When temporal attention interacts with expectation

Abstract

Temporal attention is voluntarily deployed at specific moments, whereas temporal expectation is deployed according to timing probabilities. When the target appears at an expected moment in a sequence, temporal attention improves performance at the attended moments, but the timing and the precision of the attentional window remain unknown. Here we independently and concurrently manipulated temporal attention-via behavioral relevance-and temporal expectation-via session-wise precision and trial-wise hazard rate-to investigate whether and how these mechanisms interact to improve perception. Our results reveal that temporal attention interacts with temporal expectation-the higher the precision, the stronger the attention benefit, but surprisingly this benefit decreased with delayed onset despite the increasing probability of stimulus appearance. When attention was suboptimally deployed to earlier than expected moments, it could not be reoriented to a later time point. These findings provide evidence that temporal attention and temporal expectation are different mechanisms, and highlight their interplay in optimizing visual performance.

Figure 1

Illustration depicts experimental manipulations of temporal attention, hazard rate and precision. The brain’s temporal processing capacity is limited, and it is a challenge to process the two brief sequential targets perfectly well. (A) Temporal Attention is prioritization of a behaviorally relevant moment, and is allocated based on the instructions to selectively process a target that briefly appears (first target, T1, depicted here as an example), and ignore the subsequent one (T2). Given an event with a temporal probability distribution (distribution of possible visual event onsets), we can characterize time relative to the most expected time point: “early”, “expected”, and “late” time points. Temporal Precision describes the inverse of the variability of temporal distribution (shown with a purple horizontal arrow overlaid on T1 temporal probability distribution). (B) Hazard Function is a function of time that characterizes the conditional probability of target appearance at every time point, which increases as time passes and the target onset is delayed. The figure illustrates the shape of the hazard function in the current study (arbitrary time, does not represent actual experimental manipulations).

Figure 2

(A) Psychophysical procedure (adapted from Denison et al., 2017 to test visual performance at the attended and unattended time points when the stimulus onset is variable. Precue was either neutral, or indicated which target’s orientation will be asked to report at the end of the trial (response cue). Observers had unlimited time to respond, and received feedback based on the accuracy of the response. (B) Temporal precision of the stimulus onset was manipulated across sessions, and ranged from certain (no variability, highest precision) to uniform (lowest precision) (adapted from Todorovic et al., 2015)). Values inside the purple bars represent the percentage of trials that the stimuli appeared at the expected moment. For the certain condition, the targets appeared at the expected moment in 100% of the trials. With decreasing precision, targets could also appear earlier or later than this expected time point. Percentage of expected trials dropped with decreasing precision, 86% of the narrow, 42% for the wide and 33% for the uniform conditions.

Figure 3

d’ and reaction time computed at the expected time point of all precision conditions. Only trials where T1 and T2 appeared at the expected time point (1400 and 1650, respectively) were included in the analysis. Error bars above each group of bars denote within-subject error rate with Morey correction^,. Diamond icons and the horizontal bars overlaid on the violin plots represent mean and median values, respectively. (A) Main effect of temporal precision on discriminability (d’). The results suggest a gradual increase in overall performance with increasing precision. (B) The interaction between cue validity and temporal precision on reaction time. The benefit of temporal attention was present across the levels, although it gradually increased with precision.

Figure 4

We analyzed visual performance with respect to temporal precision by focusing on the trials in which the targets appeared at the expected moment/middle point of each temporal distribution (T1 onset 1400 ms, and T2 onset 1650 ms), such that the stimulus timings were equivalent in those trials, and the only difference was the temporal distribution (precision) in which those trials were embedded. Diamond icons and the horizontal bars overlaid on the violin plots represent mean and median values, respectively. Significant interactions between cue validity and target were present for (A) discriminability (d’), (B) response time (RT) and (C) Balanced Integration Score (BIS) across temporal precision levels. Higher d’, BIS and faster RT emerged in the valid than neutral condition for T1, but there was either smaller (for RT) or no (for d’ and BIS) such difference for T2. (D) The differential attentional modulations were visualized by subtracting visual performance (d’ and RT) in valid trials from neutral trials. Dashed horizontal line marks 0 difference, and large deviation from 0 indicates attentional benefit.

Figure 5

Trials in which the targets appeared at the expected moment/ middle point of the temporal distributions (T1 onset 1400 ms, and T2 onset 1650 ms) were analyzed, such that the stimulus timings in these trials were the same, but the temporal precision was variable. Temporal precision is indicated with the darkness of the colors in the figures. (A) There was a significant interaction between target and temporal precision. The level of temporal precision (temporal variability of the experimental session) affected T1, but not T2 at the same expected moment. (B) There was a significant interaction between temporal precision and cue validity. Temporal attention benefits performance when target timing is precisely predictable in narrow and certain conditions, and the magnitude of benefit increases with precision.

Figure 6

Performance was analyzed across the time window that the stimuli could appear (wide 42% and uniform 33% temporal precisions). (A,B) Analyses on d’ and BIS revealed three-way significant interactions between target, stimulus onset and cue validity. Temporal attention improved performance for T1, but not T2. The improvements of temporal attention (the difference between valid and neutral at each time point) for T1 were absent after the expected moment. (C) Benefits on RT were present across the time window (all ps < 0.001), although this effect got smaller as the stimulus onset was delayed (revealed by an interaction between the stimulus onset and cue validity). (D) Temporal attention benefits on RT were present for both targets (both ps < 0.001), although was stronger for T1 than it was for T2, indicated by a significant interaction between target and cue validity. (E) Post-hoc analysis on T1 performance in neutral trials. Same data from (B) replotted to show the statistical findings.