Experimental results have no built-in validity. In a fundamental sense the experiment is the point of departure for all other methods. It offers the most rigorous solution to the causality problem because the investigator can directly control the three features of the causal proposition: time order between variables, covariance, and the exclusion of rival causal factors. The other four methods—the social survey, participant observation, life histories, and unobtrusive measures—are situations of decreasing control over these factors and are better seen as strategies of analysis. As noted earlier, the true experimental design contains at least one control and one experimental group. Two sets of observations are made, and most typically subjects are randomly assigned to the observational groups. Orne suggests that demand characteristics can never be removed from the experimental setting. Friedman has shown that experimenter-subject interaction falls naturally into three phases: a face-sheet phase, an instructions phase, and the data collection phase.