Differential contribution of task conflicts to task switch cost and task mixing cost in alternating...
Transcript of Differential contribution of task conflicts to task switch cost and task mixing cost in alternating...
ORIGINAL PAPER
Differential contribution of task conflicts to task switch costand task mixing cost in alternating runs and cued task-switching:evidence from ex-Gaussian modeling of reaction time distributions
Nitzan Shahar • Nachshon Meiran
Received: 15 December 2013 / Accepted: 8 April 2014
� Springer-Verlag Berlin Heidelberg 2014
Abstract Task switching involves switch cost (poorer
performance switch trials than in task-repetition trials) and
mixing cost (poorer performance in task-repetition trials
than in trials from blocks without task switching). These are
mainly studied with the alternating runs task-switching
(ARTS) paradigm (in which the task changes every constant
number of trials) or the cued task-switching (CTS) para-
digm, in which the tasks change randomly. The authors
tested the hypothesis that dealing with actual or potential
conflicts regarding which task is currently required con-
tribute to mixing cost in CTS more than in ARTS and
contribute to switch costs more in ARTS than in CTS. This
hypothesis was tested using ex-Gaussian modeling of reac-
tion time (RT) distributions, in which the heaviness of the
right tail marks task conflicts (Steinhauser and Hubner in J
Exp Psychol Human Percept Perform 35:1398–1412 2009).
As predicted, a heavier RT-distribution tail marked switch
cost more strongly in ARTS than in CTS and marked mixing
costs more strongly in CTS than in ARTS. These results
indicate that switch cost and mixing cost reflect different
processes in different task-switching paradigms.
Introduction
Task switching (TS) has become the method-of-choice to
study the ability to flexibly change mental sets (Kiesel
et al., 2010; Meiran, 2010; Monsell, 2003; Van-
dierendonck, Liefooghe, & Verbruggen, 2010). Most of the
task-switching studies employ one of two paradigms. In the
cued task-switching (CTS) paradigm, the tasks are ordered
randomly, and a cue appearing in the beginning of the trial
instructs which task to execute (Shaffer, 1965; Meiran,
1996). In the alternating runs task-switching (ARTS) par-
adigm, the tasks are ordered in series of trials of fixed
length (called ‘‘runs’’) and the tasks alternate between runs
(e.g., with Tasks A, B and run length = 2, the trial
sequence would be AA-BB-AA…, Rogers & Monsell,
1995).
Regardless of the paradigm being used, switching often
(although not always, e.g., Jersild, 1927) involves a cost.
This cost is often broken down into two components
(Fagot, 1994; Meiran, Chorev, & Sapir, 2000). Switching
cost is the performance decrement in task switch trials
relative to task-repetition trials (both taking place in blocks
involving task switching), presumably reflects the influence
of immediate task transition. The other component, mixing
cost is based on comparing two types of task-repetition
trials. It indexes the performance decrement seen in task-
repetition trials coming from blocks involving task
switching as compared to task-repetition trials coming
from blocks without task switching (‘‘single-task’’ blocks).
Mixing cost therefore reflects the contextual influence on
performance, and is therefore taken to reflect more sus-
tained forms of control (Braver, Reynolds, & Donaldson,
2003).
Several attempts were made to offer a much more
detailed process account of the costs. Regarding mixing
cost, one explanation is that it reflects the higher working
memory (WM) load in blocks involving task switching
(especially Rogers & Monsell, 1995). Despite its intuitive
appeal, this account is incompatible with several
N. Shahar � N. Meiran (&)
Department of Psychology and Zlotowski Center for
Neuroscience, Ben-Gurion University of the Negev,
Be’er Sheva, Israel
e-mail: [email protected]
N. Shahar
e-mail: [email protected]
123
Psychological Research
DOI 10.1007/s00426-014-0569-1
observations. One comes from Experiment 2 in Rubin and
Meiran’s (2005) paper. These authors used a CTS para-
digm and required participants to switch between four
tasks. Two of the four tasks, the color task and the shape
task, were executed on colored shapes. The two other tasks,
vertical (up vs. down) and horizontal (right vs. left) were
executed on a stimulus presented inside a 2 9 2 grid. Thus,
the objective WM load involved four task rules. None-
theless, performance in the 4-tasks condition was similar to
that seen when participants switched between two tasks:
the color task and the shape task (or between the vertical
task and the horizontal task). Hence, the results suggest that
WM load (4 tasks vs. 2 tasks) did not contribute to mixing
cost. Similar conclusions regarding the lack of influence of
WM load on behavior in the CTS paradigm were reached
by other investigators (see also Kessler & Meiran, 2010;
Kiesel, Wendt, & Peters, 2007; van ‘t Wout, Lavric, &
Monsell, 2013).
