Content uploaded by Gouret Florian
Author content
All content in this area was uploaded by Gouret Florian on Jul 20, 2020
Content may be subject to copyright.
Action-Effect Learning and Anticipatory Saccades 1
Learning to Expect the Future: Linking learning, monitoring, and action control by
assessing how fast anticipatory saccades towards future action consequences
emerge
Florian Gouret & Christina U. Pfeuffer
Cognition, Action, and Sustainability Unit, Department of Psychology, Albert-Ludwigs-
University of Freiburg, Freiburg, Germany
Draft version 07/20/2020. This paper has not yet been peer reviewed. Please do not copy or
cite without author's permission.
Author Note
This research was supported by a grant of the Baden-Württemberg Stiftung awarded to
Christina Pfeuffer who is indebted to the Baden-Württemberg Stiftung for the financial support
of this research project by the Eliteprogramme for Postdocs.
The data as well as experiment files and syntaxes are publicly available on OSF:
https://osf.io/457gy/ - DOI 10.17605/OSF.IO/457GY
Word count: 6,123
Correspondence:
Florian Gouret
Albert-Ludwigs-Universitaet Freiburg
Cognition, Action, and Sustainability Unit, Department of Psychology
Engelbergerstrasse 41
79085 Freiburg, Germany
Tel. +49-761-203-5689
Email: florian.gouret@psychologie.uni-freiburg.de
Action-Effect Learning and Anticipatory Saccades 2
Abstract
While performing an action that contingently yields the same effect, we form action-
effect associations that allow us to anticipate the effects of our actions. Importantly, our eyes
already move towards the expected future location of our actions´ effects in anticipation of
them, that is, we perform anticipatory saccades. These anticipatory saccades are linked to a
proactive effect monitoring process that prepares the comparison of expected and actual
effect. However, how fast such anticipatory saccades emerge (i.e., how fast learning leads to
monitoring) is unknown. To address this question, here, correct left/right responses were
followed by a visual effect either on the same side (action-effect compatible) or on the
opposite side (action-effect incompatible). In Experiment 1, action-effect compatibility
switched after sequences of 4, 8, or 12 trials (randomly allocated; partly predictable
environment). In Experiment 2, random trials (2 to 7) separated sequences of 3, 5, or 7
experimental trials. Again, action-effect compatibility switched after a sequence and
sequences were randomly allocated (unpredictable environment). In both experiments,
participants started to perform anticipatory saccades towards future effects after
experiencing a new action-effect mapping just once to twice. Thus, one to two action-effect
learning instances were sufficient to develop action-effect associations that trigger attentional
shifts towards the expected future consequences of our actions (i.e., monitoring processes),
whereas influences on action selection are only observed after a much larger number of
learning instances. This pattern might suggest that monitoring processes modulate the
expression of action-effect associations in action planning based on observed action-effect
contingencies.
Keywords: ideomotor theory; action effect; anticipatory saccades; action control;
learning; monitoring
Action-Effect Learning and Anticipatory Saccades 3
Public Significance statement
When an action predictably yields a certain consequence, we anticipate the future
effect and already move our eyes to where it will appear. These eye movements prepare us
for comparing expected and actual consequences of our actions and allow us to adapt if
things did not go as planned. Here, we demonstrate that we just need to experience our
action and its consequence once to twice to start performing such anticipatory eye
movements that guide our subsequent actions.
Action-Effect Learning and Anticipatory Saccades 4
Introduction
Working with our computer, we have learned that, pressing a key makes the
corresponding character appear on the screen and moving the mouse moves the cursor on
the monitor accordingly. Therefore, when performing these actions, as we anticipate what is
going to happen, we already look at the corresponding locations on screen and are surprised
when the expected effect does not occur. As simple as this behaviour appears, it raises the
question how we actually develop such anticipatory eye movements.
How we control our actions based on their consequences was already addressed by,
for instance, Harleß (1861) and James (1890/1981) and their research on ideomotor action
control has experienced resurgent interest in more recent years (e.g., Bunlon et al., 2015;
Elsner & Hommel, 2001; Greenwald, 1970; Prinz, 1990; Hommel, 1997, 2009; Hommel et al.,
2001; Kunde, 2001; Sun et al., 2020; see e.g., Shin, Proctor, & Capaldi, 2010, for a review).,
When we perform a goal-directed action, an effect follows. After an action has yielded the
same effect multiple times, a bi-directional association between the action and the effect is
formed (e.g., Dutzi & Hommel, 2009, Elsner & Hommel, 2001; Hommel et al., 2001; Kunde,
2001). Thanks to this bi-directional action-effect association, by anticipating desired future
effects, we can select the appropriate actions to produce them (e.g., Hommel, 2009; Hommel
et al., 2001; Kunde, 2001).
Bi-directional action-effect associations can be assessed with the acquisition-test
paradigm of Elsner and Hommel (2001). During the acquisition phase (200 trials),
participants left/right key presses were contingently followed by a high/low tone. In the
following test phase, prior effect tones were used as imperative stimuli. One group of
participants was instructed to respond to the tones with the responses the respective tones
were paired with during the acquisition phase (compatible), whereas the other group was
instructed to respond with the opposite responses (incompatible). Elsner and Hommel (2001)
found that the compatible group responded faster than the incompatible group. This suggests
that participants had formed bi-directional action-effect associations during the acquisition
phase and retrieved them during the test phase, supporting performance when action-effect
Action-Effect Learning and Anticipatory Saccades 5
mappings remained the same in the compatible condition rather than switched in the
incompatible condition.
Evidence for the idea that we actually anticipate action effects to select appropriate
actions comes from a study by Kunde (2001). Kunde (2001) suggested that if effects were
anticipated prior to acting, action selection should also be affected by stimuli that predictably
follow, and not precede, response execution (i.e., effects). To assess this idea, he asked
participants to press four different keys to trigger four different horizontally-aligned visual
action-effects. In one half of the experiment, responses and their effects were spatially
response-effect (R-E) compatible (e.g., left responses > visual effects on the left), in the
other half, response and their effects were spatially R-E incompatible (e.g., left response >
visual effect on the right). When response and effect were spatially R-E compatible, the
reaction time (RT) was on average shorter than when the spatial R-E mapping was
incompatible. This finding suggests that action effects were anticipated before the action and
had an influence on action selection (see e.g., Koch & Kunde, 2002; Kunde, 2003; Pfister, et
al., 2010, 2014, for later assessments).
