The program of work reported within the thesis constitutes the first quantitative analysis of the observational modelling (OM) literature. It was motivated by the major limitations that restrict qualitative reviews. These have been common in the skill acquisition literature (McCullagh et af., 1989; Williams et af., 1999; Williams, 1993). Qualitative reviews typically involve only a limited sample of independent investigations, and selection and subsequent interpretation processes are highly susceptible to various types of biases (Copper and Hedges, 1994). An extensive search of the literature was conducted including (n = 293) sources associated with all types of behaviour modification involving OM. Because the thesis was aimed at understanding the effect of constraints on movement skill acquisition, only modelling effects associated with movement behaviour modification were considered (n = 65). In chapter 1, a qualitative review of the movement based OM literature was included. This revealed that experiments on behaviour modification associated with OM have used various experimental designs (e.g., between and within-groups), and typically movement effect (ME) and/or movement dynamics (MD) outcomes as dependent measures. These qualitative findings provided the rationale for the meta-analyses that followed. In chapter 2, current meta-analytic procedures were reviewed to clarify the protocols required to synthesize overall mean effects of OM treatments from diverse designs. Effect size estimates derived from (n = 69) primary investigations were used within two major meta-analytic summaries. The first review (chap.3) clarified the overall mean treatment effect of OM for ME (0 = 0.27) and MD (0 = 0.77) measures over and above that gained through practice only / discovery learning conditions. Both treatment effects represent significant (p<0.01) modelling benefits over control conditions, with additional benefits clearly evident for MD outcomes. These results are consistent with the Visual Perception Perspective (Scully and Newell, 1985) for OM, and suggest that, primarily, demonstrations convey the MD (i.e. relative motions) required to approximate modelled movement skills. Although, ME (Le., performance outcomes) can benefit, modelling treatment effects are typically more modest, suggesting an increased role of 'practice' in skill acquisition. To quantify task constraint influences during OM a new task classification measure was developed (chapA). The classification used two difficulty components, novelty and complexity, which were defined using 3 and 7 descriptive variables respectively. The inter-rater reliability and test-retest objectivity of each descriptive variable rating produced an intraclass correlation coefficients ranging from r = 0.81 to 1.00 and r = 0.87 to 1.00 respectively. The second review (chap.5) reported the mean treatment effects for high and low movement novelty and complexity for MD and ME modelling outcomes. MD results indicated a marked difference in overall treatment effects gained for high (0 = 1.02) and low (0 = 0.57) novelty. Similar, yet more modest novelty effects were obtained for ME outcomes (Ohigh = 0.42 and Olow = 0.11). These results were in direct contrast to previous predictions and conclusions (Gould, 1978). Results suggest facilitative MD modeling outcomes occurred with increased task novelty. The complexity analyses highlighted no discernable difference in MD treatment effects for high (0 = 0.72) and low (8 = 0.74) movement complexity. ME measures were generally more trivial, but also showed little difference resulting from high (0 = 0.07) or low (0 = 0.12) task complexity. Comparable estimates were obtained for an overall difficulty analysis which combined novelty and complexity components. These results indicated that whilst complexity might be expected to influence OM outcomes, further analysis and refinement of the current complexity classification may be warranted within future research efforts.