Perceptual sequence learning in a serial reaction time task

Sunbin Song; James H Howard, Jr; Darlene V Howard

doi:10.1007/s00221-008-1411-z

. Author manuscript; available in PMC: 2009 Aug 1.

Published in final edited form as: Exp Brain Res. 2008 May 14;189(2):145–158. doi: 10.1007/s00221-008-1411-z

Perceptual sequence learning in a serial reaction time task

Sunbin Song ^1,^*, James H Howard Jr ^2,³, Darlene V Howard ¹

PMCID: PMC2672106 NIHMSID: NIHMS104496 PMID: 18478209

Abstract

In the serial reaction time task (SRTT), a sequence of visuo-spatial cues instructs subjects to perform a sequence of movements which follow a repeating pattern. Though motor responses are known to support implicit sequence learning in this task, the goal of the present experiments is to determine whether observation of the sequence of cues alone can also yield evidence of implicit sequence learning. This question has been difficult to answer because in previous research, performance improvements which appeared to be due to implicit perceptual sequence learning could also be due to spontaneous increases in explicit knowledge of the sequence. The present experiments use probabilistic sequences to prevent the spontaneous development of explicit awareness. They include a training phase, during which half of the subjects observe and the other half respond, followed by a transfer phase in which everyone responds. Results show that observation alone can support sequence learning, which translates at transfer into equivalent performance as that of a group who made motor responses during training. However, perceptual learning or its expression is sensitive to changes in target colors, and its expression is impaired by concurrent explicit search. Motor-response based learning is not affected by these manipulations. Thus, observation alone can support implicit sequence learning, even of higher order probabilistic sequences. However, perceptual learning can be prevented or concealed by variations of stimuli or task demands.

Keywords: Implicit, Explicit, Perceptual Learning, Sequence Learning, Motor Learning

Introduction

Motor sequence learning is often studied with the serial response time task (SRTT) which was first introduced by Nissen and Bullemer in 1987. In it, participants make a motor response by pressing a button which corresponds to a target appearing in one of four locations on a screen. Unbeknownst to the subject, the sequence of targets follows a fixed order which repeats many times. Subjects exhibit sequence learning by being faster and more accurate on blocks of patterned trials as opposed to a block of randomly ordered stimuli. Learning in this task is often considered implicit when it is unaccompanied by conscious awareness of the pattern. Since its inception, the SRTT has become increasingly influential for psychological, cognitive, and clinical studies of motor sequence learning. Despite this widespread use, controversy still surrounds this task, because performance measures may reflect explicit (declarative) knowledge of the sequence which can arise spontaneously, and also because there is disagreement regarding whether performance on this tasks reflects perceptual, in addition to motor, learning (Robertson, 2007).

Even though spontaneous explicit awareness of the pattern often develops with training, multiple studies have established that sequence learning on the SRTT task can be implicit (Willingham et al., 1989; Willingham et al., 2002; Robertson, 2007). It has also become generally accepted that implicit sequence learning on the SRTT can be supported by motor responses and/or motor response locations (Willingham et al., 1989; Willingham, 1999; Rüsseler and Rösler, 2000; Willingham et al., 2000; Deroost and Soetens, 2006a). It has proven much more difficult however to determine the extent to which perceptual learning contributes to performance on the SRTT task, because observation without motor responding results in increased spontaneous explicit awareness of the pattern (Willingham et al., 1989; Howard et al., 1992).

Many different types of stimuli have been used to direct motor responses in the SRTT including auditory cues (Dennis et al., 2006), color cues (Willingham, 1999), letter cues (Rüsseler and Rösler, 2000), spatial cues (Nissen and Bullemer, 1987), or combinations of these various choices (Robertson and Pascual-Leone, 2001). The classic and most often employed version however uses spatially oriented targets. For spatial targets, some studies have shown that purely perceptual sequence learning does occur (Howard et al., 1992; Mayr, 1996; Remillard, 2003) while some studies have shown that it does not (Willingham et al., 1989; Willingham, 1999; Kelly and Burton, 2001). In instances where perceptual learning has been found to occur, it has been unclear whether increases in performance were in fact attributable to perceptual learning, or were reflections of spontaneous increases in explicit awareness or oculomotor sequence learning.