However, this raises the question, how many action-effect pairings are needed to
form an action-effect association strong enough to affect action selection (i.e., show an R-E
compatibility effect). Wolfensteller and Ruge (2011), explored this question using the
acquisition-test paradigm of Elsner and Hommel (2001). During acquisition phases of
differing length (8 vs. 16 vs. 24 vs. 32 trials – half of the trials left/right responses,
respectively) participants responded to visual stimuli with a left or right key press. Once
participants had responded, a distinct, coloured background was added to the stimulus (e.g.,
blue/red for correct right/left responses, grey for incorrect responses). During the test
phases, stimuli were presented with one of the effect background colours and responses
were only followed by verbal feedback (correct, false, miss). In half of the test trials per test
phase, stimuli were presented with compatible/incompatible effects. Extending previous
studies where action and effect were paired at least 100 times (e.g., Elsner & Hommel, 2001;
Action-Effect Learning and Anticipatory Saccades 6
Ziessler, et al., 2004), Wolfensteller and Ruge (2011) found that R-E learning was rather fast,
as R-E compatibility effects were observed after eight to 12 repetitions of a R-E pairing.
Interestingly, action effect anticipation has also been studied in eye movements.
However, in the majority of studies on action effect anticipation in eye movements, saccadic
responses were instructed and used as a substitute for manual responses to assess
ideomotor influences on action selection (e.g., Herwig & Horstmann, 2011; Huestegge &
Kreutzfeldt, 2012; Riechelmann et al., 2017). In opposition to this trend, Pfeuffer et al. (2016)
suggested that spontaneously-occurring, uninstructed eye movements could carry additional
information about action effect anticipation. In their forced-choice and free-choice
experiments, correct answers were followed by a visual effect (coloured circle) predictably
located either at a spatially R-E compatible location (right key press > effect on the right) or
at a spatially R-E incompatible location (right key > circle on the left) after a short delay.
Importantly, no instructions regarding eye movements were provided. In addition to the
typical spatial R-E compatibility effect in manual responses (e.g., Kunde, 2001), Pfeuffer et
al. (2016; see also Pfeuffer et al., 2020/preprint) demonstrated that participants´ anticipation
of their actions´ consequences also showed in their eye movements. Before effect onset,
participants performed effect-congruent saccades toward the future effect location
significantly more frequently than effect-incongruent saccades in the direction opposite of the
effect (saccade-effect-congruency, SEC, effect) under both spatially R-E compatible and R-E
incompatible conditions. These anticipatory eye movements occurred mostly after manual
responses and irrespective of whether an R-E compatibility effect was observed in manual
responses in the respective experiment. Thus, Pfeuffer et al. (2016) concluded that the
anticipatory saccades they observed reflected a process dissociable from influences of
action effect anticipation on action selection which can be assessed via R-E compatibility
effects in manual, effect-generating responses. Instead, Pfeuffer et al. (2016) argued that
anticipatory saccades reflected the preparation for upcoming outcome evaluation, that is, the
comparison of expected and actual effect (i. e., a proactive effect monitoring process).
Action-Effect Learning and Anticipatory Saccades 7
Two findings of the study of Pfeuffer et al. (2016) are remarkable when considered in
the context of learning. First, anticipatory saccades towards future effects were observed
regardless of the current R-E compatibility and irrespective of whether an R-E compatibility
effect was observed in manual responses. Furthermore, SEC effects in spontaneous eye
movements were, surprisingly, substantially stronger than R-E compatibility effects in manual
responses in this and similar studies. Second, anticipatory saccades emerged already within
the first block of trials and were also present during the first block after the R-E compatibility
switched. Thus, it seems that only a few learning instances are sufficient to establish action-
effect associations strong enough to trigger proactive effect monitoring processes and
resulting anticipatory saccades.
Prior studies investigating how many learning instances were necessary to establish
action-effect associations strong enough to affect action selection suggested that eight
action-effect pairings were required at minimum (Wolfensteller & Ruge, 2011). Here, we
addressed the question how many learning instances were necessary to establish action-
effect associations strong enough to allow for proactive effect monitoring, that is, cause SEC
effects in participants´ spontaneous eye movements. Furthermore, we assessed whether
influences of action effect anticipation on manual action selection (R-E compatibility effects)
emerged in a similar range of learning instances, earlier, or later when assessed in the same
paradigm.
In order to explore this question, we conducted two experiments using an adaptation
of the eye tracking paradigm developed by Pfeuffer et al. (2016). In both experiments, we
varied the number of trials each action (left/right key presses) was paired with one effect
(coloured circles appearing on the left or on the right) before the R-E mapping unpredictably
reversed. We assessed both instructed manual responses (forced-choice) as well as
uninstructed, spontaneous eye movements occurring between target offset and effect onset.
We determined how many action-effect pairings (i.e., learning instances) were necessary to
observe influences of spatial R-E learning on action selection (manual responses; spatial R-
E compatibility effect) and proactive effect monitoring (eye movements; SEC effect). As the
Action-Effect Learning and Anticipatory Saccades 8
findings of Pfeuffer et al. (2016; see also Pfeuffer et al., 2020/preprint) suggested that SEC
effects (reflecting proactive effect monitoring) are stronger than manual R-E compatibility
effects (reflecting action selection), we hypothesized that less than eight action-effect
pairings (see Wolfensteller & Ruge, 2011) should be sufficient to observe SEC effects, but
not R-E compatibility effects.
In Experiment 1, we explored the number of action-effect pairings sufficient for
anticipatory saccades to emerge by randomly intermixing sequences of four, eight, and
twelve R-E compatible/incompatible trials. There, trial sequences or more precisely trials with
a possible switch in R-E compatibility were rather predictable and participants could have
adapted their behaviour accordingly. Thus, we set up a second experiment to confirm our
findings in a less predictable environment. In Experiment 2, trial sequences were less
predictable as sequences of three, five, or seven trials were randomly intermixed with
random sequences of two to seven trials.
