Lasting first impressions: a conservative bias in automatic filters of the acoustic environment

Even in the unattended auditory environment, what we learn first appears resistant to re-evaluation based on experience. Mismatch negativity (MMN) is an auditory event-related potential elicited to rare deviation from automatically generated predictions about the sound environment. MMN amplitude is thought to reflect the potential importance of a sound for further processing. This study was designed to explore the degree to which past experience with a sound can alter automatic attributions about that sound's importance. MMN was elicited to rare (p=.125) physical "deviants" amongst a sequence of highly probable (p=.875) "standard" sounds. Sound identity alternated across blocks within the sequence (i.e., the former deviant became the new standard and the former standard the new deviant). The time period over which a standard remained the more probable tone was varied over Fast (0.8 min), Medium (1.6 min) and Slow (2.4 min) change conditions. Given that local within-block probabilities remained constant across conditions, any change in MMN size was considered a reflection of more rostral brain regions enabling a longer time scale (across-block) representation of event-probability extraction. Larger MMNs were expected to deviations in blocks with longer standard-stability. Although a significant increase in MMN amplitude was observed with increased rule stability, MMN amplitude was heavily dependent on the initial sequence structure. A "primacy bias" was observed such that prolonged stability produced large increases in the MMN to deviations from the first established standard but substantially smaller MMN to this first standard as a later deviant. The primacy effect in these data implies that the automatic filtering of sound relevance is biased toward a confirmation of initial expectations. Initial experience therefore altered the perceived salience of subsequent events.