Sei sulla pagina 1di 8

www.sciencemag.

org/cgi/content/full/317/5839/823/DC1

Supporting Online Material for


Characterizing the Limits of Human Visual Awareness
Liqiang Huang,* Anne Treisman, Harold Pashler *To whom correspondence should be addressed. E-mail: lhuang@princeton.edu
Published 10 August 2007, Science 317, 823 (2007) DOI: 10.1126/science.1143515

This PDF file includes: Materials and Methods SOM text Fig. S1

Supporting on-line material Experiment 1. Simultaneous Access to Two Color Values or Two Location Values Observers. Twenty-six college undergraduate students participated in this project, 14 in the color task and 12 in the location task. All had normal or corrected-to-normal vision. Apparatus. Stimuli were presented on a 1024 768 color monitor. Observers viewed the displays from a distance of about 60 cm and responded using the keyboard. The program was written in Microsoft Visual Basic 6.0, and was run in Microsoft Windows XP using timing routines that were tested using the Black Box Toolkit designed for calibration of computerized stimulus and response timing procedures. Method. Sample displays and procedures are illustrated in Fig. 1.a. Four colors (highly saturated red, green, blue, and yellow) and four locations (left-top, right-top, left-bottom, right-bottom) were used. In each trial, two of the four colors and two of the four locations were randomly chosen, and the two colored squares were presented in these locations. Each trial began with a small white fixation cross presented for 400 ms in the center of the screen. After a short interval during which the observer saw a blank screen (400 ms), the stimuli were presented in one frame for the simultaneous condition and two for the successive condition (inter-frame interval from offset of one square to onset of the next square =700 ms). The exposure duration for the squares was adjusted with a staircase for each observer to achieve accuracy of about 70%. (In the color task, the exposure ranged from 28 ms to 165 ms, with an average of 74 ms. In the location task, the exposure ranged from 23 ms to 127 ms, with an average of 43 ms. All frames of the stimuli were immediately masked with four 2 2
1

checkerboards in Red, Green, Blue, and Yellow in the stimulus locations (see Fig. S1 in this supporting on-line material). The mask duration was 200 ms.

Fig. S1. A sample of the mask in Experiment 1

There were two tasks that were each tested on a different group of observers. In the color task the 14 observers were asked to respond according to whether a probe square matched either of the display stimuli in color. In the location task the 12 observers were required to respond by indicating whether a probe square shared its location with either of them. The probe square was always presented 700 ms after the offset of the last frame (i.e., the single frame containing both squares in the simultaneous condition and the second of the two frames in the successive condition) and this probe square remained present until the response. In the color task, the probe square was randomly assigned to have one of the four possible colors (thus it shared the color of one of the two presented squares in 50% of the trials) and was presented in the center of the display (thus, it was neutral in location). In the location task, the probe square was randomly presented in one of the four possible locations (i.e. it shared the location with one of the two presented squares in 50% of the trials) and was grey (i.e. neutral) in color. Observers responded by pressing one of two adjacent keys with fingers of the right hand ('j' & 'k') after all stimuli had been presented. They were asked to respond as accurately as possible (unspeeded response). A tone sounded to indicate whether the response was correct, and the next
2

trial began after a 400 ms delay. Observers performed 10 blocks of 50 trials, alternating between trials of simultaneous and successive presentations within blocks. The first block was excluded from data analysis as practice. Results. In this and the next experiment, ANOVA will be used for data analysis (One-way ANOVA for main effects, and two-way ANOVA for testing the interaction). In the color task, the average accuracy of the successive condition was significantly better than that of the simultaneous condition (successive condition 73.1% vs. simultaneous condition 65.5%, F (1, 13) = 11.04, p < 0.005). In the location task, the average accuracy in the successive condition was slightly worse than in the simultaneous condition (successive condition 73.9% vs. simultaneous condition 75.2%). The interaction was significant (F (1, 24) = 6.64, p < 0.02). In sum, these results suggest that perception of the colors of two objects is significantly improved by successive compared to simultaneous presentation whereas perception of their locations is not. Thus, multiple colors can only be consciously accessed one by one, whereas multiple locations can be accessed at the same time. Additional Analysis Beyond That Reported in the Manuscript. We observed a substantial difference between the successive and simultaneous conditions in color task and we attribute this to the fact that observers could only have access to one color value at any given moment. However, it is not clear how much decrement in performance should occur in the simultaneous condition in comparison to the successive condition, assuming this limitation. If the size of the decrement in the data is less than the theoretical prediction, then our claim would not be warranted and a more modest interpretation would be required. Therefore, a model was
3

