Implicit visuospatial sequence representations are accessible in both the practice and the transfer hand

A serial reaction time task was used to test whether the representations of a probabilistic second-order sequence structure are (i) stored in an effector-dependent, effector-independent intrinsic or effector-independent visuospatial code and (ii) are inter-manually accessible. Participants were trained either with the dominant or non-dominant hand. Tests were performed with both hands in the practice sequence, a random sequence, and a mirror sequence. Learning did not differ significantly between left and right-hand practice, suggesting symmetric intermanual transfer from the dominant to the non-dominant hand and vice versa. In the posttest, RTs were shorter for the practice sequence than for the random sequence, and longest for the mirror sequence. Participants were unable to freely generate or recognize the practice sequence, indicating implicit knowledge of the probabilistic sequence structure. Because sequence-specific learning did not differ significantly between hands, we conclude that representations of the probabilistic sequence structure are stored in an effector-independent visuospatial code.


Introduction
Implicit learning can be measured using the serial reaction time task (Nissen & Bullemer, 1987).In this task, participants react as quickly and accurately as possible to a simple, visually presented stimulus that can appear at different predetermined locations on the screen.In intentional (explicit) learning conditions, participants are explicitly told that the stimulus follows a predefined pattern that will be repeated (Jimenez et al., 1996).In incidental (implicit) learning conditions, participants are unaware that the stimulus follows a predetermined pattern of stimulus locations; participants are simply told to respond as quickly and accurately as possible to the stimulus as it appears.The complexity and hence the awareness of the sequence pattern behind the stimulus presentations under implicit learning conditions can be increased, for example, by increasing the predictability of the next stimulus position through second-order sequence structures, i.e., by making the next stimulus presentation predictable only on the basis of the two previous positions.The interesting finding is that under incidental learning conditions, even when participants are unaware of the repeated sequence structure, reaction times to the stimulus progressively decrease over time in the practice condition compared to trials in which the stimulus does not follow a predefined structure and appears randomly at a position (control condition).In this case, participants have acquired implicit knowledge of the underlying sequence structure (Deroost et al., 2006).The aim of the present study is to investigate the types of representation (effector-dependent, effector-independent intrinsic, effector-independent visual-spatial) that are acquired with probabilistic second-order sequence structures (Deroost et al., 2006) and whether they are accessible to the (unpracticed) transfer hand.

Representation types
Implicit learning of sequence structures is often regarded as motor learning (Grafton et al., 2002;Rüsseler & Rösler, 2000).However, in addition to motor learning, the serial reaction time task also involves perceptual learning (Pedraza et al., 2023;Vékony et al., 2023).There is evidence that representations of sequential actions contain information in an effector-dependent (motor) code, an effector-independent intrinsic (motor) code, an effector-independent extrinsic (visuospatial) code, or a combination of them (Dahm, Weigelt et al., 2023;Dahm & Rieger, 2023).
Effector-dependent representations have not always been rigorously defined (for an overview see Shea et al., 2011).For instance, sometimes effector-independent intrinsic representations have been named 'effector-dependent' (e.g., Panzer et al., 2009) although the homologous fingers in the transfer hand are not the same fingers (there is undoubtedly transfer to the other hand).Further, effector-dependent representations can be mistakenly interpreted as being shown by effects in the practice hand only.However, effector-independent representations can influence performance in the transfer hand as well as performance in the practice hand.This is why performance differences between practice and transfer hand are essential to assess effector-dependent representations that are not intermixed with effector-independent representations (Dahm, Weigelt et al., 2023).Effector-dependent sequence representations and effector-independent intrinsic representations are based on learning of the sequence structure of the responses (Koch & Hoffmann, 2000a, 2000b).Effector-independent visuospatial sequence representations are based on learning of the sequence structure of the stimuli (Remillard, 2003;Soetens et al., 2004;Verwey & Clegg, 2005), the sequence structure of external response locations (Willingham et al., 2000).Further, sequence-unspecific representations may evolve by improved stimulus-response coupling, i.e., associations between responses and the upcoming stimulus (Dahm, Hyna, et al., 2023;Stöcker & Hoffmann, 2004;Ziessler & Nattkemper, 2001).

