**Implicit sequence learning and cognitive control: a pilot study**
**Abstract**
The goal of this pilot study is to replicate the results of Jiménez, Lupiáñez and Vaquero (2009), and to demonstrate that implicit sequence learning (ISL) occurs even after practice with a smaller number of trials (80%) appearing following the trained sequence, and that it is modulated by the sequential congruency effect (SCE) even though the sources of conflict are acquired and implicit. Non sequential trials will be randomly inserted in a trial by trial basis.
After a training period, participants will be transferred to a new context (Jimenez, Vaquero, Lupiáñez, 2006) in which cognitive control engagement could be either be necessary or not to efficiently perform the new task. In this new situation we want to explore 1) whether our manipulation of cognitive control recruitment is effective and 2) its consequences on the expression of both ISL and SCE.
**1. Procedure**
**1.1 Participants**
Considering the effect size of the SCE (η2p = .26) in Jiménez, Lupiáñez and Vaquero (2009), we have computed with G*power 3.1 the minimum number of participants necessary to observe these effects, with 1-β = .90 and α = .05, and with a correlation among repeated measures of 0.6, which provided a minimum number of 23 participants per group. Thus, we decided that 48 participants (24 in each group) will take part in this pilot study in exchange for course credits and after giving their informed consent. However, since we preliminarily want to check if our manipulations are suitable to produce reliable ISL, the learning effect will be analyzed when data are collected from only 12 participants, considering that G *power 3.1 provides an estimation of a minimum of 8 participants to have a significant effect of ISL (η2p = .63, in Jiménez et al., 2009). If learning is observed after 12 participants, then data collection will continue until we complete the whole sample.
All participants will go through a training period during Blocks 1-8 (cfr. 1.2 Training blocks). Then they will be divided in two groups and transferred to two different tasks during Blocks 9-11 (cfr. 1.3 Transfer blocks), while in Blocks 12-13 the learning context will be re-enacted for all of them, to re-establish the learning contingencies (cfr. 1.2 Training blocks). During the final Block 14 all the subjects will perform a Generation task to assess (explicit) sequence awareness (cfr. Generation task).
**1.2 Training blocks**
All the participants will perform a localization task of an “X” target, appearing over four horizontal placeholders on a computer screen, by pressing the spatially correspondent Z, X, N, M keys on a QWERTY keyboard.
Unbeknownst to the participants the target will be presented through a trial-by-trial substitution procedure according to two different and complex spatial sequences, each composed by twelve locations. Specifically, 80% of the trials on average will follow one sequence (i.e. Training sequence, or sequential trials), and 20% of the trials on average will follow the other one (i.e. Control sequence, or non sequential trials). The sequence that is used as training sequence will be counterbalanced across participants.
**1.3 Transfer blocks**
Half of the participants (n=24) will be randomly assigned to the Cognitive Control recruitment group (C+), and half (n=24) to the Non Cognitive Control recruitment group (C-).
C+ group will be instructed to perform a No-go task: in 80% of the trials they will localize again the “X” targets (Go trials), but in 20% of the trials an “O” will appear on the screen following the trained locations, and they will have to withhold the response (No-go trials).
C- group will perform instead a localization task of the “X” targets in 80% of the trials (Go trials), but they will be instructed that in some occasions (i.e. 20% of the times and numerically part of the sequential trials) the placeholders will be presented without any target and they will have to naturally wait for the next trial to appear (Wait trials). Given that no target will show up, the presence of these trials will be marked by a flashing light on the background.
Therefore, each subject in both groups will have on average the same amount of sequential (80%), non-sequential (20% ), and Go trials (80%, both sequential and non-sequential), requiring the same type of localization response, and on average the same percentage of No-go/Wait trials (20%, always sequential) requiring the same non-response. We surmise that the two groups will principally differ in the type of control set behind this non-response: the C+ would probably activate cognitive control not just during No-go trials, but during the whole task, while the C- group could transiently recruit just the amount of cognitive control necessary to override the acquired ISL rhythm which, after eight blocks of training, might have pre-activated the subsequent response in the learned sequence. Indeed, we assume that the participants will perceive these Wait trials as trials that, if presented, would have followed the learned sequence.
**1.4 Generation task**
During this task the participants will be presented twice with twelve pairs of “X” targets appearing in the trained sequence locations and they will localize them. After each pair, four question marks will be presented above the placeholders and they will be asked to press the key corresponding to the third natural subsequent location according to their explicit knowledge about the trained sequence.
**2. Data analyses**
Mean correct RTs per condition will be analyzed with different ANOVA designs, depending on the analyses of the different hypotheses. Outliers (correct RT 3SD slower/faster than average RTs for each participant) will be excluded before computing the means.
