Content uploaded by Nicholaus Brosowsky
Author content
All content in this area was uploaded by Nicholaus Brosowsky on May 22, 2018
Content may be subject to copyright.
Running head: MEMORY-GUIDED SELECTIVE ATTENTION
Memory-guided selective attention: Single experiences with conflict have long-lasting
effects on cognitive control
Nicholaus P. Brosowsky
The Graduate Center of the City University of New York
Matthew J.C. Crump
Brooklyn College and the Graduate Center of the City University of New York
Word count: 13684
Author Note
Data in this manuscript were previously presented, in part, at the 57th Annual
Meeting of the Psychonomic Society, November 2016.
Stimulus and data files for all experiments and code for constructing the
experiments are openly accessible at osf.io/gr8zv.
Correspondence concerning this manuscript should be addressed to Nicholaus
P. Brosowsky, Department of Psychology, The Graduate Center, CUNY, 365 5th Ave,
New York, NY 10016
E-mail: nbrosowsky@gradcenter.cuny.edu
MEMORY-GUIDED SELECTIVE ATTENTION
2
Abstract
Adjustments in cognitive control, as measured by congruency sequence effects,
are thought to be influenced by both external stimuli and internal goals. However, this
dichotomy has often overshadowed the potential contribution of past experience stored
in memory. Here, we examine the role of long-term episodic memory in guiding
selective attention. Our aim was to demonstrate new evidence that selective attention
can be modulated by long-term retrieval of stimulus-specific attentional control settings.
All the experiments used a modified flanker task involving multiple unique stimuli.
Critically, each stimulus was only presented twice during the experiment: first as a
prime, and second as a probe. Experiments 1 and 2 varied the number of intervening
trials between prime and probe and manipulated the amount of conflict using a
secondary task. Experiment 3 ensured that specific colors assigned to prime stimuli
were not repeated when presented as probes. Across both experiments 1 and 2, we
consistently found smaller congruency effects on probe trials when its associated prime
trial was incongruent compared to congruent, demonstrating long-term congruency
sequence effects. However, experiment 3 showed no evidence for long-term effects.
These findings suggest long-term preservation of selective attention processing at the
episodic level, and implicate a role for memory in updating cognitive control.
Keywords: Attention, memory, cognitive control, conflict adaptation, conflict monitor
MEMORY-GUIDED SELECTIVE ATTENTION
3
Memory-guided selective attention: Single experiences with conflict have long-lasting
effects on cognitive control
Cognitive control enables flexible goal-directed behavior via attention and action
selection processes that prioritize goal-relevant over irrelevant information. Attention is
known to be strongly influenced by both external stimuli and internal goals. However,
the strict dichotomy between stimulus-driven and goal-driven influences (Posner &
Snyder, 1975; Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977) has downplayed
the role of memory in guiding attention (Awh, Belopolsky, & Theeuwes, 2012;
Hutchinson & Turk-Browne, 2012). People often re-encounter similar objects, tasks, and
environments that require similar cognitive control operations. A memory-retrieval
process could shortcut the slow, effortful, and resource-demanding task of updating
control settings by retrieving and reinstating the control procedures used in the past.
Here we examine the role of long-term episodic memory in guiding selective attention.
Evidence for long-term, cue-driven retrieval of control operations has been
reported in multiple attention paradigms, suggesting a general phenomenon. However,
evidence within paradigms is limited to a small number of reports, and remains absent
in conventional selective attention tasks, such as Stroop (1935) and Flanker (Eriksen &
Eriksen, 1974), commonly used to make inferences about cognitive control processes.
Our aim was to demonstrate new evidence that selective attention can be modulated by
long-term retrieval of stimulus-specific attentional control settings, and then discuss
implications of these findings for theories of cognitive control.
Long-term retrieval of control settings
MEMORY-GUIDED SELECTIVE ATTENTION
4
Early evidence for long-term, cue-driven retrieval of attentional control settings
was developed in the negative priming literature (for recent reviews, see D’Angelo,
Thomson, Tipper, & Milliken, 2016; Frings, Schneider, & Fox, 2015). Negative priming
refers generally to the finding that reaction times to identify a previously ignored target
are slowed compared to a target that was not previously ignored (Tipper, 1985). In a
typical design, a prime display might include a to-be-named green target word (e.g.,
TRUCK) interleaved with a to-be-ignored red distractor word (e.g., PIANO). An
immediately following probe display then presents a target/distractor pair, involving a
target that was previously attended (attended repetition: TRUCK), previously ignored
(ignored repetition: PIANO), or a word that was not attended or ignored (control:
MOCHA). Negative priming is observed when ignored repetition reaction times are
slower than control trials. Early explanations of negative priming invoked a short-term,
transient inhibitory process: ignoring a stimulus causes it to be briefly inhibited, and
negative priming reflects the extra time needed to recover from inhibition during
responding (Tipper, 1985; Tipper & Driver, 1988). However, two classes of findings
were difficult to reconcile with the short-term inhibition explanation, and were formative
for the idea that long-term, cue-driven memory processes may play a role in re-instating
prior attentional control settings.
First, negative priming is sensitive to the match between probe and prime tasks,
and can disappear when the probe task does not require selective attention to the
target. The above task description involves selection in both prime and probe trials, as
both trials present an interleaved target/distractor pair. If negative priming reflects carry-
over of inhibition from the ignored distractor on the prime trial, then that inhibition ought
MEMORY-GUIDED SELECTIVE ATTENTION
5
to be detected on a following probe trial that presented the ignored distractor alone, as a
single target. In this case, the probe trials do not require selection because only a single
target is displayed. However, several experiments showed that negative priming is
abolished when the probe display contains a single target (D. G. Lowe, 1979; Milliken,
Joordens, Merikle, & Seiffert, 1998; Moore, 1994; Tipper & Cranston, 1985).
Second, negative priming can persist for long temporal intervals between a prime
and probe trial. DeSchepper and Treisman (1996) demonstrated that negative priming
in a shape discrimination task is observed up to 30 days between a prime trial (including
a target and distractor shape), and a probe trial (including the previously ignored shape
as the target). We are aware of only two other investigations of long-term negative
priming. Lowe (1998) demonstrated negative priming persisting for 5 minutes, and
Grison, Tipper, and Hewitt (2005), showed negative priming persisting over 54
intervening trials between a prime and probe.
Taken together, the findings that negative priming is sensitive to the match
between probe and prime tasks, and that negative priming persists over the long-term,
provided evidence suggesting a role for memory-based retrieval processes in negative
priming. For example, inspired by instance-theories of memory (Hintzman, 1984; Logan,
1988), Neill and colleagues (Neill, 1997; Neill & Valdes, 1992; Neill, Valdes, Terry, &
Gorfein, 1992) proposed an episodic retrieval account of negative priming. Here, an
ignored distractor presented during a prime trial is tagged with a "do-not-respond"
control operation. If the ignored distractor is presented as a target on the following
probe trial, it could then retrieve its associated "do-not-respond" control operation, which
would interfere with responding to that stimulus on the probe trial. Furthermore,
MEMORY-GUIDED SELECTIVE ATTENTION
6
because control operations associated with prime processing are preserved in an
instance-based memory, they could be available (under the appropriate retrieval
conditions) over the long-term.
Evidence for long-term retrieval of attention control settings, like those observed
in negative priming, has been shown in a few different attention paradigms. These
include long-term inhibition of return (Tipper, Grison, & Kessler, 2003), long-term
retrieval of task-sets in task-switching (Waszak, Hommel, & Allport, 2003), long-term
priming-of-pop out in visual search (Thomson & Milliken, 2012, 2013), and long-term
response inhibition in stop-signal tasks (Verbruggen & Logan, 2008). It remains unclear
whether this collection of evidence points to a general role for memory retrieval of
control operations linked with specific prior processing episodes to update and adjust
control operations in the present moment.
However, evidence for long-term retrieval of attention control settings has not
been established in classic selective attention paradigms, such as Stroop and Flanker,
commonly used to make inferences about cognitive control processes. A demonstration
would be useful in its own right to further establish the generality of the phenomena and
would test theories of control processes used to explain modulations to congruency
effects. We outline theoretical implications for explanations of n-1 congruency sequence
effects, and proportion congruent effects; and, then overview the procedures we
adopted to measure long-term memory based control of attention.
Congruency effects
Congruency tasks measure target identification in the presence of potentially
conflicting distractors. For example, in the Flanker task (Eriksen & Eriksen, 1974)
MEMORY-GUIDED SELECTIVE ATTENTION
7
participants are faster and more accurate to identify a center letter (e.g., “HHFHH”)
when flanking letters are congruent (e.g., “HHHHH”) versus incongruent (e.g., “FFHFF”)
with the response. Modulations to the size of congruency effects can index the gain of
attentional control assigned to target and distractor dimensions. For example, target
information is assumed to be prioritized over distractor information when smaller versus
larger congruency effects are observed.
Importantly, congruency effects are modulated by the history of previously
experienced conflict. Congruency effects are reduced immediately following an
incongruent trial, and when the proportion of incongruent trials is greater than the
proportion of congruent trials. It is possible that both trial history effects could be
explained by common principles, and some existing accounts have forwarded unified
theories (Abrahamse, Braem, Notebaert, & Verguts, 2016; Egner, 2014; Verguts &
Notebaert, 2008). We consider whether common principles invoked by the notion of
long-term, cue-driven retrieval of attention control settings could explain congruency
sequence and proportion congruent effects. Alternatively, memory-driven control could
reflect a distinct influence that clarifies how different processes acting over the long and
short-term use prior experience with conflict to update control settings.
Congruency Sequence effects. Congruency effects on trial n are smaller when
trial n-1 contains an incongruent versus congruent trial (Gratton, Coles, & Donchin,
1992, for a review see Egner, 2007). Early explanations invoked voluntary control
(Gratton et al., 1992), but recent findings suggest volition is not necessary. For
example, congruency sequence effects can be produced despite contradictory
expectations about the likelihood of conflict on the next trial (Jiménez & Méndez, 2013,
MEMORY-GUIDED SELECTIVE ATTENTION
8
2014) and in the absence of awareness (Desender, Van Lierde, & Van den Bussche,
2013). Congruency sequence effects also occur over short timescales, persisting only
for one or two trials (Akçay & Hazeltine, 2008; Mayr, Awh, & Laurey, 2003), quickly
decaying with increased inter-stimulus or response-to-stimulus intervals, and eliminated
all-together after three- to seven-second intervals (Duthoo, Abrahamse, Braem, &
Notebaert, 2014; Egner, 2010).
All accounts of congruency sequence effects assume that influences from a
recent trial on current trial performance are transient and decay rapidly. Debate focuses
on whether or not congruency sequence effects are driven by processes that change
attentional control settings. Rapid decay is assumed by non-control accounts based on
feature integration or event-binding processes (Hommel, 1998; Hommel, Müsseler,
Aschersleben, & Prinz, 2001; Hommel, Proctor, & Vu, 2004), repetition priming (Mayr et
al., 2003), and sequential contingency biases (Schmidt & De Houwer, 2011). Rapid
decay is also assumed by control accounts based on conflict-monitoring theory
(Botvinick, Braver, Barch, Carter, & Cohen, 2001). Here, a conflict-monitoring unit
registers a transient conflict signal that triggers adjustments to attentional control
settings which carry-forward to influence performance on the next trial.
There are notable parallels between the congruency sequence effect and
negative priming. Like the congruency sequence, negative priming was assumed to
operate on a transient, short-term basis. Although the congruency sequence can
dissipate over the short-term, it remains unclear whether experiencing conflict on one
trial can have long-term influences over congruency effects on future trials. There is
some evidence that congruency sequence effects can accumulate in strength as a
MEMORY-GUIDED SELECTIVE ATTENTION
9
function of the number of preceding incongruent trials (Aben, Verguts, & Van den
Bussche, 2017; Jiménez & Méndez, 2013; Rey-Mermet & Meier, 2017). However, there
is no evidence, akin to long-term negative priming, showing that control operations
applied on a single trial to a specific stimulus can be retrieved on a long-term basis to
influence control operations to similar stimuli in the future. Another parallel is that
congruency-sequence effects, like negative priming, can depend on the match between
tasks performed on trial n-1 and trial n. For example, conflict experienced on trial n-1 in
one interference task does not always cause modulations to congruency effects for a
different task presented on trial n (for a review, see Braem, Abrahamse, Duthoo, &
Notebaert, 2014).
These parallels motivated us to determine whether congruency sequence-like
effects could extend across many intervening trials well beyond trial n-1. On the one
hand, a finding of this nature could identify a memory-based attentional control process
that is distinctly different from other short-term processes also capable of producing
congruency sequence effects. On the other hand, perhaps memory-based retrieval of
attention control settings could explain the short-term n-1 congruency sequence effect,
especially if temporal similarity, along with item and context features are assumed to act
as retrieval cues to apply control settings from recent trials (for similar perspectives, see
Egner, 2014; Spapé & Hommel, 2008, 2014).
Proportion Congruent effects. Proportion congruent effects show larger
congruency effects for conditions associated with high rather than low proportions of
congruent trials (for a review, see Bugg & Crump, 2012), and are demonstrated in list-
wide, item-specific, and context-specific designs. In a Stroop variant, item-specific
MEMORY-GUIDED SELECTIVE ATTENTION
10
designs assign one set of items (e.g., Red and Blue combinations) to a high proportion
congruent condition, and another set (e.g., Green and Yellow combinations) to a low
proportion congruent condition. Both item types are intermixed randomly, so subjects
cannot accurately predict whether the next trial will be congruent or incongruent. In
these designs, congruency effects are found to be larger for high versus low proportion
congruent item. Similarly, Context-specific proportion congruent (CSPC) designs
manipulate proportion congruent between two different contexts in which items can
appear, again in a randomized, intermixed fashion. CSPC effects have been shown
using location (Brosowsky & Crump, 2016; Corballis & Gratton, 2003; Crump, 2016;
Crump, Brosowsky, & Milliken, 2017; Crump, Gong, & Milliken, 2006; Hübner & Mishra,
2016; Weidler & Bugg, 2016), font (Bugg, Jacoby, & Toth, 2008; Crump, 2016), shape
(Crump, Vaquero, & Milliken, 2008), color (Vietze & Wendt, 2009), social categories
(Cañadas, Rodríguez-Bailón, Milliken, & Lupiáñez, 2013), and incidental semantic cues
(Blais, Harris, Sinanian, & Bunge, 2015). Again, congruency effects are larger for items
appearing in high than low proportion congruent contexts. These trial history effects
imply that item and context-specific cues become associated with attentional control
settings, and that changes to attentional control can be triggered in a cue-driven
manner.
