There Are Limits to the Effects of Task Instructions: Making the
Automatic Effects of Task Instructions Context-Specific Takes Practice
Senne Braem, Baptist Liefooghe, Jan De Houwer, Marcel Brass, and Elger L. Abrahamse
Ghent University
Unlike other animals, humans have the unique ability to share and use verbal instructions to prepare for
upcoming tasks. Recent research showed that instructions are sufficient for the automatic, reflex-like
activation of responses. However, systematic studies into the limits of these automatic effects of task
instructions remain relatively scarce. In this study, the authors set out to investigate whether this
instruction-based automatic activation of responses can be context-dependent. Specifically, participants
performed a task of which the stimulus-response rules and context (location on the screen) could either
coincide or not with those of an instructed to-be-performed task (whose instructions changed every run).
In 2 experiments, the authors showed that the instructed task rules had an automatic impact on
performance—performance was slowed down when the merely instructed task rules did not coincide, but,
importantly, this effect was not context-dependent. Interestingly, a third and fourth experiment suggests
that context dependency can actually be observed, but only when practicing the task in its appropriate
context for over 60 trials or after a sufficient amount of practice on a fixed context (the context was the
same for all instructed tasks). Together, these findings seem to suggest that instructions can establish
stimulus-response representations that have a reflexive impact on behavior but are insensitive to the
context in which the task is known to be valid. Instead, context-specific task representations seem to
require practice.
Keywords: cognitive control, task sets, instructions, context-sensitivity
The ability to use and share symbolic representations is com-
monly thought to separate humans from other animals, equipping
us with unique language abilities that guide our everyday action
and perception (Deacon, 1997). For example, using language we
can learn without trial and error the route to a new city, how to
build a cabinet, or how to prepare a meal. Although verbal instruc-
tions are omnipresent in daily life and psychological research, the
mechanisms via which they influence behavior are still poorly
understood. In line with this observation, Cole, Laurent, and
Stocco (2013) recently stressed the need for a more systematic
investigation into this unique ability, to help accelerate the insights
in what is now still a relatively small research domain.
Intriguingly, recent expeditions into the domain of instruc-
tion learning already demonstrated how the presentation of a
stimulus can result in the automatic activation of a response that
was assigned to that stimulus via instructions, even when the
stimulus-response (S-R) instruction was never before executed
(e.g., Liefooghe, Wenke, & De Houwer, 2012; Liefooghe, De
Houwer, & Wenke, 2013; Meiran, Pereg, Kessler, Cole, &
Braver, 2015a; Theeuwes, Liefooghe, De Houwer, 2014;
Wenke, Gaschler, & Nattkemper, 2007). Mere S-R instructions
were also shown to influence automatic motor activation during
stimulus presentation as indexed via electroencephalography
and electromyography (Bardi, Bundt, Notebaert, & Brass, 2015;
Everaert, Theeuwes, Liefooghe, & De Houwer, 2014; Meiran,
Pereg, Kessler, Cole, & Braver, 2015b). These findings suggest
that instructions can establish S-R associations that are activated in
an almost reflexive manner upon stimulus presentation (Meiran et
al., 2015b). Although evidence suggests that practice does further
strengthen these instructed associations (Wenke, De Houwer, De
Winne, & Liefooghe, 2015), the boundary conditions of the auto-
matic impact of task instructions remain largely unknown. In the
current study, we set out to investigate to which extent instruction-
based automatic behavior is restricted to the context in which the
instructions are said to be valid.
In our everyday lives, we are often required to learn that
certain actions or tasks are only applicable in certain contexts.
For example, the sound of your doorbell might trigger an
automatic tendency to stand up when you hear the sound in your
own house, but not when hearing it at a friend’s house. This
impact of task context has received ample attention in the
domain of trained (rather than merely instructed) task sets,
demonstrating that responses can be automatically activated in
a stimulus-driven and context-dependent manner (e.g., Abrahamse
This article was published Online First September 12, 2016.
Senne Braem, Department of Experimental Psychology and Department
of Experimental Clinical and Health Psychology, Ghent University; Baptist
Liefooghe and Jan De Houwer, Department of Experimental Clinical and
Health Psychology, Ghent University; Marcel Brass and Elger L. Abra-
hamse, Department of Experimental Psychology, Ghent University.