Experiment 1
In Experiment 1, the R-E mapping switched after each sequence of four, eight, or
twelve trials. Trial sequences were randomly intermixed and equally frequent in each block
and R-E compatible/incompatible trials were equally frequent across the experiment.
Method
Participants. A priori sample size estimations assuming at least an effect of 𝜂 =.30
for the SEC effect on the basis of the mean effect size (𝜂= .60) reported in Pfeuffer et al.
(2016) suggested that 22 participants were necessary to find a significant effect (α < .05) with
80% power (GPower; Erdfelder et al., 1996, Faul et al., 2007). For reasons of
counterbalancing, twenty-four participants were recruited for Experiment 1 (7 males, 17
females, mean age = 22.9 years, SD = 3.4 years, 3 left-handed, 6 left eye dominant). All
participants had normal or corrected-to-normal vision and were naive to the purpose of the
experiment. All participants provided written informed consent prior to their participation and
received either course credit or 12€.
Action-Effect Learning and Anticipatory Saccades 9
Stimuli and Apparatus. The experiment was conducted in a sound attenuated, dark
laboratory room. Participants sat approximately 60 cm from a 24” LCD screen (1920 pixels x
1080 pixels, 144 Hz) with their index fingers on a left and right key (distance 13.5 cm),
respectively. The experiment was run via EPrime (PST, Sharpsburg, USA; version
2.0.10.356).
Eye movements were tracked with an EyeLink 1000 Plus Desktop Mount (SR
Research Ltd., Ontario, Canada). Corneal reflection and pupil diameter were measured via
an infrared camera and eye movements (dominant eye) were sampled at 1000 Hz with a
spatial resolution of 0.01° visual angle. At the beginning of each block, calibration and
validation were performed.
Design and procedure. During the entire experiment, the background of the screen
remained black. Each experimental trial began with the presentation of a white fixation cross
(0.5° visual angle) in the middle of the screen for a duration of 1200ms to 1500ms. The
jittered intertrial interval prevented any temporal anticipations of the target stimuli. The
fixation cross was then replaced by the target. In the first trial of each block, the target was a
white arrow (0.7° visual angle) pointing either to the left or to the right. It was displayed for
100ms and its direction indicated the first key to press. In the following trials, the target was
either “=” or “x” (0.7° visual angle). The “=” target indicated that participants had to press the
same key as in the previous trial and the “x” target indicated that participants had to press
the opposite key as compared to the previous trial (reference: correct key in the previous
trial). The target was followed by a blank screen (maximum duration: 1400ms) and
participants could respond within 1500ms from target onset. Participants were instructed to
respond fast and accurately. Correct answers were followed by another 300ms blank screen
R-E interval before the action effect appeared and remained on screen for 500ms. Eye
movements were assessed in the blank screen interval between target offset and action
effect onset (see Figure 1).
Action effects were orange or blue circles (1.3°) appearing at 12.6° to the left/right of
the screen centre and moving further in the respective direction (additional movement: 6.6°,
Action-Effect Learning and Anticipatory Saccades 10
end position: 19.1°). The colour of the action effect depended on the key pressed (e.g.,
orange for the left and blue for the right key). Action effects appeared spatially R-E
compatible to manual responses in half of the trials (R-E compatible: left manual response >
action effect on the left) and spatially R-E incompatible in the other half of the trials (R-E
incompatible: left manual response > action effect on the right). R-E compatibility remained
constant for either four, eight, or twelve trials and then switched. Per sequence, half of the
trials required right and left manual responses. The very first R-E compatibility condition and
action effect colour mappings were counterbalance across participants. Participants were
only told that their responses would be followed by coloured circles. They were not informed
about R-E compatibility conditions or changes in R-E compatibility.
In case of an incorrect response, feedback was displayed in red in the centre of the
screen (“zu früh!”/”too early!” for premature responses, “Fehler!”/”error!” for incorrect
responses, and “zu langsam!”/”too slow!” for response omissions; duration: 1000ms) and the
trial was aborted.
After each block, participants could take a break and were informed about the number
of premature and erroneous responses they had committed as well as about the number of
response omissions. They were then reminded to try and respond as fast and accurately as
possible.
The experiment consisted of 12 blocks of 72 trials (3 sequences of 4, 8, and 12 trials
each per block, randomly intermixed) and a preceding practice of 36 trials (without action
effects). R-E compatible and R-E incompatible four, eight, and twelve trial sequences
appeared equally often across the entire experiment. Half of the blocks started with an R-E
compatible/R-E incompatible sequence.
Participants were not informed about the positions the action effects would appear at
beforehand and no information regarding eye movements was given to avoid biases. Thus,
any saccades observed can be consider as spontaneous and uninstructed.
Results
Action-Effect Learning and Anticipatory Saccades 11
All statistical analyses were performed using Python version 3.7, Pengouin (Vallat,
2018) and Numpy version 1.18. The first trial of each sequence was excluded from all
analyses. Participants were not informed about changes in R-E compatibility and thus
performance (manual responses as well as saccades) during the first trial of a new sequence
were not informative, as they had not experienced the change in R-E compatibility yet.
Furthermore, as R-E compatibility predictably switched between blocks, the first sequence of
each block was excluded from all analyses.
Manual responses. Trials with premature (0.1 %) and omitted (0.5 %) responses
were excluded. For analyses of manual RTs, trials containing errors (5.6%) were excluded.
Furthermore, trials with RTs deviating by more than three SDs from their individual cell
means were considered as outliers and excluded from the RT analyses (1.0%).
A 2x3 repeated measures analysis of variance (ANOVA) was conducted to assess
the effect of the factors R-E compatibility (R-E compatible vs. R-E incompatible) and
sequence trial (2-4 vs. 5-8 vs. 9-12) on error rates within-subjects. Neither of the main
effects, R-E Compatibility, F (1, 23) = 1.97, p = .174 𝜂 = .08, and sequence trial, F < 1, nor
their interaction, F(2,23) = 1.72, p = .190, 𝜂 = .07, reached significance (see Figure 2C).
Another 2x3 repeated measure ANOVA was conducted to assess the effects of the
within-subjects factors R-E compatibility and sequence trial on RTs. There was a significant
main effect of sequence trial on manual RTs, F(2, 23) = 22.81, p <.001, 𝜂 = .50. RTs
decreased with increasing sequence trial number. Neither the main effect of R-E
compatibility, F(1, 23) = 2.93, p =.100, 𝜂 = .11, nor the interaction between R-E
compatibility and sequence trial, F(2, 23) = 1.02, p =.370, 𝜂 = .04, reached significance
(see Figure 2A).
