What a working day actually looks like

Lesson 2 asked you to count your discretionary hours twice, as a total and as a shape, and some of you will have found them arriving in pieces shorter than you expected. This lesson is about why the day does that, and about the most quoted number in the whole subject, which turns out to have no printed source anybody has been able to produce.

What was measured, and on whom

Everything in the first half of this lesson comes from one study, so its scope goes first and the figures follow.

Mark, Gonzalez and Harris, published at CHI 2005, followed twenty-four information workers at a single outsourcing company and observed them to the second.¹ It is a field study rather than an experiment: nobody was assigned to anything, and the researchers observed people directly and recorded what they did.

That's a small number of people in one office in one industry, which means the figures below are a description of that office rather than a law about work. Lesson 1 asked you to write the scope beside every number you carry, and this is a lesson where it matters, because these particular numbers travel further from their source than almost any others in this subject.

Three findings, in the authors' order.

People had about a dozen pieces of work on the go. Each informant worked in an average of 11.7 different working spheres (standard deviation 2.4, which is a measure of how spread out the individual counts were around that average).¹ The source says they worked in that many; it does not say all of them were live at the same moment. A working sphere is a chunk of work with its own purpose and its own people, roughly what you would call a project or a case.

They spent about eleven minutes in one before leaving it. "The average length of time that the informants spent in central and peripheral working spheres was 11 min. 4 sec. (sd=18 min. 9 sec.) before switching to another working sphere or being interrupted."¹

Read the bracket as well as the figure. The spread is larger than the average. That isn't a tidy day of eleven-minute blocks; it is a day with a great many very short pieces and a few long ones, which is exactly the shape lesson 2 asked you to look for in your own week.

And more than half of those stretches ended because something broke in. 57.1 percent of working sphere segments were interrupted, which the abstract reports as "57% of their working spheres are interrupted".¹ Central spheres ran at 60.3 percent, peripheral ones at 41.7. This course has read those two figures and not the study's definition of the two terms,⁴ so read "central" as the study's own label rather than as anything this lesson can unpack.

Time spent is time exposed

This is worth slowing down on, because it quietly demolishes a piece of advice you have certainly been given.

The explanation isn't complicated. Time spent is time exposed. Every extra minute you are in a piece of work is another minute in which the phone can ring, and the longer you have been at something the more visibly involved in it you are. Involvement is legible to other people. They can see you are working, they can often see roughly what you're working on, and for a colleague with a question that makes you findable rather than protected.

Put the two findings side by side and the picture is coherent rather than strange. The most central work is interrupted most, because it is the work you're most obviously doing. The longest stretches are interrupted most, because they are the longest. Neither of those is anybody behaving badly, and neither is going to be fixed by trying harder.

Worked: a day counted, against the study's day

Sam's day is constructed, and so is every figure in it.⁴ He estimates jobs for a building firm. He has the record from lesson 1 and has gone back over Tuesday with a pen.

He marks a new working sphere each time the purpose of what he is doing changes. That is this course's working rule for the exercise and not the study's own definition,⁴ which this course has read described but not specified.

From 8:40 to 17:20, with forty minutes for lunch, he counts 31 switches across 9 working spheres.

Do the sum carefully, because it has two places to go wrong. 8:40 to 17:20 is 520 minutes, and less the 40 for lunch that is 480. And 31 switches make 32 stretches, not 31, because the first stretch starts before any switch has happened. 480 over 32 gives 15 minutes.

Set that beside the study's 11 minutes 4 seconds and the honest conclusion is that his day looks much like their day. Not because fifteen is near eleven, though it is, but because he had to count to find out, and before he counted he would have told you he spent the morning on the tender.

The number isn't the finding either. What Sam's sheet tells him is where the two long stretches were: 9:10 to 9:55 and 15:30 to 16:20, both while his manager was out of the building. That is a condition, in the sense lesson 2 used the word, and it is the kind of thing a week's record can show you and an impression can't.

Worked: when interruption is the job

Nadia is a ward nurse. She is also constructed.⁴

Her day has more switches in it than Sam's and she isn't going to reduce them. Being interruptible is not a failure of her arrangements; it's a large part of what she is for, and a ward where the nurse could not be interrupted would be a worse ward.

Almost all of the advice she will meet on this subject is useless to her, and the reason is worth naming: the advice treats interruption as a defect in a day that could have been otherwise. Hers couldn't.

