The Validated Touch-Video Database

Abstract

Visually observing a touch quickly reveals who is being touched, how it might feel, and the broader social or emotional context, shaping our interpretation of such interactions. Investigating these dimensions is essential for understanding how tactile experiences are processed individually and how we empathise with observed sensations in others. Here, we expand available resources for studying visually perceived touch by providing a wide-ranging set of dynamic interactions that specifically focus on the sensory qualities of touch. The Validated Touch-Video Database (VTD) consists of a set of 90 videos depicting tactile interactions with a stationary left hand, viewed from a first-person perspective. In each video, a second hand makes contact either directly (e.g., with fingers or an open palm) or using an object (e.g., a soft brush or scissors), with variations across dimensions such as hedonic qualities, arousal, threat, touch type, and the object used. Validation by 350 participants (283 women, 66 men, 1 non-binary) involved categorising the videos as 'neutral', 'pleasant', 'unpleasant', or 'painful' and rating arousal and threat levels. Our findings reveal high inter-subject agreement, with painful touch videos eliciting the highest arousal and threat ratings, while neutral touch videos serve as a baseline. Exploratory analyses indicate that women rated the videos as more threatening and painful than men, suggesting potential gender differences in the visual perception of negatively valenced touch stimuli. The VTD provides a comprehensive resource for researchers investigating the sensory and emotional dimensions of observed touch.

Keywords: Arousal; Dynamic touch stimuli; Gender differences; Sensory qualities; Tactile interactions; Threat perception; Videos; Visually perceived touch.

Fig. 1

Still frames for a subset of videos from the Validated Touch-Video Database. These frames illustrate a variety of touch interactions captured from a first-person perspective, depicting different hedonic qualities including neutral, pleasant, unpleasant, and painful touch. The database includes 90 systematically controlled videos that vary across dimensions such as arousal, threat, and touch type, providing a comprehensive resource for investigating the emotional and sensory dimensions of observed touch

Fig. 2

Arousal and threat ratings by gender. The box plots display the median ratings for (A) arousal and (B) threat across all 90 videos, separated by gender. Participants rated the touch on a scale from 1 (not at all) to 10 (extremely), with a total of 283 women and 66 men

Fig. 3

Hedonic categorisation of touch videos by gender. The bar graphs display the categorisation of all 90 videos into one of four hedonics categories, separated by gender

Fig. 4

Hedonic strength ratings for the top 10 videos in each category. Each bar graph displays the average hedonic strength ratings for the entire sample, accompanied by 95% confidence intervals. Participants rated the touch on a scale from 1 (not at all) to 10 (extremely). The top 10 most distinctly categorised videos are organised from left to right, with pleasant (green), unpleasant (orange), and painful (red)

Fig. 5

Hedonic strength ratings for each category divided by gender. Each bar shows the average hedonic strength ratings for a subset of videos (top 10 per category), differentiated by gender (light bars for men and dark bars for women) with 95% confidence intervals. Videos are categorised as pleasant (green), unpleasant (orange), and painful (red)

Fig. 6

Average ratings for arousal (A) and threat (B) across the top 10 videos in each hedonic category. Participants rated each video on a scale from 1 (not at all) to 10 (extremely). Scatter plots within the violins represent mean scores for individual videos

Fig. 7

Overview of the videos along with their validation data. Preview of the overview table which can be found on GitHub (https://sophiesmit1.github.io/VTD/). The 90 videos in this table can be arranged based on the highest or lowest values for each index, such as hedonic quality (percentage of participants), threat, and arousal (1 = not at all, 10 = extremely)