Switching tasks increases time: scientifically proven?

Question

Is there any scientific data available regarding the impact on delivery time due to switching between tasks?

Peopleware (IIRC) suggests it's half an hour per switch, but I feel it could be a lot higher.

Answer 1

Joel on Software: Human Task Switches Considered Harmful
Coding Horror: The Multi-Tasking Myth

Answer 2

Two articles from Kathy Sierra (from the Creating Passionate Users's blog) on multitasking. 1, 2

She doesn't cite articles explicitly, but does list some authors that may provide what you're after.

Answer 3

The general consensus among all industries is that task switching is detrimental. The more complex the tasks, the greater the detriment. If you are looking for scholarly arguments for this try these:

1. A Diary Study of Task Switching and Interruptions, Microsoft Research
2. Concerning Interruptions, Stephen B. Jenkins, IEEE Comp. Society, 2006
3. Information Needs in Collocated Software Development Teams, ICSE Proceedings of the 29th international conference on Software Engineering - need to be an ACM member to access
4. From Scatterbrained to Focused: UI Support for Today’s Crazed Information Worker, Microsoft Research, 2006
5. Predicting User Tasks: I Know What You’re Doing!, School of Electrical Engineering Oregon State University

All of the papers listed above contains very many citations and references if you are interested in reading more. I'm not sure if any of these sources might make raw data available.

On the lighter and shorter side, these are not scholarly, but not in the blogoshpere either:

1. The Art of Lean Software Development: A Practical and Incremental Approach, page 19
2. Worker, Interrupted: The Cost of Task Switching, FastCompany, 28 July 2008
3. Continuous Partial Attention and Software Productivity, QSM Associates, 14 Nov 2002

Answer 4

An excellent general summary (not limited to programming tasks) of the cost of task switching is here -- the summary's written in terms accessible to the non-expert, but it's published at a pretty reputable site (the online site of the American Psychological Association), and ends with a bibliography of a dozen good references to prestigious peer-reviewed journals. You'll have to go to said journals for most numbers -- the summary only quotes Meyer's famous measurement that

even brief mental blocks created by shifting between tasks can cost as much as 40 percent of someone's productive time

That summary is from 2006, and research in the fields has of course continued since. A drier and less accessible, but more recent research survey from earlier this year is here (a site at Michigan State University) -- it quotes and summarizes a dozen of the survey author's works (Erik M. Altmann, associate professor with the Department of Psychology at Michigan State University). All links in this summary are to freely downloadable PDFs, so starting from here may be handier than the study I previously quoted if you don't have access to JSTOR through your library or research institution; however, what you get here is basically a good survey of the work of just one researcher.

Of course, no summary can replace going right to the peer-reviewed scholarly journals and studying (and critically assessing) all the relevant results. It is, however, a work that will take you years, because task-switching costs and human multitasking issues are an extremely popular research field these days -- from the neuroscientists to the behavioral psychologists and management scientists, everybody wants a slice of the action. A Google Scholar search for task-switching costs shows 16,400 results -- many of those are freely accessible even without JSTOR, most all with be with JSTOR or other similar gateways to published research, though probably understanding most of them thoroughly will require a few PhDs in related disciplines (peer-reviewed journals do tend to be that way;-).

BTW, answering the question you ask in your Q's title: way scientifically proven (as solidly as, say, evolution, or global warming, though of course you'll always find deniers for any of those too;-) -- the hot debate and enormous volume of research is about understanding what are the exact numeric parameters of task-switching costs, how they depend on the tasks and on other circumstances, what neurological or other pathways explain what part of them, how to organize to reduce or minimize them, and so forth. Whether the famous "40% productivity loss" estimate is roughly in the ballpark (as I suspect it is) may still be somewhat contentious, but the existence of the effect as a some pretty relevant number (and with enormous economic costs all over the world, once you consider that it applies to every task human beings perform -- permanently raising every human being's productivity by even 10 % would mean a huge jump ahead for the world!-) is essentially beyond doubt.