"Constant, constant, multi-tasking craziness": Managing multiple working spheres

Abstract

Most current designs of information technology are based on the notion of supporting distinct tasks such as document production, email usage, and voice communication. In this paper we present empirical results that suggest that people organize their work in terms of much larger and thematically connected units of work. We present results of fieldwork observation of information workers in three different roles: analysts, software developers, and managers. We discovered that all of these types of workers experience a high level of discontinuity in the execution of their activities. People average about three minutes on a task and somewhat more than two minutes using any electronic tool or paper document before switching tasks. We introduce the concept of working spheres to explain the inherent way in which individuals conceptualize and organize their basic units of work. People worked in an average of ten different working spheres. Working spheres are also fragmented; people spend about 12 minutes in a working sphere before they switch to another. We argue that design of information technology needs to support people's continual switching between working spheres.

Full text

“Constant, Constant, Multi-tasking Craziness”:
Managing Multiple Working Spheres
Victor M. González and Gloria Mark
School of Information and Computer Science
University of California, Irvine
Irvine, CA 92697
{vmgyg, gmark}@ics.uci.edu
ABSTRACT
Most current designs of information technology are based on
the notion of supporting distinct tasks such as document
production, email usage, and voice communication. In this
paper we present empirical results that suggest that people
organize their work in terms of much larger and thematically
connected units of work. We present results of fieldwork
observation of information workers in three different roles:
analysts, software developers, and managers. We discovered
that all of these types of workers experience a high level of
discontinuity in the execution of their activities. People
average about three minutes on a task and somewhat more
than two minutes using any electronic tool or paper
document before switching tasks. We introduce the concept
of working spheres to explain the inherent way in which
individuals conceptualize and organize their basic units of
work. People worked in an average of ten different working
spheres. Working spheres are also fragmented; people spend
about 12 minutes in a working sphere before they switch to
another. We argue that design of information technology
needs to support people’s continual switching between
working spheres.
Categories & Subject Descriptors: H.4.1. [Information
Systems Applications]: Office Automation—Time
Management; H.5.2. [Information Interfaces and
Presentation]: User Interfaces — Theory and Methods.
General Terms: Human Factors
Keywords: Time management, attention management,
information overload, interruptions, empirical study
INTRODUCTION
More than ever, information workers are facing highly
demanding workloads. In some cases, with budget cuts and
layoffs, companies are struggling to do the same work with
less people and consequently are increasing the number of
tasks and responsibilities for each employee. In other cases,
such as academia, multiple activities are just part of the
inherent nature of the work: supervising students, performing
departmental duties, and participating in different research
projects. In both small start-up companies and large complex
organizations, information work in general is typically
characterized as requiring involvement in multiple projects,
initiatives and teams. This is the kind of work performed by,
e.g. administrators, managers, financial analysts, consultants,
and accountants. To add to the complexity, workers also use
a variety of digital and physical devices to conduct their
work: e.g. electronic mail, instant messaging, PDAs, cell
phones and paper documents.
The purpose of this study is to examine how people manage
their work on multiple projects each with different goals,
deadlines, and resource constraints using a variety of
technologies. Yet paradoxically, most current designs of
information technology support distinct tasks such as writing
and editing documents, using email, or sending text or phone
messages. Technology is not organized in terms of larger
themes that are associated with separate projects or duties. It
is left up to people to integrate their information into
cohesive task structures that make sense to them.
MANAGING MULTIPLE ACTIVITIES
In the last forty years in fields as diverse as CSCW,
organizational behavior, and management science, many
studies have been conducted to understand how people
distribute their time at the workplace. Most studies have
focused on managers, describing how they spend their time
both in terms of the structure and content of their activities
[4, 5, 7, 12]. For a review see [10]). These studies have
consistently shown how managers engage in multiple
activities. In the late 1960’s, Horne found that middle
managers spend most of their time in managing various
activities and very little time in reflection and solitary
decision-making [4]. Mintzberg found that the activities of
CEOs can be characterized by their brevity, variety and
fragmentation [6, 7]. Sproull, who found that managers spend
80% of their time talking with people, proposed that
managerial work be considered as multitask processing, and
not in terms of tasks and interruptions [12]. A recent study by
Perlow [11] of software engineers found that they
conceptually separate their activities into engineering work
and all else. The earlier studies have looked at time
distribution before IT played a major role in the workplace.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CHI 2004, April 24–29, 2004, Vienna, Austria.
Copyright 2004 ACM 1-58113-702-8/04/0004…$5.00.
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
113

Another line of studies has examined the nature of
interruptions. Gillie and Broadbent [3] found that the nature
and complexity of an interruption affects how much
performance will be disrupted. Surprisingly, O’Connaill and
Frohlich [9] found that 41% of the time people do not resume
their original task after an interruption. Interruptions have
been found to be beneficial as well as disruptive [5].
