Teixeira, Wedel, and Pieters (2012) measure how joy and surprise keep people watching internet video ads. The dependent behaviors are attention concentration (eye-tracking) and retention versus zapping. The paper is a primary scholarly source for moment-to-moment creative engagement in skippable video.

Joy and surprise both concentrate attention and retain viewers, but through different routes. The level of surprise most strongly concentrates attention across viewers. The velocity (change) of joy most strongly reduces zapping. Emotion measures explain 72% of zapping variation, with changes in joy alone accounting for 58%. Early surprise reduces zapping more strongly than surprise later in the ad.

Sample and method

Controlled experiment with university participants exposed to 28 internet video advertisements (14 emotional, 14 neutral buffers) in a free-zapping web setting. Facial expressions of joy and surprise were scored by automated classifiers at 4 Hz. Eye movements were recorded on a Tobii infrared tracker. The resulting panel is approximately 145,000 frames of simultaneous emotion, attention, and zapping data. A content pretest of 106 online video ads found joy as the most common targeted emotion (38%), then surprise (26%).

Main contrasts

  • Surprise → attention. Surprise level concentrates interperson attention more strongly than joy. Early surprise is especially protective against zapping.
  • Joy → retention. Joy velocity retains viewers more strongly than joy level. Gains from rising joy outweigh losses from falling joy (asymmetric effect).
  • Attention → retention. Higher attention concentration reduces zapping, so emotions also retain viewers indirectly by binding gaze.

Why this is in the Darwin index

This is the peer-reviewed video-feed engagement map closest to Darwin’s hook-retention question. Darwin’s field study (No. 104) measures 24-hour message recall for text-overlay versus audio hooks in silent-start auctions. Teixeira et al. measure second-by-second emotion, gaze, and skip decisions in a lab web setting. The constructs sit next to each other: early attention binding, then later memory. They are not the same measurement.

Source of record: doi:10.1509/jmr.10.0207.