Content uploaded by Deborah Boehm-Davis
Author content
All content in this area was uploaded by Deborah Boehm-Davis
Content may be subject to copyright.
The Effect of Interruption Duration 1
Running head: INTERRUPTION DURATION AND DEMAND
The Effect of Interruption Duration and Demand on Resuming Suspended Goals
Christopher A. Monk
George Mason University, Fairfax, VA
J. Gregory Trafton
Naval Research Laboratory, Washington, DC
Deborah A. Boehm-Davis
George Mason University, Fairfax, VA
Address Correspondence to: Christopher A. Monk
Department of Psychology
George Mason University
4400 University Drive, MSN 3F5, Fairfax, VA 22030-4444
703-993-3408 phone
703-993-1330 fax
E-mail: cmonk@gmu.edu
The Effect of Interruption Duration 2
Abstract
The time to resume task goals after an interruption varied depending on the duration and
cognitive demand of interruptions, as predicted by the memory for goals model (Altmann &
Trafton, 2002). Three experiments using an interleaved tasks interruption paradigm showed that
longer and more demanding interruptions led to longer resumption times in a hierarchical,
interactive task. The resumption time profile for durations up to one minute supported the role of
decay in defining resumption costs, and the interaction between duration and demand supported
the importance of goal rehearsal in mitigating decay. These findings supported the memory for
goals model, and had practical implications for context where tasks are frequently interleaved
such as office settings, driving, emergency rooms, and aircraft cockpits.
Keywords: Interruption; Goals; Interleaved tasks; Memory
The Effect of Interruption Duration 3
The Effect of Interruption Duration and Demand on Resuming Suspended Goals
For most people, dealing with interruptions is not a problem to be overcome as much as it
is an inevitable part of life. In fact, the ability to “multitask” is considered a desirable job skill by
many employers, which is not surprising given that, on average, workers shift between tasks
every three minutes (Gonzalez & Mark, 2004). By shifting between tasks every few minutes, it
appears that people are managing interruptions by interleaving them with their primary tasks. For
example, many people engage in conversations through instant message applications while
working on other projects on the computer. The need to understand how interruptions and
multitasking behaviors impact performance in the workplace has spawned several studies in
recent years (e.g., Czerwinski, Horvitz, & Wilhite, 2004; Iqbal & Horvitz, 2007; McFarlane &
Latorella, 2002).
A study that investigated the use of instant message communications in the workplace
found that conversations lasted nearly 4.5 minutes on average, with exchanges every 15 seconds
or so (Isaacs, Walendowski, Whittaker, Schiano, & Kamm, 2002). The study also showed that
workers who heavily used instant messaging covered multiple topics in an exchange, and
frequently shifted attention to other tasks while conversing. Avrahami and Hudson (2006) found
that 92% of messages were responded to within 5 minutes, with 50% responded to within 15
seconds. They also found that students and interns used instant messaging at about double the
rate of researchers in the sample.
The impact of interruptions is not merely an inconvenience for people going about their
work and home lives. Interruptions can have devastating consequences. Multiple plane crashes
have been attributed in part to interruptions to the pre-flight checklists pilots perform prior to
take-off (NTSB, 1969, 1988). Studies have also shown that interruptions can affect driving safety
The Effect of Interruption Duration 4
(Monk, Boehm-Davis, & Trafton, 2004) and emergency room care (Chisholm et al., 2000;
Chisholm et al., 2001). Given the prevalence of interruptions and their potential for harmful
consequences, it is not surprising that researchers have turned their attention to understanding
how people perform when interrupted.
Although interruptions research dates back to the 1920’s when Zeigarnik (1927) reported
that people recalled details of interrupted tasks better than uninterrupted tasks, there was a long
gap in experimental studies of interruptions until those conducted by Kreifeldt and McCarthy
(1981) and Gillie and Broadbent (1989). The Kreifeldt and McCarthy and Gillie and Broadbent
studies concluded that people performed post-interruption tasks more slowly compared to pre-
interruption performance. They also found that people made more errors in post-interruption
performance; results corroborated initially by Cellier and Eyrolle (1992) and later by Zijlstra,
Roe, Leonova, & Krediet (1999).
Subsequent interruptions studies primarily focused on determining the characteristics that
make interruptions disruptive (see McFarlane & Latorella, 2002 for a comprehensive review).
Several characteristics have been shown to affect primary task performance, including task
similarity to the primary task (Cellier & Eyrolle, 1992; Czerwinski, Chrisman & Rudisill, 1991;
Edwards & Gronlund, 1998; Oulasvirta & Saariluoma, 2004), interruption complexity (Cades,
Trafton, Boehm-Davis, & Monk, 2007; Gillie & Broadbent, 1989; Hodgetts & Jones, 2006a;
Zijlstra, et al., 1999), the relatedness of the primary and interruption tasks (Cutrell, Czerwinski,
& Horvitz, 2000; Zijlstra, et al., 1999), control over interruption onset (McFarlane, 2002), and
the availability of primary task retrieval cues (Cutrell, Czerwinski & Horvitz, 2000; Czerwinski,
Cutrell & Horvitz, 2000). Unfortunately, some of the findings in these studies have been
contradictory. For example, some found that interruptions slowed down performance on the
The Effect of Interruption Duration 5
primary task (Gillie & Broadbent, 1989), and some found that performance was faster when
interrupted (Zijlstra et al., 1999). Speier, Valacich, and Vessey (1999) found that decision-
making on simple tasks was aided by interruptions, but hindered for complex tasks. Recent
evidence suggests that primary task performance is not the only victim of interruptions;
secondary task performance can suffer in addition to primary task performance (Einstein et al.,
2003; McFarlane, 2002).
These studies are useful for understanding how people might better deal with
interruptions to work more efficiently, but they lack a cohesive theoretical approach to
understand how people manage the multiple and temporary goals that result from interruptions.
Altmann and Trafton (2002) introduced such a theoretical model for memory for goals that is
particularly suited for the study of interruptions. This model has been tested in multiple
interruptions studies (Altmann & Trafton, 2007, Hodgetts & Jones, 2006a; Hodgetts & Jones,
2006b; Li et al., 2006; Monk, Boehm-Davis, & Trafton, 2004; Trafton, Altmann, Brock, &
Mintz, 2003), and it is the basis for the predictions for interruption recovery in this study.
Memory for Goals
Altmann and Trafton’s (2002) memory for goals model is a formal model of goal
encoding and retrieval in memory. In their work, Altmann and Trafton successfully applied this
model to simulating reaction time and error data from the Tower of Hanoi, a task that depends
heavily on suspension and resumption of goals during problem solving. The suspension and
resumption of goals is a fundamental aspect of interrupted task performance. For example, a
person’s current “train of thought” (primary task goal) when writing a report must be halted or
suspended when an instant message arrives from an important source. As with all conversations,
there is turn taking in instant message conversations that allows the person to return attention to
The Effect of Interruption Duration 6
the report-writing task while waiting for a response. With each shift of focus, the person must
regain the suspended train of thought to resume writing the report. Because the model was
developed to handle such suspended and resumed goals, it is well suited to predict the impact of
interruptions on primary task resumption.
The memory for goals model is based on the activation model of memory items and is
instantiated within the ACT-R cognitive architecture (Anderson, 1993; Anderson & Lebiere,
1998; Anderson et al., 2004). The fundamental processing assumption in this theory is that when
central cognition queries memory, the chunk that is most active at that instant is returned.
Returning to the example above, the writer’s current goal or action is that with the highest level
of activation at that moment in time. It is this goal that directs behavior (Anderson & Lebiere,
1998; Newell, 1990). Altmann and Trafton (2002) used an adapted version of ACT-R’s Base
Level Learning Equation (Anderson & Lebiere, 1998) to determine levels for goal memories.
Within the ACT-R framework, a memory element’s base-level activation represents its activation
without any associations or cues (Anderson & Lebiere, 1998; Lovett, Reder, & Lebiere, 1999). A
goal’s retrieval history plays a significant role in its activation level, and therefore when it directs
behavior (see Altmann & Trafton, 2002 for a detailed explanation). Frequently sampled goals
will have higher levels of activation, as will recently encoded or retrieved goals. For example,
the report writer in the above example will have more success in resuming a suspended train of
thought if it was the focus of attention just before the interruption (recency) or for long periods
before the interruption (frequency). The activation time course for a goal is depicted in Figure 1.
The effect of interruptions on task performance can be examined with the memory for
goals framework (Altmann & Trafton, 2002) as a theoretical explanation of the determinants of
goal activation, and therefore behavior. For example, when a goal is interrupted by another goal
The Effect of Interruption Duration 7
the original goal memory will immediately begin to suffer activation decay (assuming that the
interrupting task engages the cognitive resources that would otherwise be used to rehearse such
information). The time required to resume the suspended goal after the interruption is directly
related to its level of activation (Altmann & Trafton, 2002). Goals that have been suspended for
longer periods will have decayed to lower activation levels and therefore will take longer to
resume, assuming no intervening rehearsal. In other words, the report writer will have greater
difficulty resuming the suspended train thought when the instant message conversation persists
for longer periods without opportunity to shift focus back to the report. Therefore, the memory
for goals model predicts that longer interruptions should result in longer times to resume the
primary task (or goal). Hodgetts and Jones (2006a) recently demonstrated support for this
prediction.