Taken together, these studies show high amounts of
interruptions in information work. However, we maintain it is
not only interruptions that tax an individual, but also the
process of frequently switching activities. Though the
popular and academic press has given much recent notice to
the “attention economy”, there has been a lack of empirical
studies to understand how people actually switch between
activities and use of paper and digital devices in the
workplace. To our knowledge, no study has ever focused on
the role of information technology as people switch between
work activities. Our study differs from these earlier studies of
time distribution in that we are studying people in a very
technology-rich environment. We explore to what extent
work and use of tools in a technology-rich setting is
fragmented and we examine the strategies that people use to
maintain continuity in their work.
RESEARCH SETTING AND METHODOLOGY
In order to gain a detailed understanding of how information
workers manage multiple activities we conducted an
observational study at ITS1, an investment management
company located on the west coast of the U.S. ITS acts as an
outsourcer, providing information technology and accounting
services for a major fund manager. Given the size and extent
of operations performed by their current client, ITS services
only this company. More than 250 employees work at the
ITS branch where we conducted the study. We concentrated
our work on the day-to-day operations of one team that we
call the JEB team. JEB is in charge of developing, testing
and supporting major financial software modules to be used
by the client. Twenty-five information workers form the team
including software developers, database administrators,
financial analysts, and managers.
JEB supports the client by guaranteeing at all times the
transmission, validation and booking of all the financial
transactions performed by brokers in the U.S., Europe and
Asia. The JEB team also acts as an intermediary between the
client and a financial market information company that
provides financial terminals and transaction support. Records
of transactions are accessed through applications developed
by JEB and used by other teams within ITS, who are in
charge of the accounting system and daily consolidation with
banks.
Twenty-two members of the JEB team work in cubicles in an
open office environment; three have their own offices. Each
individual has a cubicle with a networked computer, phone
1 This name, and all other names, are pseudonyms.
unit, and resources such as books, binders, stationary, etc.
They all have commercial email and activity management
tools. Six individuals have a financial terminal to monitor the
status of transactions and they can perform tests for the
software modules they develop. Printer and fax machines are
shared and are located at the end of an aisle.
This kind of open office setting allows team members to
interact and communicate easily with other colleagues even
without the need to move from their own cubicles. It is
common that people chat with each other through the walls,
or even walk over to join conversations in other cubicles. At
the same time, the height of the cubicles is high enough to
provide privacy for the occupants. The employees generally
concentrate on their work within the cubicle.
Methodology
The study was based on two main ethnographic techniques:
participant observation and use of long interviews. The level
of detail required for our research demanded that we be able
to capture the details of the informant’s behavior with respect
to the structure and the content of the activities that they
performed. We felt that indirect observation techniques such
as asking subjects to keep diaries or to generate estimates of
activities at the end of the day would be less appropriate as
they would be disruptive and inaccurate. Also, videotaping
was not permitted by ITS. Thus we decided to use a
“shadowing” observation technique similar to the ones used
in previous time management studies [11, 12]. In our case,
the researcher sat with the informant at her cubicle and
followed her, whenever possible, to meetings or other
activities. The researcher sat just behind the informant where
it was possible to fully observe what she did and to some
extent to be able to read documents displayed on the
computer screens, the ID caller display on the phone unit, the
content of print outs, sticky notes and binders on the desk,
etc.
Whenever the individual performed an action such as
opening a computer application, making a phone call, writing
down a note on her planner or pulling a paper note from the
cubicle wall, the researcher annotated the time (to the
second) and other details of the event. All interactions with
others were also documented, including details about the
topic of the conversation, documents used and persons
involved. The researcher was very careful to capture as much
detail as possible. Whenever something was not clear, the
researcher noted it and asked the informant at the end of the
day. Inspired by Mintzberg’s structured observation method
[6], we designed an activity tracking log where we
transcribed the observation notes collected during the day. In
the tracking log we included the time stamps, data about the
type of event (e.g. “responding email to AMX”, or
“modifying the Java code for CEW module), the resources
using during the action (e.g. phone, Excel, planner, sticky
note, calendar), and the people participating in the event. The
tracking logs were used in the analysis. A total of 477 hours
was spent in observation at the field site.
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
114

Fourteen people were observed over a seven-month period.