Senne Braem (12K6316N) and Elger L. Abrahamse (12C4715N) were
supported by FWO-Research Foundation Flanders. The research reported
in this paper was funded by the Interuniversity Attraction Poles Programme
initiated by the Belgian Science Policy Office (IUAPVII/33) and by the
Ghent University Methusalem Grant (BOF09/01M00209). We thank Nach-
shon Meiran, Matthew Crump, and an anonymous reviewer for their
helpful suggestions on an earlier version of this article.
Correspondence concerning this article should be addressed to Senne
Braem, Department of Experimental Psychology, Ghent University, Henri
Dunantlaan 2, B–9000 Ghent, Belgium. E-mail: senne.braem@ugent.be
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Experimental Psychology:
Learning, Memory, and Cognition
© 2016 American Psychological Association
2017, Vol. 43, No. 3, 394– 403
0278-7393/17/$12.00 http://dx.doi.org/10.1037/xlm0000310
394
& Verwey, 2008; Braem, Abrahamse, Duthoo, & Notebaert, 2014;
Bugg, Jacoby, & Toth, 2008; Cañadas, Rodríguez-Bailón, Mil-
liken, & Lupiáñez, 2013; Chun & Jiang, 1998; Crump, Gong, &
Milliken, 2006; Crump, Vaquero, & Milliken, 2008; Crump &
Logan, 2010; King, Korb, & Egner, 2012; Mayr & Bryck, 2005,
2007; Reuss, Desender, Kiesel, & Kunde, 2014; Rubin & Koch,
2006; Ruitenberg, Abrahamse, De Kleine, & Verwey, 2012;
Schouppe, de Ferrerre, Van Opstal, Braem, & Notebaert, 2014). In
fact, context-dependence seems to be an almost automatic conse-
quence of systematically repeating a certain task within a specific
context, and applies, among others, to the processing of response
conflict (e.g., Crump et al., 2006), the deployment of spatial
attention (e.g., Chun & Jiang, 1998), or the learning of sequential
motor skills (e.g., Abrahamse & Verwey, 2008). These studies
often differ both in terms of paradigm and specific effect of
interest. However, the manipulations and general motivation be-
hind these studies is very similar. These studies have in common
that they seem to evidence how task or goal representations can
become bound to the context in which they were most frequently
experienced (as reviewed in Abrahamse, Braem, Notebaert, &
Verguts, 2016). In this study, we are particularly interested in the
context-specificity of (instructed) S-R representations (e.g., Mayr
& Bryck, 2007).
In the lab, this context-dependency is often tested by presenting
a certain task or stimulus more in one over the other context (e.g.,
location on the screen, background color). After a certain amount
of practice, it is then observed that context starts to function as a
cue for the application of the task. For example, in the study by
Mayr and Bryck (2007) participants alternated between two tasks
in which they judged either the orientation (i.e., when vertical
press left, when horizontal press right), or the color (i.e., when
green press left, when red press right) of a rectangle. Note that
stimuli could vary on both dimensions of which only one was
relevant. This way, congruent (i.e., a vertical green rectangle) and
incongruent (i.e., a horizontal green rectangle) stimuli were cre-
ated. Crucially, participants responded more slowly to incongruent
stimuli, but only when both tasks were always presented at the
same location. When the location on the screen was predictive of
the task, the congruency effect disappeared, suggesting that the
location started indicating the applicability of the task rules. Inter-
estingly, this context-dependent task performance is often hypoth-
esized to be a product of implicit, habitual learning, which on its
turn is thought to require practice (Crump et al., 2008; Reuss et al.,
2014). Based on these studies, one might postulate that context-
dependent task performance could not be achieved on the basis of
mere instructions (without any prior task experience). This would
suggest that context-dependence requires actual experience. Yet,
such a potential boundary condition of what can be achieved on the
basis of mere verbal instructions remains to be tested. Importantly,
earlier studies on the context-dependence of task sets only used
paradigms in which the context was not instructed, and thus had to
be learned through practice.
Second, similar to how automaticity was thought to result ex-
clusively from practice but can actually be established by instruc-
tions (see above), we hypothesize that instructions can establish
context-dependence too. Specifically, the above-mentioned in-
struction studies (e.g., Liefooghe et al., 2012) suggest that working
memory is capable of maintaining and implementing stimulus-
response mappings, but it remains unclear to what extent working
memory can also integrate contextual cues into task-set represen-
tations. If working memory can bind task-sets to contextual fea-
tures, then environmental cues that match or mismatch these
contextual features may control whether the contents of working
memory interfere with performance. Therefore, the present study
aims to be a first test of whether people can integrate contextual
cues into their task-set representations on the basis of verbal
instructions alone.
