Structure of Electrodermal Responses during Sleep

Taylor, S., Sano, A., and Picard, R. "Structure of Electrodermal Responses during Sleep," Sleep2017, June 2017.



Spontaneous electrodermal responses (EDRs) during sleep, or sleep storms, have been observed since the 1960s. With results counter-intuitive to an emotional arousal interpretation, these studies have found that sleep storms occurred most frequently during slow wave sleep and least frequently during REM sleep. However, little is known about the sleep EDR structure. We used the state-space generalized linear model (SSGLM) to investigate the structure of EDRs during sleep.


Eleven individuals (10 male) wore two sensors to capture actigraphy and electrodermal activity and reported their sleep timing via online surveys for 30 days. The EDRs during sleep (sleep was determined by actigraphy and surveys) for each participant were modeled using the SSGLM using the Kolmogorov-Smirnov test. This approach allowed us to model the rate of EDRs during different time periods during sleep. Using Monte-Carlo techniques, we computed a comparison between EDR event rates between pairs of time periods during sleep, without needing to correct for multiple comparisons. Previous work found that periods of high frequency EDRs were most probable in the first half of the night; thus, we chose to compare the first two 90min of sleep. In particular, we computed the probability that the rate during one period was greater than during another period within a night.


When comparing the first two 90min of sleep, we found that for 8 of the 11 participants there were a high percentage of nights (>80%) where one of the two 90min periods had a significantly higher probability of being greater than the other. Furthermore, we found that 8 of the 11 participants had more days with the higher rates during the second 90min than the first.


Using the SSGLM provides a framework to compare the rates of EDRs during different time periods of the night. This work provides a baseline for what pattern of EDRs are typical in a healthy college-age population. The methodology can be extended to model EDRs during sleep in other populations and used to compare differences.

Related Content