In his book, The Scientific Image, van Fraassen lucidly draws an alternative to scientific realism, which he calls “Constructive Empiricism”. In this epistemological theory, the concept of observability plays the pivotal role: acceptable theories may be believed only where what they say solely concerns observables. Van Fraassen develops a concept of observability which is, as he admits, vague, relative, science-dependent, and anthropocentric. I draw out unacceptable consequences of each of these aspects of his concept. Also, I argue against his assumption that “empirical adequacy” is the same thing as “saving the phenomena”, according to his sense of the expressions.