Cognition 111 (2009) 98–113

Contents lists available at ScienceDirect

Cognition
journal homepage: www.elsevier.com/locate/COGNIT

Anatomy of an error: A bidirectional state model of task

engagement/disengagement and attention-related errors
J. Allan Cheyne, Grayden J.F. Solman, Jonathan S.A. Carriere, Daniel Smilek *
Department of Psychology, University of Waterloo, 200 University Ave. West, Waterloo, Ontario, Canada N2L 3G1

a r t i c l e i n f o a b s t r a c t

Article history: We present arguments and evidence for a three-state attentional model of task engage-
Received 15 February 2008 ment/disengagement. The model postulates three states of mind-wandering: occurrent
Revised 30 December 2008 task inattention, generic task inattention, and response disengagement. We hypothesize
Accepted 31 December 2008
that all three states are both causes and consequences of task performance outcomes
and apply across a variety of experimental and real-world tasks. We apply this model to
the analysis of a widely used GO/NOGO task, the Sustained Attention to Response Task
Keywords:
(SART). We identify three performance characteristics of the SART that map onto the three
Mind-wandering
Attention
states of the model: RT variability, anticipations, and omissions. Predictions based on the
Attentional engagement model are tested, and largely corroborated, via regression and lag-sequential analyses of
Attentional disengagement both successful and unsuccessful withholding on NOGO trials as well as self-reported
SART mind-wandering and everyday cognitive errors. The results revealed theoretically consis-
Inattention tent temporal associations among the state indicators and between these and SART errors
Attentional error as well as with self-report measures. Lag analysis was consistent with the hypotheses that
Action slips temporal transitions among states are often extremely abrupt and that the association
Sustained attention
between mind-wandering and performance is bidirectional. The bidirectional effects sug-
gest that errors constitute important occasions for reactive mind-wandering. The model
also enables concrete phenomenological, behavioral, and physiological predictions for
future research.
Ó 2009 Elsevier B.V. All rights reserved.

1. Introduction Smallwood, O’Connor, Sudberry, Haskell, & Ballantine,

The type of error in question is an attention-related ac-
tion or cognitive error caused by a loss of attentional
engagement in an ongoing primary task (Reason, 1990;
Reason & Mycielska, 1982). Task disengagement and its
sources have been studied under a number of rubrics
including absent-mindedness (Cheyne, Carriere, & Smilek,
2006; Reason & Lucas, 1984), mind-wandering (Kane
et al., 2007; Smallwood & Schooler, 2006; Wegner, 1997),
stimulus-independent thought (Antrobus, 1968; Teasdale,
Lloyd, Proctor, & Baddeley, 1993; Teasdale, Segal, &
Williams, 1995), task-unrelated/related thought (Giambra,
1995; Smallwood, Baracaia, Lowe, & Obonsawin, 2003;
* Corresponding author.
E-mail address: dsmilek@uwaterloo.ca (D. Smilek).
2004; Smallwood, Obonsawin, & Heim, 2003; Smallwood,
Obonsawin, & Reid, 2003; Smallwood et al., 2004), tune
outs and zone outs (Schooler, 2002; Schooler, Reichle, &
Halpern, 2005; Smallwood, McSpadden, & Schooler,
2007), and mind pops (Kvavilashvili & Mandler, 2004).
Whatever label is used, current approaches to attentional
task disengagement agree that the core phenomenon con-
sists of a state of reduced allocation of attentional re-
sources to environmental task-related stimuli (Manly,
Robertson, Galloway, & Hawkins, 1999; Robertson, Manly,
Andrade, Baddeley, & Yiend, 1997; Smallwood & Schooler,
2006; Smallwood et al., 2007). Conditions promoting such
task disengagement arising from mind-wandering include
protracted, unvarying, familiar, repetitive, and undemand-
ing tasks and task environments (see Smallwood and
Schooler (2006) for a review). Although distinctions

0010-0277/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.cognition.2008.12.009
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
between mind-wandering with and without awareness
(Hester, Foxe, Molholm, Shpaner, & Garavan, 2005;
Schooler, 2002; Schooler et al., 2005; Smallwood et al.,
2007) imply different modes or states of attentional disen-
gagement or decoupling (Smallwood & Schooler, 2006),
these have not been conceptually analyzed in terms of
underlying cognitive processing failures. We propose a
model of attentional engagement/disengagement consist-
ing of three distinct states, describe their cognitive,
phenomenological, and performance characteristics, and
suggest real-world examples of each. We propose and as-
sess behavioral markers for each state using an established
sustained attention task and relate these markers to
behavioral performance and to self-reported everyday
mind-wandering and cognitive failure.
2. States of attentional disengagement: a three-state
model
2.1. Task engagement: to what do we attend when we attend
to a task?
The present model of mind-wandering and task disen-
gagement assumes that successful task performance in
everyday tasks such as reading a book, listening to a speak-
er, or driving a vehicle, has the following attentional
requirements. First, at the most demanding level, atten-
tional resources must be allocated to occurrent, dynamic,
and focal features of a task that change from moment to
moment. In the language of intentionality theories, occur-
rent attentional concerns correspond to intentions-in-ac-
tion (Searle, 1983), immediate intentions (Brand, 1984),
present-direct intentions (Bratman, 1987), and proximal
(Mele, 1992) or P-intentions (Pacherie, 2008). Dynamic
features are cues that guide moment-to-moment task per-
formance including feedback regarding the adequacy of
ongoing performance. These are often, though not exclu-
sively, semantic features such as specific word/phrase/sen-
tence meanings during a reading/listening task, or critical
meaningful changes in a driving environment such as traf-
fic signals, object visibility, surface features, turns, and the
behavior of other vehicles.
Second, additional attentional resources must be dedi-
cated to monitoring the stable background task environ-
ment and goals (e.g., reading an instruction manual,
listening to a lecture, and driving to work). As Pacherie
(2008) notes, P-intentions, or what we have called occur-
rent attention, must integrate overarching prior intentions
(i.e., distal or D-intentions) concerning task goals and per-
formance with the current task conditions. That is, in
maintaining task set some attentional resources must be
given over to general task parameters and goals.
Third, successful task engagement typically has implicit
and explicit performative aspects. Attentional resources
must therefore also be allocated to the on-line recruitment,
guidance, and maintenance of task responses (e.g., mainte-
nance of saccades down a page; taking notes during a lec-
ture, manipulation of the control surfaces of a vehicle).
Clearly anything that interferes with, or places chal-
lenges on, any of these three aspects of task-related atten-
tion will increase task failures. Note that the foregoing
99
analysis, as well as the subsequent model of task disen-
gagement, assumes adequate knowledge, skill, and motiva-
tion on the part of the performer. As noted at the outset,
the error in question occurs despite optimal training, com-
petence, and motivation of the performer. The present
analysis is focused on the withdrawal of attentional re-
sources from an external task rather than on important is-
sue of the nature of the allocation of resources to internal
processing (see Smallwood, Baracaia, Lowe, and Obonsa-
win (2003), Smallwood, Heim, Rigby, and Davies (2006),
and Smallwood et al. (2004, 2007)).
2.2. Task disengagement: states of inattention
State 1: The model proposes three discrete states of
attentional disengagement. Based on the foregoing task
analysis, it follows that the attentionally most demanding
dynamic level will be most vulnerable to mind-wandering
leading to occurrent task inattention. Occurrent task inat-
tention (State 1) consists of an often brief and/or partial
waning of detailed processing of moment-to-moment
stimulus meaning. The essential characteristic of State 1
is a transient disengagement of attention from the dy-
namic features of the task (e.g., the content of the present
reading passage or lecture or the current movements and
signals of the vehicle ahead). Individuals are aware of such
inattention some of the time during State 1 (Hester et al.,
2005; Schooler, 2002; Schooler et al., 2005; Smallwood
et al., 2007) and likely experience dual consciousness or di-
vided attention (Smallwood et al., 2007) in which they are
aware of the transient intrusions of trains of interfering
thought, which they may attempt to suppress. The phe-
nomenology of this state is nicely captured in the meta-
phor of ‘‘tuning out” (Schooler, 2002; Schooler et al.,
2005; Smallwood et al., 2007). State 1 inattention is brief
and unstable, and likely to be affected by incidental events,
errors, and near-misses. Near-misses will be common at
this early state compared to later states and attentional re-
sources will often be sufficient to detect incipient slips be-
fore they become overt errors. Nonetheless, because of its
ubiquity in human performance, transient focal task inat-
tention may well be responsible for the majority of errors
despite its comparatively brief and mild character.
Conflicted division of attention can occur at any time,
even very early in a novel task, although it is more com-
mon after extended periods on routine tasks and in famil-
iar settings. Yet, although attention lapses tend to increase
as a function of task features such as the elapsed time, rep-
etition, or familiarity of the ongoing task (Mackworth,
1948; Manly, Lewis, Robertson, Watson, & Datta, 2002;
Manly et al., 1999; Robertson et al., 1997; Smallwood
et al., 2003, 2004), mind-wandering can occur even at the
beginning of a novel task as the individual considers
thoughts of the upcoming task and potential quality of per-
sonal performance. A novice driver will, for example, fre-
quently be quite explicitly preoccupied with recent
errors, or aspects of the task of driving that are not relevant
to the immediate task context (e.g., thinking, while practic-
ing braking, about how to avoid grinding the gears next
time gear shifting is required). Nonetheless, novel and
challenging tasks, requiring frequent monitoring of chang-
100 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
ing task conditions will tend to keep mind-wandering lim-
ited to State 1.
State 2: In State 1 the individual maintains awareness of
the general task environment. That is, attention to task set
remains active but with loss of sensitivity to moment-to-
moment task variations. In State 2, attention to the generic
task-relevant aspects of the environment decreases though
the individual continues to show well-practiced automatic
responding. In everyday terms we characterize this as
‘‘going through the motions”. For simple repetitive tasks
this minimal engagement may suffice for adequate perfor-
mance for brief, and sometimes surprisingly extended,
periods. Mind-wandering in this state is less likely to be
conscious (cf. O’Connell et al., 2007; Smallwood et al.,
2007) and has been succinctly characterized as ‘‘zoning
out” (Smallwood & Schooler, 2006). The individual may
be alerted by blatant errors but will be less sensitive to
near-misses. State 2 can normally be maintained for more
than a few seconds only in a relatively quiet, uneventful
environment and/or during very tedious and undemanding
tasks. Such tasks and environments are, however, not
uncommon. Reading, for example, is generally performed
in the absence of spontaneously changing conditions. We
therefore sometimes find ourselves well beyond that last
point of recall of what we have been reading, but clearly
have continued to scan down the page some way before
recovering task awareness. Another such experience is
waiting in a car at traffic lights for a green light. When traf-
fic beside us begins to move and we may move with the
traffic in response to peripheral background features of
the driving environment rather than appropriate cues
(i.e., traffic signals) for correct task performance. Thus, if
the light change signaled a left turn for vehicles in the lane
beside us, we might find that we have anticipated the rel-
evant light change to find ourselves in the middle of the
intersection moving against a red light – at which point
we are likely to quickly recover full lucidity. Note, how-
ever, that attentional resources at such points are required
to process the error itself. Such error-induced task-relevant
processing can, ironically, further compromise task perfor-
mance, and hence lead to further errors (see bidirectional-
ity thesis below). This is one way in which errors beget
errors.
State 3: Discussions of mind-wandering have tradition-
ally stressed the ‘‘decoupling” of conscious processing from
on-line environmental sensory information as attention is
directed inward to thoughts and feelings (Schooler et al.,
2005; Singer, 1966; Smallwood & Schooler, 2006). This
has led to an important focus on changes in the individual’s
access to representations of environmental stimuli during
mind-wandering (Smallwood & Schooler, 2006). Attention
can, however, also be drawn away from motor behavior as
evidenced in action slips (Reason, 1979; Reason & Myciels-
ka, 1982). Hence, in State 3 there are also gross behavioral
indicators of mind-wandering. The individual may be
unresponsive to all but the most intrusive aspects of the
task environment. In real-world situations, such as
reading, we cease even moving down the page and stare
– blank, unmoving, and unseeing – at the page; or we fail
to drive on at all, even after the light has turned green
and the traffic around us has moved on, until we are roused
by an insistent honking behind us or the crunching of me-
tal. Ultimately, the internal train of thought may initiate a
new stream of action. In some cases, the individual may
even wander physically, getting up from a work station,
for example, to engage a colleague in a discussion concern-
ing the intruded thought. Thus, conditions permitting,
mind-wandering can lead to a change of the primary task.
Relations among states: In addition to specifying three
discrete states of attentional engagement/disengagement,
the three-state model also postulates specific relations
among states and the relations of the states to specific
errors. First, the model holds that the three states of
mind-wandering are often sequential in that an individual
will move from state to state. Second, because the states,
though discrete, are functionally and sequentially related,
the three-state model holds that individuals with a pro-
pensity of being in one state of disengagement should also
show a bias to enter other states of disengagement. None-
theless, the hypothesis of qualitative differences among
the different states the model leads to the prediction that
each level of attentional disengagement will potentially
contribute independently to cognitive and action failures.
3. Testing the three-state model
We next describe the Sustained Attention to Response
Task (SART) that constitutes a controlled context to assess
the model and provide distinct indicators for each of the
three states. Although we claim considerable generality
for the state model of task disengagement, the specific
behavioral indicators of each of the three states of mind-
wandering will vary according to the nature of the task
and hence critical features of each task must be analyzed
to determine such indicators for that task. Thus, we next
propose indicators relevant to the SART for each of the
three states and test a set of novel predictions that emerge
from the three-state model.
3.1. The SART
The Sustained Attention to Response Task (SART: Rob-
ertson et al., 1997) is a measure of sustained attention
and mind-wandering. Failures of sustained attention as
measured by the SART have been related to an increasing
number of real-world problems and psychological vari-
ables, including self-reported everyday attention failures
(Cheyne et al., 2006), general cognitive failures (Manly
et al., 1999; Robertson et al., 1997), as well as specifically
attention-related cognitive errors (Cheyne et al., 2006),
conditions such as ADHD (Bellgrove, Hawi, Gill, & Robert-
son, 2006; Bellgrove, Hawi, Kirley, Gill, & Robertson,
2005; Johnson, Kelly, et al., 2007; Johnson, Robertson,
et al., 2007; Manly et al., 2001; Mullins, Bellgrove, Gill, &
Robertson, 2005), TBI (Dockree et al., 2004; Manly et al.,
2003; O’Keeffe, Dockree, & Robertson, 2004; Robertson
et al., 1997 but see Whyte, Grieb-Neff, Gantz, & Polansky,
2007), boredom (Cheyne et al., 2006), and depression
(Carriere, Cheyne, & Smilek, 2008; Smallwood et al.,
2007), as well as the neurophysiology of attention (Bell-
grove, Hester, & Garavan, 2004; Dockree, Kelly, Foxe, Reilly,
& Robertson, 2007; Dockree, Kelly, Robertson, Reilly, &
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113 101