So the question changes shape. Not "how do I reduce interruptions", which is not hers to decide, but which of my work actually needs a continuous stretch, and when in the week does such a stretch exist at all?

For Nadia the answer to the first is short, and everything in this paragraph is stipulated about her rather than established about nursing.⁴ Most of her job is meant to be interruptible and, as this case is built, is done no worse for it. Two things aren't: the drug round, which in this constructed case the ward already protects by procedure, and the written handover, which she currently does in fragments and which takes her twice as long that way.

The answer to the second is that a stretch exists in exactly one place in her week, the half hour after the handover meeting on the days she is on early. That is the whole of the discretionary capacity her job offers for work needing continuity, and knowing that it is one half hour is more use to her than any technique, because it tells her that the written handover is the only thing that can go there.

The number nobody can find in the paper

Now the part of this lesson that is a method rather than a finding.

You have almost certainly met the claim that an interruption costs you twenty-three minutes and fifteen seconds, or twenty-three minutes, or "about twenty-five minutes to get back on task". It is the most quoted figure in this subject. It is on slides in corporate training rooms and in more articles than anybody has counted.

It's normally attributed to one of two papers, and this course went looking in both.

The 2005 field study, the one every measured figure above comes from. This course has read its abstract, its research setting and its results on fragmented work.¹ The figure is not in what this course read, and the paper's finding about resumption is about something else entirely: the abstract says that "most interrupted work is resumed on the same day" but that "more than two intervening activities occur before it is".¹ That's a count of activities, not a number of minutes.

The 2008 laboratory experiment, by Mark with different colleagues. This course has read its abstract and its experimental design.² The figure is not there either, and the paper's actual finding is close to the opposite of the popular one, which the next section is about.

And a third party went looking properly. A 2023 review by a writer working through five papers and twenty-three blog posts concluded: "So in the end, where do the 23 minutes and 15 seconds come from? They are mentioned in interviews multiple times by Gloria Mark. But I wasn't able to find a primary printed source."³ This course cites that piece for one claim only, which is that its author looked and did not find one. It is a blog post, not a study, and the honest use of it is as a record of a search rather than as an authority.

What an interruption does cost, measured

The 2008 experiment is the honest answer to the question the twenty-three minutes pretends to answer, and the first thing to say about it is what it was.²

Forty-eight subjects, 81 percent of them German university students with a mean age of 26, answering twelve simulated emails as a fictional human resources manager, interrupted every two minutes by the experimenter, by telephone or instant message, and paid twelve euros. A laboratory task of twelve emails. It is not a study of a working day and this lesson isn't going to let it be read as one.

Its finding, verbatim from the abstract: "We found that context does not make a difference but surprisingly, people completed interrupted tasks in less time with no difference in quality. Our data suggests that people compensate for interruptions by working faster, but this comes at a price: experiencing more stress, higher frustration, time pressure and effort."²

Less time. No difference in quality. More stress, frustration, time pressure and effort.

That is a real cost and it isn't the one the folklore describes. Nothing was lost in minutes and nothing was lost in the work. What was paid was paid in how the work felt, and people bought the speed with it.

If you have ever come out of a heavily interrupted day having got everything done and feeling as though you'd been hit by a bus, that combination is the measured one, and there is no contradiction in it to be resolved.

Four things people believe about interruption

"An interruption costs twenty-three minutes." Traced above. Not in what this course read of either paper, no primary printed source found by somebody who went looking, and the papers it is attributed to report a count of intervening activities and a cost in stress.

"Interruptions are what other people do to me." The 2005 study records internal interruptions, the ones you start yourself, as a large share of the total.¹ This course has not read a percentage for it and isn't going to print one, which is the rule lesson 1 set. The practical form is that the exercise below asks you to mark each switch internal or external, because the two have entirely different remedies, and this course expects the ratio to surprise you[4].

"Interruption is simply bad." The authors are careful in a way the genre is not: switching may refresh a person and bring new ideas, and too much of it across too many activities may be harmful. The 2008 experiment found interrupted tasks finished faster and no worse.² A day with no interruption in it is not the target, and a course that implied otherwise would be selling something its own sources do not support.

