Second, and most paradoxically, the exploratory analysis revealed a marginally significant interaction effect between manipulated profile source and past experience on reported attraction, F (1, 201) = 2.94, p = .088, η p 2 = .014 ( Figure 3 ).
A supplementary linear regression confirmed this pattern: although the interaction term did not reach conventional significance, p = .173, a simple slope analysis showed that at high levels of prejudice (mean +1 SD), the difference between the Facebook and Tinder conditions approached significance, p = .054. 2.2.3.