Purchase this article with an account.
Martin Arguin, Justine Massé; The time course of novel visual object recognition.. Journal of Vision 2019;19(10):61a. doi: https://doi.org/10.1167/19.10.61a.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The novel method of random temporal sampling served to investigate the temporal progression of visual encoding/processing effectiveness in novel object recognition. 12 observers first learned to associate each of six novel synthetic 3-D objects displayed from variable viewpoints to a keyboard key serving to signal its recognition. Objects were made by attaching together two elongated blobs with variable widths, curvature, and tapering. For the recognition test, one object overlaid with white visual noise was displayed for 200 ms (1200 trials/participant) and the participant pressed its associated key. The signal-to-noise ratio (SNR) varied randomly throughout exposure duration at a rate of 120 Hz. The SNR temporal profile was generated on each trial by the integration of 5–60 Hz sinewaves (5 Hz steps) of random amplitudes and phases and contrast energy was equated across image frames. Z-scored classification images of the temporal SNR profile and of the SNR frequency content through time supporting correct performance were calculated. The average temporal classification image shows that encoding effectiveness reaches its peak at 17 ms after stimulus onset, to then decline to a minimum at 75 ms. It then increases to reach an intermediate value at the end of exposure. The average time-frequency classification image shows a peak of encoding effectiveness for 15–20 Hz SNR oscillations lasting throughout the 200 ms exposure. A lower peak occurs at 45–50 Hz, centred at 100 ms. Minimum effectiveness occurs at 35–45 Hz frequencies early after stimulus onset. This minimum persists throughout exposure duration while shifting towards lower frequencies, ending at 25–30 Hz at 200 ms. Individual temporal classification images were very variable (average of pairwise correlations: .03) whereas the time-frequency profiles were highly similar across participants (.65). The time-frequency classification images capture fundamental features of novel visual object processing that are largely shared across individuals.
This PDF is available to Subscribers Only