ArticlePDF Available

Implicit Sequence Learning and Contextual Cueing Do Not Compete for Central Cognitive Resources

Authors:

Abstract and Figures

Sequence learning and contextual cueing explore different forms of implicit learning, arising from practice with a structured serial task, or with a search task with informative contexts. We assess whether these two learning effects arise simultaneously when both remain implicit. Experiments 1 and 2 confirm that a cueing effect can be observed under a continuous setting and that there is no interference between contextual cueing and sequence learning. Experiments 3a and 3b tested whether an interference arises specifically when the sequence becomes explicit. Results show that the expression of contextual cueing disappeared in those conditions but that context information is still acquired, and it affects performance when the sequence is removed. The results are discussed in relation to the current debates about the automaticity of implicit learning, and about the role of attention in the acquisition and expression of contextual cueing.
Content may be subject to copyright.
Implicit Sequence Learning and Contextual Cueing Do Not Compete for
Central Cognitive Resources
Luis Jime´nez and Gustavo A. Va´zquez
Universidad de Santiago
Sequence learning and contextual cueing explore different forms of implicit learning, arising from
practice with a structured serial task, or with a search task with informative contexts. We assess whether
these two learning effects arise simultaneously when both remain implicit. Experiments 1 and 2 confirm
that a cueing effect can be observed under a continuous setting and that there is no interference between
contextual cueing and sequence learning. Experiments 3a and 3b tested whether an interference arises
specifically when the sequence becomes explicit. Results show that the expression of contextual cueing
disappeared in those conditions but that context information is still acquired, and it affects performance
when the sequence is removed. The results are discussed in relation to the current debates about the
automaticity of implicit learning, and about the role of attention in the acquisition and expression of
contextual cueing.
Keywords: implicit learning, sequence learning, contextual cueing, attention
The concept of automaticity appears too wide and fuzzy to allow
for a single test of that property. Describing a cognitive process as
automatic has been taken to mean different things such as that the
process runs without intention or unconsciously, that it proceeds
together with other nonautomatic processes without incurring cog-
nitive costs, that it goes on in an efficient manner, that it is goal
independent or stimulus driven, or that it fulfills just a particular
subset of all these attributes (see the work of Moors & De Houwer,
2006, for an extensive analysis of the term). However, despite a
persistent failure to define automaticity, the concept has survived
its own ambiguity, as if it holds a promise to provide researchers
with an important tool to understand cognition (Bargh & Morsella,
2008).
In the learning literature, the difference between automatic and
nonautomatic processes has been identified with the distinction
between implicit and explicit learning. Implicit learning is often
characterized as automatic, in that it takes place unintentionally as
a side effect of processing and responding to structured stimuli. In
contrast, explicit learning is taken as the result of an intentional
attempt to encode and retain the relevant information contained in
a structured environment. It has also been argued that the results of
implicit learning processes can remain unconscious under certain
circumstances and that they do not rely on resource-demanding
operations such as testing hypotheses or actively trying to commit
the relevant information to memory. The issue about the indepen-
dence of implicit learning processes from the availability of central
cognitive resources has been tested thoroughly within the sequence
learning paradigm (Jime´nez, 2003; Nissen & Bullemer, 1987;
Shanks, 2003), which has become one of the most popular para-
digms to investigate implicit learning.
Sequence Learning
The sequence learning paradigm was initially developed by
Nissen and Bullemer (1987). These authors devised the serial
reaction-time (SRT) task, in which participants were told to re-
spond to a series of stimuli by pressing on the key corresponding
to its current location. Unbeknownst to participants, the series of
locations was structured to follow a predictable sequence, and
participants learned about the sequence as attested by a slowdown
contingent to its removal. Although learning was incidental in
these conditions, and it resulted in more knowledge than that
which could be reported on conventional measures of awareness,
it has been argued that the underlying processes could not qualify
as automatic, because performing the SRT task together with a
secondary tone-counting task resulted in a decrease in the effect of
sequence learning (e.g., Nissen & Bullemer, 1987; Rowland &
Shanks, 2006; Shanks, 2003; Shanks & Channon, 2002; Shanks,
Rowland, & Ranger, 2005). Alternative accounts attributed this
pattern of results either to a specific interference on explicit
acquisition processes or to an increased distortion of the sequence
provoked by interspersing a series of random tones that became
integrated with the sequence. According to these arguments, Jime´-
nez and Va´zquez (2005) reported that the effect of the secondary
task decreased when the series of tones covariated with the loca-
tions, hence reducing the disruption caused by the integration of
both series (see also Schmidtke & Heuer, 1997). Moreover, the
authors also showed that contrary to a resource-dependence view,
the impact of a secondary task did actually decrease and appeared
This article was published Online First August 23, 2010.
Luis Jime´nez and Gustavo A. Va´zquez, Universidad de Santiago.
This research was supported by the Spanish Ministerio de Educacio´n y
Ciencia (MEC) with research grants SEJ2005-25754-E and SEJ2006-
27564 to Luis Jime´nez, and with a predoctoral grant to Gustavo A.
Va´zquez. The authors wish to thank Herbert Heuer and Sander Los for their
thoughtful comments on previous versions of the article.
Correspondence concerning this article should be addressed to Luis
Jime´nez, Facultad de Psicologı´a, Universidad de Santiago, 15782-
Santiago, Spain. E-mail: luis.jimenez@usc.es
Journal of Experimental Psychology: © 2010 American Psychological Association
Human Perception and Performance
2011, Vol. 37, No. 1, 222–235 0096-1523/10/$12.00 DOI: 10.1037/a0020378
222
later when the sequence was made more complex by arranging
only probabilistic contingencies.
A few studies have approached the question about the role of
central cognitive resources in implicit sequence learning by adding
a second learning task, instead of a counting task unrelated to
learning. The rationale underlying this strategy is that if all learn-
ing processes are assumed to depend on a central pool of resource-
demanding operations, then learning simultaneously about two
sequential contingencies should be harder than learning about a
single one. In turn, if two effects of learning arose simultaneously
without cost, then the most parsimonious conclusion would be that
at least one of them was acquired automatically, without taxing a
shared and limited amount of cognitive resources.
A number of experimental reports have shown that two inde-
pendent learning effects can be obtained without cost under certain
circumstances. Mayr (1996) used a variant of the SRT task in
which participants responded to the color of a stimulus appearing
on each trial at one of several possible locations, and in which both
the sequence of locations and the sequence of colors were predict-
able. The author showed that participants learned simultaneously
about each of these two sequences to the same extent as did two
control groups that were exposed to a single sequence. Analogous
effects have been found by arranging a sequence of locations
together with a temporal sequence in the response-to-stimulus
intervals (Shin, 2008; Shin & Ivry, 2002).
Dual learning effects have been reported not only when two
independent cues predict different dimensions of the next trial, but
also when two cues predict the same relevant outcome (Cleer-
emans, 1997; Jime´nez&Me´ndez, 1999, 2001). In Cleeremans’
(1997) study, a series of targets followed a complex, probabilistic
series of locations, but a secondary stimulus also predicted the
following location, by appearing precisely at the location on which
the next target was more likely to appear. In these conditions,
participants learned to predict the location of the next target by
relying on the additional cue, but they also learned about the
complex series of target locations. Jime´nez and Me´ndez (1999,
2001) used a symbolic procedure in which the next location was
predicted by the identity of the current stimulus rather than by the
location of an additional cue. In this case, indeed, learning about
that secondary contingency was acquired exclusively when partic-
ipants performed an additional task on that dimension. However,
neither performing that secondary task nor learning about the
additional contingency interfered with the expression of sequence
learning. Moreover, learning about a complex sequence of loca-
tions was obtained even when the participants were informed
about the predictive relation arranged between the identity of each
stimulus and the following location (Jime´nez&Me´ndez, 2001).
Likewise, Cleeremans reported that sequence learning was ac-
quired even when the contingency between the alternative cue and
the following location was perfectly reliable. In this case, however,
the expression of sequence learning was concealed by the presence
of the more reliable cue, but learning was expressed over a transfer
phase in which the alternative cue had been removed.
Together, these results attest to the automatic nature of these
implicit learning processes, which arise simultaneously with other
learning effects, even in conditions in which the presence of
alternative, explicit cues could arguably occupy most of the learn-
ers’ cognitive resources, and remove any obvious motivation to
keep looking for sequential contingencies. Moreover, these results
illustrate how such dual-learning settings could be used as a tool to
demonstrate the automatic nature of other sources of implicit
learning.
Contextual Cueing
The aim of the present study is to apply a similar procedure to
assess whether the processes underlying contextual cueing could
be construed as automatic in precisely this sense. Contextual
cueing refers to a relatively recent addition to the implicit learning
paradigms, which has received increasing interest over the last
decade (e.g., Chun & Jiang, 1998, 1999, 2003; Endo & Takeda,
2005; Jiang & Chun, 2001, 2003; Junge´, Scholl, & Chun, 2007;
Olson & Chun, 2001, 2002; Rausei, Makovski, & Jiang, 2007;
Smyth & Shanks, 2008). In this paradigm, first developed by Chun
and Jiang (1998), participants are repeatedly presented with a
search task in which they are required to look for a rotated Tshape
among a number of rotated Lshapes. Participants are instructed to
respond by pressing one of two keys depending on the orientation
of the target T. When training is structured so that the spatial
configurations of distractors are often associated with particular
target locations, the authors found that participants learn about
these regularities, responding faster to repeated configurations
rather than to variable ones. Contextual cueing, thus, has been
interpreted as showing an improvement in the search for the
location of a target when the configuration of distractors is sys-
tematically associated with the location of the target.
In contrast to the strategy used in the sequence learning para-
digm, the role of attention in contextual cueing has been studied
mainly by manipulating filtering factors. Jiang and Chun (2001),
for instance, manipulated the color of the repeated distractors and
found that contextual cueing effects arose only when the informa-
tive distractors were drawn in the same color as that of the target,
but not when they were drawn in a different color, thus allowing
participants to successfully ignore them. Further experiments re-
fined this conclusion by showing that some latent context learning
could be acquired even in these unattended conditions (Jiang &
Leung, 2005). In this case, indeed, learning was not expressed over
the training phase when the informative distractors were drawn in
a different color. However, switching colors over a transfer phase
revealed that participants had learned about these repeated distrac-
tors, because this change did immediately lead to an effect of
contextual cueing. Thus, it is possible that the repeated configu-
ration of distractors was encoded even without selective attention,
but that their effects were not expressed when the benefits derived
from using that contextual information were outweighed by those
derived by filtering it out. Recent results reported by Rausei et al.
(2007) pointed in the same direction, by showing that comparable
effects of contextual cueing were obtained regardless of the sim-
ilarity between targets and distractors, even though larger similar-
ities resulted in larger dwelling times and thus, arguably, in more
attention devoted to the processing of distractors.
