A study of the cognitive processes involved in achieving coherence in discourse examined a major factor affecting the accessibility of referents: the length of absence of the referent from the distribution of different types of grammatical referential devices (definite nouns and anaphoric pronouns) in discourse. Six versions of a set of twenty short stories were developed by combining two factors: referential distance (zero to five clauses between subsequent references to a character) and anaphoric device (an identical noun, a pronoun, or an epithet). Six experiments were conducted with college students, each using a different group of students and a different version of the stories. In each experiment, the students were asked to read the stories and were periodically interrupted with an inquiry about coreference. The amount of time taken by the subjects to respond to the inquiries was measured by computer, and the response patterns were analyzed in terms of the cognitive processes at work for each of the factor combinations during the decision-making delay. The results are interpreted as an interaction between automated/unattended and analytic/attended components in the processing of referents in discourse. (MSE)