Purchase this article with an account.
Guy Snir, Yaffa Yeshurun; Temporal attention selects compound representations in a strategic manner: Evidence from the attentional blink. Journal of Vision 2016;16(12):596. doi: https://doi.org/10.1167/16.12.596.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When two targets are embedded in a rapid serial visual presentation (RSVP) the ability to report the second target (T2) is impaired if it is temporally close to the first target (T1): The attentional blink (AB). This deficit diminishes when the targets appear consecutively, even if three targets are employed. However, T1 identification deteriorates (a cost) when trailing targets are spared from the blink. The present study evaluated a modification of classic 2-stage theories of the AB that addresses both sparing and cost effects. The modified theory suggests that sequential RSVP stimuli are represented together in volatile "snapshots" formed during a 1st stage of processing. These snapshots may be selected and sent to a limited-capacity 2nd stage of processing where stimuli are individuated and consolidated. Snapshot selection is subject to strategic constraints. With multiple-target RSVP, two opposing constraints are balanced: Early snapshot selection might prevent identification of later targets, while late snapshot selection might prevent early targets individuation. These principles suggest that T1 identification would deteriorate with successful identification of consecutive targets, and that snapshot selection timing will vary with target-distractor distribution. Experiment 1 tested this prediction using RSVP with 3 targets. Three types of streams were employed: T1T2T3, T1T2DT3, and T1DT2T3 (D=distractor). Timing of snapshot selection was estimated based on T2 and/or T3 identification accuracy. As predicted, regardless of stream type, T1 identification was better when earlier snapshots were selected. Additionally, later snapshots were observed with the T1T2T3 stream compared to other streams, and with T1T2DT3 stream compared to T1DT2T3 stream. In Experiment 2, temporal expectancies were manipulated using RSVP with 2 targets. In line with the theory, with long inter-target-intervals, which afford multiple snapshots, T2 identification was better when T1 was correctly identified, and vice versa with short inter-target-intervals. Interestingly, only T1 benefited from temporal uncertainty reduction.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only