We roughly group theories of item and context-specific proportion congruent
effects into memory-based and conflict-monitoring accounts. Memory-based accounts
invoke instance-based, long-term, cue-driven retrieval processes (Logan, 1988). Some
proportion congruent designs are confounded by item-frequency, and may be explained
simply by an event-learning process sensitive to the frequency of events (Schmidt,
MEMORY-GUIDED SELECTIVE ATTENTION
11
2013; Schmidt & Besner, 2008). At the same time other designs show evidence that
cues associated with proportion congruent can bias congruency effects even for
frequency unbiased items (Crump et al., 2017; Crump & Milliken, 2009; though, see
Hutcheon & Spieler, 2016). Here, memory-based accounts argue that attentional control
settings are encoded during each processing experience, and are retrieved to update
ongoing control operations in the present moment (Bugg & Hutchison, 2013; Crump,
2016; Crump et al., 2008). Conflict-monitoring accounts can explain item-specific
proportion congruent effects by assuming that conflict-signals trigger adjustments to
attentional control settings on an item-specific basis (Blais, Robidoux, Risko, & Besner,
2007; Verguts & Notebaert, 2008), and this kind of account could in principle be
extended to explain context-specific proportion congruent effects.
There are clear parallels between early item-specific proportion congruent
designs (Jacoby, Lindsay, & Hessels, 2003), and negative priming designs manipulating
the application of attentional control sets on an item-specific basis (Milliken, Lupianez,
Debner, & Abello, 1999). Indeed, the idea from negative priming that episodic retrieval
processes are used to retrieve and reinstate prior attentional control sets was borrowed
to explain proportion congruent effects. In the proportion congruent literature however,
there is no direct evidence supporting the core assumption of episodic retrieval theories
that control operations from single-trials are stored in traces, or that single-traces could
be retrieved to influence control operations for specific items on a long-term basis. For
example, most proportion congruent designs use a small number of stimuli that are
repeatedly presented over an experiment. It is unknown whether cues retrieve a single
MEMORY-GUIDED SELECTIVE ATTENTION
12
instance from among the available item repetitions or multiple instances that are
aggregated during retrieval.
A demonstration that congruency effects could be modulated by the long-term
retrieval of item-specific attention control settings has theoretical implications for
proportion congruent effects. A positive demonstration would corroborate predictions
from memory-based accounts, and challenge conflict-monitoring accounts that
aggregate over item-specific control settings (Botvinick et al., 2001; Braver, 2012; De
Pisapia & Braver, 2006; Jiang, Heller, & Egner, 2014).
Overview of present studies
Our experiments test whether a single experience with applying attentional
control to a unique stimulus can be retrieved on a long-term basis to influence how
attentional control is applied when the same stimulus is re-presented later. We
reasoned that if the single prior experience is retrieved, it will influence performance on
the current trial in a manner similar to the n-1 congruency sequence effect where
smaller congruency effects are found following an incongruent as compared to
congruent trial. In other words, we asked whether a congruency sequence-like effect
could be observed on a long-term basis, when there are many intervening trials
between a first and second experience with a unique stimulus.
All the experiments used a modified flanker task involving multiple unique stimuli.
The designs were inspired by long-term negative priming where a unique
target/distractor pair could be presented once as a prime stimulus, and once as a probe
stimulus after any number of intervening trials. We created unique stimuli using a large
bank of natural objects that could be displayed in different colors (Brady, Konkle, Gill,
MEMORY-GUIDED SELECTIVE ATTENTION
13
Oliva, & Alvarez, 2013). Each trial involved a row of objects, and the task was to identify
the color of the central object as quickly and accurately as possible. Like other context-
specific designs, the object feature dimension was irrelevant to the color-identification
task. Each object was only presented once as a prime, either in a congruent or
incongruent format, and once as a probe, either in a congruent or incongruent format.
Across experiments we varied the number of intervening trials between prime and probe
presentations. Our design allowed us to determine whether congruency effects for
probe stimuli would vary as a function of prime congruency, indicating a long-term
congruency sequence-like effect. Specifically, we measured whether the congruency
effect for probe stimuli preceded by incongruent primes would be smaller than the
congruency effect for probe stimuli preceded by congruent primes.
Experiments 1a, b, and c varied the number of intervening trials between prime
and probe by five to eleven trials and manipulated the amount of conflict using a
secondary task. Experiments 2a and b increased the number of intervening trials to an
average of 160 trials. To foreshadow our results, we found clear evidence of a long-term
congruency-sequence-like effect. Congruency effects for probes preceded by
incongruent primes were smaller than congruency effects for probes preceded by
congruent primes. Experiments 3a and 3b were conducted to test a long-term feature
integration account, and ensured that specific colors assigned to prime stimuli were not
repeated when presented as probes. These experiments showed no evidence of long-
term congruency sequence-like effects.
Experiment 1A, 1B, and 1C
MEMORY-GUIDED SELECTIVE ATTENTION
14
For experiment 1, we report three replications of the same experimental design
(see Figure 1). In all three experiments, the primary task was to identify the color of a
central image (either blue or green) flanked on the left and right by the same image
presented in either the same (congruent) or alternate color (incongruent). Each image
was only presented twice during the experiment: once as a prime stimulus, and once as
a probe stimulus. The trial order was constructed such that the distance between any
given prime and probe stimulus always ranged from 5 to 11 trials (8 trials, on average).
We chose to use a color flanker task so that congruency could be manipulated
independently of the image representing the target and flanker stimuli such that we
could repeat contextual images while alternating congruency.
The amount of conflict has been shown to influence the size of the n-1
congruency sequence effect (Forster, Carter, Cohen, & Cho, 2011; Wendt, Kiesel,
Geringswald, Purmann, & Fischer, 2015; though see Weissman & Carp, 2013). It was
unclear however, whether the amount of conflict would influence our ability to detect
long-term influences. For experiment 1A, we used the basic design described above.
For experiments 1B and 1C, we included a secondary task to increase conflict and
potentially improve our ability to detect the presence of long-term sequence effects. For
the secondary task, we required participants to press the spacebar if the identity of the
center image differed from the flankers. We reasoned that having participants
continuously monitor for differing flanker and target images would cause them to attend
more to the flanking images throughout the experiment and increase the overall level of
conflict. This alternative task was randomly presented once for every 8 normal trials.
MEMORY-GUIDED SELECTIVE ATTENTION
15
Experiment 1C was a replication of experiment 1B. A Monte-Carlo simulation
analysis of the results from experiment 1A suggested that doubling our trial count from
216 to 432 and increasing our subject count to 50 would increase our power to detect
the long-term sequence effect from an estimated .7 to .95 (for a complete description of
this procedure, see Crump et al., 2017). Therefore, for experiment 1C the trial count
was doubled and we collected data until we had 50 participants who completed all trials
and maintained an error rate less than 20%.
Methods
Participants. All participants were recruited from Amazon Mechanical Turk
(AMT) and compensated $1.00 (experiment 1A & 1B) or $3.00 (experiment 1C) for
participating. The amount compensated was calculated by estimating the maximum
amount of time required to complete each experiment and multiplying by $6.00 per
hour. For each experiment the number of HITs (Human intelligence tasks, an Amazon
term for a work-unit) refers to the number of participants who initiated the study.
Participants were included in the study if they completed all trials and each experiment
consisted of unique participants. For experiment 1A, 40 HITs were posted, and 40
participants completed all trials. For experiment 1B, 40 HITs were posted, and 39
participants completed all trials, and for experiment 1C, 55 HITs were posted, and 54
participants completed all trials.
Apparatus & Stimuli. The experiments were programmed using JavaScript,
CSS and HTML. The program allowed participants to complete task only if they were
running Safari, Google Chrome, or Firefox web browsers. Flanker stimuli were
constructed using the 540 images created by Brady, Konkle, Gill, Oliva, and Alvarez
MEMORY-GUIDED SELECTIVE ATTENTION
16
(2013). Images were color rotated to either blue or green (for a more detailed
description see Brady et al., 2013) and presented at 200 x 200 pixels. Each experiment
ran as a pop-up window that filled the entire screen. The background was white, and
stimuli were presented in the center of the screen.
Design. Experiment 1 used a 2x2x3 mixed design with prime congruency
(congruent vs. incongruent) and probe congruency (congruent vs. incongruent) as
within-subject factors, and experiment (1A, 1B, and 1C) as the between-subject factor.
Experiments 1A, 1B, and 1C were all constructed using the same general
method (see Figure 1). Every block of 16 trials was divided into four sub-blocks, each
consisting of four trials (referred to as the Prime A, Prime B, Probe A, and Probe B sub-
blocks). The images presented in the Prime A sub-block were then repeated in the
Probe A sub-block and images presented in the Prime B sub-block, repeated in the
Probe B sub-block. The trial order of each sub-block was randomized. The use of the
interleaved A/B sub-blocks ensured that the distance between any probe (trial n) and
prime stimulus pair ranged from n-5 to n-11. Importantly, the congruency of each
prime/probe pair was randomized and counterbalanced across each block with an equal
number of each congruency combination (i.e., Con – Con, Con – Inc, Inc – Con, and Inc
– Inc), and an equal number of response repetition and alternation prime/probe pairs.
Additionally, images were randomly selected for every participant from the total 540
images (Brady et al., 2013) and randomly assigned a color and condition. Each image
was only presented twice during the experiment: once in a prime block and once in a
probe block.
MEMORY-GUIDED SELECTIVE ATTENTION
17
Experiment 1A consisted of 192 trials constructed using this basic method.
Experiment 1B used the same general design but included a secondary task where
participants were instructed to press the spacebar if the center image differed in identity
to the flanking images. This alternate task occurred once for every 8 flanker trials,
bringing the total trials to 216. Experiment 1C was identical to experiment 1B except the
number of trials was doubled, bringing the total to 432 trials.
Procedure. All participants were AMT workers who found the experiment using
the AMT system. The participant recruitment procedure and tasks were approved by the
Brooklyn College Institutional Review Board. Each participant read a short description of
the task and gave consent by pressing a button acknowledging they had read the
displayed consent form. Participants then completed a short demographic survey, and
proceeded to the main task, which was displayed as a pop-up window. Participants
were instructed to identify the color of the center image on each trial as quickly and
accurately as possible by pressing ‘g’ if the image was green, and ‘b’ if the image was
blue. For experiments 1B and 1C, participants were further instructed to press the
spacebar if the identity of the center image differed from the identity of the flanking
images. Throughout the course of the experiment the upper left corner of the display
indicated the number of completed and remaining trials, as well as an instruction
reminder button that displayed the instructions in a new pop-up window.
Each trial began with a fixation cross presented in the center of the screen for
1,000 ms, followed by a blank inter-stimulus interval (ISI) of 250 ms. Next, the flanker
stimulus appeared in the center of screen, and remained on screen until a response
was made. Following a response, feedback indicating whether the response was correct
MEMORY-GUIDED SELECTIVE ATTENTION
18
or incorrect was presented above the target stimulus for 500 ms. For experiments 1B
and 1C, if the participant failed to press the spacebar on a secondary task trial, a
message appeared below the target stimulus reminding the participant of the secondary
task instructions. A response automatically triggered the next trial.
Halfway through experiments 1A (96 trials) and 1B (108 trials), participants were
instructed to take a short break, and to press the button on-screen when they were
ready to continue. In experiment 1C they received this message three times, each after
they had completed 108 trials.
Figure 1. Figure 1A shows examples of the stimuli and basic prime/probe structure
used in all experiments. Figure 1B shows the trial block structures from Experiments 1
and 2. In Experiment 1, every block of 16 trials was divided into four sub-blocks, each
MEMORY-GUIDED SELECTIVE ATTENTION
19
consisting of four trials (referred to as the Prime A, Prime B, Probe A, and Probe B sub-
blocks). The images presented in the Prime A sub-block were then repeated in the
Probe A sub-block and images presented in the Prime B sub-block, repeated in the
Probe B sub-block. In Experiment 2, there were two blocks of trials, each consisting of
160 trials. The images presented in the Prime block were then repeated in the Probe
block.
Results
Participants with mean error rates greater than 20% were excluded from the
analyses. For experiment 1A, this eliminated five participants, for 1B this eliminated
seven participants, and for 1C this eliminated four participants. For all remaining
participants, the RTs from correct trials in each condition were submitted to an outlier
removal procedure (the non-recursive procedure; Van Selst & Jolicoeur, 1994) that
eliminated an average of 3.58%, 3.53%, and 3.11% of the observations from
experiments 1A, 1B, and 1C respectively.
Long-term congruency sequence effects. The primary question of interest was
whether the repetition of unique stimuli after a single presentation (trial n-5 to n-11)
would produce sequential-like effects. To address this question, mean RTs from correct
responses on the probe trials and error rates were submitted to a mixed analysis of
variance (ANOVA) with prime congruency (congruent vs. incongruent) and probe
congruency (congruent vs. incongruent) as within-subject factors, and experiment (1A,
1B, and 1C) as the between-subject factor.
The results of the RT analysis revealed a significant two-way interaction between
prime congruency and probe congruency, F(1,114) = 10.05, MSE = 1508.82, p = .002,
MEMORY-GUIDED SELECTIVE ATTENTION
20
ηp2 = .08, demonstrating a smaller congruency effect when the prime stimulus was
incongruent rather than congruent. Furthermore, the three-way interaction between
prime congruency, probe congruency, and experiment, was non-significant, F(2,114) =
.11, MSE = 1508.82 p = .90, ηp2 = .002, showing no significant difference between the
size or direction of the long-term sequence effects across experiments.
The results of the error analysis revealed no significant effects of interest. The
three-way interaction between experiment, prime congruency, and probe congruency
was non-significant, F(1,114) = .48, MSE = 11.17, p = .62, ηp2 = .008, and the two-way
interaction between prime congruency and probe congruency was non-significant,
F(1,114) = 1.39, MSE = 11.17, p = .24, ηp2 = .01. Average error rates from experiments
1A, 1B, and 1C (probe trials only), were 4.38%, 3.29%, and 2.9% respectively.
