ИСТИНА

Current understanding of feature binding remains controversial. Studies measuring mismatch negativity (MMN) demonstrate a low level of binding, while behavioral experiments suggest a higher level. We examined the possibility that the two levels of feature binding coexist and can be revealed within one experiment. EEG was recorded while participants were engaged in an auditory two-alternative choice task, which was a combination of the oddball and condensation tasks. Two types of deviant target stimuli were used – complex stimuli, which required feature conjunction to be identified, and simple stimuli, which differed from standard stimuli in a single feature. We measured MMN and P300, as well as behavioral accuracy and response time. Responses to complex stimuli were slower (p<.001) and less accurate (p<.0001) than responses to simple stimuli. For simple stimuli, errors were associated with increased response time (p<.0001), while there was no such effect for complex stimuli (p=.46). MMN was prominent and its amplitude was similar for both simple and complex stimuli (p=.84) – although the stimuli differed from standards in a single feature or two features correspondingly. Errors in response only to complex stimuli were associated with decreased MMN amplitude (p=.03). P300 amplitude was greater for complex stimuli than for simple stimuli (p<0.0001). For simple stimuli, P300 amplitude was reduced before errors (p=.04). Thus, we have replicated within one experiment the major effects reported in both lines of binding research. Our observation that during errors MMN amplitude was reduced only for complex stimuli (with no such effect for simple stimuli) hints that the neuronal population encoding feature conjunction is closely associated with (or coincides with) the neuronal population that generates MMN. Our P300 data are compatible with the explanation that higher processing levels receive sensory representations of conjoined features as well as of separate features. Thus, the increased informational load created by complex stimuli prolongs processing time – leading to increased response time for the stimuli requiring conjunction. In summary, our data show that binding in auditory modality involves two distinct levels of processing. The early level related to MMN generation seems to be a critical stage. Yet a later analysis is also performed, affecting P300 amplitude and response time. Our findings provide resolution to conflicting views concerning the nature of feature binding and support the notion that feature binding is a distributed multi-level process.