Trained-feature specific offline learning in an orientation detection task

Abstract

It has been suggested that sleep provides additional enhancement of visual perceptual learning (VPL) acquired before sleep, termed offline performance gains. A majority of the studies that found offline performance gains of VPL used discrimination tasks including the texture discrimination task (TDT). This makes it questionable whether offline performance gains on VPL are generalized to other visual tasks. The present study examined whether a Gabor orientation detection task, which is a standard task in VPL, shows offline performance gains. In Experiment 1, we investigated whether sleep leads to offline performance gains on the task. Subjects were trained with the Gabor orientation detection task, and re-tested it after a 12-hr interval that included either nightly sleep or only wakefulness. We found that performance on the task improved to a significantly greater degree after the interval that included sleep and wakefulness than the interval including wakefulness alone. In addition, offline performance gains were specific to the trained orientation. In Experiment 2, we tested whether offline performance gains occur by a nap. Also, we tested whether spontaneous sigma activity in early visual areas during non-rapid eye movement (NREM) sleep, previously implicated in offline performance gains of TDT, was associated with offline performance gains of the task. A different group of subjects had a nap with polysomnography. The subjects were trained with the task before the nap and re-tested after the nap. The performance of the task improved significantly after the nap only on the trained orientation. Sigma activity in the trained region of early visual areas during NREM sleep was significantly larger than in the untrained region, in correlation with offline performance gains. These aspects were also found with VPL of TDT. The results of the present study demonstrate that offline performance gains are not specific to a discrimination task such as TDT, and can be generalized to other forms of VPL tasks, along with trained-feature specificity. Moreover, the present results also suggest that sigma activity in the trained region of early visual areas plays an important role in offline performance gains of VPL of detection as well as discrimination tasks.