In previous studies using magnetoencephalograms (MEGs) to monitor tactile [1
], auditory [2
], visual [3
] and pain [5
] systems, we found very similar mechanisms of sensory processing among these sensory modalities. In brief, several 'early' activities appear serially with a time delay of about 4 ms at each step followed by one or two 'late' activities. In general, the 'early' activity reverses polarity twice with an interval of 10 ms, which results in a characteristic triphasic waveform, while the 'late' activity is long-lasting without a polarity reversal at such a short interval [7
]. For example, following tactile stimulation, 'early' activations are elicited in area 3b, area 1 and the posterior parietal cortex in this order with a delay of 3–4 ms between each step, and then a long-lasting 'late' activity is evoked in the secondary somatosensory area. We postulate that a basic role of the 'early' activity is to receive inputs from the thalamus or convergent inputs from the thalamus and/or adjacent cortical areas and to send this information to the next point quickly, while the long-lasting 'late' activity is involved in recognition of the stimuli [2
In the present study, we sought to compare mechanisms of sensory processing at latencies later than the 'late' activity among these sensory modalities (vision, audition, touch and pain). At first, we expected there to be unimodal and multimodal activities. Although there is a large and growing number of studies on multimodal interaction using electroencephalography (EEG) and MEG [8
] as well as multimodal activation and interaction using functional magnetic resonance imaging (fMRI) [14
], it is unclear whether or not the timing of the transition from unimodal to multimodal cortical activations is different among modalities. The present study manipulated the interstimulus interval (ISI) at three levels to examine the transition from unimodal and multimodal activations. In general, evoked potentials around and after 100 ms, like N1 and P3a/P3b, are more sensitively increased by increasing the ISI than earlier responses. If the difference in response amplitude between different ISIs shows the same scalp distribution among modalities, the difference in amplitude should originate from the same generator. In addition, a source analysis is helpful to estimate the location of the generator. In general, the manipulation of the ISI more strongly affects P3a and P3b, and the non-specific activities of N1, which are considered indices of orienting attention [18
], compared with activities ("late" activity) within 100 ms. Because the original N1 response in any modality largely includes modality-specific activities showing a different scalp distribution depending on the modality, a comparison between the scalp distributions of the original N1 might not provide a clear-cut result. Therefore, we chose a simple manipulation of the ISI to extract more effectively the non-specific activities, to obtain a clearer result and to enable us to estimate more reliably and simply the location of the activity. Another reason for this choice is that the activities obtained by manipulating the ISI may be associated with orienting attention and later processes reflected by the non-specific N1 and P3a/P3b.
We expected the non-specific, possibly multimodal, activities obtained by manipulating the ISI are clearly found at later than 100 ms and have the same scalp distribution among modalities, but the "late" activities within 100 ms to have a different scalp distribution among modalities. The multimodal activations were expected to be in the anterior cingulate gyrus or hippocampus on the basis of a large number of previous studies performing source analyses [5
] and intracranial recordings [26
], whereas unimodal activities ("late" activities) are estimated to be in areas specific to each modality. Of final and special interest was whether or not the timing of the transition from unimodal to multimodal activations is the same among modalities.