (e.g., Blue-Red and Red-Green) induced stronger and longer-lasting constrictions than luminance-modulated flicker (e.g ., Red-Black and White-Black). Blue-Green and Blue-Red also induced stronger and longer-lasting constrictions (up to 50% of control diameter) than other luminance-modulated flicker stimuli (35% or less of control diameter).

Conclusions: Pupil response can be affected by chromatic temporal modulation, as well as by temporal luminance modulation. The parameters which induce constrictions are partially consistent with known risky parameters in photosensitive patients. Lower frequencies produce more powerful contractions than higher frequencies. Hue-modulated flicker, particularly Blue-Green and Blue-Red, produces more powerful constrictions than luminance-modulated flicker.

Short term visual spatial memory


  • Coordinate Transformations that can help or hurt accuracy
    B.R. Sheth and S. Shimojo,

In a pointing task, human observers (Os) had to estimate target position (TP), while the reliability of eye position and object cues was varied. Our past experiments showed that the presence of stable objects in the environment enhanced accuracy of estimation. However, we find here that when objects added no extra knowledge, Os still relied on them, only to degrade in accuracy. The O (n=5) had to maintain gaze on a fixation point (FP), while a target (0.4o dia.) was flashed (10 ms) in a 6o X 6o area centered 14.7o left/right of the FP. The area was either a) blank, or b) enclosed by four objects. 500 ms following target offset, the entire screen flashed (45 ms), after which the FP re-appeared at the same location on an otherwise blank screen. The O then had to point a mouse to the TP. Ideally, Os could estimate TP from identical, stable eye position cues in both conditions, and yet, accuracy was significantly worse in b) than in a). In experiment II (n=6), a target was flashed briefly in between the FP and an object. After the screen flash, in separate conditions, either the object and/or FP was displaced, or both FP and object disappeared (eyes can move freely). TP estimates were compared to a baseline in which the FP and object were re-displayed at the same locations.

Estimates in all conditions were significantly worse than the baseline, except when the FP and object disappeared after the screen flash, i.e. computation of TP in absolute, world coordinates was nearly as accurate as when reliable eye position and object cues were at hand (baseline). The visual system is, in principle, capable of accurately computing TP in absolute coordinates, yet relies on cue-based coordinate systems even when they are unreliable, reducing accuracy.

Perceptual Learning