Wrist actigraphy is increasingly used to track circadian rest-activity cycles and to identify states of wakefulness and sleep, yet the measurement characteristics of activity recorders have never been compared. Two widely used recorders are compared here: the MotionLogger from Ambulatory Monitoring, Inc (AM) and the Gaehwiler (G). They were worn together on the same wrist for periods averaging 41.5 hours by five members of a research team. Activity counts were stored every half-minute. Pairwise comparisons between recorders of each type showed both types to be reliable. Each also validly detected circadian rest/activity cycles. Both types suffered, however, from insensitivity. For the lower 75% of activity levels, the variance of data from the G was indeed so small as to be essentially uninformative. Since these levels include over 95% of all nocturnal data, the G must be less sensitive than the AM to small nocturnal movements, including those signifying arousal. An additional difference is that data from the AM but not the G were distributed in biphasic fashion. Biphasic activity levels are consistent with the common assumption that activity/wakefulness and rest/sleep are distinct neurobehavioral states. As the use of actigraphy increases, the important differences found here between two leading instruments point to an urgent need for standards by which activity recorders can be compared. Aspects of instrument design that could be quantitatively rated are reliability, validity, ruggedness and artifact rejection.