The benefits and costs of temporal attention

Attending to specific moments in time improves the quality of sensory information presented at these moments. Behavioural and neural benefits of temporal orienting have been shown in several contexts, including rhythmic regularities of the presented stimuli, cues informing about the relevance of specific time windows, and evolving probabilities of events occurring over time. However, it is not clear to what extent the effects and mechanisms of temporal attention are shared with other domains of attentional selection. For example, spatial attention not only improves the processing of stimuli presented at the attended location, but also has detrimental effects on processing stimuli presented elsewhere, as compared to neutral baseline conditions. This is exactly the focus of a recent paper by Denison, Heeger and Carrasco, who investigate whether similar perceptual trade-offs characterise temporal attention.

In a series of behavioural experiments, the authors present cues that inform participants of the latency of targets about which they will be most likely prompted for a response. Specifically, participants are asked to discriminate the visual orientation of gratings presented at various latencies after the auditory cue. In the simplest scenario (Experiment 1), the cue informs (with 75% validity) whether the orientation of the first (1000 ms after the cue) or the second target (250 ms later) will need to be reported. Crucially, 20% of the trials were left neutral and contained an uninformative cue, allowing for a baseline comparison. An analysis of accuracy and reaction times showed that when participants believe they will be asked about the first target, they are better and faster at discriminating its orientation than when they can’t predict which target will be relevant, and their performance is weakest if they are invalidly cued to the second target. This result suggests that participants attending a later time window will perform worse when asked about stimuli presented before this time window. But is this effect symmetric – i.e., is task performance worse for targets presented late if participants attend an earlier time window? While the behavioural effects show a pattern consistent with such an effect, the pairwise comparisons of the three experimental conditions (valid, neutral, and invalid cues) are not significant. This is not entirely surprising, given that at the moment of report the late targets might be accessed more easily since they are more recent and have not been interrupted by other irrelevant targets.

Thus, in a second experiment, the same task was administered but this time with three consecutively presented targets. However, the results were only partly consistent with the simpler version of the experiment. For example, attentional costs for targets presented before an attended time window were marginally significant or not observed. Interestingly, while attentional benefits of cueing were not observed for the intermediate latency, targets presented at this latency showed robust costs when participants attended an earlier time window, suggesting that in some contexts temporal attention might disrupt processing of targets presented after the attended time window, perhaps similar to an attentional blink.

To address the question whether temporal attention actually improves the quality of visual representations (as previously shown for rhythmic orienting) or can be explained by other factors such as missing the unattended targets (as in attentional blink) or mistakes in which targets are reported, another version of the task was run. This time, again using two targets, participants had to reproduce the orientation of targets instead of only reporting whether they were tilted clockwise or counter-clockwise. The results of these experiment suggested that attention primarily affected the precision of sensory representation but not the rates of guesses (expected if unattended targets were completely missed) or target swaps. However, as in the first experiment, these effects were largely confined to targets presented early, with no evidence of benefits or costs at longer latencies.

Taken together, while these studies provide further evidence for behavioural benefits of temporal attention and for the first time directly address attentional costs in the unattended time windows, the results aren’t always symmetric or consistent across experiments. Some of these differences can be explained by task specifics; however, it would be interesting to see whether similar effects can be identified in a more continuous version of the task, where the influence of each consecutively presented stimulus – attended or not – on the final orientation report could be quantified using established modelling techniques. Finally, the paper by Denison et al., does provoke further questions about the possible neural implementation of the observed perceptual trade-offs – where any differences that might be observed between the neural mechanisms of temporal and spatial attention will likely transform our understanding on attentional selection.

Ryszard Auksztulewicz, Oxford Centre for Human Brain Activity

Source article: Denison RN, Heeger DJ, Carrasco M (2017) Attention flexibly trades off across points in time. Psychon Bull Rev, Jan 4. doi: 10.3758/s13423-016-1216-1.