The specific aim of this study is to jointly explore the role of
attention in contextual cueing and in sequence learning, by assess-
ing whether contextual cueing effects could arise together with the
effects of sequence learning. In so doing, we also expect to obtain
informative results with respect to the ongoing discussion about
the neural bases of these two learning processes. The most ac-
cepted views concerning the neural structures underlying these two
223
SEQUENCE LEARNING AND CONTEXTUAL CUEING
processes assume that implicit sequence learning involves struc-
tures associated with procedural memory, such as the supplemen-
tary motor area, parietal regions, and basal ganglia (Hazeltine &
Ivry, 2003). In contrast, contextual cueing processes, even though
they have been as assumed to be unconscious, have been shown to
be impaired in amnesia cases, and have been related to the action
of areas traditionally associated with declarative memory, such as
the hippocampus and related medial temporal structures (Chun &
Phelps, 1999; Greene, Gross, Elsinger, & Rao, 2007). This non-
overlapping distribution over cortical and subcortical structures
could be taken as predicting a complete independence between
them. However, some imaging studies of sequence learning have
also found activation of these medial temporal structures associ-
ated with sequence learning (Schendan, Searl, Melrose, & Stern,
2003), especially for those learning conditions involving higher-
order associations, and when learning results in more explicit
knowledge. The traditional role of the medial temporal lobe in
declarative memory thus makes it more important to ascertain
whether this kind of context information can be encoded and
retrieved without cognitive cost, simultaneously with other pro-
cesses of sequence learning involving relational, second-order
conditional information. If, as was claimed by Schendan and
co-workers (2003), the role of the hippocampus in higher-order
implicit sequence learning involves binding sequential events into
a unique episode, regardless of the learner’s awareness, then it
becomes more relevant to ascertain whether encoding such diach-
ronic information could be done simultaneously with the encoding
of the synchronic, spatial configuration that is required to produce
contextual cueing.
Methodologically, the present study follows up on Jime´nez and
Va´zquez’s (2008) proposal to show that implicit sequence learning
can be acquired in the context of a search task, in which partici-
pants were to look for a target surrounded by distractors, and to
respond in terms of the target identity. In the present experiments
we replicated these sequence learning effects, and added contex-
tual information, to assess whether both learning results could arise
simultaneously without mutual interference.
In Experiment 1 we adapted the standard contextual cueing
paradigm to a continuous presentation pattern in which successive
trials were presented without pauses or fixation points, and in
which the number of possible responses was augmented with four
alternatives instead of the usual two, so as to allow the merging of
the contextual cueing and the sequence learning paradigms. In
Experiment 2a group of participants was trained with such a
blended paradigm, by arranging the targets to follow a second-
order conditional (SOC; Reed & Johnson, 1994) sequence, while
their locations were cued, on half of the trials, by the configuration
of distractors. The simultaneous effects of contextual cueing and
sequence learning were compared with those obtained by control
groups, which were presented exclusively with one of these con-
tingencies.
Finally, in Experiments 3A and 3B we arranged either simpler
deterministic sequences or more complex probabilistic structures
to assess the effect of explicit sequence learning on the acquisition
and expression of contextual cueing. In Experiment 3A a simpler
sequence was repeated consistently over training, and we predicted
that participants could become fully aware of it, and hence that
they could end up responding to the task without paying much
attention to the context. In contrast, when the sequence was made
probabilistic in Experiment 3B, we expected that knowledge
would not become explicit, and therefore that participants would
need to keep paying continuous attention to each display in order
to accurately identify the target. If attention is necessary to learn
about an informative context, then stronger contextual cueing
effects could be expected to arise in these latter conditions, when
the sequence was only probabilistic. In contrast, if only minimal
attention to the context is required to produce contextual learning,
then it could be acquired and expressed regardless of whether the
identity of each target was reliably predicted by the series of
previous targets.
In all of these experiments, we assessed sequence learning and
contextual cueing indirectly through the measures of performance,
but we also included direct measures of the same learning effects,
arranging a cued-generation task and a location-guessing task at
the end of each experiment. These two tasks were not taken as pure
measures of conscious knowledge, but their results were taken into
account under the assumption that all the conscious knowledge
that could be used to improve performance over the indirect
tasks could also be applied, under direct instructions, either to
anticipate the next target on the basis of the previous ones, or to
guess the location of a removed target in terms of the configu-
ration of the context. So, as far as these two direct measures fail to
reveal the acquisition of any relevant knowledge, we will be
inclined to accept that the knowledge expressed through the indi-
rect measures is not conscious.
Experiment 1: Adaptation of the Contextual Cueing
Paradigm
To adapt the standard contextual cueing paradigm to conditions
allowing for the inclusion of a sequence, we modified some
features from the original setting. In the standard procedure there
were only two possible responses (left vs. right orientation of the
target T), but sequence learning is usually assessed with tasks
involving more responses. Thus we adopted a procedure from
Jime´nez and Va´zquez (2008) in which participants were instructed
to search for, and respond to, the identity of one of four possible
targets (even numbers), which appeared surrounded by seven
distractors. This arrangement allowed for fast responses, and pro-
duced salient configurations of the context, while avoiding an
automatic detection of the target. With this procedure, Jime´nez and
Va´zquez found that participants learned about a SOC sequence
when the series of trials were presented continuously over 12
blocks of trials. The goal of this experiment was to ascertain
whether contextual cueing could be obtained under these circum-
stances, and whether the continuous presentation could affect the
results, in comparison with those obtained with more standard
temporal parameters.
Method
Participants. Forty-eight students from the University of
Santiago participated in the experiment and were paid 5. They
had never participated before in any implicit learning experiment,
and declared to have normal or corrected-to-normal vision.
Twenty-four participants were randomly assigned to a Discrete
condition, in which each trial was preceded by a fixation point of
500 ms, and in which the intertrial interval was fixed to 1000 ms.
224 JIME
´NEZ AND VA
´ZQUEZ
The remaining 24 participants were presented with a Continuous
condition in which the fixation point was removed, and the
response-to-stimulus interval (RSI) was reduced to 200 ms.
Apparatus and materials. The experiment was run on Pen-
tium III PCs connected to color monitors. Responses were made on
Spanish QWERTY keyboards. The experimental program was
designed with INQUISIT 1.31 (Inquisit, 2004). The stimuli con-
sisted of a set of colored digits printed in Garamond font, 1.1 cm
high by 0.7 cm wide, over a grey background. Target stimuli were
even numbers (2, 4, 6, and 8) presented in one of four possible
colors (red, blue, green, or yellow). The target appeared on each
trial at one of the 16 locations defined by an invisible 4 4 matrix,
8.4 cm wide by 8.6 cm high, accompanied by seven distractors.
Vertical and horizontal lines divided the matrix into four quad-
rants. Between neighboring slots there was a horizontal separation
of 1.9 cm and a vertical separation of 1.4 cm. The location of the
target was decided randomly without replacement over each series
of 16 successive trials. After a series of 16 trials, the next trial
could appear at any other location except at the location sampled
on the previous trial. The distractors for each trial were seven
instances of the same odd number, randomly chosen from the set
1, 3, 5, 7. The seven distractors plus the target stimulus were
colored and located pseudo-randomly for each trial, so that two of
them were drawn in each possible color (red, blue, green, and
yellow), and two items were located at each one of the four matrix
quadrants (see Figure 1).
The identity of the target, and therefore the response required for
each trial, was decided pseudo-randomly with the constraint that
all four possible targets were equally frequent, and repetition of the
same target was not allowed on consecutive trials. As for the
relation between the configuration of the context and the location
of the target, the program designated two positions from each
quadrant as being associated with repeated contexts for each
participant (i.e., eight repeated-context trials in total), whereas the
other two positions from each quadrant were assigned to variable
contexts. For repeated-context trials, the location of the target was
associated with a global configuration of the context, including the
identity, colors, and locations of the distractors, as well as the color
(but not the identity) of the target. As for the variable-context
trials, a different configuration of colors and locations was ran-
domly generated for each of the four possible identities of the
distractors, so that the same location of the target was not associ-
ated with a single configuration of distractors.
Procedure. Participants responded on the identity of the even
number (2, 4, 6, 8) by pressing the keys V,B,N, and M, respec-
tively, with the middle and index fingers of each hand. Training
consisted of 1152 trials, divided into 24 blocks of 48 trials. Each
block comprised three series of 16 trials, which presented the
target once on each of the 16 possible locations over the matrix.
Participants assigned to the Discrete condition were presented with
a fixation point (i.e., the letter xcentered over the display) during
500 ms at the beginning of each trial, followed by the presentation
of the matrix until response. A high or low tone of 250 ms
provided positive or negative feedback on performance, and the
next trial came after 1000 ms. In the Continuous condition, the RSI
was reduced to 200 ms, and feedback was presented only on errors
(a high 100-ms tone).
Learning measures were indirectly taken from performance in
responding to repeated- versus variable-context trials. If partici-
pants learned about the information provided by the context with
respect to the target location, we expected that participants would
respond progressively faster to repeated-context trials than to
variable-context trials. If learning was implicit, we expected this
learning to arise even when participants were not able to guess the
location of the target over a direct task in which the same contexts
were presented, but in which the target was replaced by another
distractor (Chun & Jiang, 2003; cf. Smyth & Shanks, 2008).
According to this rationale, a direct measure was arranged at the
end of training, in which participants were instructed to guess the
location of the target in a series of 32 trials comprising two
presentations of each of the eight repeated-context displays and of
eight displays selected from the variable contexts. In all cases, the
target was replaced by another distractor. Rather than guessing the
specific location of the removed target, participants were asked
only to guess the quadrant at which the target would have appeared
in that particular context. Responses were issued by pressing keys
4, 5, 1, or 2 from the numeric keypad to indicate upper-left,
upper-right, lower- left, and lower-right quadrants, respectively.
To assess whether this measure showed any conscious knowledge
about the contextual regularities, we compared the proportion of
correct responses produced respectively for repeated- and variable-
context trials.
Results
For all experiments, the alpha level was set at .05. We report on
partial eta squared (
p
2
) as an index of effect size, as well as
Greenhouse–Geisser epsilon and adjusted pvalues whenever it is
relevant, but nominal degrees of freedom are maintained. Reaction
times (RTs) corresponding to the first trial of each block, as well
as error responses, were excluded from the analyses. Accuracy
data were analyzed in order to discard any of the reported effects
Figure 1. Illustration of a trial corresponding to Experiment 1. Upper and
lower black circles (not shown to participants) indicate target and correct
response, respectively.
225
SEQUENCE LEARNING AND CONTEXTUAL CUEING
that could result from a trade-off between speed and accuracy.
Because there was no evidence of this kind of trade-off, we report
only on the overall accuracy levels, and concentrate the discussion
on the effects observed on RTs.
Training. The percentage of errors in the Discrete and Con-
tinuous conditions was 3.9 and 3.6, respectively, and they did not
differ between groups, t(46) .43, p.66. Analyses of practice
were conducted by collapsing data into successive epochs of two
blocks (i.e., 96 trials, or six repetitions of each repeated context).