Interruption Duration
Interruption duration has produced mixed results in the literature. Earlier studies that
manipulated duration failed to show an effect (Gillie & Broadbent, 1989; Einstein et al., 2003; Li
et al., 2006). Recently, Hodgetts and Jones (2006a) were successful in finding an interruption
duration effect using a Tower of London task when testing predictions from Altmann and
Trafton’s (2002) memory for goals model. Participants were prompted after three moves to click
on a ‘mood’ button appearing at the bottom of the computer display to open a mood checklist
task (interruption). The interruptions lasted either 6 or 18 seconds, consisting of one or three
non-repeated mood checklists. In addition, the ‘mood’ button indicated the length of interruption
by noting the number of checklists to be completed. Resumption times (time to make the next
move on the Tower of London task after the interruption) were longer in the 18-second
interruption condition, however there was no effect for knowing the length of interruption in
The Effect of Interruption Duration 8
advance. This evidence from Hodgetts and Jones was the first empirical support of Altmann and
Trafton’s predictions for greater retrieval times for goal memories suspended for longer periods.
Despite Hodgetts and Jones’s (2006) findings, other studies failed to find a duration
effect. In their landmark interruptions study, Gillie and Broadbent (1989) conducted a series of
experiments investigating why some interruptions are more disruptive than others. In the first
two experiments, participants performed a computer game task that required them to navigate
through an environment and “pick-up” objects from a memorized list. Interruptions occurred
after designated objects were picked-up and consisted of simple arithmetic problems. The first
experiment used a 30-second interruption and the second experiment used a 2.75-minute
interruption; neither duration resulted in post-interruption task performance decrements. The
authors claimed, “the length of an interruption on its own does not seem to be the critical factor
in determining whether or not it will prove disruptive” (p. 246).
Recent research in the prospective memory domain provided additional data regarding
the interruption duration question (Einstein et al., 2003; McDaniel et al., 2004). Einstein et al.
(2003) found that people were able to maintain intentions over brief intervals ranging from 5 to
40 seconds. In a subsequent study, McDaniel et al. (2004) added a manipulation of interruption
duration in the 40-second intention execution delay condition. They compared 10- and 20-second
interruption durations. The results once again showed no effect for intention execution delay, nor
did they reveal an effect for interruption duration. McDaniel et al. argued that a maintenance
rehearsal explanation should have resulted in a decline in prospective memory performance for
the longer interruption; however, this prediction was not supported. Interestingly, the digit
monitoring interruption task used by Einstein et al. (2003) and McDaniel et al. probably reduced
participants’ ability to rehearse intentions during the interruption. These prospective memory
The Effect of Interruption Duration 9
findings were yet more evidence that contradicted the predictions of the memory for goals
model.
The challenge for the Altmann and Trafton (2002) model was to explain why one of its
fundamental predictions for interrupted task performance had not been supported in the
interruptions literature until recently (Hodgetts & Jones, 2006a). A review of the literature
revealed two reasons for the failure to find consistent evidence for an interruption duration
effect. First, the measures used in many interruptions studies were global, and therefore
insensitive to the effects associated with goal resumption. For example, Gillie and Broadbent
(1989) compared pre- and post-interruption task times and error rates, which did not address the
time participants required to resume the task after being interrupted. Czerwinski et al. (2000)
measured total task time, the time to respond to the interruption notification, and the time spent
on the interruption notification. Zijlstra et al. (1999) similarly measured task times and total
interruption time, in addition to other performance measures. Other interruptions studies used
measures like error rates in primary task performance (Cellier & Eyerolle, 1992; McFarlane,
2000; Oulasvirta & Saariluoma, 2004), decision-making performance (Speier et al., 1999; Speier,
Vessey, & Valacich, 2003), and proportion of correct prospective memory responses (Einstein et
al., 2003; McDaniel et al., 2004). The lack of sensitive measures for how quickly people resume
the primary task after the interruption may have been one of the key reasons why previous
studies failed to find an effect for interruption duration. It was not until Hodgetts and Jones
(2006a) implemented Altmann and Trafton’s resumption lag measure that evidence for the effect
materialized. As a result, the resumption lag measure, which is a response time measure
capturing the time required to resume a goal, was adopted for the present experiments. The intent
The Effect of Interruption Duration 10
was to capture the changes in resumption time using the resumption lag measure as predicted by
the memory for goals model.
The second reason why past research failed to find a consistent effect for interruption
duration was the manipulation of interruption duration. The interruption duration effect predicted
by the memory for goals model (Altmann & Trafton, 2002) occurs when goals are still in the
initial stages of decay. Although the 30-second and 2.75-minute interruptions used by Gillie and
Broadbent (1989) seemed reasonable in terms of face validity, these interruption durations may
have masked resumption time effects. The memory for goals decay function (see Figure 1)
indicates that the rate of decay slows down dramatically over time, and therefore if a goal had
reached asymptotic levels of activation decay after 30 seconds, then the activation level for a
goal suspended for more than 2 minutes would be similar, assuming the same level of initial
activation. Therefore, the only way to detect the predicted effect was to use much shorter
interruptions like the 6 and 18 second interruptions used by Hodgetts and Jones (2006a).
Interruption Demand
Because the theoretical explanation for the duration effect focuses on goal memory
decay, the issue of goal rehearsal must also be addressed. Goals left unrehearsed during an
interruption will decay, resulting in longer resumption times (Altmann & Trafton, 2002;
Hodgetts & Jones, 2006a). However, there are many interruption tasks that afford opportunities
to rehearse the suspended goal. For example, the report writer could make quick glances to the
document in the text editor when waiting for responses in the instant message exchange. These
quick “reminder” glances to the report would help maintain the writer’s train of thought for when
the instant message conversion concludes. Alternatively, the writer may engage in an instant
message conversation wherein a long, detailed response is made over several seconds, preventing
The Effect of Interruption Duration 11
any glances to the open report. In this scenario, the suspended thought would be difficult to
resume without recreating the thought processes by reading the previous report entry. Therefore,
it follows that interrupting tasks that prevent or inhibit goal maintenance should result in
unmitigated goal decay manifested as longer resumption times. Alternatively, interrupting tasks
that allow people to rehearse their suspended goals should show shorter resumption times in
comparison (see Trafton, et al., 2003).
The literature also suggested that ability to rehearse during an interruption is influenced
by the cognitive demand of the interrupting task. Gillie and Broadbent (1989) showed that
additional decoding requirements to an arithmetic task resulted in worse primary task
performance. Zijlstra et al. (1999) found that document editing tasks resulted in more time to
“reorient” to the primary editing task compared to interruptions consisting of unrelated menial
tasks such as looking up a phone number. Recently, Cades et al. (2007) showed that 1-back and
3-back versions of the n-back task resulted in longer resumption times than a shadowing
interruption task. Finally, Hodgetts and Jones (2006a) also manipulated interruption task
complexity with a single digit addition task (simple) and a double-digit addition task requiring
carrying (complex). They found that both the simple and complex interruptions resulted in longer
resumption times than the no-task interruption condition. These studies provided evidence to
support an effect for interruption complexity, both at the more sensitive resumption time measure
and at the more global task measures as well.
Although the interruptions literature has generally supported an effect for interruption
task complexity, the term complexity has been inconsistently defined. Gillie and Broadbent
(1989), Hodgetts and Jones (2006a), and Cades et al. (2007) all used a processing requirements
definition for complexity. These manipulations were consistent with the definition by Byrne and
The Effect of Interruption Duration 12
Bovair (1997), who noted that a number of characteristics appear to determine complexity,
including the number of actions to be performed, the difficulty of executing those actions, the
number of subgoals to be remembered, and the amount of information to be managed and
maintained. For the purposes of this series of experiments, we used the term “demand” rather
than complexity because it referred more directly to the processing demands on working memory
that prevented or allowed rehearsal of suspended task goals.
Because the manipulation of rehearsal was complicated, it was assumed that the cognitive
demand of an interruption task was directly related to the amount of resources available for
rehearsal. In other words, more demanding interruption tasks would leave few resources, if any,
for goal maintenance or rehearsal. Accordingly, working memory processing demands were
varied in the interruption task in attempt to manipulate the available resources for rehearsal.
Overview of the Experiments
The present set of experiments was designed to test predictions from Altmann and
Trafton’s (2002) memory for goals model regarding the resumption of suspended memories (i.e.,
task goals). The fundamental prediction addressed by this set of experiments was that memory
for task goals decays over time, resulting in longer resumption times for those task goals that
have had more time to decay. The resumption lag measure introduced by Altmann and Trafton
has shown to be sensitive to differences in task goal resumption times due to interruption
complexity, duration, the interval between interruption alert and engagement (interruption lag),
and cues in previous research (see Hodgetts & Jones, 2006a; Hodgetts & Jones, 2006b; Monk,
Boehm-Davis, & Trafton, 2004; Trafton, et al., 2003). Multiple and frequent interruptions within
each trial characterized the interleaved tasks interruption paradigm used in this set of
experiments. Although the majority of past interruptions studies used few interruptions per trial,
The Effect of Interruption Duration 13
the present focus was on interleaved interruptions similar to the instant messaging example.
However, the tasks used in these experiments were not intended to simulate instant message
interruptions. Instead, the intention was to use computer-based tasks that could be used to create
situations in which interrupting tasks were interleaved with the primary task. In addition, the
focus on resuming suspended task goals required a primary task with many subgoals that users
typically perform linearly. A VCR programming task was selected because it has served this
purpose well in past research (Monk, Boehm-Davis, & Trafton, 2004).
The first experiment tested the prediction that longer interruptions lead to longer
resumption times. The second experiment attempted to replicate the findings from Experiment 1
while extending the interruption duration manipulation to further characterize the decay trend for
suspended goals. Finally, the third experiment added levels of task demand to the original
interruption duration manipulation to test the rehearsal explanation of the duration effect.
