Results
Emotion Ratings. The video set was assembled to evoke both
pleasant and unpleasant emotion states across a range of intensities.
Consequently, a repeated measures ANOVA found a strong difference
between SAM valence ratings between the 5 video categories (F
( 4,164) = 135.60, p< .001;ηG2 = .690). As shown in Figure
1, the pleasant videos elicited the largest valence ratings, followed by
highly arousing pleasant, neutral, unpleasant, and highly arousing
unpleasant videos, with all differences significant. This pattern of
valence ratings is quite consistent with ratings of comparable
categories of scenes (Farkas & Sabatinelli, 2023; Frank & Sabatinelli,
2019). Emotional arousal was also strongly modulated by video content
(F ( 4,164) = 46.64, p < .001;ηG2 = .285). Arousal ratings
were equivalently enhanced by highly arousing pleasant, highly arousing
unpleasant, and unpleasant videos. These were followed by arousal
ratings for pleasant videos, while neutral videos were rated as least
arousing. Table 1 lists the emotion rating means for each video, in
addition to luminance, sound intensity, entropy, entropy standard
deviation, and pixel motion.
Steady-State visual evoked potential. Steady state visual evoked
potential amplitude did not differ between the first and second
presentation of the videos (see Table 2) and were averaged together.
Shown in Figure 2, raw steady-state evoked amplitude was maximal over
occipital regions and significantly modulated by video content from 1-9
seconds after video onset (F ( 4,164) = 14.21, p< .001; ηG2 = .013),
with decreases in 7.5 Hz power evident during emotional, compared to
neutral videos. Standardizing ssVEP amplitude across participants
yielded the same pattern of modulation, with considerably greater effect
size (F ( 4,164) = 16.17, p < .001;ηG2 = .283). Shown in Figure 3,
video categories sharing the same letter in the legend (a, b, or c) do
not significantly differ. Thus, highly arousing unpleasant, unpleasant,
and highly arousing pleasant videos reduced ssVEP amplitude
equivalently, followed by ssVEP amplitude during pleasant videos, which
differed only from ssVEP amplitude during high unpleasant and unpleasant
videos. Lastly, neutral videos evoked the smallest reduction in ssVEP
amplitude, different from ssVEP amplitude during all other video
categories but pleasant, which was marginal (t = 2.72, p =
.056). Also shown in Figure 3 is a histogram of individual participants’
ssVEP amplitude across the 5 categories, representing the consistency of
ssVEP modulation across video content across the sample.
Pearson’s correlations were used to assess the relationship between raw
ssVEP amplitude and emotional and perceptual video features. Thus the
averaged ssVEP amplitude for the subset of participants with data for
each of the 45 videos (average 34.2, SD = 4.6) was compared with the
emotional ratings and quantitative measures from all 42 participants.
This analysis could reveal relationships between ssVEP amplitude and
video features that may be confounded with 5 video categories. For
example, if videos with high luminance, sound intensity, or motion
tended to reduce ssVEP amplitude, and occurred more often in emotional
videos, correlations of these values may emerge if these features were
not sufficiently distributed across the 5 categories. Shown in Table 3,
ssVEP amplitude correlated only with arousal ratings
(R2 = .308, p < .001), while
valence ratings and all perceptual features showed no significant
association. A trend suggests a relationship between entropy SD
(variability in perceptual complexity over time) across the video and
ssVEP amplitude (R2 = .074, p = .07).
Shown in Figure 4, when averaged by video category across the sample of
42, the correlation of arousal ratings and ssVEP amplitude was very high
(R2 = .97, p < .01).