A mixed-model analysis of variance (ANOVA) was conducted on
RTs, with condition (Discrete vs. Continuous) as a between-
participants factor, and with practice (Epochs 1–12) and context
(repeated vs. variable) as repeated measures. The analysis showed
a significant effect of condition, F(1, 46) 9.84,
p
2
.18, p
.01, indicating that performance was faster when participants had
time to prepare in advance for each trial (765 vs. 858 ms). The
effect of practice showed that RTs improved with training, F(11,
506) 137.84,
p
2
.75, ε.51, p.0001, and the significant
effect of context confirmed the production of a contextual cueing
effect, F(1, 46) 116.96,
p
2
.72, p.0001. Significant
interactions Condition Practice, F(11, 506) 2.62,
p
2
.05,
ε.51, p.05, and Practice Context, F(11, 506) 3.01,
p
2
.06, ε.76, p.01, showed, respectively, that the Continuous
condition improved more with practice, and that the effect of
context grew significantly with training (see Figure 2). There was
no significant interaction involving condition and context simul-
taneously, thus suggesting that the effect of contextual cueing did
not differ between groups.
Direct measure of contextual cueing (location guessing task).
An ANOVA conducted over the percentage of hits, with condition
(Discrete vs. Continuous) and context (repeated vs. variable) as
independent factors did not produce any significant effect or
interaction (all Fs1). Thus, participants guessed all locations at
random, and the proportion of hits did not differ from that expected
by chance (.25) either on repeated-context (.24) or on variable-
context (.25) trials.
Discussion
The results of Experiment 1 showed that the effect of contextual
cueing is robust, in that it resisted the procedural changes intro-
duced in this new version of the paradigm. Importantly, this
phenomenon has often been assessed under conditions in which
the allocation of attention is heavily controlled before the start of
each trial, presenting a fixation point and using a large interval
between successive trials. Our finding that analogous effects can
be obtained in a continuous version of this task, even when there
is no explicit realignment of attention between trials, lends support
to the argument that this experimental phenomenon may affect the
dynamics of attention in natural domains in which people have to
process and respond to a series of regular contexts. In many of
these domains, both synchronic and diachronic information (i.e.,
the context of each trial plus the sequence of successive trials)
could be helpful for learners to prepare and to react optimally to a
continuous flow of structured events. In the following experiments
we set up a design to investigate the interaction between contextual
cueing and sequence learning effects by merging these two para-
digms.
Experiment 2: Contextual Cueing and Second-Order
Sequence Learning
Method
Participants. Ninety students from the same pool partici-
pated in the experiment. Thirty participants were randomly as-
signed to each condition: In the Cueing condition only the con-
textual contingency was presented; in the Sequence condition a
SOC structure was included in the series of responses, but there
was no relation between contexts and targets; finally, in the Cueing
plus Sequence group both contingencies were included at the same
time. Unless otherwise stated, all features of the design followed
those described for the Continuous condition from Experiment 1.
Procedure. The Cueing condition constituted a close repli-
cation of the Continuous condition from Experiment 1. In contrast,
participants in both the Sequence and the Cueing plus Sequence
conditions were presented with a series of structured targets that
called for a repeated sequence of responses. This sequence of
targets followed a 12-element, SOC sequence in which all indi-
vidual items and all first-order transitions were equally probable,
but in which the correct successor of a sequence could be predicted
by considering two previous trials. Specifically, the series we used
as training and control sequences were 2-4-2-8-6-4-8-2-6-8-4-6
and 6-4-6-8-2-4-8-6-2-8-4-2. As can be observed, both sequences
are structurally analogous and are related to each other by a simple
replacement of Elements 2 and 6. However, they are maximally
discriminative in that the successor of any two-event series is
always different between training and control sequences.
Training was structured in 12 blocks of 96 trials. Each training
block (i.e., Blocks 1 to 9, plus Block 12) contained eight repeti-
tions of the training sequence. Blocks 10 and 11 were arranged as
test blocks, in which only 50% of the trials were generated ac-
cording to the training sequence, whereas the remaining trials
followed the control sequence. Thus, over the test blocks each
individual trial was generated according to the two previous trials,
by following at chance either the prescriptions of the control
650
750
850
950
1050
123456789101112 123456789101112
suounitnocetercsid
experiment 1: contextual cueing (RT)
repeated
varia ble
Figure 2. Mean reaction times (RTs) for repeated- and variable-context
trials in Discrete and Continuous conditions throughout training (Blocks
1–12) in Experiment 1. Vertical bars represent standard errors.
226 JIME
´NEZ AND VA
´ZQUEZ
sequence or those of the training sequence. In other words, after
each fragment of two trials (e.g., 8-2) the successor was randomly
assigned to be that following this fragment according to the train-
ing (i.e., 8-2-6), or the control sequence (i.e., 8-2-4). During Block
12 the training sequence was restored, right before introducing the
direct measures of learning.
To indirectly assess sequence learning, we compared the aver-
age RTs in response to control trials over the test Blocks 10 and 11
with the average RTs in response to training trials over the neigh-
boring Blocks 9 and 12. Contextual cueing was assessed in the
Cueing and Cueing plus Sequence conditions over the training
phase (Blocks 1 to 9) as was described for Experiment 1. However,
in the Cueing plus Sequence condition we separately assessed the
effect of contextual cueing over the test blocks, to ascertain
whether changing the sequence affected the expression of contex-
tual cueing.
After training, direct measures of contextual cueing and se-
quence learning were presented in the appropriate groups. In the
Cueing plus Sequence condition, the order of these measures was
counterbalanced across participants. The location-guessing task
was exactly as that described for Experiment 1. As for the se-
quence learning, we included a cued generation task comprising 24
tests. Each test consisted of three trials, the first two of which were
analogous to those presented over training, whereas the third one
was a prediction trial. Over the first two trials of each test,
participants were presented with 1 of the 12 possible two-trial
fragments that could appear over training, and were asked to
respond to them as they did over the standard task. On the third
trial, the target was replaced by another distractor, and participants
were asked to guess the identity of the removed target by relying
on what they saw on the two previous trials. Each of the 12
possible two-trial fragments was presented twice, for a total of 24
generation tests. To evaluate the expression of sequence learning
over this generation task, we compared the proportion of predic-
tion trials in which participants generated the successor corre-
sponding to either the training sequence or the control sequence.
During this task, accuracy was emphasized over RTs, and no
feedback was provided.
Results
Percentage of errors over training on Cueing, Sequence, and
Cueing plus Sequence conditions were 2.8, 3.1, and 3.3, respec-
tively. These values did not differ significantly among groups, F(2,
87) .76,
p
2
.02, p.46.
Contextual cueing. A repeated-measures ANOVA over RTs
with condition (Cueing vs. Cueing plus Sequence) as a between-
participants factor, and with practice (9) and context (repeated vs.
variable) as repeated measures showed effects of practice, F(8,
464) 70.66,
p
2
.55, ε.57, p.0001, and context, F(1,
58) 77.95,
p
2
.57, p.0001, but no overall differences
between conditions ( p.26). A significant Practice Context
interaction, F(8, 464) 3.57,
p
2
.06, ε.87, p.001,
indicated that the contextual cueing effect arose with practice. The
Condition Practice interaction just missed significance, F(8,
464) 2.28,
p
2
.04, ε.57, p.052, but it suggested that the
existence of a sequence in the Cueing plus Sequence condition
tended to produce larger improvements with practice than those
observed in the Cueing condition. No other interaction involving
condition approached significance ( ps.18). Moreover, and
against the hypothesis that both learning effects compete for cog-
nitive resources, the overall effect of context was larger in absolute
terms for the Cueing plus Sequence condition than for the Cueing
group (45 vs. 33 ms; see Figure 3, top panel). A further ANOVA
conducted over the test blocks showed that the effect of context did
not change in the Cueing plus Sequence condition when the
validity of the sequence was reduced. The overall effect of
context was still significant, F(1, 58) 114.25,
p
2
.66, p
.0001, and there was no significant Condition Context inter-
action ( p.20).
Sequence learning. The effect was assessed in both the
Sequence and Cueing plus Sequence conditions, by comparing RT
in response to the control trials over the test blocks, with RT in
response to the training trials over their neighboring blocks. An
ANOVA with condition (Sequence vs. Cueing plus Sequence) as
a between-participants factor and with sequence (training vs. con-
trol) as a repeated measure, showed a significant effect of se-
700
800
900
1000
1100
123456789t12 123456789 t12
gnieucgnieuc plus sequence
experiment 2: contextual cueing (RT)
repeated
vari able
700
800
900
1000
1100
123456789t12 123456789t12
sequence cueing plus s equence
experi ment 2: se quence l earning (RT)
control sequence
trai ning seque nce
Figure 3. Mean reaction times (RTs) for repeated- and variable-context
trials in Cueing and Cueing plus Sequence conditions (top panel), and for
training and control sequence trials in Sequence and Cueing plus Sequence
conditions (bottom panel) throughout training (Blocks 1–12) in Experiment
2. Results for the test (t) Blocks 10 and 11 are collapsed. Vertical bars
represent standard errors.
227
SEQUENCE LEARNING AND CONTEXTUAL CUEING
quence, F(1, 58) 12.33,
p
2
.18, p.001, but not an effect of
condition, nor a Condition Sequence interaction (Fs1; see
Figure 3, bottom panel), indicating that both groups learned the
sequence in a similar way.
Contextual cueing and sequence learning. To further assess
the effect of one contingency over the other in the Cueing plus
Sequence condition, we compared the effect of contextual cueing
over the test blocks with that observed over the neighboring
Blocks 9 and 12. The ANOVA with sequence validity (low vs.
high) and context (repeated vs. variable) as within-participant
factors showed significant effects of sequence validity, F(1, 29)
4.45,
p
2
.13, p.05, and context, F(1, 29) 78.15,
p
2
.73,
p.0001, but no interaction between them (F1). Finally, an
ANOVA conducted specifically over the test blocks to assess
whether the effect of contextual cueing was affected by the se-
quential status of each trial showed no interaction between context
and sequence (F1). The effect of context was clearly significant
in this analysis, F(1, 29) 69.30,
p
2
.70, p.0001. Somewhat
surprisingly, however, no significant expression of sequence learn-
ing was observed specifically within these test blocks, F(1, 29)
1.88,
p
2
.06, p.18.
Direct measure of contextual cueing (location guessing task).
An ANOVA conducted over the percentage of correct guesses
with condition (Cueing vs. Cueing plus Sequence) as a between-
participants factor and context (repeated vs. variable) as a repeated
measure did not show any significant effect or interaction. Indeed,
the proportion of correct guesses for the repeated trials (.24) was
not different from that obtained for variable trials (.22), and it did
not exceed that expected by chance (.25). This strongly indicates
that context knowledge was implicit.
Direct measure of sequence learning (cued-generation task).
