Evidence-based decision-making in education increasingly relies on research syntheses. However, when a body of evidence contains seemingly disparate findings, decision-makers without statistical expertise may not reason according to meta-analytic norms. One known pitfall is vote-counting, where one judges evidence by tallying positive and negative effects rather than relying on meta-analytic summaries. Because many education interventions have small effects, negative study-level effects are statistically expected even when the true effect is positive, so vote-counting heuristics may lead decision-makers to undervalue evidence. We present findings from a randomized survey experiment investigating how non-researchers reason about meta-analytic evidence displayed visually, particularly when negative effects are present, and whether displaying only the meta-analytic summary improves reasoning. Quantitative and qualitative findings show that visible negative study-level effects led participants to systematically undervalue interventions, even when overall summary evidence was unchanged. Showing only the summary mitigated this bias, while showing the summary first attenuated the bias for users later shown the study-level effects. Participants also reported greater trust in simpler displays, suggesting that providing more information can sometimes reduce both understanding and trust in research evidence.