Ratings
For valence, the linear mixed models yielded a significant main effect of Stimulus, F (5, 425) = 106.31, p < .001,R 2 = .556, no difference for Group,F (2,85) = 0.57, p = .567, R 2 = .013, and a nearly significant Stimulus x Group interaction,F (10, 425) = 1.76, p = . 066, R 2= .040. Simple contrasts demonstrated that compared to CS-, participants found the CS+, b(CS-, CS+) = 3.33, SE = 0.18,t (435.00) = 18.38, p < .001, GS1,b(CS-, GS1) = 2.31, SE = 0.18, t (435.00) = 12.73, p < .001, GS2, b(CS-, GS2)= 1.31, SE = 0.18, t (435.00) = 7.21, p < .001, and GS3, b(CS-, GS3) = 0.79, SE = 0.18,t (435.00) = 4.39, p < .001, significantly more unpleasant than the CS-, but not GS4, b(CS-, GS4)= 0.09, SE = 0.18, t (435.00) = 0.50, p = .616. Trend analysis for valence ratings revealed significant Linear, F (1, 438) = 494.11, p < . 001 and Quadratic trends,F (1, 438) = 24.79, p < . 001. Exploratorily, we found significant Linear trend x Group interaction, F (2,434) = 7.19, p < .001 but no Quadratic trend x Group interaction, F (2,434) = 0.36, p = .695. All groups showed significant Linear trends, HU: F (1, 149) = 213.98, p< . 001, LU: F (1, 129) = 169.72, p < . 001, MU: F (1, 159) = 110.19, p < . 001, but HU and MU had the biggest difference, and this might have caused the significant interaction. Figure 5 shows that the generalization responses for MU and HU look almost identical with lower responses for CS+, and higher responses for GS4 and CS- in MU.
Similarly, for arousal there was a significant main effect for Stimulus,F (5, 425) = 68.77, p < .001,R 2 = .447, but no main effect of Group,F (2,85) = 0.10, p = .902, R 2 = .002, and a nearly significant Stimulus x Group interaction,F (10, 425) = 1.84, p = .052, R 2 = .042. Follow-up simple contrasts showed that compared to CS-, CS+,b(CS-, CS+) = 3.12, SE = 0.22, t (435.00) = 14.49, p < .001, GS1, b(CS-, GS1)= 2.43, SE = 0.22, t (435.00) = 11.28, p < .001, GS2, b(CS-, GS2) = 1.60, SE = 0.22,t (435.00) = 7.43, p < .001, and GS3,b(CS-, GS3) = 0.81, SE = 0.22, t (435.00) = 3.74, p < .001 were found significantly more arousing, but there was no difference with GS4, b(CS-, GS4)= 0.25, SE = 0.22, t (435.00) = 1.16,, p = .247. Trend analysis for arousal revealed significant Linear, F (1, 438) = 325.10, p < .001 and Quadratic trends, F (1, 438) = 5.62, p = .018. Again, we exploratorily followed-up the nearly significant Stimulus x Group interaction and we found a significant interaction with the Linear trend x Group, F (2, 434) = 4.60,p = .011 with all groups showing a significant Linear trend, LU:F (1,129) = 160.80, p < .001, HU: F (1,149) = 123.74, p < .001, MU: F (1,159) = 67.41,p < .001. Similar to valence, CS+ responses in MU were lower than in HU and LU and slightly higher for GS4 which might have caused the nearly significant interaction. However, the Quadratic trend x Group interaction was not significant, F (2, 434) = 2.18,p = .115.
Lastly, for US-expectancy there was a significant main effect of Stimulus, F (5, 425) = 62.16, p < .001,R 2 = .422, but neither the main effect of Group, F (2,85) = 0.98, p = .379,R 2 = .023 nor the Stimulus x Group interaction reached significance, F (10, 425) = 0.91, p = .522,R 2 = .021. Following the main effect of stimulus, simple contrasts with CS- as reference, revealed that participants expected significantly more threat after CS+:b(CS-, CS+) = 37.09, SE = 2.39, t (435.00) = 15.46, p < .001, GS1: b(CS-, GS1) = 20.52, SE = 2.39, t (435.00) = 8.56, p < .001, GS2: b(CS-, GS2) = 9.61, SE = 2.39,t (435.00) = 4.01, p < .001, GS3:b(CS-, GS3) = 11.42, SE = 2.39, t (435.00) = 4.76, p < .001, but not GS4:b(CS-, GS4) = 4.14, SE = 2.49, t (435.00) = 1.72, p = .085. Trend analysis for US-expectancy yielded both a Linear, F (1,434) = 261.96, p < .001 and Quadratic trend, F (1,434) = 23.66, p < .001. US-expectancy ratings showed a steep, linear decrease from CS+ to GS2 and a less steep decrease from GS2 to CS-.