Another account of mixing cost refers to task uncertainty.
Namely, when task switching takes place, the probability
that a given task will be required is lower (say 50 %) than
when the task is executed in isolation (100 %). Nonetheless,
the aforementioned results of Rubin and Meiran (2005) seem
to also rule out task uncertainty as an explanation since the
likelihood of each task to appear was 50 % in the 2-task
condition as compared to 25 % when all four tasks were
intermixed. Thus, despite the large uncertainty difference,
performance remained roughly unchanged.
Of greatest relevance in the present work is Koch, Prinz,
& Allport, (2005) as well as Rubin and Meiran’s (2005; cf.
Steinhauser & Hubner, 2009) suggestion that task mixing
costs involve task conflict. The term ‘‘task conflict’’ has
been distinguished from ‘‘response conflict’’ (or ‘‘infor-
mation conflict’’, e.g., Goldfarb & Henik, 2007). It refers to
a conflict between abstract task rules (each rule referring to
more than one response, such as the rule ‘‘respond to color
by pressing the right key in response to green color and
pressing the left key in response to red color’’). Basically,
Koch et al.’s and Rubin and Meiran’s findings indicate
greater mixing costs when the stimuli cued the competing
task and the account is therefore in terms of the resultant
conflict between task rules. In other words, when two (or
more) tasks are executed on the same stimuli, the stimulus
leads to the retrieval of the wrong task rule and generates a
conflict regarding which task rule is currently required
(Waszak, Hommel, & Allport, 2003; cf. Gilbert & Shallice,
2002). Notably, mixing costs are observed even with uni-
valent trials that afford only one task, as long as bivalent
trials (affording both tasks) are also included (Woodward,
Meier, Tipper, & Graf, 2003) suggesting that mixing costs
may also reflect dealing with the potential task conflict, not
only with an actual conflict.
Nonetheless, Koch et al. (2005) and Rubin and Meiran’s
hypothesis is based on results coming from the CTS par-
adigm, and the causes for mixing cost may depend on the
paradigm, especially given the known processing differ-
ences between the ARTS and CTS paradigms (Altmann,
2007, 2013; Andreadis & Quinlan, 2010; Koch, 2003,
2005; Monsell, Sumner, & Waters, 2003; Pereg, Shahar, &
Meiran, 2013; Schmitz & Voss, 2012; Tornay & Milan
2001). More specifically, Altmann (2007) referred to a
process in which participants activate the abstract task
representation. He suggested that, given the task uncer-
tainty that characterizes the CTS paradigm, participants
employ this process (activate the abstract task representa-
tion) in each and every trial. However, in the ARTS par-
adigm, participants make use of task predictability. What
they appear to do is to activate the abstract task represen-
tation only in the first trial of the run, and then skip this
process in the remaining trials of the run (see also Meiran,
Kessler, & Adi-Japha, 2008). So, for example, if the run
length is 4, participants know that once the run has begun,
the task will remain unchanged in the next three trials.
Therefore, they activate the abstract task representation
only in the first trial in the run (which is the switch trial)
and do not activate it in the second through fourth trial of
the run (which are task-repetition trials).
This strategy which differentiates ARTS from CTS has
specific implications to the processes which are responsible
for switch costs and mixing costs in the two paradigms.
Because switch costs are computed as the difference
between task switch trials (which is the first task in the run
in ARTS) and task-repetition trials (subsequent trials in the
run), switch cost in the ARTS paradigm is influenced by
the need to activate task representation (present only in
switch trials, and absent in task-repetition trials, hence
contributing to their difference). In contrast, in CTS, par-
ticipants activate the abstract task representation in each
and every trial. Consequently, both switch trials and task-
repetition trials involve the process of task-representation
activation and the difference between them does not reflect
this process. For that reason, this process of activating
abstract task representation does not contribute (as much)
to switch costs in the CTS paradigm.