constructed to estimate the predicted size of the difference between successive and simultaneous conditions based on the postulated limitation, in order to compare that with the size of the difference in the empirical data. Assume that performance is limited by the strength of an early sensory signal and also by the postulated limitations of conscious visual access. To simplify the question, we assume a dichotomy, whereby in a certain proportion of trials (p) the signal is strong enough to allow conscious access, and otherwise (1-p) it is not. We also assume the strength of the early sensory signal is independent for each color. In the successive condition each color is perceived in p trials therefore the accuracy should be 0.5 + p/2. In the simultaneous condition, the early sensory signal allows 0 color to be perceived in (1-p)2 trials, 1 color to be perceived in 2p(1-p) trials, and 2 colors to be perceived in p2 trials. The restriction of conscious access will apply to the last condition when the early sensory signals allow both colors to reach conscious access, the 2 colors will compete for conscious access and only 1 color can be perceived in those trials. Taken together, in the simultaneous condition, one color can be perceived in 2p(1-p)+p2 =2p- p2 trials; therefore, each of the two colors can be perceived in p- p2 /2 trials. Given that chance level is 0.5, the accuracy should be 0.5 + (p- p2/2)/2 = 0.5 + (p/2)-(p2/4). The value of p is estimated to fit the accuracies in successive and simultaneous conditions (Successive condition: 0.5+p/2=0.731; Simultaneous condition: 0.5+ (p/2)-(p2/4) = 0.655). We find the value of p that generates the least sum of square errors for the two conditions (p = 0.442). Thus, the predicted accuracy for the simultaneous and successive conditions is 67.2% (i.e. use p=0.442 in 0.5 + (p/2)-(p2/4)) and 72.1% (i.e. use p=0.442 in 0.5 + p/2) respectively. These closely resemble the actual data of
4

65.5% and 73.1%, respectively. In short, the size of the significant difference between performance in the simultaneous and successive conditions is consistent with the hypothesis that color access is truly limited to one color at a time. In the location task, we predict that performance should be the same in successive and simultaneous condition (both equal to 0.5 + p/2) and the data indeed confirmed that they are very close. Therefore, detailed modeling of this aspect of the data is unnecessary.

Experiment 2. Does Prior Information Improve Performance More for Color than for Location Judgements? Observers. There were 52 new observers, 26 for the color task and 26 for the location task, all drawn from the same pool as in Experiment 1. Apparatus The equipment and software were the same as in Experiment 1.

Method. Two squares were presented in each trial. Square 1 was either blue or yellow and it was presented either on top or at the bottom. Square 2 was either red or green and it was presented either on the left or on the right. The two squares were presented simultaneously and then rapidly masked (duration of mask = 200 ms) after a very brief exposure. The mask consisted of four 4-square checkerboards in Blue, and Yellow in the top and bottom locations and in Red and Green in the left and right locations, covering the possible locations of the stimuli. The duration of stimulus exposure was adjusted for each observer to achieve accuracy of about 70% (In the color task, the exposure ranged from 44 ms to 244 ms, with an average of 95 ms; In the

location task, the exposure ranged from 19 ms to 166 ms, with an average of 49 ms.). In each trial, one of the two squares was tested; we will call it the target square. There were two different tasks and they were tested with different observers. In the color task (26 observers), observers were required to press the key corresponding to the color of the target square. In the location task (26 observers), observers were required to press the key corresponding to the location of the target square. A probe was presented 700 ms after the offset of the display and remained present until response. The probe included a rectangle to indicate which pair was being tested, and some instructions in words (e.g. left j right k would be shown for a location test on Square 2. The responding rules were constant throughout the experiment and they were presented only to aid the observers.) There were two types of blocks. In the no information block, the probe rectangle, which indicated the pair of relevant locations and colors (i.e. left-right & red-green or top-bottom & blue-yellow), randomly varied from trial to trial. In the prior information block, the probe rectangle remained constant through an entire block, and alternated between blocks with the order counterbalanced across observers. Thus the observers could narrow down their attention in advance and disregard the other two locations or colors. This prior information narrowed down 4 possible locations to 2 possible locations, and 4 possible colors to 2 possible colors. Observers performed 10 blocks of 100 trials each. The first 2 blocks were excluded from data analysis as practice. Results. In the color task, the average accuracy in the prior information condition was

significantly higher than in the no information condition (80.3 % vs. 72.6 %, respectively, F (1, 25) = 39.34, p < 0.00001). In the location task, the average accuracy in the prior information condition was slightly higher than in the no information condition (75.4 % vs. 73.2 % respectively, F (1, 25) = 6.09, p < 0.025). The interaction was significant (prior information effect = 7.7 % in the color task and 2.2% in the location task, F (1, 50) = 12.85, p < 0.001). These results suggest that prior knowledge aids perception substantially in the color task, but only slightly in the location task. Again, there is evidently a severe difficulty in accessing two colors simultaneously, whereas there is little or no such difficulty in accessing two locations simultaneously.

Potrebbero piacerti anche