Each person was observed for a period of three and a half
days. The first half day was used to get familiar with their
activities and their working style. For the next three days
formal data collection was done and the average time of
formal observation for each individual was 26 hours. Some
days after the observation we conducted a two-hour semi-
structured long interview where we asked about their activity
management strategies and about some of the events that
happened during the observation. Those interviews were
tape-recorded and transcribed. We also collected documents
such as email print outs, project descriptions, group calendars
and software specifications. Using a grounded theory
approach [13] we analyzed our data to understand how
individuals spend their time, the usage of digital and physical
artifacts, the different kind of interactions and how activities
switch throughout the day.
Our study started with a period of ten non-consecutive days
of observation of one of the managers to become familiar
with the practices, projects and people related with the JEB
team. His location in the office and his position in the
organization’s hierarchy allowed us to gain a good overview
in the early stages of the study. This observation was also
useful for the other team members because they gradually
became used to the researcher’s presence.
The employees
Among the fourteen team members observed, six of these
were business analysts who were the main point of contact
with the client both for specifying long-term requirements
and for reporting problems during the daily operation of the
systems. Four team members were developers who were
more involved in the actual design and implementation of
software modules. Four individuals were managers who
coordinated and planned the work of other team members, as
well as interacting with the client and providers.
RESULTS: THE NATURE OF INFORMATION WORK
An overview of the data
We begin by dividing our data into five main groups of
events in order to compare our results with those of earlier
studies [4-6, 12] (see Table 1). Deskwork, which is defined
as the time that individuals work with computers, and other
physical artifacts, clearly consumes a main portion of the
day. Our results reveal more time spent in deskwork
compared to earlier non-intensive IT environments, and is
most similar to Hudson’s data [5] of managers in IT-rich
environments. Unscheduled meetings (going to other
cubicles, people entering cubicles, or chatting through the
wall) constitutes the second largest category. This is not
surprising if we consider how easy is for the JEB members to
walk to another cubicle or to chat from their own. On the
average, our informants spent less time in formal planned
meetings which is consistent with Panko’s observation about
the relationship between upper hierarchical role and more
time spent in meetings [10]. We studied analysts, developers
and managers, while the other studies focused on managers.
The Other category includes time for lunch, personal
activities (e.g. getting coffee alone), and non-identified
activities outside the cubicle.
In summary, our data describes individuals who spend much
of their time in cubicles and have a high level of informal
interaction with nearby colleagues.
% time
Avg.
time/day
(s.d.)
Horne
1965a
Minzt-
berg
1970a
Sproull
1984a
Hudson
2002b
Desk
work
36.6
3:10:40
(1:22:51)
26%
22%
19%
42%
Phone1
5.8
0:30:22
(0:19:14)
9
6
13
E-mail
9.2
0:47:46
(0:21:18)
Sched-
uled
meetings
14.4
1:14:58
(1:17:40)
10
59
34
27
Unsched-
uled
meetings
18.9
1:38:40
(0:40:31)
55
10
34
19
Other
15.1
1:18:39
(0:34:26)
3
Total
100%
8:41:05
(1:03:08)
100%
100%
100%
88% 2
a Pre-email study b Post-email study
1 Includes time spent on cell phones
2 For this study 12% of the time subjects were “to busy to respond”
Table 1: Average percentage of time spent on activities in
current study compared to previous studies (hour:min:sec).
The fragmented nature of information work
Since the first days at the site we noticed that it was a very
fast-paced environment with multiple conversations,
telephones ringing constantly and people walking
unannounced into other cubicles or calling through the walls.
The high monetary value of information that the team
manages plays an important role in shaping the work rhythm,
as one analyst explained: “[The client] expects 100%
accuracy. They don’t want 99.9%. A lot of people think: ‘Oh,
we can fix it tomorrow’. ITS does not work like that. In this
kind of industry you have to correct it right away. Every time
with one issue is like a major issue. You know, every trade is
a minimum of a million bucks.”
We also noticed that the nature of their work was
characterized by a constant switching among physical and
digital artifacts as well. Their work could be described as
chains of short-term events. We coded our data into events
that we defined as any continuous use of a device or
engagement in an interaction with other individuals (e.g.
phone conversation, using a spreadsheet with the PC,
annotating documents, or talking “through the wall”).
Following Sproull [12], we considered that in any particular
event neither the structure nor the content changes. Table 2
shows the average time spent on an event per person, per
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
115

day, for all three roles combined (analyst, developer,
managers). This time reflects the amount of time that people
spent in continuous uninterrupted work on events.
Events
% entire
day
Avg.