Anticipatory saccades. Saccades were detected according to a combined velocity
(30°/s), motion (0.1°), and acceleration (8,000°/s2) threshold. As the experiment is
concerned with anticipatory saccades, only saccades occurring within the anticipatory
interval between target offset and action effect onset were selected. Furthermore, saccades
occurring in trials with incorrect or omitted responses were not. Moreover, saccades
Action-Effect Learning and Anticipatory Saccades 12
occurring during the first trial of a block or the first trial of a sequence were not assessed.
Only saccades in trials in which the first saccade started around the fixation in the screen
centre (±1.0° visual angle) were considered (9,773 of 12,039 saccades) to ensure that
participants had perceived the target. Furthermore, only saccades that crossed a minimum
horizontal distance of 1.0° were considered (7,722 of 12,039 saccades). In sum, 6,447
saccades were included of the 12,039 saccades detected in correct trials (53.5 %) during the
anticipatory phase between target offset and effect onset (see Figure 6 for the frequency
distribution of saccade end positions).
Relative saccade frequencies (SEC effect).
The SEC effect was computed by considering the frequency of saccades toward the
future effect´s location over the total frequency of saccades performed per condition.
Therefore, a value of 50% represents the chance level of saccades toward the future effect.
Values above 50% indicate that participants anticipated future effects and thus directed their
eyes toward them (i.e., participants proactively monitored their actions´ effects).
Overall, 70.8% of saccades were effect-congruent and directed towards the future
effect (73.5% for R-E compatible and 68.1% for R-E incompatible trials).
First, we conducted three one sample t-tests to compare SEC effects for different
sequences trial positions (trials 2-4 vs. 5-8 vs. 9-12) against 50% irrespective of the R-E
compatibility condition. SEC effects were significantly larger than 50% for all sequence trial
positions, sequence trial 2-4: t(23) = 5.50, p < .001, d=1.12 (M = 66.6%, SD = 21.9%),
sequence trial 5-8: t(23) = 6.50, p < .001, d = 1.33 (M = 73.3%, SD = 22.5%), sequence trial
9-12: t(23) = 5.53, p < .001, d = 1.13 (M = 72.4%, SD = 25.4%; see Figure 3A).
A 2x3 repeated measure ANOVA was conducted to further assess the influence of
the factors R-E compatibility and sequence trial on the relative frequency of effect-congruent
saccades (SEC effects) within-subjects. The main effect of R-E compatibility did not reach
significance, F(1,23) = 1.90, p = .179, 𝜂 = .07. However, there was a significant main effect
of sequence trial, F(2,23) = 5.10, p = .009, 𝜂 = .18. The later in the sequence, the more
likely participants were to perform an anticipatory saccade toward the future effect location.
Action-Effect Learning and Anticipatory Saccades 13
Finally, there was a significant interaction between sequence trial and R-E compatibility,
F(2,23) = 3.83, p = .029, 𝜂 = .14.
To further assess this interaction between R-E compatibility and sequence trial, we
conducted three paired t-tests comparing the R-E compatibility conditions for each of the
sequence trial positions. For sequence trials 2-4, a significant difference between R-E
compatible and R-E incompatible trials was obtained, t(23) = 2.52, p = .019, d = 0.72. For
sequence trials 5-8, t(23) = 0.38, p = .745, d = 0.08, and trials 9-12, t(23) = 0.06, p = .952, d
= 0.02 no significant differences emerged between R-E compatible and R-E incompatible
trials.
Saccade latency.
As multiple saccades could occur during one trial, for the following analyses of
saccade latency, only the first effect-congruent saccade per trial fulfilling the selection criteria
was assessed. Six participants who did not provide saccades for each condition were
excluded from saccade latency analyses.
A 2x3 repeated measure ANOVA was conducted to assess the effects of the factors
R-E compatibility and sequence trial on saccade latencies within-subjects. Neither main
effect nor their interaction reached significance, Fs < 1 (see Figure 4A).
Saccade latency and manual RT strongly correlate (Pfeuffer et al., 2016,
2020/preprint) and anticipatory saccades appear to be postponed until after manual
response execution. Thus, influences on saccade latencies could have been overshadowed
by diverging influences on manual RTs. We therefore additionally assessed saccade-manual
latency differences between saccade latencies and manual RTs on corresponding trials (Δ
latency = latency –RT).
A 2x3 repeated measure ANOVA was conducted to assess the effects of R-E
compatibility and sequence trial on saccade latency differences within-subjects. There were
no significant main effects of sequence trial, F(2,18) = 2.44, p = .102, 𝜂 = 0.72, or R-E
compatibility, F(1,18) = 1.70, p = .209, 𝜂 = .09, as well as no significant interaction, F < 1
(see Figure 4C).
Action-Effect Learning and Anticipatory Saccades 14
Discussion
In Experiment 1, spatial R-E compatibility unpredictably switched after four, eight, or
twelve trials and we investigated, how many action-effect pairings were required to first
observe a reliable SEC effect (> 50%), indicating proactive effect monitoring.
Our results showed that within the first four sequence trials (i.e., the first two action-
effect pairings), participants already looked significantly more often towards the location of
the future effect than expected by chance (50%). Taking into account that participants had
not yet experienced the effect a second time on the second action-effect pairing, this means
that a single action-effect pairing was sufficient to start proactively monitoring future effects
via anticipatory saccades. This finding demonstrates how fast humans (re-)learn action-effect
associations and form corresponding action effect anticipations that guide their eye
movements.
Importantly, SEC effects larger than 50% were observed for all sequence trials and
under both R-E compatible and R-E incompatible conditions. This finding shows that
participants adapted with just one prior learning instance and were able to anticipate and
proactively monitor their actions´ future effects successfully under both R-E compatible and
R-E incompatible conditions.