Participants in this task can generate successors in accordance with
either the training sequences or the control sequences, but they
could also produce responses inconsistent with both of them. Both
groups of participants tended to generate the training successors
slightly more often than they did the control successors (.34 vs. .29
in the Sequence group, and .40 vs. .37 in the Cueing plus Sequence
group). An ANOVA comparing these proportions of generation
responses between conditions showed a significant effect of con-
dition, F(1, 58) 6.64,
p
2
.10, p.05, indicating that
participants in the Cueing plus Sequence condition generated
responses consistent with either the control or the training se-
quence more often than did participants in the Sequence group.
More important, neither the effect of sequence, F(1, 58) 2.56,
p
2
.04, p.11, nor the Condition Sequence interaction, F(1,
58).17,
p
2
.00, p.67, reached significance in this analysis.
Thus, although there was a trend to generate the training succes-
sors of each fragment more often than their control counterparts,
this effect was not statistically reliable, thus suggesting that the
effects of sequence learning obtained in the indirect measures were
not attributable to explicit sequence knowledge.
Discussion
The results of this experiment show that contextual cueing and
sequence learning effects can proceed simultaneously without cost
and that both of them are implicit, in the sense that they could be
expressed through the indirect measures of performance, even
when they do not produce significant effects in comparable direct
measures. This is especially clear for the location guessing task,
which directly assessed participants’ ability to infer the location of
a removed target by relying on a repeated context (Chun & Jiang,
2003; Smyth & Shanks, 2008). A similar, but somewhat less clear
pattern was observed in the sequence learning task. On the one
hand, the indirect measures taken during training did clearly indi-
cate that sequence learning took place, even though its expression
(i.e., the difference between responses to training and control
trials) disappeared within the test blocks. This tendency of the
effects of learning to disappear upon a sudden decrease in the
proportion of predictable trials has been previously found, and it
has been associated with participants’ becoming aware of the
change introduced over the test blocks (Jime´nez, Vaquero, &
Lupia´n˜ez, 2006; Jime´nez&Va´zquez, 2008). On the other hand,
the results from the cued-generation task do not support the con-
clusion that sequence knowledge became explicit in this experi-
ment, because participants were not able to generate the training
successors significantly more often than they generated the control
successors when they were directly instructed to do so. Alterna-
tively, we surmise that the low reliance on the training sequence
observed over the test blocks could be attributed to the trial-by-
trial substitution procedure adopted in this experiment to introduce
control trials (cf. Schvaneveldt & Gomez, 1998). Whereas in
previous studies (e.g., Jime´nez et al., 2006; Jime´nez&Va´zquez,
2008) we adopted a series substitution strategy, in which complete
12-trial series of the training or control sequence were mixed with
each other over the test blocks, in the present experiment each
individual trial was arranged to follow either the control or the
training structure. As has been recently demonstrated by Jime´nez,
Lupia´n˜ez, and Vaquero (2009), a trial-by-trial substitution proce-
dure could result in some automatic sequential congruency effects,
which would reduce the expression of sequence learning right after
an incongruent (i.e., a control) trial in a way similar to that found
in other congruency tasks (see Egner, 2007, for a review of those
sequential congruency effects). In any case, given that these results
are not completely conclusive on whether sequence learning was
implicit, and therefore on whether the observed lack of competi-
tion between sequence learning and contextual cueing could be
obtained regardless of the explicit versus implicit nature of this
learning, two additional experiments were designed to address this
issue. In Experiment 3A and 3B, we arranged either a simpler
deterministic sequence or a more complex probabilistic sequence,
and we assessed the impact of the resulting knowledge on the
acquisition and expression of contextual cueing.
Experiments 3A and 3B: Contextual Cueing and
Deterministic Versus Probabilistic Sequence Learning
Method
Participants. Forty-eight students from the same population
participated in Experiment 3A (deterministic sequence) and were
randomly assigned to one of two conditions. Half of them were
assigned to the Sequence condition and were presented with a
deterministic hybrid sequence, which we expected to result in
more explicit sequence knowledge. The other 24 participants were
presented with the same sequence together with a contextual
cueing preparation (Cueing plus Sequence condition), so as to
assess the interaction between contextual cueing and such explicit
228 JIME
´NEZ AND VA
´ZQUEZ
sequence learning. Another pool of 48 students took part in Ex-
periment 3B (probabilistic sequence). Participants in this experi-
ment were also assigned to either Sequence or Cueing plus Se-
quence conditions, but now the training sequence was respected in
only 75% of the training trials, whereas it was replaced by a
control sequence in the remaining 25% of the trials.
Procedure. The procedure for Experiments 3A and 3B fol-
lowed the lines described for the corresponding conditions from
Experiment 2, with the exceptions described below. First, the
series of target digits were generated according to a simpler hybrid
sequence (Cohen, Ivry, & Keele, 1990). The training sequence was
composed by eight elements, comprising two presentations of each
digit. To control for the possibility that certain transitions could be
easier to learn or to respond to, we developed four versions of the
training sequence and presented each version to 6 participants
within each group. Table 1 presents these four training sequences,
together with their corresponding control sequences. As can be
observed, all sequences are structurally analogous. They contain
two repetitions of a single unique transition, which allows partic-
ipants to predict the next event by relying on a single previous
target (e.g., 2-6 in the first training sequence, 2-4 in its correspond-
ing control). The two repetitions of this unique transition are
separated by two intervening events, which give place to six
ambiguous transitions in which a single event predicts a different
successor on each of its two occurrences (e.g., 6-4 and 6-8 in the
first training sequence, 6-2 and 6-8 in its corresponding control).
No immediate repetitions (e.g., 6-6) or reversals (e.g., 6-2-6) are
allowed, and each control sequence was generated by replacing
specific digits from the training structure, so that the two occur-
rences of the unique transition and half of the ambiguous transi-
tions were modified.
In both experiments, the training phase comprised 12 blocks,
with a test at Block 11. In Experiment 3A (deterministic training),
a training block was made up of 96 trials, featuring 12 repetitions
of the training sequence. The test Block 11 contained six repeti-
tions of the training sequence, randomly interspersed with six
repetitions of the control sequence. In Experiment 3B (probabilis-
tic training) each training block contained nine repetitions of the
trained sequence randomly interspersed with three repetitions of
the control sequence. The test Block 11 was analogous in both
experiments. Thus, in both of them we adopted a series substitu-
tion strategy, using complete series of eight trials conforming to
the relevant sequence, rather than relying on the trial-by-trial
substitution procedure described for Experiment 2. To make sure
that connections between training and control sequences could be
made in accordance with the second-order transitions stipulated by
the upcoming sequence, the starting point of each sequence at the
beginning of each block was not selected completely at random,
but only among the three transitions that were shared between each
pair of control and training sequences. Thus, for instance, the
Training Sequence 1 could start with the successor of the transition
8-2 (i.e., with digit 6) because in this way the sequence could end
up precisely with the fragment 8-2, which was also legal according
to the corresponding control sequence. In that case, the first trial of
the control sequence would be the successor of this legal path, thus
respecting the second-order conditionals stipulated by the upcom-
ing sequence.
To assess the amount of explicit knowledge, we included a
generation task, together with a location-guessing task presented to
those groups in which the context was informative. In the gener-
ation task, participants were presented twice with each of the 11
two-trial fragments that could appear over either the training or the
control sequences experienced by each participant. After reacting
to these two trials, they were required to predict the most likely
successor of that fragment. Generation responses were scored as
consistent with the training sequence, consistent with the control
sequence, or inconsistent with both.
Results from Experiment 3A (Deterministic Sequences)
Percentages of errors in the Sequence and Cueing plus Sequence
conditions were 3.1 and 2.5, respectively. Both groups performed
the task with a similar level of accuracy, t(46) 1.16, p.24.
Sequence learning. An ANOVA conducted over RTs with
condition (Sequence vs. Cueing plus Sequence) as a between-
participants factor, and practice (Blocks 1–10) as a repeated mea-
sure showed only a significant effect of practice, F(9, 414)
81.27,
p
2
.64, ε.64, p.0001. To assess sequence learning,
we compared performance over the control trials during the test
block with performance over the adjacent training blocks. An
ANOVA with condition (Sequence vs. Cueing plus Sequence) and
sequence (training Blocks 10/12 vs. control trials from test block)
showed significant effects of sequence, F(1, 46) 88.23,
p
2
.66, p.0001. As is shown in Figure 4 (top panel), the effect was
larger than that usually found in other sequence learning tasks, thus
suggesting that the acquired knowledge was actually explicit and
that consciously anticipating the next event produced strong ben-
efits in this task. Neither the effect of condition nor the Condi-
tion Sequence interaction approached significance ( ps.15).
Generation performance confirmed that sequence knowledge
was explicit, by showing that participants produced successors
consistent with the training sequence significantly more often than
those consistent with the control sequence (.53 vs. .36), F(1, 46)
19.84,
p
2
.30, p.0001. Neither the effect of condition nor
the Condition Sequence interaction approached significance
(ps.70).
Contextual cueing. To assess whether the acquisition or
expression of contextual cueing was affected by the presence of
explicit sequence learning in the Cueing plus Sequence condition,
we conducted an ANOVA over RTs with practice (Blocks 1–10)
and context (repeated vs. variable) as repeated measures. Results
showed significant effects of practice, F(9, 207) 37.13,
p
2
.62, ε.54, p.0001, and context, F(1, 23) 9.20,
p
2
.29,
p.01. Figure 4 (middle panel) shows that there was a small but
significant effect of contextual cueing, which contrary to that
observed in previous experiments, did not grow with practice ( p
.50). Interestingly, the effect seemed to disappear by the end of
Table 1
Pairs of Training and Control Sequences Counterbalanced
Across Participants in Experiments 3a and 3b
Training Control
Sequence 1 2–6-4–8-2–6-8–4 2–4-6–8-2–4-8–6
Sequence 2 4–8-2–6-4–8-6–2 4–2-6–8-4–2-8–6
Sequence 3 2–4-6–8-2–4-8–6 2–6-4–8-2–6-8–4
Sequence 4 4–2-6–8-4–2-8–6 4–8-2–6-4–8-6–2
229
SEQUENCE LEARNING AND CONTEXTUAL CUEING
training, but it reappeared strongly over the test, when the validity
of the sequence decreased suddenly. Within the test block, a strong
effect of 68 ms was observed, F(1, 23) 25.23,
p
2
.53, p.01.
To assess whether explicitly following the sequence interfered
with the expression of contextual cueing, we compared these
effects between the test and the adjacent blocks. An ANOVA with
type of block (training vs. test) and context (repeated vs. variable)
showed significant main effects of type of block, F(1, 23) 44.33,
p
2
.66, p.0001, and context, F(1, 23) 23.13,
p
2
.50, p
.0001, as well as a significant interaction between them, F(1,
23) 10.30,
p
2
.31, p.01. Thus, the effect of contextual
cueing was observed selectively over the test block, when the
validity of the sequence decreased. A repeated-measures ANOVA
conducted within the test block to assess the interaction between
450
550
650
750
850
950
1050
123456789101112 123456789101112
gnieucecneuqes plus seque nce
experiment 3a: sequence learning (RT)
control sequence
training sequence
450
550
650
750
850
950
1050
123456789101112
cueing plus sequence
experiment 3a: contextual cueing (RT)
repeated
variable
650
750
850
950
repeated variable
control sequence
training sequence
experiment 3a: contextual cueing and sequence learning during test (RT)
Figure 4. Mean reaction times (RTs) for training and control sequence trials in Sequence and Cueing plus
Sequence conditions (top panel) and for repeated- and variable-context trials in the Cueing plus Sequence
condition (middle panel) during training (Blocks 1–12) in Experiment 3a. Bottom panel represents mean RTs for
training versus control sequence trials, respectively, for repeated versus variable contexts over the test phase
(Block 11) for the Cueing plus Sequence condition from Experiment 3a. Vertical bars represent standard errors.
