In addition to addressing response-time outliers, we also considered the potential differences in the rate of response as influenced by the individual and technological differences particular to each participant. For example, some people are naturally faster than others in answering questions or have faster Internet connections that may influence baseline response latencies. To control for this, Mulligan et al. (2003) make the following recommendation:

Accordingly, time scores used for our analyses were normalized by dividing time spent responding to the item battery by overall time spent completing preexperimental questions.

Following accepted practice in response latency measurement (Mulligan et al., 2003), outliers were assumed not to represent the actual time participants spent in answering questions and, instead, were replaced with corresponding sample mean scores. Although seemingly arbitrary, “trimming the tail of the latency distribution in this manner results in the loss of a very small proportion of the latencies and improves analysis by reducing the signal-to-noise ratio, allowing researchers to assess more clearly associations between accessibility and substantive variables of interest” (Mulligan et al., 2003, p. 293).