Signals (or cues) are added to multimedia learning materials to guide learners' attention to critical elements of the materials. Yet, research on signaling has produced mixed findings on learning outcomes. On the one hand, some studies have reported positive effects of signaling on the performance of learning outcomes (e.g., Jamet in Hum Behav 32:47-53, 10.1016/j.chb.2013.11.013, 2014). On the other hand, some studies have found that signaling did not improve performance on learning-related outcomes (e.g., Mayer and DaPra in J Exp Psychol 18(3):239-252, 10.1037/a002861670, 2012). The present meta-analysis seeks to (a) resolve the mixed findings in signaling research, (b) examine the effects on signaling on learning outcomes, and (c) identify potential moderating variables. Following an exhaustive search for studies meeting specified design criteria, 44 independent effect sizes were extracted from 29 experimental studies involving 2726 participants. Studies were coded on features, such as participants (e.g., grade level), presentation (e.g., pacing), and methodology (e.g., quality of the study). Results indicated that signaling is associated with increased learning outcomes (d = 0.38, p < 0.01), with effect sizes varying from small to large. This overall effect was moderated by study, participant, presentation, and methodological features. For example, beneficial effects on learning outcomes were found when studies were high in quality, reported the reliability of outcomes, use pretest, and control for differences in prior knowledge. The findings have significant implications for educators and instructional designers as well as for multimedia researchers.