N-1 congruency sequence effects. In our experimental design, specific stimuli
never repeated trial-to-trial. Another question of interest was whether this design would
still produce n-1 sequence effects when using non-repeating stimuli. Some previous
work has demonstrated that sequence effects were eliminated when contextual features
alternate rather than repeat (Spapé & Hommel, 2008) whereas other studies using non-
repeating stimuli have successfully produced sequential effects (Egner, 2010; King,
Korb, & Egner, 2012). To address this question, mean RTs from correct responses and
error rates were submitted to a mixed analysis of variance (ANOVA) with trial n-1
congruency (congruent vs. incongruent) and trial n congruency (congruent vs.
incongruent) as within-subject factors, and experiment (1A, 1B, and 1C) as the
between-subject factor (see Figure 2).
MEMORY-GUIDED SELECTIVE ATTENTION
21
Two results of the RT analysis are of particular interest. First, the two-way
interaction between trial n congruency and experiment was significant, F(2,114) = 11,
MSE = 1743.37, p = .04, ηp2 = .06, suggesting the size of the congruency effect differed
across experiments. Specifically, the congruency effect was smallest in experiment 1A
(M = 33 ms), then experiment 1C (M = 50 ms), and largest in experiment 1B (M = 60
ms).
Second, the critical two-way interaction between trial n-1 congruency and trial n
congruency was significant, F(1,114) = 11, MSE = 1023.99, p = .001, ηp2 = .09 showing
a smaller congruency effect when trial n-1 was incongruent compared to congruent.
However, this interaction was qualified by a significant three-way interaction between
trial n-1 congruency, trial n congruency, and experiment, F(2,114) = 3.65, MSE =
1023.99, p = .03, ηp2 = .06.
To further probe the three-way interaction, we analyzed each of the experiments
separately. The analysis of experiment 1A resulted in no significant interaction between
trial n-1 congruency and trial n congruency, F(1,34) < .01, MSE = 1364.48, p = .79, ηp2
< .01, suggesting no sequence effects. However, there were significant two-way
interactions between trial n-1 congruency and trial n congruency for both experiment
1B, F(1,31) = 8.56, MSE = 1226.6, p = .006, ηp2 = .22, and experiment 1C, F(1,49) =
13.87, MSE = 659.55, p < .001, ηp2 = .22, showing a smaller congruency effect following
incongruent rather than congruent trials.
The results of the error analysis revealed no significant effects of interest. The
three-way interaction between experiment, prime congruency, and probe congruency
was non-significant, F(1,114) = .97, MSE = 7.1, p = .38, ηp2 = .02, and the two-way
MEMORY-GUIDED SELECTIVE ATTENTION
22
interaction between trial n-1 congruency and trial n congruency was non-significant,
F(1,114) = .91, MSE = 7.1, p = .34, ηp2 = .008. Average error rates from experiments
1A, 1B, and 1C, were 3.62%, 3.18%, and 2.67% respectively.
Figure 2. Results of Experiment 1. Figure 2A shows congruency effects in reaction
times as a function of Prime Congruency (congruent and incongruent) and Experiment
(A, B, and C). Figure 2B shows congruency effects in reaction times as a function of
Trial N-1 Congruency (congruent and incongruent) and Experiment (A, B, and C). Error
bars represent the Standard Error of the Mean (SEM).
Discussion
Across three replications, we found that a single experience with a unique
stimulus could influence performance 5 to 11 trials after the initial presentation.
Specifically, we consistently found smaller congruency effects for the probe when the
first prime presentation was incongruent as compared to congruent, demonstrating a
long-term congruency sequence effect. This result is consistent with the instance-based
MEMORY-GUIDED SELECTIVE ATTENTION
23
memory account suggesting that contextual features (image identity) could cue the
rapid adjustment of attentional priorities after only a single prior presentation.
An unlikely, but alternative interpretation is that the decaying control signal
carried forward over trials from the first presentation to influence the second. N-1
congruency sequence effects are often interpreted as the result of control settings or
conflict signals from trial n-1 carrying forward to influence trial n. Various studies have
shown that sequence effects, given the right conditions, can persist longer than one
trial, from two to four trials (Jiménez & Méndez, 2013; Mayr et al., 2003), and up to five
seconds (Egner, 2010) after the initial presentation. On the one hand, this interpretation
seems unlikely given the intervening length in our experiments was much longer than
previous demonstrations. On the other hand, the rate of decay is not well understood
and certainly conflict-monitoring models are flexible in terms of the speed of decay (e.g.,
Botvinick et al., 2001; Braver, 2012). Additionally, there is evidence that under some
conditions the rate of decay could be slowed. For example, one study demonstrated
that the use of proactive strategies could prevent the sequence effect from decaying as
rapidly as previously demonstrated (Duthoo et al., 2014). It is possible that the use of
contextual cues combined with the frequency and regularity by which they repeated
created some expectation for when contextual cues would repeat. This may have
promoted the use of proactive strategies that slowed the decay rate long enough to
produce our long-term sequence effect.
An additional consideration is whether the secondary task influenced
performance in experiments 1B and 1C. The secondary task had participants monitor
the flanking images and press the spacebar when the flanking images differed in
MEMORY-GUIDED SELECTIVE ATTENTION
24
identity to the target image. The use of contextual cues in attention tasks is often
thought to develop automatically (e.g., Chun & Jiang, 1998). However, there have been
demonstrations using proportion congruent designs where context-dependency fails to
develop without specific task instructions to engage in specific strategies (e.g.,
Brosowsky & Crump, 2016; Crump et al., 2008). The nature of our secondary task could
have caused participants to attend more to the identities of the images and encouraged
the use of contextual cues. Regardless, the long-term congruency sequence effect was
also found in experiment 1A where participants did not have the secondary task. So,
although it may be possible that the secondary task contributed to the effects in
experiments 1B and 1C, removing the secondary task was not sufficient for eliminating
the long-term sequence effect.
Finally, experiments 1B and 1C included a secondary task to increase the
amount of conflict, as measured by the congruency effect. Consistent with that
manipulation, we found a smaller congruency effect in experiment 1A as compared to
1B and 1C. However, the long-term sequence effect appeared to be insensitive to the
conflict manipulation as we found no significant differences in the size of the long-term
sequence effect across experiments. In contrast, we only found n-1 congruency
sequence effects in experiments 1B and 1C. These findings are consistent with prior
work demonstrating the n-1 sequence effect despite the use of non-repeating stimuli
(Egner, 2010; King, Korb, & Egner, 2012), and consistent with prior work showing a
sensitivity to the amount of conflict (Forster et al., 2011; Wendt et al., 2015).
Experiment 2A and 2B
MEMORY-GUIDED SELECTIVE ATTENTION
25
In experiment one, across three replications, we found long-term congruency
sequence effects when there were 5 to 11 intervening trials between the first and
second presentation of a unique stimulus. The goals of experiment two were to
conceptually replicate and extend the findings from experiment one by increasing the
number of intervening trials between the prime and probe pairs, increasing the
variability in the frequency of stimulus repetition, and including an alternate conflict
manipulation.
For both experiments 2A and 2B, the primary task was the same as experiment
one which involved identifying the color of a central image (either blue or green) flanked
on the left and right by the same image presented in either the same (congruent) or the
alternate color (incongruent). Each image was only presented once as a prime stimulus,
and once as a probe stimulus. Importantly, the experiment consisted of two blocks of
160 trials; A prime block followed by a probe block. Each block was randomized such
that the distance between any given prime and probe stimulus ranged from 1 to 319
trials (160 trials, on average). To increase conflict in experiment 2B, the flanking images
preceded the target image by 100 ms, a manipulation known to increase the
congruency effect (Wendt et al., 2015).
Methods
Participants. All participants were recruited from Amazon Mechanical Turk
(AMT) and compensated $2.00 for participating. The amount compensated was
calculated by estimating the maximum amount of time required to complete each
experiment and multiplying by $6.00 per hour. For each experiment the number of HITs
(Human intelligence tasks, an Amazon term for a work-unit) refers to the number of
MEMORY-GUIDED SELECTIVE ATTENTION
26
participants who initiated the study and each experiment consisted of unique
participants. Participants were included in the study if they completed all trials. For
experiment 2A, 40 HITs were posted, and 39 participants completed all trials. For
experiment 2B, 40 HITs were posted, and 40 participants completed all trials.
Apparatus & Stimuli. The apparatus and stimuli were identical to those used in
experiment 1.
Design. Experiment 2 used a 2x2x2 mixed design with prime congruency
(congruent vs. incongruent) and probe congruency (congruent vs. incongruent) as
within-subject factors, and experiment (2A and 2B) as the between-subject factor.
Experiments 2A and 2B were both constructed using the same general method.
Both experiments consisted of 320 total trials divided into two halves, a prime block and
probe block. The prime block was constructed using 160 unique images randomly
selected for each participant from the total 540 images (Brady et al., 2013). The images
presented in the prime block were then repeated in the probe block. The trial order for
each block was randomized, so the distance between any given probe (trial n) and
prime stimulus paired ranged from n-1 to n-319. Each experiment consisted of 50%
congruent/incongruent trials, an equal number of each congruency combination
between prime/probe pairs (i.e., Con – Con, Con – Inc, Inc – Inc, and Inc – Con), and
an equal number of response repetition and response alternation prime/probe pairs.
Procedure. All participants were AMT workers who found the experiment using
the AMT system. The participant recruitment procedure and tasks were approved by the
Brooklyn College Institutional Review Board. Each participant read a short description of
the task and gave consent by pressing a button acknowledging they had read the
MEMORY-GUIDED SELECTIVE ATTENTION
27
displayed consent form. Participants then completed a short demographic survey, and
proceeded to the main task, which was displayed as a pop-up window. Participants
were instructed to identify the color of the center image on each trial as quickly and
accurately as possible by pressing ‘g’ if the image was green, and ‘b’ if the image was
blue. Throughout the course of the experiment the upper left corner of the display
indicated the number of completed and remaining trials, as well as an instruction
reminder button that displayed the instructions in a new pop-up window.
For experiment 2A, each trial began with a fixation cross presented in the center
of the screen for 1,000 ms, followed by a blank ISI of 250 ms. Next, the flanker stimulus
appeared in the center of screen, and remained on screen until a response was made.
Feedback indicating whether the answer was correct or incorrect was presented above
the target stimulus following a response and remained on-screen for 500 ms which
automatically triggered the next trial. For experiment 2B, each trial began with a fixation
cross presented in the center of the screen for 1,000 ms, followed by a blank ISI of 250
ms. Next, the flanking images appeared for 100 ms followed by the presentation of the
center image. All images remained on screen until a response was given. Feedback
indicating whether the answer was correct or incorrect was presented above the target
stimulus following a response and remained on-screen for 500 ms which automatically
triggered the next trial. In both experiments, after every 80 trials, a message appeared
on-screen that instructed participants to take a short break and to press the button when
they were ready to continue.
Results
MEMORY-GUIDED SELECTIVE ATTENTION
28
Participants with mean error rates greater than 20% were excluded from the analyses.
For experiment 2A, this eliminated three participants and for 2B this eliminated two
participants. For all remaining participants, the RTs from correct trials in each condition
were submitted to an outlier removal procedure (the non-recursive procedure; Van Selst
& Jolicoeur, 1994) that eliminated an average of 3.2% and 3.3% of the observations
from experiments 2A and 2B, respectively.
Long-term congruency sequence effects. Mean RTs from correct responses
on probe trials and error rates were submitted to a mixed analysis of variance (ANOVA)
with prime congruency (congruent vs. incongruent) and probe congruency (congruent
vs. incongruent) as within-subject factors, and experiment (2A and 2B) as the between-
subject factor (see Figure 3).
The results of the RT analysis revealed a significant two-way interaction between
prime congruency and probe congruency, F(1,72) = 6.99, MSE = 980.13, p = .01, ηp2 =
.09, demonstrating a smaller congruency effect when the prime stimulus was
incongruent rather than congruent. Additionally, the three-way interaction between
prime congruency, probe congruency, and experiment, was non-significant, F(1, 72) =
.09, MSE = 980.13, p = .77, ηp2 = .001, showing no difference between the size or
direction of the long-term sequence effects across experiments.
The results of the error analysis revealed no significant effects of interest. The
three-way interaction between experiment, prime congruency, and probe congruency
was non-significant, F(1,72) = 1.23, MSE = 6.77, p = .27, ηp2 = .02, and the two-way
interaction between prime congruency and probe congruency was non-significant,
MEMORY-GUIDED SELECTIVE ATTENTION
29
F(1,72) = .09, MSE = 6.77, p = .76, ηp2 = .001. Average error rates from experiments 2A
and 2C (probe trials only), were 3.52% and 3.17% respectively.
N-1 congruency sequence effects. Mean RTs from correct responses and
mean error rates were submitted to a mixed analysis of variance (ANOVA) with trial n-1
congruency (congruent vs. incongruent) and trial n congruency (congruent vs.
incongruent) as within-subject factors, and experiment (2A and 2B) as the between-
subject factor (see Figure 3).
The RT analysis resulted in a significant two-way interaction between trial n
congruency and experiment, F(1,72) = 9.9, MSE = 851.61, p = .002, ηp2 = .12; The size
of the congruency effect was significantly smaller in experiment 2A (M = 33 ms), as
compared to experiment 2B (M = 54 ms).
The critical two-way interaction between trial n-1 congruency and trial n
congruency was also significant, F(1,72) = 15.92, MSE = 481.26, p < .001, ηp2 = .18,
demonstrating a smaller congruency effect when trial n-1 was incongruent rather than
congruent. However, this was qualified by a three-way interaction between trial n-1
congruency, trial n congruency, and experiment, F(1,72) = 5.99, MSE = 481.26, p = .02,
ηp2 = .08.
A separate analysis of experiment 2A showed no significant interaction between
trial n-1 congruency and trial n congruency, F(1,35) = .91, MSE = 719.14, p = .35, ηp2 =
.03. However, the analysis of experiment 2B showed a significant two-way interaction,
F(1,37) = 28.87, MSE = 355.08, p < .0001, ηp2 = .44, with a smaller congruency effect
when trial n-1 was incongruent rather than congruent.