230 JIME
´NEZ AND VA
´ZQUEZ
context and sequence learning showed significant effects of con-
text, F(1, 23) 21.02,
p
2
.48, p.0001, and sequence, F(1,
23) 22.75,
p
2
.50, p.0001, but not a significant interaction
between them ( p.30; see Figure 4, bottom panel).
Finally, the direct measure of contextual cueing obtained in the
Cueing plus Sequence condition did not show location guessing
levels above those expected by chance (the proportion of hits were
.23 and .24, respectively, for repeated and variable trials). This
again indicated that contextual learning remained implicit.
Results of Experiment 3B
Participants in both Sequence and Cueing plus Sequence con-
ditions from Experiment 3B produced only a small percentage of
errors (3.6 and 3.1, respectively). These two levels of accuracy
were not significantly different from each other, t(46) .85,
p.39.
Sequence learning. Because in this experiment the sequence
was probabilistic, it was possible to assess sequence learning
on-line over the whole training period. An ANOVA on RTs with
condition (Sequence vs. Cueing plus Sequence) as a between-
participants factor, and with practice (Blocks 1–10) and sequence
(training vs. control) as repeated measures showed significant
effects of practice, F(9, 414) 61.79, ε.64.
p
2
.57, p
.0001, and sequence, F(1, 46) 19.52,
p
2
.30, p.0001, but
only a nonsignificant trend for the Practice Sequence interac-
tion, F(9, 414) 1.79,
p
2
.04, ε.75, p.09 (see Figure 5,
top panel). No effects or interactions involving condition ap-
proached significance in this analysis.
In keeping with the practice adopted in previous experiments,
our main conclusions concerning sequence learning relied on the
comparison between responding to control trials over the test block
and responding to the sequence trials over the adjacent training
blocks. The ANOVA comparing these two scores with condition
as a between-participants factor showed a significant effect of
sequence, F(1, 46) 25.91,
p
2
.36, p.0001, but neither the
effect of condition nor the Condition Sequence interaction
approached significance ( ps.50).
The generation results indicated that sequence knowledge was
less explicit in this probabilistic setting than it was in Experiment
3A. Although in absolute terms, participants tended to generate the
training successor of each fragment somewhat more often than
they generated its control counterpart (.47 vs. .42), an ANOVA
conducted over these proportions of generation responses, with
condition as a between-participants factor, did not reach significant
effects of sequence, F(1, 46) 2.89,
p
2
.06, p.09. Neither
the effect of condition nor the Sequence Condition interaction
approached significance in this analysis ( ps.50).
Contextual cueing. A repeated-measures ANOVA con-
ducted on the RTs for participants in the Cueing plus Sequence
condition, with practice (Blocks 1–10) and context (repeated vs.
variable) as independent factors showed significant effects of
practice, F(9, 207) 36.44,
p
2
.61, ε.47, p.0001, and
context, F(1, 23) 40.19,
p
2
.64, p.0001, indicating both an
improvement with training and a contextual cueing effect (see
Figure 5, bottom panel). The Practice Context interaction was
not significant ( p.25). When the validity of the sequence was
reduced over the test Block 11, the effect of context remained
present, F(1, 23) 21.11,
p
2
.48, p.0001.
Contextual cueing and sequence learning. The interaction
between both contingencies was assessed over training in the
Cueing plus Sequence condition by means of an ANOVA over
RTs, considering practice (Blocks 1–10), context (repeated vs.
variable), and sequence (training vs. control) as repeated measures.
Results showed significant main effects of practice, F(9, 207)
30.09,
2
.57, ε.50, p.0001, context, F(1, 23) 25.91,
p
2
.53, p.0001, and sequence, F(1, 23) 10.23,
p
2
.31, p
.01, as well as a two-way significant interaction of Practice
Sequence, F(9, 207) 2.31,
p
2
.09, ε.64, p.05. No other
interaction reached significant levels.
To assess whether this lack of interaction between learning
effects could be extended over the test block, and whether both
learning effects remained unchanged despite the decrease in se-
quence validity, we conducted an ANOVA considering type of
block (training 10/12 vs. test 11), context (repeated vs. variable),
and sequence (training vs. control). The results showed significant
main effects of type of block, F(1, 23) 12.27,
p
2
.35, p
.001, context, F(1, 23) 41.66,
p
2
.64, p.0001, and
sequence, F(1, 23) 6.36,
2
p
.22, p.05, but no significant
interactions. Thus, even though performance got slowed by the
decrease in sequence validity, the main effects of learning re-
mained essentially unchanged. Finally, the direct measure of con-
textual cueing suggested that this learning was implicit, because
600
700
800
900
1000
1100
123456789101112 123456789101112
sequence cueing plus se quence
experiment 3b: sequence learning (RT)
control sequence
training sequence
600
700
800
900
1000
1100
123456789101112
cueing plus sequenc e
experiment 3b: contextual cueing (RT)
repeated
variable
Figure 5. Mean reaction times (RTs) for training and control sequence
trials in Sequence and Cueing plus Sequence conditions (top panel), and for
repeated- and variable-context trials in the Cueing plus Sequence condition
(bottom panel) over training (Blocks 1–12) in Experiment 3b. Block 11
corresponds to test phase. Vertical bars represent standard errors.
231
SEQUENCE LEARNING AND CONTEXTUAL CUEING
there were no significant differences between the proportion of
correct guesses for the repeated (.27) and variable (.24) trials,
t(23) .83, p.41.
Discussion of Experiments 3A and 3B
Together, these experiments indicate that the contextual cueing
effect may be acquired and expressed simultaneously with the
implicit sequence learning produced by training participants with a
probabilistic sequence (Experiment 3B), but that learning about a
simpler, deterministic sequence does somehow interfere with the
expression of contextual cueing (Experiment 3A). The average
effects of contextual cueing over the whole training period were 33
and 18 ms, respectively, for each of these two experiments, cor-
responding to partial
2
of .64 and .29. An ANOVA comparing
these two effects did not reach significant differences between
them, although the Context Experiment interaction indicated a
trend in that direction, F(1, 46) 3.41,
p
2
.07, p.07. By
inspecting the corresponding patterns in Figure 4 (middle panel)
and Figure 5 (bottom panel), one can see that the difference in the
context effect arises mainly over the second half of training, when
participants in Experiment 3A could be increasingly relying on the
sequence. An ANOVA restricted to Blocks 6 to 10 confirmed that
the Context Experiment interaction was significant at this point,
F(1, 46) 9.13,
p
2
.17, p.005.
The simplest interpretation of the difference in the expression of
contextual cueing between Experiments 3A and 3B does rely on
the assumption that participants trained with a short, deterministic
sequence learned explicitly about it, and thus either stopped
searching through the display, or at least restricted their processing
efforts to a mere confirmatory check for the presence of the
expected target. According to this interpretation, it is worth noting
that RTs over training became significantly faster in Experiment
3A, as was indicated by the Practice Experiment interaction,
F(9, 38) 3.22,
p
2
.43, p.01, and that the direct measure
also showed explicit sequence knowledge only in Experiment 3A.
An ANOVA comparing the generation results over the two exper-
iments confirmed that the effect of sequence was significantly
larger in Experiment 3A than that in Experiment 3B, F(1, 92)
5.00,
p
2
.05, p.05.
The interaction between sequence learning and contextual cue-
ing thus appears to be restricted to conditions in which sequence
knowledge becomes explicit. However, it is important to note that
even though explicit sequence learning interfered with the expres-
sion of contextual cueing, the acquisition of context information
was not affected by that manipulation. Indeed, the effect of con-
textual cueing was comparable between experiments when they
were assessed over the test block, in which the learned sequence
became unreliable. Actually, over this test block participants in
Experiment 3A showed an effect of contextual cueing that was
larger in absolute terms than that found in Experiment 3B (68 vs.
51 ms). Thus, explicit sequence knowledge may have affected the
expression of contextual cueing, but not the amount of context
knowledge acquired by the end of this period.
General Discussion
In three experiments, we investigated whether contextual cueing
could affect performance in conditions in which successive trials
proceeded continuously without pauses, and in which both diach-
ronic (i.e., sequence) and synchronic (i.e., context) information
were useful to improve responding to a stream of events. In
Experiment 1, we confirmed that contextual cueing effects arose
similarly with discrete trials separated by a large response-to-
stimulus interval (RSI) and a fixation point, and in continuous
conditions in which successive trials appeared without transition.
In Experiment 2, we showed that it is possible to learn simulta-
neously about this context information, and about a relatively
complex sequence of targets and responses, thus suggesting that
these two learning processes do not compete for a limited pool of
central cognitive resources. Finally, Experiments 3A and 3B, de-
signed to ascertain the differential effects of explicit versus im-
plicit sequence learning on the acquisition and expression of
contextual cueing, confirmed that implicit learning about a prob-
abilistic sequence does not interfere with the effects of contextual
cueing, whereas explicit learning about a simpler, deterministic
structure does affect the expression but not the acquisition of
context knowledge. In the following paragraphs, we discuss the
main conclusions to be drawn from this pattern of results, with
special emphasis on the processes underlying contextual cueing,
and on the automatic nature of these learning effects.
Automatic Nature of Implicit Learning
The main conclusion to be drawn from Experiments 2 and 3B is
that sequence learning and contextual cueing can proceed simul-
taneously without cost, as long as both of them remain implicit.
We made sure that sequence learning processes were implicit by
using either complex (i.e., second order) or probabilistic contin-
gencies, and we assessed their explicit results by including direct
measures of the relevant knowledge by the end of the experiments.
In accordance with most previous research on contextual cueing
(with the notable exception of that of Smyth & Shanks, 2008), our
measures of location guessing did not provide any evidence that
participants could guess the location of a removed target more
often when they were presented with a repeated context than when
they responded to a variable context. This was so despite the fact
that, following Smyth and Shanks’s suggestions, we arranged two
presentations of each context, instead of a single one, so as to
improve the sensitivity of the location-guessing task.
As for the direct measure of sequence learning, we relied upon
a cued-generation task that showed, as was expected, that partic-
ipants acquired some explicit sequence knowledge when the se-
quence was simple and deterministic (Experiment 3A), but not
when it was probabilistic (Experiment 3B), or when it involved
only second-order conditionals (Experiment 2).
