Mixing Cost

What it is

Kiesel and colleagues, in their 2010 review in Psychological Bulletin, describe the two costs together. A switch cost is the difference in reaction time between switch trials and repetition trials, a definition given with the caption to the review's second figure (p. 853). Mixing costs are the global slowdown in mixed-task blocks compared with single-task blocks. The review calls these alternation costs or mixing costs (p. 850) and contrasts the global costs with the local switch cost (pp. 850, 852).

The earliest experiments measured the global cost. The review describes Arthur Jersild's work, published in 1927 as Mental set and shift, as having people alternate between two arithmetic tasks in a block and compared with performing just one task (p. 850). It cites Spector and Biederman (1976) alongside Jersild for the same fixed-sequence approach, in which the task changed on every trial. Both are also cited as early sources for the idea of a task set.

Later designs changed the schedule. In the alternating-runs paradigm the task changes after a fixed number of trials, and in Rogers and Monsell's 1995 experiments that number was two in five experiments and four in the sixth. Because a run of repeats sits inside the same block as the switches, the design lets the experimenter compare switch trials with repeat trials without leaving the mixed block. This is the local cost. The comparison with a pure block is the global one.

In effect

Jersild's lists and the block comparison In the 1927 design as the review reports it, the quantity of interest was the time to complete a block. The comparison is between a block of one task and a block alternating two, so the number it produces is a block-level difference and not a trial-level one. The review does not give Jersild's figures, and this page does not either.

The cost that persists in simple tasks Kiesel and colleagues note that costs in the alternating-runs paradigm were substantial even though the tasks were rather simple (p. 852). Digit and letter judgements of the kind Rogers and Monsell used are about as simple as tasks come, and the local switch cost showed up on the first trial after each change.

Why the two are reported separately An article that says task switching costs time can mean either cost. The residual switch cost, the part that survives preparation, is a statement about a single trial. A mixing cost is a statement about a block. When a popular source says that costs persist when a switch is predictable, as The Distracted Mind does at p. 79, the supporting experiments are not all the same measurement. Rubinstein and colleagues (2001) defined the switching cost as the difference in mean reaction time between alternating and repetitive blocks, which is a block-level comparison. Rogers and Monsell's 1995 experiments located the cost on the first trial after each change.

What it does not say

It does not say how large the mixing cost is. The sources read for this page describe it as a global slowdown and give no magnitude for it, so none is given here.

It does not say what causes it. A comparison between a mixed block and a single-task block cannot by itself separate the cost of changing from the cost of being in a situation where a change might come. The review treats global and local costs as distinct, and this page follows it.

It does not say the cost applies to working life. The experiments use laboratory tasks lasting seconds. Whether a mixed afternoon costs more than a pure one in an office is a question for studies of work, and the interruption literature covered by our articles on attention residue and self-interruption addresses it with different methods.

Who developed it

Arthur Jersild is credited with an early comparison of alternating and single-task blocks in 1927. The phrase mixing cost is later vocabulary, used in the literature reviewed by Kiesel and colleagues, and we do not name a coiner. Our confidence in the attribution of the comparison to Jersild is medium, because it rests on the review's description and we did not read the 1927 monograph.

---