Experiment 1
The first experiment was designed to test the predictions of Altmann and Trafton’s
(2002) model regarding time to resume suspended task goal memories using the interleaved tasks
interruption paradigm. For the purposes of this experiment, goals were defined as low-level, next
action goals. For example, the memory of what button to click next in a computer interface
would be the suspended goal during an interruption or shift in attention. Consistent with the 6
and 18 second interruption durations used by Hodgetts and Jones (2006a), Experiment 1 used
interruption durations of 3, 8, and 13 seconds. In addition, Experiment 1 included uninterrupted
control trials to assess the effect of interruptions on primary task performance. First, the general
disruptiveness of interruptions was predicted to be evident in longer resumption lags compared
to the average time between uninterrupted clicks (called inter-action intervals). Second, the
The Effect of Interruption Duration 14
resumption times were predicted to increase from 3 to 13 seconds. Although the memory for
goals model predicted a log function for resumption times over increased interruption durations,
the segment of the function captured between 3 and 13 seconds was expected to appear linear.
Method
Participants
Twelve students from George Mason University received partial course credit for
participating in this study. The participants (5 men and 7 women) ranged in age from 17 to 32,
with an average age of 20 years.
Tasks and Equipment
The primary task was a VCR programming task using a simulated VCR built in
Macintosh Common Lisp (Gray, 2000; Gray & Fu, 2001). The interruption task was a pursuit-
tracking task that required subjects to track a moving target. These tasks were presented side-by-
side on a Macintosh G4 computer with a 17-inch VGA monitor. The VCR task was displayed on
the left side of the monitor and the tracking task on the right side. Participants programmed show
information into the VCR for randomly selected intervals between 3, 5, or 7 seconds at a time.
The random VCR times were used to prevent participants from predicting the onset of
interruptions. The VCR task was interrupted by the tracking task for 3, 8, or 13 seconds,
alternating back and forth until the VCR program show was completed. Both tasks required only
the computer mouse for input, and only one of the tasks was visible at a time.
VCR task. Programming a show consisted of four sub-tasks: entering the show’s start-
time, end-time, day of week, and channel number. The VCR interface can be seen in Figure 2.
All interactions with the VCR were based on simulated VCR buttons; there were no field entries.
To enter the start-time, the participant first clicked the column button above the hour buttons
The Effect of Interruption Duration 15
(leftmost square button under the Enter button). The participant then clicked the start-hour
button, before clicking on the up or down arrow multiple times until the displayed hour number
reached the target. Next, the participant clicked on the enter button to save the start-hour setting.
Finally, to end this subtask, the participant clicked the column button again (to “deselect” it)
before moving onto the next subtask. The same process was completed for each subtask element
of the end-time, day of week, and channel number tasks, respectively. The VCR display was
blank when no setting was selected. The participants had access to target show information (the
show name, start-time, end-time, day of week, and channel number) at all times as the
information was posted to the right of the monitor on a 3x5-index card.
Interruption task. The pursuit-tracking task required the participant to track an airplane
symbol (target) moving around the right half of the screen. The target’s movement algorithm
randomly updated each of the x and y coordinates by no more than 100 pixels (either direction)
at a rate of 10 Hz. The resulting movement was rapid and somewhat erratic. The airplane
symbol’s visual angle was estimated at .37 degrees high by .79 degrees wide. The circle that
corresponded to the participant’s mouse position was estimated at .97 degrees of visual angle.
Design
A single factor repeated measures design was used to test the interruption duration
hypothesis. There were three interruption durations, 3-, 8-, and 13-seconds, which were varied
between trials. Each participant completed two trials for each interruption duration, resulting in
six interruption trials. In addition, each participant completed six uninterrupted trials that served
as a comparison for the interrupted trials for determining the magnitude of any disruption effect,
for a total of 12 experimental trails. The dependent measure was the resumption lag after each
interruption, as measured by the time from the switch from the tracking task to the VCR task
The Effect of Interruption Duration 16
until the participant’s first click on a button in the interface. Participants tended to establish a
consistent sequence of operations or “path” when programming the VCR. Therefore, resumption
errors were identified as those clicks that deviated from the expected next-action based on each
participant’s established path through the task. Tracking task and resumption error performances
were also recorded to assess any potential speed-accuracy trade-off in performance. Trial order
was randomized and balanced with a Latin-square.
Procedure
Each participant was tested individually. The sessions, which lasted approximately one
hour, began with the experimenter explaining the VCR task through demonstration. The
participants were then given two practice trials where they programmed the VCR without
interruption, followed by two 60-second practice trials with the tracking task alone. The
participants were then introduced to the interruption condition, where they alternated performing
the VCR and interruption tasks within a trial. The participants were instructed that the cursor
position for each respective task would be repositioned to its saved location upon each switch so
that dragging the mouse back and forth between the two sides was unnecessary. Accordingly, the
cursor position, along with various state indicators in the VCR interface (e.g., column button in
“selected” state), acted as environmental cues that aided resumption. Participants were instructed
to treat both tasks as equally important, and to focus on the task that was “on” at any given
moment. Because the trials began and ended with the VCR task, there was implicit emphasis on
this task as the primary task. After the two practice interruption trials, the participants completed
the 12 experimental trials, each with new show information to be programmed. Participants
began each trial with the VCR programming task. After completing the experimental trials, the
participants were debriefed and dismissed.
The Effect of Interruption Duration 17
Results and Discussion
The resumption lag data were screened for errors to isolate VCR task actions that
represented successful post-interruption goal resumption. There were two categories of
resumption errors. The first category consisted of resumption actions that deviated from the
participant’s established task path. Due to the nature of the VCR task, participants generally
performed the task in the same sequence of actions across all trials. This reliable behavioral
pattern provided a definition of path-deviation for each participant. The second error category
consisted of resumption lags less than 100 ms, which were assumed to be due to incidental
mouse-clicks timed coincidentally with the VCR task onset. Both resumption error categories
were eliminated from the data. The path-deviation resumption error rate was 5.3%. Table 1
shows that the error rate was lowest in the 3-second condition, but consistent between the 8- and
13-second conditions. Repeated measures ANOVA revealed no significant difference between
the error rates across the three conditions, F(2, 22) = 1.17, p = .33. There were no resumption
lags less than 100 ms.
Data were lost for one participant’s second 8-second interruption trial so five values from
the population of 8-second trial resumption lags were randomly selected and imputed for the
missing cell. The following results represent the mean values of the five calculations and
analyses for each imputation (see McKnight et al., 2006 for missing data procedures).
To first demonstrate the presence of a disruption effect for interruption trials, the
resumption lags for the interrupted condition were compared with randomly sampled inter-action
intervals (IAIs) in the uninterrupted condition. The IAIs were the time elapsed between interface
actions or button clicks, and were viewed as an appropriate comparison for the resumption lags
to quantify the relative disruptive effect of interruptions in the VCR task. By comparing the
The Effect of Interruption Duration 18
mean resumption lags (M=1,548 ms, SD=231) to the mean IAIs for the uninterrupted trials
(M=949 ms, SD=283), it was evident that the interruptions resulted in a delay in the execution
time for the next action or goal in the interface compared to when uninterrupted. Subtracting the
mean IAI from the mean resumption lag reveals an estimated cost of 599 ms on the VCR
programming task. These data showed that resumption lags were longer than inter-action
intervals for uninterrupted trials, indicating the basic interruption disruption effect on task
resumption with the interleaved tasks interruption paradigm.
To test the interruption duration prediction, the resumption lags from the interrupted trials
were entered into a single-factor repeated measures ANOVA. Recall that the following is the
mean F-value of the five ANOVAs corresponding to each randomly imputed value as detailed in
the previously described missing data procedures (McKnight et al., 2006). The main effect for
interruption duration was significant, F(2, 22) = 10.92, p < .01,
η
p2 = .50. As can be seen in
Figure 3, the resumption lags increased from 3- to 8- to 13-second interruptions (b = 30.72),
thereby supporting Altmann and Trafton’s goal decay predictions. This finding was consistent
with the interruption duration results by Hodgetts & Jones (2006a) and therefore represented a
significant addition to the growing body of empirical support for the memory for goals model as
a framework for studying interrupted task performance (Altmann & Trafton, 2002; Trafton et al.,
2003). T-tests showed that the 3-second condition was reliably shorter than the 8- and 13-second
conditions (both p < .01), and that the 8-second condition was shorter than the 13-second
condition (p < .05).
The x-y coordinates for the mouse and target positions were recorded at 10 Hz during the
tracking task. The Euclidean distance between the target and mouse positions was calculated for
each sampling record. The root mean square (RMS) of the distance calculations was used as a
The Effect of Interruption Duration 19
measure of accuracy. The first second of data (i.e., the first 10 data points) after each switch to
the tracking task were excluded due to high variability while the participants readjusted to the
tracking task. The RMS scores were averaged across the tracking task switches within trials, and
again for participants within interruption duration conditions. The data were trimmed using a cut-
off of three standard deviations above the mean. With this criterion, 2% of the tracking data were
excluded. Tracking task performance was significant for the interruption duration, F(2, 22) =
15.90, p < .01,
η
p2 = .59. Table 2 shows that RMS was higher for the 3-second condition, though
a Tukey HSD post-hoc analysis only revealed significant difference between the 3- and 13-
second conditions (p < .05). The worse performance in the 3-second condition was likely due to
less time for tracking performance to stabilize compared to the longer durations.
Taken together, the relationship between interruption duration and resumption time from
this experiment and the duration effect found by Hodgetts and Jones (2006a) provided strong
evidence for the existence of an interruption duration effect despite the null findings of duration
in previous interruptions studies (Gillie & Broadbent, 1989; Li et al., 2006; McDaniel et al.,
2004). Without question part of the reason for the discrepancy was tied to the specific tasks in
the different studies, but there were two reasons to suggest why both the present experiment and
Hodgetts and Jones detected an effect of duration whereas past studies had not. The first reason
was connected to using the resumption lag measure, which was appropriately sensitive to the
theoretical predicted outcomes. The second reason was linked to the manipulation of interruption
duration. Gillie and Broadbent were perhaps the least likely to detect an effect of interruption
duration because they used longer durations (minimum 30 seconds) and global measures.