MEMORY-GUIDED SELECTIVE ATTENTION
30
The results of the error analysis revealed a significant two-way interaction
between trial n-1 congruency and trial n congruency, F(1,72) = 13.69, MSE = 4.35, p <
.001, ηp2 = .16, showing a larger congruency effect following a congruent (M = 2.37%),
as compared to an incongruent trial (M = 0.57%). However, the three-way interaction
between experiment, trial n-1 congruency, and trial n congruency was non-significant,
F(1,72) < .01, MSE = 4.35, p = .98, ηp2 < .0001. Average error rates from experiments
2A and 2C, were 3.53% and 3.3% respectively.
Figure 3. Results of Experiment 2. Figure 3A shows congruency effects in reaction
times as a function of Prime Congruency (congruent and incongruent) and Experiment
(A and B). Figure 3B shows congruency effects in reaction times as a function of Trial
N-1 Congruency (congruent and incongruent) and Experiment (A and B). Error bars
represent the Standard Error of the Mean (SEM).
MEMORY-GUIDED SELECTIVE ATTENTION
31
Discussion
The critical result in experiment two was that congruency effects were
significantly smaller on probe trials paired with an incongruent as compared to
congruent prime trial. Experiment 2 therefore conceptually replicates experiment 1, and
demonstrates long-term congruency sequence effects with 1 to 319 intervening trials,
increased variability in the frequency of stimulus repetition, and an alternate conflict
manipulation.
Additionally, the level of conflict was manipulated across experiments. Consistent
with our manipulation, the congruency effect was significantly larger in experiment 2B
as compared to 2A. However, this manipulation did not modulate the size or direction of
the long-term congruency sequence effect. In contrast, we only found n-1 congruency
sequence effects in experiment 2B, suggesting a sensitivity to the level of conflict, and
replicating the results of experiment 1.
Experiment 3A and 3B
Across experiments 1 and 2, we have demonstrated long-term congruency
sequence effects with as many as 160 intervening trials between the first and second
presentation of a unique stimulus. However, both experiments used a 2-choice flanker
task resulting in some feature-overlap between the prime and probe trial. Feature
integration accounts have proposed that differences in match between features
presented on trial n-1 and trial n could account for congruency sequence effects by way
of event files and a memory retrieval process (Hommel, 1998; Hommel et al., 2001;
Hommel et al., 2004). This issue will be discussed in greater detail in the general
MEMORY-GUIDED SELECTIVE ATTENTION
32
discussion, however, the goal of experiment 3 was to test whether the long-term
congruency effect would persist when the prime and probe trials consist entirely of non-
overlapping color features.
For both experiments 3A and 3B, the primary task was the same as experiments
1 and 2 identifying the color of a central image flanked on the left and right by the same
image presented in either the same (congruent) or the alternate color (incongruent).
Each image was only presented once as a prime stimulus, and once as a probe
stimulus. However, in contrast to experiments 1 and 2, images could appear in one of
four colors (red, blue, green, or yellow). For each participant, colors were randomly
assigned to two mutually exclusive color sets. Each prime/probe stimulus pair used both
color sets ensuring that colors did not overlap between the prime and probe trial.
Except for the image colors, experiment 3A followed the same methods as
experiment 1A such that the distance between any given prime and probe stimulus
ranged from 5 to 11 trials (8 trials, on average). Similarly, experiment 3B followed the
same methods as experiment 2A such that the distance between any given prime and
probe stimulus ranged from 1 to 319 (160 trials, on average).
Methods
Participants. All participants were recruited from Amazon Mechanical Turk
(AMT) and compensated $1.00 for participating. The amount compensated was
calculated by estimating the maximum amount of time required to complete each
experiment and multiplying by $6.00 per hour. For each experiment the number of HITs
(Human intelligence tasks, an Amazon term for a work-unit) refers to the number of
participants who initiated the study and each experiment consisted of unique
MEMORY-GUIDED SELECTIVE ATTENTION
33
participants. Participants were included in the study if they completed all trials. For
experiment 3A, 50 HITs were posted, and 50 participants completed all trials. For
experiment 3B, 50 HITs were posted, and 47 participants completed all trials.
Apparatus & Stimuli. The apparatus and stimuli were identical to those used in
experiments 1 and 2.
Design. Experiment 3 used a 2x2 within-subjects design with prime congruency
(congruent vs. incongruent) and probe congruency (congruent vs. incongruent) as
factors.
Experiment 3A was constructed using the methods as described in experiment 1A.
Therefore, experiment 3A consisted of 192 total trials with the distance between each
prime and probe stimulus pair ranging from n-5 to n-11. Experiment 3B was constructed
using the methods as described in experiment 2. Therefore, experiment 3B consisted of
320 total trials with the distance between each prime and probe stimulus pair ranged
from n-1 to n-319. Each experiment consisted of 50% congruent/incongruent trials, an
equal number of each congruency combination between prime/probe pairs (i.e., Con –
Con, Con – Inc, Inc – Inc, and Inc – Con). Images were randomly selected for every
participant from the total 540 images (Brady et al., 2013) and randomly assigned a color
and condition. Each image was only presented twice during the experiment: once in a
prime block and once in a probe block.
The colors of the images however, differed from experiments 1 and 2. For
experiment 3, images could appear in one of four colors: blue, green, red, or yellow. For
each participant, the four colors were randomly assigned to two color sets (e.g.,
blue/green, red/yellow), such that colors in differing sets were never presented together
MEMORY-GUIDED SELECTIVE ATTENTION
34
on a single trial (e.g., green/yellow never appeared together). Additionally, each
prime/probe pair always consisted of colors from both sets to ensure that colors did not
repeat from the prime to probe trial. The assignment of colors to prime/probe trials was
counterbalanced for each participant. Therefore, on 50% of trials, color set 1 was
assigned to the prime stimuli and color set 2 to the corresponding probe, and on the
other half, color set 2 was assigned to the prime and color set 1 to the probe.
Procedure. The procedure was identical to experiments 1 and 2. However,
because of the use of four colors, participants were instructed to identify the color of the
center image on each trial as quickly and accurately as possible by pressing ‘b’ if the
image was blue, ‘g’ if the image was green, ‘r’ if the image was red, and ‘y’ if the image
was yellow.
Results
Participants with mean error rates greater than 20% were excluded from the
analyses. For experiment 3A, this eliminated six participants and for 3B this eliminated
four participants. For all remaining participants, the RTs from correct trials in each
condition were submitted to an outlier removal procedure (the non-recursive procedure;
Van Selst & Jolicoeur, 1994) that eliminated an average of 3.19% and 2.89% of the
observations from experiments 3A and 3B, respectively.
Experiment 3A: N – 8 trials
Long-term congruency sequence effects. Mean RTs from correct responses
and error rates were submitted to a repeated measures analysis of variance (ANOVA)
with prime congruency (congruent vs. incongruent) and probe congruency (congruent
vs. incongruent) as factors. As a result, the two-way interaction between prime
MEMORY-GUIDED SELECTIVE ATTENTION
35
congruency and probe congruency was non-significant, F(1,43) = .86, MSE = 2119.31,
p = .36, ηp2 = .02, showing no differences between the congruency effects when the
prime was congruent versus incongruent.
The results of the error analysis also revealed no significant effects of interest.
The two-way interaction between prime congruency and probe congruency was non-
significant, F(1,43) = .34, MSE = 9.85, p = .56, ηp2 = .008.
N-1 congruency sequence effects. Mean RTs from correct responses and
mean error rates were submitted to a repeated measures analysis of variance (ANOVA)
with trial n-1 congruency (congruent vs. incongruent) and trial n congruency (congruent
vs. incongruent) as within-subject factors. As a result, the two-way interaction between
trial n-1 and trial n congruency was marginal, though non-significant, F(1,43) = 3.66,
MSE = 1882.08, p = .06, ηp2 = .08, showing no differences between the congruency
effects when trial n-1 was congruent versus incongruent.
The results of the error analysis also revealed no significant effects of interest.
The two-way interaction between trial n-1 congruency and trial n congruency was non-
significant, F(1,43) = .68, MSE = 4.03, p = .41, ηp2 = .02. Average error rates were
2.36%.
Experiment 3B: N – 160 trials
Long-term congruency sequence effects. Mean RTs from correct responses
and mean error rates from probe trials were submitted to a repeated measures analysis
of variance (ANOVA) with prime congruency (congruent vs. incongruent) and probe
congruency (congruent vs. incongruent) as factors (see Figure 4). As a result, the two-
way interaction between prime congruency and probe congruency was non-significant,
MEMORY-GUIDED SELECTIVE ATTENTION
36
F(1,42) < .01, MSE = 973.11, p = .96, ηp2 < .0001, showing no differences between the
congruency effects when the prime was congruent versus incongruent.
The results of the error analysis also revealed no significant effects of interest.
The two-way interaction between prime congruency and probe congruency was non-
significant, F(1,42) = .21, MSE = 8.48, p = .65, ηp2 = .005.
N-1 congruency sequence effects. Mean RTs from correct responses and
mean error rates were submitted to a repeated measures analysis of variance (ANOVA)
with trial n-1 congruency (congruent vs. incongruent) and trial n congruency (congruent
vs. incongruent) as within-subject factors. As a result, the two-way interaction between
trial n-1 and trial n congruency was non-significant, F(1,42) = 2.06, MSE = 1273.95, p =
.16, ηp2 = .05, showing no differences between the congruency effects when trial n-1
was congruent versus incongruent.
Similarly, the results of the error analysis also revealed no significant effects of
interest. The two-way interaction between trial n-1 congruency and trial n congruency
was non-significant, F(1,42) = .26, MSE = 4.28, p = .61, ηp2 = .006. Average error rates
were 2.58%.
MEMORY-GUIDED SELECTIVE ATTENTION
37
Figure 4. Results of Experiment 3. Figure 4A shows congruency effects in reaction
times as a function of Prime Congruency (congruent and incongruent) and Experiment
(A and B). Figure 4B shows congruency effects in reaction times as a function of Trial
N-1 Congruency (congruent and incongruent) and Experiment (A and B). Error bars
represent the Standard Error of the Mean (SEM). In both experiments, the interaction
was non-significant (p > .05).
Discussion
The critical result of experiment three was the failure to find long-term
congruency sequence effects. Experiment 3, therefore failed to replicate experiments 1
and 2 when color features did not overlap between prime and probe stimuli. A positive
finding would have convincingly ruled out a potential long-term feature integration
account of the findings from experiments 1 and 2. However, the failure to find the effect
is more ambiguous. Any theory that relies on memory-retrieval might make the
prediction that decreasing the similarity between the prime and probe could diminish or
MEMORY-GUIDED SELECTIVE ATTENTION
38
eliminate the effect because the probe is no longer an adequate retrieval cue.
Therefore, the finding in experiment 3 does not provide direct evidence for a long-term
feature integration account though it does fail to rule out such a possibility. These issues
are discussed in greater detail in the general discussion.
Stimulus-Response Repetition Analyses
Previous work has demonstrated that stimulus-response repetition biases
confounded with congruency manipulations may contribute to, or account entirely for,
sequential modulations of congruency effects (e.g., Mayr et al., 2003). This issue will be
discussed in more detail in the general discussion, however to determine the
contribution of repetition biases, we compared response repeat to response change
trials for both the long-term and n-1 congruency sequence effects.
Long-Term Congruency Sequence Effects
Collapsing across experiments 1 and 2, mean RTs from correct probe trials were
submitted to a repeated-measures analysis of variance (ANOVA) with response (repeat
vs. change), prime congruency (congruent vs. incongruent), and probe congruency
(congruent vs. incongruent) as factors (see Figure 5).
There was a significant main effect of response showing speeded responses for
response repeat versus change trials, F(1,190) = 40.25, MSE = 3536.39, p < .0001, ηp2
= .17. The two-way interaction between prime congruency and probe congruency was
also significant showing a smaller congruency effect when the prime trial was
incongruent as compared to congruent, F(1,190) = 16.09, MSE = 2582.77, p < .0001,
ηp2 = .08. Critically, three-way interaction between response, prime congruency, and
MEMORY-GUIDED SELECTIVE ATTENTION
39
probe congruency was non-significant, F(1,190) = .14, MSE = 1717.33, p = .71, ηp2 <
.001.
Therefore, although we found an overall long-term response priming effect, we
found no significant difference between the long-term congruency sequence effects for
response repeat versus change trials.
N-1 Congruency Sequence Effects
Collapsing across experiments 1 and 2, mean RTs from correct trials were
submitted to a repeated measures analysis of variance (ANOVA) with response (repeat
vs. change), trial n-1 congruency (congruent vs. incongruent), and trial n congruency
(congruent vs. incongruent) as factors (see Figure 5). One participant was removed
prior to the analysis due to missing data in one condition.
There was a significant main effect of response showing speeded responses for
response repeat versus change trials, F(1,189) = 28.66, MSE = 3892.71, p < .0001, ηp2
= .13. The two-way interaction between trial n-1 congruency and trial n congruency was
significant showing a smaller congruency effect when the prime trial was incongruent as
compared to congruent, F(1,189) = 28.99, MSE = 2055.37, p < .0001, ηp2 = .13.
However, the critical three-way interaction between response, trial n-1 congruency, and
trial n congruency was also significant, F(1,189) = 42.15, MSE = 1741.5, p < .0001, ηp2
= .18.
To probe the three-way interaction, the response change and repeat trials were
analyzed separately. The analysis of the change trials revealed no significant interaction
between trial n-1 congruency and trial n congruency, F(1,189) = .18, MSE = 1953.09, p
= .67, ηp2 = .001. However, the analysis of the repeat trials revealed a significant
MEMORY-GUIDED SELECTIVE ATTENTION
40
interaction with a smaller congruency effect when trial n-1 was incongruent as
compared to congruent, F(1,189) = 71.94, MSE = 1843.78, p < .0001, ηp2 = .28.
Therefore, the n-1 congruency sequence effect was only found for response repeat and
not for response change trials.
Figure 5. Results of the stimulus-response repetition analyses. Figure 5A shows
congruency effects in reaction times collapsed across all experiments as a function of
Prime Congruency (congruent and incongruent) and Response (change and repeat).
Figure 5B shows congruency effects in reaction times collapsed across all experiments
as a function of Trial N-1 Congruency (congruent and incongruent) and Response
(change and repeat). Error bars represent the Standard Error of the Mean (SEM).