The opposite applies to mixing costs, defined as the
difference between task-repetition trials and single-task
trials. Task-repetition trials in CTS involve the activation
of task representation. In single-task trials this is not the
case, given the perfect task certainty in these blocks. Thus,
mixing cost in CTS reflects this process (see Altmann &
Gray, 2008). In contrast, in ARTS, task-repetition trials
like single-task trials do not involve the activation of task
representation. Thus, mixing costs in ARTS do not reflect
this process (as much).
Psychological Research
123
Altmann’s (2007) hypothesis can be combined with
Koch et al.’s (2005) and Rubin and Meiran’s (2005)
hypothesis as follows. On trials in which task representa-
tion is activated, greater interference from competing task
representations (task conflict) is expected, and/or greater
effort associated with potential task conflicts is expected.
Thus, the two hypotheses, combined, predict that processes
dealing with actual or potential task conflicts will differ-
entially influence mixing cost and switch cost in CTS and
ARTS, depending on which cost-component mostly
reflects the activation of task representation. Accordingly,
in CTS, mixing cost is predicted to demonstrate larger task
conflict-related processing; whereas, in ARTS task con-
flict-related processing is predicted to primarily influence
switch cost. One line of evidence supporting the afore-
mentioned hypothesis comes from a study comparing
bivalent trials (affording both tasks) with univalent trials
(affording just one task). Usually, performance is poorer in
bivalent trials than in univalent trials (e.g., see Rogers &
Monsell’s, 1995, ‘‘task cuing effect’’, cf. Woodward et al.
2003, for a related finding), a finding that is interpreted as
evidence for a conflict taking place at the level of the
abstract task representations (e.g., Goldfarb & Henik,
2007). Accordingly, Andreadis and Quinlan (2010) found
greater involvement of bivalent vs. univalent costs in task-
repetition trials in conditions in which the task was
unpredictable (CTS) than in conditions when the task was
predictable (ARTS).
Another very relevant study has been reported by Ste-
inhauser and Hubner (2009), who used a CTS paradigm in
which there were bivalent and univalent trials and mixing
costs were also assessed. Importantly, the authors modeled
the shape of the RT distributions with the ex-Gaussian
distribution (e.g., Ratcliff, 1993). This distribution is a
combination of two distributions: Gaussian, described by
the parameters l (mu) and r (sigma), and exponential,
described by the parameter s (tau). In this distribution, the sparameter is mostly responsible for the typical heavy right
tail, so that when s = 0, the distribution no longer has such
a tail and becomes Gaussian, and when s increases, so does
the heaviness of the right tail. Two important additional
characteristics of this distribution are: (1) that
RTmean = l ? s, and (2) RTvar = r2 ? s2.
Steinhauser and Hubner’s (2009) results show (a) that
the difference between bivalent and univalent trials is
mostly seen in s; and, (b) that the same holds true for
mixing costs. This has led the authors to conclude that
increased s characterizes task conflicts. We therefore have
decided to use this index to test our hypothesis concerning
differential contribution of task conflicts to mixing costs
and switch costs in CTS vs. ARTS.
Accordingly, the present work was based on ex-Gauss-
ian analysis of RT distributions taken from an experiment
in which the participants performed in comparable ARTS
and CTS paradigms. We predicted that, for switch costs, swill be influential mostly in ARTS whereas for mixing
costs, s will be mostly influential in CTS.
Method
Participants
A total of 16 healthy undergraduate students
(Mage = 23.25, SD = 1.81, 10 females, 6 males) took part
in the experiment in return for 25 NIS per hour (*6$).
Participants were pre-screened based on self-reported for
not previously suffering from head injury, psychiatric dis-
orders, drug/alcohol use, color blindness or diag-
nosed learning disabilities.
Stimuli and apparatus
Four stimuli were used as targets including a circle or a
triangle, either in red or green printed with a 2-mm outline.
Task cues in the CTS conditions included the Hebrew word
‘‘ ’’ or ‘‘ ’’ (i.e., ‘‘shape’’ or ‘‘color’’). Each shape
stimulus was 40 9 40 mm in size. Text was presented in a
white 18-point Courier New font in the center of the
screen. The experiment was programmed using E-Prime
2.0 (Psychology Software Tools, Pittsburgh, PA, USA). All
stimuli were presented in the center of a black 19-in.
computer screen.