That action effect anticipation as evidenced by proactive effect monitoring (i.e.,
anticipatory saccades) emerged within four trials (i.e., basically after a single action-effect
pairing), is especially interesting, as previous studies have shown that at least eight action-
effect pairings were necessary to observe reliable R-E compatibility effects in participants´
effect-generating manual responses (Wolfensteller & Ruge, 2011). This finding that
influences of action-effect associations on action selection are only observed after a
substantially larger number of learning instances replicates in our results. We did not observe
a significant R-E compatibility effect within our sequences of up to twelve trials (i.e., up to 6
action-effect pairings).
Thus, it appears that rather weak action-effect associations are sufficient to derive
action effect anticipations that lead to proactive effect monitoring (i.e., anticipatory saccades
Action-Effect Learning and Anticipatory Saccades 15
towards future effects), whereas much stronger action-effect associations (i.e., a larger
number of learning instances) are required to observe influences of action effect anticipation
on action selection.
The aim of experiment 1 was to assess the number of action-effect pairings needed
to start observing a reliable SEC effect, that is, proactive effect monitoring. However, in
Experiment 1, sequences (4, 8, or 12 R-E compatible vs. R-E incompatible trials) were
selected in a rather predictable way with low variability even though their order was
randomized per block. To investigate if our findings replicate in a less predictable
environment, we conducted a second experiment in which random sequences were inserted
between sequences of one R-E compatibility.
Experiment 2
In Experiment 2, R-E compatibility switched after sequence of three, five, or seven
trials. Trial sequences were randomly intermixed and equally frequent in each block and all
sequences were interdispersed with random sequences. Again, overall R-E compatible and
R-E incompatible trials were equally frequent.
Methods
Participants. Based on the mean effect size for SEC effect across sequence trials (d
= 1.19) in Experiment 1, we assumed at least an effect of d =0.50 for the SEC effect in
Experiment 2. Based on this estimation a sample size of at least 27 participants was
suggested to find a significant SEC effect (α < .05) with 80% power (GPower; Erdfelder et al.,
1996, Faul et al., 2007).
Twenty-eight participants were recruited for this experiment (7 males, 21 females,
mean age = 23 years, SD = 5.8 years, 1 left-handed, 5 left eye dominant,). All participants
had normal or corrected-to-normal vision and were naive to the purpose of the experiment.
All participants provided written informed consent prior to their participation and received
either course credit or 12€. Two additional participants were recruited but had to be excluded
as their eyes could not consistently be detected during the experiment.
Action-Effect Learning and Anticipatory Saccades 16
Stimuli and Apparatus. Stimuli and apparatus of Experiment 2 were the same as for
Experiment 1.
Design and procedure. The design of Experiment 2 was the same as for Experiment
1 with one exception. Per block, three sequences of three, five and seven trials (experimental
sequences) were presented twice and were separated by a random trial sequence of two to
seven trials (random sequences). In the random sequences, there were no more than two
trials in a row that had the same R-E compatibility and the R-E compatibility condition of the
first and last trial of a random sequence was not the same as in the previous or following
experimental sequence.
The order of sequences was randomly allocated, so that experimental sequences
were separated by random sequences. Each time the number of trials per block exceeded
68, a new block was created after the end of the respective sequence. Each block started
with a random sequence.
In total, there were twelve blocks containing 68 to 75 trials each, totalling 848 trials.
The number of R-E compatible and R-E incompatible action effects following left/right manual
responses was balanced across the entire experiment. Per sequence trial, there were overall
an equal number of left and right responses with R-E compatible and R-E incompatible
effects.
Results
Only trials from experimental sequences were considered for the following analyses.
Additionally, the first trial per block and the first trial of each sequence were excluded from all
analyses.
Manual responses. Trials with premature (< 0.1 %) or omitted (0.3 %) responses
were excluded. For the analysis of manual RTs, trials containing errors (6.3 %) were
excluded. Trials with RTs deviating by more than three SDs from their individual cell means
were considered as outliers and excluded from the RT analysis (0.3 %).
A 2x3 repeated measures ANOVA was conducted to assess the effects of the factors
R-E compatibility and sequence trial (2-3 vs. 4-5 vs. 6-7) on manual error rates within-
Action-Effect Learning and Anticipatory Saccades 17
subjects. Neither of the main effects, R-E compatibility: F(1,27) = 1.81, p = .189, 𝜂 = 0.06,
sequence trial: F < 1, nor their interaction, F < 1, reached significance (see Figure 2D).
The same 2x3 repeated measures ANOVA was performed on manual RTs. Neither
the main effect of sequence trial, F(2,27) = 2.67, p = .114, 𝜂 = 0.09, nor the main effect of
R-E compatibility, F(1,27) = 1.87, p = .164, 𝜂 = 0.07, nor their interaction, F < 1, reached
significance (see Figure 2B).
Anticipatory saccades. Only saccades that occurred in sequence trial 2-7 were
considered for the following analyses.
The pre-processing performed on saccades was the same as in Experiment 1.
Saccades occurring in trials with incorrect or omitted responses were not selected. Only
saccades in trials in which the first saccade started around the fixation in the screen centre
(±1.0° visual angle) were considered (3,888 of 8,345 saccades) to ensure that participants
had perceived the target and only saccades that crossed a minimum horizontal distance of
1.0° were considered (5,818 of 8,345 saccades). In sum, 2,950 saccades were included of
the 8,345 saccades detected in correct trials (35.4%) during the anticipatory phase between
target offset and effect onset (see Figure 6 for the frequency distribution of saccade end
positions).
Relative saccade frequencies (SEC effect).
We conducted three one sample t-test to compare the different sequence trial
positions (sequence trial 2-3 vs. 4-5 vs. 6-7) against 50% (see Figure 3B). For sequence
trials 2-3, the t-test did not reach significance, t(27) = 1.43, p = .165, d = 0.27 (M = 54.4%,
SD = 27.2%). However, both for sequence trials 4-5, t(27) = 3.40, p = .002, d = 0.64 (M =
62.0%, SD = 27.2%)., and 6-7, t(27) = 3.65, p = .001, d = 0.69 (M = 65.5%, SD = 30.5%),
SEC effects were significantly larger than 50%.