Foxe, 2005; Dockree et al., 2004; Fassbender et al., 2006;

1000
Garavan, Hester, Murphy, Fassbender, & Kelley, 2006;
800
Hester, Fassbender, & Garavan, 2004; Hester et al., 2005;

RTs (ms)
Kaufman, Ross, Stein, & Garavan, 2003; O’Connell et al., 600

2007). 400

200
3.2. SART Indices of attentional disengagement
0
0 50 100 150 200 250
The SART is a GO/NOGO continuous performance task in
which the NOGO stimulus appears infrequently. SART 1000

NOGO errors are typically interpreted as analogs to real- 800

RTs (ms)
world action slips caused by failures of sustained attention 600
rather than failures of inhibition (Manly et al., 1999; 400
Robertson et al., 1997). SART performance also provides
200
several behavioral indices of off-task mind-wandering that
0
could potentially lead to NOGO errors. 0 50 100 150 200 250
The first and most critical of these indices is the speed-
ing of response times (RTs), especially on GO trials immedi- 1000

ately preceding NOGO errors. The suggestion that fast RTs 800
index off-task mind-wandering is supported by (1) obser-

RTs (ms)
600
vations of negative correlations between overall RTs and
total NOGO errors across subjects (Cheyne et al., 2006; 400

Manly et al., 1999; Robertson et al., 1997), (2) evidence 200

of shorter mean sRTs in the trials immediately prior to
NOGO trials leading to SART errors (Manly et al., 1999;
Robertson et al., 1997) and shorter RTs prior to error NOGO
trials than prior to successful NOGO trials (Cheyne et al.,
2006), (3) reports of an association between task-unrelated
thought and speeded RTs during SART performance
(Smallwood et al., 2004), (4) the finding that as target
probability decreases GO RTs decrease and errors increase
(Manly et al., 1999) presumably because infrequent NOGO
trials provide opportunities for increased mind-wandering,
and (5) reports of shorter RTs and more errors with ab-
sence of awareness during probe-caught mind-wandering
(Smallwood et al., 2007).
Second, we draw attention to another potential index of
mind-wandering in the SART, namely, responses on GO tri-
als that are too fast to be responses to the GO stimuli, but
rather appear to be anticipations of the presentation of GO
stimuli. In reexamining our own SART data, we noted a
striking pattern of RTs across trials (see Fig. 1). It is evident
in Fig. 1, which provides RT data across 225 trials for each
of three subjects, that the records reveal many impossibly
fast RTs (Mean = 55.85; SD = 19.52). Anticipations are qual-
itatively different from normal RTs evidenced by a clear
band of latencies essentially free of data points below the
majority of RTs and above the extremely short RTs. Extre-
mely fast RTs (less than 100 ms) have been noted on the
SART before (e.g., Johnson, Robertson, et al., 2007) and
these anomalous values have been replaced with RTs from
adjacent trials. It occurred to us, however, that these
events might provide distinctive information about task
disengagement. Anticipations may reflect speed–accuracy
trade-offs, possibly encouraged by the fixed inter-trial
interval generally used in the SART. Some recent versions
of the SART have employed variable ITIs to discourage
speed–accuracy trade-offs. This apparent refinement may,
however, conceal an important phenomenon. Anticipa-
tions would seem to fit well our description of State 2
mind-wandering. Anticipations are too fast for the individ-
0
0 50 100 150 200 250
Trials
Fig. 1. Three sample cases of reaction times for individual subjects over
225 SART trials including anticipations (RTs below 100 ms).
ual to have processed the immediately relevant stimulus
and hence are consistent with the notion of responding
without the moment-to-moment task-relevant stimulus
monitoring (‘‘going through the motions”). This interpreta-
tion is also consistent with the view that anticipations are
fundamental form of action slips and spoonerisms (cf. pro-
lepsis; Douse, 1900; see also Norman, 1981) indicating
mind-wandering or absent-mindedness (Reason, 1979,
1984; Reason & Mycielska, 1982). Finally, the distinct gap
in the data points evident in Fig. 1 is clearly consistent with
abrupt qualitative state changes.
A third potential index of mind-wandering in the SART
task is the failure to respond to GO stimuli (i.e., omissions).
Errors of omission are among the most common form of
human error (Norman, 1981; Rasmussen, 1980; Reason,
1990, 1998). Errors of omission have been observed in
the SART with both fixed and random inter-trial intervals
and have been interpreted as a break from task engage-
ment and thus reflective of lapsing attention (Johnson,
Robertson, et al., 2007; Manly et al., 1999). This interpreta-
tion of omissions is supported by the finding that omis-
sions tend to increase along with NOGO errors in the
second half of the SART task when attentional engagement
is thought to be low (Johnson, Robertson, et al., 2007). In
addition, relative to controls, more omissions and NOGO
errors have been observed for individuals with ADHD
(Johnson, Robertson, et al., 2007; O’Connell, Bellgrove,
Dockree, & Robertson, 2004) and patients with traumatic
brain injury (Manly et al., 2003). Again, it is evident from
Fig. 1 that there is a substantial gap in the upper portion
of each record, even for the slowest responder, between
102 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
the longest RTs and the end of each trial (at 1150 ms).
Thus, transition from normal RTs to omissions appears to
be as abrupt as that for anticipations, consistent with a
qualitative change in state.
It is important to acknowledge that whether a given
response latency is an anticipation or an omission is
determined by the temporal parameters of the task. The
argument for different states, however, hinges on whether
there is a discontinuity between acceptable variations in
response times for the task. In the model, an important
executive function is to keep responding within task
parameters. Hence, in State 1, variation is kept within
acceptable ranges. Moreover, when performance approaches
the limits of acceptable response speed or precision the
data suggest that they are either brought back from that
limit or they ‘‘pop out” of that range into extremely fast
(automaticity) or very slow (in the limit – non-response).
We therefore suggest that the three SART indices of
mind-wandering described above (i.e., RT changes, antici-
pations, and omissions) are potentially valid indicators of
the three states of attentional disengagement as described
by the model.
Applying the model in the context of the SART, speeding
RTs indicate State 1. This pattern of alternating speeding
and slowing would generate increased RT variability
(Johnson, Kelly, et al., 2007). We therefore suggest that
an appropriate index of State 1 is the coefficient of variabil-
ity (CV = SD/Mean), a measure of variability independent of
mean differences. We further propose that State 2 of atten-
tional disengagement in the SART is manifest in anticipa-
tions of the SART stimuli. Anticipations occur when the
focal semantic features of the stimulus (i.e., the value of
the stimulus) are not fully processed and behavior (i.e.,
the key press) occurs automatically according to the
rhythm of stimulus presentation. Finally, we suggest that
omissions index State 3; namely, a cessation of responding
within the temporal bounds of adequate task performance.
3.3. Task disengagement and bidirectionality
By task disengagement we mean the shifting of
resources from on-line processing of task cues to off-line
processing of materials not relevant to immediate task
demands. Importantly, processing during task disengage-
ment need not be entirely task irrelevant (Smallwood
et al., 2004); merely irrelevant to occurrent task demands.
Mind-wandering is often more than idle wool-gathering
and may involve serious off-line problem solving (Kane
et al., 2007; Singer, 1966, 1978). Indeed, because of the
assumption of sufficient background competence and
motivation, our model is compatible with the likelihood
that much off-line processing is relevant to task perfor-
mance and is, indeed, often precipitated by variations in task
performance. Thus, even under optimal task conditions of
competence and motivation, both successful and especially
unsuccessful performance events, will induce off-line
analysis of performance. This leads to the prediction that
the performance – mind-wandering association is
bidirectional.
It has been commonly noted that errors redirect atten-
tion back to the primary task at hand (Robertson et al.,
1997; Sarason, Sarason, Keefe, Hates, & Shearin, 1986;
Smallwood, Davies, et al., 2004; Smallwood, O’Connor,
et al., 2004). Recent evidence suggests, however, that
reflection on task performance accounts for substantial
off-task processing, sometimes referred to as Task-Related
Interference (Smallwood, Davies, et al., 2004; Smallwood
et al., 2003). Smallwood and colleagues specifically report
high levels of Task-Related Interference following errors
(Smallwood, Davies, et al., 2004). Thus, although it may
be the case that errors redirect attention back to the task,
the redirection may often be in the form of off-line internal
analysis, including evaluation of performance, itself a form
of reactive mind-wandering, that may interfere with
ongoing task performance. In the extreme, the bidirectional
relation between mind-wandering and attention-related
errors in performance can snowball with disastrous conse-
quences in which off-line analysis leads to complete
neglect of the task at hand (e.g. flying an aircraft or moni-
toring a nuclear facility, cf. Reason, 1979; Vincente, 2004).
3.4. Predictions
Following are general sets of predictions that emerge
from a three-state bidirectional model of attentional disen-
gagement when applied to SART performance.
Relations among SART indicators and general measures of
mind-wandering: In general, the model predicts that RT CV,
anticipations, and omissions will all be correlated with
NOGO errors and, because they represent distinct states
of mind-wandering, each of these measures will contribute
independent variance to the prediction of NOGO errors.
Thus, the first set of specific predictions concerns associa-
tions at the level of the individual differences between
the proposed SART indicators of the states of disengage-
ment (i.e., RT CV, anticipations, and omissions) and self-re-
ported indices of everyday mind-wandering and
attentional failure. The model predicts positive correla-
tions among RT CV, anticipations, and omissions on GO tri-
als as well as between each of these and SART errors, self-
reported mind-wandering measured by the Mindful
Awareness and Attention Scale – Lapses Only (MAAS-LO),
and self-reported attentional failures measured by the
Attention-Related Cognitive Errors (ARCES) scale. These
predictions fundamentally assume stable individual differ-
ences in propensities for mind-wandering and the external
validity of the SART measures.
We note that some of these predictions are counter-
intuitive from an inhibition perspective, given that antici-
pation would seem to index strong disinhibition whereas
omissions would appear to reflect complete inhibition.
According to the model, however, anticipations and omis-
sions represent successive states in attentional disengage-
ment related to task performance/nonperformance and
hence should be positively mutually associated and both
should be positively associated with commission errors
on NOGO trials. In our earlier analyses (Cheyne et al.,
2006), we found, consistent with previous and subsequent
research, that overall RTs were significantly and robustly
negatively associated with NOGO errors (r = .64,
p < .001). Previously, however, RTs included the anticipa-
tions discussed above and, as argued above. For the present
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
study, we therefore calculated RT and RT CV only for trials
