Are the neural correlates of conscious contents stable or plastic?

It is often an implicit assumption of consciousness research that the neural correlate of a particular conscious experience is both universal (shared between subjects) and stable over time. Recently, we demonstrated that the first assumption is only partially true, and in the present study we examine the second assumption. We recorded the magnetoencephalographic (MEG) signal from healthy human participants while they viewed an intermittently presented binocular rivalry stimulus consisting of a face and a grating. During binocular rivalry, the stimulus is kept constant, but the conscious content alternates between two possibilities. Using a multivariate classification algorithm, we found that it was possible to predict the conscious experience of a participant from early event-related field components (100-300ms after stimulus presentation) using data gathered on different trials of the same recording session. Very similar accuracies were obtained when the data used to train and test the classifier were gathered on different days within a week. However, when training/testing data were separated by 2.5 years, prediction accuracy was reduced drastically, to a level comparable to when the classifier was trained on a different participant. We discuss whether this drop in accuracy can best be explained by changes in the predictive signal in terms of timing, topography or underlying sources. Our results thus show that the neural correlates of conscious perception of a particular stimulus are stable within a time frame of days, but not across years. This may be taken as an indication that our experience of the same visual stimulus changes slowly across long periods of time, or alternatively the results may be understood in terms of multiple realizability.