Two passages were written, one on the topic of graphs and one on the topic of sonnets. Sixteen versions of each passage were constructed. Each version contained identical content but different textual manipulations. One hundred twelve community college students and 48 Cornell students participated in the experiment. Each read a single version of both passages, recorded time to read, wrote free recalls of both versions, and then answered a series of probe questions. The community college subjects recalled less of the passages; recalled only isolated pieces of information; and produced recalls which were greatly affected by the presence or absence of explicitly stated logical and relative relations. The Cornell subjects recalled the passages more completely; tended to recall more whole propositions or inter-propositional units; and produced recalls that were not affected by the textual manipulations. It was concluded that logical and relative relations between propositions are part of memory for text, and that the Cornell subjects reflect a population of truly-fluent readers while the community college subjects reflect a population of non-so-fluent readers. (Author)