within-subjects experiment

Terms from Statistics for HCI: Making Sense of Quantitative Data

When you ask the same person to take part in two or more conditions of an experiment, in contrast to between-subjects design where each person takes part in a single condition. For example, if you are comparing two user interfaces, you might ask each subject to perform tasks on both. The advantage of this is that if an individual is particularly skilled or unskilled at the tasks or has simply had a bad day, this cancels out between the two conditions – that is it cancels out individual differences. Crucially this increases the statistical power of your study by reducing noise. The downside is that there may be order effects or other forms of interference between the conditions; for example, a positive learning effect where the fact that someone has done a task with the first system makes it easier to perform a similar task on the second system. There are ways to ameliorate these effects, for example to randomise the order of conditions. If a within-subjects design is not possible and you need to use a between-subjects design, you might try matching subjects.

Defined on page 110

Used on pages 110, 111, 112, 118

Also known as within-subjects design, within-subjects designs, within-subjects designs , within-subjects studies