Page 47 - Emotions through the eyes of our closest living relatives- Exploring attentional and behavioral mechanisms

Emotional attention is modulated by familiarity
evidence for the null-hypothesis (Jeffreys, 1961). Thus, given our data, we found no evidence that bonobos have an attentional bias towards human facial expressions of emotion.
To verify whether our null-findings could be due to any inherent qualities of the 2 used stimuli, we conducted a control experiment with a new group of zoo visitors
and using the same stimuli as the ones presented to the bonobos, here. Crucially,
the visitors had no prior experience with the individuals on the stimuli. If we find
an attentional bias towards emotions using this stimulus sample, then we can at
least say that the stimuli are salient enough to elicit an attentional bias in humans.
In total, we tested N = 150 zoo visitors (75 men and 75 women, 18-88 years old, Mage
= 39.79, SD = 14.98). Note that for clarity, we stick to using ‘familiar’ and ‘unfamiliar’
to denote the two stimulus sets (Caretaker vs. NimStim), but keep in mind that for
the participants, none of the stimuli were familiar. We created two versions of the
task each containing 72 trials with 36 trials of ‘familiar’ humans and 36 ‘unfamiliar’
humans. The only difference between the two versions was that the probe location
was mirrored (i.e., if in version 1 it appeared behind one of the emotional pictures, it
appeared behind the neutral stimulus in version 2, and vice versa). Per participant,
every stimulus was only shown once. To filter our data, we first divided every
participant into an age category as reaction times can be higher in older versus
younger individuals (i.e., 18-20, 21-25, 26-30, etc.). Next, we filtered out extreme RTs
(RT < 250 ms and RT >5000 ms) and then calculated the median absolute deviation
for reaction times per age category. Finally, we used the following data filter: [RT <
(Median RT + (2.5 * Mean Absolute Deviation))]. 606 Trials (5.61%) were subsequently
removed for further analysis.
We performed a GLMM with Congruency, Familiarity, and their interaction, with random intercepts per ID, and using a gamma distribution with a log-link function (as AICnormal = 111557, AICgamma = 111624). We found the expected main effect of Congruency (c2(1) = 4.00, p = .046) and, importantly, neither an effect of Familiarity (c2(1) = .01, p = .931) nor an interaction between the two (c2(1) = .03, p = .866). As such, participants had a faster reaction time to a probe replacing an emotional stimulus (M = 434.87, SD = 85.41) than to a probe replacing a neutral stimulus (M = 437.72, SD = 84.87, b = -0.007 (log scale), t = -2.00, p = .046. See Table S8 for further model output), irrespective of the stimulus set. This is important because it shows that the expressions of the caretakers attracted as much attention as the ones from the NimStim set, thus, the null-result in bonobos is unlikely to be attributable to any qualitative characteristics of the stimuli used, at least to the human eye.
45