Time/Day
(sd)
Avg. Time/
Event (sd)
Using phone1
5.83
0:30:22
(0:19:14)
0:02:25
(0:00:42)
Using email
9.17
0:47:46
(0:21:18)
0:02:22
(0:00:27)
Using PCs2
29.48
2:33:36
(1:11:23)
0:02:53
(0:01:10)
Using paper documents/books
6.80
0:35:25
(0:29:48)
0:01:47
(0:00:31)
Using other tools3
0.31
0:01:38
(0:03:08)
0:01:04
(0:00:15)
Talking through the walls
2.94
0:15:18
(0:14:12)
0:01:40
(0:00:24)
Interacting with people in their
own cubicle
6.88
0:35:53
(0:29:25)
0:03:34
(0:01:57)
Formal meetings
14.39
1:14:58
(1:17:40)
0:41:47
(0:12:46)
Going to other cubicles
9.11
0:47:29
(0:27:21)
0:07:37
(0:03:24)
Other (unknown, personal)
15.09
1:18:39
(0:34:26)
0:17:27
(0:06:27)
All events except “Formal
meetings” and “Other”
70.52%
0:45:56
(0:52:03)
0:03:08
(0:02:27)
All events total
100%
0:52:07
(0:55:25)
0:08:55
(0:13:23)
1 Includes time spent on cell phones
2 Includes both PCs and financial terminals – does not include email.
3 ‘Other tools’ include: handheld calculator, planners, and address
books
Table 2: Average continuous time spent on events before
switching (hour:min:sec).
What was most surprising to us was that people spend on the
average slightly over three minutes on an event (leaving out
formal meetings, unknown, and personal events) before
another event is initiated. An ANOVA between analysts,
developers, and managers shows no significant difference for
all but two of the events. We found a trend that developers
participate in shorter formal meetings (mean=30 min. 16 sec.,
sd=5 min. 54 seconds) compared with analysts and managers
(combined mean=46 min. 23 sec., sd=11 min. 54 sec.),
F(2,13)=3.4, p<0.07. A Tukey post-hoc test (95% confidence
interval) however, shows no significant differences. Also we
found that the three roles differ significantly in the
continuous time spent using the computer, F(2,13)=6.74,
p<0.01. A Tukey post-hoc test (95% confidence interval)
showed that developers spend significantly more time using a
PC (mean=4 min. 0 sec., sd=29 sec.) compared with
managers (mean=1 min. 58 sec., sd=46 sec.). Four minutes
on the PC before being interrupted or switching events is still
quite a short period of time for developers.
People spend an average of less than two and a half minutes
reading email before they switch to another event, or are
interrupted. The data also show, that in fact, people spend
very little time in one continuous stretch when they engage in
any kind of informal interaction with another person
(combined mean=4 min. 29 sec., sd= 3 min. 23 sec.).
Device
% entire
day
% of
device
usage
only
Avg. Time/
Day (sd)
Avg. Time/
Event (sd)
PC1
37.01
72.37
3:12:52
(1:13:48)
0:02:52
(0:00:51)
Financial
terminals2
1.64
3.19
0:16:59
(0:13:13)
0:01:20
(0:00:36)
Paper documents
and formats
5.01
8.92
0:26:06
(0:22:21)
0:01:33
(0:00:28)
Books, manual and
other references
1.79
3.50
0:09:20
(0:12:16)
0:01:57
(0:00:55)
Hand-held
calculator3
0.05
0.10
0:01:13
(0:01:29)
0:00:48
(0:00:18)
Daily-Monthly
planner (paper)3
0.19
0.38
0:04:40
(0:04:18)
0:00:50
(0:00:15)
Address books
(paper)3
0.07
0.14
0:01:45
(0:03:04)
0:01:00
(0:00:42)
Phone unit
5.16
10.08
0:26:52
(0:18:23)
0:02:17
(0:00:43)
Cell Phone
0.67
1.31
0:04:53
(0:06:06)
0:04:13
(0:04:24)
All devices4
51.59%
100%
0:44:57
(1:13:27)
0:02:11
(0:01:52)
1Includes using email 2Only seven informants have terminals.
3Only three informants used each of these. 4Weighted average
Table 3: Average time usage per device per person
(hour:min:sec).
Looking closer at the usage of the different information
artifacts, we can clearly see how a short-term duration usage
pattern prevails across different technologies. Table 3 shows
the length of time that people spent using different electronic
devices and paper documents before they were interrupted or
switched to another activity. People spent an average of 2
min. 11 sec. working with any device or paper before they
switched to another device or event.
An ANOVA between the three different roles shows a
significant difference in phone use (F(2,13)=7.09, p<.01). A
Tukey post-hoc analysis (95% confidence interval) shows
that developers (mean=10 min. 13 sec., sd=3 min. 55 sec.)
speak significantly less time on the phone per day than
analysts or managers (combined mean=33 min. 32 sec.,
sd=11 min. 27 sec.). With the exception of PC use described
in Table 2, no other significant differences between roles
were found.