For example, Howard et al. (1992) found that subjects who had simply observed a fixed sequence during a training phase showed as much reaction time difference between patterned and random trials upon responding as those who had made motor responses the entire time. This seemed to suggest that learning on the SRTT could be purely perceptual. However, subjects who only observed the sequence also had more explicit awareness of the pattern, making it unclear whether performance improvements were due to implicit perceptual learning or to increased explicit awareness of the sequence. Evidence in favor of the latter interpretation was reported by Willingham (1999), who first replicated Howard et al, but then removed subjects with greater explicit awareness. No perceptual sequence learning was revealed by the remaining subjects.

For studies in which spontaneous explicit awareness had not developed, oculomotor sequence learning, rather than visuo-spatial sequence learning, appeared to be a likely explanation for performance improvements. For example, in Mayr (1996), two independent sequences were determined by a succession of locations and objects. Subjects made motor responses only to object sequences but they displayed learning for both the location-based and the motor-based object sequences, which seemed to suggest visuo-spatial sequence learning had occurred. However, stimuli were placed far apart along four corners of an imaginary square with a side length of 22cm, and so eye movements could have led to sequence learning. In Willingham (1989), targets were horizontally separated by no more than 4.7° of the visual field to minimize ocular movements, and subjects responded to the color of a target which appeared in one of the four locations. Subjects displayed sequence learning when the color-based motor responses followed a sequence and the targets had appeared in random locations. In contrast, they failed to reveal sequence learning when the targets followed a spatial sequence but color-based motor responses were randomly determined. This study suggested that without large eye movements, perceptual sequence learning did not occur.

The present experiments were conducted to see if observation alone could support sequence learning even when eye movements were minimized and when spontaneous increases in explicit awareness did not affect performance measures. Therefore, in both experiments, participants completed a training phase and a transfer phase of an SRTT paradigm in which the same stimulus sequence occurred throughout both phases. Half of the participants responded during both phases, while the other half just observed during the training phase and then responded during transfer.

To minimize oculomotor learning, adjacent locations were horizontally separated by 3.8°, which is comparable to the horizontally separated (4.7°) targets used in the study by Willingham et al. (1989), and corresponds to foveal or near parafoveal space. Second, to prevent the development of spontaneous explicit awareness, probabilistic sequences were used instead of fixed ones. Unlike fixed sequences which rely on repetitions of 6- to 12-unit regularities, probabilistic sequences depend on differing relative frequencies, and learning is shown via faster and more accurate performance on high- versus low frequency sequences. Probabilistic sequences have advantages over fixed sequences in that spontaneous explicit awareness does not readily occur (Howard and Howard, 1997; Schvaneveldt and Gomez, 1998). Here we used the Alternating Serial Response Time (ASRT) task in which predictable Pattern trials alternate with unpredictable Random ones which causes certain “triplets” (a sequence of three trials) to occur at a higher frequency than others. Previous research has shown that participants do not gain explicit knowledge spontaneously in this task, and performance is sensitive to the relative frequencies of triplets, not to the alternating regularity (Howard et al. 2004).

The present experiments also addressed one further concern in observation based studies – the problem of attention to stimuli. For example, Kelly and Burton (2001) found no perceptual learning in an SRTT task, but this negative result was obtained from subjects who had no task to perform during the observation session. Thus, it was unclear whether these subjects were paying attention to the stimuli. For this reason, in both of the current experiments, all subjects were given tasks which would assure attention to the stimuli during the training phase.

We addressed an additional question relevant to the SRTT literature, namely whether or not perceptual learning is influenced by an explicit mode of learning. Some evidence suggests that explicit knowledge of the sequence can in fact increase perceptual sequence learning. Using stationary letter cues, Rüsseler and Rösler (2000) found that implicit learners showed ERP components in response to motor sequence deviants only, whereas explicit learners showed ERP components when confronted with either motor or perceptual sequence deviants. This additional ERP effect for perceptual deviants occurred only when subjects had successfully acquired explicit sequence knowledge, and not when they were only searching for the pattern (Rüsseler et al., 2003). It was unclear from these experiments, however, whether explicit knowledge improved implicit perceptual learning, or if ERP effects simply reflected the effects of explicit knowledge without having altered implicit perceptual learning. In the first experiment reported here, we explored these questions using a variant of the ASRT in which implicit memory could be measured in both explicit and implicit learners. From these implicit memory measures, implicit learning could be inferred even in subjects who had learned in an explicit mode.

