Using a SRT task, we exposed young adult participants to a predictable sequence of hand shifts while they had to respond to alphabetical stimuli, which were largely unpredictable. We expected that following a sufficient amount of practice (180 repetitions of a sequence consisting of twelve elements) participants would improve their performance in terms of reduced response latency and reduced percent error. Overall, our expectations appeared to be confirmed as both outcome parameters demonstrated clear learning effects when comparing final performance in the predictable against the unpredictable series. However, predicting shifts between hands was not universally beneficial but the benefit was largely limited to finger actions that repeated the same hand. When a regular shift between hands occurred, irrespective of whether a homologous finger to the previous one was activated or not, finger responses showed the tendency of a small performance advantage in terms of response latency compared to a random series of hand shifts.
To our knowledge, the present study may be the first to demonstrate hand sequence learning in a SRT task, in which the hands took part of the required actions only indirectly. Learning to benefit from predictable hand sequences when only the fingers are in the focus of action suggests some abstraction from the sequence of individual stimuli and responses. Compared, for example, to predictable stimulus modality (visual vs. auditory) sequences in the context of an otherwise unpredictable manual response sequence, where no clear performance benefits could be detected (Koch et al., 2020), the present hand sequence learning paradigm showed robust learning at the level of the hand sequence.
It should be noted though that the constraints imposed on the stimulus sequence, disallowing immediate response repetitions, introduces some partial predictability of the response in the context of a hand repetition. That is, if a hand repetition could be predicted, it could also be inferred, after sufficient exposure to the sequence, that the next stimulus would be a different stimulus, thus calling for the other finger response on the same hand. Hence, predicting a hand repetition, relative to unpredictable hand sequences in the random sequence block, allows reducing the response alternatives from three (completely random, such as in Block 10, where it could only be predicted that the same response is not possible and thus a hand shift is likely) to a single option, so that this response could be prepared during the response-stimulus interval. In contrast, when being able to predict a hand shift, this would allow reducing the number of possible responses in the next trial only from three to two. It is interesting that this partial response predictability still yielded small performance benefits (but just not significant, p = .06 for the learning effect for same finger/different hand), suggesting that some learning may actually have taken place which might have become significant with more statistical power (e.g., a larger sample size). Hence, while we currently can only speculate about the benefit of predicting a hand shift, we can clearly state that the opportunity to predict not only the hand but also, in the case of a hand repetition, the specific response, gives a substantial performance gain.
The present SRT task represents a paradigm in which the hand sequence can be acquired in an incidental learning situation, that is, without explicit instruction to use the predictable hand sequence for response preparation. It is thus interesting to relate the present learning effects to those performance benefits that can be observed in an explicit response preparation paradigm using the so-called finger-cuing task (e.g., Adam et al., 2003; Adam & Koch, 2014; Miller, 1982; Rosenbaum, 1983). In this task, using a four-choice task a response pre-cue indicates a subset of possible responses in the subsequent trial. In this finger-cueing paradigm, there is usually a clear benefit with hand pre-cues relative to finger cues (see, e.g., Adam & Koch, 2014, for discussion), which enable preparation of homologous fingers of the two hands. Note though that in explicit finger cuing paradigm, the information conveyed by the explicit pre-cues is statistically identical across the pre-cuing conditions. This is different in the present situation, where incidental learning may occur but in which the predictability of the hand sequence represents an “implicit” pre-cue that needs to be learnt before it can be used for response preparation.
In the present incidental learning context, it is notable that participants did not mention the partial predictability very clearly, suggesting that this particular effector sequence learning effect is mostly implicit. Similar observations have also been made in the context of learning other abstract sequences (e.g., Goschke & Bolte, 2012; Koch et al., 2006), but we would like to note that our study was not designed to dissociate implicit learning from explicit learning (see Esser et al., 2022, for a recent review and discussion). Therefore, we might tentatively assume that the present case of effector-sequence learning was not associated with high degree of explicit sequence awareness, leaving it to future studies to examine the more specific contributions of explicit vs. implicit learning processes to the observed learning benefits in performance.
Learning to predict the repetition of a hand or an imminent shift between hands may require an integration of the two contralateral two segments, which needs to be distinguished from strategies or mechanisms involved in bilateral or cross-limb transfer. In the context of motor skill acquisition, for example, bilateral transfer has been related to the application of cognitive strategies that can be generalized across limbs (Malfait & Ostry, 2004; Yadav & Mutha, 2020). Similar assumptions of abstract representations, such as representations in terms of visuo-spatial and motor coordinates, have been proposed to enable transfer between effectors and effector-independent knowledge when learning action sequences (Panzer et al., 2009; Park & Shea, 2002). We believe it unlikely that any generalizable representation of the hand sequence was acquired in the context of the present SRT task. Instead, it may be conceivable that incidental acquisition of the hand sequence is analogous to “model-free” habit learning that may resemble a distinct associative learning mechanism, which determines which actions lead to favourable outcome, e.g. in terms of correctly predicted target stimuli, and which may occur in the context of goal-directed associative learning (Daw et al., 2005; Doll et al., 2012).
In conclusion, our study demonstrates that humans are able to learn a covert hand sequence while engaged in a SRT task, where finger actions are performed in response to largely unpredictable target stimuli. Participants appeared to be more sensitive to hand repetitions, where a succeeding target stimulus and the associated response were theoretically predictable, in contrast to shifts between hands, which nevertheless demonstrated a small benefit compared to random hand sequence.