with RTs above 200 ms and counted as anticipations all RTs
less than 100 ms. The very few ambiguous RTs (i.e., be-
tween 100 and 200 ms) were ignored.
In our earlier research, we reported that a self-report
measure of mind-wandering propensity, the Mindful
Awareness of Attention Scale (MAAS) was significantly
associated with SART GO RTs and accounted for the associ-
ation between the self-report scale of Attention-Related
Cognitive Errors (ARCES) and RTs (Cheyne et al., 2006).
On the other hand, the ARCES contributed unique variance
to the prediction of SART NOGO Errors. This pattern of re-
sults was consistent with the hypothesis that the MAAS
and SART RT speeding directly reflect mind-wandering,
whereas ARCES and SART NOGO errors are cognitive per-
formance consequences of mind-wandering. We therefore
predicted that a slightly revised scale, focusing on lapses
only, the MAAS-LO (Carriere et al., 2008) and the ARCES
would be significantly correlated with RT CV, as well as
with anticipations and omissions but that the MAAS-LO
would account for the association of the ARCES with each
of these measures.
Temporal relations and transitions between states: A crit-
ical aspect of the three-state model of attentional disen-
gagement is the hypothesis that the three states are
sequentially related. Hence, speeding of RT should occur
on trials immediately prior to anticipations and omission
as well as following anticipations. Second, the probability
of anticipations and omissions should increase on trials
just before and after omissions and anticipations.
Bidirectionality of mind-wandering and NOGO errors: The
model states that mind-wandering and the resulting inat-
tention to the task leads to errors and to reflection on those
errors, consequently directing attentional resources away
from current task conditions. Thus, the bidirectional aspect
of the model predicts that all performance tests (NOGO tri-
als in the SART), successful as well as unsuccessful, will
lead to reflection on task performance and hence provide
occasions for mind-wandering. To assess these predictions,
we examined RT changes, anticipations, and omissions fol-
lowing as well as preceding both correct and incorrect
NOGO trials. We predict that, in the context of the SART,
all indicators of mind-wandering (RT variability, anticipa-
tions, and omissions) will serve as both precursors and
consequences of both successful and unsuccessful task
performance.
The bidirectional model leads to one further prediction.
If performance tests (i.e., NOGO trials) lead to increased
mind-wandering then closely spaced NOGO trials should
also lead to increased probability of errors, especially fol-
lowing a recent NOGO error.
4. Method
4.1. Participants
Subjects were selected from among prior respondents
to a WWW survey on sleep paralysis (see Cheyne et al.,
2006). The final selection of participants who completed
the full study included 349 females, 155 males (N = 504).
Participants ranged from adolescence to senescence, with
103
a mean age of 32.23 (SD = 11.23). Blank or incomplete sub-
missions were removed from analysis and any instances of
multiple submissions from the same participant were
removed. Similarly, results from participants who failed to
complete all 225 trials of the SART, or responded in opposite
manner to that in which they were instructed (i.e., key press
provided only when the target appeared), were removed
from the data set. Of an original 717 participants, total of
213 participant responses were removed in this manner.
4.2. Measures
RT CVs were calculated for all response latencies over
200 ms. RTs less than 100 ms were coded as anticipations.
If no response was made during a GO trial this was re-
corded as an omission. Responses to the NOGO stimulus
(‘‘3”) were coded as NOGO errors and successful withhold-
ing of responses as NOGO successes. The self-report mea-
sures included the 12-item ARCES and the 13-item
revised Mindful Awareness of Attention with Lapses Only
(MAAS-LO; see Carriere et al., 2008).
The SART procedure, as described by Robertson et al.
(1997), involves 225 single digits, 25 each from 1 to 9, pre-
sented for 250 ms and followed by an encircled  mask ()
for 900 ms, for a total digit-to-digit duration of 1150 ms.
Participants were instructed to respond via a key press to
each digit, unless that digit is a 3. The digits are randomly
distributed, as are the font sizes of all 225 trials, with equal
representation of 48, 72, 94, 100, and 120 point size Sym-
bol font in white against a black background. The SART was
recreated in Macromedia Flash MX 2004 according to these
specifications, in order to allow on-line presentation of the
SART with the rest of the questionnaires. Actual stimulus
displays were dependent on the computer equipment used
by each participant and viewing distance was not con-
trolled. On a Compaq Presario R3140CA series notebook,
with 15.4 in WXGA (1280  800) display, used for scale
development, stimulus height varied from 8.74 mm
(34 px) to 21.33 mm (83 px) and the circular mask diame-
ter was 21.07 mm (82 px).
4.3. Procedures
Participants received an informational e-mail regarding
the study, including a link to the study website and a pre-
assigned personal identification number. Upon visiting the
study website, participants received further instructions
regarding the purposes of the study and were given the
opportunity to participate by completing, in random and
undisclosed order, the ARCES, MFS, and MAAS. Participants
completed the three measures in a single session, followed
by the SART, at the end of which they were given a feed-
back page thanking them for their participation and pro-
viding further information about the purposes of the
study and links to a website with additional information
on sleep paralysis.
4.4. Lag analyses
Lag RTs were calculated as follows. For each NOGO trial,
separately for correct and incorrect responses, the RTs for
104 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
the four preceding trials and the four trials following the
NOGO trial were tabulated. RTs were then averaged within
each lag. The same procedure was repeated for four trials
preceding and four trials following anticipations and omis-
sions. Lag probabilities for anticipations and omissions
were calculated as follows. For each NOGO trial the four
preceding and the four following trials were analyzed sep-
arately for correct and incorrect responses. Probabilities for
a given trial type were generated by taking the number of
occurrences of that type at a given lag, divided by the total
number of trials counted at that lag (in most cases this was
equal to the number of errors considered, except when an
error occurred within four trials of the beginning or end of
the experiment). These analyses were performed using a
command line tool written in C.
To ensure that our measures of RTs, anticipations and
omissions before and after NOGO errors did not overlap,
it was necessary to exclude some trials and participants
from the analyses. Hence, sample sizes vary slightly from
measure to measure and between comparisons within
measures and are reflected in the changing degrees of free-
dom of the F-tests. For RT, for example, we recorded as
missing data trials preceding or following NOGO errors
that included omissions, anticipations, or other NOGO
trials. Missing data also arose in cases in which RT was
not available because the end or beginning of the session
occurred within four trials of the target NOGO trial. For
anticipations and omissions, conditional probabilities for
mutual transitions could be calculated only for individuals
committing anticipations or omissions.
Given the statistical power of the large sample size and the
multiple repeated measures the lower-bound degrees-of-
freedom were used for all repeated-measures analyses, lead-
ing to further variations in degrees of freedom across tests.
5. Results and discussion
5.1. Associations among indicators and consequences of
mind-wandering
We began by testing our prediction that the three target
variables (RT CVs, anticipations, and omissions on GO tri-
als) as measures of states of mind-wandering generalize
beyond the SART. Hence, by hypothesis, each of the SART
indices will be correlated with one another, with NOGO Er-
rors, and with self-reported mind-wandering measured by
the MAAS-LO and self-reported attentional failures mea-
sured by the ARCES. The correlations among each of these
measures are shown in Table 1.
Table 1
Means, SDs, and Pearson product-moment correlations among SART measures (GO RTs, anticipations, omissions, and NOGO errors) and self-report measures of
mind-wandering (MAAS-LO) and Attention-Related Cognitive Errors (ARCES). Means are in italics on diagonals. SDs are in parentheses, for all correlations,
p < .01, N = 504.
RT CV Anticipations
RT CV 0.28 0.70
Anticipations (0.12) 3.13
Omissions (6.76)
NOGO errors
MAAS-LO
ARCES
As can be seen in Table 1, the three SART indicators (RT
CV, anticipations, and omissions) were significantly and
strongly correlated with one another. Also as predicted,
RT CV, anticipations, and omissions were each robustly cor-
related with NOGO Errors. These correlations corroborate
the interpretation of, and the model predictions for, RT
CV, anticipations and omissions as indices of mind-wan-
dering in the SART task. Also consistent with predictions,
RT CV, anticipations, and omissions were significantly cor-
related with self-reported mind-wandering as measured
by the MAAS-LO and with self-reported action slips as-
sessed by the ARCES. Moreover, extending our previous
findings (Cheyne et al., 2006) that the association of the
MAAS-LO with the RTs completely accounted for associa-
tion of the ARCES and RT, a series of regression analyses,
replicated this result for RT CV, anticipations, and omis-
sions. The ARCES contributed no additional significant pre-
dictive power for any of these variables beyond that of the
MAAS-LO, though, also consistent with previous analyses,
it did for NOGO errors, b = .13, t = 2.02, p < .04. This finding
is consistent with the conclusion that anticipations and
omissions, like SART RTs, are best interpreted as direct
indices of mind-wandering.
A further prediction arising from the three-state model
is that different states of mind-wandering make indepen-
dent contributions to cognitive errors and action slips.
Consistent with this prediction, a multiple regression anal-
ysis with NOGO errors as the dependent variable, yielded
significant coefficients for RT CV, anticipations, and omis-
sions (see Table 2). Anticipations made slightly stronger
contributions than the other two variables and omissions
made slightly weaker contributions to the prediction of
SART errors. The latter finding may reflect a peculiarity of
omissions in the SART task. Because the critical feature of
successful task performance requires the withholding of a
response, omissions will occasionally lead to spuriously
successful NOGO trials, attenuating the general positive
association of omissions with NOGO errors. It is predicted
that omission will provide even better prediction for a var-
iant of the SART that requires an alternative positive re-
sponse (rather than none).
Temporal associations among the SART indicators of mind-
wandering RTs and anticipation: RTs for the four trials (lags)
immediately preceding and following anticipations were
analyzed (Fig. 2). The data were analyzed in a 2 (pre-antic-
ipations vs. post-anticipations) by 4 (lag) repeated-mea-
sures ANOVA for individuals making anticipations. This
analysis yielded significant temporal order and lag effects,
which were superseded by a significant temporal order by
Omissions NOGO errors MAAS-LO ARCES
0.62 0.50 0.21 0.16
0.53 0.52 0.17 0.12
3.40 0.42 0.19 0.13
(5.79) 10.89 0.31 0.28
(5.83) 54.76 0.73
(12.05) 49.40
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113 105