Working Spheres: a practical unit of work
Our data so far conveys the very short-term nature of
interactions with people and devices, and shows how people
constantly switch between different events. In formal
interviews and informal comments, the informants confirmed
this behavior as typical in their day and pointed out that it
does not always mean it is detrimental for their work.
Constant switching at the level of events is not challenging
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
116

per se; it is the switching at higher levels of activity that we
envision can be problematic.
We decided to examine how often people switch between
what we consider to be higher levels of activity, such as
projects, of which events are a subset. Observations gave us
the opportunity to understand the different ways in which
individuals define, delimit and differentiate their activities.
Many times during observations, comments arose about their
working on “the TRK stuff” or the “the R6 spec” or the
“Clear Quest app”, or “Bill is over the Commercial Paper
thing”. We analyzed our data (photocopies, email printouts,
documents, pictures, observation notes, tracking logs, and
transcripts). This led us to emerge with the grounded concept
of working sphere which describes higher levels of units of
work or activities that people divide their work into on a
daily basis.
Type of
WS
Condition
Average
WS/day (sd)
Avg. Time/
WS (sd)
Avg. Time /
segment
(sd)
Default
5.00 (1.90)
0:46:57
(1:02:40)
0:12:52
(0:11:41)
Central
Urgent
1.00 (1.15)
0:39:38
(0:45:45)
0:15:42
(0:11:01)
Default
2.90 (1.76)
0:15:10
(0:22:02)
0:08:53
(0:09:26)
Peripheral
Urgent
0.90 (1.10)
0:11:40
(0:11:03)
0:07:16
(0:07:14)
Metawork
--
1 (0.0)
0:44:29
(0:29:34)
0:06:22
(0:03:59)
Personal
--
1 (0.0)
1:05:13
(0:32:48)
0:48:38
(0:30:13)
Unknown
--
1 (0.0)
1:22:24
(0:55:31)
0:12:00
(0:10:06)
Central
and
Peripheral
WS only
9.81 (3.39)
0:33:32
(0:50:59)
0:11:28
(0:10:54)
All
All
12.81 (3.39)
0:40:41
(0:50:58)
0:13:49
(0:16:14)
Table 4. Avg. no. of working spheres (WS) per person, and avg.
time spent in each WS per day (hour:min:sec).
We define a working sphere as a set of interrelated events,
which share a common motive (or goal), involves the
communication or interaction with a particular constellation
of people, uses unique resources and has its own individual
time framework. With respect to tools, each working sphere
might use different documents, reference materials, software,
or hardware. It is the whole web of motives, people,
resources, and tools that distinguishes it from other working
spheres. Examples of working spheres of the JEB team
include a training effort for new UNIX programmers, an
implementation of a new feature in a compliance module, a
trip to a regional Park for JEB teammates, or the
documentation of modules.
We characterized working spheres in two main ways: with
respect to the importance or centrality that they hold for the
individual, and with respect to the way that they are handled.
Working spheres can be shared with others which means that
the same sphere can have different importance for different
people. We identify two main levels of engagement with a
working sphere: central and peripheral. A working sphere is
central when it is of primary importance for an individual;
otherwise it is peripheral. Working spheres are also handled
in two basic ways: as urgent cases that have to be solved
right away or as default cases that involve expected work
with expected time schedules. The same working sphere can
vary at times between being urgent (when something goes
wrong) or involving its expected state of affairs. We
identified and counted the number of working spheres in
which each of our informants was involved in during the
three days of observation and measured the length of time
they spent in each. Table 4 shows the results.
We identified that each person worked on an average of ten
working spheres per day, during the three days of our
observation. (Note that we observed a three-day “slice of
time”, so it is possible that we did not count all the working
spheres that an individual might have). Table 4 shows that
individuals clearly spend more time in working spheres
which are central to their work. Work in peripheral working
spheres generally involves answering questions or providing
feedback. An individual spent about 33 min. 32 sec. per
working sphere; however it does not occur as a continuous
period of time. Instead the individual attends to a working
sphere in small periods of time that we call segments. People
frequently switch from one working sphere segment to
another. Table 4 shows that the actual average duration of a
working sphere segment is quite short (11 min, 28 sec.). An
ANOVA between analysts, developers, and managers
showed no significant differences in working sphere segment
length.
Condition
Type of W.S.
Av. Time p/ segment
(S.D.)
Default
0:14:13 (0:12:55)
Central
Urgent
0:16:24 (0:12:25)
Default
0:09:03 (0:09:38)
Peripheral
Urgent
0:07:41 (0:07:30)
Both
All
0:12:18 (0:11:53)
Table 5. Avg. time per WS segment without “nonsignificant”
disruptions (hour:min:sec).