Intermanual transfer
In intermanual transfer, it is assumed that both hands (also the unpracticed transfer hand) benefit from the acquisition of effectorindependent representations.In serial reaction time tasks, learning effects have been observed for the practice as well as for the transfer hand, which indicates effector-independent components (Dahm & Rieger, 2023;Deroost et al., 2006;Grafton et al., 2002;Japikse et al., 2003).For instance, after unimanual practice, participants reacted faster (and with fewer errors) with the transfer hand to the practice sequence than to a random control sequence (Japikse et al., 2003).However, the practice hand is assumed to benefit more than the transfer hand, suggesting effector-dependent learning (Dahm, Weigelt, et al., 2023;Verwey & Clegg, 2005).Separate mechanisms have been proposed for effector-dependent and effector-independent components of sequence learning because both types of representations can be observed independent from each other (Berner & Hoffman, 2009).
In the transfer hand, in addition to extrinsic visuospatial representations, intrinsic representations of the sequence can be observed in a mirror sequence.In the mirror sequence, the use of homologous fingers is indicative for effector-independent intrinsic representations.In contrast, the practice sequence in the transfer hand (using spatially equivalent stimuli and response locations) is indicative for effector-independent extrinsic visuospatial representations.For instance, participants react faster with the transfer hand when performing a practice and a mirror sequence than in a random control sequence (Deroost et al., 2006;Grafton et al., 2002).Effector-independent intrinsic representations have been further supported by positron emission tomography showing that identifying the mirror sequence engages the motor cortex (Grafton et al., 2002).
Intermanual transfer may differ in regard to its direction.Transfer from the dominant to the non-dominant hand (D to ND) may differ from transfer from the non-dominant to the dominant hand (ND to D) due to hemispheric asymmetry.Hemispheric asymmetry has been observed in brain imaging (Schluter et al., 2001) and in deficits following brain lesions (Haaland & Harrington, 1996).These studies demonstrated that actions of the dominant limb involve mainly the dominant hemisphere, whereas actions of the nondominant limb involve both hemispheres.It has been argued that action representations are primarily stored in the dominant hemisphere, irrespective of the hand used during practice (Taylor & Heilman, 1980).This would provide direct access for the dominant hand, but indirect access across the Corpus Callosum for the non-dominant hand.Indeed, more ND to D transfer than D to ND transfer has been observed (Bagesteiro & Sainburg, 2005;Chase & Seidler, 2008).
Alternatively, representations of actions may be stored independently in each hemisphere (Parlow & Kinsbourne, 1989).Following this, representations in the corresponding hemisphere are upgraded with unimanual practice.Those upgrades are then accessible to the other hemisphere.However, during practice of the non-dominant hand, those upgraded representations may still be inferior to the already existing representations in the dominant hemisphere.Hence, storing superior action representations may result in an advantage of D to ND transfer.This has been supported by studies observing D to ND transfer, but no ND to D transfer (Balitsky Thompson & Henriques, 2010;Mostafa et al., 2014).
These inconclusive observations on the direction of intermanual transfer are even more ambiguous when considering observations of symmetric intermanual transfer which is similar in both directions (Chase & Seidler, 2008;Dahm & Rieger, 2023;Panzer et al., 2009;Wohldmann et al., 2008).For instance, no differences were observed in the direction of intermanual transfer when typing practiced and unpracticed 4-digit numbers (Wohldmann et al., 2008, Exp. 2).
One factor influencing the direction of intermanual transfer may be explicitness of learning.For instance, symmetric intermanual transfer was observed in a serial reaction time task, but not after visuomotor adaptation in a reaching task, where more ND to D transfer than D to ND transfer was observed (right-handers in Chase & Seidler, 2008).However, participants were informed in advance to each sequence whether the stimuli would appear sequentially or randomly.This may have led to explicit learning strategies in the serial reaction time task, whereas learning in the visuomotor adaptation task was implicit.