Table 2 0.16 Anticipation->Omission p(O|A)

Multiple regression predicting NOGO errors with RT CV, anticipations, and Omission->Anticipation p(A|O)
omissions. 0.14
Anticipation Baseline p(A)
Dependent variable: NOGO errors b t p 0.12
Omission Baseline p(O)

p(A|O) or p(O|A)
RT CV 0.21 3.71 .001 0.10
Anticipations 0.30 5.72 .001
Omissions 0.13 2.77 .006 0.08
R2 = .32, F(3, 500) = 77.74, p < .001
0.06

0.04

400 0.02

0.00
350
t-4 t-3 t-2 t-1 t0 t+1 t+2 t+3 t+4

300 Pre A or O A or O Post A or O

250 Fig. 3. Mean conditional probabilities of anticipations/omissions (with

SEs) for four trials preceding and four trials following omissions/
RT (ms)

PrePost Anticipation anticipations.

200
150
100
50
0 ing (t  1) and following (t + 1) anticipations and omis-
t-4 t-3 t-2 t-1 t0 t+1
Pre-Anticipation Ant.
Fig. 2. Mean SART RTs (with SEs) for four trials preceding and four trials
following anticipations.
lag interaction, F(1, 162) = 23.42, p < .001, g2 = .31. This
interaction was disambiguated by analyses of the lag ef-
fects for pre- and post-anticipation RTs. RTs preceding
anticipations showed a robust significant linear decrease,
F(1, 162) = 110.17, p < .001, g2 = .39, and a smaller qua-
dratic component, F(1, 162) = 47.88, p < .001, g2 = .22. RTs
following anticipations appeared to show a near mirror-
image linear decline, although RT speed was significantly
faster immediately prior to, than following, anticipations.
t(197) = 9.44, p < .001 (see Fig. 2).
Anticipations and omissions: To assess the sequential
associations of anticipations and omissions, the condi-
tional probabilities of anticipations for the four trials
immediately preceding and following omissions and the
conditional probabilities of omissions preceding and fol-
lowing anticipations were analyzed in a 2 (anticipation
vs. omission) by 2 (pre vs. post) by 4 (lag) repeated-mea-
sures ANOVA for individuals making anticipations (see
Fig. 3). This analysis yielded significant temporal order
and lag effects, which were superseded by a significant
temporal order by lag interaction, F(1, 207) = 49.68,
p < .001, g2 = .19 (see Fig. 3). The interaction was disambig-
uated by analyses of the lag effects for pre- and post-antic-
ipations and omissions combined. Preceding conditional
probabilities showed a robust significant linear increase
with a marked inflection between t  2 and t  1, well
above the mean baseline, yielding significant linear
F(1, 210) = 62.36, p < .001, g2 = .23, quadratic, F(1, 210) =
32.69, p < .001, g2 = .13, and cubic, F(1, 210) = 10.48,
p < .001, g2 = .05, effects. Probabilities following anticipa-
Baseline
tions and omissions showed the expected near mirror-
image pattern, with virtually identical results, including
linear, F(1, 208) = 62.90, p < .001, g2 = .23, quadratic,
F(1, 172) = 27.56, p < .001, g2 = .12, and cubic, F(1, 208) =
9.49, p < .001, g2 = .04, effects. Trials immediately preced-
t+2 t+3 t+4
sions were, significantly above the overall baseline
Post-Anticipation probabilities.
RTs and omissions: To assess the temporal association of
RTs and omissions, the RTs for the four trials immediately
preceding and following omissions were analyzed in a 2
(pre vs. post) by 4 (lag) repeated-measures ANOVA for
individuals making anticipations (Fig. 4). This analysis
yielded significant order and lag effects, which were super-
seded by a significant order by lag interaction,
F(1, 205) = 22.31, p < .001, g2 = .25. This interaction was
disambiguated by analyses of the sequential lag effects
for pre- and post-omission RTs. Preceding RTs showed a
robust significant increase leading up to t0, well above
the mean baseline, yielding significant linear F(1, 220) =
62.36, p < .001, g2 = .23, and quadratic, F(1, 210) = 31.69,
p < .001, effects. Conversely, there was a significant de-
crease following t0, yielding significant linear, F(1, 208) =
62.90, p < .001, g2 = .23, and quadratic, F(1, 208) = 27.56,
p < .001, g2 = .12, effects. The deviation from linearity in
both cases is clearly evident at t  1 and t + 1. There was
substantial slowing of RTs on the trial immediately before
an omission. Hence, extremely slow RTs are harbingers of
omissions. Had our inter-stimulus interval been shorter,
the pre-omission RTS responses would have themselves
been omissions. Ultimately, omissions are simply very long
response times that exceed the available time for the task.
Nonetheless, it clear that there is an abrupt transition from
moderately slower RTs in the normal range to very long
RTs. This result and the post-omission speeding of RTs
are consistent with abrupt transitions between states.
5.2. Bidirectionality of mind-wandering and attentional
errors
The final set of predictions that we tested concerned the
bidirectionality of mind-wandering and task challenges
106 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113

550 420 Correct

PrePost Omission RTs
Baseline Error
500 400 Baseline

450 380

RT (ms)
360
RT (ms)

400

350 340

320
300

300
250
t-4 t-3 t-2 t-1 t0 t+1 t+2 t+3 t+4

Pre-NOGO NOGO Post-NOGO

200
t-4 t-3 t-2 t-1 t0 t+1 t+2 t+3 t+4
Fig. 5. Mean SART RTs (with SEs) for four trials preceding and four trials
Pre-Omission O Post-Omission following NOGO incorrect and correct trials.

Fig. 4. Mean SART RTs (with SEs) for four trials preceding and four trials
following omissions.
such as attentional errors. Specifically, we predicted that
each of the SART indicators of mind-wandering should
both precede and follow NOGO trials, particularly those
involving errors. To evaluate these bidirectional relations,
we assessed RTs, anticipations, and omissions on four trials
(lags) before and four trials (lags) after correct and incor-
rect NOGO trials.
RTs preceding and following NOGO trials: The data were
analyzed in a 2 (correct vs. incorrect) by 2 (pre- vs. post-
NOGO trial) by 4 (lag) repeated-measures ANOVA. All ef-
fects were highly significant, including the three-way
interaction, F(1, 481) = 36.18, p < .001, g2 = .07 (see Fig. 5).
This interaction was disambiguated by a series of analyses
of the lag effects for correct and incorrect responses, pre-
and post-NOGO trials. RTs preceding successful NOGO tri-
als, were all significantly above baseline and characterized
by increasing RTs, with significant linear, F(1, 494) =
111.70, p < .001, g2 = .18, and quadratic components,
F(1, 481) = 16.12, p < .001, g2 = .03. Conversely, RTs preced-
ing error NOGO trials were all below baseline RTs and de-
creased prior to a NOGO error, with significant linear,
F(1, 492) = 6.68, p < .01, g2 = .01, and cubic components,
F(1, 492) = 12.12, p < .001, g2 = .02. RTs following successful
NOGO trials, dropped markedly below baseline and then
rapidly recovered, yielding significant linear, F(1, 498) =
126.42, p < .001, g2 = .20, quadratic, F(1, 481) = 75.74,
p < .001, g2 = .13, and cubic components, F(1, 481) = 13.54,
p < .001, g2 = .03. In contrast, RTs following error NOGO tri-
likelihood of an anticipation following an error was greater
than for the other three conditions, p < .001. The finding
that anticipations were not significantly more likely pre-
ceding errors than correct NOGO trials was somewhat
counter-intuitive, especially given that anticipations dur-
ing a NOGO trial create an automatic error. Nonetheless,
the increased probability of anticipations following errors
is consistent with the bidirectional model.
Omissions preceding and following NOGO trials: 2 (correct
vs. incorrect) by 2 (temporal order: pre-NOGO trial vs.
post-NOGO trial) by 4 (lag) repeated-measures ANOVA
for omissions yielded a significant three-way interaction,
F(1, 494) = 21.51, p < .001, g2 = .04 (see Fig. 7). A series of
lag analyses for pre-error trials, but not pre-success trials
yielded a marginal cubic effect, F(1, 481) = 7.70, p < .006,
g2 = .02. There was a significant increase in omissions in
the trial immediately preceding an error. Omissions fol-
lowing a success decreased and then returned to near
baseline, producing a significant quadratic effect,
F(1, 481) = 8.33, p < .001, g2 = .02. Omissions following an
error increased then declined, yielding significant linear,
F(1, 481) = 56.07, p < .001, g2 = .10, and quadratic compo-
nents, F(1, 481) = 19.44, p < .001, g2 = .04.
0.04 Anticipations
Error p(A|E)
Correct p(A|C)
Baseline p(A)
0.03
p(A|E) or p(A|C)

als increased to baseline levels and then resumed signifi-

0.02
cantly below baseline and not different from preceding
RTs yielding significant quadratic, F(1, 497) = 23.08,
p < .001, g2 = .04, and cubic components, F(1, 497) = 7.55, 0.01
p < .001, g2 = .02. The mean RT for NOGO errors is also
shown in Fig. 5. The mean NOGO error RT is significantly
0.00
faster than mean RTs for any lag.
t-4 t-3 t-2 t-1 t0 t+1 t+2 t+3 t+4
Anticipations preceding and following NOGO trials: A 2
(correct vs. incorrect) by 2 (pre- vs. post-NOGO trial) by Pre-NOGO NOGO Post-NOGO

4 (lag) repeated-measures ANOVA of anticipations yielded

Fig. 6. Mean conditional probabilities of anticipations (with SEs) for four
a significant two-way accuracy by temporal order interac- trials preceding and four trials following NOGO incorrect and correct
tion, F(1, 494) = 14.84, p < .001, g2 = .03 (see Fig. 6). The trials.

