In contrast to traditional linguistic analysis, a model based on the empirical agent is presented and tested. A text is regarded as an intentionally produced cognitive process. The analysis has to take the agent (perspective) into account to facilitate an adequate processing of its objectives (viewpoints). Moreover, the model is surface-oriented and assumes that the cognitive relevance of an utterance is defined by the dynamics of the text production and not by artificial criteria. The model has been tested on responses to questionnaire items constructed by a multinational industry. The responses were produced by 35 randomly selected workers (mechanics) from England, Italy, Sweden, West Germany, and the United States; seven subjects were chosen from each country. The differences were tested for the power of normal curve test of P(sub 1) equals P(sub 2) via arcsine transformation at a(sub 1) equals 0.05. By agent control of the question-response relationship, differences in coherence could be demonstrated. English and Swedish workers showed a significantly higher coherence with the perspective of the industry, whereas Italian, German, and American workers assume their own agent function in their response behavior. (TJH)