The Present Study
In the present study, participants performed probabilistic second-order structures (Deroost et al., 2006, Exp. 2) because probabilistic structures (i.e., complex conditional associations) are less likely to become explicit than deterministic sequences (i.e., straightforward predictable associations) (Howard & Howard, 1997;Vékony et al., 2022), although motor and perceptual learning does occur (Nemeth et al., 2009).Participants performed a practice, a mirror, and a random structure, each before and after practice.It was expected to replicate the D to ND transfer of probabilistic structure representations (Deroost et al., 2006).Although ND to D transfer has not yet been investigated using a probabilistic structure, it was expected to occur similarly as in deterministic sequences (Dahm & Rieger, 2023;Exp. 2 in Wohldmann et al., 2008).
It has been suggested that effector-independent visuospatial representations are acquired independently of the practiced (dominant or non-dominant) hand and are therefore available in both transfer directions (Dahm et al., 2023;Dahm & Rieger, 2023).In contrast, effector-dependent and effector-independent intrinsic representations may differ between transfer directions.Sequence-specific learning was expected to be observable in the practice hand by shorter Reaction Times (RTs) in the Practice Structure (same visuospatial location and effectors) than in the Control and Mirror Structures (different visuospatial location and effectors).Effectordependent representations were expected to be observable in the Practice Structure by shorter RTs in the practice hand than in the transfer hand.Effector-independent intrinsic representations were expected to be observable in the transfer hand by shorter RTs in the Mirror Structure (different visuospatial locations, but homologous effectors) than in a Random Control Structure (different visuospatial locations and effectors).Effector-independent visuospatial representations were expected to be observable in the transfer hand by shorter RTs in the Practice Structure (same visuospatial locations, but different effectors) than in a Random Control Structure.

Participants
Originally, 87 undergraduate university students participated in the study.Two participants were excluded from analysis due to extreme outliers either in RTs or in error rates (above 5 SD).The remaining 85 participants (13 male) had a mean age of 22.5 years (SD = 4.3, range: 18-46 years) and were all strongly right-handed (Laterality Index above 80, M = 99; Oldfield, 1971).A maximum of 15 participants were assessed simultaneously on separate computers for approximately 1 h.All participants gave informed consent and received course credit.

Experimental procedure
The experiment was programmed using the software OpenSesame version 3.1.6(Mathôt et al., 2012).The experiment file is available at https://osf.io/s3emg/.The experimental procedure is shown in Fig. 1.
Participants started the experiment with a Familiarization Block (66 trials, in Random Structure, Hand counterbalanced).At the beginning of each block, participants were asked to put four fingers (index, middle, ring, and little finger) either of the right hand or left hand on the response keys ('f', 'g', 'h', and 'j').On each trial, four horizontally aligned white circles were presented (diameter: 2.5 cm, distance: 0.5 cm).The positions of the circles (leftmost, left, right, and rightmost) were spatially aligned to the response keys.After 50 ms a black X (font size = 44) appeared in one of the circles.The stimulus remained on screen until the participant pressed the key corresponding to the location of the X.Incorrect responses did not influence the upcoming stimulus of the sequence.With correct responses, the X disappeared for 50 ms and appeared in one of the circles, initiating the next trial.Due to the possibility of repeating S.F.Dahm et al. stimuli a Response-to-Stimulus Interval (RSI = 50 ms) was used after each trial.During the RSI the circles were empty (see Fig. 2).The RSI was kept short because it has been shown that participants' sequence knowledge is reflected better in RTs with short intervals than with long intervals (Kiss et al., 2022).Participants were instructed to press the corresponding response keys as fast as possible.After a block, participants received feedback about their error rate and the average RT (Fig. 2).
The following Pretest consisted of Practice, Random, and Mirror Structure Blocks (counterbalanced order, 98 trials each).Each Structure (Practice, Random, Mirror) was performed with the dominant and non-dominant hand (blocked and counterbalanced order).Even in Random Structures the X appeared equally often in each circle and the percentage of first-order repetitions was 25 %.Second order transitions in the Random Structure were compatible to 50 % with the Practice Structure and to 50 % with the Mirror Structure.The Practice Structure is shown in Table 1.Here, the first two Xs of a block appeared randomly in one of the circles.The following 96 Xs were quasi-randomized depending on the two preceding trials.Two consecutive stimuli could be followed by only two of the four possible stimuli (Table 1; Deroost et al., 2006).In the Mirror Structure, the Practice Structure was reversed.As a consequence, two consecutive stimuli in the Mirror Structure were always followed by one of the two stimuli that did not occur in the Practice Structure.
The pretest was followed by a Practice Phase in which participants performed 12 blocks of 98 trials (see RTs of the Practice Blocks in Appendix B) of the Practice Structure either with the dominant (right) hand (N = 44) or the non-dominant (left) hand (N = 41).The Practice Structure was the one shown in Table 1 or the reverse structure (counterbalanced across participants).For instance, participant Jane Doe, who practiced the structure in Table 1, was requested to respond with 'g' in the first response and 'j' in the second response.Hence, the third response in her practice structure would have been either 'h' or 'j' (example structure: g-j-h-j-f-g-h-… see Fig. 2).
Practice was followed by a Posttest that consisted of the three structures in the same order as in the Pretest.After the Posttest,  participants were asked whether they could freely generate and recognize the practiced response pattern (Practice Structure).To do this, participants reported whether they had noticed a certain structure in the order of stimuli or responses.Then, in a Free Generation Test, irrespective of their preceding answer, they pressed 34 responses in an order that might have occurred during practice.Finally, in a Recognition Test, participants were asked to focus on the structure in the order of the stimuli and the responses they were going to
Corresponding responses of preceding stimuli Corresponding responses of possible next stimuli Note. "f", "g", "h", and "j" represent the required keyboard keys.perform.After completing a block of 34 trials, participants rated (from 1 to very unlikely to 9 -very likely) whether the performed structure occurred during practice or not.This was done with the Practice Hand with the Practice Structure, the Random Structure, and the Mirror Structure (in randomized order).