Procedure
Participants were seated in front of the computer screen in
a small room in the lab. After a short instruction screen,
participants performed four trials of each condition
including the shape as a single task, color as a single task,
and two mixed blocks (i.e., either ARTS followed by CTS
or vice versa, counterbalanced between participants). The
test phase included the same sequence of conditions,
repeated twice. Each single-task block in the test phase
included two mini-blocks of 24 trials, and each mixed
block included four mini-blocks of 24 trials, with a short
recess after each mini-block. Thus, each participant per-
formed in the test phase 96 trials of each single task (i.e.,
shape or color) and 192 trials for each mixed block (i.e.,
CTS or ARTS).
In shape or color single-task blocks, the participants
were asked to perform the same task without switching
until the end of the block. In CTS blocks, the participants
were asked to perform either the shape task or the color
task according to a cue presented before the target. In
ARTS blocks, the participants were asked to switch tasks
Psychological Research
123
after every second trial, beginning with the shape task (i.e.,
shape, shape, color, color, etc.).
In shape trials, participants were asked to decide whe-
ther the shape is a circle or a triangle. In the color trials
participants were asked to report whether the color of the
outline of the shape is green or red. In the single-task
blocks and ARTS mixed blocks, each trial included a fix-
ation screen (1,000 ms) followed by the target that was
presented until the response was given or until 6 s had
elapsed. In CTS blocks, each trial included a fixation
screen (500 ms), a randomly chosen cue screen (switch
probability = 50 %, as in ARTS; 500 ms), followed by the
target. Participants reacted by pressing the ‘‘s’’ (left) or the
‘‘k’’ (right) key. A 400-ms beep signaled errors.
Results
For RT analyses, error and post-error trials, ten first trials in
each condition and the first trial after each recess were
discarded. RTs below 200 ms or above 4 SDs from the
participant’s mean in the respective condition were con-
sidered as outliers and thus omitted. We re-calculated the
RT mean in each condition and repeated trimming until all
outliers were removed (resulting in 0–4 trials omitted in
each conditions; Schmiedek, Oberauer, Wilhelm, Suss, &
Wittmann, 2007). The ex-Gaussian distribution fitting was
performed using the DISTRIB toolbox in MATLABTM
(Lacouture & Cousineau, 2008). In order to ensure that the
ex-Gaussian distribution fits the data, we simulated 10,000
data points from each set of the ex-Gaussian parameters in
each condition. We then calculated four quantile means
(e.g., 0.20 0.40 0.60 0.80) for both the simulated and the
empirical data. A quantile–quantile plot was generated to
allow a visual comparison between the simulated and
empirical data (e.g., Steinhauser & Hubner, 2009). A good
fit between the theoretical ex-Gaussian distributions and
the empirical data was demonstrated (see Fig. 1).
Summary of the means in all conditions is presented in
Table 1. Because the same single-task condition served as
baseline for both ARTS and CTS, the conditions did not
form a factorial design. We therefore conducted the anal-
yses on costs by calculating switch cost (i.e., subtracting
switch from repeat trials) and mixing cost (i.e., subtracting
repeat trials from single-task trials). A series of repeated
measures analyses of variance (ANOVAs) were conducted
with Cost-Type (switch cost vs. mixing cost), Paradigm
(CTS vs. ARTS) and Task (Shape vs. Color) as indepen-
dent variables (see Fig. 2). The first ANOVA was per-
formed on the mean RTs. It revealed a significant
interaction between cost-type and block-type
[F(1,15) = 17.35, p \ 0.001., g2p = 0.54] showing a larger
mixing cost in the CTS than in ARTS and a larger
switching cost for ARTS than for CTS (see Fig. 2).
Similar ANOVAs were performed on each one of the three
ex-Gaussian parameters (l, r, s, see Fig. 2). A significant
main effects for Cost-Type was found in l [F(1, 15) = 5.80,
p \ 0.05, g2p = 0.29] and r [F(1,15) = 5.16, p \ 0.05,
g2p = 0.25], but not in s [F(1,15) = 0.98, n.s., g2
p = 0.06].