In order to assess the impact of sequence trial and R-E compatibility on SEC effects,
we conducted a 2x3 repeated measures ANOVA. Only the main effect of sequence trial
reached significance, F(2,27) = 3.74, p = .030, 𝜂 = 0.12. Participants performed more eye
movements towards the effect location for later sequence trials as compared to earlier
Action-Effect Learning and Anticipatory Saccades 18
sequence trials. Both the main effect of R-E compatibility, F(1,27) = 1,90, p = .179, 𝜂 =
0.07, and the interaction of sequence trial and R-E compatibility failed to reach significance,
F < 1.
We conducted three planned paired t-tests comparing the R-E compatibility
conditions for each of the sequence trial conditions. Significant differences emerged for
neither sequence trial condition, sequence trial 2-3: t(27) = 1.18, p = .249, d = 0.35,
sequence trials 4-5: t(27) = 1.22, p = .232, d = 0.33, sequence trial 6-7: t(27) = 0.57, p =
.574, d = 0.15.
Saccade latency. For the following analyses of saccade latency, only the first effect-
congruent saccade occurring in a trial and fulfilling the selection criteria was selected. Twelve
participants who did not include saccades for each condition were excluded from saccade
latency analyses.
A 2x3 repeated measure ANOVA was conducted to assess the effects of the factors
R-E compatibility and sequence trial on saccade latency within-subjects. Neither the main
effect of R-E compatibility, F(1,15) < 1, nor the main effect of sequence trial, F(2,15) = 1.46,
p = .247, 𝜂 = 0.002, nor their interaction reached significance, F(2,15) = 1.83, p = .178, 𝜂
= 0.11 (see Figure 4B).
Like in Experiment 1, we additionally assessed saccade-manual latency differences
with the same 2x3 repeated measures ANOVA. Neither the main effect of R-E compatibility,
F(1,15) < 1, nor the main effect of sequence trial, F(1,15) < 1, nor their interaction reached
significance, F(2,15) = 2.98, p = .066, 𝜂 = 0.17 (see Figure 4D).
Discussion
In Experiment 2, we unpredictably switched the spatial R-E compatibility after
sequences of three, five, or seven trials. Between experimental sequences, we additionally
inserted random sequences of two to seven trials to decrease the predictability of a switch in
R-E compatibility. This way, we were able to explore whether a similar number of action-
effect pairings was required to first observe a reliable SEC effect indicating proactive effect
monitoring even in an unpredictable environment.
Action-Effect Learning and Anticipatory Saccades 19
To verify that the experimental setting was unpredictable, at the end of the
experiment, participants were asked, to guess the number of different trial sequences they
had faced. None of the participants were able to accurately provide the number of sequences
or any information on their length or regularities. Thus, we assume that our manipulation
successfully created an unpredictable environment.
Furthermore, again we found that participants anticipatorily looked towards their
actions’ future effects. Whereas, the SEC effect did not yet significantly differ from 50% for
sequence trials 2-3, it was significantly larger than 50% from sequence trial 4-5 onward.
Thus, we replicate the findings of Experiment 1 that only a few action-effect pairings are
sufficient for effect monitoring to emerge. Given the sequence trial positions, we observed a
significant SEC effect for, again between one and two prior learning instance were sufficient
for anticipatory saccades (i.e., proactive effect monitoring) to emerge even under
unpredictable environmental conditions.
Again, we did not observe R-E compatibility effects in manual responses within our
sequences of up to seven trials, whereas SEC effects could be observed even under the
unpredictable conditions in Experiment 2. This supports the conclusion that influences of
action effect anticipation on action selection emerge much later than proactive effect
monitoring.
General Discussion
The present study investigated the number of action-effect pairings (i.e., R-E learning
instances) required to form action-effect associations that are sufficiently strong to trigger
proactive effect monitoring (i.e. anticipatory saccades towards future effects). Two
experiments were conducted in which participants manually responded to a forced-choice
repeat/switch target. In case of a correct response, a spatially R-E compatible or R-E
incompatible visual effect followed after a brief delay. In both experiments, the action-effect
mapping switched after a few trials. In Experiment 1, sequences of four, eight, and twelve R-
E compatible/R-E incompatible trials were randomly intermixed. Here, the timing of switches
in R-E compatibility could have been partly predicted by participants. In contrast, in
Action-Effect Learning and Anticipatory Saccades 20
Experiment 2, experimental sequences of three, five, and seven trials were separated by
random sequences to create an environment with unpredictable switches in action-effect
mappings. We assessed both participants´ manual responses and their anticipatory
saccades between target offset and effect onset.
In both experiments, we observed that anticipatory saccades toward future effects
emerged directly within the first four to five sequence trials (i.e., trial two to three per new
action-effect mapping). This means that one to two prior action-effect learning instance were
sufficient to establish action-effect associations strong enough to trigger anticipatory
saccades (i.e., proactive effect monitoring). This finding highlights how little learning
experience is necessary for proactive monitoring processes to emerge. Crucially, this was
the case both under partly predictable as well as unpredictable environmental conditions.
Wolfensteller and Ruge, (2011) found that at least eight action-effect pairings per
response were required to establish action-effect associations strong enough to affect action
selection. In line with their findings, we did not observe a significant effect of R-E
compatibility on manual action selection for our trial sequences of up to six action-effect
pairings per response. Compared to our finding that proactive effect monitoring was evident
in anticipatory saccades after one to two action-effect pairings, this finding further supports
the notion that R-E compatibility effects in manual responses and SEC effects in eye
movements reflect dissociable, functional processes, that is, action selection and proactive
effect monitoring (i.e., attentional shifts toward future effect locations preparing the
comparison of actual to expected effects), respectively (e.g., Pfeuffer et al., 2016,
2020/preprint). Both action selection and proactive effect monitoring are based on
associative learning of bi-directional action effect associations (see e.g. Abrahamse et al.,
2016, for a review on how cognitive control/monitoring processes are grounded in
associative learning in other domains; see e.g., Awh et al., 2012; Feldmann‐Wüstefeld et al.,
2015; Le Pelley et al., 2011; Mitchell et al., 2012, for evidence that attentional orienting can
be influenced by associative learning).
Action-Effect Learning and Anticipatory Saccades 21
Extending the model of Pfeuffer et al. (2016, see also 2020/preprint), we highlight that
proactive effect monitoring, through anticipatory saccades, emerges after just one to two
learning instances when action-effect associations are still rather weak. Our findings illustrate
not only how fast humans can learn action-effect associations and start monitoring their
actions’ effects, but also how fast they can modify these processes to re-learn and adapt to
changes.
