We studied electrophysiological correlates of change blindness in the oddball variant of the flicker paradigm during change blindness, i.e. stimulus presentation during which the subjects did not notice the change, and during which they could not anticipate when the changes would occur. Intriguingly, and compatible with our hypothesis, we found that even during change blindness the occurrences of random changes modulated electrical brain responses at all electrode sites. There was a global positive difference at the latency of less than 100 ms in ERPs to infrequent images containing changes (deviants) compared to frequently presented images without changes (standards), indicating that the changed features are somehow represented in the brain even in the absence of the ability to anticipate or report on the occurrences of the changes. This difference in ERPs was confined to the change blindness situation.
Our results cannot be directly compared to the results of those experiments in which the S1-S2 and traditional continuous flicker of paradigms were applied. This is because search behavior in S1-S2 and traditional continuous flicker paradigms may differ from those in the oddball paradigm. In that sense, changes differ in salience and infrequent changes may require different comparison mechanisms, e.g. searching for a violation in a rule rather than serial comparison of elements, or more sustained attention to specific locations in images. However, the search for implicit representations of changed features during change blindness is not affected, even if search behavior may differ from each other in these different conditions. Any registration of changes still indicates that the changed features are represented at some level in the brain. Infrequent changes may also render such effects visible that would go unnoticed in successive presentation of original and modified pictures. Since visual search mechanisms differ in the manner described above, the oddball paradigm can reveal different aspects of brain processing, such as effects of neural dishabituation. Therefore, introducing the oddball paradigm may be an important methodological addition to the investigation of the change blindness phenomenon.
The early latency of the ERP effect (60-100 ms post-stimulus) suggests that the difference is unlikely to reflect any implicit processing of changes per se . Instead, it may be due to preliminary processing of low-level features of images or effects of dishabituation in response to changed features after repetitive identical stimulation. Nevertheless, any such difference indicates that some neural representation exists for the features in which unnoticed changes occur.
The present results cannot be due to the effects of anticipation or task-preparation, which may be the case in a previous study with comparable results from the S1-S2 paradigm, namely, that by Eimer and Mazza (; see also ). In their study, Eimer and Mazza also observed a wide-spread positive modulation in brain responses at the early latencies of 30-80 ms and 90-130 ms during change blindness in the S1-S2 paradigm with natural and complex stimuli (groups of faces) with large changes. The authors were unable to interpret this finding simply as a genuine stimulus-related modulation, but instead they proposed that it might be an instance of task-preparation related contingent negative variation, normally elicited by differences of expectations that would be present already before stimulus onset. They conceded that their change blindness trials might have included more trials from sequences with worse task preparation as compared to trials in which participants correctly reported the absence of change, as this could have resulted in the kind of differences in ERPs they observed. In the present study, however, such a bias in change blindness trials is not possible. Namely, because of the pseudo-random presentation of the stimulus types and the fact that data were analyzed only for the standards immediately preceding the deviants (because the occurrence of a standard after the deviant could be expected), both stimulus types were from the same sequence and preceded by numerous identical pictures. Thus, there could have been no systematic difference in subjects' state of preparation, as is possible in the S1-S2 paradigm.
The temporal analyses showed that the onsets and offsets of the differences between responses to (modified) deviant and those to (original) standard images were rather similar at the electrode sites of Fz, Cz, and Pz (significant differences in responses observed between 70 and 104 ms). Unlike these electrode sites, the difference did not quite reach statistical significance by the criterion that we used (p < .01) at the electrode site of Oz. Nonetheless, the offset of the effect seems, on the basis of the temporal analysis and the difference waves, more abrupt at Oz. This may indicate that the responses at this electrode site reflect different brain processes from the ones reflected in responses at the more anterior electrode sites.
In studies of conscious change detection using a S1-S2 or "one-shot" flicker paradigm, a difference in ERP amplitudes at latencies between of 60-150 ms from stimulus onset related to detected stimulus changes in comparison to stimuli containing no change has been observed [10, 15, 29–32]. However, the polarity of the difference in ERPs varies across studies. In most of the studies [15, 29–32], images with detected changes elicited more positive ERPs than those without changes or with undetected changes. In the study of Eimer & Mazza , identified changes evoked a negative difference compared to the no-change situation. In the present study, the difference in ERPs to identified changes as compared to no-change images did not reach significance in MANOVA or in the point-by-point temporal analysis, although there was some hint of differential activity at the Oz electrode in the grand average waveforms (Figure 2). The reason why the change related modulation did not reach significance in the present study may be in the differences of psychological states of the subjects. In other studies subjects focused on changes, while in the present study the participants were instructed to ignore the previously identified changes. The results, however, suggest that the modulation of ERPs by unnoticed changes observed in the change blindness trials reflects neural processes that are different from those related to explicit change identification.
One purpose of the study was to make the experimental conditions resemble those of behavioral studies, and therefore participants were allowed to search freely for the change. Moreover, since the main interest and analyses were on the latencies that precede even the most rapid eye-movements evoked by sensory stimulation , we decided not to constrain them. We endeavored to avoid any compromising effects of covert visual spatial attention or inhibition of eye-movements on responses to changes that might result from such restrictions. Constraining eye-movements has been suggested to result in obtaining data on active inhibition of eye-movements rather than responses to visual stimulation , and it has also been shown to affect change detection performance in the flicker paradigm .