Since the early 1900s, researchers have investigated the relative effectiveness of part versus whole methods, and interest in the topic persists today. This line of inquiry has examined if or when it is best to present material (or practice a skill) as a whole versus in some "non-whole" simplified manner. A number of non-whole variations have been developed, categorized as part (e.g., forward chaining, fractionation) or simplification (adaptive training) (Mané, Adams, & Donchin, 1989). Despite considerable study, identifying the conditions favoring part or whole strategies continues to be evasive, and conclusions to date, relative to motor tasks, are based on limited data (Lintern, 1989). We sought to critically examine the available part-whole data on the acquisition/learning of motor skills through meta-analysis. Using computer searches and reference lists, we identified studies that compared motor skill acquisition/learning under whole and non-whole conditions. Effect size (ES) was calculated from studies that provided means and standard deviations of groups, or the results of two-group comparisons (t-test, ANOVA). Effect sizes were calculated from 13 studies; 14 additional studies did not provide enough information for ES calculation but served to supplement the ES data. Studies were coded relative to type of treatment (chaining, fractionation, simplification), motor task, and participant characteristics. A total of 24 acquisition effect sizes (AES) and 10 learning effect sizes (LES) resulted. AES ranged from .15 to 2.50, mean=.92; LES ranged from -1.44 to .98, mean=-.32. This indicated a clear advantage of non-whole training in acquisition, but much variety in learning. Subgroups of studies were used to examine the relative effectiveness of forms of non-whole training, and with respect to skill types. Two of these subgroups are reported here. One used chaining (forward or backward) with a serial task. Again, there was a consistent advantage of non-whole training in acquisition (ES range=.31 to 1.20), but no clear pattern in learning (ES range=-1.40 to .40). The other set of studies used either fractionation part training or simplification with a non-serial (integrated) task. Here, non-whole training had a clear advantage in acquisition (ES range=.44 to 1.76) and learning (ES range=.16 to .98). Our findings call into question the research base on which part/whole recommendations are made for motor skills, and support the conclusion that the clearest evidence for a part advantage is in complex integrated tasks where fractionation or simplification is used. Keyword(s): performance, research