Experiment 1

In Experiment 1, we asked whether perceptual sequence learning could occur by observation alone during a training phase and be reflected in performance measures in a transfer phase. Additionally, we asked whether perceptual learning was influenced by the absence or presence of explicit knowledge during the training phase. For the training phase, subjects were assigned to one of two response groups: either they made motor responses during the training phase (the Respond group) or they did not (the Observe group). In the transfer session, all subjects made motor responses. We assumed that by comparing these two groups during transfer, we could determine the extent to which perceptual processing in the absence of motor responding could support implicit sequence learning.

Within these two response groups, subjects were assigned to one of two Instruction groups: Intentional or Incidental. In an earlier study, it was found that cues could facilitate explicit awareness of the sequence in Intentional learners who would have otherwise had much difficulty in gaining explicit knowledge of the pattern (Song et al., 2007). When these color cues were removed, Intentional subjects displayed no explicit knowledge of the pattern in Uncued blocks, even though it was a pattern identical to that of the Cued blocks. Incidental learners showed no explicit awareness in either Cued or Uncued blocks. Since the removal of cues presumably removed explicit awareness, these Uncued blocks assessed implicit memory which was assumed to reflect implicit learning. By placing Uncued blocks at the start of the transfer session, we attempted to assess both whether or not observation alone led to implicit perceptual learning, and also, if an explicit mode of learning influenced perceptual learning.

Materials and Methods

Participants

Fifty-six right-handed volunteers (aged 18 to 22 years) were recruited from Georgetown University to participate in the study for credit or payment. All subjects were naïve to SRTT based tasks and all signed informed consent approved by the Georgetown University IRB. Subjects were randomly assigned to one of four groups representing the combination of two response conditions (Observe or Respond) by two instruction conditions (Intentional and Incidental). A schematic of the study design is presented in Figure 1. The entire test visit for each subject took approximately one hour.

For all subjects, there were two types of epochs: Cued (1, 2, 3, and 6) and Uncued (4 and 5). For all four groups, all three epochs in the training session were Cued epochs in which Pattern trials were marked by grey targets while Random trials were marked by black targets. In the transfer session however, the first two epochs were Uncued epochs. For Uncued epochs, color cues were removed such that all targets were black. Half of the subjects in each response condition were given Incidental Instructions, and half were given Intentional Instructions. Uncued epochs measure implicit learning for all groups regardless of instruction, whereas Cued epochs measured implicit and/or explicit sequence learning, depending on instruction (Song et al., 2007).

Subjects in the Observe Intentional condition reported the 4-unit pattern after each block in the 1st session. Subjects in the Observe Incidental condition were given a comparable simple task in which they counted 0 to 7 red targets which appeared in each block instead of black targets. Subjects in the Respond groups made button press responses throughout both sessions. In addition, Respond Intentional subjects reported the 4-unit pattern after each block in the 1st session. However, all subjects in all four groups were treated identically during transfer session, and all four groups responded.

Design and Procedure

All subjects completed a training and transfer session of the Alternating Serial Response Time task (ASRT). Those in the Observe condition watched the display during the training session and responded via button press only during the transfer session. Those in the Respond condition responded via button press during both sessions. For all blocks and all conditions, participants saw a filled in circle (target) that appeared in one of four open circles arranged horizontally (3cm from center to center, 1cm diameter, the four cues together occupied approximately 3.8° of the visual angle) on a computer screen. Each block began with 5 random trials (for warm-up), followed by 80 experimental trials. These 80 trials consisted of 10 repetitions of an 8-item sequence, in which four Pattern trials alternated with four Random trials (e.g., 1r3r4r2r for an imbedded pattern of 1342 where 1 refers to the left-most position and 4 to the right-most). Each subject received one of the six possible patterns (1r2r3r4r, 1r2r4r3r, 1r3r4r2r, 1r3r2r4r, 1r4r2r3r, 1r4r3r2r) and for a given subject, this same pattern occurred throughout all blocks. There were 15 blocks in the training session, and 20 blocks in the transfer session. These blocks were grouped for purposes of analysis into 7 epochs containing 5 blocks each.

