Abstract
For cases such as stereodepth, edges, and motion direction, our percepts afford conscious access to high temporal frequency information. For example, percepts of motion direction are accurate even for gratings drifting at 20 Hz- indicating the binding of information from intervals of less than 50 msec. But other judgements, such as which spatially separated colors and orientation are presented simultaneously, are only accurate at slower than 5 Hz (Holcombe & Cavanagh 2001). Why the difference? Investigation of the perception of words should help decide between competing theories. Words are identified late in visual processing, leading one theory to predict temporal resolution will be low. Yet we have lifelong experience with processing words in rapid series, leading another theory to predict temporal resolution will be high.
METHODS: To measure the temporal resolution of binding letters into words, in each trial two four-letter strings alternated in the same location for several cycles. Observers discriminated between two pairs of letters strings which were indistinguishable at high temporal frequencies, e.g. observers discriminated (2AFC) “tank” alternating with “mope” from “tape” alternating with “monk”. For successful performance, observers had to bind the letters presented simultaneously (forming a word).
RESULTS: Temporal frequency thresholds were remarkably poor for each subject, ranging from 5 Hz (100 ms/word) to as low as 2 Hz (250 ms/word). Binding letters into strings which were not words yielded even slower thresholds.
CONCLUSIONS: The low temporal resolution of binding letters into words indicates that decades of training is not sufficient for high-resolution access to stimulus information. Instead, conscious access to high temporal frequencies may be limited to those cases where a visual mechanism turns rapidly changing information (e.g. stimulation from rapidly drifting gratings) into a constant percept (e.g. motion direction or depth).