Affective prosody recognition is an important area of research in autism spectrum conditions where difficulties in social cognition have been frequently observed. To probe into the mixed results reported in the literature, we conducted a systematic review with meta-analysis and examined potential factors that could explain the inconsistent results. Our literature search included six electronic databases for studies that compared the affective prosody recognition performance in individuals with autism spectrum condition with typically developing participants, which yielded 23 papers eligible for quantitative synthesis. Using a random-effects model, we obtained a moderate-to-large pooled effect (Hedges' g = -0.63) for the overall affective prosody recognition performance of autism spectrum condition participants, which, however, reduced substantially (to -0.26) and became non-significant after the correction for publication bias. The number of answer codes was found to be a significant moderator for the effect estimate, whereas the number of speakers was not. Moreover, the magnitude of the pooled effect estimate varied across emotions. The findings suggested moderate differences in affective prosody recognition ability between autism spectrum condition and typically developing individuals, which reduced to marginal difficulties for autism spectrum condition when the impact of publication bias was taken into account. Diversity in the number of answer codes could have differential effects on affective prosody recognition performance in autism spectrum condition, which varied across emotions. The present review and meta-analysis demonstrated the insufficiency of research on affective prosody recognition in autism spectrum condition, highlighting a need for further exploration of the contributors and underlying mechanisms for specific affective prosody recognition difficulties.