Experimenters' assumptions and animals' perceptions: observer bias in temporal-discrimination tasks

In laboratory studies of animal behaviour, interpretations of how a task is learned or of which stimulus characteristics are attended to are, ultimately, inferences. So, there is always potential for experimenter assumptions to be at odds with what an animal learns from a task. The study of choice biases has been fertile ground for revealing another kind of bias: observer bias. This paper reviews a series of studies featuring a matching-to-sample task where, following the presentation of a stimulus, a choice had to be made between two options. The option to be chosen depended on the duration of the initial stimulus. An intertrial interval traditionally separates trials and, in a variation of the task, a delay is introduced before choice. In some cases, task components traditionally considered secondary (such as the intertrial interval) actually become part of what the animals learn from the task. Not only that, but individual animals may also differ in how they use different task components. In addition, delays appear to be interpreted differently across tasks. This paper presents attempts to control for confounds and to clarify what is learned in each task. Biases and task limitations highlight the importance of employing a variety of procedures to study a given phenomenon. • In laboratory tasks, the perceptions of experimenters and animals may not coincide. • Choice biases are commonly found in temporal discrimination tasks. • These biases suggest that animals attend to cues not predicted by the experimenter. • Observer bias may lead to misinterpretations of the animals' behaviour.