Most importantly, there was a significant interaction between
Paradigm and Cost-Type only for s [F(1, 15) = 16.50,
p \ 0.001, g2p = 0.52].
The significant interaction found for s was probed by
planned contrasts comparing between paradigms. The dif-
ference between paradigms was significant for mixing cost
[t(15) = 2.78, p \ 0.01, g2p = 0.34] and switch cost
[t(15) = 3.48, p \ 0.001, g2p = 0.44].
A similar ANOVA on accuracy demonstrated a signifi-
cant main effect for Paradigm [F(1,15) = 8.66, n.s.,
g2p = 0.37], Cost-Type [F(1, 15) = 29.54, p \ 0.001,
g2p = 0.66], and Task. The latter effect indicates lower,
negative costs for the shape task as compared with positive
costs in the color task (see Fig. 2).
Task-rule congruency effect (TRCE)
TRCE results from the competition between response
tendencies generated by the required (relevant) task rule
and the irrelevant task rule (see Meiran & Kessler, 2008).
For example, the task rule in the current trial may be
Fig. 1 To obtain an estimation of how well the ex-Gaussian
distribution fits the data, we first simulated a set of 10,000 data
points using the previously extracted ex-Gaussian parameters. We
then calculated the mean RT for four bins (0.2 0.4 0.6 0.8), once for
the empirical and once for simulated data sets. A quantile–quantile
plot was created. Each data point describes the mean RT for the
relevant bin in the empirical data (i.e., x axis) and simulated data (i.e.,
y axis). A poor fit between the data and the ex-Gaussian should be
observed by a discrepancy from the diagonal line. As can be seen, an
excellent fit was observed
Psychological Research
123
SHAPE and this rule may indicate the right key as the
correct response. The currently irrelevant rule (COLOR)
may indicate the same key as the correct response (in
congruent trials) or the opposite key (in incongruent trials).
The performance difference between congruent and
incongruent trials (the TRCE) thus reflects the aforemen-
tioned competition. Note, however, that the TRCE con-
founds task conflict and response conflict, since the
response competition takes place between responses that
were generated by two competing task rules (see Braver-
man & Meiran, 2010).
We analyzed the TRCE in the mixed blocks despite the
fact that participants were not reminded of the required task
in ARTS. This fact makes it conceivable that they may
have lost track of the task sequence, something that could
not have occurred in CTS. To ameliorate this problem, we
used very short blocks, but this may have been only a
partial solution. Therefore, one could argue that the para-
digms were not equated in this regard (namely, the chance
of committing the wrong task). Fortunately, it is possible to
estimate the rate of task errors by comparing error rates
between incongruent and congruent trials. While making a
task error in incongruent trials would lead to an error, it
would not lead to an error in congruent trials (assuming
that making the color or shape judgment is trivial, see
Meiran & Daichman, 2005). An analysis of the mixed-tasks
blocks revealed a robust TRCE in errors, F(1, 15) = 73.38,
MSE = 0.0019, g2p = 0.83, but the interaction with Para-
digm was clearly non-significant, F(1,15) = 1.44,
p = 0.25, g2p = 0.09. PE-TRCE was 0.07 in CTS and 0.06
in ARTS. Note that the direction of this (non-significant)
interaction is opposite to that predicted by the alternate
account (more task errors in ARTS).
We began by an analysis of raw RT, which indicated a
significant main effect for TRCE, F(1, 15) = 33.52,
MSE = 1,235.72, p \ 0.0001, g2p = 0.69, but a clearly
non-significant interaction between TRCE and Paradigm,
F(1, 15) = 0.87, MSE = 1,155.36, p = 0.36, g2p = 0.05
(42 and 30 ms in CTS and ARTS, respectively). We have
also fitted the ex-Gaussian model to RT distributions in the
mixed-tasks blocks, this time classifying the trials
according to Congruency (and ignoring the Switch vari-
able). The TRCE was neither significant in l (-2 and 3 ms
in CTS and ARTS, respectively) nor in r (-4 and 0 ms,
respectively), but has reached significance in s, F(1,
15) = 17.16, MSE = 2,149.14, p \ 0.001, g2p = 0.53.