Data analysis
RT and accuracy were recorded for each trial.Error rates ranged from 3 to 5 % per block (SD = 2-3 %).RTs of incorrect responses as well as RTs of the two responses following an error were not included in the analyses.For the remaining responses median RTs were calculated for each block depending on structure and hand.Statistical significance was set at p < 0.05.Dependent variables were analyzed using mixed model ANOVAs (Kassambara, 2021) for which we report the effect size ƞ p 2 .Further comparisons were conducted using t-tests with Holm-adjusted p-values for multiple pairwise comparisons (effect size d).Data and data preparation scripts are available at: https://osf.io/s3emg/.significantly longest in the Mirror Structure (M ± SD = 429 ± 40 ms, t( 84) ≥ 2.6, p Holm ≤ .017,d ≥ 0.28).All remaining effects were not significant (Test x Test Hand x Structure: F(2, 168) = 1, p = .367,ƞ p 2 = .01;all others: F < 1).

Detection of the structure
24 participants reported having detected a certain structure in the order of the stimuli and responses.To control for the amount of explicit learning of the structure, we calculated the percentage of responses in the Free Generation Test that were compatible with the Practice structure.The probability of a compatible response by chance was 50 % because two responses out of four matched with the practiced sequence structure.A one-sample t-test of the mean (M ± SD = 50.1 ± 14 %) against chance was not significant, t(84) = 0.1, p = .926,CI 95 % [47.2; 53.1], d = 0.01, indicating that the responses did not differ significantly from the guessing rate.An unpaired two-sample t-test, t(83) = 1.3, p = .193,CI 95 % [-10.8; 2.2], d = 0.32, indicated that the compatibility of the freely generated responses did not significantly differ between participants who reported noticing a structure (M ± SD = 51 ± 14 %) and those who reported not noticing a structure (M ± SD = 47 ± 13 %).
To control for Recognition of the structure, a mixed ANOVA with the between-subjects factor Report (Structure Detected, Structure Not Detected) and the within-subject factor Structure (Practice, Random, Mirror) was conducted on Recognition Ratings.Means and standard errors of Recognition Ratings are shown in Fig. 4. The main effect of Structure, F(2, 166) = 1.9, p = .152,ƞ p 2 = .02,the main effect of Report, F(1, 83) = 1.1, p = .289,ƞ p 2 = .01,and the interaction between Report and Structure, F(2, 166) = 1, p = .388,ƞ p 2 = .01,were not significant.

