On the epistemological status of semantic networks