Naturalism holds that there is no higher access to truth than empirically testable hypotheses. Still it does not repudiate untestable hypotheses. They fill out interstices of theory and lead to further hypotheses that are testable.

A hypothesis is tested by deducing, from it and a background of accepted theory, some observation categorical that does not follow from the background alone. This categorical, a generalized conditional compounded of two observation sentences, admits in turn of a primitive experimental test.

The observation sentences themselves, like ape cries and bird calls, are in holophrastic association with ranges of neural intake. Denotation of determinate objects figures neither in this association nor in deducing the categorical from the scientific hypotheses. Hence the indeterminacy of reference; ontology is purely auxiliary to the structure of theory. Truth, however, is seen still as transcendent at least in this sense: we say of a superseded scientific theory not that it ceased to be true, but that it is found to have been false.