Many findings concerning reading comprehension during the past decade have been based on the study of reading in the laboratory. Laboratory reading, however, may involve shallower processing or entail different goals than natural reading. To determine whether laboratory reading is a satisfactory approximation of natural reading, college students were assigned to natural or laboratory reading conditions depending on whether they had read a particular newspaper article or not. The laboratory readers viewed the article on television screens. Tests of free recall, recognition, and interestingness were administered immediately (to laboratory readers only), 1 day, 2 wk., or 6 wk. after reading.The free recall analyses used the principles of text analysis, scoring, and evaluation of Kintsch and van Dijk (1978). Natural and laboratory readers were highly similar in the proportion of macro-and micropropositions they recalled, in the ratio of reproductive to reconstructive recall, and in the memory parameters provided by Kintsch and van Dijk's (1978) coherence graph model. In recognition, natural and laboratory readers showed similar profiles of performance, although laboratory readers scored higher. It is proposed that this advantage resulted from the laboratory readers' particular concern with the superficial details of the message. Finally, natural readers rated the article more interesting than laboratory readers did.It is important that the analyses revealed, for both laboratory and natural readers, familiar effects including (a) better memory for macro-than micropropositions, (b) decreasing performance with test delay, and (c) increasingly reconstructive recall with increased delay. These effects coupled with the similarity of the natural and laboratory readers, provide reassurance that it is reasonable to treat laboratory reading as an approximation of natural reading.