Foundational research on infant social evaluations (e.g., Hamlin et al., 2007; Hamlin et al., 2011; Hamlin & Wynn, 2011) has been cited over 2,500 times and infant researchers suggest these data show infants have an unlearned preference for prosocial others. However, several failed replications have been published, which might be attributable to the type of research methods used to investigate this question. A single measure of the dependent variable is ubiquitous among these studies; within-subject repeated measures are rarely used. In the current study, we adapted methods used by Hamlin and Wynn (2011) to a video-only format, due to COVID-19 restrictions; we extended their methods by including four puppet shows and four corresponding puppet choices to assess for choice stability within and across participants. Six infants were assessed; all but one infant failed to make all four choices and three sessions had to be terminated early due to fussiness. Among the four infants who made at least two choices, no infant showed a robust preference for the helper puppet, two infants chose a puppet on the same side at least three times, and one infant chose the hinderer on three of four opportunities. Our data suggest that a completely virtual method might not be feasible for assessing infants’ choices between two puppets presented on a screen. Suggestions for addressing the limitations of the current study and directions for future research are described.