0.06
Error p(O|E)
0.05 Correct p(O|C)
Baseline p(O)
p(O|C) or p(O|E)
0.04
0.03
0.02
0.01
0.00
t-4 t-3 t-2 t-1 t0 t+1
Pre-NOGO NOGO
Fig. 7. Mean conditional probabilities of omissions (with SEs) for four
trials preceding and four trials following NOGO incorrect and correct
trials.
The results for omissions following a NOGO error are
perhaps the most striking of all, particularly at t + 1. Not
only are RTs lengthened, and anticipations increased, but
omissions are markedly increased, with incomplete recov-
ery over the four post-error trials (see Fig. 7).
Interval length and errors: The final prediction concerns
the effect of mind-wandering induced by NOGO trials.
The previous results reveal increases in the indices of
mind-wandering. Therefore, error trials following soon
after a NOGO trial should show increased error rates. This
hypothesis was tested by assessing the association of
length of interval between NOGO trials with probability
of an error at the end of the interval (i.e., the proportion
of subjects making an error after varying interval lengths).
The bidirectional hypothesis suggests, however, that the
effect should be especially strong for short intervals and
hence this relation should show significant non-linear
components. Separate linear and quadratic regression
analyses were conducted regressing probability of NOGO
errors following a correct prior NOGO trial, P(E|C), and
NOGO errors following an incorrect prior NOGO trial,
P(E|E). Analysis of successful prior challenges reveals inde-
pendent linear and quadratic effects, linear: b = .76,
t = 6.53, p < .001; quadratic: b = .33, t = 2.82, p < .010. The
overall model generated a robust effect of R2 = .72,
F(1, 22) = 26.45, p < .001. The R2 change produced by add-
ing the quadratic component was also significant,
R2change ¼ :11, F(1, 21) = 7.97, p < .010. Analysis of unsuccess-
ful prior challenges also revealed independent linear and
quadratic effects, linear: b = .76, t = 7.08, p < .001; qua-
dratic: b = .26, t = 2.28, p < .033. The overall model gener-
ated a substantial effect of R2 = .80, F(1, 22) = 26.45,
p < .001. The R2 change produced by adding the quadratic
component was also significant, R2change ¼ :07, F(1, 21) =
5.18, p < .033.
The function for errors following shorter intervals ap-
pears somewhat steeper than that following correct with-
holds. To assess the effect the difference in the linear
component for correct and incorrect trials a SEM model
was constructed in which the linear components for Inter-
val length and both P(E|C) and P(E|E) were constrained to
be equal. The results of the analysis of this model were
J. Allan Cheyne et al. / Cognition 111 (2009) 98–113 107
compared to those of an unconstrained (saturated) model
Omissions
in which all parameters were free to vary. This comparison
resulted in a significantly poorer fitting model,
v2(1) = 11.81, p < .001. Hence, the linear component of
the association of interval length and P(E|C) and P(E|E) dif-
fer significantly. The linear component was significantly
stronger for P(E|E) that P(E|C). In contrast, a parallel analy-
sis constraining the quadratic components to equality
yielded a well fitting model, v2(1) = 0.02, p = .90. Hence,
the quadratic component of the associations of interval
length and P(E|C) and P(E|E) did not differ significantly.
Summary: The present results are generally consistent
t+2 t+3 t+4
with the hypothesis that the three indices of mind-wan-
Post-NOGO
dering are associated with errors on NOGO trials, and
two of these indices (RT speeding and omissions) preceded
errors at a rate significantly above baseline. Errors, in turn,
lead to faster RTs and increased probabilities of both antic-
ipations and omissions. Successful trials are less likely than
errors to lead to mind-wandering as indicated by a brief,
one-trial speeding of RTs, a brief, one-trial increase in
anticipations, but no evidence of an increase in omissions.
Errors lead, in contrast, to a sustained increase in anticipa-
tions and an especially dramatic increase in omissions. The
results generally support the conclusion that the relation
between mind-wandering and attentional errors on NOGO
trials is bidirectional.
In addition, consistent with the model, many of the
transitions between indices appear to be rather abrupt.
Thus, an abrupt increase in RT presaged the occurrence of
an omission, but this occurred at t  1, as reflected in a sig-
nificant deviation from linearity in the trials preceding an
omission. The transition between anticipations and omis-
sions was similarly qualitative.
6. General discussion
We have described and tested a three-state bidirec-
tional model of attentional engagement/disengagement.
According to this model, and consistent with the evidence
presented, attentional disengagement during a sustained
attention task can be described as three distinct states of
mind-wandering. Indices of these three states are posi-
tively associated with one another, with NOGO errors,
and with self-reported measures of mind-wandering and
attention-related cognitive errors in everyday settings
and tasks in theoretically consistent patterns. The relations
between SART and self-report measures suggest that the
SART measures generalize beyond the SART task and pro-
vide potentially useful individual difference behavioral
measures of mind-wandering.
We also found evidence that indices of these states pre-
cede and follow one another as well as NOGO errors at
above chance levels. Particularly striking is the fact that
anticipations and omissions, which represent opposite ex-
tremes of RT, are positively correlated with one another
across individuals, are sequentially associated in time
and, although also positively correlated with variation
within the normal range of RT (RT CV), and are associated
with NOGO errors independently of such variation and
one another. One alternative possibility is that anticipa-
tions reflect very late responses that carry over to the next
108 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
trial. This is consistent with the positive associations found
both across individuals and over time (trials). It is, how-
ever, not consistent with the finding of independent asso-
ciations of each with NOGO errors. Moreover, although the
carry-over interpretation is consistent with the strong
transition from omissions to anticipations (i.e., that an
anticipation is the much delayed response to the previous
trial, which would have been an omission), it is clearly
inconsistent with the significantly stronger likelihood of
transition from anticipation to omission (see Fig. 3).
The results for the lag analysis of RTs preceding NOGO
errors were consistent with prior findings and interpreta-
tions of the attentional implications of speeding of SART
GO RTs. This pattern is particularly consistent with the
interpretation of SART RT change directly reflecting atten-
tional task disengagement. The present results also add the
finding that longer RTs, within the normal range, prior to a
NOGO trial are associated with an increasing probability of
a successful withholding, suggesting that changing RTs
within the normal range reflect both increases and de-
creases in attentional engagement in the task. In addition,
although mean NOGO error RTs were faster than RTs at any
lag they were still within normal RT range and appear to
reflect a linear trend of increasing State 1 task disengage-
ment prior to, and during, an error. Nonetheless, the most
significant task events for the SART are errors on NOGO tri-
als. These appear to produce substantial task interference
as indexed by increased anticipations and omissions fol-
lowing errors. This interference appears to persist for sev-
eral trials (Figs. 7 and 8). Persisting interference is further
indexed by the mutual and self-sequencing of anticipations
and omissions (Fig. 3) as well as the effects of NOGO trial
interval length. Moreover, although the immediate effect
on RT was actually greater for successful NOGO trials, the
effect on RT was more protracted following an error
(Fig. 5). Thus, RT results are also consistent with the
hypothesis that errors may lead to longer off-task rumina-
tion than successful performance.
In the trial immediately following a NOGO error there
was a significant slowing of RTs. This would appear to be
consistent with the hypothesis of increased caution and
on-task attentional engagement (Manly et al., 1999;
Robertson et al., 1997). Alternatively, however, and consis-
tent with the foregoing arguments, this one-trial effect
may reflect a brief disruptive reaction to the error, that
is, a task-relevant distraction. Interestingly, in the trial
immediately following a successful NOGO trial, there was
a significant and substantial speeding of RT. Indeed, this
was the only trial around a NOGO trial in which RTs asso-
ciated with correct response were faster than those associ-
ated with an error. This may also reflect a brief distraction
from the task as subjects evaluate their performance.
Catastrophic mind-wandering: There were extremely
abrupt transitions in response times indexing state transi-
tions, anticipations and omissions, following lesser degrees
of acceleration and deceleration of RTs. That is, abrupt
state transitions to anticipations and omissions were pre-
saged in the immediately preceding trials by smaller devi-
ations in the same direction. Thus, the precipitous drop in
RTs (350 ms) that defined anticipations were signaled by
smaller but significant speeding of RT (Fig. 2). Conversely,
Fig. 8. Quadratic regression of probability of NOGO error on length of
interval following successful, P(E|C), and unsuccessful, P(E|E), NOGO
trials.
slowing in RT signaled an upcoming omission, which, in
the context of the present task, entailed a dramatic single
trial mean change of over 500 ms (Fig. 3). Such abrupt
shifts provide corroboration for the claim of qualitatively
different states. The pattern of findings is also consistent
with the existence of an underlying process with a transi-
tional period of quantitative changes (speeding and slow-
ing of task-relevant responses within task parameters)
and a reduction of on-line attentional processing followed
by an abrupt collapse of on-task processing. According to
the model, in the case of anticipations, the collapse repre-
sents a cessation of task analysis and, in the case of omis-
sions, a failure of response preparation.
Bidirectionality: We found considerable evidence for
bidirectionality between mind-wandering and attention-
related errors. All three indicator variables occurred as
reactions to errors (Figs. 6–8). Anticipations and omissions
were particularly sensitive to errors, suggesting that NOGO
trials and particularly errors on NOGO trials constitute
important occasions for reactive mind-wandering. In all
cases, it seems possible that reactive mind-wandering
reflects attempts to reengage with the task, but which,
paradoxically, at least briefly adversely affect ongoing
performance.
Bidirectionality may be relevant to a recent distinction,
similar that proposed in the present paper, between
sustained and transient aspects of control processing
(Botvinick, Braver, Barch, Carter, & Cohen, 2001; Braver,
Barch, Gray, Molfese, & Snyder, 2001; Braver, Reynolds, &
Donaldson, 2003; Burgund, Lugar, Miezin, & Peterson,
2003; Cohen, Botvinick, & Carter, 2000; Donaldson, Peter-
son, Ollinger, & Buckner, 2001; MacDonald, Cohen, Stenger,
& Carter, 2000; Miller & Cohen, 2001; Mitchell et al., 2007;
Velanova et al., 2003; Visscher et al., 2003). A sustained or
tonic mode of attention involves monitoring of task condi-
tions and behavioral performance and has been associated
with prefrontal activity. The second, transient mode, is an
occurrent form of attention, reactively engaged by behav-
ioral and cognitive (cf. Mitchell et al., 2007) conflicts and
has been associated neurologically with anterior cingulate
cortex activity (Carter et al., 1998; Cohen et al., 2000). Sus-
tained and transient control processes work in tandem.
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
The first process engages in active, ongoing monitoring of
