Towards a visual recognition threshold: New instrument shows humans identify animals with only 1ms of visual exposure

Abstract

The human visual system is very adept at extracting categorical information from complex scenes with only the briefest of exposure. Here we show that information from visual scenes can be processed to the level of identification with formally unattainable, ultra-brief (1ms) presentations. This brief presentation time is afforded by a new instrument, the light-emitting diode (LED) tachistoscope, in which a liquid crystal display (LCD) is illuminated externally by a brief LED flash after LCD steady-state is reached, such that image onset and offset timing can be precisely controlled. Photographs of animals were presented with or without backgrounds for 1ms and 10ms. The results indicate that visual recognition of objects benefits from presenting them in isolation rather than with a background at smaller (1ms) durations. In both conditions, however, animals could be recognised at 1ms at least 83% of the time, possibly due to iconic memory and top-down, feedback mechanisms.

Visual information representation and rapid-scene categorization are simultaneous across cortex: An MEG study

“…When the visual system arrives at a preconscious perceptual decision about the scene category, it combines developmental priors with incoming natural image statistics (Bar, 2004;Kersten et al, 2004;Hegdé, 2008). These priors result in behavioral biases; for instance, animals are recognized faster than non-living objects (Thorpe et al, 1996;Thurgood et al, 2011) and natural versus man-made category judgments occur faster than basic-level category judgments (Loschky & Larson, 2010). A potential weakness in our study is that whereas behavior and brain activity reflect these biases, our visual decoders do not.…”

The unsteady eye: an information-processing stage, not a bug

“…Our proposal that the visual system uses oculomotor behavior to represent space in time may appear at odds with the observation that some aspects of the visual scene can be extracted even with extremely brief stimulus exposures [67, 68, 69, 70]. But in these experiments, the stimulus presentation itself produces very sharp transients—an extreme amplification of the modulations normally caused by saccades—which effectively redistribute spatial information across temporal frequencies.…”

Shapes Displayed with Durations in the Microsecond Range Do Not Obey Bloch's Law of Temporal Summation

“…Similarly, with that brief exposure respondents can distinguish with the same level of accuracy whether a scene contains an animal or a transportation vehicle (Van Rullen & Thorpe, 2001 ). Thurgood, Whitfield, and Patterson ( 2011 ) used brief flashes from a bank of white light-emitting diodes (LEDs) to backlight a liquid crystal display, making the images visible with exposures as short as 1 ms. They found that respondents could identify photos of animals with 96% accuracy when they were presented alone, and with 83% accuracy if they were embedded within a background scene.…”