Short- and Long-Term Comparison Analysis
The results thus far suggest that the short- and long-term congruency effects are
the result of independent processes. However, to directly test their independence, we
analyzed the congruency effects on probe trials as a function of the previous trial
MEMORY-GUIDED SELECTIVE ATTENTION
41
congruency. We collapsed across experiments 1 and 2 and submitted the mean RTs
from correct probe trials to a repeated-measures analysis of variance (ANOVA) with
prime congruency (congruent vs. incongruent), trial n-1 congruency (congruent vs.
incongruent), and trial n (probe) congruency (congruent vs. incongruent) as factors. Two
participants were removed prior to the analysis due to missing data.
The critical three-way interaction, was non-significant, F(1,188) = 1.55, MSE =
4320.43, p = .21, ηp2 = .003. Furthermore, the two-way interaction between trial n-1 and
trial n (probe) congruency was significant, F(1,188) = 8.64, MSE = 4566.08 p = .004, ηp2
= .04, and the two-way interaction between prime and trial n (probe) congruency was
also significant, F(1,188) = 6.62, MSE = 3742.42 p = .01, ηp2 = .03. Therefore, the
presence of the long-term congruency sequence effect did not depend on the previous
trial congruency.
General Discussion
The current study investigated whether congruency sequence effects could be
observed on a long-term basis. Across experiments 1 and 2, congruency effects were
significantly smaller on probe trials when the prime was incongruent versus congruent.
Long-term congruency sequence effects were demonstrated with an average of eight
intervening trials (experiment one), and 160 intervening trials (experiment two); and
were observed in both response-repeat and response-change trials. In experiment 3,
when specific colors assigned to prime stimuli were not repeated when presented as
probes, we failed to find long-term congruency sequence effects.
MEMORY-GUIDED SELECTIVE ATTENTION
42
In addition to the long-term congruency sequence effect we also found the
traditional n-1 congruency sequence effect. However, the n-1 sequence effect was only
observed in experiments including a high-conflict manipulation. Last, the n-1
congruency sequence effect was only found for response repeat trials and not for
response change trials, suggesting that these effects could be explained entirely by
stimulus-response repetitions.
The finding that a single presentation of a unique stimulus can influence
performance on the second, much later presentation, is consistent with the instance-
based memory account. Per this account, the attentional priorities adopted in the
presence of unique stimulus features on the prime trial were retrieved and reinstated
when those features were presented again. This work contributes to a small, but
growing body of evidence suggesting that adjustments in cognitive control processes
like attention can be guided by memory representations (Awh et al., 2012; Egner, 2014;
Hutchinson & Turk-Browne, 2012). Memory influences on cognitive control have largely
been studied in the context of negative priming (D’Angelo et al., 2016; Frings et al.,
2015) and visual spatial attention using visual search tasks (e.g., Chun & Jiang, 1998,
2003; Hutchinson & Turk-Browne, 2012). Our findings complement and extend this prior
work demonstrating long-term memory can influence selective attention in a conflict
task.
Although the instance-based memory account provides one explanation for the
current results, we now discuss other accounts of the congruency sequence and
proportion congruent effects and the implication of our findings. Whether modulations of
congruency effects indexes adjustments in cognitive control or instead, emerges from
MEMORY-GUIDED SELECTIVE ATTENTION
43
lower level learning and memory processes, remains one major point of disagreement.
We have organized our discussion around these two perspectives.
Control perspectives
Expectation and voluntary control accounts. The expectation account
postulates that participants develop expectations about the congruency of upcoming
trials and engage in compensatory voluntary control strategies (Gratton et al., 1992;
Logan & Zbrodoff, 1979). For example, the n-1 congruency sequence effect could
reflect deliberate controlled adjustments following participants’ expectation that trial n
will be the same congruency as trial n-1 (Gratton et al., 1992). In proportion congruent
designs, participants could become aware that the trials are mostly congruent or mostly
incongruent and then engage in global control strategies in anticipation of the more
likely item type (Logan & Zbrodoff, 1979).
It is possible that setting voluntary attentional priorities could explain the long-
term congruency effect. For example, when a stimulus is presented, participants might
expect the stimulus will repeat and prepare a strategy for the next time they see that
stimulus. When that stimulus is presented a second time, they would then voluntarily
adopt the prepared strategy. However, this proposal relies on several assumptions that
make such an explanation unlikely.
First, we must assume participants had become aware that each prime stimulus
would be repeated as a probe. Although awareness was not measured in the current
experiments, participants in experiment 1 could easily have noticed that primes are
repeated as probes every five to eleven trials. However, in experiment 2 all the prime
MEMORY-GUIDED SELECTIVE ATTENTION
44
trials were presented first, followed by all the probe trials. Therefore, during the prime
block, participants have no reason to expect stimulus repetition, and during the probe
block, once stimuli start repeating, it would be too late to rely on previously unprepared
stimulus-specific voluntary strategies. Second, participants would have to actively
maintain multiple stimulus-specific strategies simultaneously. In experiment 1, they
would have to maintain 5 to 11 at any given point and in experiment 2 they would need
to maintain 160. Although it is unknown how many stimulus- or context-specific
voluntary strategies can be maintained simultaneously, 160 seems implausible. Finally,
participants would have to rapidly adjust attentional control in a voluntary manner at the
time of stimulus onset. However, voluntary control is traditionally thought of as slow and
effortful (Shiffrin & Schneider, 1977). Taken together, the expectation or voluntary
control account is not a viable explanation of the long-term congruency sequence
effects.
Conflict-monitoring accounts. According to the conflict-monitoring model,
modulations of congruency effects reflect conflict-driven adjustments in cognitive control
(Botvinick et al., 2001). That is, the detection of response-conflict – the simultaneous
activation of competing responses – triggers an up-regulation of cognitive control which
biases attentional priority towards the target dimension and away from the distractor
dimension. Thus, the influence of the distractor dimension is reduced following a high-
conflict, incongruent trial producing the congruency sequence effect.
The conflict-monitoring model and many of its variants (e.g., Botvinick, 2007;
Braver, 2012; Dreisbach & Fischer, 2012; Egner, 2008; Hazeltine, Lightman, Schwarb,
& Schumacher, 2011; Jiang et al., 2014) are incapable of explaining the long-term
MEMORY-GUIDED SELECTIVE ATTENTION
45
congruency effects for the same reasons that they have difficulties explaining item- and
context-specific proportion congruent effects. Namely, adjustments in control operate on
task-level representations, pre-specified by the model as relevant versus irrelevant. For
example, in a Stroop task, the detection of conflict would trigger an attentional bias
toward the task-relevant dimension of color, regardless of the item presented. Of
course, item- and context-specific proportion congruent effects, and now the long-term
congruency sequence effect, demonstrate that control can be adjusted differentially for
specific items. Traditional conflict-monitoring models cannot discriminate between items
while detecting conflict or adjusting control, and therefore, cannot produce item-specific
effects.
The adaptation-by-binding account proposes one remedy: aggregate conflict-
driven learning at the level of item features (Blais et al., 2007; Verguts & Notebaert,
2008, 2009). Like other conflict-monitoring accounts, the detection of response conflict
provides a signal that task-relevant connections should be strengthened. In contrast to
the previous models however, this is accomplished through a Hebbian learning rule that
only strengthens connections between active representations. Representations consist
of item-level features (e.g., the color red, the word “BLUE”), and considered active if
they are task-relevant and currently presented. Therefore, item-specific features can
selectively become associated to the current task representation if it is frequently paired
with conflict. Such computational models have been successful in simulating item-
specific proportion congruent effects as well as more generalized forms of the
congruency sequence effects (Blais et al., 2007; Verguts & Notebaert, 2008, 2009).
MEMORY-GUIDED SELECTIVE ATTENTION
46
The adaptation-by-binding account may be able to produce the long-term
congruency sequence effects, but doing so may stretch the model beyond its plausible
limits. For example, it is not clear whether the model could produce single-trial, long-
term learning, or whether repeated presentations are required to produce measurable
changes in performance. Additionally, the irrelevant contextual features that defined the
unique stimuli would need to be considered “task-relevant” by the model for them to be
active during learning. Each unique stimulus would also need to receive its own input
layer in which case the model would require at most 160 input layers. The stimulus-set
is never specified prior to the experiment so we would also have to assume that each
new stimulus presented creates the new required input layers. If we accept these
assumptions, it is possible that the model could produce the long-term congruency
sequence effects. However, it is not clear that this model is compatible with these
assumptions. Furthermore, once these additional assumptions are made, it is not clear
how different this account is from the instance-based memory account.
Conflict-monitoring with memory selection. We propose a new alternative
conflict-monitoring model that could account for the long-term congruency sequence
effect. The congruency task literature has had difficulty explaining why cognitive control
adjustments have been demonstrated to be at times, specific, failing to generalize (e.g.,
item-specific), and at other times, nonspecific, successfully generalizing across stimuli
(for reviews, see Abrahamse et al., 2016; Braem et al., 2014; Egner, 2014). The
conflict-monitor, as specified by traditional accounts (Botvinick et al., 2001; Braver,
2012) can detect response conflict and aggregate recent or frequent conflict within a
conflict-signal, but lacks the ability to select what experiences are aggregated or the
MEMORY-GUIDED SELECTIVE ATTENTION
47
ability to preserve multiple conflict signals. To incorporate the ability to select and store
multiple conflict signals into the conflict monitor is problematic because it would require
the monitor to know in advance, what items should be aggregated over and which
signals preserved (e.g., Egner, 2008).
Instead, we suggest a memory-retrieval process could provide a mechanism by
which prior experiences are selected and then aggregated over by a conflict-monitor.
For example, the word “RED” in blue ink, would cue the retrieval of any other similar
experiences: any trial containing the word “RED” or the color blue. The conflict-monitor
then detects and aggregates the conflict across the retrieved item-set and adjusts
control accordingly. By offloading the selection to a memory system, the conflict-monitor
does not need to distinguish between items and can bias attentional priority along task-
dimensions, as originally specified (Botvinick et al., 2001). However, by allowing
memory to select what prior experiences are evaluated by the conflict-monitor, the
model becomes extremely flexible in determining when control should be adjusted.
Such an account for example, can easily explain item- and context-specific
proportion congruent effects. In a typical item-specific design (e.g., Jacoby et al., 2003),
items are organized into two distinct sets associated with different proportions of
congruency (e.g., the words “RED” and “BLUE” could be high proportion congruent, and
the words “GREEN” and “YELLOW” could be low proportion congruent). Importantly,
the individual features do not overlap between item sets (though, see Bugg &
Hutchison, 2013; Bugg, Jacoby, & Chanani, 2011). For example, if the word “RED” is
the high proportion item set and the word “GREEN” in the low proportion, then the word
and/or color red will never be presented with the word and/or color green. Presenting
MEMORY-GUIDED SELECTIVE ATTENTION
48
“RED” in blue will then only cue the retrieval of items from the high proportion item set.
Similarly, items that appear in one context will be associated to items that have also
appeared in that context by virtue of their shared contextual features (e.g., same
location). Therefore, context-specific effects (Crump et al., 2006), including
generalization to frequency unbiased items (Crump et al., 2017; Crump & Milliken,
2009; Weidler & Bugg, 2016) would also be predicted by such an account. Similarly,
memory retrieval could contribute to traditional congruency sequence effects if we
accept that that more recent memories are more easily retrievable than distant
memories (e.g., Egner, 2014).
There are however, some potential remaining issues. For example, memory-
based theories beg questions about the active features and/or dimensions controlling
memory retrieval. Prior work in the item-specific proportion congruent literature has
suggested that single features, rather than conjunctions of features, drive proportion
congruent effects (e.g., Bugg & Hutchison, 2013; Jacoby et al., 2003). Similarly,
context-specific transfer effects suggest that single, context-features can also drive
memory retrieval (e.g., Crump et al., 2017; Crump & Milliken, 2009; Weidler & Bugg,
2016). In the current study, we found long-term congruency sequence effects when
stimuli shared contextual features. We only found this effect however, when stimuli
appeared in the same color set (experiments 1 and 2). When the prime and probe
appeared in different colors (experiment 3) we failed to find evidence for the long-term
congruency effect. These findings suggests that the conjunction of features may have
served as a retrieval cue in the current study. A task for future work is to clarify the
conditions that enable a feature or conjunction of features to drive retrieval.
MEMORY-GUIDED SELECTIVE ATTENTION
49
Non-control perspectives
Although the control perspective remains popular, several accounts have
challenged the underlying premise that modulations of the congruency effect index
cognitive control adjustments. Instead, some have argued that learning and memory
processes could produce the same effects without the need for notions of conflict-driven
control (Mayr et al., 2003; Schmidt, 2013; Schmidt & Besner, 2008). For example, many
congruency task designs contain item- or feature-repetition biases that are confounded
with congruency manipulations. These confounds have produced alternative
explanations of congruency phenomena, two of which are of importance to the current
study.
Contingency learning. First, the contingency learning account suggests that the
frequency of item presentation can produce predictive relationships between item
features and responses. Responses are thought to be speeded for stimuli that contain
features that are highly predictive of a response regardless of congruency. For example,
if the word “BLUE” is most often presented in red ink, then the word “BLUE” becomes
predictive of the red response, and responses would be quicker relative to non-
predictive items. In many proportion congruent designs, the frequency of item
presentation is confounded with the proportion congruent manipulations. Therefore, the
contingency learning account has been sufficient for explaining many proportion
congruent effects, although still unable to explain transfer to frequency unbiased items
(Crump et al., 2017; Weidler & Bugg, 2016). In the current study, however, we used a
two-choice flanker task and all potential contingencies were held constant. That is, there
MEMORY-GUIDED SELECTIVE ATTENTION
50
were no predictive relationships between stimulus features and responses that could
explain our results.
Stimulus-specific repetition priming and feature integration. The stimulus-
specific repetition priming account was proposed to explain congruency sequence
effects. Mayr, Awh, and Laurey (2003) noted that the frequency of complete stimulus-
response repetitions in trial-to-trial transitions are unbalanced in two-choice congruency
tasks. Specifically, some congruent-to-congruent and incongruent-to-incongruent
transitions contain complete stimulus-response repetitions which could speed
responses selectively for those conditions (Hommel, 1998; Pashler & Baylis, 1991).