It is striking how fast monitoring processes emerged in the present context preceding
influences on action selection by far. The temporal relations seem to imply that monitoring
might be a precursor of action control (e.g., influences on manual action selection) based on
prior associative learning. This notion is also in line with findings by Hommel et al. (2014),
suggesting that attentional control modulates the retrieval of associative bindings. Taken
together, these findings suggest that information on action-effect contingencies derived from
monitoring processes might be used to modulate the expression of action-effect associations
in the planning of effect-generating actions.
It would therefore be of outmost interest to research on cognitive control and
associative learning in other areas to further assess the interplay of associative learning and
monitoring (see e.g., Abrahamse et al., 2016, for previous considerations) including its
timeline to determine whether the early emergence of proactive effect monitoring in the
present study can also be observed for comparable cognitive control operations in other
contexts. Regardless of what such investigations will find, here, we present first evidence
that one to two prior learning instances are sufficient to derive proactive effect monitoring as
demonstrated by anticipatory attentional orienting towards future effects that precedes the
expression of associative learning in manual action control by far.
Conclusion
In conclusion, our findings show that one to two action-effect pairings are sufficient for
proactive effect monitoring processes to emerge regardless of the R-E compatibility. This
suggests that rather weak action-effect associations are sufficient to derive action effect
Action-Effect Learning and Anticipatory Saccades 22
anticipations leading to proactive effect monitoring. In contrast, much stronger action-effect
associations are required to observe influences of action effect anticipation on action
selection. Both proactive effect monitoring and influences on manual action selection derive
from bi-directional action-effect associations (i.e., associative learning). Therefore, this
pattern of results might suggest that monitoring processes modulate the expression of
action-effect associations in action planning based on observed action-effect contingencies.
The present findings thus extend our knowledge of the interplay between learning and
monitoring in human action control.
Action-Effect Learning and Anticipatory Saccades 23
Figure 1: Trial structure of Experiments 1 and 2: A forced-choice repeat/switch target was
followed by a blank screen response frame. Participants’ correct responses produced effects
on the left/right side after a response-effect (R-E) interval (the action-effect displayed
corresponds to the scenario of a correct left key press). For R-E compatible trials, effects
were spatially compatible with responses (e.g., left key press > effect on the left side), and
for R-E incompatible trials, effects were spatially incompatible with responses (e.g., left key
press > effect on the right side). Trials were separated by an intertrial interval (ITI). In
Experiment 1, R-E compatibility switched after a sequence of either 4, 8, or 12 trial and
sequences were randomly intermixed. In Experiment 2, R-E compatibility switched after a
sequence of 3, 5, or 7 trials. Again, trial sequences were randomly intermixed and
additionally separated by 2 to 7 with random R-E compatibility allocation.
Action-Effect Learning and Anticipatory Saccades 24
Figure 2: Manual reaction times (RTs) in A) Experiment 1 and B) Experiment 2 and error
rates in C) Experiment 1 and D) Experiment 2 per R-E compatibility condition (R-E
compatible vs. R-E incompatible) and sequence trial condition (Experiment 1: 2-4 vs. 5-8 vs.
9-12 ; Experiment 2: 2-3 vs. 4-5 vs. 6-7). Error bars depict the standard error (SE) of the
mean.
Action-Effect Learning and Anticipatory Saccades 25
Figure 3: Saccade-effect congruency (SEC) effects, that is, the percentage of saccades
toward the effect’s future location in A) Experiment 1 and B) Experiment 2 per R-E
compatibility condition (R-E compatible vs. R-E incompatible) and sequence trial condition
(Experiment 1: 2-4 vs. 5-8 vs. 9-12; Experiment 2: 2-3 vs. 4-5 vs. 6-7). A value of 50% (grey
dashed line) indicates that an equal number of saccades were directed towards and away
from the future effect, as would be expected by chance. Values above 50% indicate that
participants looked more often towards future effects than away from them, that is,
participants anticipatorily saccaded towards future effects. Error bars depict the standard
error (SE) of the mean.
Action-Effect Learning and Anticipatory Saccades 26
Figure 4: Latency of first effect-congruent saccades in A) Experiment 1 and C) Experiment 2
and saccade-manual latency differences (Δ latency = latency –RT) in B)
Experiment 1 and D) Experiment 2 per R-E compatibility condition (R-E compatible vs. R-E
incompatible) and sequence trial condition (Experiment 1: 2-4 vs. 5-8 vs. 9-12 ; Experiment
2 : 2-3 vs. 4-5 vs. 6-7). Error bars depict the standard error (SE) of the mean.
Action-Effect Learning and Anticipatory Saccades 27
Figure 5: Violin plot depicting the frequency distribution of saccade end positions in Experiment
1 (saccadic gain: 0 = screen center, 1 = effect position, -1 = opposite position) per R-E
compatibility condition (R-E compatible vs. R-E incompatible) and sequence trial condition (2-
4 vs. 5-8 vs. 9-7). Saccade end positions were mapped, so that positive values indicate
saccades towards the effect (green dashed line = effect position) and negative values indicate
saccades away from the effect (orange dashed line = opposite position).
Action-Effect Learning and Anticipatory Saccades 28
Figure 6: Violin plot depicting the frequency distribution of saccade end positions in Experiment
2 (saccadic gain: 0 = screen center, 1 = effect position, -1 = opposite position) per R-E
compatibility condition (R-E compatible vs. R-E incompatible) and sequence trial condition (2-
3 vs. 4-5 vs. 6-7). Saccade end positions were mapped, so that positive values indicate
saccades towards the effect (green dashed line = effect position) and negative values indicate
saccades away from the effect (orange dashed line = opposite position).
Action-Effect Learning and Anticipatory Saccades 29
References
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
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.
