According to Hull, all selection is "a process in which the differential extinction and proliferation of interactors cause the differential perpetuation of the replicators that produced them" (p. 134). The natural is one type of selection, the conceptual another. In conceptual selection in science, the interactors are scientists, and the replicators "the substantive content of science," whose chief vehicles are "books, journals, computers, and of course human brains" (p. 140). When interactors face their environment, there is a test of survival. As a result, some replicators survive; others do not. These differential rates of survival produce two lineages, that of replicators and that of interactors. To describe the evolution of science is to trace the genealogy of these two lineages (p. 150).
In this model of evolutionary epistemology, what does testing against an environment mean? Early in his paper, Hull implies that testing in science is an exceptional procedure: "testing is reserved for those findings which threaten one's own research" (p. 131). Indeed, wholesale testing would be counterproductive: "I cannot conceive of worse advice [than checking everything]. The whole point of scientists working together is to pool conceptual resources" (p. 131). Writing in this vein, Hull refers to "the possibility of empirical testing" (p. 129; see also 154), the implication being that this possibility alone is the hallmark of the epistemological superiority of the sciences. But Hull's model seems to require that every replicator be tested against an environment. If there is no testing, there can be no differential survival: "Replicators are generated, recombined, and tested by scientists interacting with the relevant portion of the natural world," a "non-conceptual world" (p. 141). This match between concepts and the non-conceptual world is essential: "testing is essential to science"; scientists "spend an appreciable amount of time testing their views" (p. 140; p. 148).
The solution to this seeming inconsistency is in part semantic. Two senses of test must be distinguished: the trial and the repetition. Science proceeds by means of empirical trials of theory; at some point, even theoretical advance must answer to empirical trial. A repetition is a trial of a trial: were the original results correct; if not, was their source error or fraud? Hull uses "test" in two senses: when he alludes to its frequency, he is referring to trials; when he alludes to its infrequency, he is referring to repetitions.
But what is the environment against which replicators are tested? This question also yields to disambiguation. Crucial to each sense of test is the need to interpret. When the scientist conducts a trial, her criterion is the non-conceptual world. But she must interpret her viewings of that world; she must turn them into the candidate scientific knowledge that Hull calls information: "In their investigations, scientists learn about the structure of the empirical world. They record this knowledge in a language of some sort. This characterization of the natural regularity counts as information, but the natural regularity itself cannot without making the notion of information vacuous" (p. 142). The data of trials must be read critically; in this sense, a repetition is a reading of a reading. All testing, then, is a match between theory and the non-conceptual world; but no testing can count as science until the data are read and interpreted. This means that testing involves hermeneutic understanding.