Consistent with this proposition, Mayr et al. found that the congruency sequence effect
disappeared when response repetition trials were either removed from the analysis, or
prevented from occurring in the trial sequence. Though, many studies have now
demonstrated congruency sequence effects while controlling for stimulus-response
repetitions suggesting stimulus-response repetitions cannot entirely account for
sequential effects (Akçay & Hazeltine, 2007; Kerns et al., 2004; Kunde & Wühr, 2006;
Ullsperger, Bylsma, & Botvinick, 2005; Weissman, Jiang, & Egner, 2014).
To determine whether stimulus-response repetitions played a role in producing
the current result we compared response repeat to response change trials. We found an
overall long-term response repetition effect, in that performance was facilitated when
responses repeated from prime to probe trials. However, the size of the long-term
congruency sequence effect did not differ between response repeat and change trials
suggesting that the current result could not be explained by stimulus-response
repetitions.
MEMORY-GUIDED SELECTIVE ATTENTION
51
The feature integration account makes a similar proposal. According to this
account, stimuli and responses that co-occur in time are bound together in a common
episodic memory representation called an event file (Hommel, 1998; Hommel et al.,
2001, 2004); a more general form of the “object file” proposed by Kahneman, Treisman,
and Gibbs (1992). The subsequent re-occurrence of any features automatically
retrieves the entire event file which could either help or hinder performance depending
on the match between the currently presented features and the features contained in
the event file. Across two consecutive trials, features are either completely matched,
partially matched, or completely mismatched. Critically, an effortful “unbinding” process
is necessary whenever features are partially matched. That is, feature representations
must be unbound from the associated event file so that they can be re-used in the
creation of a new event file. Therefore, performance would be predicted to be slowed on
partial match trials relative to complete match or complete mismatch trials. In many
congruency task designs, feature overlap is confounded with congruency sequences
and as such, feature integration can explain trial-to-trial effects in many cases (for
reviews, see Egner, 2007, 2014). Similarly, experiments 1 and 2, we utilized a two-
choice flanker task that contains these same feature overlap confounds.
Event files are typically referred to as “transient” or “temporary” memory
structures (Hommel, 1998; Hommel et al., 2001; 2004), and are generally invoked to
explain short-term, trial-to-trial effects (Egner, 2007; 2014). However, event files are
based on instance-based memory theories (Hintzman, 1986; Logan, 1988, 1990), and
typically, the timescale is not explicitly defined. If we assume that event files are stable
episodic memory structures, then feature integration, like the other memory-retrieval
MEMORY-GUIDED SELECTIVE ATTENTION
52
accounts proposed above, could also explain long-term congruency sequence effects.
The evidence for long-term feature integration across our experiments however, is
mixed and largely inconclusive.
On the one hand, across experiments one and two we found long-term
congruency sequence effects for response-change trials. This result is generally
inconsistent with feature integration theories. One possibility, as others have suggested,
is that event files may not be limited to stimulus-response associations. That is, other
aspects of an experience like perceived conflict and control processes may also be
encoded in the event file (Bugg & Hutchison, 2013; Spapé & Hommel, 2008). We could
speculate that the added contextual features could have provided additional support for
event file retrieval, even in the absence of response repetitions (for a similar proposal,
see Spapé & Hommel 2008), or perhaps response outcomes are forgotten more rapidly
than degree of conflict. On the other hand, in experiment three we failed to find long-
term congruency sequence effects when colors did not repeat from prime to probe trials.
This result is consistent with feature integration theory. Although our speculation about
why we found long-term effects in response-change trials could have also applied here,
so these two results are at odds. Furthermore, any memory-based explanation might
predict that altering the similarity between the prime and probe trials would influence the
long-term effects. Therefore, the failure to find long-term effects when colors do not
repeat, does not allow us to discriminate between any of the memory-based theories we
have proposed. Finally, to the extent that you allow event files to be permanent memory
representations and allow them to encode many aspects of our experience like stimulus
and context features, responses, perceived conflict, and control processing, it is not
MEMORY-GUIDED SELECTIVE ATTENTION
53
clear how different feature integration theories are from instance-based memory
theories.
Perceptual learning and attentional control. Finally, we propose an alternative
non-control account. Perceptual learning refers to experience-dependent changes in
perception and is thought to reflect perceptual or neural plasticity in visual
representations (Goldstone, 1998; Lu, Hua, Huang, Zhou, & Dosher, 2011; Roelfsema,
van Ooyen, & Watanabe, 2010; Sasaki, Nanez, & Watanabe, 2010). Perceptual
learning has been demonstrated across a wide variety of perceptual tasks including the
discrimination and detection of stimulus orientation (Dosher & Lu, 1998; Shiu & Pashler,
1992; Vogels & Orban, 1985), motion direction (Ball & Sekuler, 1987; Ball, Sekuler, &
Machamer, 1983), and object recognition (Furmanski & Engel, 2000), to name a few (for
a review, see Watanabe & Sasaki, 2015). Importantly, across tasks, selective attention
has been shown to influence perceptual learning, such that learning is enhanced for
task-relevant, or attended features as compared to irrelevant, unattended features
(Ahissar & Hochstein, 1993; Gutnisky, Hansen, Iliescu, & Dragoi, 2009; Shiu & Pashler,
1992; Szpiro & Carrasco, 2015).
One possible explanation of the long-term congruency sequence effect is that
selective attention influences perceptual learning on the first presentation, which in-turn,
influences how the stimulus is attended on the second presentation. For example, when
presented with an incongruent stimulus, attention is shifted towards the target and away
from the flankers facilitating perceptual learning of the target features relative to the
flanker features. On the second presentation, the altered visual representation and
MEMORY-GUIDED SELECTIVE ATTENTION
54
enhanced perceptual processing of the target could cue attention towards the target,
facilitating performance if the second presentation is incongruent.
There is some evidence that increased selective attention demands from
incongruent stimuli may enhance target representations on a long-term basis. As noted
earlier, in perceptual learning tasks, learning is enhanced for attended versus
unattended features (Ahissar & Hochstein, 1993; Gutnisky, Hansen, et al., 2009; Shiu &
Pashler, 1992; Szpiro & Carrasco, 2015). However, recognition memory has also been
shown to be improved for items previously presented with incongruent versus congruent
distractors. Here, the increased need for cognitive control is thought to facilitate target
encoding at the time of study improving later memory recognition (Krebs, Boehler, De
Belder, & Egner, 2013; Rosner, D’Angelo, MacLellan, & Milliken, 2015; Rosner &
Milliken, 2015).
Perceptual learning could provide an important mechanism for informing how
attention changes through experience and learning. Perceptual learning has been
shown to be highly stimulus-specific and produces long-lasting effects (Watanabe &
Sasaki, 2015). Though in contrast to the immediate effects in our study, measuring
changes in performance on perception tasks often requires extensive training (Dosher &
Lu, 1999). Therefore, changes in perception alone could not account for the current
results. Instead, we are suggesting that small changes in perceptual representations
could help guide attention, causing more immediate and measurable effects in attention
tasks.
MEMORY-GUIDED SELECTIVE ATTENTION
55
Short- and long-term congruency sequence effects: Single or multiple
processes?
In the current experiments, we found both short- (n-1) and long-term congruency
sequence effects. All current accounts of the short-term congruency sequence effects
posit rapidly decaying representations and are generally incapable of accounting for the
long-term congruency sequence effects (Egner, 2007). The alternative accounts we
have proposed above however, could accommodate both long- and short-term effects.
For example, if memory retrieval mediates shifts in attentional control then we might
expect that similarity in temporal context could cue retrieval (e.g., Egner, 2014). One
possibility is that both short- and long-term effects are produced via a single, memory-
driven process. Alternatively, we might speculate the contribution from two independent
processes: A memory-driven process and a short-term, conflict-driven process.
On the one hand, there were some clear differences between the short- and
long-term effects found in the current study. First, the long-term effects were insensitive
to increased response conflict, whereas the short-term effects were only present in the
high conflict experiments. This might be expected, if we assume that the short-term
effect reflects changes in the conflict-signal which dissipates over time (e.g., Botvinick,
et al., 2001). Second, the long-term effects were present for both response change and
response repeat trials while the short-term effects were only present for response
repeat trials. As such, the short-term effect could be explained entirely by stimulus-
response repetitions (e.g., Mayr et al., 2003), while the long-term effect cannot. Third,
and most importantly, we found no interaction between the short- and long-term effects.
MEMORY-GUIDED SELECTIVE ATTENTION
56
Taken together, these differences suggest that the two phenomena measured in the
current study do not reflect the same underlying process.
On the other hand, several other studies have reported n-1 congruency
sequence effects while controlling for repetition biases (e.g., Kunde & Wühr, 2006;
Weissman et al., 2014), and others have found n-1 congruency sequence effects to be
insensitive to changes in the degree of conflict (e.g., Weissman & Carp, 2013). One
possible resolution to these inconsistencies, is that a memory-driven process can, under
the right circumstances, contribute to short-term congruency effects. That is, if the
stimuli presented on trial n provides an effect retrieval cue for trial n-1, then we might
expect that a memory-driven process could influence performance on trial n. Of course,
if trial n is a poor retrieval cue for trial n-1, then we might expect no influence from the
memory-retrieval process, leaving only the short-term, conflict-driven effect. In the
current study, we alternated the context trial-to-trial such that the same contextual cue
never repeated from trial n-1 to trial n. Therefore, it could be the case that trial n, while
an effective retrieval cue for the prime trial (e.g., n-8, n-160), was a poor retrieval cue for
trial n-1. As such, the short-term effects observed in the current study reflect only the
influence of a short-term, conflict-driven process, which was sensitive to the degree of
conflict, and response repetition.
Consistent with this interpretation, Spapé and Hommel (2008) found that the n-1
congruency sequence effect was eliminated on trials that alternated contextual cues,
but preserved when contextual cues repeated. The authors suggested that the
alternation of contextual cues selectively disrupted episodic retrieval on those trials.
However, others have found n-1 congruency sequence effects with non-repeating
MEMORY-GUIDED SELECTIVE ATTENTION
57
contextual cues (Egner, 2010; King et al., 2012). One possibility is that the combination
of making a single prior experience (the prime trial) distinctly similar to trial n and the
non-repeating contextual cues, could have disrupted memory retrieval of the n-1 trial.
This would also suggest that memory retrieval in this context, is a competitive process,
whereby only the most similar experiences are retrieved. What constitutes an effective
versus ineffective retrieval cue however, remains unclear. Furthermore, whether only a
single, most-similar previous experience is retrieved, or if a collection of similar
experiences is aggregated over and used to guide attention remains an open question.
Broader implications
A global aim of this research program is to determine how memory for specific
prior experiences guides performance in the present moment. Previous work has
focused on evidence that contextual cues can rapidly modify cognitive control settings.
The instance-based memory account of contextual control (Crump, 2016) is the general
hypothesis that memory not only preserves a record of the details of specific
experiences (Hintzman, 1986; Jacoby & Brooks, 1984; Logan, 1988), but also
preserves a record of the control procedures involved in processing those experiences
(Kolers & Roediger, 1984). Our aim here was to supply evidence showing that
attentional control in the present moment can be modified on a long-term basis by
memories of specific prior processing experiences. Beyond the implications of this
finding for theories of cognitive control from the congruency literature, we are optimistic
the idea behind our results will spur more work into the memorial basis of cognitive
control. In everyday life, we expect that memory for prior cognitive control operations
MEMORY-GUIDED SELECTIVE ATTENTION
58
routinely optimizes performance in familiar environments. In these situations, people
gain the benefit of applying memory-based procedures for regulating information without
the normal cost of effortful deliberation. We also expect that problems in regulating the
flow of information can result from the inappropriate use of, or failure to rely on memory.
For example, deliberate control may need to override memory based control when
situations act as strong cues for prior memory procedures that may not be appropriate
for the present moment. Or, when memory fails to encode or retrieve cognitive control
procedures, people may be forced to rely on taxing voluntary control processes to
supply the control they normally receive from memory for free. The present results show
that some aspects of the attentional control procedures used during a fleeting encounter
with a unique stimulus in a flanker task have long-term influences on responding to that
stimulus in the future. Everyday life presents many more fleeting and meaningful
experiences, and the capacity of memory to preserve and reinstate past control
procedures to regulate cognition, behavior, and performance points to a healthy avenue
for future work.
MEMORY-GUIDED SELECTIVE ATTENTION
59
References
Aben, B., Verguts, T., & Van den Bussche, E. (2017). Beyond trial-by-trial adaptation: A
quantification of the time scale of cognitive control. Journal of Experimental
Psychology: Human Perception and Performance, 43(3), 509.
Abrahamse, E., Braem, S., Notebaert, W., & Verguts, T. (2016). Grounding cognitive
control in associative learning. Psychological Bulletin, 142, 693–728.
https://doi.org/10.1037/bul0000047
Ahissar, M., & Hochstein, S. (1993). Attentional control of early perceptual learning.
Proceedings of the National Academy of Sciences, 90(12), 5718–5722.
Akçay, Ç., & Hazeltine, E. (2007). Conflict monitoring and feature overlap: Two sources
of sequential modulations. Psychonomic Bulletin & Review, 14, 742–748.
https://doi.org/10.3758/BF03196831
Akçay, Ç., & Hazeltine, E. (2008). Conflict adaptation depends on task structure.
Journal of Experimental Psychology: Human Perception and Performance, 34,
958–973. https://doi.org/10.1037/0096-1523.34.4.958
Awh, E., Belopolsky, A. V., & Theeuwes, J. (2012). Top-down versus bottom-up
attentional control: a failed theoretical dichotomy. Trends in Cognitive Sciences,
16, 437–443. https://doi.org/10.1016/j.tics.2012.06.010
Ball, K., & Sekuler, R. (1987). Direction-specific improvement in motion discrimination.
Vision Research, 27, 953–965. https://doi.org/10.1016/0042-6989(87)90011-3
MEMORY-GUIDED SELECTIVE ATTENTION
60
Ball, K., Sekuler, R., & Machamer, J. (1983). Detection and identification of moving
targets. Vision Research, 23, 229–238. https://doi.org/10.1016/0042-
6989(83)90111-6
Blais, C., Harris, M. B., Sinanian, M. H., & Bunge, S. A. (2015). Trial-by-trial
adjustments in control triggered by incidentally encoded semantic cues. The
Quarterly Journal of Experimental Psychology, 68, 1920–1930.
https://doi.org/10.1080/17470218.2014.1000346
Blais, C., Robidoux, S., Risko, E. F., & Besner, D. (2007). Item-specific adaptation and
the conflict-monitoring hypothesis: A computational model. Psychological
Review, 114, 1076–1086.