McDaniel et al. manipulated delays in the shorter duration range (5 and 15 seconds), but they
used global measures of intention execution. Li et al. found a trend for more post-completion
The Effect of Interruption Duration 20
errors with a 45-second interruption compared to a 15-second interruption; however, this
difference was not reliably different. Our results, combined with those from Hodgetts and Jones
(2006a), suggested that the interruption duration effect was best detected with the resumption lag
measure and durations less than 15 seconds.
An unanswered question was if the resumption lag trend would continue to increase
linearly with longer interruption durations, or if the trend would resemble a log function as
predicted by the memory for goals model. Absent additional strengthening (e.g., rehearsal)
during an interruption, a suspended goal’s activation level should decay as a function of delay
(see Figure 1). As an indicator of goal memory activation, the resumption lag trend should be
characterized as an inverse of the decay function, rapidly climbing in the shorter duration range
(i.e., the duration effect) before approaching asymptote. Experiment 2 was designed to test this
prediction.
Experiment 2
Experiment 2 was designed with two objectives in mind: To replicate the interruption
duration effect in Experiment 1 and to extend the resumption lag profile beyond 13 seconds to
nearly 1 minute. Based on the Altmann and Trafton (2002) model, it was predicted that rate of
resumption times would rise rapidly in the short duration range (i.e., 3-13 seconds) before the
diminishing over the next 45 seconds, approaching asymptote (i.e., a log function). To meet
these objectives, three additional longer interruption durations were added to the 3-, 8-, and 13-
second interruptions used in Experiment 1. The longer durations were specified using increasing
intervals of 10, 15, and 20 seconds, which resulted in 23-, 38-, and 58-second interruption
durations. The increased variability in the interruption duration manipulation was intended to
provide a resumption lag profile for durations ranging between 3 seconds and 1 minute. We
The Effect of Interruption Duration 21
hypothesized that the resumption lag profile would best fit a log trend, resembling an inverse of
the decay function (see Figure 1) as predicted by the memory for goals model.
Method
Participants
Twelve undergraduates from George Mason University received partial course credit for
participating in this study. The participants (8 men and 4 women) ranged in age from 18 to 23,
with an average age of 21 years.
Tasks and Equipment
The VCR and tracking tasks were identical to those used in Experiment 1.
Design
A single factor repeated measures design was used with six levels of interruption
durations. The six durations were 3-, 8-, 13-, 23-, 38, and 58-seconds. There were no matched
uninterrupted trials in this experiment. Participants completed two trials for each duration,
resulting in a total of 12 experimental trials. The shorter (3-, 8-, and 13-second) and longer (23-,
38-, and 58-second) interruption duration trials were blocked and counterbalanced with a Latin
square across participants.
Procedure
The procedure was identical to that in Experiment 1 except that all 12 experimental trials
were interruption trials.
Results and Discussion
One participant’s data were excluded from the analyses because of failure to perform the
tracking task during the interruption. As with Experiment 1, both categories of resumption errors
were removed from the data. The overall path-deviation resumption error rate was 7%. Table 1
The Effect of Interruption Duration 22
shows that the error rate was again lowest in the 3-second condition and gradually increased with
interruption duration. A repeated measures ANOVA revealed a significant difference between
the error rates across the six conditions, F(5, 50) = 3.55, p < .01,
η
p2 = .26. Tukey HSD post-hoc
comparisons showed that the 3-second condition was significantly lower than the 38-second and
58-second conditions. With the 100 ms criterion for incidental resumption actions, 0.3% of the
resumption lag data were excluded from the analyses.
The purpose of this experiment was to show that resumption times followed a log
function corresponding to activation decay over time. The memory activation formula as
expressed in Altmann and Trafton (2002) was a log function that resulted in the familiar decay
pattern (see Figure 1). Accordingly, a log model was fit to the data from this experiment. As seen
in Figure 4, the model fit the means data very well (R2 = .989). As interruption duration
increased, the resulting resumption lag times grew at a slower rate. This finding was particularly
important because it showed that the memory for goals model’s explanation for interrupted task
performance could account for not only the presence of the interruption duration effect at the
shorter durations, but it also offered an explanation regarding the absence of this effect in
previous literature that used interruption times longer than 30 seconds (e.g., Gillie & Broadbent,
1989).
Although the model fit provided strong support for the theoretical prediction, there were
two additional questions that required attention. First, the initial three durations were examined
to determine if the duration effect from Experiment 1 was replicated within this range of
durations. Second, we attempted to identify the range at which the resumption lag curve began to
approach asymptote to provide an indicator of when interruption duration ceases to have
substantial effect on resumption time.
The Effect of Interruption Duration 23
As in Experiment 1, the 3-, 8-, and 13-second interruption durations resulted in a
significant main effect for duration, F(2, 20) = 11.95, p < .01, ηp2 = .54. Paired comparisons (t-
tests) showed that only the 3-second condition was reliably shorter than the 8- and 13-second
conditions (both p’s < .01). The resumption lag slopes for the 3- to 13-second durations were
very similar for Experiments 1 and 2 (b = 30.98 and b = 32.54, respectively). Combined with the
interruption duration findings of Hodgetts and Jones (2006a), the main effects for the 3- to 13-
second interruption duration conditions in Experiments 1 and 2 provided compelling support for
the authenticity of the duration effect and the goal-activation explanation.
To identify if and when the resumption lag curve began to asymptote, the 58- and 38-
second conditions were compared and were found to not be significantly different, t(21) = -.626,
p = .54. Moving one duration shorter, the linear contrast between the 23-, 38-, and 58-second
conditions was marginally significant at best, F(1, 10) = 3.81, p = .08. Because the linear trend
was close to being reliably different from zero, confidence was low in declaring the 23-second
range as the point of asymptote. Accordingly, the 13-second condition was added to the analysis,
which yielded a significant linear contrast for the 13- through 58-second conditions, F(1, 10) =
10.37, p < .01. These results indicated that the resumption lag curve began to asymptote some
time between 13 and 23 seconds for the VCR and tracking task pairing.
The tracking task performance data were trimmed and analyzed in the same manner as in
Experiment 1, resulting in exclusion of 3% of the tracking data. Tracking task performance (see
Table 2) failed to show a significant effect across the six interruption durations, F < 1.
The implication for interruptions and interleaved task situations is that brief interruptions
will be less disruptive in terms of resuming the interrupted task, but only for interruptions lasting
up to roughly 15 to 25 seconds when the effect appeared to approach asymptote. Conclusions
The Effect of Interruption Duration 24
beyond one minute cannot be drawn from the present results, but they suggest that people
desiring to interleave tasks should strive to shift attention at least every 15 seconds for optimal
resumption times in computer-based, hierarchical tasks. Recall that Gonzalez and Mark (2004)
found that information workers shifted between tasks every three minutes on average, and Isaacs
et al. (2002) found that instant message turn-taking occurred every 15 seconds on average.
Taken together, the results of Experiments 1 and 2 provided compelling evidence that the
memory for goals framework can accurately describe the role of goal decay in interruption
recovery. The model was further tested in Experiment 3, which focused on the interaction
between interruption duration and varying levels of interruption task demand, which was
assumed to be related to the ability of participants to engage in goal rehearsal during the
interruptions. None of the recent interruptions studies based on Altmann and Trafton’s (2002)
memory for goals model manipulated both interruption duration and demand. Whereas Hodgetts
and Jones (2006a) provided important empirical findings related to duration and demand, they
did not manipulate these factors in the same experiment to test the interaction between the two.
Experiment 3
Because the memory for goals model explains the interruption duration effect in terms of
memory activation decay, the role of rehearsal becomes important in understanding how people
manage suspended goals. Theoretically, persistent rehearsal during an interruption, regardless of
duration, should minimize the interruption duration effect. In other words, if the duration effect
is due primarily to activation decay, then the rehearsal process of strengthening a memory’s
activation trace over the course of the interruption should make resuming that goal easier and
faster, thereby minimizing the duration effect. Alternatively, if the ability to maintain the
suspended goal through rehearsal is minimized with a demanding interruption task, then the
The Effect of Interruption Duration 25
duration effect may reveal higher rates of decay compared to the tracking task condition where
some rehearsal was assumed to be possible. Experiment 3 was designed to compare these three
conditions to test the predictions of the memory for goals theory regarding the strengthening
constraint and the interruption duration effect.
Three levels of interruption task demand were included to test the rehearsal prediction.
These levels of demand were assumed to directly impact opportunity for goal rehearsal during
the interruptions. Rehearsal was assumed to be uninhibited in the low-demand condition,
moderately inhibited in the medium-demand condition, and severely inhibited in the high-
demand task. For the low-demand condition, the interruption did not consist of a task. Rather, the
interruption was a blank screen. Participants were free to rehearse goals during the no-task
interruptions and it was assumed they would (see Trafton et al., 2003), although they were not
specifically instructed to do so.
The medium-demand condition consisted of the tracking task used in Experiments 1 and
2. It was considered to be moderately demanding because of its largely perceptual-motor nature,
which afforded opportunities for participants to rehearse their VCR task goal while tracking.
For the high-demand interruption condition, the selected task was a verbal version of the
n-back task that required participants to listen to, remember, and make decisions about verbally
presented letters. Although different from the verbal n-back task used by Smith and Jonides
(1999), the same assumptions about executive processes and storage of verbal material applied.