We adapted the procedure introduced by Liefooghe and col-
leagues (Liefooghe et al., 2012, 2013) to include a context manip-
ulation (see Figure 1). As in previous work, the paradigm consisted
of an inducer task and a diagnostic task. In a series of runs, each
separate run started with the presentation of the instructions for the
inducer task, which consisted of two highly frequent Dutch four-
letter words (see the appendix) and their left versus right response
button assignment (e.g., “if lamp press left, if wolf press right”).
Participants were instructed to only respond to the word identity
when the word was presented in green (i.e., inducer task). In the
retention interval between the inducer task instructions and the
actual inducer task, several trials of a diagnostic task were per-
formed. Diagnostic trials were indicated by the fact that the four-
letter words were presented in a white font type (as opposed to the
green font inducer trials). While the instructions for the inducer
task changed each run, the instructions for the diagnostic task were
always the same: “When presented in upright font, press left; when
presented in italic font, press right.” Importantly, the identity of the
words was irrelevant for the diagnostic task.
Using this paradigm, Liefooghe and colleagues (2012, 2013) ob-
served congruency effects that resulted from the instructions of the
inducer task during performance on the diagnostic task. For example,
the word lamp presented in italic requires a right button response in
the diagnostic task, but a left button response in the later-to-be-
performed inducer task, thereby slowing down performance in the
diagnostic task—even though the inducer instructions were formally
irrelevant at that point in the run. As argued above, these findings
suggest that merely instructed, but never executed, S-R mappings
can automatically influence behavior. To investigate whether con-
text instructions could modulate this effect, we added the instruc-
tion that the inducer task would be presented either above or below
a central fixation cross. The diagnostic trials, however, were al-
ways randomly presented above or below the fixation cross. This
design allowed us to investigate whether the instructed task con-
text (i.e., stimulus location) would impact the automatic interfer-
ence from merely instructed task rules. Specifically, if context-
dependence can be instructed, we would expect to observe the
impact of the instructed task-rules on diagnostic trials only at the
location of the later-to-be-performed inducer trial.
Experiment 1
Method
Participants. Thirty-two students (range ⫽ 18 –20 years, 29
women, 28 right-handed) took part in return for course credits or
10€. Although earlier studies on the context-sensitivity of trained
task sets (e.g., Crump et al., 2006; Mayr & Bryck, 2005, 2007)or
the instruction-based congruency effect (e.g., Liefooghe et al.,
2012) all used around 14 to 18 subjects per experiment, we decided
to test 32 subjects for our first experiment to ensure sufficient
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
395
LIMITS TO THE EFFECTS OF TASK INSTRUCTIONS
power to detect context effects. For all of the hereafter presented
experiments we recruited 20 subjects.
Stimuli and material. A list of 98 highly frequent four-letter
Dutch nouns (see the appendix) were selected from the
SUBTLEX-NL database (Keuleers, Brysbaert, & New, 2010). For
each participant, a randomly determined set of 49 unique pairs of
words was chosen, of which one was assigned to the practice run
and the other 48 to the six blocks of experimental runs (blocks
were interspersed with self-paced brakes). Specifically, eight pairs
per block were randomly assigned to the eight different runs within
a block, of which two runs counted either four, eight, 12, or 16
trials of the diagnostic task, which were presented in a random
order.
The words were presented either above or below the centrally
presented fixation cross, halfway between the fixation cross and
the edge of the screen, in Arial 24-point font. The words in the
diagnostic task could be presented either in italic or in upright font
which determined the correct response (see above). Importantly,
the location and congruency of the words were randomized but
balanced across trials (all participants received an equal number of
congruent and incongruent trials at either location, per run). The
instructions for the inducer task always consisted of three lines and
were similarly presented above or below the fixation cross. The
location of these instructions corresponded to the location on
which the stimulus for the inducer task would be presented. This
location was determined randomly across runs. All instructions
and stimuli were presented in white on a black background, except
for the words during the inducer task which were presented in
green. Stimuli were presented on a CRT monitor located 60 cm
away from the eyes using Tscope software (Steven, Lammertyn,
Verbruggen, & Vandierendonk, 2006). The left and right response
key were the letters F and J, respectively, on a standard QWERTY
keyboard.