In sessions requiring a motor response, subjects responded by pressing one of four corresponding keys on the computer keyboard using the four fingers of the dominant hand. The response-to-stimulus interval was 120ms, and the circle remained filled in until participants pressed the correct key, at which time the circle cleared and another target appeared after a delay of 120ms. By giving end-of-block feedback to focus more on accuracy or on speed, subjects were guided to 92% overall accuracy. For observe sessions, stimuli remained onscreen for 380ms so as to be similar to the on-time in the Respond condition, and the inter-stimulus interval was 120ms.

For the training session, all Pattern targets were grey while Random targets were black, such that epochs in the training session were Cued (Epochs 1, 2, 3). In the transfer session, the beginning two epochs (4 and 5) were Uncued in that all targets were black and the last two epochs (6 and 7) were again Cued, with grey targets demarcating Pattern trials and black targets demarcating Random ones.

Half of the subjects in each response condition were given Incidental instructions, and half were given Intentional instructions. Those given Incidental instructions were not alerted to the presence of any regularity and were not instructed to look for one. Those given Intentional instructions were told to learn a repeating 4-unit pattern that occurred on every other trial on the Cued epochs and were told that these patterned trials were marked by grey targets whereas black targets were always randomly positioned. After each block, they were asked to report the 4-unit pattern followed by the trials with grey targets. For Uncued epochs, no subject was alerted to the presence of the regularity although it was identical to the pattern in the Cued epochs.

A previous study demonstrated that subjects given Intentional instructions gained explicit knowledge of the repeating pattern in Cued epochs but were unaware of it during the Uncued epochs (Song et al., 2007). In contrast, those given Incidental instructions did not notice the pattern in the Cued nor Uncued epochs. Thus, explicit knowledge did not affect performance measures on Uncued epochs for any groups regardless of instruction, whereas performance on Cued epochs reflected explicit sequence knowledge, depending on instruction (Song et al., 2007).

To ensure that subjects in the Observe groups were paying attention to the stimuli, the following manipulations were employed during the training session. For subjects in the Observe-Intentional condition, having to explicitly learn and report the 4-unit pattern after each block in the training session ensured that they were observing the stimuli. Subjects were not told that the sequence was identical each time, in order to give them reason to always pay attention to targets. Subjects in the Observe-Incidental condition were given a comparable task in which they counted 0 to 7 red targets which appeared randomly in each block instead of black targets. Accuracy on this detection task was 96.7 +/− 4.3% which confirms that subjects were paying attention to the stimuli. Subjects in the Observe-Intentional condition also saw the red targets but were told to ignore them. During the transfer session, all subjects in all four groups were treated identically, and all groups made button press responses.

Although explicit knowledge on this cued variant of the ASRT is deterministic (Pattern trials follow a four-unit regularity), earlier research indicates that motor performance measures reflect probabilistic learning. That is, performance is not sensitive to the alternating regularity, but rather to the relative frequencies of “triplets” (a sequence of three trials) (Howard et al. 2004). The sequence structure in the ASRT, where predictable Pattern trials alternate with unpredictable Random ones, causes certain triplets of trials to occur at a higher frequency than others. For example, when the pattern is 1r3r2r4r, triplets beginning with 1 and ending with 3 (e.g. 123, 143, 113 etc.) are high frequency whereas those beginning with 3 and ending with 1 (e.g. 321, 341, 311 etc.) are low. Performance improves on the last trial of high- versus low-frequency triplets, regardless of whether this high-frequency triplet ends on a Pattern or Random trial (Howard et al. 2004). Pattern trials are always high frequency, whereas one fourth of Random trials are high frequency by chance, such that 16 high frequency triplets occur 62.5% of the time and 46 low frequency triplets (including 4 repetitions and 8 trills) occur 37.5% of the time. High frequency triplets occur 5 times more often than any other triplet type. Sequence learning is demonstrated by increasing differences in reaction time and accuracy between high and low frequency triplet sequences (triplet-type-effect) with continued practice. Learning is typically implicit in this task, in that subjects (without specific instruction to guide them otherwise) do not show any awareness of the alternating regularity or of the varying triplet frequencies (Howard et al., 2004). For all results on motor responding reported in the following sections, only high and low frequency triplets are reported. Trills (i.e. 121) and repetitions (i.e. 111) are removed from analysis because they are not counterbalanced across subjects in that they are always low frequency for all subjects. Performance on trills and repetitions could also reflect pre-existing biases, rather than sequence-specific learning (Remillard and Clark, 2001).

Tests of Explicit Awareness

The following end-of-block sequence report task was only given to Intentional subjects. After each Cued block (i.e. in epochs 1, 2, 3, 6 and 7), subjects were asked to verbally report the four-sequence pattern. The four positions corresponded to numbers 1 through 4, and subjects were required to guess after each Cued block. Awareness was said to have been achieved on the block when subjects reported the correct pattern and continued to do so for the remainder of the blocks.

At the end of the ASRT testing, all groups also received the following tests. First, they were given a generation task in which they were asked to generate a typical 16-unit long pattern that occurred. They did so by numerically writing out a typical 16 unit sequence. Thus, if they were able to correctly insert numerically the 4-digit repeating regularity into every other position, then generation was coded as correct. If they were unable to do so, then it was coded as incorrect. Each participant was asked to separately generate a typical 16-unit long pattern that occurred during the Cued epochs and a typical 16-unit long pattern that occurred during the Uncued epochs.

Next, subjects in both instruction groups were given questionnaire interviews which involved increasingly specific questions regarding their knowledge of any regularity in the sequences they had encountered. If they revealed any awareness of the alternating pattern or of triplet frequencies, they were coded as having explicit awareness. They were separately asked questions for Cued and Uncued blocks regarding their knowledge of any regularity in the sequences they had encountered.

Lastly, they were given a card sorting task. For this task, participants were given a stack of 64 index cards, each of which contained three rows of four circles. One circle on each row was filled in with black. Each row represented one trial hence each card denoted three consecutive trials on the ASRT task (a triplet). Participants were told to sort each card into one of two piles labeled “Occurred More Often,” and “Occurred Less Often” based on how often they thought these triplets of trials occurred (Japikse et al., 2001). Participants were specifically asked to sort for Uncued blocks.

Results

Awareness Analysis

Incidental Groups

No subject in either the Incidental Observe or Respond condition showed or reported any explicit awareness of the correct pattern as measured by generation, questionnaire, or card sort. Results for card sort are summarized in Table 1.

Table 1.

End-of Session Awareness Test for Experiment 1

Group	Card Sort (% accurately sorted)
Observe-Incidental	48.6 +/− 5.2, t=−0.98, p=0.35
Respond-Incidental	51.6 +/− 5.1, t=1.17, p=0.26
Observe-Intentional	47.7 +/− 5.4, t=−1.62, p=0.13
Respond-Intentional	51.3 +/− 5.3, t=−0.71, p=0.50

Open in a new tab

Intentional Groups

All Intentional subjects were able to gain explicit awareness of the correct pattern, as measured by end-of-block sequence report. Five subjects reported some recognition of the same pattern in the Uncued epochs as measured by generation, questionnaire and/or the card sorting task, and so these subjects were removed from the analysis. For this criteria, a conservative approach was taken and subjects who even reported a suspicion that the pattern was the same on the Uncued epochs as the Cued epochs were removed from analysis. Thus, n=14 for the Intentional-Observe group, and n=9 for the Intentional-Respond group. For the remaining Intentional subjects, no explicit awareness of the pattern was reported for the Uncued epochs as measured by generation, questionnaire and the card sorting task (Table 1). For these remaining subjects, as assessed by end-of-block sequence report, explicit awareness of the pattern was reached on average by block 1.0 +/− 0.0 of the first epoch for the Observe group, and by block 2.8 +/− 1.6 of the first epoch for the Respond group.

In sum, on Cued epochs, Intentional subjects had explicit knowledge of the regularity but the Incidental subjects did not. On Uncued epochs, no subject included in the final analysis was aware of the regularity present in them.

Performance Measures

Means of median reaction times for High and Low frequency triplets for Incidental (Figure 2a) and Intentional (Figure 2b) groups are shown for both Observe and Respond conditions. Figure 3 shows triplet-type-effects (reaction time differences between High and Low frequency triplets) for all four groups across all epochs. We assume that performance measures for Cued epochs (1, 2, 3, 6 and 7) reflect explicit knowledge for Intentional but not Incidental subjects. In contrast, Epochs 4 and 5, the first two epochs of the 2nd session, are Uncued epochs, and in all four groups, triplet-type-effects seen in these epochs measure implicit learning.

Mean of Median Reaction Times for Incidental subjects are plotted in Figure 2a (Incidental Observe and Respond) split by High (closed triangles) and Low (closed circles) frequency triplets and for Intentional subjects in Figure 2b (Intentional Observe and Respond) split by High (open triangles) and Low (open circles) frequency triplets. Observe groups are plotted with dashed lines whereas Respond groups are plotted with solid lines. Analyses of the Uncued epochs revealed that those who responded for the training session showed more sequence learning than those who observed only.

Reaction Time Triplet-type-effects (RT for Low minus RT for High frequency triplets) are plotted for all groups. Incidental subjects are plotted with filled diamonds and Intentional subjects are plotted with unfilled diamonds. Observe groups are plotted with dashed lines whereas Respond groups are plotted with solid lines. The Intentional and Incidental Observe groups cannot be seen separately for epochs 4 and 5 because these measures overlap completely.

Training Session

Since only the Respond groups made motor responses in the Training session, only the Respond-Intentional and Respond-Incidental groups could be analyzed here. These blocks were all cued, and so, this training session would reveal the effects of explicit knowledge on performance measures. For reaction time, a triplet-type (High or Low) by instruction (Intentional versus Incidental) by epoch (1, 2, 3) ANOVA, revealed a significant main effect of triplet-type, F(1,21) = 37.12, MSE = 250.68, p<0.0001, a trend for a triplet-type by epoch interaction, F(2,42) = 3.07, MSE = 94.21, p<0.06 which reflect that high-frequency triplet trials become faster than low-frequency ones with practice, and a triplet-type by Instruction interaction, F(1,21) = 7.91, MSE = 250.68, p<0.02, which showed that Intentional instructions enabled people to use explicit knowledge to increase measures of sequence learning for reaction time in Cued epochs.

For accuracy, a triplet-type (High or Low) by epoch (1 to 3) by instruction (Intentional versus Incidental) ANOVA revealed only a main effect of triplet-type, F(1,22) = 16.19, MSE = 0.002, p<0.0006 where high-frequency triplets were more accurate than low-frequency ones. These findings are consistent with Song et al. (2007) in which it was found that explicit awareness affected triplet type effects for reaction time but not for accuracy. Mean accuracy for the groups was 91.8% (Intentional) and 92.5% (Incidental).

Transfer Session

Our main questions here were concerned with the transfer of perceptual learning which occurred during training and how it was influenced by explicit modes of learning. Therefore, our analyses focused on the Uncued epochs at the start of the transfer session, and these are reported first below.

Uncued epochs (4-5)

A triplet-type (High or Low) by response condition (Observe versus Respond) by instruction (Intentional versus Incidental) ANOVA was conducted on Reaction Time for these Uncued epochs. A main effect of response condition, F(1,47) = 8.52, MSE = 2909.23, p<0.006 indicated that the Respond groups were faster than the Observe groups, and a significant main effect of triplet-type, F(1,47) = 87.76, MSE = 116.68, p<0.0001, indicated that High frequency triplets were faster than Low frequency ones reflecting sequence learning. Importantly, a triplet-type by response condition interaction, F(1,47) = 8.85, MSE = 116.69, p<0.005, indicated that the Respond groups showed a larger triplet type effect than the Observe groups, as can be seen in Figure 3. This indicated that motor responding contributed to sequence learning.

There was no significant triplet-type by instruction interaction, F(1,47) = 0.75, MSE = 116.68, p=0.39, which suggests that explicit knowledge during the training phase had no effect on performance measures of sequence learning in Uncued epochs in the transfer phase. Since explicit awareness was not present in these Uncued epochs, we assume that these blocks reflect underlying implicit-based learning processes even in Intentional learners. This result replicates our previous finding that implicit learning occurred independent of explicit awareness (Song et al., 2007). There was also no triplet-type by instruction by response condition interaction, F(1,47) = 0.59, MSE = 116.68, p=0.45, suggesting that instruction did not alter the relationship between perceptual and motor processes.

Since the Uncued epochs in the transfer session were 10 blocks long, to more directly test whether Observe and Respond groups displayed triplet-type differences from the very start of responding, the first two blocks of the transfer session were analyzed separately (Figure 4). Observe groups showed no significant triplet-type-effect in these initial blocks (Observe Incidental: t(13) = −0.11, p= 0.91, Observe Intentional: t(13) = 0.63, p=0.54) whereas Respond groups showed significant triplet-type-effects (Respond Incidental: t(13) = 4.13, p<0.001, Respond Intentional: t(8) = 4.63, p<0.002). Thus, in this first experiment, no evidence for sequence learning by observation alone was found, in that participants who had only observed during training showed no triplet-type effects when they began to respond. In contrast, sequence learning did occur when motor responses were made during training, consistent with the idea that motor responses support sequence learning (Willingham, 1999).

Those who observed the 1st session showed no significant triplet-type effect in the 1st block of the 2nd session whereas those who responded to the 1st session showed significant triplet-type effects in the 1st block of the 2nd session. There was therefore no evidence for perceptual learning in Experiment 1

Accuracy analyses of Cued epochs 4 and 5 revealed findings similar to those on reaction times reported above. A triplet-type (High or Low) by response condition (Observe versus Respond) by instruction (Intentional versus Incidental) ANOVA on accuracy revealed a significant main effect of triplet-type, F(1,47) = 66.58, MSE = 0.001, p<0.0001 as high-frequency triplets were more accurate than low-frequency ones, and a trend for a triplet-type by response condition interaction, F(1,47) = 3.80, MSE = 0.001, p<0.06 reflected the greater sequence learning in Respond groups as compared to Observe groups. Mean accuracy for the four groups in the Uncued epochs ranged from 91.6% to 92.9%.

Cued epochs (6 and 7)

Although our primary questions concerned the Uncued epochs as analyzed above, we also examined the Cued epochs for two purposes. First, we wanted to see how reintroducing cues and explicit awareness affected performance measures in Intentional learners. Second, we wanted to see whether transitioning back to Cued blocks from Uncued blocks led to any changes in performance for Incidental learners, because these Cued blocks reinstituted the perceptual conditions all groups encountered during the training blocks which were Cued.

For Reaction Time, a triplet-type (High or Low) by instruction (Intentional versus Incidental) by response condition (Observe versus Respond) ANOVA revealed a significant main effect of triplet-type, F(1,47) = 66.01, MSE = 848.86, p<0.0001, and a triplet-type by Instruction interaction, F(1,47) = 16.65, MSE = 848.85, p<0.0001. This finding along with Figure 3 indicates that Intentional subjects were able to recall and reinstate their explicit knowledge to affect their performance for reaction time in Cued epochs.

For accuracy in Cued epochs 6 and 7, a triplet-type (High or Low) by instruction (Intentional versus Incidental) by response condition (Observe or Respond) ANOVA revealed only a main effect of triplet-type, F(1,36) = 41.34, MSE = 0.001, p<0.001, which reflects that high-frequency trials were more accurate than low-frequency ones Mean accuracy for groups ranged between 90.6% and 91.6%.