Metawork
Individuals spend part of their day on a set of activities that is
not connected with any specific working sphere but rather
related to the management of all of them. We call these
activities metawork. People periodically conduct metawork
throughout the day, which involves coordination, checking
activities, organizing email, organizing their desk at the start
or end of a working day, and catching up with teammates on
what they have missed. People spend an average of 44 1/2
min. per day conducting metawork, and similar to working
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
117

spheres, this work is also conducted in shorter chunks of
about six and a half minutes at any one time (see Table 4).
Generally, we found that individuals engage in metawork
whenever they conclude large activities, or when they return
from a meeting.
Working spheres without “nonsignificant” disruptions
We noticed that our informants, sometimes while working in
one working sphere, were interrupted by work from another
working sphere, switched to the second sphere briefly, and
then resumed work in the first sphere. Examples of this can
be when somebody brings a document to be signed, or when
they get a quick call. We realized that some disruptions are
not significant and would not introduce a large overhead to
resume work. We conducted a further analysis of our data
where we explored the effects of disregarding short
distractions on the segment length. We used a criteria of two
minutes after reviewing the data and decided that this would
make a feasible heuristic to begin with. We then coded the
data also considering how much attention the disruption
required. For example, a disruption from a central working
sphere required more concentration than from a peripheral
working sphere. Thus, a segment was considered as
“continuous” even if people turned to another working sphere
for less than two minutes. Table 5 shows that even after
removing what we consider “nonsignificant” disruptions,
people still spent only a short average time in a working
sphere segment (somewhat more than 12 minutes).
External and internal interruptions
People can switch to a different working sphere or they can
be interrupted. About two-thirds of the time people resume
work in their working sphere after interruption. Most
“significant” interruptions can range from two minutes up to
40 minutes. About one-third of the time people switched to
another working sphere without returning to the first working
sphere. In some cases, for example, when deadlines are
similar, people move back and forth between working
spheres. In general, interruptions could be categorized into
two main groups: Internal and External, following [8]. An
Internal interruption refers to self-initiated switching among
working spheres. An External interruption is a condition in
the environment that motivates switching. In Table 6 we can
see some of the actions that resulted from internal
interruptions (e.g. making a phone call or leaving the cubicle)
as well as kinds of external interruptions (e.g. a person enters
the cubicle).
Our data confirms previous studies indicating that people
interrupt themselves as often as they are interrupted [12]. Our
informants continually switched on their own volition to
another working sphere and most internal interruptions were
due to people leaving their cubicle to interact with other
individuals. With external interruptions, individuals switch
between working spheres more often due to verbal-based
interruptions (such as visitors or phone calls) than to
notification mechanisms from their e-mail or voice mail.
Type
Average
Interruptions
per day (S.D)
% all
types
Internal /
External
Checking/Using
Paper Docs
0.52 (0.86)
1.87
49.11%
Checking/Using
Computer
1.54 (1.47)
10.98
Talking t/wall
1.93 (2.15)
6.89
Phone call
1.14 (1.56)
4.09
Email use
1.04 (1.47)
7.40
Internal
Leaves cubicle
5.00 (2.56)
17.87%
New email notif.
3.55 (3.18)
12.68%
50.89%
Person arrives
6.00 (3.03)
21.45%
Status on terminals
0.36 (0.82)
1.28%
Phone ringing
2.62 (2.01)
9.36%
Voice message light
0.19 (0.45)
0.68%
Call through wall
1.33 (1.75)
4.77%
External
Reminder
notification
0.19 (0.40)
0.68%
Total
25.40 (8.23)
100%
100%
Table 6. Avg. number and types of interruptions per day.
Strategies for maintaining continuity in working spheres
People employ strategies to help them maintain continuity as
they continually switch between their working spheres. Most
of our informants commented that their preference is to work
in a single working sphere until the job is completed.
However, this is rarely the case because interruptions
(internal or external) lead people to switch their attention into
different working spheres or urgent issues arise that lead
people to switch working spheres. People develop strategies
to adjust to the unpredictability of their environment, such as
knowing they will need to respond to urgent requests.
Because frequent interruptions are expected we noticed that
some of our subjects use special artifacts that help them to
prioritize and maintain their attention over their working
spheres. These artifacts function like containers in that they
hold information about central working spheres. The
information included in the artifact plays the role of a
reminder and, as pointed out by Miyata and Norman [8], it
both signals the working sphere to be attended to and
describes with some detail what has to be remembered. The
artifact is often updated across the day with results when
work within a sphere has to be postponed.
The artifact takes on different forms, each providing different
properties in terms of the signaling or descriptive power. A
common implementation uses a special inbox folder in the
email client to contain messages related to central working
spheres to be attended to. Messages are from others or from
the individual herself with summaries of other messages or
other documents. A second form of the artifact used by three
of our informants uses printouts of email messages. People
keep piles on their desks with printouts of email messages or
meeting notices generated by their team’s scheduling system.