Procedure. Participants read the general instructions, after
they were seated behind the computer in a dimly lit room and filled
in the informed consent form. The instructions explained the
general procedure as well as the specific response rules for the
diagnostic task. These response rules were the same for all partic-
ipants: Participants had to press left when the stimulus was pre-
sented in upright font and right when presented in italic. The
instructions further emphasized to use the location information of
the inducer task instructions to respond as fast as possible to the
inducer task. After a first example run with 12 diagnostic trials, the
participants read the instructions once more, and completed six
blocks of eight runs.
The trial procedure is visualized in Figure 1. First, the inducer
task instructions were presented on the screen above or below the
fixation cross, depending on the inducer task location. The first
line mentioned whether the task would be presented above or
below the fixation cross with the location printed in uppercase
letters (i.e., the Dutch translation of “The task will be presented
ABOVE”), under which the next two lines indicated the run-
specific response mapping (e.g., the Dutch translation of “if wolf,
press right”). The order of the lines indicating the left or right
Figure 1. General paradigm and trial procedure. At the start of each run, participants were instructed about the
inducer task which they had to perform whenever the word appeared in green. After 750 ms the first of four,
eight, 12, or 16 trials of the diagnostic task appeared. The duration of the inter-trial-interval for the diagnostic
task randomly varied between 500 and 1,000 ms and was indicated by the presence of a black fixation cross.
During this task participants had to respond to the orientation of the font type, while the meaning of the word
was irrelevant. The inducer task was preceded by a 500-ms presentation of a green fixation cross. In Experiment
2, the inducer task could appear at the alternative location in 12.5% of the trials, indicating that participants
should not respond to the inducer task. In Experiment 3, 69 trials of the inducer task preceded the diagnostic task.
In Experiment 4, the location of the inducer task was the same for all runs and was counterbalanced across
participants. In Experiments 2 to 4 a rectangle surrounded each stimulus to further emphasize its location. See
the online article for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
396
BRAEM, LIEFOOGHE, DE HOUWER, BRASS, AND ABRAHAMSE
button response was randomized. These instructions remained on
the screen for 20 s or until the participant pressed the spacebar.
After the instructions and a 750-ms interval, the first stimulus of
the diagnostic task appeared. Each stimulus of the diagnostic task
was presented until response or for 2,000 ms. When the participant
responded incorrectly, a red screen appeared for 200 ms. In be-
tween trials, there was a random intertrial interval that was be-
tween 500 and 1,000 ms long. Following the final diagnostic trial
of each run, the same intertrial interval was presented after which
the fixation cross turned green for another 500 ms. Finally, the
inducer task stimulus appeared which was also presented in green
and remained on screen until response or for 2,000 ms. A new run
started 1,500 ms after the participant performed the inducer task.
The fixation cross remained on the screen throughout the entire
run.
Results
Three participants were excluded from further analyses because
they performed at chance level on either the diagnostic task (n ⫽
2), or the inducer task (n ⫽ 1). Mean accuracy for the remaining
29 participants was 92.3% (SD ⫽ 6.0%) on the diagnostic task and
86.5% (SD ⫽ 9.7%) on the inducer task. All further analyses
focused on the diagnostic task only. Trials following an error, or
trials that were part of a run on which the inducer task was
performed inaccurately, were removed from analyses. The former
is to avoid posterror (and postfeedback) slowdown effects in our
data, whereas the latter is to exclude trials where we cannot
guarantee that participants successfully implemented the inducer
task. For the reaction time (RT) analysis, only RTs of correct
responses that were slower than 200 ms were considered. As a
result of these exclusion criteria, 24.1% of the diagnostic trials
were removed for RT analyses. Median RT and mean accuracy
results were analyzed statistically in repeated-measures analyses of
variance (rANOVA) with two within-subject factors for congru-
ency (congruent vs. incongruent) and context (same vs. different
location as inducer task). To evaluate whether the following results
depended on the experimenter’s choice of using median RTs
instead of the more conventional mean RTs, we also analyzed
mean RT data using an outlier criterion of 2.5 SD (see also,
Everaert et al., 2014; Theeuwes et al., 2014). Importantly, similar
significance levels were obtained, which rendered the same statis-
tical conclusions as below. All rANOVA and follow-up tests were
two-tailed. To determine whether a nonsignificant finding could be
considered support for the null hypothesis (i.e., the location rele-
vance does not affect the automatic activation of task instructions),
we also performed Bayesian analyses on null findings. Specifi-
cally, we computed the Bayes factor BF
01
, which quantifies the
evidence for the null hypothesis against the evidence for the
alternative hypothesis. To this end, we used the open source
statistical program JASP Version 7.1 (Love et al., 2015) and ran
one-sided Bayesian tests (with Cauchy prior width ⫽ 1, as spec-
ified by Rouder et al., 2009).