All the predicted and the exploratory analyses will be performed with JASP software after collecting the data from all the subjects.
**2.1 Predictions**
We predict that after the training blocks participants will acquire ISL: hence, they will need shorter mean RTs to localize the “X” targets in the sequential trials compared to the non sequential trials (mainly, a main effect of *Trial Type*, which might also be accompanied by a main effect of *Trial Block*, and/or the *Trial Type x Trial Block* interaction).
Moreover, we expect that with practice subjects will process the sequential trials as congruent and the control trials as incongruent, and that their performance will be mediated by the SCE: hence, reduced ISL after an incongruent trial, compared to ISL after a congruent trial (interaction *Trial Type x Previous Trial Type*).
Regarding the Generation task, we don’t expect a significant difference in the number of successors predicted in the sequential locations compared to the non sequential locations.
**2.2 Exploratory analysis**
Furthermore, we want to explore whether our manipulation during the transfer blocks is effective and the directionality of these effects.
Indeed, in Jiménez, Vaquero and Lupiáñez (2006), subjects were trained with the digits 2,4,6,8 appearing always in the congruent locations and with 80% of sequential trials, and then they were transferred to a context in which three odd numbers (from 1, 3, 5, or 9) appeared over the nontarget locations, so that participants were required to search and respond to the location of one even number (target) among three of these odd numbers (distractors). In this transfer block meaan RTs increased enormously (from around 500 to 1000 ms) and no implicit learning was expressed. Results were explained on the basis of the adoption of a control set that eliminated the expression of automaticity (ISL).
In the current study, we seek to further investigate the consequences of adopting a control set on ISL, by forcing participants to inhibit the response to a type of trial never seen before, when transferred to a Go-NoGo task. As a comparison group, other participants will have to avoid responses to blank but marked trials, that could still break the rhythmic sequence of responses. Taking into account the results and the conclusions in Jiménez, Vaquero, Lupiáñez (2006), with our manipulations we have the following exploratory predictions:
*a.* We could find that implicit sequence learning expression is reduced in the C+ group (e.g., or eliminated a tleast in the first block), and unaffected in the C- group. In this situation, we’ll be particularly interested in how is the SCE in the C+ group. If it is reduced or eliminated, it could be that the control mechanism that governs the proactive/tonic cognitive control engagement during the whole Go/No-go task (i.e., the control set), and the control mechanism that governs its phasic recruitment after conflict detection are the same. If instead the SCE is unaffected as in the C- group, then probably proactive and reactive control mechanisms are different.
*b.* We could find that in the C+ group ISL expression is completely eliminated, and that is unaffected in the C- group. We will investigate again the effects on the SCE in both groups, keeping in mind that probably our manipulation was too strong.
*c.* Instead, we could find that in both groups ISL expression is unaffected. We will investigate again the effects on the SCE, considering that probably our manipulation was weak.
*d.* Alternatively, we could find that ISL expression is also unexpectedly influenced in the C- group. We will then explore in a specific analysis how ISL expression is differently affected by No-go versus Wait trials.
**3. Conclusion remarks**
The effect size (when significance is reached) or the non significance of all our predicted or exploratory analyses will be used to adjust the experimental design of subsequent studies, i.e., the number of participants, the proportion of sequential and non sequential trials, the number of training blocks, the number of transfer blocks, the transfer task in the C+ group, the transfer task in the C- group, and the proportion of Go and No-go/Wait trials.
Indeed, the procedure of this pilot study corresponds to the first phase of an experiment in which we seek to investigate how Cognitive Control influences ISL expression on multiple levels: 1) transiently, when trials are presented following an acquired and implicit incongruency, 2) proactively, when transferred to a task requiring tonic cognitive control engagement, and 3) how both proactive and reactive effects are affected when a brain region involved in cognitive control recruitment is inhibited by transcranial magnetic stimulation. However, the hypotheses and the experimental design of the latter depends on the first two findings, which are the main objectives of this pilot study.
**References**
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, 1149-1160.
JASP Team (2018). JASP (Version 0.9)[Computer software].
Jiménez, L., Vaquero, J., & Lupiáñez, J. (2006). Qualitative differences between implicit and explicit sequence learning. Journal of Experimental Psychology-Learning Memory and Cognition, 32(3), 475-490. doi:10.1037/0278-7393.32.3.475
Jiménez, L., Lupiáñez, J. & Vaquero, J. M. M. (2009). Sequential congruency effects in implicit sequence learning. Consciousness and cognition, 18, 690–700. doi:10.1016/j.concog.2009.04.006