Purchase this article with an account.
Alison Campbell, James W. Tanaka; Face identity processing at 33 ms and 100 ms with 4 ms of stimulus exposure. Journal of Vision 2020;20(11):1478. doi: https://doi.org/10.1167/jov.20.11.1478.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Measuring the processing time from presentation to identification has proved difficult: Reaction times are slowed by decision-making and motor response processes, and although advances in neural representational analysis provide new insight on the speed of neural response, the relationship between these representations (e.g., decoding accuracy) and behaviour is not well understood. We used a psychophysical approach to measure the threshold for conscious access to face identity at 4 stages of processing: the minimum exposure duration, access to low-level information, access to high-level category information, and access to high-level identity information. In Experiment 1, 4ms of exposure was sufficient for identification and increasing exposure did not improve performance. In subsequent experiments, targets were presented for 4ms and the time available for processing the target was constrained by backward-masking and varying SOA between 8-213ms. In Experiment 2, a diffeomorphic transformation of the target image that preserves basic perceptual properties but obliterates high-level properties was used to effectively mask and limit access to the low-level properties of the target (Stojanoski & Cusack, 2014). With this mask, we found a threshold of only 33ms for above-chance identification. Targets were then masked by unfamiliar faces (Experiment 3) and familiar faces (Experiment 4) to mask the high-level face category and face familiarity properties of the target, respectively. The main findings were that unfamiliar faces were as effective as familiar faces in masking target identity; moreover, the psychometric function observed when masking with a diffeomorphic scramble indicates a narrow window of time needed to process low-level perceptual properties, whereas the functions observed when masking with another face show a gradual accumulation of evidence. The data indicate that as little as 33ms of uninterrupted processing is required to extract the low-level properties and ~100ms to extract the high-level properties needed for conscious access to a familiar face identity.
This PDF is available to Subscribers Only