Pilot participants reported being unable to think about the VCR task while performing the verbal
n-back task, suggesting that participants would have few remaining cognitive resources for
rehearsal during the n-back task interruptions.
The Effect of Interruption Duration 26
The interruption duration effect was predicted for both the tracking and n-back task
conditions, with the latter demonstrating a steeper trend because of unmitigated goal decay.
Alternatively, the uninhibited opportunity to rehearse in the no-task condition was predicted to
minimize the duration effect as evidenced by a flatter slope than the tracking and n-back task
conditions. Corollary predictions were that the mean resumption lags for the n-back condition
would be longer than the other two conditions, and the resumption error rates would be highest
in the n-back condition. In addition, it was predicted that the no-task condition would yield
shorter resumption lags than the tracking task and n-back conditions because of uninhibited
opportunities for goal rehearsal.
Method
Participants
Thirty-six undergraduates from George Mason University received partial course credit
for participating in this study. The participants (9 men and 27 women) ranged in age from 18 to
30, with an average age of 21 years.
Tasks and Equipment
The VCR task was identical to those used in the previous experiments. The interruption
task levels consisted of a no-task condition, the tracking task, and the n-back task. For the no-
task condition, the interruption consisted of a blank screen. The tracking task was the same as in
Experiments 1 and 2. A verbal version of the n-back task was used in the high-demand
interruption condition.
The n-back task involves the serial presentation of digits where the participant must
respond whether the current digit is higher or lower than the previously presented digit (e.g.,
Lovett, Daily, & Reder, 2000). For the verbal version of the n-back task, single letters were
The Effect of Interruption Duration 27
presented serially and subjects are required to respond if the letter came before or after the 1-
back letter in the alphabet. For this experiment, the letters were presented aloud by the computer
and subjects responded by clicking on either a “higher” or “lower” button, corresponding to
closer to Z or closer to A, respectively. For example, if the letter sequence was G followed by T
the correct response was “higher.” The letters were “spoken” by the computer at a rate of one
letter every 1.6 seconds. The response buttons were located on the right half of the screen in
place of the tracking task. As with the tracking task condition, the cursor was automatically
repositioned to the saved position on the right or left half of the screen upon a switch.
Design
This experiment was a 3 x 3 mixed within-between design. The 3-, 8-, and 13-second
interruption durations were used as the within-subjects factor. The between-subjects factor was
interruption demand, which included the no-task, tracking task, and n-back task conditions.
Participants were randomly assigned to one of these three interruption demand conditions and
performed six experimental trials, two for each level of interruption duration.
Procedure
The procedure was the same as in Experiment 1 with the exception that the no-task
condition participants did not receive any interruption task practice, and the n-back task
participants received two 60-second practice trials. In addition, the interruptions occurred in
fixed 5-second intervals.
Results and Discussion
As with the previous experiments, path-deviation resumption errors and resumption lags
less than 100 ms were screened from the data. The overall path-deviation resumption error rate
was 3%. Table 1 presents the error rates for level of the interruption duration and demand
The Effect of Interruption Duration 28
factors. The error rate data were entered into a mixed within-between ANOVA. There was a
significant main effect for interruption demand, F(2, 27) = 10.56, p < .01,
η
p2 = .44. Tukey HSD
post-hoc comparisons revealed the error rate in the n-back condition (M = .06, SD = .23) was
greater than the error rates in the no-task condition (M = .02, SD = .15) and the tracking task
condition (M = .01, SD = .12), both p < .01. The no-task and tracking task conditions were not
reliably different (p = .62). Neither the main effect for duration nor the interaction between
duration and demand was significant (F < 1). With the 100 ms criterion for incidental resumption
actions, 0.3% of the resumption lag data were excluded from the analyses.
A 3 x 3 mixed within-between ANOVA revealed the predicted main effect for task
condition, F(2, 33) = 19.92, p < .01,
η
p2 = .55. The n-back condition resulted in the longest
resumption lags (M = 1789 ms, SD = 340), followed by the tracking task condition (M = 1605
ms, SD = 244), and then the no-task condition (M = 1322 ms, SD = 239). Planned t-test
comparisons showed that each of these conditions was reliably different from the others (all p <
.01). The fact that the three levels of interruption demand resulted in the predicted ordinal
resumption lag outcome along with the greater resumption error rate in the n-back condition
indicated that the task demand manipulation was a successful proxy for manipulating goal
rehearsal opportunity. In addition, the faster resumption lags in the no-task condition supports
the assumption that participants took advantage of the opportunity to rehearse.
The omnibus ANOVA also revealed a significant main effect for interruption duration,
F(2, 66) = 19.96, p < .01,
η
p2 = .38. However, the predictions concerned the individual demand
conditions rather than the overall main effect. To determine if the interruption duration effect
was present within demand condition, separate repeated measures ANOVAs were conducted for
the no-task, tracking, and n-back task conditions. The no-task condition resulted in a significant
The Effect of Interruption Duration 29
main effect for duration, F(2, 22) = 4.57, p < .05,
η
p2 = .29. In contrast with the findings from
Experiments 1 and 2, the main effect for duration was not significant in the tracking task
condition, F(2, 22) = 2.08, p = .15,
η
p2 = .16. Finally the n-back task condition resulted in the
predicted main effect for duration, F(2, 22) = 17.24, p < .01,
η
p2 = .61.
The absence of the duration effect in the tracking condition was surprising given its
reliability in the previous experiments. However, the lower effect size compared to those in
Experiments 1 and 2 (.50 and .54, respectively) suggested that the lack of effect might have been
due to greater variability in subjects. Further examination of the tracking condition results
revealed a marginally significant linear trend, F(1, 11) = 4.62, p = .055,
η
p2 = .30, hinting of the
duration effect. Considering the overall main effect for duration combined with the effects and
trends at the task demand condition level, there was compelling evidence to accept the duration
effect despite its modest presence in the tracking condition. The presence of the duration effect in
the no-task condition suggested that despite uninhibited opportunity for goal rehearsal, goal
activation still showed evidence of decay as interruption durations increased.
The significant interaction between interruption duration and demand conditions, F(4, 66)
= 3.92, p < .01 ,
η
p2 = .19, was also of interest because the n-back and tracking conditions were
predicted to produce steeper duration effect trends than the no-task condition. Specifically, it was
hypothesized that the n-back task condition would result in a steeper trend than the tracking task
and no-task conditions because of limited available cognitive resources while performing the
cognitively demanding n-back task, and that both the n-back and tracking task conditions would
product steeper slopes than the no-task condition. Linear contrast interactions were conducted
between the three demand conditions to test this hypothesis.
The Effect of Interruption Duration 30
As seen in Figure 5, the linear contrast interaction between the n-back and the no-task
conditions was significant, F(1, 22) = 11.17, p < .01,
η
p2 = .48, as was the interaction between
the n-back and tracking task conditions, F(1, 22) = 8.26, p < .05,
η
p2 = .27 as predicted. These
differences were confirmed with an analysis of the slopes. The slope for the n-back condition (b
= 39.14) was significantly greater than the slope for the no-task condition (b = 11.39), t(11) = -
3.61, p < .01. The n-back slope was also greater than the tracking task condition slope (b =
12.60), t(11) = -2.80, p < .05, as predicted. However, the no-task and tracking task condition
slopes were not significantly different, t < 1. The linear contrast interaction between the tracking
and no-task conditions was also not significant, F < 1. Whereas the n-back task produced a
greater resumption lag slope than the tracking and no-task conditions as predicted, the tracking
task condition slope was much lower than it was in Experiments 1 and 2 (b = 30.98 and b =
32.54, respectively). The marginal duration effect in the tracking task condition, as evidenced by
the smaller slope, suggested that the slope interactions with the tracking task condition be
considered cautiously.
The tracking task performance data were analyzed as in the previous experiments, along
with the n-back task performance. The tracking task performance data were trimmed and
analyzed in the same manner as in the previous experiments, resulting in exclusion of 1% of the
tracking data. Tracking task performance (see Table 2) did not vary reliably across the three
interruption durations, F(2, 22) = 2.90, p = .076. The n-back task accuracy scores were computed
for each trial. Because the letter presentation rate (every 1.6 seconds) prevented a response to the
second stimulus in the 3-second condition, related no-response errors were screened out of the
data. Accuracy rates showed no difference between the three interruption durations, F < 1. The
The Effect of Interruption Duration 31
mean accuracy rate was 73% (SD=12%) for the 3-second condition, 71% (SD=7%) for the 8-
second condition, and 74% (SD=4%) for the 13-second condition.
It is important to note the consistent performance in the n-back and tracking tasks across
the three interruption durations. Combined with no differences in the resumption error rates
across duration, there was no evidence of a speed-accuracy trade-off to explain the interruption
duration effect. The large differences in resumption times between task demand conditions
support the view that rehearsal is key to efficient resumption of suspended goals in an
interruption situation (see Trafton et al., 2003). However, the presence of the duration effect in
the no-task condition suggested that even in optimal rehearsal conditions, decay processes
appeared to win out to some degree.
General Discussion
The goal of this set of experiments was to demonstrate that interruption duration and
demand affect post-interruption task resumption, and that goal decay and opportunity to rehearse
play an important role in these effects. Experiment 1 showed that interruptions were disruptive,
and that longer interruptions were associated with longer resumption times, as predicted by the
memory for goals model (Altmann & Trafton, 2002). This finding was consistent with the
duration effect between 6- and 18-second interruptions demonstrated by Hodgetts and Jones
(2006a). Experiment 2 extended the interruption duration manipulation to nearly one minute and
supported the predicted log function for resumption lags. The trends observed in Experiments 1
and 3 were not inconsistent with the log trend observed in Experiment 2 because the 3- to 13-
second segment of the resumption curve captured the steep incline period that resembles a linear
trend. Finally, Experiment 3 showed that resumption lags were longer when available resources
for rehearsal were minimized by a high-demand interruption task. The results also showed an
The Effect of Interruption Duration 32
interaction between duration and demand manifested by a steeper resumption lag trend across
durations in the high-demand condition. The combined evidence from all three experiments
supported the veracity of the interruption duration effect, its log function characteristic over
interruptions up to one minute, and highlighted the importance of goal rehearsal during
interruptions for better resumption performance. These findings will be discussed in terms of the
their theoretical and practical implications.
