In 1868, the Dutch physiologist and ophthalmologist F. C. Donders suggested that such mental processes as sensory discrimination, perceptual identification, and motor selection might occur serially, each consuming a certain amount of time. If so, wrote Donders, “interposing into the process some new components of mental action would reveal the time required for the interposed item” (Donders, 1868/1969, p. 418). In other words, if one could devise a pair of tasks differing only in that one required an extra mental process, then the difference in reaction time (RT) to the two tasks would be an estimate of the duration of the extra mental process. To illustrate his subtraction method for studying cognition he compared three tasks: Simple Reaction Time, Go/No-Go Response, and Choice Reaction Time.

Donders assumed that the Simple RT task is identical to the Go/No-Go task, except for an extra process of stimulus discrimination in the latter: Go/No-Go requires the discrimination of squares from diamonds, whereas Simple RT does not. Similarly, the Go/No-Go and Choice RT tasks are identical, except for the extra process of response selection in the latter. Therefore, subtracting mean Simple RT from mean Go/No-Go RT should provide an estimate the duration of the stimulus discrimination stage. Subtracting mean Go/No-Go RT from mean Choice RT should provide an estimate the duration of response selection. Your data in these three tasks likely confirm that Choice RT is more difficult than Go/No-Go, which in turn is more difficult than Simple RT. However, are the RT differences accurate estimates of stimulus discrimination and response selection timing? Or do the three tasks differ in ways other than the “interposed items” proposed by Donders? The validity of any application of the subtraction method depends on the assumption of pure insertion: that a mental process can be added or omitted without in any way altering the speed of the other processes.

This assumption has long been criticized on many grounds (Boring, 1929; Ilan & Miller, 1994). For instance, changing a simple task to a more complicated one changes the subject’s strategy, and the entire conscious information processing pattern is therefore altered. Introspective accounts have been used to argue that the subtraction method is invalid for studying cognition, because increasing a task’s complexity always affects other stages, both qualitatively and quantitatively (Kuelpe, 1909). Despite such criticisms, Donders’ subtraction method continues to influence modern cognitive psychology. Sternberg’s seminal Additive Factors Method (Sternberg, 1969) is based on Donders’ study, and modern brain-imaging procedures such as PET and fMRI critically rely on subtraction logic to infer what parts of the brain are used to perform basic mental processes.