Botvinick, M. M. (2007). Conflict monitoring and decision making: reconciling two
perspectives on anterior cingulate function. Cognitive, Affective, & Behavioral
Neuroscience, 7, 356–366. https://doi.org/10.3758/CABN.7.4.356
Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001).
Conflict monitoring and cognitive control. Psychological Review, 108, 624.
https://doi.org/10.1037/0033-295X.108.3.624
Brady, T. F., Konkle, T., Gill, J., Oliva, A., & Alvarez, G. A. (2013). Visual long-term
memory has the same limit on fidelity as visual working memory. Psychological
Science, 24, 981–990. https://doi.org/10.1177/0956797612465439
Braem, S., Abrahamse, E. L., Duthoo, W., & Notebaert, W. (2014). What determines the
specificity of conflict adaptation? A review, critical analysis, and proposed
synthesis. Frontiers in Psychology, 5, 1134.
https://doi.org/10.3389/fpsyg.2014.01134
MEMORY-GUIDED SELECTIVE ATTENTION
61
Braver, T. S. (2012). The variable nature of cognitive control: a dual mechanisms
framework. Trends in Cognitive Sciences, 16, 106–113.
https://doi.org/10.1016/j.tics.2011.12.010
Brosowsky, N. P., & Crump, M. J. C. (2016). Context-specific attentional sampling:
Intentional control as a pre-requisite for contextual control. Consciousness and
Cognition, 44, 146–160. https://doi.org/10.1016/j.concog.2016.07.001
Bugg, J. M., & Crump, M. J. C. (2012). In support of a distinction between voluntary and
stimulus-driven control: A review of the literature on proportion congruent effects.
Frontiers in Psychology, 3, 367. https://doi.org/10.3389/fpsyg.2012.00367
Bugg, J. M., & Hutchison, K. A. (2013). Converging evidence for control of color–word
Stroop interference at the item level. Journal of Experimental Psychology:
Human Perception and Performance, 39(2), 433.
Bugg, J. M., Jacoby, L. L., & Chanani, S. (2011). Why it is too early to lose control in
accounts of item-specific proportion congruency effects. Journal of Experimental
Psychology: Human Perception and Performance, 37(3), 844.
Bugg, J. M., Jacoby, L. L., & Toth, J. P. (2008). Multiple levels of control in the Stroop
task. Memory & Cognition, 36, 1484–1494. https://doi.org/10.3758/MC.36.8.1484
Cañadas, E., Rodríguez-Bailón, R., Milliken, B., & Lupiáñez, J. (2013). Social categories
as a context for the allocation of attentional control. Journal of Experimental
Psychology: General, 142, 934–943. https://doi.org/10.1037/a0029794
Chun, M. M., & Jiang, Y. (1998). Contextual cueing: Implicit learning and memory of
visual context guides spatial attention. Cognitive Psychology, 36, 28–71.
https://doi.org/10.1006/cogp.1998.0681
MEMORY-GUIDED SELECTIVE ATTENTION
62
Chun, M. M., & Jiang, Y. (2003). Implicit, long-term spatial contextual memory. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 29, 224–234.
https://doi.org/10.1037/0278-7393.29.2.224
Corballis, P. M., & Gratton, G. (2003). Independent control of processing strategies for
different locations in the visual field. Biological Psychology, 64, 191–209.
https://doi.org/10.1016/S0301-0511(03)00109-1
Crump, M. J. C. (2016). Learning to selectively attend from context-specific attentional
histories: A demonstration and some constraints. Canadian Journal of
Experimental Psychology/Revue Canadienne de Psychologie Expérimentale, 70,
59–77. https://doi.org/10.1037/cep0000066
Crump, M. J. C., Brosowsky, N. P., & Milliken, B. (2017). Reproducing the location-
based context-specific proportion congruent effect for frequency unbiased items:
A reply to Hutcheon and Spieler (2016). The Quarterly Journal of Experimental
Psychology, 70, 1792–1807. https://doi.org/10.1080/17470218.2016.1206130
Crump, M. J. C., Gong, Z., & Milliken, B. (2006). The context-specific proportion
congruent Stroop effect: Location as a contextual cue. Psychonomic Bulletin &
Review, 13, 316–321. https://doi.org/10.3758/BF03193850
Crump, M. J. C., & Milliken, B. (2009). The flexibility of context-specific control:
Evidence for context-driven generalization of item-specific control settings. The
Quarterly Journal of Experimental Psychology, 62, 1523–1532.
https://doi.org/10.1080/17470210902752096
Crump, M. J. C., Vaquero, J. M. M., & Milliken, B. (2008). Context-specific learning and
control: The roles of awareness, task relevance, and relative salience.
MEMORY-GUIDED SELECTIVE ATTENTION
63
Consciousness and Cognition, 17, 22–36.
https://doi.org/10.1016/j.concog.2007.01.004
D’Angelo, M. C., Thomson, D. R., Tipper, S. P., & Milliken, B. (2016). Negative priming
1985 to 2015: a measure of inhibition, the emergence of alternative accounts,
and the multiple process challenge. The Quarterly Journal of Experimental
Psychology, 69(10), 1890–1909.
https://doi.org/10.1080/17470218.2016.1173077
De Pisapia, N., & Braver, T. S. (2006). A model of dual control mechanisms through
anterior cingulate and prefrontal cortex interactions. Neurocomputing, 69, 1322–
1326. https://doi.org/10.1016/j.neucom.2005.12.100
DeSchepper, B., & Treisman, A. (1996). Visual memory for novel shapes: Implicit
coding without attention. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 22, 27–47. https://doi.org/10.1037/0278-7393.22.1.27
Desender, K., Van Lierde, E., & Van den Bussche, E. (2013). Comparing conscious and
unconscious conflict adaptation. PLoS One, 8, e55976.
https://doi.org/10.1371/journal.pone.0055976
Dosher, B. A., & Lu, Z.-L. (1998). Perceptual learning reflects external noise filtering and
internal noise reduction through channel reweighting. Proceedings of the
National Academy of Sciences, 95, 13988–13993.
https://doi.org/10.1073/pnas.95.23.13988
Dosher, B. A., & Lu, Z.-L. (1999). Mechanisms of perceptual learning. Vision Research,
39, 3197–3221. https://doi.org/10.1016/S0042-6989(99)00059-0
MEMORY-GUIDED SELECTIVE ATTENTION
64
Dreisbach, G., & Fischer, R. (2012). Conflicts as aversive signals. Brain and Cognition,
78, 94–98. https://doi.org/10.1016/j.bandc.2011.12.003
Duthoo, W., Abrahamse, E. L., Braem, S., & Notebaert, W. (2014). Going, going, gone?
Proactive control prevents the congruency sequence effect from rapid decay.
Psychological Research, 78, 483–493. https://doi.org/10.1007/s00426-013-0498-
4
Egner, T. (2007). Congruency sequence effects and cognitive control. Cognitive,
Affective, & Behavioral Neuroscience, 7, 380–390.
https://doi.org/10.3758/CABN.7.4.380
Egner, T. (2008). Multiple conflict-driven control mechanisms in the human brain.
Trends in Cognitive Sciences, 12(10), 374–380.
Egner, T. (2010). Going, going, gone: characterizing the time-course of congruency
sequence effects. Frontiers in Psychology, 1, 154.
https://doi.org/10.3389/fpsyg.2010.00154
Egner, T. (2014). Creatures of habit (and control): a multi-level learning perspective on
the modulation of congruency effects. Frontiers in Psychology, 5, 1247.
https://doi.org/10.3389/fpsyg.2014.01247
Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of
a target letter in a nonsearch task. Perception & Psychophysics, 16, 143–149.
https://doi.org/10.3758/BF03203267
Forster, S. E., Carter, C. S., Cohen, J. D., & Cho, R. Y. (2011). Parametric manipulation
of the conflict signal and control-state adaptation. Journal of Cognitive
Neuroscience, 23, 923–935. https://doi.org/10.1162/jocn.2010.21458
MEMORY-GUIDED SELECTIVE ATTENTION
65
Frings, C., Schneider, K. K., & Fox, E. (2015). The negative priming paradigm: An
update and implications for selective attention. Psychonomic Bulletin & Review,
22(6), 1577–1597.
Furmanski, C. S., & Engel, S. A. (2000). Perceptual learning in object recognition:
Object specificity and size invariance. Vision Research, 40, 473–484.
https://doi.org/10.1016/S0042-6989(99)00134-0
Goldstone, R. L. (1998). Perceptual learning. Annual Review of Psychology, 49, 585–
612. https://doi.org/10.1146/annurev.psych.49.1.585
Gratton, G., Coles, M. G., & Donchin, E. (1992). Optimizing the use of information:
strategic control of activation of responses. Journal of Experimental Psychology:
General, 121, 480–506. https://doi.org/10.1037/0096-3445.121.4.480
Grison, S., Tipper, S. P., & Hewitt, O. (2005). Long-term negative priming: Support for
retrieval of prior attentional processes. The Quarterly Journal of Experimental
Psychology Section A, 58(7), 1199–1224.
Gutnisky, D. A., Hansen, B. J., Iliescu, B. F., & Dragoi, V. (2009). Attention alters visual
plasticity during exposure-based learning. Current Biology, 19, 555–560.
https://doi.org/10.1016/j.cub.2009.01.063
Hazeltine, E., Lightman, E., Schwarb, H., & Schumacher, E. H. (2011). The boundaries
of sequential modulations: Evidence for set-level control. Journal of Experimental
Psychology: Human Perception and Performance, 37, 1898–1914.
https://doi.org/10.1037/a0024662
Hintzman, D. L. (1984). MINERVA 2: A simulation model of human memory. Behavior
Research Methods, Instruments, & Computers, 16(2), 96–101.
MEMORY-GUIDED SELECTIVE ATTENTION
66
Hintzman, D. L. (1986). Schema abstraction in a multiple-trace memory model.
Psychological Review, 93, 411–428.
Hommel, B. (1998). Event files: Evidence for automatic integration of stimulus-response
episodes. Visual Cognition, 5, 183–216. https://doi.org/10.1080/713756773
Hommel, B., Müsseler, J., Aschersleben, G., & Prinz, W. (2001). The Theory of Event
Coding (TEC): A framework for perception and action planning. Behavioral and
Brain Sciences, 24, 849–937. https://doi.org/10.1017/S0140525X01000103
Hommel, B., Proctor, R. W., & Vu, K.-P. L. (2004). A feature-integration account of
sequential effects in the Simon task. Psychological Research, 68, 1–17.
https://doi.org/10.1007/s00426-003-0132-y
Hübner, R., & Mishra, S. (2016). Location-specific attentional control is also possible in
the Simon task. Psychonomic Bulletin & Review, 23, 1867–1872.
https://doi.org/10.3758/s13423-016-1057-y
Hutcheon, T. G., & Spieler, D. H. (2017). Limits on the generalizability of context-driven
control. Quarterly Journal of Experimental Psychology, 70, 1292-1304.
https://doi.org/10.1080/17470218.2016.1182193
Hutchinson, J. B., & Turk-Browne, N. B. (2012). Memory-guided attention: Control from
multiple memory systems. Trends in Cognitive Sciences, 16, 576–579.
https://doi.org/10.1016/j.tics.2012.10.003
Jacoby, L. L., & Brooks, L. R. (1984). Nonanalytic cognition: Memory, perception, and
concept learning. Psychology of Learning and Motivation, 18, 1–47.
https://doi.org/10.1016/S0079-7421(08)60358-8
MEMORY-GUIDED SELECTIVE ATTENTION
67
Jacoby, L. L., Lindsay, D. S., & Hessels, S. (2003). Item-specific control of automatic
processes: Stroop process dissociations. Psychonomic Bulletin & Review, 10(3),
638–644.
Jiang, J., Heller, K., & Egner, T. (2014). Bayesian modeling of flexible cognitive control.
Neuroscience & Biobehavioral Reviews, 46, 30–43.
https://doi.org/10.1016/j.neubiorev.2014.06.001
Jiménez, L., & Méndez, A. (2013). It is not what you expect: dissociating conflict
adaptation from expectancies in a Stroop task. Journal of Experimental
Psychology: Human Perception and Performance, 39, 271–284.
https://doi.org/10.1037/a0027734
Jiménez, L., & Méndez, A. (2014). Even with time, conflict adaptation is not made of
expectancies. Frontiers in Psychology, 5, 1042.
https://doi.org/10.3389/fpsyg.2014.01042
Kahneman, D., Treisman, A., & Gibbs, B. J. (1992). The reviewing of object files:
Object-specific integration of information. Cognitive Psychology, 24, 175–219.
https://doi.org/10.1016/0010-0285(92)90007-O
Kerns, J. G., Cohen, J. D., MacDonald, A. W., Cho, R. Y., Stenger, V. A., & Carter, C.
S. (2004). Anterior cingulate conflict monitoring and adjustments in control.
Science, 303, 1023–1026. https://doi.org/10.1126/science.1089910
King, J. A., Korb, F. M., & Egner, T. (2012). Priming of control: Implicit contextual cuing
of top-down attentional set. The Journal of Neuroscience, 32, 8192–8200.
https://doi.org/10.1523/JNEUROSCI.0934-12.2012
MEMORY-GUIDED SELECTIVE ATTENTION
68
Krebs, R. M., Boehler, C. N., De Belder, M., & Egner, T. (2013). Neural conflict–control
mechanisms improve memory for target stimuli. Cerebral Cortex, 25(3), 833–843.
https://doi.org/10.1093/cercor/bht283
Kolers, P. A., & Roediger, H. L. (1984). Procedures of mind. Journal of Verbal Learning
and Verbal Behavior, 23, 425–449. https://doi.org/10.1016/S0022-
5371(84)90282-2
Kunde, W., & Wühr, P. (2006). Sequential modulations of correspondence effects
across spatial dimensions and tasks. Memory & Cognition, 34, 356–367.
https://doi.org/10.3758/BF03193413
Logan, G. D. (1988). Toward an instance theory of automatization. Psychological
Review, 95(4), 492–527.