The main hypothesis arising from the assumption that these two
learning processes result automatically from performing the ori-
enting task was that we should not find competition between them
for a limited pool of central processing resources. In line with this
hypothesis, we found that both learning effects proceeded without
mutual interference when they remain implicit (i.e., Experiment 2
and Experiment 3B). Importantly, however, results from Experi-
ment 3A showed that when the sequence information became
explicit, it hindered the expression of contextual cueing, even
though the acquisition appeared not to be affected, as was shown
by the observation of comparable levels of context knowledge over
the test block, in which the sequence was no longer reliable.
232 JIME
´NEZ AND VA
´ZQUEZ
Overall, these results are compatible with the conclusion that
learning about informative contexts does not call for the same
cognitive resources needed to learn about a sequence of targets,
neither when the participants become aware of the sequence nor
when this sequence knowledge remains implicit.
Selective Attention and Contextual Cueing
In Experiment 3A, participants learned about informative con-
texts even when they were responding to a completely predictable
sequence of targets. Under these conditions, the search task could
become simplified to the point of allowing participants to perform
the task without paying much attention to the context, by simply
verifying the presence of the anticipated target. This suggests that
selective attention was not necessary for learning about the con-
text. In this sense, the results could be taken as consistent with
those reported by Jiang and Leung (2005), who showed that
participants learned about informative contexts even without se-
lective attention to these contexts, but that their effects were
expressed only when the conditions changed so that participants
could no longer ignore these informative contexts.
Alternatively, one might argue that knowledge about repeated
contexts could have been acquired early in training, when full
selective attention was still devoted to the search task, and that this
context learning could have reached the asymptote before the
sequence knowledge became explicit. This argument is consistent
with some intriguing results showing that contextual cueing effects
appear to rely heavily on the first experiences with a search task,
to the point of being absent when participants are exposed to a
number of noninformative patterns before training with structured
configurations (Junge´ et al., 2007). Although this early- learning
account remains a possibility that calls for a systematic analysis, at
this point it seems premature to accept that the first experiences
with a search task could have such a disproportionate, and ulti-
mately nonadaptive, impact on performance.
Expression Deficit
Results from Experiments 3A and 3B showed that context
knowledge can be acquired to a similar extent regardless of the
presence of a simultaneous process of explicit sequence learning,
but also that exploiting a reliable sequence does hinder the expres-
sion of contextual cueing. We have proposed that such expression
deficit could be explained in terms of the reduced amount of
selective attention deployed to the informative contexts, but there
remain at least two alternative accounts for this expression deficit.
First, one might argue that the absence of contextual cueing could
not depend as much on the reduction in selective attention caused
by explicit sequence learning, but rather on the size of this se-
quence learning effect, which might simply mask the expression of
cueing. In other words, it is possible that contextual cueing could
only arise above a certain “floor level” in performance, and that
explicit sequence learning could have led response latencies below
this level. Alternatively, although the acquisition of context infor-
mation does not appear to depend on a deliberate decision to
encode that context, it is in principle possible that deliberately
trying to anticipate the next target could interfere directly with the
retrieval of relevant information about repeated contexts. We sur-
mise that an account based on the reduction of selective attention
is more likely than any of the two alternatives described, but
further research is needed to assess them.
Processes Underlying Learning
Finally, the absence of interference showed in this study be-
tween implicit sequence learning and contextual cueing also bears
some more general implications with respect to the nature of the
underlying processes. There are admittedly many possible ways of
thinking about the basis for dual-task interference. To end this
article, we would like to briefly refer to three of these perspectives,
which can be built in terms of either the neural or the functional
bases of these processes. First, in terms of the neural networks
involved in these two processes, one might speculate that an
independent course for these two effects could be expected as far
as each of them is subserved by the action of nonoverlapping
networks. Contextual cueing, on the one hand, has been related to
the activation of the hippocampus and associated medial temporal
structures, which are involved in the encoding of complex epi-
sodes, and in exploiting their lawful configurations regardless of
awareness (Chun, 2000; Chun & Phelps, 1999; Greene et al.,
2007). On the other hand, implicit sequence learning has been
identified with the action of a procedural network, involving
mainly the supplementary motor area, basal ganglia, and some loci
within the parietal cortex (Hazeltine & Ivry, 2003; Keele, Ivry,
Mayr, Hazeltine, & Heuer, 2003). Thus, just as these two brain
networks involve mostly nonoverlapping structures, it comes as no
surprise that the functions subserved by each of these subsystems
could be developed independently from each other. An interesting
twist in this regard, however, comes from results reported by
Schendan el al. (2003), who found activation of the medial tem-
poral lobe in a sequence learning task early in training, specifically
when it involved second-order, relational contingencies, and even
in conditions in which learning remained implicit. If the role of the
hippocampus in this process is to bind sequential events into
unique episodic experiences, as was suggested by these authors,
then the results of Experiment 2 could be especially relevant,
because they would show that these medial temporal structures
could simultaneously contribute to two different binding pro-
cesses, a diachronic one that binds successive targets, and a syn-
chronic one that associates each context with a specific location of
the target.
At a functional level, the observed lack of competition between
sequence learning and contextual cueing effects would be surpris-
ing only if one conceives these processes as based either on a
general problem-solving ability or on the result of some complex
integration of independent mechanisms that, however, need to
share some specific modules at certain processing levels. The
observed lack of competition between these two learning processes
is thus consistent with the conclusions that these processes are not
sustained by a shared pool of nonmodular working-memory re-
sources, nor are they requiring the use of overlapping modules at
the same time or for the same purpose. Concerning the sequence
learning process, for instance, there is evidence indicating that,
even if it involves multicomponent influences, one of their main
components has to do with a representation of the sequence of
response locations (Willingham, Wells, Farrel, & Stemwedel,
2000). In contrast, the locus of the effect of contextual cueing has
been located earlier, at a perceptual level, and it has been described
233
SEQUENCE LEARNING AND CONTEXTUAL CUEING
as a facilitation of the search process, resulting from a guidance of
attention toward the location of the cued target (Jiang & Chun,
2003). Given that these two effects arise at different processing
stages and at different moments, it follows that they could affect
performance in an additive manner. Indeed, the absence of an
interaction between implicit sequence learning and contextual cue-
ing could be seen as problematic for a recent attempt to locate the
contextual cueing effects at the response selection stage (Kunar,
Flusberg, Horowitz, & Wolfe, 2007; Kunar, Flusberg, & Wolfe,
2008). According to these authors, when a target appears in a
familiar context, this context should not necessarily guide partic-
ipants’ attention toward the location of the target, but it could
lower their response threshold in comparison with conditions in
which the context is not familiar, thus requiring less evidence
about the identity of the target before committing to a response. If
this were the case, indeed, we surmise that the effect of contextual
cueing could be expected to interact with the effect of implicit
sequence learning, so that when the identity of the target is already
predicted by the previous series, the effect of a lowered response
threshold should become minimized. Because this pattern of in-
teraction has not been observed in the present study, our results do
not lend support to such a response-selection account of contextual
cueing. We hope that future designs could use the paradigm
introduced in this article as a useful tool to further investigate the
locus of these contextual cueing effects. Moreover, we surmise
that the specific effect provoked by explicit sequence learning on
the expression of contextual cueing could be used as an indirect
way to assess the explicit versus implicit nature of sequence
learning.
References
Bargh, J. A., & Morsella, E. (2008). The unconscious mind. Perspectives
on Psychological Science, 3, 73–79.
Chun, M. M. (2000). Contextual cueing of visual attention. Trends in
Cognitive Sciences, 4, 170–178.
Chun, M. M., & Jiang, Y. V. (1998). Contextual cueing: Implicit learning
and memory of visual context guides spatial attention. Cognitive Psy-
chology, 36, 28–71.
Chun, M. M., & Jiang, Y. V. (1999). Top–down attentional guidance based
on implicit learning of visual covariation. Psychological Science, 10,
360–365.
Chun, M. M., & Jiang, Y. V. (2003). Implicit, long-term spatial contextual
memory. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 29, 224–234.
Chun, M. M., & Phelps, E. A. (1999). Memory deficits for implicit
contextual information in amnesic subjects with hippocampal damage.
Nature Neuroscience, 2, 844–847.
Cleeremans, A. (1997). Sequence learning in a dual-stimulus setting.
Psychological Research, 60, 72–86.
Cohen, A., Ivry, R. I., & Keele, S. W. (1990). Attention and structure in
sequence learning. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 16, 17–30.
Egner, T. (2007). Congruency sequence effects and cognitive control.
Cognitive, Affective, & Behavioral Neuroscience, 7, 380–390.
Endo, N., & Takeda, Y. (2005). Use of spatial context is restricted by
relative position in implicit learning. Psychonomic Bulletin & Review,
12, 880–885.
Greene, A. J., Gross, W. L., Elsinger, C. L., & Rao, S. M. (2007).
Hippocampal differentiation without recognition: An fMRI analysis of
the contextual cueing task. Learning & Memory, 14, 548–553.
Hazeltine, E., & Ivry, R. (2003). Neural structures that support implicit
sequence learning. In L. Jime´nez (Ed), Attention and implicit learning
(pp. 71–107). Philadelphia, PA: John Benjamins.
INQUISIT (Version 1.31) [Computer program]. (2004). Seattle, WA:
Millisecond Software. (http://www.millisecond.com)
Jiang, Y. V., & Chun, M. M. (2001). Selective attention modulates implicit
learning. Quarterly Journal of Experimental Psychology, 54A, 1105–
1124.
Jiang, Y. V., & Chun, M. M. (2003). Contextual cueing: Reciprocal
influences between attention and implicit learning. In L. Jime´nez (Ed),
Attention and implicit learning (pp. 277–296). Philadelphia, PA: John
Benjamins.
Jiang, Y. V., & Leung, A. W. (2005). Implicit learning of ignored visual
context. Psychonomic Bulletin & Review, 12, 100–106.
Jime´nez, L. (2003). Intention, attention, and consciousness in probabilistic
sequence learning. In L. Jime´nez (Ed), Attention and implicit learning
(pp. 43–68). Philadelphia, PA: John Benjamins.
Jime´nez, L., Lupia´n˜ ez, J., & Vaquero J. M. M. (2009). Sequential congru-
ency effects in implicit sequence learning. Consciousness and Cogni-
tion, 18, 690–700.
Jime´nez, L., & Me´ndez, C. (1999). Which attention is needed for implicit
sequence learning? Journal of Experimental Psychology: Learning,
Memory, and Cognition, 25, 236–259.
Jime´nez, L., & Me´ndez, C. (2001). Implicit sequence learning with com-
peting explicit cues. Quarterly Journal of Experimental Psychology,
54A, 345–369.
Jime´nez, L., Vaquero, J. M. M., & Lupia´n˜ ez, J. (2006). Qualitative differ-
ences between implicit and explicit sequence learning. Journal of Ex-
perimental Psychology: Learning, Memory, and Cognition, 32, 475–
490.
Jime´nez, L., & Va´zquez, G. A. (2005). Sequence learning under dual-task
conditions: Alternatives to a resource-based account. Psychological Re-
search, 69, 352–368.
