Decoding working memory contents from induced alpha power was largely successful in the short delay (left), but importantly, deprioritized items eventually vanished completely in the long delay (right). Work lead by Yuanyuan Weng and in collaboration with Jelmer Borst. 3/3 #neuroskyence #cogpsyc

Our task included a longer-than-usual delay period of 3 seconds, where working memory maintenance should have stabilized. Voltage decoding dropped to zero even in the short delay (left panel), and re-emerged only just before the probe in the long delay (right panel), presumably in anticipation. 2/3

We do not integrate because our eyes or brains are slow. We integrate for efficiency, to get as much information as possible at the lowest costs. Integration is also adaptive, tuned to our current needs, from the fastest perceptual events to conceptual ones. 3/3 #neuroskyence #cogpsyc #VisionScience

Surprisingly, perceptual acceleration was not constant, but varied over time. Acceleration started early at 100 msec after stimulus onset, speeding up further around 200 msec, only to slow down again around 400 msec. Cross-temporal generalization on the left, and temporal shift on the right. 2/3