Logan, G. D. (1990). Repetition priming and automaticity: Common underlying
mechanisms? Cognitive Psychology, 22(1), 1–35. https://doi.org/10.1016/0010-
0285(90)90002-L
Logan, G. D., & Zbrodoff, N. J. (1979). When it helps to be misled: Facilitative effects of
increasing the frequency of conflicting stimuli in a Stroop-like task. Memory &
Cognition, 7, 166–174. https://doi.org/10.3758/BF03197535
Lowe, D. (1998). Long-term positive and negative identity priming: Evidence for episodic
retrieval. Memory & Cognition, 26(3), 435–443.
Lowe, D. G. (1979). Strategies, context, and the mechanism of response inhibition.
Memory & Cognition, 7(5), 382–389.
MEMORY-GUIDED SELECTIVE ATTENTION
69
Lu, Z.-L., Hua, T., Huang, C.-B., Zhou, Y., & Dosher, B. A. (2011). Visual perceptual
learning. Neurobiology of Learning and Memory, 95, 145–151.
https://doi.org/10.1016/j.nlm.2010.09.010
Mayr, U., Awh, E., & Laurey, P. (2003). Conflict adaptation effects in the absence of
executive control. Nature Neuroscience, 6, 450–452.
https://doi.org/10.1038/nn1051
Milliken, B., Joordens, S., Merikle, P. M., & Seiffert, A. E. (1998). Selective attention: A
reevaluation of the implications of negative priming. Psychological Review,
105(2), 203.
Milliken, B., Lupianez, J., Debner, J., & Abello, B. (1999). Automatic and controlled
processing in Stroop negative priming: The role of attentional set. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 25(6), 1384.
Moore, C. M. (1994). Negative priming depends on probe-trial conflict: Where has all
the inhibition gone? Attention, Perception, & Psychophysics, 56(2), 133–147.
Neill, W. T. (1997). Episodic retrieval in negative priming and repetition priming. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 23(6), 1291–
1305.
Neill, W. T., & Valdes, L. A. (1992). Persistence of negative priming: Steady state or
decay? Journal of Experimental Psychology: Learning, Memory, and Cognition,
18(3), 565.
Neill, W. T., Valdes, L. A., Terry, K. M., & Gorfein, D. S. (1992). Persistence of negative
priming: II. Evidence for episodic trace retrieval. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 18(5), 993–1000.
MEMORY-GUIDED SELECTIVE ATTENTION
70
Pashler, H., & Baylis, G. C. (1991). Procedural learning: I. Locus of practice effects in
speeded choice tasks. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 17, 20. https://doi.org/10.1037/0278-7393.17.1.20
Posner, M. I., & Snyder, C. R. R. (1975). Facilitation and inhibition in the processing of
signals. Attention and Performance V, 669–682.
Rey-Mermet, A., & Meier, B. (2017). How long-lasting is the post-conflict slowing after
incongruent trials? Evidence from the Stroop, Simon, and flanker tasks.
Attention, Perception, & Psychophysics, 1–23.
Roelfsema, P. R., van Ooyen, A., & Watanabe, T. (2010). Perceptual learning rules
based on reinforcers and attention. Trends in Cognitive Sciences, 14, 64–71.
https://doi.org/10.1016/j.tics.2009.11.005
Rosner, T. M., D1Angelo, M. C., MacLellan, E., & Milliken, B. (2015). Selective attention
and recognition: effects of congruency on episodic learning. Psychological
Research, 79, 411-424. https://doi.org/10.1007/s00426-014-0572-6
Rosner, T. M., & Milliken, B. (2015). Congruency effects on recognition memory: A
context effect. Canadian Journal of Experimental Psychology/Revue Canadienne
de Psychologie Exp A context, 69, 206–212. https://doi.org/10.1037/cep0000049
Sasaki, Y., Nanez, J. E., & Watanabe, T. (2010). Advances in visual perceptual learning
and plasticity. Nature Reviews Neuroscience, 11, 53–60.
https://doi.org/10.1038/nrn2737
Schmidt, J. R. (2013). The Parallel Episodic Processing (PEP) model: Dissociating
contingency and conflict adaptation in the item-specific proportion congruent
paradigm. Acta Psychologica, 142(1), 119–126.
MEMORY-GUIDED SELECTIVE ATTENTION
71
Schmidt, J. R., & Besner, D. (2008). The Stroop effect: Why proportion congruent has
nothing to do with congruency and everything to do with contingency. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 34(3), 514–523.
Schmidt, J. R., & De Houwer, J. (2011). Now you see it, now you don’t: Controlling for
contingencies and stimulus repetitions eliminates the Gratton effect. Acta
Psychologica, 138, 176–186. https://doi.org/10.1016/j.actpsy.2011.06.002
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information
processing: I. Detection, search, and attention. Psychological Review, 84(1), 1–
66.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information
processing: II. Perceptual learning, automatic attending and a general theory.
Psychological Review, 84, 127–190. https://doi.org/10.1037/0033-295X.84.2.127
Shiu, L.-P., & Pashler, H. (1992). Improvement in line orientation discrimination is
retinally local but dependent on cognitive set. Attention, Perception, &
Psychophysics, 52, 582–588. https://doi.org/10.3758/BF03206720
Spapé, M. M., & Hommel, B. (2008). He said, she said: Episodic retrieval induces
conflict adaptation in an auditory Stroop task. Psychonomic Bulletin & Review,
15, 1117–1121. https://doi.org/10.3758/PBR.15.6.1117
Spapé, M. M., & Hommel, B. (2014). Sequential modulations of the Simon effect
depend on episodic retrieval. Frontiers in Psychology, 5, 855.
https://doi.org/10.3389/fpsyg.2014.00855
Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of
Experimental Psychology, 18, 643. https://doi.org/10.1037/h0054651
MEMORY-GUIDED SELECTIVE ATTENTION
72
Szpiro, S. F. A., & Carrasco, M. (2015). Exogenous attention enables perceptual
learning. Psychological Science, 26, 1854–1862.
https://doi.org/10.1177/0956797615598976
Thomson, D. R., & Milliken, B. (2012). Perceptual distinctiveness produces long-lasting
priming of pop-out. Psychonomic Bulletin & Review, 19(2), 170–176.
Thomson, D. R., & Milliken, B. (2013). Contextual distinctiveness produces long-lasting
priming of pop-out. Journal of Experimental Psychology: Human Perception and
Performance, 39(1), 202–215.
Tipper, S. P. (1985). The negative priming effect: Inhibitory priming by ignored objects.
The Quarterly Journal of Experimental Psychology, 37(4), 571–590.
Tipper, S. P., & Cranston, M. (1985). Selective attention and priming: Inhibitory and
facilitatory effects of ignored primes. The Quarterly Journal of Experimental
Psychology, 37(4), 591–611.
Tipper, S. P., & Driver, J. (1988). Negative priming between pictures and words in a
selective attention task: Evidence for semantic processing of ignored stimuli.
Memory & Cognition, 16(1), 64–70.
Tipper, S. P., Grison, S., & Kessler, K. (2003). Long-term inhibition of return of attention.
Psychological Science, 14(1), 19–25.
Ullsperger, M., Bylsma, L. M., & Botvinick, M. M. (2005). The conflict adaptation effect:
It’s not just priming. Cognitive, Affective, & Behavioral Neuroscience, 5, 467–472.
https://doi.org/10.3758/CABN.5.4.467
MEMORY-GUIDED SELECTIVE ATTENTION
73
Van Selst, M., & Jolicoeur, P. (1994). A solution to the effect of sample size on outlier
elimination. The Quarterly Journal of Experimental Psychology Section A, 47,
631–650. https://doi.org/10.1080/14640749408401131
Verbruggen, F., & Logan, G. D. (2008). Long-term aftereffects of response inhibition:
memory retrieval, task goals, and cognitive control. Journal of Experimental
Psychology: Human Perception and Performance, 34, 1229–1235.
https://doi.org/10.1037/0096-1523.34.5.1229
Verguts, T., & Notebaert, W. (2008). Hebbian learning of cognitive control: dealing with
specific and nonspecific adaptation. Psychological Review, 115(2), 518.
Verguts, T., & Notebaert, W. (2009). Adaptation by binding: a learning account of
cognitive control. Trends in Cognitive Sciences, 13, 252–257.
https://doi.org/10.1016/j.tics.2009.02.007
Vietze, I., & Wendt, M. (2009). Context specificity of conflict frequency-dependent
control. The Quarterly Journal of Experimental Psychology, 62, 1391–1400.
https://doi.org/10.1080/17470210802426908
Vogels, R., & Orban, G. A. (1985). The effect of practice on the oblique effect in line
orientation judgments. Vision Research, 25, 1679–1687.
https://doi.org/10.1016/0042-6989(85)90140-3
Waszak, F., Hommel, B., & Allport, A. (2003). Task-switching and long-term priming:
Role of episodic stimulus–task bindings in task-shift costs. Cognitive Psychology,
46, 361–413. https://doi.org/10.1016/S0010-0285(02)00520-0
MEMORY-GUIDED SELECTIVE ATTENTION
74
Watanabe, T., & Sasaki, Y. (2015). Perceptual learning: Toward a comprehensive
theory. Annual Review of Psychology, 66, 197–221.
https://doi.org/10.1146/annurev-psych-010814-015214
Weidler, B. J., & Bugg, J. M. (2016). Transfer of location-specific control to untrained
locations. The Quarterly Journal of Experimental Psychology, 69, 2202–2217.
https://doi.org/10.1080/17470218.2015.1111396
Weissman, D. H., & Carp, J. (2013). Congruency sequence effects are driven by
previous-trial congruency, not previous-trial response conflict. Frontiers in
Psychology, 4. https://doi.org/10.3389/fpsyg.2013.00587
Weissman, D. H., Jiang, J., & Egner, T. (2014). Determinants of congruency sequence
effects without learning and memory confounds. Journal of Experimental
Psychology: Human Perception and Performance, 40, 2022–2037.
https://doi.org/10.1037/a0037454
Wendt, M., Kiesel, A., Geringswald, F., Purmann, S., & Fischer, R. (2015). Attentional
adjustment to conflict strength. Experimental Psychology, 61, 55–67.
https://doi.org/10.1027/1618-3169/a000227
MEMORY-GUIDED SELECTIVE ATTENTION
75
Table 1.
Long-term congruency sequence effects for Experiments 1-3
Probe
Congruency
Effect
Long-Term
CSE
Con
Inc
(I - C)
(C(I - C) - I(I - C))
Prime
RT
ER
RT
ER
RT
RT
Exp. 1A
Con
623 (23)
5.00 (.08)
664 (24)
5.12 (.71)
41 (11)
27 (14)
Inc
632 (22)
3.69 (.74)
646 (22)
3.69 (.68)
14 (10)
Exp. 1B
Con
766 (22)
3.78 (.71)
831 (25)
2.21 (.53)
65 (15)
23 (16)
Inc
779 (27)
3.65 (.69)
829 (21)
3.52 (.90)
42 (13)
Exp. 1C
Con
774 (26)
2.88 (.44)
831 (28)
2.21 (.42)
58 (9)
19 (9)
Inc
773 (26)
3.12 (.47)
812 (27)
3.38 (.51)
39 (6)
Exp. 2A
Con
566 (20)
2.99 (.55)
605 (21)
4.17 (.68)
39 (6)
21 (9)
Inc
575 (21)
2.64 (.43)
593 (19)
4.31 (.55)
18 (7)
Exp. 2B
Con
589 (22)
1.97 (.38)
647 (21)
4.34 (.66)
58 (8)
17 (11)
Inc
590 (23)
2.43 (.38)
630 (19)
3.95 (.68)
41 (11)
Exp. 3A
Con
842 (21)
2.37 (.50)
880 (22)
3.03 (.63)
38 (8)
13 (14)
Inc
846 (24)
2.84 (.55)
871 (21)
2.94 (.56)
25 (11)
Exp. 3B
Con
837 (22)
2.73 (.45)
867 (20)
2.50 (.48)
30 (7)
0 (10)
Inc
840 (22)
1.92 (.39)
870 (23)
2.15 (.47)
30 (7)
*Note: RT = Reaction times (ms); ER = error rates (%); Con/C = congruent; Inc/I = incongruent;
standard errors are presented in parentheses.
MEMORY-GUIDED SELECTIVE ATTENTION
76
Table 2.
N-1 congruency sequence effects for Experiments 1-3
Trial N
Congruency
Effect
N-1 CSE
Con
Inc
(I - C)
(C(I - C) - I(I - C))
Trial N-1
RT
ER
RT
ER
RT
RT
Exp. 1A
Con
626 (22)
2.97 (.55)
658 (22)
3.48 (.48)
32 (7)
-3 (13)
Inc
635 (21)
3.55 (.46)
671 (25)
4.48 (.72)
36 (9)
Exp. 1B
Con
753 (21)
2.40 (.51)
832 (25)
3.48 (.53)
78 (10)
36 (12)
Inc
791 (24)
3.58 (.54)
671 (25)
3.23 (.71)
42 (14)
Exp. 1C
Con
771 (27)
2.46 (.37)
834 (27)
2.66 (.39)
62 (6)
27 (7)
Inc
794 (25)
2.91 (.40)
829 (26)
2.65 (.38)
35 (7)
Exp. 2A
Con
557 (17)
2.36 (.38)
593 (16)
4.36 (.57)
36 (6)
8 (8)
Inc
576 (18)
3.59 (.47)
605 (19)
3.80 (.60)
28 (6)
Exp. 2B
Con
568 (20)
1.91 (.37)
638 (24)
4.67 (.61)
70 (5)
33 (6)
Inc
606 (24)
2.83 (.41)
643 (23)
3.78 (.58)
37 (7)
Exp. 3A
Con
837 (24)
2.18 (.45)
880 (20)
2.65 (.41)
43 (8)
22 (12)
Inc
855 (21)
2.83 (.40)
876 (22)
2.85 (.56)
20 (8)
Exp. 3B
Con
860 (23)
2.54 (.38)
889 (24)
2.38 (.42)
30 (7)
13 (10)
Inc
873 (24)
2.28 (.36)
889 (23)
2.43 (.35)
16 (8)
*Note: RT = Reaction times (ms); ER = error rates (%); Con/C = congruent; Inc/I = incongruent;
standard errors are presented in parentheses.