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
118

On those printouts are annotations such as clarification notes
or contact information. A third form of artifact relies on more
traditional activity management tools. Two of our informants
used planners extensively to manage their working spheres
usually by integrating conversations or email messages into
the planner. For each day they listed the central working
spheres to cover and if necessary they transferred pending
actions. A fourth form of artifact is the post-it note. One
informant places up to seven notes with references to actions
to be taken in different working spheres. The notes stay on
the monitor shelf for as long as the working sphere lasts and
are annotated to indicate the status of the working sphere. For
instance, while working on a software update for one of his
colleagues, an informant wrote down comments regarding
progress in that working sphere (e.g. “Test PB patch DAN’s
PC –Waiting for AL”).
DISCUSSION AND CONCLUSIONS
Our study confirms what many of our colleagues and
ourselves have been informally observing for some time: that
information work is very fragmented. What surprised us was
exactly how fragmented the work is. In a typical day, we
found that people spend an average of three minutes working
on any single event before switching to another event.
Informal interactions average four and a half minutes each.
Further, people spend on the average somewhat more than
two minutes on any use of electronic tool, application, or
paper document before they switch to use another tool. The
longest duration of tool use is with PCs, yet this averages
only slightly more than three minutes at any one time. One
informant’s quote used in the title of this paper captures the
extent of this fragmentation of work.
We argue that it makes more sense to understand how time is
distributed among working spheres, activities that are
thematically connected for the individual. A working sphere
consists of a string of events. We found that working spheres
are also highly fragmented: people spend on the average
eleven and a half minutes in continuous work on a project or
theme before they switch to another. After removing what we
considered to be “nonsignificant” disruptions, we found that
these segments of time were not much longer, averaging
somewhat longer than 12 minutes.
People interrupt their work themselves (internal
interruptions) about as much as they are interrupted by
external influences. Most interruptions are due to face-to-face
interactions, similar to what O’Conaill and Frohlich [9]
found. People are just as likely to interrupt themselves as to
be interrupted by external sources. An interesting area for
future research is to understand the reasons that lead
information workers such as these to interrupt themselves so
frequently.
We examined the work of people in three different roles:
analysts, developers, and managers and found (except for PC
use) that no significant differences existed in the
fragmentation of their work. Our work expands on the past
studies of time distribution [4,5,7,11,12] that looked only at
managers’ work. We discovered that switching between
multiple spheres of work is more pervasive.
Our results indicate that throughout their day, individuals are
constantly moving from one topic to another and managing
information streams from a myriad of sources. Our results
build upon Sproull’s claim [12] that managers are multi-task
processors. We provide empirical data to show that people
mostly spend very little continuous time on any single
activity. We found that our informants’ work can be
described more specifically not as multi-task processing, but
as requiring attentional resources to constantly change
between different events, tools, and working spheres.
The three kinds of strategies that we found to manage
different working spheres share some characteristics. First,
they result from an explicit effort to help organize the work
that has to be done. Information related to different working
spheres is aggregated into a single type of artifact. Second,
the prioritization of working spheres is not something that is
explicitly indicated in the artifacts. The order in which
information is listed or organized does not express an order
of execution or importance. People prefer to talk of those
items contained in the artifact as important items that need to
be done. A third characteristic is that the artifacts always
appear in a visible spot of the working space so they can be
consulted constantly. For printouts and post-it notes,
visibility is achieved by placing them within the normal field
of view (e.g. nearby the keyboard). The inbox folders are
located where they can be seen whenever email is used. For
instance one of our subjects labeled his inbox “Z Immediate
Attention Issues” so he would have it at the end of all his
alphabetically-sorted inboxes. A final important
characteristic is that individuals leave the artifacts only while
the working sphere is active. When the working sphere is
finished, items are transferred to other archives.
Implications for information technology design
Current information technology is designed to support
individual events such as word processing or e-mail use
rather than to provide mechanisms to integrate the multiple
information objects required by some working spheres. The
design of information technology needs to consider how
information workers switch constantly among working
spheres. Mechanisms should be flexible to enable people to
group particular documents and applications but at the same
time there should be recognition that many applications are
shared among working spheres (e.g. an e-mail client or
scheduling tools). In particular, when working spheres use
very different information resources and applications, it will
be convenient to have mechanisms that save the state of the
information device particular to that working sphere, making
it easier to resume work. In contrast with a similar idea
proposed by Bannon [1], we argue that we need to re-
examine the role of artifacts such as notepads, post-its or
planners and derive hybrid designs which include
computerized and non-computerized tools.
