1. Scientific theories seldom entail precise predictions by themselves; background assumptions are usually required. If a prediction turns out to be false, it might make more sense to reject an assumption than to reject the theory.
Example: The discovery of Neptune (cf. Putnam 1974)
2. Observations seldom match predictions exactly. When they don't match exactly, then (assuming that a given observation is correct) the prediction is, strictly speaking, false. Falsificationism therefore wrongly implies that scientists should reject theories (or background assumptions - see #1) more often than they actually do.
Potential response to this argument: OK, let's modify falsificationism so that we only reject a theory (or assumption) if a prediction fails to come "close enough" to an observation.
Objection to this response: But there is no absolute standard by which to judge whether a prediction is close enough. There is often, however, a relative standard: does a prediction made with the help of one theory come closer to an observation than a prediction made using another theory does? This suggests the following thesisÉ
It is not a matter of each theory standing or falling alone in the face of the evidence. Instead, one theory fits the evidence better or worse than its competitors. This is a different way of looking at the relationship between theories and evidence than the views offered by the logical positivists, the logical empiricists, or Popper.
The "classical" paradigm of statistical hypothesis testing is not compatible with it (Mikkelson 1997).
Regarding the question of which cards shown in Figure 3.1 are useful for testing the hypothesis described in the text, Godfrey-Smith (2003) states flatly that "The right answer is A and D." He says that in particular, "[u]nmasking Card C is evidentially useless." But he fails to note that this judgment about the relevance or irrelevance of certain cards depends on certain background assumptions. In particular, we must assume the following:
Necessary background assumption: It is possible that some card has only one circle, and it is possible that some card has either zero or two circles.
If this assumption is false, then the alternative assumption that I described in class is true, in which case Card C does turn out to be evidentially useful. (Think about it!)
Godfrey-Smith, P. 2003. Theory and Reality: An Introduction to the Philosophy of Science. University of Chicago. Chicago, IL.
Mikkelson, G. M. 1997. Other Things Being Equal: Counterfactuals, Natural Laws, and Scientific Models; With Case Studies from Ecology. Ph.D. Dissertation, University of Chicago. Chicago, IL.
Putnam, H. 1974. The "corroboration" of theories. In: Schilpp, P. A., editor. The Philosophy of Karl Popper. Open Court. Lasalle, IL.