However, the s-TRCE did not differ significantly between
paradigms, F(1, 15) = 1.12, p = 0.30, g2p = 0.07. It was
40 ms in CTS and 27 ms in ARTS.
The present results differ somewhat from Steinhauser
and Hubner’s (2009) results showing a significant Stroop
congruency effect (that resembles the TRCE in a sense) in
all the three ex-Gaussian parameters. Nonetheless, even in
their experiments, it was numerically larger in s than in the
Gaussian parameters. Moreover, Meiran and Kessler
(2008) found that the TRCE increases with response
slowness, a result that seems compatible with the current
trend.
In summary, the TRCE analyses do not suggest any
differential contribution of this variable to the paradigm
differences that we observed in the preceding analyses.
Discussion
The present work tested the hypothesis that, in the ARTS,
processes associated with dealing with actual or potential
task conflicts influence switch costs, mostly; whereas, for
CTS these processes influence mixing costs, mostly. This
hypothesis was based on Altmann’s (2007) hypothesized
Table 1 Means (and standard deviations) in ms
Trial-type Task RT Mean l r s Accuracy
(proportion)
Single tasks Shape 418 (49) 345 (30) 34 (8) 74 (35) 0.96 (0.03)
Color 420 (75) 329 (32) 30 (19) 91 (66) 0.98 (0.02)
Cued task
switching
Repeat Shape 560 (127) 357 (35) 34 (13) 202 (129) 0.96 (0.03)
Color 569 (158) 344 (45) 32 (19) 230 (157) 0.95 (0.03)
Switch Shape 656 (172) 409 (69) 60 (55) 246 (158) 0.91 (0.05)
Color 688 (216) 402 (118) 57 (58) 286 (148) 0.91 (0.04)
Alternating runs
task switching
Repeat Shape 515 (104) 348 (34) 31 (15) 166 (92) 0.97 (0.02)
Color 512 (103) 344 (50) 29 (14) 168 (92) 0.97 (0.02)
Switch Shape 702 (194) 420 (89) 67 (35) 283 (154) 0.95 (0.04)
Color 690 (184) 382 (116) 41 (50) 308 (135) 0.91 (0.05)
Psychological Research
123
strategic difference between the two paradigms, whereby
the abstract task representation of the task is encoded only
in the first trial of the run, whereas it is encoded in each and
every CTS trial. We additionally hypothesized that task
conflicts will influence performance in trials in which the
abstract task identity is encoded (Koch et al., 2005; Rubin
& Meiran, 2005). With these two hypotheses, combined,
we predicted that task conflicts will be seen in switch costs
in the ARTS and in mixing costs in CTS. Finally, following
Steinhauser and Hubner (2009), we used the s parameter
from the ex-Gaussian RT distribution (describing the
heaviness of the right RT-distribution tail) as an empirical
marker of task conflicts. As predicted, we found that s
contributed to mixing cost more in the CTS than in the
ARTS and it contributed more to switch cost in ARTS than
in CTS.
In addition to supporting the hypothesis (and the
hypotheses on which it was based), our results provide an
important replication to Steinhauser and Hubner’s (2009)
results. Their results were based on Stroop task switching,
which is a less commonly used version of the CTS para-
digm, and we showed similar results in a much more
standard CTS paradigm. More importantly, we found the
exact opposite pattern in the ARTS paradigm. This aspect
of our results, in fact, provides important support to Ste-
inhauser and Hubner given that our hypothesis was partly
Fig. 2 Mean cost as a function of cost-type and paradigm: a in reaction time (ms); b in proportion of errors; c in g (ms); d in r (ms); e in s (ms).
Error bars represent within-subject confidence intervals (Hollands & Jarmasz, 2010; Jarmasz & Hollands, 2009)
Psychological Research
123
based on their conclusions that the s parameter signals
processes associated with task conflict.
Importantly, the literature often refers to switch cost and
mixing cost as indications of specific processes. The
present results show that these processes change as a
function of the constraints afforded by the experimental
paradigm. Thus, for example, the fact that a given popu-
lation such as normally aged adults show enlarged mixing
costs, may tell us quite different things when the finding
comes from ARTS (e.g., Kray & Lindenberger, 2000) or
from CTS (e.g., Meiran, Gotler, & Perlman, 2001).
