Counterbalancing

Order and Sequence Effects

The goal of counterbalancing is to ensure internal validity by controlling the potential confounds [Page 278]created by sequence and order effects. A sequence effect (e.g., practice) occurs when responses to a condition are influenced by the sequence in which conditions are presented. Order effects occur when the position a condition occupies in the research protocol (e.g., 1st, 2nd) influences the response (e.g., fatigue effects). Suppose that participants in a laboratory experiment are asked to interact with a remote partner three times, each time through a different channel (video, chat, and text message), and measured after each interaction on feelings of self-efficacy. The researcher’s goal is to assess which channel obtains the highest ratings of perceived self-efficacy. Further, suppose that all participants interact through the three channels in the same order: first video, then chat, and finally text message. Because the sequence is the same for all participants, low self-efficacy attributed to video cannot be univocally attributed to the channel, since scores may have been tainted by the participants’ lack of practice with the study protocol (order effect); similarly, low ratings for text message (the last condition) may reflect fatigue due to a long experimental session (order effect), or may be due to a comparison with the preceding condition (sequence effect). Thus, the main effects of channel (the independent variable) will be confounded by order and sequence effects unless counterbalancing is used.

Counterbalanced Designs

Counterbalancing can be obtained through different designs. The major distinctions are between intrasubjects and intersubjects designs, and complete and incomplete designs. The first distinction refers to exposure of participants to the conditions. Intrasubjects counterbalancing allows for order and sequence effects to be balanced within subjects by exposing each participant to all conditions multiple times and in different orders, and is obtained through either ABBA counterbalancing or block randomization.

...