Theoretical Implications
The results of this study contributed to the growing body of interruptions literature in
terms of the effects of interruption duration and cognitive demand, and how these two factors
interact to impact resumption of suspended task goals. The findings indicated that the time to
resume a task after an interruption depended both on the duration of that interruption and the
cognitive demand of the interrupting task. We argued that the duration effect was primarily due
to goal memory decay, and that the demand effect was directly related to the ability to rehearse
the suspended goal during the interruption. Each of these factors was found to affect resumption
performance by Hodgetts and Jones (2006a), and the present studies confirmed and expanded
upon their interruption duration findings to create a resumption lag profile from 3 to 58 seconds.
In addition, unlike Hodgetts and Jones’s study, Experiment 3 manipulated both duration and
demand to show how interruption task demand impacts the duration effect. Specifically, the
results showed that a more cognitively demanding interruption task produced a steeper
resumption lag trend across the 3-, 8-, and 13-second durations. This interaction showed for the
first time how opportunities to rehearse not only helped to speed-up resumption times, but also
showed how rehearsal opportunities help mitigate goal memory decay as interruption duration
The Effect of Interruption Duration 33
increases. This finding highlighted the importance of interleaving quick “reminders” of the
primary task state for reducing resumption costs.
In the instant messaging example, the report writer could occasionally steal a glance to
the report or quickly think about the suspended goal while waiting for a quick response, or even
before reading a response. In other words, people can interleave rehearsal within just about any
task that does not consume the available cognitive resources (see Trafton et al., 2003). The
presence of the decay trend in the no-task condition strongly suggested that despite optimal
rehearsal opportunities, decay effects still manifest in resumption times.
Quality of goal rehearsal may be partly responsible for the slight decay trend in the no-
task condition. A mismatch between the type of rehearsal executed and the actual task goal could
have weakened the strengthening of the task goal (see Nairne, 2002). Another possibility was
that the type of rehearsal that people engaged in was somehow shallow or ineffective for
maintaining activation levels above the interference threshold. Einstein et al. (2003) attempted to
deal with this issue by instructing participants to use implementation intentions as a means of
having participants form detailed plans for accomplishing intentions after a delay.
Implementation intentions are the more detailed when, where, and how aspects of accomplishing
the goal intention (Gollwitzer, 1999) rather than the intention to accomplish a goal. Einstein et al.
predicted better prospective memory performance by instructing participants to form
implementation intentions rather than simple goal intentions. The assumption was that by
generating implementation intentions the participants would be encoding more detailed and
therefore stronger intentions. However, the implementation intentions proved no better than
simple rehearsal instructions for remembering to execute an intention over brief delays. Further
The Effect of Interruption Duration 34
research is required to fully explore the rehearsal characteristics that produce optimal goal
strengthening in memory.
Goal decay is an important component of the memory for goals model (Altmann &
Trafton, 2002) and is consistent with the findings from classic short-term memory studies such
as Brown (1958) and Peterson and Peterson (1959), which showed longer retention intervals led
to more forgetting. The present study provided strong support for the role of decay in the
memory for goals over short interruptions. However, a common criticism of the decay process of
forgetting is that interference can be used to explain the same effects. The interference that
occurs during the interruption may better explain our findings rather than the time-based process
of memory decay and goal rehearsal. Perhaps people were more likely to experience proactive
interference as time away from the primary task increased because of the build-up of previous
task goals in memory. Recent evidence suggested that intrusion errors were greater for an
interrupted task, but the intrusions were based on prior-knowledge rather than on the interruption
itself (Oulasvirta & Saariluoma, 2004).
Contrary to the proactive interference explanation, Monk (2004) found that resumption
lags were actually shorter when people were interrupted more frequently. More frequent
interruptions should result in greater proactive interference because more goals have been
suspended and resumed. Monk suggested that the rapid switching between the VCR and tracking
tasks may have compelled participants to adopt a strategy to actively rehearse their suspended
goals during the interruptions, leading to faster resumption times. Whether this finding was
viewed as lack of evidence for proactive interference using the same empirical method or as
evidence for the active rehearsal strategy explanation, the results were consistent with the
memory for goal model’s decay explanation. In addition, Altmann and Schunn (2002) made a
The Effect of Interruption Duration 35
compelling argument for the role of decay in short-term forgetting. They did not argue that decay
is the principal mechanism for forgetting; rather that it plays a secondary but important role
compared to interference. Likewise, the importance of interference as a strong contributor to
forgetting is not disputed here; however, the present evidence shows that goal decay also plays
an important role.
The results from interruption studies are inevitably compared to those from task-
switching studies in which switch costs have been explored extensively (see Monsell, 2003 for a
brief review). This comparison is particularly tempting with the interleaved interruptions
paradigm from the current study. However, interruption studies involve the suspension and
resumption of task goals rather than the switching of stimulus-response mappings between trials,
which we argue is a fundamentally different operation. Hodgetts and Jones (2006a) noted that
time-based determinants of goal retrieval cannot be attributed to task-switching costs, and that
the memory for goals model provided a more compelling explanation for resumption costs.
Mixing cost evidence from the task-switching literature (see Monsell, 2003; Rubin & Meiran,
2005), however, provided an alternative theoretical explanation for the duration effect that was
important to consider. Mixing costs are those costs associated with maintaining multiple task sets
in working memory, resulting in longer response latencies in switching trials versus single-task
trials. The duration effect, therefore, could have been the result of different resource allocation
strategies when maintaining two task goals in memory in the shorter verses longer interruption
durations.
In the present experiments, mixing costs would translate to longer IAIs in the interrupted
versus uninterrupted VCR programming trials in Experiment 1. Knowing that they would need to
interleave the VCR and tracking tasks, participants may have maintained both task sets in
The Effect of Interruption Duration 36
working memory to foster better switching performance. The differential allocation of resources
would be an overall effort-saving strategy to produce more efficient task switching and thus
better dual-task performance overall when interleaving two tasks. However, the dual-task
strategy loses its utility with longer interruptions because the switches seem few and far between
(though there were actually the same number of switches on average because the VCR times
consistently ranged between 3 and 7 seconds). The strategy changes to exclusively allocate
resources to the tracking task until the shift back to the VCR task. The change to single-task
resource allocation would result in longer resumption times when switching back to the VCR
task because the VCR task set was not actively maintained during the interruption.
When comparing the IAIs from the interrupted and uninterrupted trials from Experiment
1, we found the opposite results from those predicted by the resource allocation explanation. The
IAIs in the interrupted condition averaged 510 ms (SD = 81), whereas they averaged 949 ms (SD
= 284) in the uninterrupted condition (using the same sampling procedure as in Experiment 1).
However, this finding did not rule out the resource allocation explanation entirely because Rubin
and Meiran (2005) showed that mixing costs were eliminated when the two task sets were
unambiguous (i.e., clearly distinct tasks) as they were in this study.
If the resource allocation explanation was correct, then we would have expected to see
consistently short resumption lags until the interruption duration was sufficiently long to elicit
the strategy shift, producing longer, asymptotic resumption lags. One would expect to see a
resumption lag trend resembling a logistic s-curve across the six durations in Experiment 2 rather
than the observed log function (see Figure 5). As long as people were working to maintain both
tasks sets in working memory, faster resumption lags should have resulted. However, once the
dual-task strategy was abandoned for the single-task strategy, one would expect asymptotic
The Effect of Interruption Duration 37
resumption lags. In fact, the observed resumption lag trend from Experiment 2 supported the
goal decay explanation over the resource allocation explanation.
The present findings also added to a growing body of empirical evidence (e.g., Altmann
& Trafton, 2007; Cades et al., 2007; Hodgetts & Jones, 2006a; Hodgetts & Jones, 2006b; Li et
al., 2006; Monk, Boehm-Davis, & Trafton, 2004; Trafton et al., 2003) supporting the use of the
memory for goals model (Altmann & Trafton, 2002) as a framework for studying interruptions.
When the interruptions problem was approached with this cognitive theory, we were able to
make specific predictions that were confirmed by using a theory-driven metric that is sensitive to
the subtle effects of goal decay. Corroborating and extending Hodgetts and Jones’ (2006a)
duration effect evidence while also demonstrating why this effect has gone undetected in
previous research (e.g., Gillie & Broadbent) was indeed a powerful expression of how sound
cognitive theory can significantly contribute to the interruptions problem.
One issue that was unaddressed by this research was the role of environmental cues in
helping to retrieve suspended goals. In the VCR task, there were several available cues to help
the participant re-establish the suspended task state. For example, the cursor arrow remained in
the same location as when the switch occurred, providing a powerful cue as to where the
participant was in the task and what action/goal was to be accomplished next. Other display
features such as button activation highlights and display feedback were also available to aid the
participant in resuming the task. However, these cues were available in all conditions and still
the interruption duration and inhibited rehearsal effects persisted. More research is required to
fully isolate the role of environmental cues from rehearsal, recency, and frequency.
Practical Implications
The Effect of Interruption Duration 38
The application of these findings to real-world tasks exceeds the simple conclusion that
longer and more demanding interruptions will result in longer primary task resumption times.
