Response Selection Multitasking Bottleneck: Can it be bypassed? Part 3

This post is the final installment in my discussion of a study by researchers François Maquestiaux, Maude Laguë-Beauvais, Eric Ruthruff, and Louis Bherer titled "Bypassing the central bottleneck after single-task practice in the psychological refractory period paradigm: Evidence for task automatization and greedy resource recruitment" published in October of 2008 in the journal Memory and Cognition. Having explained in my first post the central bottleneck model of response selection, and having reviewed in my second post the experimental design refinements that researchers made to more accurately investigate whether the bottleneck can be bypassed, I am now ready to share the results of Maquestiaux et al.'s study.

Maquestiaux et al. performed two experiments. The first one was designed to detect the bypass of the response selection bottleneck. They set up this experiment to control for all of the known confounding factors relevant to detecting whether the bottleneck has been bypassed or not (some of which I discussed in yesterday's post). Among other important factors, they ensured that subjects received sufficient training with the task they wanted subjects to automatize (over 4,000 practice repetitions). They also ensured that the experiment included verbal and manual response components, and that the first task was sufficiently difficult that it would produce a measurable delay in the subjects' processing of the second task.

In the first experiment, subjects were instructed to verbally identify whether an audible tone was high-pitched or low-pitched. Then they practiced discriminating the two tones several thousand times until the response became automatic. After the training for the tone task, the subjects then were instructed to identify one of four visually displayed letters by manually pressing corresponding buttons on a keyboard. At the same time, they were instructed to continue discriminating between the high and low tones verbally. The letters were always displayed first, and the tones were played second, so subjects were expected to respond to the letter task first and the tone task second. Maquestiaux et al. manipulated the timing of the tone task to see its effect on the length of time it took subjects to respond. If bypassing were possible, Maquestiaux et al. expected that the timing of the tone would not have an effect on the length of time it took subjects to respond to it.

The results of the first experiment provide strong evidence that it is possible to bypass the response selection bottleneck. Eleven out of 20 subjects demonstrated the ability to respond to the tone at the same time as the letter response was being processed, and six subjects even responded to the tone before the letter--even though the letter stimulus was always presented first. Additionally, when Maquestiaux et al. increased the difficulty of the letter task to increase its processing time by a specific number of milliseconds (they had measured the increased response time), it did not increase the quantity of time it took for subjects to respond to the tone, contrary to the bottleneck model's predictions. From these results, Maquestiaux et al. concluded that their subjects had bypassed the response selection bottleneck.

For the second experiment, Maquestiaux et al. retested the five fastest subjects, to find out whether Ruthruff, Van Selst, Johnston, and Remington's "greedy resource recruitment" hypothesis was accurate. If it were, then when the tone task was presented first, even though it had been automatized, it would monopolize the brain's response selection resources and cause a measureable delay in the subjects' responses to the letter task, which was to be presented second. Sure enough, that is exactly what they found.

There were a number of things about this study that struck me, but the one thing impacted me most was Maquestiaux et al.'s careful explanation for the design of their experiment. Every aspect of it was based on the findings of previous studies. Because of the care they exercised in justifying their experimental design, I was able to take their results and conclusions seriously.

This study has furnished fairly reliable evidence for the trend observed so far in this blog: multitasking is possible, and technology can help train people to do it better. This study also helped me to better define multitasking and task switching. At this point, I think multitasking involves the automatic performance of one or more tasks concurrently with a task that requires response selection--i.e. only one response can be selected at a time. If tasks are not automatized, then one must switch between them.

Questions: Do tasks performed concurrently "count" as multitasking even if the responses underlying them are performed consecutively? Is it possible for people to learn to select two (or more) responses simultaneously? How complicated can a task become before it is un-automatizable? How might "complicated" be defined? How does this research relate to the dual-processing model of working memory? With all these questions, it looks like I've got my work cut out for me.