task conditions and engages a reactive process when out-
comes conflict with task goals (i.e., errors). Reactive pro-
cessing then interferes with analysis of occurrent task
demands (State 2) or produces failure of task performance
(State 3).
The present findings also shed light on a recently re-
ported a principled components analysis of SART RTs in
blocks between critical events (NOGO trials or thought
probes) (Smallwood, McSpadden, Luus, & Schooler, 2008).
The first of three components extracted appeared to be a
conventional ‘‘size” factor and, as such, likely reflects con-
sistency in individual differences in mean RT across trials,
possibly reflecting individual differences in Stage 1 pro-
pensities. There appear to have been differences in this
component between correct and incorrect trials as well
as between mind-wandering with and without awareness
(Smallwood et al., 2008, Fig. 1). The second and third com-
ponents appeared to be more bipolar and contrasted trials
following one critical event and, particularly the second
component, those preceding the next critical event. That
is, early RTs appeared to be mutually positively correlated
with one another, as are those of late trials, whereas early
and late trials are negatively correlated. Thus, individuals
who show reactive mind-wandering appear to have a dif-
ferent pattern of RTs prior to the next critical event from
those of less reactive individuals. If this interpretation is
correct, the bidirectional aspect of the present model pre-
dicts that this component should be stronger in blocks ter-
minated by an error. Consistent with this, error blocks
produced higher weightings on the second component
than did baseline blocks. Moreover, correct blocks appear
to generate intermediate weights on Factor 2, also consis-
tent with present model.
A paradox observed, explained, and generalized: The in-
creased probability of errors following very short intervals
following prior NOGO trials, especially unsuccessful NOGO
trials, may seem at odds with prior research indicating that
tasks with long intervals or slow presentation rates
increase error rates (Antrobus, 1968; Smallwood et al.,
2003, 2004; Teasdale et al., 1993; but see Smallwood
et al., 2008). There is an important difference however, be-
tween tasks with long intervals providing many mind-
wandering opportunities, and episodes of repeated chal-
lenges interspersed among long periods of inactivity. Thus,
combining the present results with previous findings, we
hypothesize that situations most conducive to mind-
wandering are those with long periods of inactivity (ran-
dom or periodic mind-wandering) interspersed with bouts
of distracting critical events (reactive mind-wandering).
Such situations are not uncommon in reading and driving
tasks, both of which may have long ‘‘fallow” periods inter-
spersed with critical events (passages critical for future
understanding of text, encountering construction zones
or accident scenes). In addition, real-world catastrophes
often entail series of closely spaced cumulative errors,
which would likely compound the reactive effects sug-
gested by our bidirectional model (Three Mile Island,
Bhopal, Chernobyl: see Reason, 1990, Appendix).
The patterns of RTs, anticipations, and omissions could
arise through a combination of at least three factors: (1)
109
exogenous task and non-task events (2) endogenous rhyth-
mic attentional oscillations (see below), and (3) explicit
strategic modifications of attention based on expectations.
Thus, it is possible that the high rate of errors found for clo-
sely spaced NOGO trials might reflect a lowered expecta-
tion of another NOGO trial immediately following
preceding NOGO trials. Perhaps, subjects strategically relax
vigilance following NOGO trials and only gradually in-
crease on-task attention as the expectation of a NOGO trial
grows over succeeding trials. Studies underway attempt to
assess this possibility by manipulating the probability of
NOGO trials at different trial intervals. Preliminary results
suggest that the function presented in Fig. 8 is independent
of such manipulations.
6.1. Future studies: further predictions from the three-state
model
In what follows we consider further potential corre-
lates, causes, and effects of the three states of attentional
disengagement at behavioral, phenomenological, physio-
logical, cognitive, and neural levels.
The influence of endogenous attentional rhythms: The lin-
ear trends found in almost all analyses likely reflect
increasing inattention across trials preceding NOGO errors.
In a number of cases for all attentional indices, however,
there were also frequent quadratic and cubic trends. These
latter results may reflect unstable, fluctuating attentional
engagement throughout the task, suggesting that task
engagement is an unstable and constantly fluctuating set
of states. A further possibility that should be acknowledged
is that the state transitions may be affected by an underly-
ing periodicity of attentional states or arousal, or underly-
ing neural oscillator driving manual responses, or
indirectly affecting attention (Gilden & Hancock, 2007;
Johnson, Kelly, et al., 2007; Smallwood et al., 2008).
Smallwood and colleagues, for example, using a principal
components analysis, report systematic fluctuations in
RTs leading up to SART errors or probe-detected mind-
wandering.
The increasing RTs prior to successful NOGO trials may
suggest the possibility of an explicit strategy of allocating
attentional resources to the task as time passes and the
expectation of a NOGO trial increases. Speeding following
errors, conversely, might reflect a decreased expectation
of an imminent NOGO trial, though this would not explain
the differential speeding following correct and incorrect
trials. The present results do not provide direct evidence
for anti-correlated endogenous oscillations such as have
been suggested for the ‘‘default network” (Sonuga-Barke
& Castellanos, 2007). Sonuga-Barke and Castellanos do
not explicitly consider the effects of task errors on the
power of endogenous rhythms. It is possible, however, that
the current pattern of results might arise because task
active (exteroceptive) cycles of the default network are
suppressed and/or task negative (introspective) cycles are
actually induced by the occurrence of errors, leading to
the patterns observed in the present study (see also
Smallwood et al., 2008). It is, however, impossible to eval-
uate these alternative interpretations in the absence of
neurological data. Given the temporal parameters of the
110 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
task, imaging techniques such as MEG will likely be re-
quired for future studies.
Near-misses: During SART performance subjects some-
time report perceived incipient movements as muscle ten-
sion in the arm or they may detect small hand or finger
movements too small to be conventionally recorded as
overt responses (i.e., key presses). These are potentially
informative behavioral cues that could be recorded elec-
tromyographically (EMG) and potentially related to
changes in the lateralized readiness potential, such as
those reported for nontarget trials (Smallwood et al.,
2008), and possibly indexing incipient attention-related
errors. According to the model, State 1 of mind-wandering
consists of brief and partial processing of stimulus mean-
ing of the sort likely to lead to near-misses. Thus, being
presented with a NOGO trial while in State 1 of mind-wan-
dering should lead to muscle activity in the arm as if the
subject were about to respond, but with sometimes suc-
cessful inhibition of the overt response (a near miss). In
State 2 of mind-wandering task monitoring is no longer
effective. Accordingly, we predict that being presented
with a NOGO trial while in State 2 will lead to a larger
EMG response relative to State 1 and that this will more of-
ten lead to an overt commission error. Finally, because
State 3 of mind-wandering is characterized by a complete
failure to engage to task response system, we predict that
presenting a NOGO trial during State 3 of mind-wandering
will be associated with no EMG response. In addition,
when detected by subjects in State 1, near miss cues
should also lead to off-line ruminations about performance
in the same way as overt errors. Such ruminations could be
assessed by probes (e.g., Smallwood et al., 2004) designed
to detect awareness of near-misses following speeding,
anticipations, and omissions. Smallwood and colleagues
(2007) also suggest the possibility of using probes sepa-
rated by different intervals to assess the timing and dura-
tion of state transitions as well as the comparison of self-
caught and probe-caught mind-wandering as potential
indices of early and late (e.g., States 1 and 2) mind-
wandering.
GSR: The galvanic skin response (GSR) has been fairly
consistently associated with SART performance (O’Connell
et al., 2007; Smallwood et al., 2004). Also, for example, and
consistent with our finding of behavioral responsiveness to
both successful and unsuccessful NOGO trials, O’Connell
and colleagues reported GSR to both errors and successful
withholds on NOGO trials (O’Connell et al., 2007). Also con-
sistent with the model, GSR was slightly greater to errors
than to successful withholds. Finally, across subjects, SART
error detection rates and GSR were significantly correlated.
Thus, GSR appears to index task-related mind-wandering
following NOGO trials. This suggests that GSR will be dif-
ferentially related to indices of the three mind-wandering
states following both success and error, but especially fol-
lowing error, on NOGO trials.
ERP: Several predictions also follow from the three-state
model regarding brain activity (as measured by event-re-
lated potentials – ERP). Error processing has long been
thought to involve two separate mechanisms, one for error
detection and another for error correction (Rabbitt, 1966).
Error negativity (ERN) appears to be related to implicit
detection (but not necessarily conscious awareness) of er-
ror, whereas late error positivity (Pe) appears to be related
to the (conscious) evaluation of response strategies in re-
sponse to error awareness (Falkenstein, Hohnsbein,
Hoorman, & Blanke, 1990; O’Connell et al., 2007).
Smallwood and colleagues (2008) also report a reduction
in P300 prior to errors and off-task reports. We hypothe-
sized above that the post-error changes in RTs and in-
creases in the probability of anticipations and omissions
likely also reflect a cognitive response to error in the form
of off-line processing (i.e., task-relevant mind-wandering).
Hence, we predict that, whereas ERN will be associated
with errors generally, only late error positivity (Pe) will
be specifically associated with RT change, anticipations,
and omissions that follow an error on NOGO trials.
Memory effects: The available evidence suggests that
mind-wandering also has detrimental effects on recall of
environmental- and task-relevant events (Smallwood
et al., 2006). More specifically, mind-wandering is
associated with a decrease in recall (explicit memory) rel-
ative to familiarity (implicit memory) (Smallwood et al.,
2003, 2004). Recall should decline as mind-wandering
moves from State 1 to 3. We also predict an increasing shift
from explicit to implicit remembering. It is an interesting
possibility that both explicit and implicit recall will be
reduced during state 3.
Special populations: Attentional deficits are a common
feature of debilitating psychiatric and neurological pathol-
ogy (Sonuga-Barke & Castellanos, 2007). One of the central
characteristics of the three-state model is that it can be
used to describe qualitative differences in attentional dis-
engagement across individuals. That is, the model makes
specific predictions regarding how attentional disengage-
ment is affected by various attention-related disorders
such as traumatic brain injury (TBI), attention deficit
hyperactive disorder (ADHD), and anxiety disorders such
as obsessive compulsive disorder (OCD). ADHD individuals
often have much more variable RTs (Klein, Wendling,