We will discuss some of the contexts in which people interleave interrupting tasks, and where
additional time costs when shifting cognitive effort can have significant ramifications. In
addition, we will discuss how the duration and demand findings generalize to each of these
situations.
Quick switches between the primary tasks and interrupting tasks, or task interleaving, is a
common behavior observed in contexts such as driving, emergency rooms, and aviation cockpits,
among others. Studies describing glance duration and frequency behavior when engaging an in-
vehicle tasks go back decades (e.g., Dingus et al. 1989; Mourant & Rockwell, 1972; Wierwille,
1993). The results from these studies and others showed that voluntary eyes-off-road times rarely
exceed two seconds. Wierwille argued that tasks requiring more than 1.5 seconds to complete
push drivers to adopt a time-sharing strategy shifting visual and cognitive attention between the
driving and in-vehicle tasks. Gellatly and Kleiss (2000) showed that people were remarkably
consistent in shifting attention between the road and in-vehicle task every second. This time-
sharing scenario showed how people interleave tasks in a similar time scale as studied in the
present experiments. The resumption costs in the present experiments were on the order of
hundreds of milliseconds. The time costs certainly were inextricably connected to the VCR
programming task used in this study, along with the tracking and n-back interruption tasks;
however, Lee et al. (2001) showed that reaction latencies as short as 300 ms can greatly increase
the odds of a collision. Therefore, quick shifts of attention can potentially have consequences in
both driver reactions to unexpected events, as well as time to complete the in-vehicle task. The
longer a task takes to complete, the more time the driver spends engaged in a distracting and
The Effect of Interruption Duration 39
potentially dangerous situation. The connection between driver reactions and resumption costs
should be considered cautiously until further research using driving tasks and resumption lag
measures are conducted.
The interruption duration effect has less of an impact on the driver distraction situation
because of the small range of observed glance durations (see Horrey & Wickens, 2007).
However, the demand effect does have implications for the kinds of tasks that drivers engage in
while driving. Our results suggest that simple tasks such as tuning the radio dial (visual and
motor requirements only) would have lower time costs than a complex task like finding a
particular song in an mp3 player or entering a destination into a GPS navigation system (visual,
cognitive, and motor requirements). Because our findings rely on the resumption of suspended
task goals, generalizing to reactions to driving-related events that do not involve goal resumption
should be made with caution. Further research is required to quantify switch costs on driver
reactions. Alternatively, our findings help to understand total time to complete a task like
destination entry because a task goal must be suspended and resumed with each shift of attention.
Even if the resumption lags were very short for each shift, the costs would be additive over the
course of the entire task, resulting in longer task times. Longer task times are typically associated
with more eyes-off-road time because attention must be shifted a greater number of times.
Emergency rooms are another environment in which people shift visual and cognitive
attention frequently and rapidly. Chisholm et al. (2001) reported that emergency room physicians
spent 37.5 minutes out of every hour managing three or more patients and were interrupted 9.7
times per hour. Although it is impossible to estimate from our data how resumption costs
manifest in emergency rooms, our data showed that repeated suspension and resumption of task
goals may be costly in such a time-critical context. For example, an alarm may sound during a
The Effect of Interruption Duration 40
procedure requiring several seconds of a nurse’s attention. Once the urgent matter is resolved,
the nurse then returns to the previous task of assisting the doctor, potentially with a brief time
delay as the nurse retrieves the suspended goal from memory. As in the driving example, the
demand effect has the potential to be greater than the duration effect because of the range of
cognitive tasks in such complex, life and death situations. Further research investigating
resumption performance in emergency rooms and other healthcare environments is crucial for
understanding how interruptions affect performance and ultimately patients’ lives.
Another critical situation in which interruptions can have a significant impact is the
aircraft cockpit. Air traffic controllers, other personnel in the cockpit, and flight attendants
frequently interrupt pilots going through pre-flight checklists and other critical tasks.
Loukopolous, Dismukes, and Barshi (2001) reported that frequent interruptions in the cockpit
required pilots to continuously make task management decisions, including adding, shedding,
and rescheduling actions. Perhaps more important than the time costs associated with
interruptions in the cockpit are the potential error costs such as missed items on the pre-flight
checklist. As previously noted, multiple plane crashes have been attributed in part to
interruptions to the pre-flight checklists (NTSB, 1969, 1988).
There are countless other situations in which people interleave tasks. The instant
messaging example was used earlier to show how interruption duration and demand could affect
the resumption of a writer’s performance. This example does not typically involve life-
threatening situations as with the driving and emergency room examples; however, the additive
time costs can have significant economic impact in loss of productivity over time.
Conclusions
The Effect of Interruption Duration 41
The goal of this study was to apply a well-specified theory of memory for goals to the
real-world problem of resuming tasks after being interrupted. The results helped define the role
of interruption duration and demand in determining resumption costs. Duration was shown to
result in increased resumption costs when the interruptions lasted between 3 and 13 seconds;
however, a log function pattern emerged when the duration manipulation was extended to nearly
one minute. This finding supported the role of decay in Altmann and Trafton’s (2002) theory.
Demand was also shown to have a substantial impact on resumption costs, indicating that
opportunities to rehearse suspended task goals are an important determinant in defining
resumption times. The interaction between duration and demand, while needing further
exploration, provided additional insight into how opportunity to rehearse task goals during an
interruption can help mitigate decay processes, though it appeared that decay cannot be
completely eliminated even with optimal opportunity for goal rehearsal. These results added to
the growing body of empirical support for the memory for goals model and its application to the
study of interruptions (see Altmann & Trafton, 2007; Cades et al., 2007; Hodgetts & Jones,
2006a; Hodgetts & Jones, 2006b; Li et al., 2006; Monk, Boehm-Davis, & Trafton, 2004; Trafton
et al., 2003). The current findings also provided insight into the practical costs of interleaving
interruption tasks with the primary task. The added resumption times associated with
interruptions have important consequences for overall task efficiency and productivity in office
settings; however, these costs can have far more serious consequences in situations like driving,
emergency rooms, and aircraft cockpits.
The Effect of Interruption Duration 42
References
Altmann, E. M., & Schunn, C. D. (2002). Integrating decay and interference: A new look at an
old interaction. Proceedings of the 24th annual meeting of the Cognitive Science Society
(pp. 65-70). Hillsdale, NJ: Erlbaum.
Altmann, E. M., & Trafton, J. G. (2002). Memory for goals: An activation-based model.
Cognitive Science, 26, 39-83.
Altmann, E. M., & Trafton, J. G. (2007). Timecourse of recovery from task interruption: Data
and a model. Psychonomic Bulletin & Review, 14, 1079-1084.
Anderson, J. R. (1993). Rules of the mind. Hillsdale, NJ: Erlbaum.
Anderson, J. R., Bothell, D., Byrne, M. D., Douglass, S., Lebiere, C., and Qin, Y. (2004). An
integrated theory of the mind. Psychological Review, 111, 1036-1060.
Anderson, J. R. & Lebiere, C. (1998). Atomic components of thought. Mahwah, NJ: Erlbaum.
Avrahami D. & Hudson S. E. (2006) Responsiveness in instant messaging: predictive models
supporting inter-personal communication. Proceedings of the SIGCHI Conference on
Human Factors in Computing Systems (CHI'06) (pp. 731-740). New York: ACM Press.
Brown, J. (1958). Some tests of the decay theory of immediate memory. Quarterly Journal of
Experimental Psychology, 10, 12-21.
Byrne, M. D., & Bovair, S. (1997). A working memory model of a common procedural error.
Cognitive Science, 21, 31-61.
Cades, D. M., Trafton, J. G., Boehm-Davis, D. A., & Monk, C. A. (2007). Does the difficulty of
an interruption affect our ability to resume? Proceedings of the 51st Annual Meeting of
the Human Factors and Ergonomics Society (pp. 234-238). Santa Monica, CA: HFES.
Cellier, J., & Eyrolle, H. (1992). Interference between switched tasks. Ergonomics, 35(1), 25-36.
The Effect of Interruption Duration 43
Chisholm, C. D., Collison, E. K., Nelson, D. R., & Cordell W. H. (2000). Emergency department
workplace interruptions: Are emergency physicians “interrupt-driven” and
“multitasking”? Academic Emergency Medicine, 7 (11), 1239-1243.
Chisholm, C. D., Dornfeld, A. M., Nelson, D. R., & Cordell, W. H. (2001). Work interrupted: A
comparison of workplace interruptions in emergency departments and primary care
offices. Annals of Emergency Medicine, 38, 146-151.
Cutrell, E., Czerwinski, M. & Horvitz, E. (2001). Notification, disruption and memory: effects of
messaging interruptions on memory and performance. In Human-Computer Interaction--
Interact '01 (pp. 263-269). Amsterdam: IOS Press.
Czerwinski, M., Christman, S., & Rudisill, M. (1991) Interruptions in Multitasking Situations:
The Effects of Similarity and Warning. Technical Report JSC-24757, NASA Johnson
Space Center, Houston, Texas.
Czerwinski, M., Cutrell, E., & Horvitz, E. (2000). Instant messaging: Effects of relevance and
time. In S. Turner and P. Turner (Eds.), People and Computers XIV: Proceedings of HCI
2000 (pp. 71-76). British Computer Society.
Czerwinski, M., Horvitz, E., & Wilhite, S. (2004). A diary study of task switching and
interruptions. In CHI 2004 Proceedings (pp. 175-182). New York: ACM Press.
Dingus, T., Antin, J., Hulse, M., & Wierwille, W. (1989). Attentional demand requirements of an
automobile moving-map navigation system. Transportation Research, 23(A), 301-315.
