| 其他摘要 | In the field of category learning, there has been much debate about whether there is a dissociation between implicit and explicit category learning systems. The "dual system" theory suggests that there are two types of category learning systems, i.e.,implicit and explicit category learning systems, which have different cognitive processing characteristics and different neural mechanisms. The "single-system" theory, on the other hand, suggests that there is only a single system, i.e., the explicit category learning system. Most of these arguments are limited to uni-modal learning, whereas cross-modal learning is closer to the real environment of human learning than unimodal learning. Moreover, the human brain does not simply add up information from different modalities but integrates them to form a unified experience and representation.The cognitive characteristics and neural mechanisms regarding cross-modal learning remains unclear. Therefore, this study investigates whether there are two different category learning systems in cross-modal learning by using audiovisual cross-modal stimuli varied on discrete and continuous variables to define categories. The related findings may not only provide new research evidence for the controversy of category learning systems, but also help to reveal the time and neural mechanisms of cross-modal information in the brain, and thus give some inspiration to the application of crossmodal learning in artificial intelligence.
Study 1 used a weather prediction task paradigm to examine whether subjects could acquire categorical knowledge across auditory and visual modalities, and whether cross-modal category knowledge could be implicit. Additionally, Study 1 manipulated the stimulus modality to investigate the characteristics of uni-modal and cross-modal category learning. The results of Experiment 1a suggested that subjects could acquire some unconscious knowledge in the auditory and cross-modal conditions and showed a dominant effect of the visual modality on category learning performance. Experiment 1b extended the training blocks to three times that of Experiment 1a and revealed that there still existed a visual dominance effect and that all subjects acquired conscious category knowledge. Experiments 2a and 2b adopted complex visual stimuli and the combination of complex visual stimuli and auditory feedback, respectively, and still observed the visual dominance effect in category learning. Experiment 3 used auditory stimuli with two dimensions from one object, and observed an increased optimal response rate in the auditory condition and a visual dominance effect. These results suggest that participants can combine information from auditory and visual modalities to acquire category knowledge and that this cross-modal knowledge can be implicit; cross-modal category learning is less well performed than visual learning and there exists a stable visual dominance effect.
Study 2 used behavioral and functional magnetic resonance imaging techniques to investigate characteristics and neural mechanisms of implicit and explicit category learning by setting up rule-based and information-integrated category structures. The fMRI results revealed that the explicit learning group had significantly stronger activation in brain regions such as hippocampus and anterior cingulate gyrus; while the implicit learning group had significantly stronger activation in brain regions such as bilateral inferior parietal lobule, supplementary motor area, and left thalamus. In addition, compared to the baseline task, the categorization task elicited significantly stronger activation in the caudate nucleus and substantia nigra in both conditions. The results suggest that the basal ganglia are not specifically involved in implicit category learning, but rather play a role in both implicit and explicit learning. These results not only provide new supporting evidence for a "dual system", but also suggest that the implicit and explicit category learning systems are not totally distinct. Especially, brain regions such as the caudate nucleus play an important role in both implicit and explicit category learning.
Study 3 used rule-based and information-integrated category structures to examine transfer mechanisms and time courses of implicit and explicit learning through combining behavioral and event-related potential techniques. The results of Experiment 6 showed that subjects were able to show transfer effects to new stimuli within modalities during the transfer phase in both implicit and explicit category learning groups, but the spontaneous transfer effect was significantly larger in the implicit category learning group than in the explicit category learning group. The results of Experiment 7, on the other hand, showed that the implicit category learning group could not transfer cross-modality knowledge to new stimuli across modalities, but the explicit category learning group could learn new categories faster with feedback, indicating transfer effects to new stimuli across modalities. The results suggest that implicit category learning may complete transfer based on similarity, in which the transfer effect to new stimuli within modalities can be observed, while explicit category learning may complete transfer based on abstract rules and hypothesis testing, in which the transfer effect to new stimuli both within and across modalities can be observed. The responselocked ERP results of Experiment 8 revealed a larger FRN (feedback-related negativity) in the implicit category learning group than in the explicit category learning group, and there was a significant correlation between FRN amplitudes and learning performance only in the implicit group. Moreover, the P3 component associated with memory expectancy was larger in the explicit category learning group than in the implicit category learning group and there was a significant correlation between P3 amplitudes and learning performance only in the explicit group. These results suggest that implicit and explicit cross-modal category learning have distinct transfer mechanisms and time courses.
The present study systematically examined the cognitive neural mechanisms of cross-model category learning, providing not only behavioral and neural evidence for the dissociation between implicit and explicit category learning from a cross-modal processing perspective, but also revealing distinct transfer mechanisms and time courses in implicit and explicit cross-modal category learning. The results revealed that there is dissociation between implicit and explicit category learning not only in brain regions such as the hippocampus, the left thalamus, and the inferior parietal lobule, but also in transfer mechanisms and time courses of cross-modal knowledge. These findings enrich and improve the theories related to category learning and are conducive to further investigate the cognitive characteristics and neural mechanisms of crosschannel category learning. |
修改评论