Discussion
The present study investigated the acquired representation types that lead to performance gains in a probabilistic serial reaction time task.In addition, the direction of intermanual transfer was investigated.Independent from Structure and Hand, Posttest RTs were shorter than Pretest RTs, indicating sequence-unspecific learning, e.g., task habituation.Further, RTs were shorter with the right hand than with the left hand indicating hand dominance in the right-handed sample.RTs did not significantly differ between Practice Hand and Transfer Hand indicating an absence of effector-dependent representations.In the Posttest, RTs were shorter in the Practice Structure than in the Random and Mirror Structure, even though participants were unable to freely generate or recognize the Practice Structure.This suggests sequence learning and implicit representations of the probabilistic structure.Because this was observed in both the Practice and Transfer Hand, such representations were effector-independent and visuospatial in nature indicating perceptual codes (Pedraza et al., 2023;Vékony et al., 2023).There was no significant difference in the direction of intermanual transfer suggesting that implicit visuospatial representations were acquired and accessible to both the dominant and non-dominant hand.
Irrespective of the structure and tested hand, participants' responses were shorter after practice than before practice.Such general improvements in responses have been observed in similar tasks (Dahm, Hyna, et al., 2023;Dahm & Rieger, 2023;Wohldmann et al., 2008).For instance, when using the practice hand, participants' responses were shorter not only for practiced (4-digit) numbers, but also for unpracticed numbers (Wohldmann et al., 2008).Participants may have internalized the association of the stimuli with the corresponding keypresses.Furthermore, participants may have optimized their perception and encoding speed of the visual stimuli (Wilms et al., 2013).Thus, hand and structure independent adaptations played a crucial role in optimizing performance in the serial reaction time task (Dahm, Hyna, et al., 2023).
Although some participants reported noticing a sequential pattern during the practice blocks, their performance was not significantly different from those who did not report noticing a pattern.Both groups of participants were unable to freely generate the Practice Structure.Similarly, both groups were unable to distinguish the structures after performance.Hence, the structure was not explicitly available (Vékony et al., 2022).Most importantly, although the effect size was small, RTs in the Posttest were shorter for the Practice Structure than for the Random Structure.Hence, the structure was not only implicitly learned, but also implicitly represented.Such implicit representations without explicit awareness of the sequence structure have been observed in several previous studies (Japikse et al., 2003;Reber & Squire, 1998;Rüsseler & Rösler, 2000).For instance, RTs were shorter on trials that were congruent with the practiced sequence than on incongruent trials, even though participants reported in post-experimental interviews that a pattern search strategy was not helpful (Japikse et al., 2003).These results are also consistent with the notion that implicit learning is revealed by performance improvements that participants cannot explain explicitly (Reber, 1989;Vékony et al., 2022).
Implicit sequence learning has been considered as a form of motor learning, i.e., the learning of a sequence of the effectors (Grafton et al., 2002;Rüsseler & Rösler, 2000).In the case of the transfer hand, effector-independent intrinsic representations may have provided an advantage to Mirror Structures requiring homologous muscles over Random Structures.However, the opposite was observed.In the transfer hand, shorter RTs in the Practice Structure than in the Random and Mirror Structure indicate effectorindependent visuospatial representations.Hence, the sequence structure is stored in a visuospatial code rather than an intrinsic motor code (Dahm & Rieger, 2023;Soetens et al., 2004).Participants may have implicitly learned where the subsequent cue will not appear.In other words, they may have learned to anticipate the position of the subsequent cues by ignoring improbable positions, which may have optimized visual stimulus processing (Koch & Hoffmann, 2000b;Pedraza et al., 2023;Vékony et al., 2023).Future studies could investigate the representation types using auditory stimulus material (Zhuang et al., 2009).
As expected, responses with the dominant hand were faster than responses with the non-dominant hand.This finding is consistent with faster responses with the dominant hand than with the non-dominant hand for instance in typing 4-digit numbers (Wohldmann et al., 2008).Although handedness had a global effect on RTs, no significant interactions were observed that would have shown differences between the group that practiced with the non-dominant hand and the group that practiced with the dominant hand.
Hence, irrespective of the transfer direction, sequence-unspecific learning was transferred to the transfer hand.Furthermore, irrespective of the transfer direction, implicitly acquired representations of the sequence structure were available to the transfer hand.This indicates that symmetric intermanual transfer was observed in both directions: D to ND and ND to D. Such symmetric intermanual transfer is consistent with previous studies investigating implicit learning of sequences (Panzer et al., 2009;Wohldmann et al., 2008).Hence, effector-independent visuospatial representations of an action that are stored and optimized during repetitive execution, are also accessible to the unpracticed limb, irrespective of whether the non-dominant or dominant hand was used during practice.
Performance in the transfer hand did not differ significantly from performance in the practice hand.This was unexpected because performance improvements in sequence learning are not always fully reproducible in the transfer hand (compared to the practice hand).For instance, although participants were faster with the transfer hand when responding to a practiced sequence than to a control sequence, RTs in the transfer hand were longer than in the practice hand, indicating an effector-dependent component of sequence representations (Dahm, Weigelt et al., 2023;Dahm & Rieger, 2023;Verwey & Clegg, 2005).In contrast, in the present study, performance improvements in the practice hand were fully reflected in the transfer hand.Hence, practice hand and transfer hand benefited equally from unimanual practice, suggesting an absence of effector-dependent representations.Similar performance with the practice and transfer hand was previously observed in a probabilistic serial reaction time task (Deroost et al., 2006).Most likely, the complexity of the probabilistic sequence structure triggered an effector-independent visuospatial representation of the structure, which is rather perceptual in nature and therefore available in both hands.