"Open offices are the cause." The collocation finding goes both ways, in the same study, on the same people. Workers sitting near their colleagues, which the study calls collocation, spent longer in a working sphere, 11 minutes 56 seconds against 9 minutes 56, and were interrupted more, 62.9 percent against 49.3.¹ Whatever you think about open offices, this study does not settle it, and it's a good example of a result that gets quoted in one direction by people who have read half of it.

Practice

Connections

Back. Lesson 2 found that discretionary hours arrive in pieces and asked you to measure the shape as well as the total; this lesson is the measurement of that shape in somebody else's office, and the reason it happens. Lesson 1's habit of writing the scope beside a figure is what the second half of this lesson is made of. From earlier Core courses: Using AI Effectively lesson 11 taught you to ask what a claim was measured on and when, which is what the first half of this lesson does with the 2005 study's scope. The provenance tracing in the second half is Logic and Argument lesson 9's four rules for sources arriving in a new subject, including what to do when a figure turns up with no denominator.

Forward. Lesson 6 is deciding what not to do, and the internal-against-external count from the exercise above is one of the things it works on. Lesson 7 builds a week that takes the shape of your day seriously rather than assuming an empty one. Lesson 8 is the trade literature, read with the tracing move from this lesson applied to it.

Go deeper

• The field study this lesson is built on (CHI 2005). Ten pages, free. This course has read the abstract, the research setting and the results on fragmented work, which is where every measured figure above comes from. Worth reading for one thing lesson 2 did not send you there for: pages 324 and 325, where the interrupted-against-uninterrupted and collocation figures sit together, which is the passage this lesson's second section is built on.
• The Cost of Interrupted Work: More Speed and Stress (CHI 2008). This course has read the abstract and the experimental design. Read the design before the result, which is a general habit worth having and matters here: what the study did tells you most of what the finding can carry.
• Interruptions cost 23 minutes 15 seconds, right? (2023). A blog post rather than a source, linked because it is a worked example of the tracing exercise above, done properly and written up. Read it as a method, not as an authority.

Sources

1. Gloria Mark, Victor M. Gonzalez and Justin Harris, "No Task Left Behind? Examining the Nature of Fragmented Work", CHI 2005, Portland, Oregon, pages 321 to 330. Read in part: the abstract, the research setting, and the results on fragmented work, pages 321 to 325. Supports: the study design and that it is a field study; 11.7 working spheres with a standard deviation of 2.4; the quoted sentence giving 11 min. 4 sec. with a standard deviation of 18 min. 9 sec.; 57.1 percent of segments interrupted, with 60.3 percent central and 41.7 percent peripheral, and the abstract's "57% of their working spheres are interrupted"; the quoted resumption finding about more than two intervening activities; 12 min. 40 sec. against 8 min. 58 sec. for interrupted and uninterrupted segments, and the quoted sentence that the longer one works the longer the interrupting event; "the irony of the work day"; the collocation figures of 11 min. 56 sec. against 9 min. 56 sec. and 62.9 percent against 49.3 percent; the authors' own two-sided statement about whether switching is beneficial; and that internal interruptions are a large share. This course has read no percentage for the internal share and prints none. Scope: 24 information workers at one outsourcing company, published at CHI 2005. The earlier three-minute figure for individual work events, cited in quiz item 1's explanation, is the same group's earlier study as reported on page 322 of this one.
2. Gloria Mark, Daniela Gudith and Ulrich Klocke, "The Cost of Interrupted Work: More Speed and Stress", CHI 2008. Read in part: the abstract and the experimental design, pages 1 to 2. Supports: the quoted abstract in full, and the design. Scope: a laboratory experiment, 48 subjects, 81 percent German university students with a mean age of 26, twelve simulated emails, interruptions every two to five minutes, paid 12 euros. It is not a study of a working day and the body says so at the point of use.
3. oberien, "Interruptions cost 23 minutes 15 seconds, right?", 2023. Read in full. A blog post, not a source, and cited for exactly one claim: that its author reviewed five papers and twenty-three blog posts and could not find a primary printed source, which is quoted verbatim in the body. Nothing about interruption itself rests on it.
4. Three things here are the course's own and are labelled where the reader meets them. The reading of the irony finding as an argument against "protect a long block" is reasoning from the study rather than a finding in it. Sam and Nadia are constructed, along with every figure attached to them; no source read for this course records one person's day. And the rule of thumb for marking a new stretch is a working definition for the exercise, not the study's own operational definition of a working sphere, which this course has read described but not specified.