Jime´nez, L., & Va´zquez, G. A. (2008). Implicit sequence learning in a
search task. Quarterly Journal of Experimental Psychology, 61, 1650
1657.
Junge´, J. A., Scholl, B. J., & Chun, M. M. (2007). How is spatial context
learning integrated over signal versus noise? A primacy effect in con-
textual cueing. Visual Cognition, 15, 1–11.
Keele, S. W., Ivry, R., Mayr, U., Hazeltine, E., & Heuer, H. (2003). The
cognitive and neural architecture of sequence representation. Psycholog-
ical Review, 110, 316–339.
Kunar, M. A., Flusberg, S. J., Horowitz, T. S., & Wolfe, J. M. (2007). Does
contextual cueing guide the deployment of attention? Journal of Exper-
imental Psychology: Human Perception and Performance, 33, 816
828.
Kunar, M. A., Flusberg, S. J., & Wolfe, J. M. (2008). Time to guide:
Evidence for delayed attentional guidance in contextual cueing. Visual
Cognition, 16, 804–825.
Mayr, U. (1996). Spatial attention and implicit sequence learning: Evi-
dence for independent learning of spatial and nonspatial sequences.
Journal of Experimental Psychology: Learning, Memory, and Cogni-
tion, 22, 350–364.
Moors, A., & De Houwer, J. (2006). Automaticity: A conceptual and
theoretical analysis. Psychological Bulletin, 132, 297–326.
Nissen, M. J., & Bullemer, P. (1987). Attentional requirements of learning:
Evidence from performance measures. Cognitive Psychology, 19, 1–32.
Olson, I. R., & Chun, M. M. (2001). Temporal contextual cueing of visual
attention. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 27, 1299–1313.
Olson, I. R., & Chun, M. M. (2002). Perceptual constraints on implicit
learning of spatial context. Visual Cognition, 9, 273–302.
Rausei, V., Makovski, T., & Jiang, Y. V. (2007). Attention dependency in
implicit learning of repeated search context. Quarterly Journal of Ex-
perimental Psychology, 60, 1321–1328.
234 JIME
´NEZ AND VA
´ZQUEZ
Reed, J., & Johnson, P. (1994). Assessing implicit learning with indirect
tests: Determining what is learned about sequence structure. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 20, 585–
594.
Rowland, L. A., & Shanks, D. R. (2006). Attention modulates the learning
of multiple contingencies. Psychonomic Bulletin & Review, 13, 643–
648.
Schendan, H. E., Searl, M. M., Melrose, R. J., & Stern, C. E. (2003). An
fMRI study of the medial temporal lobe in implicit and explicit sequence
learning. Neuron, 37, 1013–1025.
Schmidtke, V., & Heuer, H. (1997). Task integration as a factor in
secondary-task effects on sequence learning. Psychological Research,
60, 53–71.
Schvaneveldt, R. W., & Gomez, R. (1998). Attention and probabilistic
sequence learning. Psychological Research, 61, 175–190.
Shanks, D. R. (2003). Attention and awareness in ‘implicit’ sequence
learning. In L. Jime´nez (Ed), Attention and implicit learning (pp. 11–
42). Philadelphia, PA: John Benjamins.
Shanks, D. R., & Channon, S. (2002). Effects of a secondary task on
‘implicit’ sequence learning: Learning or performance? Psychological
Research, 66, 99–109.
Shanks, D. R., Rowland, L., & Ranger, M. S. (2005). Attentional load and
implicit sequence learning. Psychological Research, 69, 369–382.
Shin, J. C. (2008). The procedural learning of action order is independent
of temporal learning. Psychological Research, 72, 376–386.
Shin, J. C., & Ivry, R. B. (2002). Concurrent learning of temporal and
spatial sequences. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 28, 445–457.
Smyth, A. C., & Shanks, D. R. (2008). Awareness in contextual cueing
with extended and concurrent explicit tests. Memory & Cognition, 36,
403–415.
Willingham, D. B., Wells, L. A., Farrell, J. M., & Stemwedel, M. E.
(2000). Implicit motor sequence learning is represented in response
locations. Memory & Cognition, 28, 366–375.
Received May 13, 2009
Revision received March 4, 2010
Accepted March 8, 2010
235
SEQUENCE LEARNING AND CONTEXTUAL CUEING
... Reaction time (RT) was our primary, and accuracy our secondary variable of interest. Statistical assessment of the RT data was done for the second half of the experiment, i.e. blocks [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36] (selected a priori). Only RT data from correct trials was used. ...
... A previous study exposed observers to sequenced information in addition to, but independent of, spatial predictive context. Signs of sequence learning were visible after 96 repetitions (33). It is possible that this many repetitions of the spatial context in our study would have allowed exploitation of temporal predictive context. ...
Preprint
Full-text available
The human visual system can rapidly extract regularities from our visual environment, generating predictive context. It has been shown that spatial predictive context can be used during visual search. We set out to see whether observers can additionally exploit temporal predictive context, using an extended version of a contextual cueing paradigm. Though we replicated the contextual cueing effect, repeating search scenes in a structured order versus a random order yielded no additional behavioural benefit. This was true both for participants who were sensitive to spatial predictive context, and for those who were not. We argue that spatial predictive context during visual search is more readily learned and subsequently exploited than temporal predictive context, potentially rendering the latter redundant. In conclusion, unlike spatial context, temporal context is not automatically extracted and used during visual search.
... A previous study exposed observers to sequenced information in addition to, but independent of, spatial predictive context. Signs of sequence learning were visible after 96 repetitions [34]. It is possible that this many repetitions of the spatial context in our study would have allowed exploitation of temporal predictive context. ...
Article
Full-text available
The human visual system can rapidly extract regularities from our visual environment, generating predictive context. It has been shown that spatial predictive context can be used during visual search. We set out to see whether observers can additionally exploit temporal predictive context based on sequence order, using an extended version of a contextual cueing paradigm. Though we replicated the contextual cueing effect, repeating search scenes in a structured order versus a random order yielded no additional behavioural benefit. This was also true when we looked specifically at participants who revealed a sensitivity to spatial predictive context. We argue that spatial predictive context during visual search is more readily learned and subsequently exploited than temporal predictive context, potentially rendering the latter redundant. In conclusion, unlike spatial context, temporal context is not automatically extracted and used during visual search.
... Conversely, attention focused on a limited visual target with high salience would centrally gate the input from the peripheral vision [99] and entail a longer time period for reaching steady state in the balancing behaviour. Under different conditions, like standing on an unstable platform [100][101][102], a cognitive task enhanced automaticity of the balancing behaviour and allowed for reduced sway, but the time course of that effect was not identified. ...
Article
Full-text available
When humans are administered continuous and predictable perturbations of stance, an adaptation period precedes the steady state of balancing behaviour. Little information is available on the modulation of adaptation by vision and perturbation frequency. Moreover, performance of supra-postural tasks may modulate adaptation in as yet unidentified ways. Our purpose was to identify differences in adaptation associated to distinct visual tasks and perturbation frequencies. Twenty non-disabled adult volunteers stood on a platform translating 10 cm in antero-posterior (AP) direction at low (LF, 0.18 Hz) and high frequency (HF, 0.56 Hz) with eyes open (EO) and closed (EC). Additional conditions were reading a text fixed to platform (EO-TP) and reading a text stationary on ground (EO-TG). Peak-to-peak (PP) displacement amplitude and AP position of head and pelvis markers were computed for each of 27 continuous perturbation cycles. The time constant and extent of head and pelvis adaptation and the cross-correlation coefficients between head and pelvis were compared across visual conditions and frequencies. Head and pelvis mean positions in space varied little across conditions and perturbation cycles but the mean head PP displacements changed over time. On average, at LF, the PP displacement of the head and pelvis increased progressively. Adaptation was rapid or ineffective with EO, but slower with EO-TG, EO-TP, EC. At HF, the head PP displacement amplitude decreased progressively with fast adaptation rates, while the pelvis adaptation was not apparent. The results show that visual tasks can modulate the adaptation rate, highlight the effect of the perturbation frequency on adaptation and provide evidence of priority assigned to pelvis stabilization over visual tasks at HF. The effects of perturbation frequency and optic flow and their interaction with other sensory inputs and cognitive tasks on the adaptation strategies should be investigated in impaired individuals and considered in the design of rehabilitation protocols.
... Naturalmente, otros autores no comparten nuestro escepticismo (e.g., Gourjon, Didierjean, & Thorpe, 2015;Pollmann, Eštočinová, Sommer, Chelazzi, & Zinke, 2016). Y, en honor a la verdad, es necesario señalar que la afirmación de que la señalización contextual es implícita no sólo se basa en la cuestión de si los participantes son conscientes o no del aprendizaje (e.g., Jiménez & Vázquez, 2011;Manginelli, Langer, Klose, & Pollmann, 2013). Ciertamente, no es imposible que estudios futuros muestren más allá de toda duda que la señalización contextual se basa en un proceso de aprendizaje inconsciente. ...
Article
Full-text available
هدف البحث الحالي إلى دراسة الفروق بين التعلم الصريح والتعلم الضمني في كم المعرفة المكتسبة ، ودراسة تأثير عبء الانتباه على كل من كم المعرفة المكتسبة من التعلم الصريح و التعلم الضمني. وقد أجرى البحث على عينه مكونه من 68 طالب وطالبة بكلية التربية جامعة عين شمس، وبلغ متوسط عمر العينة 21.2 سنه بانحراف معياري 0.79 سنه. وتم تطبيق مهمة تعلم القاعدة المصطنعة Artificial grammar learning لقياس كم المعرفة المكتسبة من التعلم الصريح والتعلم الضمني. وتم استخدام مهمة توليد الأرقام العشوائية Random number generation كمهمة عبء انتباه. وقد أجرى الباحثون المعالجة الإحصائية للبيانات باستخدام تحليل التباين ثنائي الاتجاه . وأشارت النتائج إلى اختلاف كم المعرفة المكتسبة باختلاف نوع التعلم لصالح التعلم الصريح . ويختلف كم المعرفة المكتسبة باختلاف عبء الانتباه لصالح عدم وجود عبء انتباه . ويوجد أثر لتفاعل كل من نوع التعلم (ضمني – صريح ) وعبء الانتباه (بدون – في وجود ) على كم المعرفة المكتسبة حيث توجد فروق بين التعلم الضمني بدون عبء والصريح بدون عبء في كم المعرفة المكتسبة لصالح التعلم الصريح. و أن مهمة عبء الانتباه أثرت على كم المعرفة المكتسبة من التعلم الصريح بينما لم تؤثر على كم المعرفة المكتسبة من التعلم الضمني. The present research aims at examining the difference between explicit learning and implicit learning in the amount of knowledge acquired, and examining the impact of the attentional load on both the amount of acquired knowledge from both forms of learning explicit learning and implicit learning. The research was applied on a sample of(68 (students at the Faculty of Education at Ain Shams University. The average age of the sample was 21.2 years, with a standard deviation of 0.79 years. The Artificial learning task was used to measure the amount of acquired knowledge from explicit learning and implicit learning. The random number generation was used as attentional load task. The Two-Way ANOVA analysis was used.The results indicated that there are differences between implicit and explicit learning in the amount of acquired knowledge acquired in favor of explicit learning. Also, that attentional load task affected amount of knowledge gained in explicit learning but did not affect amount of knowledge gained from implicit learning.