Limitations
While the difference between sequence structures of 5 ms was rather small in its magnitude, the sequence-specific effects were consistently observed across participants and showed a clear pattern with the Random Structure falling between the Practice Structure and the Mirror Structure.Hence, although the sequence structure was complicated (Deroost et al., 2006), included relatively little practice with only one practice session (Dahm & Rieger, 2023), and (compared to a between-design) the within-factorial design with multiple blocks in the posttest may have led to interferences (Dahm, Weigelt et al., 2023), we were able to show that the probabilistic second-order sequence structure was learned by the participants (Deroost et al., 2006).
The present results suggest that by using a probabilistic serial reaction time task with visual stimuli, effector-independent visuospatial representations were acquired rather than effector-dependent or effector-independent intrinsic representations.The lack of the latter may be caused by the amount of practice (Hikosaka et al., 1999;Verwey & Clegg, 2005).Possibly, effector-dependent or effectorindependent intrinsic representations would have been acquired by adding more practice blocks and more sessions.It has been postulated that effector-unspecific (visuospatial) representations are acquired at early stages of learning, whereas effector-specific intrinsic (motor) representations which may include parameters such as muscle activation are acquired at later stages of learning (Hikosaka et al., 1999).This may not only affect the retrieval of the sequence structure during intermanual transfer, but also the direction of intermanual transfer (Kirsch & Hoffmann, 2010).In the present study, practice was extensive enough to offset the righthand dominance in the final practice blocks of the left-hand practice group, but it remains speculative whether effector-dependent or effector-independent intrinsic representations would have been acquired at later stages of learning.
Considering generalizability of our results, it remains speculative whether applied tasks (e.g.involving whole-body actions) would reveal similar results.However, regarding the alternating serial reaction time task (ASRT: Howard & Howard, 1997;Kiss et al., 2022;Nemeth et al., 2010) or when interspersing a second order sequence with single events from the reverse structure (Shanks et al., 2003), we would expect similar results as in the probabilistic sequence task used in the present study.In the ASRT, every second element is fixed while the others are random which similarly results in a second-order contingency and a higher probability of some triplets compared to other triplets.

Conclusion
The present results indicate that effector-independent visuospatial representations are acquired when practicing a probabilistic second-order structure in the serial reaction time task.Two arguments supported effector-independent visuospatial representations.First, performance in the transfer hand did not differ significantly from performance in the practice hand, which indicates a lack of evidence for effector-dependent representations.Second, RTs in the Practice Structure were shorter than RTs in a Random Structure and lowest in the Mirror Structure indicating a lack of evidence for effector-independent intrinsic representations.Therefore, the representations of the sequence structure were most likely encoded perceptually (Pedraza et al., 2023).Apart from that, symmetric intermanual transfer was observed in both directions, suggesting that the enhanced visuospatial representations are accessible to both hands, irrespective of whether the dominant or non-dominant hand was used during practice.S.F.Dahm et al.

Fig. 1 .
Fig. 1.Depiction of the experimental design.The order of hands and sequence structures is exemplary.Familiarization was performed either with the Practice or Transfer hand.The order of the hands in the Pretest and Posttest was counterbalanced across participants.Further, the order of the Practice Block (P), Random Block (R), and Mirror Block (M) in the tests was random.Recall Questions (Q), Free Generation (F), and Recognition Tests (double filled) were always performed with the Practice Hand.

Fig. 2 .
Fig. 2. Depiction of an experimental block of 98 trials as used in the Practice Phase.Four blank circles were shown for 50 ms.Then an X appeared in one of the circles for each trial.Only with a correct response the next trial was initiated.Due to the probabilistic sequence structure, depending on the previous two locations of the X, only two locations were possible.The dashed stimuli indicate the alternative locations.Each block ended with block feedback about error rate and average RT.

Fig. 4 .
Fig. 4. Means and standard errors of the Recognition Ratings in the Practice Structure, Mirror Structure, and Random Structure in participants who reported noticing a structure and those who reported not noticing a structure.