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
119

We noticed that in spite of having a commercial task
management tool, our informants did not use its functionality
to support their work. Perhaps the most important reason is
the fact that they need to have their list of current tasks
always visible, which is not possible with the tool because
the window with their current activity is usually maximized.
The artifacts developed by the JEB team members are useful
because they are visible and available. To continually show
the working sphere status requires extending an interface
beyond that provided by the personal computer and points
toward designs that include peripheral independent displays
(or PDAs), which through wireless connections can provide
portability. People also prefer more open and flexible forms
of managing their activities which fit well with their forms of
communication. Given the preponderance of email
communication, the integration of reminder artifacts with
email should be achieved very easily. Systems such as
Taskmaster [2] (and the concept of thrask) are in line with
the kind of support required for working sphere management.
However, it is important to recognize that not all information
related to a working sphere is delivered by email and
consequently a system must include ways to integrate
information from other sources.
Limitations of the study and generalizability of results
Our study has several limitations. The challenge of the
“shadowing” technique is that the researcher herself cannot
be a source of interruptions. To avoid this we kept our
questions for the end of the day, or when walking with
informants to attend meetings. In spite of our best efforts,
15.81% of the events could not be matched with any working
sphere and were consequently put in an Unknown category.
We believe that given the nature of the observations and the
fast-pace environment we observed, the percentage of events
that we clearly identified within working spheres is
acceptable to support our findings.
Our observations are limited to one fieldsite. This is also true
of other studies of time distribution [4,5,7,10,11,12]. It is
very possible that factors unique to the organization affect the
constant shifting, such as time pressures, and the high value
of the product. We would need to investigate other
organizations to understand more completely how
organizational factors affect the management of multiple
activities. We also only observed 14 people. This is a higher
number observed than Hudson [5] and Sproull [12]. Perlow
[11] observed 17 software developers. We therefore cannot
claim that our sample represents a wide range of information
workers. We believe though that the kind of fast-pace
environment that we studied is widespread. Yet more
research is needed to investigate the work roles and kinds of
organizations where such work is typical.
ACKNOWLEDGEMENTS
We would like to thank the members of the JEB team at ITS,
and Susan Knight. This research was supported by the
National Science Foundation under grant no. 0093496, by the
Center for Research on Information Technology and
Organizations (CRITO), and UC MEXUS grant no. 32080.
REFERENCES
1. Bannon, L., Cypher, A., Greenspan, S., and Monty, M.,
Evaluation and analysis of users’ activity organization.
Proceedings of CHI 1983, ACM Press (1983), 54-57.
2. Belloti, V., Ducheneaut, N., Howard, M., and Smith, I.,
Taking email to task: The design and evaluation of a task
management centered email tool. Proceedings of CHI
2003, ACM Press (2003), 345-352.
3. Gillie, T. and Broadbent D., What makes interruptions
disruptive? A study of length, similarity and complexity.
Psychological Research, 50, (1989), 243-250.
4. Horne, J.H. and Lupton, T., The work activities of
‘middle’ managers - An exploratory study. The Journal
of Management Studies, 2, (1965), 14-33.
5. Hudson, J.M., Christensen, J., Kellogg, W.A. and
Erickson, T. “I’d be overwhelmed, but it’s just one more
thing to do:” Availability and interruption in research
management. Proceedings of CHI 2002, ACM Press
(2002), 97-104.
6. Mintzberg, H., Structured observation as a method to
study managerial work. The Journal of Management
Studies, 7, (1970), 87-104.
7. Mintzberg, H., The Nature of Managerial Work. Prentice
Hall, Englewood Cliffs N.J. USA, 1973.
8. Miyata, Y. and Norman, D.A., Psychological issues in
support of multiple activities, in User Centered System
Design, D.A. Norman and S.W. Draper (Eds). Lawrence
Erlbaum, Hillsdale, N.J., USA (1986), 265-284.
9. O’Conaill, B. and Frohlich, D. Timespace in the
workplace: Dealing with interruptions. Proceedings of
CHI 95, ACM Press (1995), 262-263.
10. Panko, R., Managerial communication patterns. Journal
of Organizational Computing, 2,1, (1992), 95-122.
11. Perlow, L.A., The time famine: Toward a sociology of
work time. Administrative Science Quarterly, 44,
(1999), 57-81.
12. Sproull, L.S., The nature of managerial attention.
Advances in Information Processing in Organizations, 1,
(1984), 9-27.
13. Strauss, A. and Corbin J., Basics of Qualitative
Research, Sage, Thousand Oaks, CA. USA, 1988.
CHI 2004 ׀ Paper 24-29 April ׀ Vienna, Austria
Volume 6, Number 1
120