Using EEG To Predict TV Show Success
Science Trends connects scientists and their research with a global audience.
Join our community of 2,800+ science contributors.
In recent years, television entertainment has become a dominant force in the cultural zeitgeist, attracting ever-growing audiences and revenues. Both viewership numbers and the extent of show-specific engagement on social media (such as Twitter) have become standard indicators of TV show popularity and success. But what makes a TV show a success?
Previous research suggests that emotional motivation, memory, and attention serve as important factors contributing to a favorable perception of information, including video advertisements, movie excerpts, or TV shows. As such, these factors are likely to represent success predictors of TV programming – TV shows have to be memorable, attentionally engrossing, and emotionally engaging to prompt viewers to stay with the program or return for more episodes.
However, current measures typically used to predict future show performance or to determine the quality of individual episodes are often not reliable. For instance, subjective self-report metrics, such as surveys and focus groups, are easily biased. Similarly, assessment metrics based on autonomic nervous system measures, including galvanic skin response and heart rate, are typically weakly associated with human behavior. Thus, accurately predicting individual TV episode success or future show performance remains a challenge.
To identify specific cognitive processes that contribute to viewership and Twitter activity for prime-time TV shows, Nielsen Consumer Neuroscience conducted a study where attention processes, emotional motivation, and memory activation in response to select TV show episodes were assessed using electroencephalography (EEG) measures, which reflect direct brain activity and are less affected by reporting biases.
In the current study, EEG recordings from a large cohort of 331 participants were used to identify cognitive processes associated with TV audience retention and show-specific Twitter volume across 9 episodes from diverse prime-time TV shows, including Big Brother, Naked and Afraid, NY Med, and others. TV viewership was assessed through Nielsen ratings data, and real-time Twitter volume related to the specific TV show was estimated through Nielsen Social. Indices of naturalistic TV show consumption (viewership and Twitter engagement) were correlated with indices of Attention, Emotional Motivation, and Memory derived from the EEG signal (see Methods).
EEG measures associated with cognitive processes of Attention, Emotional Motivation, and Memory predicted, with high degree of accuracy, population-level TV Viewership and Twitter Engagement associated with episodic prime-time TV shows (Figure 1). Although there was a significant direct correlation between TV viewership and Twitter volume, this relationship was mediated by EEG measures, with a composite EEG score being the best predictor for either of the behavioral variables.
TV viewership (a passive form of TV consumption) relied primarily on Attentional focus. In contrast, Emotional Motivation (in addition to Attentional focus) was a key factor in predicting Twitter engagement – a behavioral indicator of active TV consumption (Figure 2).
The success of all television show genres may rely on similar cognitive and emotional processes, although we are likely to find different weights for Attention, Memory, or Emotional Motivation for documentaries vs. dramas vs. reality TV. We anticipate that these cognitive measures will be able to indicate positive or negative shifts in engagement (which is often an implicit process that is not accessible to self-report) ahead of water-cooler conversations, blog reviews, and social media discussions.
To explore this hypothesis, we collected data from 4 pilot episodes of serialized shows on Indian TV. Across 18 episode segments, we observed a strong correlation between TV viewership and the combined EEG metric (r = 0.53, p = 0.02). When examining correlations between individual EEG metrics and TV viewership, we saw strong correlations for EEG measures of Emotional Motivation (r = 0.52, p = 0.03) and, to a lesser degree, Memory (r = 0.44, p = 0.07). Together, Emotional Motivation and Memory explained the most variance in TV viewership (r = 0.63, p = 0.01). Stronger contribution of Memory processes to TV viewership in this sample can be explained by the nature of the tested TV episodes. Since these were pilots, audiences had to build the narrative by learning, remembering, and keeping track of new characters, settings, and plot elements.
Together, these findings highlight the viability of using EEG measures to predict the in-market success of TV programming and to identify cognitive processes that contribute to audience engagement with television shows. These EEG measures can provide an effective optimization mechanism for in-production fine-tuning of elements and segments of TV shows to be more engaging. EEG measures can also be used as early warning signals of potential audience disengagement. Used at the segment-level diagnostics, EEG measures can enable granular understanding of characters, their relationships, and story arcs and how these show elements translate into viability and audiences’ social engagement.
For this study, 9 episodes from 8 hour-long prime-time shows were tested. Selected shows varied in genre and had a range of viewership and Twitter volume metrics, which ensured a representative stimulus sample. Minute-by-minute numbers of show-related tweets posted during the original episode airing were obtained from Nielsen Social. Minute-by-minute viewership numbers for the original episode airing were obtained from Nielsen Ratings. EEG data were recorded in-lab from 331 participants, with each watching one uninterrupted episode (without commercial breaks). Each episode was tested within 2-4 days after the original airing among participants who were regular show watchers.
EEG signals for each participant were collected from 32 channels. Power estimations were obtained in the theta (4-8 Hz), alpha (8-12 Hz), beta (13-30 Hz) and gamma (31-55 Hz) frequency ranges. Power data were z-scored within each participant. Minute-by-minute EEG measures were calculated across participants for each episode: Attention (using theta and alpha power), Emotional Motivation (using hemispheric asymmetry in the alpha and beta frequency ranges), Memory (using theta and gamma power), and the Composite EEG score, calculated as an average of Attention, Emotional Motivation, and Memory scores.
Minute-by-minute EEG, Twitter volume, and TV viewership data were converted to difference scores reflecting a change from one minute to the next. Data within each show segment (originally aired in-between commercial breaks) were averaged, yielding 49 data points for each measurement stream. The composite EEG score and individual EEG scores of Attention, Emotional Motivation, and Memory were entered into separate regression models predicting either TV viewership or Twitter volume. Additional information about data analysis and results can be found in the PLOS One publication.