ABSTRACT
This study is a direct replication of gaze-liking effect using the same design, stimuli and procedure. The gaze-liking effect describes the tendency for people to rate objects as more likeable when they have recently seen a person repeatedly gaze toward rather than away from the object. However, as subsequent studies show considerable variability in the size of this effect, we sampled a larger number of participants (N = 98) than the original study (N = 24) to gain a more precise estimate of the gaze-liking effect size. Our results indicate a much smaller standardised effect size (dz = 0.02) than that of the original study (dz = 0.94). Our smaller effect size was not due to general insensitivity to eye-gaze effects because the same sample showed a clear (dz = 1.09) gaze-cuing effect – faster reaction times when eyes looked toward vs away from target objects. We discuss the implications of our findings for future studies wishing to study the gaze-liking effect.
Acknowledgements
We are very thankful to Andrew Bayliss for providing us with the original stimuli and to Bianca Monachesi and Francesca De Luca for help with the experiment.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Studies wishing to plan sample size based on power analyses based for this effect size should note that such analyses will provide an estimate of the number of items needed to achieve a given level of power. For the frequentist t-test reported here with items as the sole random effect, a future study wishing to achieve power = .80 with alpha = 0.05, would require 93 items (93 target objects). To produce an idea of the number of participants it is necessary to model the effect with both items and participants as random effects within a single model. To do this we used a multilevel model with mean item ratings regressed onto condition (eyes-toward, eyes-away) with random effects for both participants and items. The estimated effect size for condition (eyes-toward, eyes-away) for this analysis was smaller (Cohen's dz =0.006) than the estimate from the model with participants as a single random effect.