Data Treatment and Research Design

Early published reports of “even narrowly focused laboratory studies conducted with small samples were capable of generating reams of detailed data; readers of journal reports were sometimes confronted with tables of data that ran on for pages” (Smith, Best, Cylke, & Stubbs, 2000). Summary data were presented not only in tables but also in graphic form. Graphs were a common form of data summary in turn-of-the-century scientific reports [the forgetting curve of Ebbinghaus (1885) and the learning curve of Thorndike (1898) were two influential examples of graphic representation]. In addition, graphs helped to pave the way for the later development of correlation and regression analyses (Smith et al., 2000). In attempting to assess the degree of relation between physical and mental characteristics to each other, Francis Galton (1822–1911) used scatter plots in which one set of scores was arranged as a function of another set, such as the height and weight measures of a group of individuals. From such graphic plots evolved the regression line, the steepness of which reflected the degree of relation between two variables, and, in the hands of Karl Pearson (1857–1936), developed into the mathematical technique of correlating variables and measuring the degree of their relationship by the coefficient of correlation (Fancher, 1996). The development of these statistical methods became critical to the assessment of individual differences and the use of tests in psychology. Other statistical procedures were employed to assess comparisons between different groups of individuals. Galton’s research, for example, on the efficacy of prayer asked “whether those who pray attain their objects more frequently than those who do not pray, but who live in all other respects under similar conditions” (Galton, 1872, as cited by Dehue, 2000). A control group was employed in educational research to assess the effects of transfer of training (the influence of practice in one task on performance in another), and, despite arguments over whether participants should be assigned to an experimental or control group at random or by matching individuals, the use of control groups in psychological experiments became an integral part of research design (Dehue, 1997). The comparison of control and experimental group performances led to the use of statistical procedures for testing the significance of any differences that might be obtained.