We examined the role of non-rapid eye movement (NREM) and REM sleep for consolidation of associative face memory. Subjects (N=96) were trained to associate 20 pairs of bilaterally presented faces. At 9AM training, subjects were instructed to associate each pair by imagining family relationship, without mention of face position. At immediate test and again at 4PM, subjects were presented with pairs of faces (10 old and 10 new pairs) and asked whether the faces were in the same or different pairings as training. Faces were either presented in the same or opposite position as in the training phase. At 1PM, subjects took a polysomnographically recorded nap with or without REM sleep or remained awake. The change in face association memory (d’) between the morning and afternoon test sessions was calculated for various sleep conditions. This value was contrasted for position-intact and position-flipped conditions to define a position-invariant face association index (d’Δ). Sleep did not improve retention of associative face pairs in general. However, one-way ANOVA found a difference in d’Δ between three sleep conditions (F(2,93)=3.86, p<0.05). Neither NREM or REM groups differed from wake, but NREM naps had greater d’Δ compared to REM naps (p<0.01). This suggests REM sleep is involved in consolidation of image-based learning for faces while NREM sleep is engaged in more abstract position-independent learning. Additionally, sleep spindles were associated with increased false alarm rate (r=.38, p<.01) and decreased d’ (r=-.29, p<.05), suggesting that sleep spindles may reflect mechanisms associated with false visual memory formation. In summary, we demonstrate a differential effect of sleep stages on associative face memory related to coding for spatial position.
Meeting abstract presented at VSS 2013