Edwards, M. B., & Gronlund, S. D. (1998). Task interruption and its effects on memory.
Memory, 6, 665-687.
The Effect of Interruption Duration 44
Einstein, G. O., McDaniel, M. A., Williford, C. L., Pagan, J. L., & Dismukes, R. K. (2003).
Forgetting of intentions in demanding situations is rapid. Journal of Experimental
Psychology: Applied, 9(3), 147-162.
Gillie, T. & D. E. Broadbent (1989). What makes interruptions disruptive? A study of length,
similarity, and complexity, Psychological Research, 50(4), 243-250.
Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American
Psychologist, 54, 493-503.
Gonzalez, V. & Mark, G. (2004). “Constant, Constant, Multi-tasking Craziness”: Managing
multiple working spheres. Proceedings of ACM CHI’04 (pp. 113-120). New York: ACM
Press.
Gray, W. D. (2000). The nature and processing of errors in interactive behavior. Cognitive
Science, 24(2), 205-248.
Gray, W. D. & Fu, W.-T. (2001). Ignoring perfect knowledge in-the-world for imperfect
knowledge in-the-head: Implications of rational analysis for interface design.
Proceedings of the ACM CHI’01 Conference on Human Factors in Computing Systems.
Hodgetts, H. M., & Jones, D. M. (2006a). Interruption of the tower of london task: Support for a
goal-activation approach. Journal of Experimental Psychology: General, 135 (1), 103-
115.
Hodgetts, H. M., & Jones, D. M. (2006b). Contextual cues aid recovery from interruption: The
role of associative activation. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 32 (5), 1120-1132.
Horrey, W. J., & Wickens, C. D. (2007). In-vehicle glance duration. Transportation Research
Record No. 2018, 22-28.
The Effect of Interruption Duration 45
Iqbal S. T. & Horvitz E. (2007). Disruption and recovery of computing tasks: Field study,
analysis, and directions. In Proceedings of CHI 2007, New York: ACM Press.
Isaacs E., Walendowski A., Whittaker S., Schiano D. J., & Kamm C. (2002). The character,
functions, and styles of instant messaging in the workplace. Proceedings of the ACM
Conference on Computer Supported Cooperative Work (pp. 11-20). New York: ACM
Press.
Kreifeldt J. G., & McCarty, M. E. (1981). Interruption as a test of the user-computer interface.
In: Proceedings of the 17th Annual Conference on Manual Control (California Institute
of Technology, Jet Propulsion Laboratory Publication 81-95), pp. 655-667.
Lee, J. D., Caven, B., Haake, S., & Brown, T. L. (2001). Speech-based interaction with in-
vehicle computers: The effect of speech-based e-mail on drivers’ attention to the
roadway. Human Factors, 43(4), 631-64.
Li, S. Y. W., Cox, A. L., Blandford, A., Cairns, P., Young, R. M., & Abeles, A. (2006). Further
investigations into post-completion error: the effects of interruption position and
duration. In Proceedings of the 28th Annual Meeting of the Cognitive Science Conference
(pp. 471-476). Vancouver, BC: Cognitive Science Society.
Lovett, M. C., Daily, L. Z., & Reder, L. M. (2000). A source activation theory of working
memory: cross-task prediction of performance in ACT-R. Journal of Cognitive Systems
Research, 1, 99-118.
Lovett, M. C., Reder, L. M., & Lebiere, C. (1999). Modeling working memory in a unified
architecture: An ACT-R perspective. In A. Miyake & P. Shah (Eds.) Models of Working
Memory (pp. 135-182). Cambridge, MA: Cambridge.
The Effect of Interruption Duration 46
Loukopoulos, L. D., Dismukes, R. K., & Barshi, I. (2001). Cockpit interruptions and distractions:
A line observation study. The 11th International Symposium on Aviation Psychology.
Columbus: Ohio State University.
McDaniel, M.A., Einstein, G.O., Graham, T., & Rall, E. (2004). Delaying execution time of
intentions: Overcoming the costs of interruptions. Applied Cognitive Psychology, 18,
533-547.
McFarlane, D. C. (2002). Comparison of four primary methods for coordinating the interruption
of people in human-computer interaction. Human-Computer Interaction, 17(1), 63-139.
McFarlane D.C., & Latorella K.A. (2002) The scope and importance of human interruption in
human-computer interaction design, Human-Computer Interaction, 17(1), 1-61.
McKnight, P. E., McKnight, K. M, Sidani, S. & Figueredo, A. J. (2007). Missing Data: A Gentle
Introduction. New York: Guilford.
Monk, C.A. (2004). The effect of frequent versus infrequent interruptions on primary task
resumption. In the Proceedings of the 48th Annual Meeting of the Human Factors and
Ergonomics Society (pp. 295-299). Santa Monica, CA: HFES.
Monk C. A., Boehm-Davis D. A., & Trafton J. G. (2004). Recovering from interruptions:
Implications for driver distraction research, Human Factors, 46 (4), 650-663.
Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7(3), 134-140.
Mourant, R. R., & Rockwell, T. H. (1972). Strategies of visual search by novice and experienced
drivers. Human Factors, 14 (4), 325-335.
Nairne, J. S. (2002). Remembering over the short-term: The case against the standard model.
Annual Review of Psychology, 53, 53-81.
The Effect of Interruption Duration 47
NTSB. (1969). Aircraft accident report: Pan American World Airways, Inc., Boeing 707-321C,
N799PA, Elmendorf Air Force Base, Anchorage, Alaska, December 26, 1968
(NTSB/AAR-69/08). Washington, DC: National Transportation Safety Board.
NTSB. (1988). Aircraft accident report: Northwest Airlines, Inc., McDonnell Douglas DC-9-82,
N312RC, Detroit Metropolitan Wayne County Airport, Romulus, Michigan, August 16,
1987 (NTSB/AAR-88/05). Washington, DC: National Transportation Safety Board.
Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard University Press.
Oulasvirta, A., & Saariluoma, P. (2004). Long-term working memory and interrupting messages
in human-computer interaction. Behaviour & Information Technology, 23(1), 53-64.
Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individual verbal items.
Journal of Experimental Psychology, 58, 193-98.
Rubin, O., & Meiran, N. (2005). On the origins of the task mixing cost in the cuing task-
switching paradigm. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 31(6), 1477-1491.
Smith, E. E., & Jonides, J. (1999). Storage and executive processes in the frontal lobes. Science,
283, 1657-1661.
Speier C., Valacich J. S., & Vessey I. (1999). The influence of task interruption on individual
decision making: An information overload perspective, Decision Sciences, 30 (2), 337-
360.
Speier C., Vessey I., & Valacich J. S. (2003). The effects of interruptions, task complexity, and
information presentation on computer-supported decision-making performance, Decision
Sciences, 34 (4), 771-797.
The Effect of Interruption Duration 48
Trafton, J. G., Altmann, E. M., Brock, D. P., & Mintz, F. (2003). Preparing to resume an
interrupted task: effects of prospective goal encoding and retrospective rehearsal.
International Journal of Human Computer Studies, 58(5), 583-603.
Zeigarnik, B. (1927). Das behalten erledigter und unerledigter handlungen. Psychologische
Forschung, 9, 1-85.
Zijlstra, F. R. H., Roe, R. A., Leonova, A. B., & Krediet, I. (1999). Temporal factors in mental
work: Effects of interrupted activities. Journal of Occupational and Organizational
Psychology, 72(2), 164–185.
The Effect of Interruption Duration 49
Author Note
This work was partially supported by the US Office of Naval Research to J. Gregory
Trafton: N0001405WX2001 and, N0001405WX30020. The views and conclusions contained in
this document should not be interpreted as necessarily representing the official policies, either
expressed or implied, of the U. S. Navy. The authors would like to thank Yi-Fang Tsai and
David Shin for their assistance in data collection, and the Arch Lab members for their many
helpful comments on earlier drafts.
The Effect of Interruption Duration 50
Tables
Table 1. Resumption Error Rates.
Interruption Duration
3 sec
8 sec
13 sec
23 sec
38 sec
58 sec
Experiment
M
SD
M
SD
M
SD
M
SD
M
SD
M
SD
Exp. 1
.03
.18
.07
.25
.06
.23
Exp. 2
.02
.13
.06
.23
.05
.23
.09
.28
.10
.30
.10
.30
Exp. 3
No-task
.01
.12
.03
.16
.03
.16
Exp. 3
Tracking
.01
.10
.01
.12
.02
.13
Exp. 3
N-back
.07
.25
.05
.21
.05
.22
The Effect of Interruption Duration 51
Table 2. Tracking Task Performance (RMS).
Interruption Duration
3 sec
8 sec
13 sec
23 sec
38 sec
58 sec
Experiment
M
SD
M
SD
M
SD
M
SD
M
SD
M
SD
Exp. 1
47
11
41
9
40
8
Exp. 2
56
21
51
18
50
18
57
26
54
21
56
20
Exp. 3
43
7
40
5
40
5
The Effect of Interruption Duration 52
Figure Captions
Figure 1. The time course of activation of a new goal (solid line) and the interference level from
old goals (dashed line). Adapted from Altmann and Trafton (2002).
Figure 2. Simulated VCR interface used in primary task.
Figure 3. Mean resumption lags (±SE) as a function of interruption duration.
Figure 4. Mean resumption lags (±SE) as a function of interruption duration, with model fit.
Figure 5. Mean resumption lags (±SE) as a function of interruption duration and demand.
The Effect of Interruption Duration 53
The Effect of Interruption Duration 54
The Effect of Interruption Duration 55
The Effect of Interruption Duration 56
The Effect of Interruption Duration 57
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Experimental Psychology: Applied
This content is subject to copyright. Terms and conditions apply.