Huettner, Ruder, & Peper, 2006) and have been reported
to have, occasionally, extremely long RTs (Hervey et al.,
2006). The interpretation of such findings has been
hampered by the absence of a coherent theory to integrate
such findings (Castellanos et al., 2005). The present model
provides a principled explanation for these findings in
terms of different states of mind-wandering related to
the degree of decoupling of attention from task. Future
studies employing the indices proposed here may well
contribute to a theoretically grounded understanding of
the nature of ADHD attentional deficits.
With regard to TBI, the available evidence suggests that
TBI patients perform very poorly on the SART and, in
addition, show poor error detection (Hart, Giovannetti,
Montgomery, & Schwartz, 1998; McAvinue, O’Keeffe,
McMackin, & Robertson, 2005; Stemmer, Segalowitz, Witz-
ke, & Schonle, 2004). Moreover, GSR in response to NOGO
errors is significantly reduced in TBI participants relative
to controls even with awareness of errors (O’Keeffe et al.,
2004). Given that TBI patients seem less responsive to
attention-related errors, we expect that they will have less
incentive for off-task rumination and should show weaker
behavioral reactions to unsuccessful NOGO trials. That is,
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
TBI patients should be less likely to engage in post-error
ruminations, which should decrease their post-error RT
speeding, anticipations, and omissions relative to normal
controls. Thus, compared to normal controls, the bidirec-
tional relations between mind-wandering and attention-
related errors should be reduced for TBI patients.
In contrast, individuals with anxiety disorders, depres-
sion or OCD might experience exaggerated bidirectional
relations between mind-wandering and attention-related
errors. Anxious individuals will likely be more attentive
to their errors and will tend to ruminate on them following
errors. Farrin and colleagues report that depressed individ-
uals may respond ‘‘catastrophically” to errors producing an
enhanced sense of failure (Farrin, Hull, Unwin, Wykes, &
David, 2003). It is also possible that this is because such
exaggerated reactions to error produce greater rumination
on performance (cf. task-related interference; e.g., Small-
wood, Davies, et al., 2004). This would be especially disaf-
fecting in that individuals would have a subjective sense of
increased task effort associated with increasingly poorer
performance. This should result in greater post-error RT
speeding, anticipations, and omissions in anxious individu-
als compared to normal controls. The associated task-re-
lated mind-wandering then leads to more attention-
related errors, which will lead to more task-related rumi-
nations, ultimately leading to a snowball effect. Similar
predictions can be made for individuals suffering from
sub-clinical levels of performance anxiety.
Real-world tasks: Although we have empirically as-
sessed the three-state model in the context of the SART,
our reference to real-world mind-wandering situations in
the exposition of the model reflects our conviction that
the model applies beyond the SART task to many everyday
continuous performance tasks such as reading and driving.
An important aspect of our model is its ability to make con-
crete predictions about attentional disengagement in these
real-world tasks. For example, in the context of reading,
the model makes specific predictions about readers’ eye
movements and their phenomenology as they progress
through states of task disengagement. In particular, we
predict that entering State 1 disengagement during reading
will be accompanied by sporadic increases in the speed of
eye movements and reduced memory for content. There
may also be more regression saccades (re-reading) as read-
ers catch themselves mind-wandering and attempt to get
back on task. In State 2, readers’ eyes will be moving from
word to word in a very stereotyped fashion reflecting a lack
of processing of textual meaning and, hence, if stopped and
tested at such times, readers will have very little to no
memory of the content. Finally, in State 3, readers will stop
moving their eyes altogether or look away and stare com-
pletely engaged in internal processing.
Similar predictions from the three-state model can be
made with regard to driving performance. In State 1 of
mind-wandering, a driver will be intermittently less
responsive to the focal moment-to-moment demands of
the immediate driving situation but be periodically aware
that at times he/she is fully engaged in driving. In State 2,
drivers’ responses will be more automatic and triggered
only by generic cues present in the environment. In this
state, one might, for example, intend to take a new route
111
only to find oneself automatically following a habitual,
highly familiar route. In State 3, the driver will become
unresponsive to the driving environment. If briefly falling
into State 3 while driving, an individual might completely
fail to brake in response to external cues (e.g., inadvertently
running a red light). Moreover, following from the bidirec-
tionality predictions of the model, when the driver commits
an error (e.g., begins to drive off of the road); awareness of
the error might lead to more mind-wandering in the form of
rumination about inattention and driving performance
potentially generating another bout of mind-wandering.
7. Concluding comments
Our major goals have been to present a model of atten-
tional engagement/disengagement consistent with exist-
ing data on mind-wandering and sustained attention, to
present new data to assess directly a number of predictions
from the model, and to suggest novel predictions to guide
future research. As noted in our introductory remarks indi-
ces employed in the present study were selected based on
our task analysis of the SART and that detailed task analy-
ses will be required for the application of the model to dif-
ferent tasks and contexts. At an applied level, the three-
state model of attentional disengagement also points to
possible means to prevent, or at least limit, the progression
of attentional disengagement. Given signatures for each
state, for example, we could use these as triggers to pro-
vide extrinsic cues to alert subjects to the onset of mind-
wandering during attentionally demanding tasks. Previous
research has reported that providing even non-contingent,
non-predictive, content-free ‘‘cues-to-attend” (Fassbender
et al., 2006) reduces SART errors in both healthy subjects
and stroke patients (Manly et al., 2004). Based on the pres-
ent model, similar extrinsic cues provided contingently on
RT changes, anticipations, and omissions should be even
more effective at improving performance of subjects, and
especially individuals with marked deficits in sustained
attention (e.g., ADD and TBI). Indeed, it is possible that cue-
ing based on either decreasing or increasing RTs, anticipa-
tions, and omissions might selectively reduce inattentional
states as well as errors. Use of fading techniques in which
cues become more subtle and even subliminal might effec-
tively ‘‘wean” people from the cues as they become sensi-
tized to the onset of mind-wandering states, and be less
likely to engage in reactive mind-wandering. It would also
be of interest to consider potentially more general effects
of such training on an individual’s sense of self efficacy
and self control, as well as possible changes in affect. Fu-
ture investigations of these possibilities should contribute
to the further development of the current model of atten-
tion disengagement with the ultimate goal of developing
techniques to reduce the debilitating effects and conse-
quences of failures of sustained attention and attention-re-
lated errors in the real world during attentionally
demanding tasks.
References
Antrobus, J. S. (1968). Information theory and stimulus-independent
thought. British Journal of Psychology, 59, 423–430.
112 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
Bellgrove, M. A., Hawi, Z., Gill, M., & Robertson, I. H. (2006). The cognitive
genetics of attention deficit hyperactivity disorder (ADHD): Sustained
attention as a candidate phenotype. Cortex, 42, 838–845.
Bellgrove, M. A., Hawi, Z., Kirley, A., Gill, M., & Robertson, I. H. (2005).
Dissecting the attention deficit hyperactivity disorder (ADHD)
phenotype: Sustained attention, response variability and spatial
attentional asymmetries in relation to dopamine transporter (DAT1)
genotype. Neuropsychologia, 43, 1847–1982.
Bellgrove, M. A., Hester, R., & Garavan, H. (2004). The functional
neuroanatomical correlates of response variability: Evidence from a
response inhibition task. Neuropsychologia, 42, 1910–1916.
Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D.
(2001). Conflict monitoring and cognitive control. Psychological
Review, 108, 624–652.
Brand, M. (1984). Intending and acting. Cambridge, MA: MIT Press.
Bratman, M. E. (1987). Intention, plans, and practical reason. Cambridge,
MA: Cambridge University Press.
Braver, T. S., Barch, D. M., Gray, J. R., Molfese, D. L., & Snyder, A. (2001).
Anterior cingulate cortex and response conflict: Effects of frequency,
inhibition, and errors. Cerebral Cortex, 11, 825–836.
Braver, T. S., Reynolds, J. R., & Donaldson, D. I. (2003). Neural mechanisms
of transient and sustained cognitive control during task switching.
Neuron, 39, 713–726.
Burgund, E. D., Lugar, H. M., Miezin, F. M., & Peterson, S. E. (2003).
Sustained and transient activity during an object-naming task: A
mixed blocked and event-related fMRI study. NeuroImage, 19, 29–41.
Carriere, J. S. A., Cheyne, J. A., & Smilek, D. (2008). Everyday attention
lapses and memory failures: The affective consequences of
mindlessness. Consciousness and Cognition, 17, 835–847.
Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J.
D. (1998). Anterior cingulate cortex, error detection, and the online
monitoring of performance. Science, 280, 747–749.
Castellanos, F. X., Sonuga-Barke, E. J. S., Scheres, A., Di Martino, A., Hyde,
C., & Waters, J. R. (2005). Varieties of attention-deficit/hyperactivity
disorder-related intra-individual variability. Biological Psychiatry, 57,
1416–1423.
Cheyne, J. A., Carriere, J. S. A., & Smilek, D. (2006). Absent-mindedness:
Lapses in conscious awareness and everyday cognitive failures.
Consciousness and Cognition, 15, 578–592.
Cohen, J. D., Botvinick, M. M., & Carter, C. S. (2000). Anterior cingulate and
prefrontal cortex: Who’s in control? Nature Neuroscience, 3, 421–423.
Dockree, P. M., Kelly, S. P., Foxe, J. J., Reilly, R. B., & Robertson, I. H. (2007).
Optimal sustained attention is linked to the spectral content of
background EEG activity: Greater ongoing tonic alpha (10 Hz) power
supports successful phasic goal activity. European Journal of
Neuroscience, 25, 900–907.
Dockree, P. M., Kelly, S. P., Robertson, I. H., Reilly, R. B., & Foxe, J. J. (2005).
Neurophysiological markers of alert responding during goal-directed
behavior: A high-density electrical mapping study. NeuroImage, 27,
587–601.
Dockree, P. M., Kelly, S. P., Roche, R. A., Hogan, M. J., Reilly, R. B., &
Robertson, I. H. (2004). Behavioural and physiological impairments of
sustained attention after traumatic brain injury. Brain Research
Cognitive Brain Research, 20, 403–414.
Donaldson, D. I., Peterson, S. E., Ollinger, J. M., & Buckner, R. L. (2001).
Dissociating state and item components of recognition memory using
fMRI. NeuroImage, 13, 129–142.
Douse, T. L. (1900). On some minor psychological interferences: A study of
misspellings and related mistakes. Mind: New Series, 9, 85–93.
Falkenstein, M., Hohnsbein, J., Hoorman, J., & Blanke, L. (1990). Effects of
errors in choice reaction tasks on the ERP under focused and divided
attention. In C. H. M. Brunia, A. W. K. Gaillard, & A. Kok (Eds.),
Psychophysiological brain research. Tilburg: Tilburg University Press.
Farrin, L., Hull, L., Unwin, C., Wykes, T., & David, A. (2003). Effects of
depressed mood on objective and subjective measures of attention.
Journal of Neuropsychiatry and Clinical Neurosciences, 15, 98–104.
