The processing of spatial frequencies through time in visual word recognition

Spatial frequencies (SF) optimal for word recognition are well established. However, other classes of stimuli demonstrate a rapid temporal evolution of SFs most useful for visual recognition. (Bar,2003,JcogNeurosci) This study aims to assess the time course of SF processing in visual word recognition. Word images were filtered like so: 1.2, 2.4, 4.8, and 9.6 cycles per degree, with 200 ms exposure duration and variyng signal-to-noise ratio (SNR; signal=image; noise=white noise) according to a random sampling function integrating sine waves with 5-55 Hz frequencies in steps of 5 Hz. Classification images (CI) of processing efficiency as a function of time assessed the processing of each SF band evolving through time. Response accuracy was near 50% correct for all conditions. The contrast level of targets, however, differed. The lowest and highest SF conditions required a higher target contrast than intermediate conditions. Time-domain CIs however, showed greatest efficiency for the highest SF condition, then the second highest, and the lowest. Further analyses of time-frequency CIs unveiled the impact of frequency content of SNR oscillations on processing efficiency, interacting with SF condition and time. Congruently, support vector machine classification performance using leave-one-out cross-validation demonstrated superior performance when relying solely on time-frequency CIs in mapping the Fourier transforms of individual CIs to SF condition. This fails to support coarse-to-fine processing for word recognition, showing one that is fine-to-coarse. Moreover, in the context of a word recognition task, frequency content of target visibility oscillations through time is a crucial factor to completely understand the phenomenon.