Naturally-occurring sleep choice and time of day effects on p-beauty contest outcomes

We explore the behavioral consequences of sleep loss and time-of-day (circadian) effects on a particular type of decision making. Subject sleep is monitored for the week prior to a decision experiment, which is then conducted at 8 a.m. or 8 p.m. A validated circadian preference instrument allows us to randomly assign subjects to a more or less preferred time-of-day session. The well-known p-beauty contest (a.k.a., the guessing game) is administered to examine how sleep loss and circadian mismatch affect subject reasoning and learning. We find that the subject responses are consistent with significantly lower levels of iterative reasoning when ‘sleep deprived’ or at non-optimal times-of-day. A non-linear effect is estimated to indicate that too much sleep also leads to choices consistent with lower levels of reasoning, with an apparent optimum at close to 7 hours sleep per night. However, repeated play shows that sleep loss and non-optimal times-of-day do not affect learning or adaptation in response to information feedback. Our results apply to environments where anticipation is important, such as in coordination games, stock trading, driving, etc. These findings have important implications for the millions of adults considered sleep deprived, as well as those employed in shift work occupations. Key Words: