The goal of this study is to provide evidence that implicit and explicit learning of event sequences involve different processing systems. To objectify this claim, we recorded event-related brain potentials (ERPs) while subjects performed a variant of the serial reaction time (SRT) task. In this task, subjects are confronted with a sequence of stimuli (x’s or *) appearing at different locations on a computer screen and are instructed to press a corresponding key for each location as fast and as accurate as possible. Unknown to subjects, the stimuli appear in a regular repeating sequence of positions (e.g. 4-2-3-1-3-2-4-3-2-1, with 1 corresponding to the leftmost and 4 to the rightmost position of a horizontally aligned display; Nissen & Bullemer, 1987). After several structured ‘training’ blocks subjects are transferred to an unstructured stimulus sequence. Typically, subjects show a prolongation of reaction time (RT) in the unstructured compared to the preceding structured block which is taken as evidence that the sequential stimulus structure was learned.

Sequence learning was found without the concurrent development of conscious awareness for the sequential structure of the stimulus material. However, it is still an open question which types of representations are formed during implicit learning. The available evidence is contradictory, indicating either learning of response–response (R–R), stimulus–stimulus (S–S) or stimulus–response (S–R) associations.

Willingham et al. (1989) suggested that associations between stimuli and responses (S–R learning) are of primary importance for the acquisition of sequence knowledge. In their study, subjects responded to the color of stimuli appearing at different locations. Subjects failed to show an RT-advantage for structured blocks, if the task-relevant sequence of colors and responses was unpredictable although the stimulus-locations followed a predictable sequence. In contrast, if the sequence of colors and the related responses were predictable but the stimuli appeared at randomly determined locations performance improved. However, if subjects were instructed to respond to the location of uncolored stimuli which followed the same sequence as before no transfer was found. Thus, the authors concluded that stimulus structures are learned only if they are relevant for subsequent behavior and if they can be mapped directly onto motor responses.

Howard, Mutter and Howard (1992) found that subjects who simply observed sequentially structured stimuli learned as much as subjects who responded to the stimuli with key-presses throughout the learning phase. In a more recent study, Mayr (1996) found that spatial sequences could be learned independently of response sequences. These findings indicate learning of stimulus–stimulus associations.

Finally, Nattkemper and Prinz (1997) obtained evidence in favor of a motor learning perspective (R–R learning). In their experiments, pairs of stimuli were always assigned to one response. Unexpected manipulations of the stimulus sequence which did not interrupt the response sequence were not accompanied by an RT increase whereas violations of both, stimulus and response sequences, resulted in a prolonged response latency. Converging evidence that implicit learning in the SRT task is a type of motor learning was also provided by recent positron emission tomography (PET) studies. These studies suggest that implicit and explicit learning involve different neural systems. During implicit sequence acquisition an increase in regional cerebral blood flow (rCBF) was found in motor areas, i.e. in the sensorimotor cortex, the supplementary motor cortex and in the basal ganglia. In contrast, during explicit learning, enhanced activity was found in non-motor areas as the right dorsolateral prefrontal cortex, the right premotor cortex, the right ventral putamen, and the biparietal–occipital cortex.

In the present study, we used ERPs to explore whether functionally different processes contribute to explicit and implicit learning of event sequences. In particular, we want to provide converging evidence to the PET findings that motor processes are of primary importance for implicit sequence learning whereas motor as well as perceptual processes are relevant for explicit sequence acquisition. ERPs seem to be especially suited to substantiate this perspective, because different ERP components are known to reflect either perceptual and stimulus-evaluation processes or response preparation processes, respectively. Moreover, ERPs reflect a completely different type of signal than PET. ERPs are evoked by electrical rather than blood flow changes and they are coupled much more directly to the processing of single events, because they can be measured during the short epoch which extends between stimulus presentation and response execution.

Task relevant stimuli of low probability in an otherwise regular sequence of events elicit an enhanced negativity with onset of about 200 ms post-stimulus (N200 component). This omponent is followed by an enhanced positivity with onset latency of about 350 ms.

Both components seem to reflect stimulus-evaluation processes. The amplitude of the N200 component was found to be inversely related to the probability of either attended or unattended infrequent stimulus changes. P300 amplitude was found to be sensitive to both the subjective stimulus probability and to the task relevance of the presented material.

The lateralized readiness potential (LRP) is an indicator of response selection and response activation. It is computed from the readiness potential (RP), a slow negative-going potential that starts some time before a movement and rises gradually to its maximum just before movement onset. The RPs preceding voluntary finger and hand movements are larger contralateral to the performing hand. To exclude asymmetries which are unrelated to the motor response De Jong, Wierda, Mulder and Mulder (1988) and Gratton, Coles, Sirevaag, Eriksen and Donchin (1988) suggested to average the RP asymmetries obtained for left- and right-hand movements. To this end, the activity of contra- and ipsilateral electrodes is first subtracted point by point. The two resulting difference waves are averaged to obtain the LRP, which reflects the net asymmetry of the RP preceding lateralized hand or finger movements.

Several findings qualify the LRP as a specific index of response preparation. First, part of the LRP seems to be generated in the precentral motor cortex contralateral to the activated muscle group (see Sommer, Leuthold & Ulrich 1994). Second, numerous studies demonstrated a systematic relationship between the LRP and response selection. For example, Gratton et al. (1988) asked subjects to respond as fast as possible to one of the two imperative stimuli primed by a warning tone either with their left or their right hand. They found a relationship between correctness of the responding hand and the polarity of the prestimulus LRP for fast responses. Correct responses were preceded by a negative-going LRP while incorrect responses were preceded by a transient positive LRP amplitude (positive ‘dip’).