The memory mechanism carries information forward in time. Screening for genetic distortions in this information (screening for changes in the content of memory) is likely to do a better job of distinguishing between genetic effects on memory mechanisms and genetic effects on performance mechanisms than screening for changes in the strength of learned behavior. The peak procedure is the most commonly used screen for duration memory. Mice give peak data strikingly similar to data from the rat and the pigeon. The data, however, favor a model in which the decision criterion for starting the response (putting the head into the hole) and stopping the response are independently determined. For reasons we explain, this prevents the estimation of scalar memory error and memory variability from simple peak data.