Public service announcements (PSAs) are non-commercial broadcast ads that are an important part of televised public health campaigns. "Message sensation value" (MSV), a measure of sensory intensity of audio, visual, and content features of an ad, is an important factor in PSA impact. Some communication theories propose that higher message sensation value brings increased attention and cognitive processing, leading to higher ad impact. Others argue that the attention-intensive format could compete with ad's message for cognitive resources and result in reduced processing of PSA content and reduced overall effectiveness. Brain imaging during PSA viewing provides a quantitative surrogate measure of PSA impact and addresses questions of PSA evaluation and design not accessible with traditional subjective and epidemiological methods. We used Blood Oxygenation Level Dependent (BOLD) functional Magnetic Resonance Imaging (fMRI) and recognition memory measures to compare high and low MSV anti-tobacco PSAs and neutral videos. In a short-delay, forced-choice memory test, frames extracted from PSAs were recognized more accurately than frames extracted from the NV. Frames from the low MSV PSAs were better recognized than frames from the high MSV PSAs. The accuracy of recognition of PSA frames was positively correlated with the prefrontal and temporal, and negatively correlated with the occipital cortex activation. The low MSV PSAs were associated with greater prefrontal and temporal activation, than the high MSV PSAs. The high MSV PSAs produced greater activation primarily in the occipital cortex. These findings support the "dual processing" and "limited capacity" theories of communication that postulate a competition between ad's content and format for the viewers' cognitive resources and suggest that the "attention-grabbing" high MSV format could impede the learning and retention of an ad. These findings demonstrate the potential of using neuroimaging in the design and evaluation of mass media public health communications.