Bunlon, F., Marshall, P., Quandt, L., & Bouquet, C. (2015). Influence of Action-Effect
Associations Acquired by Ideomotor Learning on Imitation. PLOS ONE, 10(3),
e0121617, https:/doi.org/10.1371/journal.pone.0121617
Dutzi, I., & Hommel, B. (2009). The microgenesis of action-effect binding. Psychological
Research, 73(3), 425-435. https:/doi.org/10.1007/s00426-008-0161-7
Elsner, B., & Hommel, B. (2001). Effect anticipation and action control. Journal Of
Experimental Psychology: Human Perception And Performance, 27(1), 229-240.
https:/doi.org/10.1037/0096-1523.27.1.229
Erdfelder, E., Faul, F., & Buchner, A. (1996). GPOWER: A general power analysis program.
Behavior Research Methods, Instruments, & Computers, 28, 1-11.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39, 175-191.
Feldmann‐Wüstefeld, T., Uengoer, M., & Schubö, A. (2015). You see what you have
learned. Evidence for an interrelation of associative learning and visual selective
attention. Psychophysiology, 52, 1483-1497.
Greenwald, A. (1970). A choice reaction time test of ideomotor theory. Journal Of
Experimental Psychology, 86(1), 20-25. https:/doi.org/10.1037/h0029960
Harleß, E. (1861). Der Apparat des Willens [The Apparatus of Will]. Zeitschrift für Philosophie
und philosophische Kritik, 38(2),50–73.
Action-Effect Learning and Anticipatory Saccades 30
Herwig, A., & Horstmann, G. (2011). Action–effect associations revealed by eye
movements. Psychonomic Bulletin & Review, 18(3), 531-537.
https:/doi.org/10.3758/s13423-011-0063-3
Hoffmann, J. (2003). Anticipatory Behavioral Control. Anticipatory Behavior in Adaptative
Learning Systems Lecture Notes in Computer Science, 44-65.
https:/doi.org/10.1007/978-3-540-45002-3_4
Hommel, B. (1997). Toward an action-concept model of stimulus-response
compatibility. Advances In Psychology, 281-320. https:/doi.org/10.1016/s0166-
4115(97)80041-6
Hommel, B. (2009). Action control according to TEC (theory of event coding). Psychological
Research, 73(4), 512-526. https:/doi.org/10.1007/s00426-009-0234-2
Hommel, B., Memelink, J., Zmigrod, S., & Colzato, L. S. (2014). Attentional control of the
creation and retrieval of stimulus–response bindings. Psychological Research, 78,
520-538.
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(5), 849-878. https:/doi.org/10.1017/s0140525x01000103
Huestegge, L., & Kreutzfeldt, M. (2012). Action effects in saccade control. Psychonomic
Bulletin & Review, 19(2), 198-203. https:/doi.org/10.3758/s13423-011-0215-5
James, W. (1981). The principles of psychology. Cambridge, MA: Harvard University Press.
(Original work published 1890)
Koch, I., & Kunde, W. (2002). Verbal response-effect compatibility. Memory &
Cognition, 30(8), 1297-1303. https:/doi.org/10.3758/bf03213411
Kunde, W. (2001). Response-effect compatibility in manual choice reaction tasks. Journal Of
Experimental Psychology: Human Perception And Performance, 27(2), 387-394.
https:/doi.org/10.1037/0096-1523.27.2.387
Action-Effect Learning and Anticipatory Saccades 31
Le Pelley, M. E., Beesley, T., & Griffiths, O. (2011). Overt attention and predictiveness in
human contingency learning. Journal of Experimental Psychology: Animal Behavior
Processes, 37, 220-229.
Mitchell, C. J., Griffiths, O., Seetoo, J., & Lovibond, P. F. (2012). Attentional mechanisms in
learned predictiveness. Journal of Experimental Psychology: Animal Behavior
Processes, 38, 191-202.
Pfeuffer, C., Kiesel, A., & Huestegge, L. (2016). A look into the future: Spontaneous
anticipatory saccades reflect processes of anticipatory action control. Journal Of
Experimental Psychology: General, 145(11), 1530-1547.
https:/doi.org/10.1037/xge0000224
Pfeuffer, C.U., Kiesel, A., Huestegge, L. (2020/preprint). Proactive Monitoring does not Differ
between Free Choice and Forced Choice Action Modes.
https:/doi.org/10.13140/RG.2.2.19875.48167/1
Pfister, R., Kiesel, A., & Melcher, T. (2010). Adaptive control of ideomotor effect
anticipations. Acta Psychologica, 135(3), 316-322.
https:/doi.org/10.1016/j.actpsy.2010.08.006
Pfister, R., Pfeuffer, C. U., & Kunde, W. (2014). Perceiving by proxy: Effect-based action
control with unperceivable effects. Cognition, 132(3), 251-261.
Pfister, R., Janczyk, M., Wirth, R., Dignath, D., & Kunde, W. (2014). Thinking with portals:
Revisiting kinematic cues to intention. Cognition, 133(2). 464-473.
https:/doi.org/10.1016/j.cognition.2014.07.012
Prinz, W. (1990). A Common Coding Approach to Perception and Action. Relationships
Between Perception And Action, 167-201. https:/doi.org/10.1007/978-3-642-75348-0_7
Riechelmann, E., Pieczykolan, A., Horstmann, G., Herwig, A., & Huestegge, L. (2017).
Spatio-temporal dynamics of action-effect associations in oculomotor control. Acta
Psychologica, 180, 130-136.
Shin, Y. K., Proctor, R. W., & Capaldi, E. J. (2010). A review of contemporary ideomotor
theory. Psychological Bulletin, 136(6), 943-974.
Action-Effect Learning and Anticipatory Saccades 32
Sun, D., Custers, R., Marien, H., & Aarts, H. (2020). Ideomotor Action: Evidence for
Automaticity in Learning, but Not Execution. Frontiers In Psychology, 11.
https:/doi.org/10.3389/fpsyg.2020.00185
Vallat, R. (2018). Pingouin: statistics in Python. Journal of Open Source Software, 3(31),
1026, https:/doi.org/10.21105/joss.01026
Wolfensteller, U., & Ruge, H. (2011). On the timescale of stimulus-based action–effect
learning. Quarterly Journal Of Experimental Psychology, 64(7), 1273-1289.
https:/doi.org/10.1080/17470218.2010.546417
Ziessler, M., Nattkemper, D., & Frensch, P. (2004). The role of anticipation and intention in
the learning of effects of self-performed actions. Psychological Research, 68(2-3), 163-
175. https:/doi.org/10.1007/s00426-003-0153-6