RTs. Overall, there was a significant congruency effect, F(1,
28) ⫽ 9.55, p ⬍ .005. Moreover, a significant context-effect was
observed, F(1, 28) ⫽ 4.93, p ⬍ .05, indicating a RT benefit when
the diagnostic task occurred on the same, relative to the alternative,
location as the later-to-be performed instructed inducer task. Im-
portantly, despite this evidence that context was processed during
the diagnostic task, the factors congruency and context did not
interact, F(1, 28) ⫽ 0.58, p ⫽ .452, indicating that the congruency
effect was not modulated by the context in which it occurred. As
can be seen on Figure 2, the effect, if anything, was numerically
larger at the alternative location.
Error rates. The error rates analysis only showed a signifi-
cant congruency effect, F(1, 28) ⫽ 16.551, p ⬍ .001, indicating
more accurate responses on congruent trials. The other effects did
not reach significance, both Fs(1, 28) ⬍ 1.
Bayesian analyses. The BF
01
for the interaction between con
-
gruency and context was 11.499 in the RT analyses (and 4.113 in
the error analyses), suggesting that these data are 11.499 more
likely to be observed under the null hypothesis. This is considered
strong evidence for the null hypothesis that the automatic location-
specific activation of task sets cannot be observed on the basis of
mere instructions (Jeffreys, 1961).
Discussion
The results are in line with earlier observations showing that
instructed, but never executed, S-R mappings can have an auto-
matic impact on task performance (Liefooghe et al., 2012, 2013).
Interestingly, the instructed inducer task context had an impact on
diagnostic task performance, as evidenced by the faster RTs at the
location where the inducer task was expected to occur. This
suggests that participants’ (spatial) attention was modulated by the
context instruction. However, although participants clearly en-
coded the instructions about both the S-R mapping and task
context, the results suggest that both types of information were not
integrated. Specifically, the impact of the task-irrelevant instructed
S-R rules was not modulated by location.
These results demonstrate that, unlike to what is the case for
practiced S-R mappings (e.g., Cañadas et al., 2013; Crump et al.,
2006, 2008; Crump & Logan, 2010; Mayr & Bryck, 2005, 2007;
Reuss et al., 2014; Rubin & Koch, 2006; Schouppe et al., 2014),
tasks cannot be triggered in a context-specific manner on the mere
basis of instructions. Possibly, the integration between task sets
and their context is something that requires practice and is not
possible on the basis of mere instructions. This interpretation
would be in line with the idea that the context-dependency of task
Figure 2. Mean median reaction times for Experiments 1 and 2. Reaction
times are depicted for each congruency and context condition separately.
As reported in the results sections, the congruency and context effects both
reached significance, but there was no interaction. All error bars are ⫾ 1
standard error of the mean.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
397
LIMITS TO THE EFFECTS OF TASK INSTRUCTIONS
sets is acquired through implicit learning mechanisms that require
practice (Crump et al., 2008; Reuss et al., 2014). However, before
drawing this conclusion we wanted to further probe the impact of
instructed task contexts.
Experiment 2
Although the main effect of context in Experiment 1 seems to
suggest that participants did implement the context information,
this information was not necessary for accurate task performance.
At best, remembering this information could speed up inducer task
performance. Therefore, in our second experiment, we further
stressed the role of task context by instructing that participants
should withhold from responding to the inducer task, whenever it
appeared at the uninstructed location. In line with this general
instruction, we presented one in eighth of the inducer trials at the
alternative location. Note that, here, our context manipulation is
different from the other experiments, as well from previous stud-
ies, in that other context studies focused on the effects of contexts
that were nonessential for accurate task completion. However, we
reasoned that making context task relevant could increase its
saliency and therefore increase the chances of observing a modu-
latory effect of context instructions.
Method
Participants. Twenty students (range ⫽ 18 –36 years, 14
women, 19 right-handed) took part in return for course credits or
10€.