Two noteworthy limitations of this study should be
acknowledged. One is that, although we tried to match the
paradigms as closely as possible, some differences
remained, nonetheless. The other limitation is that we used
a particular combination of tasks, and given the large dif-
ferences between tasks, any general conclusion should be
treated as a conjecture. Nonetheless, the critical interaction
was not modulated by Task and included an element which
replicated a previous study (Steinhauser & Hubner, 2009)
with very different tasks, two facts that suggest that a
generalization may be warranted.
In conclusion, the present experiment provides impor-
tant converging evidence for two hypotheses. The first
hypothesis is that processes dealing with actual or potential
task conflicts contribute differentially to mixing cost and
switch cost, depending on the paradigm. The second
hypothesis is that the s parameter from the ex-Gaussian
model of the RT distribution, which indexes the heaviness
of the right RT-distribution tail, apparently marks pro-
cessing related to task conflicts.
Acknowledgments We wish to thank Maayan Pereg for her help in
designing the experiment, Miriam Gade and an anonymous reviewer
for helpful comments. This research was supported by a research
grant from the Israel Science Foundation to N. Meiran.
References
Altmann, E. M. (2007). Comparing switch costs: alternating runs and
explicit cuing. Journal of Experimental Psychology. Learning,
Memory, and Cognition, 33, 475–483.
Altmann, E. M. (2013). Fine-grain episodic memory processes in
cognitive control. Zeitschrift fur Psychologie, 221, 23–32.
Altmann, E. M., & Gray, W. D. (2008). An integrated model of cognitive
control in task switching. Psychological Review, 115, 602–639.
Andreadis, N., & Quinlan, P. T. (2010). Task switching under
predictable and unpredictable circumstances. Attention, Percep-
tion, & Psychophysics, 72, 1776–1790.
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.
Braverman, A., & Meiran, N. (2010). Task conflict effect in task
switching. Psychological Research, 74, 568–578.
Fagot, C. (1994). Chronometric investigation of task switching.
Unpublished doctoral dissertation, University of California, San
Diego.
Gilbert, S. J., & Shallice, T. (2002). Task switching: a PDP model.
Cognitive Psychology, 44, 297–337.
Goldfarb, L., & Henik, A. (2007). Evidence for task conflict in the
Stroop effect. Journal of Experimental Psychology: Human
Perception and Performance, 33, 1170–1176.
Hollands, J. G., & Jarmasz, J. (2010). Revisiting confidence intervals
for repeated measures designs. Psychonomic Bulletin & Review,
17, 135–138.
Jarmasz, J., & Hollands, J. G. (2009). Confidence intervals in
repeated-measures designs: the number of observations princi-
ple. Canadian Journal of Experimental Psychology, 63,
124–138.
Jersild, A. T. (1927). Mental set and shift. Archives of Psychology,
14(89), 81.
Kessler, Y., & Meiran, N. (2010). The reaction-time task-rule
congruency effect is not affected by working memory load:
further support for the activated long-term memory hypothesis.
Psychological Research, 74, 388–399.
Kiesel, A., Steinhauser, M., Wendt, M., Falkenstein, M., Jost, K.,
Philipp, A. M., et al. (2010). Control and interference in task
switching: a review. Psychological Bulletin, 136, 849–874.
Kiesel, A., Wendt, M., & Peters, A. (2007). Task switching: on the
origin of response congruency effects. Psychological Research,
71, 117–125.
Koch, I. (2003). The role of external cues for endogenous advance
reconfiguration in task switching. Psychonomic Bulletin &
Review, 10, 488–492.
Koch, I. (2005). Sequential task predictability in task switching.
Psychonomic Bulletin & Review, 12, 107–112.
Koch, I., Prinz, W., & Allport, A. (2005). Involuntary retrieval in
alphabet–arithmetic tasks: task-mixing and task-switching costs.
Psychological Research, 69, 252–261.
Kray, J., & Lindenberger, U. (2000). Adult age differences in task
switching. Psychology and Aging, 15, 126–147.
Lacouture, Y., & Cousineau, D. (2008). How to use MATLAB to fit
the ex-Gaussian and other probability functions to a distribution
of response times. Tutorials in Quantitative Methods for
Psychology, 4, 35–45.
