Role of working memory in kinaesthetic motor imagery

Abstract

Mental imagery involves seeing with the mind's eye “or feeling without actual stimuli.” Previous studies on execution of finger tapping have indicated that the actions involve spatial attention, working memory, visuomotor control, the encoding of sequence-specific information, and storage of information about the motor sequence. The purpose of this study was to use event-related potentials (ERPs) to investigate the extent to which loading to the working memory (sensori-motor buffer) could modulate the neural processes associated with imagination of finger tapping. The experimental manipulation was complexity and length of the tapping sequence. Complexity was increased from single-digit repetition to ordered to random sequence. Length was increased from short sequence (four taps) to long sequence (eight taps). It was hypothesized that the increase in complexity of the sequence to be imagined would increase the level of generation and manipulation of images in the sensori-motor and visuo-spatial buffers. In contrast, the increase in length of the sequence would put a higher load on the sensori-motor buffer. A total of 30 healthy undergraduate students (15 females and 15 males, aged 18 to 30) were recruited to participate in the study. A right-hand finger tapping task was used during which ERPs were captured in the imagery, execution, and control (not engaging in imagery) conditions. In the execution condition, the subjects tapped at 1 Hz frequency according to a tapping sequence that had been well learnt prior to the experiment. In the imagery condition, the subjects imagined and generated the kinaesthetic images as if they had physically manoeuvred according to the same 1 Hz movement rhythm. In this study, the loading on the subjects' working memory was manipulated by the complexity of the tapping that the subjects were to imagine. The two parameters were (a) three levels of sequence order (single-digit repetition, ordered taps with four digits, and random taps with four digits), and (b) two levels of sequence length (short [four taps] and long [eight taps]). These produced 12 movement patterns that the subjects well learnt prior to the experiment. The results indicate that the ERPs captured from the single-digit repetition condition largely differentiated the imagery processes into five distinct components. These were N100 relating to attention allocation; P200 relating to retrieval from long-term memory and image generation; N250 and N400 relating to image visualization, maintenance, and manipulation in the sensori-motor buffer; and P600 relating to information integration and maintenance. These components have also been revealed in previous imagery studies involving other modalities such as visual, vibrotacti1e, and auditory. Dipole source localization was also conducted using CURRY 5, which managed to generate a few dipole sources that were further identified as being associated with different neural substrates. These neural substrates are the same as those previously reported in other studies, and include frontal areas (the precentral gyrus and the middle frontal gyrus) identified during the process of image generation and manipulation, parietal areas (the precuneus and the cingu1ate gyrus) identified during the process of active attention and manipulation, temporal areas (the middle temporal gyrus) identified during the process of memory retrieval, and subcortical areas (the cerebellum and the thalamus) identified during the process of information processing and integration. Our findings further support the model of mental imagery containing a generic imagery network. Different contrasts between the experimental conditions tested the effects of complexity and length of sequence on modulating the sensori-motor buffer in imagery. There were no significant differences found in the amplitudes of the N100 and P200 across the conditions, suggesting that attention allocation and retrieval from long-term memory might not be sensitive to the imagery of taps of different complexities and lengths. Most of the contrasts among the three levels of sequence complexity in imagery did not reveal significant differences in the ERPs. This suggests that the processes involved in the single-digit repetition, ordered-versus-random sequences did not induce different neural processes, in particular those involving the sensori-motor buffer. In contrast, significant differences in the amplitudes of N250 and N400 at P4 (right parietal) and PO4 (right occipito-parietal) were revealed between the short and long taps of the random sequence. The long random sequence was found to elicit significantly less negative-going N250 and N400 components than the short random sequence (p = 0.005-0.022), suggesting the long sequence had an increase in load of active maintenance of information in the sensori-motor buffer. The results also indicate that the imagery of the longer sequence of finger taps required an increase in demand on updating of contents, comparison operations, and possibly rehearsal. Significant differences were also revealed between the short and long taps of the random sequence in the N250 elicited at CP4 (right centro-parietal region) (p = 0.029). The increase in the number of taps resulted in a less negative-going N250 component, suggesting increases in demands in generating periodic sensori-motor images as well as action planning, updating contents, and decision making. On the basis of the present dipole analysis results, it indicated that the combined N250 and N400 component originated from the parietal (the precuneus and cingulate gyrus) and frontal temporal areas (the parahippocampal gyrus or uncus) in both conditions. The dipole sources suggest that working memory relies on a coordinated interaction between the parietal and temporal regions. The results further inform the modulation effect of demand on the working memory (motor buffer) during the manipulation and maintenance processes of motor imagery. In summary, the results of this study enable further investigation into the temporal processes involved in imagery of finger tapping. The data support the model that motor imagery is composed of memory retrieval and generation of sensori-motor images, and that these images can be maintained and manipulated within the sensori-motor buffer. The findings also inform the role of working memory, which could be modulated by changing the number of taps to be imagined, but not the types of sequence during the manipulation and maintenance processes of motor imagery.