In addition, we analyzed Reaction Time for Incidental subjects separately to explore whether the stimulus changes which occurred from Uncued to Cued blocks affected sequence learning measures. We compared reaction time triplet-type effects on the last two blocks in Uncued epoch 5 to the first two blocks in Cued epoch 6. We reasoned that if perceptual learning was contextually selective for Cued blocks, then transitioning from Uncued to Cued epochs could have led to a benefit. No such benefit was found in that triplet-type effects from the last two blocks of Uncued epoch 5 did not differ from that of the first two blocks of Cued epoch 6 (Observe-Incidental: t(13) = −1.03, p=0.32, Respond-Incidental: t(13) = −0.50, p=0.62). Further experiment however is necessary to conclusively determine whether transitioning from grey and black targets to all black targets or vice versa can affect performance.

Discussion

In this experiment, no evidence for perceptual sequence learning via observation was found, even when the subjects had full explicit knowledge of the regularity during the training phase. This was demonstrated by the fact that subjects in the Observe groups showed no differences between High and Low frequency triplet types at the beginning of the transfer session. In contrast, Respond subjects showed significant triplet-type effects from the beginning of the transfer session, and showed more sequence learning overall than Observe subjects. This latter finding supports the generally accepted conclusion that motor responding can support sequence learning (Willingham, 1999).

In addition, we found that during transfer, Intentional subjects had greater triplet-type effects than Incidental subjects on Cued epochs where they could use their explicit knowledge, but not on Uncued epochs when explicit knowledge was removed. This is consistent with our previous report that explicit knowledge affects performance, but not implicit motor sequence learning itself (Song et al., 2007).

The fact that no evidence for implicit perceptual sequence learning was reflected in our performance measures even for Intentional learners may add insight to the conclusions of Rüsseler who found that perceptual learning did occur and only when subjects had full explicit knowledge. This may be because explicit knowledge is tied to the stimuli but does not increase perceptual learning (Rüsseler and Rösler, 2000; Rüsseler et al., 2003). When explicit knowledge is removed from performance measures, as it was in the current study, an explicit awareness may have no effect on perceptual processing. It has been suggested that in the presence of explicit knowledge, motor responses are more stimulus bound (Knee et al., 2007). However, other explanations may account for these findings. The Rüsseler studies used letter cues while the current study used visuo-spatial targets.

Experiment 2

The results of Experiment 1 seem to suggest that implicit sequence learning did not occur via mere observation of the sequence, even when subjects had full explicit knowledge during training, and hence attention to the sequence had been assured. However it was also possible that the alternating color cues during the training phase could have disrupted the acquisition of perceptual learning or that the switch from alternating color targets during training to all black targets in transfer could have disrupted the expression of any perceptual learning that occurred during training.

To address these possibilities, in Experiment 2 all targets were black at all times in both training and transfer sessions. As in Experiment 1, subjects were assigned to one of two response groups, Observe or Respond, based on how they were treated during training. In Experiment 2, all subjects were given Intentional instructions during training in that they were told to search for any patterns and to report any patterns found at the end of each block. We did this to ensure attention to the stimuli. However, based on previous studies (Howard and Howard, 2001), we expected that, in the absence of the alternating color cues used in Cued epochs in Experiment 1, no subject would gain explicit knowledge with these vague instructions and this amount of training. Hence, any learning would remain implicit and would occur in the absence of any explicit knowledge of the regularity. The term “implicit” learning has been used to encompass two different concepts. First and most notable is the lack of conscious awareness of learning and a resultant inability to communicate this knowledge. However, some authors also suggest “implicit” learning must also be defined by a lack of intention to learn (Frensch, 1998). In this study, we have separated these two concepts such that “Incidental” or “Intentional” instruction refers to the intention to learn the sequence, while “implicit” or “explicit” refers to the consciousness of the resultant sequence learning.

Subjects were also assigned to one of two transfer groups. Half of the Observe and half of the Respond subjects were asked to stop searching for a pattern in the transfer session (Stop-Search), while the other half were asked to continue to look for and report a pattern in the transfer session (Continue-Search). Our goal was to see if continued search for the pattern affected the expression of perceptual or motor based learning. Engaging in intentional search often interferes with expression of implicit learning although not its acquisition (Jimenez et al., 2006; Song et al., 2007). In this manner, in addition to asking whether implicit sequence learning could occur via observation alone, we could assess whether intentional search had an effect on the expression of such learning.