Article
Persistence of superstitions in the modern era could be justified by considering them as a by-product of the brain's capacity to detect associations and make assumptions about cause-effect relationships. This ability, which supports predictive behaviour, directly relates to associative learning. We tested whether variability in superstitious behaviour reflects individual variability in the efficiency of mechanisms akin to habit learning. Forty-eight individuals performed a Serial Reaction Time Task (SRTT) or an Implicit Cuing Task (ICT). In the SRTT, participants were exposed to a hidden sequence and progressively learnt to optimize responses, a process akin to skill learning. In the ICT participants met with a hidden association, which (if detected) provided a benefit (cf. habit learning). An index of superstitious beliefs was also collected. A correlation emerged between susceptibility to personal superstitions and performance at the ICT only. This novel finding is discussed in view of current ideas on how superstitions are instated.
Article
Full-text available
Since the seminal study by Chun and Jiang (1998), a large body of research based on the contextual cueing paradigm has shown that the cognitive system is capable of extracting statistical contingencies from visual environments. Most of these studies have focused on how individuals learn regularities found within an intratrial temporal window: a context predicts the target position within a given trial. However, Ono, Jiang, and Kawahara (2005) provided evidence of an intertrial implicit-learning effect when a distractor configuration in preceding trials N-1 predicted the target location in trials N. The aim of the present study was to gain further insight into this effect by examining whether it occurs when predictive relationships are impeded by interfering task-relevant noise (Experiments 2 and 3) or by a long delay (Experiments 1, 4, and 5). Our results replicated the intertrial contextual cueing effect, which occurred in the condition of temporally close contingencies. However, there was no evidence of integration across long-range spatiotemporal contingencies, suggesting a temporal limitation of statistical learning. Keywords: intertrial association, contextual cuing, visual attention, visual search, statistical learning
Article
Full-text available
A current controversy in the area of implicit statistical learning (ISL) is whether this process consists of a single, central mechanism or multiple modality-specific ones. To provide insight into this question, the current study involved three ISL experiments to explore whether multimodal input sources are processed separately in each modality or are integrated together across modalities. In Experiment 1, visual and auditory ISL were measured under unimodal conditions, with the results providing a baseline level of learning for subsequent experiments. Visual and auditory sequences were presented separately, and the underlying grammar used for both modalities was the same. In Experiment 2, visual and auditory sequences were presented simultaneously with each modality using the same artificial grammar to investigate whether redundant multisensory information would result in a facilitative effect (i.e., increased learning) compared to the baseline. In Experiment 3, visual and auditory sequences were again presented simultaneously but this time with each modality employing different artificial grammars to investigate whether an interference effect (i.e., decreased learning) would be observed compared to the baseline. Results showed that there was neither a facilitative learning effect in Experiment 2 nor an interference effect in Experiment 3. These findings suggest that participants were able to track simultaneously and independently two sets of sequential regularities under dual-modality conditions. These findings are consistent with the theories that posit the existence of multiple, modality-specific ISL mechanisms rather than a single central one.
Article
Objects in the environment have both location and identity properties. However, it is unclear how these independent properties are processed and combined in the implicit domain. The current study investigated the development of the implicit memory of object locations and object identities, both independently and combined, and the relation between implicit memory and working memory (WM) for these properties. Three age groups participated: 6- and 7-year-old children, 9- and 10-year-old children, and adults. Children and adults completed a repeated search paradigm. In the learning phase, targets' locations were consistently predicted by both the identities and locations of the distracters. In the test phase, either both remained predictive or just the identities or just the locations of the distracters predicted the location of the target. All groups showed significant implicit learning when both the identities and locations of the distracters remained predictive. When only the locations but not the identities of the distracters were predictive, adults and 9- and 10-year-olds showed significant learning, whereas 6- and 7-year-olds did not. When only the identities but not the locations of the distracters were predictive, none of the groups showed significant learning effects. In evaluating the contributions of either visual or spatial WM to implicit learning and memory, we found that children with smaller visual WM exhibited larger implicit memory effects for object identities than did children with larger visual WM. Taken together, the results indicate that children's ability to differentiate identity and location undergoes development even in the implicit domain.
Article
The present study draws together two distinct lines of enquiry into the selection and control of sequential action: motor sequence production and action selection in everyday tasks. Participants were asked to build 2 different Lego walls. The walls were designed to have hierarchical structures with shared and dissociated colors and spatial components. Participants built 1 wall at a time, under low and high load cognitive states. Selection times for correctly completed trials were measured using 3-dimensional motion tracking. The paradigm enabled precise measurement of the timing of actions, while using real objects to create an end product. The experiment demonstrated that action selection was slowed at decision boundary points, relative to boundaries where no between-wall decision was required. Decision points also affected selection time prior to the actual selection window. Dual-task conditions increased selection errors. Errors mostly occurred at boundaries between chunks and especially when these required decisions. The data support hierarchical control of sequenced behavior.
Article
Full-text available
Invariant spatial relationships of objects may provide a rich source of contextual information. Visual context can assist localization of individual objects via an implicit learning mechanism, as revealed in the contextual cueing paradigm (Chun & Jiang, 1998). What defines a visual context? How robust is contextual learning? And is it perceptually constrained? Here we investigate whether both local context that surround a target, and long-range context that does not spatially coincide with a target, can influence target localization. In the contextual cueing task, participants implicitly learned a context by repeated exposure to items arranged in invariant patterns. Experiments 1 and 2 suggest that only local con-text facilitates target localization. However, Experiment 3 showed that long-range context can prime target location when target and context are not separated by random information. Experiment 4 showed that grouping by colour does not affect contextual cueing, suggesting that spatial features play a more important role than surface features in spatial contextual cueing. In separate analyses, visual hemifield differences were found for learning and performance. In sum, the results indicate that implicit learning of spatial context is robust across noise and biased towards spatially grouped information. Context can have a powerful influence on the processing of visual information. Objects can be recognized without context but when dealing with less familiar objects, complex scenes, or degraded information, the importance of context increases (Ullman, 1996). In a series of studies, Biederman and colleagues Please address all correspondenc e to I.
Article
Contextual cueing (CC) experiments show that when visual search displays are repeated, reaction times (RTs) to find a target decrease over time even when observers are not aware of the repetition. In other experiments, observers use information about features, like color, to guide attention to likely target locations. Do observers use implicit memory for the display to guide attention in the same manner in CC tasks? We compared CC effects to standard guidance. Firstly, guidance decreases the slope of RT x Set Size functions. When guidance is perfect (e.g. feature search for a red target among green distractors) slopes will be close to 0 msec/item. In contrast, contextually repeated trials showed little reduction in slope compared to trials that did not repeat context. Secondly, observers can use color to guide attention to multiple target locations. However, with the same training, observers could not prioritize more than two target locations in CC. Thirdly, in standard guided search there is little reduction in guidance if the location of a target on one trial is occupied by a distractor on the next. However, when a distractor appeared in a prior target location in a CC paradigm, a CC benefit was not found. Finally, even with perfect guidance (e.g., in a feature search), we found a small, but reliable additional CC effect. This suggests a contribution from other factors, such as response selection. Overall, the data suggest that the relationship between guidance and CC is weak.
Article
The visual environment is extremely rich and complex, producing information overload for the visual system. But the envi- ronment also embodies structure in the form of redundancies and reg- ularities that may serve to reduce complexity. How do perceivers internalize this complex informational structure? We present new evi- dence of visual learning that illustrates how observers learn how objects and events covary in the visual world. This information serves to guide visual processes such as object recognition and search. Our first experiment demonstrates that search and object recognition are facilitated by learned associations (covariation) between novel visual shapes. Our second experiment shows that regularities in dynamic visual environments can also be learned to guide search behavior. In both experiments, learning occurred incidentally and the memory rep- resentations were implicit. These experiments show how top-down visual knowledge, acquired through implicit learning, constrains what to expect and guides where to attend and look.
Article
ABSTRACT The unconscious mind is still viewed by many psychological scientists as the shadow of a “real” conscious mind, though there now exists substantial evidence that the unconscious is not identifiably less flexible, complex, controlling, deliberative, or action-oriented than is its counterpart. This “conscious-centric” bias is due in part to the operational definition within cognitive psychology that equates unconscious with subliminal. We review the evidence challenging this restricted view of the unconscious emerging from contemporary social cognition research, which has traditionally defined the unconscious in terms of its unintentional nature; this research has demonstrated the existence of several independent unconscious behavioral guidance systems: perceptual, evaluative, and motivational. From this perspective, it is concluded that in both phylogeny and ontogeny, actions of an unconscious mind precede the arrival of a conscious mind—that action precedes reflection.
Article
The authors introduce the contextual cueing paradigm and investigate the reciprocal influences that hold between attention and implicit learning. In this paradigm, a visual search task is manipulated by including distractors that covary with the spatial location of the target, and the authors demonstrate that participants do implicitly learn about these spatial correlations, but only to the extent that the distractors cannot be efficiently ignored. Thus the results of these experiments strongly indicate that implicit learning and attention are related in complex and reciprocal ways, so that selective attention does modulate what can be learned implicitly, and implicit learning shapes the deployment of attention. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
In this study we investigated the role of attention, sequence structure, and effector specificity in learning a structured sequence of actions. Experiment 1 demonstrated that simple structured sequences can be learned in the presence of attentional distraction. The learning is unaffected by variation in distractor task difficulty, and subjects appear unaware of the structure. The structured sequence knowledge transfers from finger production to arm production (Experiment 2), suggesting that sequence specification resides in an effector-independent system. Experiments 3 and 4 demonstrated that only structures with at least some unique associations (e.g., any association in Structure 15243… or 4 to 3 in Structure 143132…) can be learned under attentional distraction. Structures with all items repeated in different orders in different parts of the structure (e.g., Sequence 132312…) require attention for learning. Such structures may require hierarchic representation, the construction of which takes attention. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Investigated the unconfounding of learning of simple frequency information from complex structure during sequenced serial reaction time (RT) trials. Experiment 1 demonstrated that disruptions in RTs that occur when Ss are transferred to random trials (e.g., A. Cohen et al; see record 1990-08941-001) may result from a change in simple frequency information. The implication is that negative transfer effects cannot be attributed to the learning of complex information if simpler information is also allowed to vary between training and transfer trials. Exp 2 used a procedure in which simple frequency information remained constant between training and transfer trials, which provided evidence of complex sequence structure learning while subjects were engaged in a secondary tone-counting task. Results from cued-generation and recognition tests indicated that the learning was implicit. (PsycINFO Database Record (c) 2012 APA, all rights reserved)