This article describes and validates a human-centric measure of audio message complexity. Messages are coded in terms of the level of cognitive resources that would be automatically elicited and required to process the message. Indicators of automatic resources elicited come from counting the orienting eliciting audio content changes in radio messages (Acc). The indicator of resources required comes from counting the dimensions of audio information introduced (Aii) by these content changes. The combination produces an indicator of available resources that serves as the complexity variable. Messages high in available resources are low in complexity; messages low in available resources are high in complexity. Two experiments are presented exploring the empirical validity of the measures as both local (moment to moment) and global (message level) operationalizations of complexity. Results suggest the measures have high construct validity.