Stimuli and material. The stimuli and material were exactly
the same as Experiment 1 (see Figure 1), except that the task
stimuli were now surrounded by a rectangle that further empha-
sized the location of the stimulus. Centered around the stimulus,
the rectangle was 60% of the screen wide, and 20% of the screen
high. Similar to the word stimuli, the rectangle was white for the
diagnostic task, and green for the inducer task.
Procedure. The procedure was similar to Experiment 1. How-
ever, the participants were now instructed that the inducer stimulus
would not always appear at its instructed location. Consistent with
this instruction, and in contrast to the other experiments, one out of
eight inducer trials (randomly determined) did not appear at its
instructed location, but at the alternative location instead. Further-
more, whenever the inducer stimulus appeared at the alternative
location, participants were explicitly instructed to withhold their
response. This way, the location information was also necessary to
perform the task accurately.
Results
Four participants were excluded from further analysis, because
they either performed at chance level on the inducer task (n ⫽ 1),
or ignored the instruction to withhold from responding when the
inducer task was at the wrong location (n ⫽ 3). Mean accuracy for
the remaining 16 participants was 94.7% (SD ⫽ 2.6%) on the
diagnostic task and 90.2% (SD ⫽ 6.6%) on the inducer task.
Moreover, the remaining 16 participants were able to successfully
withhold their response when the inducer task was presented at the
wrong location (inhibition accuracy ⫽ 85.8%; SD ⫽ 13.3%). The
data preparation procedure was the same as for Experiment 1.
Similar to Experiment 1, we only analyzed trials that were part of
a run on which the inducer task was performed accurately. Runs
where the inducer task was presented on the wrong location (and
inducer task performance could thus not be evaluated), were also
excluded from the analyses. As a result of these exclusion criteria,
27.9% of the diagnostic trials were removed for RT analyses.
RTs. Similar to Experiment 1, we observed significant main
effects of congruency, F(1, 15) ⫽ 5.46, p ⬍ .05, and context, F(1,
15) ⫽ 16.01, p ⬍ .005. Importantly, the factors congruency and
context again did not interact, F(1, 15) ⫽ 0.02, p ⫽ .892, indicat-
ing that also in Experiment 2 the congruency effect was not
modulated by the context in which it occurred (see Figure 2).
Error rates. None of the effects in the error rates analysis
reached significance, all Fs(1, 15) ⬍ 1.
Bayesian analyses. The BF
01
for the interaction between con
-
gruency and context was 4.747 in the RT analyses (and 2.151 in
the error analyses), suggesting that these data are 4.747 more likely
to be observed under the null hypothesis. This is considered
substantial evidence for the null hypothesis (Jeffreys, 1961). Pool-
ing the data of Experiment 1 and 2 together, we observed a BF
01
of 12.048 in the RT analyses (and 2.993 in the error analyses) for
the null hypothesis that the automatic location-specific activation
of task sets cannot be observed on the basis of mere instructions.
Discussion
The findings of Experiment 2 replicated those of Experiment 1.
Again, we observed that both the instructed task rule and task
context from the inducer task had an automatic impact on the RTs
of the task-irrelevant diagnostic task, but did not interact. Together
with the results of Experiment 1, our findings suggest that partic-
ipants did not integrate the task rules with their task context, as the
automatic impact of task rules was independent of the context in
which they appeared. These results seem to contrast with what has
consistently been found with trained task sets, suggesting that the
context-sensitivity of task sets requires practice. If practice is
indeed a crucial ingredient for observing a context-dependent
automatic activation of task sets, we should be able to observe a
context-modulated congruency effect after the inducer task has
been practiced in its context. This would confirm that the absence
of context-specific effects in Experiments 1 and 2 is not simply
due to some unique feature of our procedure.
Experiment 3
In Experiment 3, we tested an alternative version of our para-
digm in which the diagnostic trials were preceded by at least 69
trials of inducer task trials. Based on the prior literature and the
idea that training allows for context-specific impact of S-R asso-
ciations, we expected to observe a context-specific congruency
effect in the present experiment.
Method
Participants. Twenty students (range ⫽ 18 –29 years, 19
women, 19 right-handed) took part in return for course credits or
10€.
Stimuli and material. The stimuli and material were exactly
the same as Experiment 2.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
398
BRAEM, LIEFOOGHE, DE HOUWER, BRASS, AND ABRAHAMSE