Fassbender, C., Simoes-Franklin, C., Murphy, K., Hester, R., Meaney, J.,
Robertson, I. H., et al (2006). The role of a right fronto-parietal
network in cognitive control: Common activations for ‘‘cues-to-
attend” and response inhibition. Journal of Psychophysiology, 20,
286–296.
Garavan, H., Hester, R., Murphy, K., Fassbender, C., & Kelley, C. (2006).
Individual differences in the functional neuroanatomy of inhibitory
control. Brain Research, 1105, 130–142.
Giambra, L. M. (1995). A laboratory method for investigating influences
on switching attention to task-unrelated imagery and thought.
Consciousness and Cognition, 4, 1–21.
Gilden, D. L., & Hancock, H. (2007). Response variability in attention
deficit disorders. Psychological Science, 18, 796–802.
Hart, T., Giovannetti, M. S., Montgomery, M. W., & Schwartz, M. F. C.
(1998). Awareness of errors in naturalistic action after traumatic
brain injury. Journal of Head Trauma, 13, 16–28.
Hervey, A. S., Epstein, J. N., Curry, J. F., Tonev, S., Arnold, L. E., Conners, C.
K., et al (2006). Reaction time distribution analysis of
neuropsychological performance in an ADHD sample. Child
Neuropsychiatry, 12, 125–140.
Hester, R., Fassbender, C., & Garavan, H. (2004). Individual differences in
error processing: A review and reanalysis of three event-related fMRI
studies using the GO/NOGO task. Cerebral Cortex, 14, 986–994.
Hester, R., Foxe, J. J., Molholm, S., Shpaner, M., & Garavan, H. (2005).
Neural mechanisms involved in error processing: A comparison of
errors made with and without awareness. NeuroImage, 27, 602–608.
Johnson, K. A., Kelly, S. P., Bellgrove, M. A., Barry, E., Cox, M., Gill, M., et al
(2007). Response variability in attention deficit hyperactivity
disorder: Evidence for neuropsychological heterogeneity.
Neuropsychologia, 45, 630–638.
Johnson, K. A., Robertson, I. H., Kelly, S. P., Silk, T. J., Barry, E., Dáibhis, A.,
et al (2007). Dissociation of performance of children with ADHD and
high-functioning autism on a task of sustained attention.
Neuropsychologia, 45, 2234–2245.
Kane, M. J., Brown, L. H., McVay, J. C., Silvia, P. J., Myin-Germeys, I., &
Kwapil, T. R. (2007). For whom the mind wanders, and when: An
experience sampling study of working memory and executive control
in daily life. Psychological Science, 18, 614–621.
Kaufman, J., Ross, T. J., Stein, E. J., & Garavan, H. (2003). Cingulate
hypoactivity in cocaine users during a GO/NOGO task as revealed by
event-related fMRI. Journal of Neuroscience, 23, 7839–7843.
Klein, C., Wendling, K., Huettner, P., Ruder, H., & Peper, M. (2006). Intra-
subject variability in attention deficit hyperactivity disorder.
Biological Psychiatry, 60, 1088–1097.
Kvavilashvili, L., & Mandler, G. (2004). Out of one’s mind: A study of
involuntary semantic memories. Cognitive Psychology, 48, 47–94.
MacDonald, A. W., Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000).
Dissociating the role of the dorsolateral prefrontal and anterior
cingulate cortex in cognitive control. Science, 288, 1835–1838.
Mackworth, N. H. (1948). The breakdown of vigilance during
prolonged visual search. Quarterly Journal of Experimental Research,
1, 6–21.
Manly, T., Anderson, V., Nimmo-Smith, I., Turner, A., Watson, P., &
Robertson, I. H. (2001). The differential assessment of children’s
attention: The test of everyday attention for children TEA-CH)
normative sample and ADHD attention. Journal of Child Psychology
and Psychiatry, 42, 1065–1081.
Manly, T., Davidson, B., Gaynord, B., Greenfield, E., Heutniki, J., & Parr, A.
(2004). An electronic knot in the handkerchief: ‘Content free cueing’
and the maintenance of attentive control. Neuropsychological
Rehabilitation, 14, 89–116.
Manly, T., Lewis, G. H., Robertson, I. H., Watson, P. C., & Datta, A. K. (2002).
Coffee in the cornflakes: Time-of-day as a modulator of executive
response control. Neuropsychologia, 40, 1–6.
Manly, T., Owen, A. M., McAvinue, L., Datta, A., Lewis, G. H., Scott, S. K.,
et al (2003). Enhancing the sensitivity of a sustained attention task to
frontal damage: Convergent clinical and functional imaging evidence.
Neurocase, 9, 340–349.
Manly, T., Robertson, I. H., Galloway, M., & Hawkins, K. (1999). The absent
mind: Further investigations of sustained attention to response.
Neuropsychologia, 37, 661–670.
McAvinue, L., O’Keeffe, F., McMackin, D., & Robertson, I. H. (2005).
Impaired sustained attention and error awareness in traumatic brain
injury: Implications for insight. Neuropsychological Rehabilitation, 15,
569–587.
Mele, A. R. (1992). Springs of action. Oxford: Oxford University Press.
Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal
cortex function. Annual Review of Neuroscience, 24, 167–202.
Mitchell, J. P., Heatherton, T. F., Kelley, W. M., Wyland, C. L., Wegner, D. M.,
& Macrae, C. N. (2007). Separating sustained from transient aspects of
cognitive control during thought suppression. Psychological Science,
18, 292–297.
Mullins, C., Bellgrove, M. A., Gill, M., & Robertson, I. H. (2005). Variability
in time reproduction: Difference in ADHD combined and inattentive
subtypes. Journal of the American Academy of Child and Adolescent
Psychiatry, 44, 169–176.
Norman, D. A. (1981). Categorization of action slips. Psychological Review,
88, 1–15.
O’Connell, R. G., Bellgrove, M. A., Dockree, P. M., & Robertson, I. H. (2004).
Reduced electrodermal response to errors predicts poor sustained
attention performance in attention deficit hyperactivity disorder.
Neuroreport, 15, 2535–2538.
 J. Allan Cheyne et al. / Cognition 111 (2009) 98–113
O’Connell, R. G., Dockree, P. H., Bellgrove, M. A., Kelly, S. P., Hester, R.,
Garavan, H., et al (2007). The role of cingulate cortex in the detection
of errors with and without awareness: A high density electrical
mapping study. European Journal of Neuroscience, 25, 2571–2579.
O’Keeffe, F. M., Dockree, P. M., & Robertson, I. H. (2004). Poor insight in
traumatic brain injury mediated by impaired error processing?
Evidence from electrodermal activity. Cognitive Brain Research, 22,
101–112.
Pacherie, E. (2008). The phenomenology of action: A conceptual
framework. Cognition, 107(1), 179–217.
Rabbitt, P. M. A. (1966). Errors and error correction in choice–response
tasks. Journal of Experimental Psychology, 71, 264–272.
Rasmussen, J. (1980). What can be learned from human error reports? In
K. Duncan, K. Gruneberg, & D. Wallis (Eds.), Changes in working life.
London, UK: Wiley.
Reason, J. T. (1979). Actions not as planned: The price of automatization.
In G. Underwood & R. Stevens (Eds.), Aspects of consciousness: Vol. 1.
Psychological issues. London, UK: Wiley.
Reason, J. T. (1984). Lapses of attention in everyday life. In R. Parasuraman
& R. Davies (Eds.), Varieties of attention. New York: Academic Press.
Reason, J. T. (1990). Human error. Cambridge, UK: Cambridge University
Press.
Reason, J. T. (1998). How necessary steps in a task get omitted: Reviving
old ideas to combat a persistent problem. International Journal of
Cognitive Technology, 3, 24–32.
Reason, J., & Lucas, D. (1984). Absent-mindedness in shops: Its incidence,
correlates and consequences. British Journal of Clinical Psychology, 23,
121–131.
Reason, J. T., & Mycielska, K. (1982). Absent-minded? The psychology of
mental lapses and everyday errors. Englewood Cliffs, NJ: Prentice Hall.
Robertson, I. H., Manly, T., Andrade, J., Baddeley, B. T., & Yiend, J. (1997).
‘Oops!:’ Performance correlates of everyday attentional failures in
traumatic brain injured and normal subjects. Neuropsychologia, 35(6),
747–758.
Sarason, I. G., Sarason, B. R., Keefe, D. E., Hates, B. E., & Shearin, E. N.
(1986). Cognitive interference: Situational determinants and trait-like
characteristics. Journal of Personality and Social Psychology, 51,
215–226.
Schooler, J. W. (2002). Re-representing consciousness: Dissociations
between experience and meta-consciousness. Trends in Cognitive
Science, 6, 339–344.
Schooler, J. W., Reichle, E. D., & Halpern, D. V. (2005). Zoning-out during
reading: Evidence for dissociations between experience and
metacognition. In D. T. Levin (Ed.), Thinking and seeing: Visual
metacognition in adults and children. Cambridge, MA: MIT Press.
Searle, J. (1983). Intentionality. Cambridge: Cambridge University Press.
Singer, J. L. (1966). Daydreaming. New York: Random House.
Singer, J. L. (1978). Experimental studies of daydreaming and the stream
of thought. In K. S. Pope & J. L. Singer (Eds.), The stream of
consciousness. New York: Plenum.
Smallwood, J. M., Baracaia, S. F., Lowe, M., & Obonsawin, M. (2003). Task
unrelated thought whilst encoding information. Consciousness and
Cognition, 12, 452–484.
Smallwood, J. M., Davies, J. B., Heim, D., Finnigan, F., Sudberry, M.,
O’Connor, R., et al (2004). Subjective experience and the attentional
lapse: Task engagement and disengagement during sustained
attention. Consciousness and Cognition, 13, 657–690.
113
Smallwood, J. M., Heim, D., Rigby, L., & Davies, J. B. (2006). Encoding
during the attentional lapse: Accuracy of encoding during the
semantic SART. Consciousness and Cognition, 15, 218–231.
Smallwood, J. M., McSpadden, M., Luus, B., & Schooler, J. W. (2008).
Segmenting the stream of consciousness: The psychological
correlates of temporal structures in the time series data of a
continuous performance task. Brain and Cognition, 66, 50–56.
Smallwood, J. M., McSpadden, M., & Schooler, J. W. (2007). The lights are
on but no one’s home: Meta-awareness and the decoupling of
attention when the mind wanders. Psychonomic Bulletin and Review,
14, 527–533.
Smallwood, J., Obonsawin, M. C., & Heim, S. D. (2003). Task-unrelated
thought: The role of distributed processing. Consciousness and
Cognition, 12, 169–189.
Smallwood, J., Obonsawin, M. C., & Reid, H. (2003). The effects of block
duration and task demands on the experience of task unrelated
thought. Imagination, Cognition and Personality, 22, 13–31.
Smallwood, J. M., O’Connor, R. C., Sudberry, M. V., Haskell, C., & Ballantine,
C. (2004). The consequences of encoding information on the
maintenance of internally generated images and thoughts: The role
of meaning complexes. Consciousness and Cognition, 13, 789–820.
Smallwood, J. M., O’Connor, R. C., Sudberry, M. V., & Obosawin, M. (2007).
Mind-wandering and dysphoria. Cognition and Emotion, 21, 816–842.
Smallwood, J. M., & Schooler, J. W. (2006). The restless mind. Psychological
Bulletin, 132, 946–958.
Sonuga-Barke, E. J. S., & Castellanos, F. X. (2007). Spontaneous attentional
fluctuations in impaired states and pathological conditions: A
neurobiological hypothesis. Neuroscience and Biobehavioral Reviews,
31, 977–986.
Stemmer, B., Segalowitz, S. J., Witzke, W., & Schonle, P. W. (2004). Error
detection in patients with lesions to the medial prefrontal cortex: An
ERP study. Neuropsychologia, 42, 118–130.
Teasdale, J. D., Lloyd, C. A., Proctor, L., & Baddeley, A. (1993). Working
memory and stimulus independent thought: Effects of memory load
and presentation rate. European Journal of Psychology, 5, 417–
433.
Teasdale, J. D., Segal, Z. V., & Williams, J. M. G. (1995). How does cognitive
therapy prevent depressive relapse and why should attentional
control (mindfulness) training help? An information processing
analysis. Behaviour Research and Therapy, 33, 25–39.
Velanova, K., Jacoby, L. L., Wheeler, M. E., McAvoy, M. P., Petersen, S. E., &
Buckner, R. L. (2003). Functional–anatomic correlates of sustained
and transient processing components engaged during controlled
retrieval. Journal of Neuroscience, 23, 8460–8470.
Vincente, K. (2004). The human factor. New York: Routledge.
Visscher, K. M., Miezin, F. M., Kelly, J. E., Buckner, R. L., Donaldson, D. I.,
McAvoy, M. P., et al (2003). Mixed blocked/event-related designs
separate transient and sustained activity in fMRI. NeuroImage, 19,
1694–1708.
Wegner, D. M. (1997). Why the mind wanders. In J. D. Cohen & J. W.
Schooler (Eds.), Scientific approaches to consciousness. Mahwah, NJ:
Erlbaum.
Whyte, J., Grieb-Neff, P., Gantz, C., & Polansky, M. (2007). Measuring
sustained attention after brain injury: Differences in key findings
from the sustained attention to response task (SART).
Neuropsychologia, 44, 2007–2014.