Meiran, N. (1996). Reconfiguration of processing mode prior to task
performance. Journal of Experimental Psychology. Learning,
Memory, and Cognition, 22, 1423–1442.
Meiran, N. (2010). Task switching: Mechanisms underlying rigid vs.
flexible self control. In R. Hassin, K. Ochsner, & Y. Trope
(Eds.), Self control in society, mind and brain (pp. 202–220).
NY: Oxford University Press.
Meiran, N., Chorev, Z., & Sapir, A. (2000). Component processes in
task switching. Cognitive Psychology, 41, 211–253.
Meiran, N., & Daichman, A. (2005). Advance task preparation
reduces task error rate in the cueing task-switching paradigm.
Memory and Cognition, 33, 1272–1288.
Meiran, N., Gotler, A., & Perlman, A. (2001). Old age is associated
with a pattern of relatively intact and relatively impaired task-set
switching abilities. Journals of Gerontology: Series B: Psycho-
logical Sciences & Social Sciences B, 56, 88–102.
Meiran, N., & Kessler, Y. (2008). The task rule congruency effect in
task switching reflects activated long term memory. Journal of
Experimental Psychology: Human Perception and Performance,
34, 137–157.
Meiran, N., Kessler, Y., & Adi-Japha, E. (2008). Control by Action
Representation and Input Selection (CARIS): a theoretical frame-
work for task switching. Psychological Research, 72, 473–500.
Monsell, S. (2003). Task Switching. Trends in Cognitive Sciences, 7,
134–140.
Monsell, S., Sumner, P., & Waters, H. (2003). Task-set reconfigu-
ration with predictable and unpredictable task switches. Memory
& Cognition, 31, 327–342.
Psychological Research
123
Pereg, M., Shahar, N., & Meiran, N. (2013). Task-switching training
effects are mediated by working-memory management. Intelli-
gence, 41, 467–478.
Ratcliff, R. (1993). Methods for dealing with reaction time outliers.
Psychological Bulletin, 114, 510–532.
Rogers, R. D., & Monsell, S. (1995). The cost of a predictable switch
between simple cognitive tasks. Journal of Experimental Psy-
chology: General, 124, 207–231.
Rubin, O., & Meiran, N. (2005). On the origins of the task mixing
cost in the cuing task-switching paradigm. Journal of Experi-
mental Psychology. Learning, Memory, and Cognition, 31,
1477–1491.
Schmiedek, F., Oberauer, K., Wilhelm, O., Suss, H. M., & Wittmann,
W. W. (2007). Individual differences in components of reaction
time distributions and their relations to working memory and
intelligence. Journal of Experimental Psychology: General, 13,
414–429.
Schmitz, F., & Voss, A. (2012). Decomposing task-switching costs
with the diffusion model. Journal of Experimental Psychology:
Human Perception and Performance, 38, 222–250.
Shaffer, L. H. (1965). Choice reaction with variable S-R mapping.
Journal of Experimental Psychology, 70, 284–288.
Steinhauser, M., & Hubner, R. (2009). Distinguishing response
conflict and task conflict in the Stroop task: evidence from ex-
Gaussian distribution analysis. Journal of Experimental Psy-
chology: Human Perception and Performance, 35, 1398–1412.
Tornay, F. J., & Milan, E. G. (2001). A more complete task-set
reconfiguration in random than in predictable task switch. The
Quarterly Journal of Experimental Psychology, 54A, 785–803.
van ‘t Wout, F., Lavric, A., & Monsell, S. (2013). Are stimulus-
response rules represented phonologically for task-set prepara-
tion and maintenance? Journal of Experimental Psychology.
Learning, Memory, and Cognition, 39, 1538–1551.
Vandierendonck, A., Liefooghe, B., & Verbruggen, F. (2010). Task
switching: interplay of reconfiguration and interference. Psy-
chological Bulletin, 136, 601–626.
Waszak, F., Hommel, B., & Allport, A. (2003). Task-switching and
long term priming: role of episodic stimulus-task bindings in
task-shift costs. Cognitive Psychology, 46, 361–413.
Woodward, T. S., Meier, B., Tipper, C., & Graf, P. (2003). Bivalency
is costly: bivalent stimuli elicit cautious responding. Experimen-
tal Psychology, 50, 233–238.
Psychological Research
123