Attention increases the temporal precision of conscious perception: verifying the Neural-ST Model

Abstract

What role does attention play in ensuring the temporal precision of visual perception? Behavioural studies have investigated feature selection and binding in time using fleeting sequences of stimuli in the Rapid Serial Visual Presentation (RSVP) paradigm, and found that temporal accuracy is reduced when attentional control is diminished. To reduce the efficacy of attentional deployment, these studies have employed the Attentional Blink (AB) phenomenon. In this article, we use electroencephalography (EEG) to directly investigate the temporal dynamics of conscious perception. Specifically, employing a combination of experimental analysis and neural network modelling, we test the hypothesis that the availability of attention reduces temporal jitter in the latency between a target's visual onset and its consolidation into working memory. We perform time-frequency analysis on data from an AB study to compare the EEG trials underlying the P3 ERPs (Event-related Potential) evoked by targets seen outside vs. inside the AB time window. We find visual differences in phase-sorted ERPimages and statistical differences in the variance of the P3 phase distributions. These results argue for increased variation in the latency of conscious perception during the AB. This experimental analysis is complemented by a theoretical exploration of temporal attention and target processing. Using activation traces from the Neural-ST(2) model, we generate virtual ERPs and virtual ERPimages. These are compared to their human counterparts to propose an explanation of how target consolidation in the context of the AB influences the temporal variability of selective attention. The AB provides us with a suitable phenomenon with which to investigate the interplay between attention and perception. The combination of experimental and theoretical elucidation in this article contributes to converging evidence for the notion that the AB reflects a reduction in the temporal acuity of selective attention and the timeliness of perception.

Figure 1. The ST² model.

(1) Input & extraction of types in stage one (2) Working memory tokens in stage two (3) Temporal attention from the blaster. Refer to for an extensive description of individual layers, and the neural circuits comprising the nodes in each layer. Adapted from with kind permission of MIT Press.

Figure 2. Human P3 ERPimages for targets seen outside and inside the AB.

The ERPimages are time-locked to T2 presentation. Trials are sorted by phase at the peak latency of the grand average of the T2 P3 (indicated by the dashed line). The solid line illustrates the variation in phase, and is plotted by mapping the circular range of phase values onto the linear range of time-points encompassed by the wavelet.

Figure 3. Human P3 ERPimages for seen T1s with T2 at lag 8 and at lag 3.

The ERPimages are time-locked to T1 presentation. Trials are sorted by phase at the peak latency of the grand average of the T1 P3 (indicated by the dashed line). The solid line illustrates the variation in phase, and is plotted by mapping the circular range of phase values onto the linear range of time-points encompassed by the wavelet.

Figure 4. Virtual P3 ERPimages for targets seen outside and inside the AB.

The ERPimages are time-locked to T2 presentation. Trials are sorted by 50% area latency (indicated by the solid line) within the window indicated by the dashed lines.

Figure 5. Experimental design.

Each trial began with a central fixation cross, which turned into an arrow indicating the side on which targets would be presented. This was followed by two simultaneous RSVP streams on either side of fixation. The central arrow was finally replaced by a dot or a comma.