SCIENTIFIC DISCOVERY 193 



(b) Cause and effect inferences. Operations involving the 

 discovery of causes and effects are commonly considered to 

 be the sole concern of science. The more detailed analysis 

 of this notion, which is to be the topic of Chapter XVI, 

 will reveal the truth which there is in this statement. Here 

 the problem is only to point out that the techniques for 

 determining causes and effects are not highly determinate. 

 They do not enable one to infer the characters of causes 

 from effects, or vice versa. Instead, they are principles for 

 telling one when and where to look for causes, given effects, 

 and when and where to look for effects, given causes. They 

 thus enable one to identify causes and effects by their 

 temporal and spatial locations. Certain hoary principles, 

 e.g., that cause resembles effect, supposed by many phi- 

 losophers to be descriptive of the cause-effect connection in 

 nature, do not seem to have any important application in 

 science, and there is reason to doubt whether they have 

 any clearly defined meaning. A precise logic of causal 

 connections has not yet been formulated. As a consequence 

 no operational construction of causes from effects, or of 

 effects from causes, is possible; specific causes and effects 

 cannot be anticipated with any high degree of certainty. 

 Causes and effects are therefore hypothetical rather than 

 constructual in character; they are given content through 

 activities which are essentially imaginative, though the oper- 

 ations are often guided by analogies. 



(c) Operations of analysis and synthesis. The demand for 

 operations of this kind is based upon the fact that a whole is 

 better understood when the properties of its parts are known, 

 and the parts are better understood when the properties of 

 the whole are known. The behavior of social groups is more 

 intelligible when the modes of behavior of the individual 

 members are known, and the modes of behavior of the in- 

 dividual members are more intelligible when the nature of 

 social behavior is understood. In the one case the data 

 constitute the complex, and the inferential route is analytic; 

 in the other case the data are the elements, and the inferen- 



