Supplementary Figure 6: Timing analyses Experiment 3.

The scale used for the rating responses was presented immediately after the food image had disappeared; participants (n = 24) were instructed to then enter their rating as fast as possible. The left panel shows the rating response time distribution across all subjects and trials for short (salmon, 0.9 s) and long (purple, 2.6 s) stimulus presentation times. The mean response times for high and low stimulus presentation times were 1.53±0.45 and 1.39±0.4 ms respectively. The small difference between these RTs (0.13 s) was statistically significant (β=0.15±0.04, P<0.001). However, the effective sampling time (image presentation time + response time) was 1.53±0.4 s longer for long exposure times (β=1.5±0.08, P<0.001). Moreover, please recall that several aspects of our design make it very unlikely that participants could control the rate of information sampling to match the presentation times. That is, participants (1) did not know that the presentation time would differ between different food images, (2) did not have advance information how long any given food image would be present on the screen, and (3) were unaware that a second round of ratings with inverted presentation times would take place. Our experimental approach and empirical results therefore support the assumption that participants were able to draw more samples (for example from memory) in the long-presentation-time condition.