Just, Carpenter, and Woolley propose that measures of single-word processing times, such as gaze durations during normal reading or durations of reader-controlled exposures, can be used both for developing theories of comprehension and for determining the course of new reading technologies. We argue that better understanding of comprehension processes does not necessarily follow from observed correlations between text variables and eye-fixation parameters. Further, recent rapid serial visual presentation (RSVP) research suggests that eye fixation data obtained from studies of normal reading might have little practical value in optimizing the readability of RSVP displays and other means of text presentation.

Experiment 1 compared paragraph comprehension for texts shown either as normal pages on a computer terminal screen or as rapid serial visual presentations (RSVPs) of small text segments to a common location. Over several days of practice, reading comprehension was equivalent in the normal presentation mode and the RSVP format. When successive RSVP segments contained some information in common, to mimic the experience of successive parafoveal and foveal views of words in normal reading, comprehension was somewhat worse than when successive segments contained no overlapping information. Experiment 2 used a variety of segment size and segment duration combinations to investigate the optimal means of presenting text in the RSVP format. Across a variety of presentation rates and text difficulties, comprehension was maximal for segments averaging about 12 character spaces in length. In Experiment 3, texts were divided into short idea units or into random segments of equal average length. Comprehension was shown to be greater in the structured condition than in the random condition. An optimal means of presenting text in the RSVP format could be superior to normal presentation methods for reading and other text-processing tasks.There are few methodologies that enable the study of ongoing reading processes without at the same time interfering with normal reading behavior. One such methodology consists of unobtrusively monitoring a reader's eye movement patterns. Eye movement studies have found that as a person reads a text he or she fixates parts of the text for varying amounts of time. The variability of fixation durations could reflect changing processing loads as a reader progresses through a text (Just & Carpenter, 1980). However, simply monitoring eye movements and fixation durations does not enable experimental control of the parts of text processed by a reader, the sequence of words a reader fixates, or the durations for which words are fixated. Rather, these are regulated by the reader and are, in fact, the variables measured in eye movement studies. Experimental control over what is seen and for

No abstract

Just, Carpenter, and Woolley propose that measures of single-word processing times, such as gaze durations during normal reading or durations of readercontrolled exposures, can be used both for developing theories of comprehension and for determining the course of new reading technologies. We argue that better understanding of comprehension processes does not necessarily follow from observed correlations between text variables and eye-fixation parameters. Further, recent rapid serial visual presentation (RSVP) research suggests that eye fixation data obtained from studies of normal reading might have little practical value in optimizing the readability of RSVP displays and other means of text presentation.

College students searched for single target letters in word and unpronounceable nonword displays of four, five, or six letters in length. The displays were typed in either intact upperor lowercase form or in an alternating mixture of upper-and lowercase letters. Response times and error rates were less for words than for nonwords. Search rates, as measured by the slopes of the functions relating response time to display length, were greater for nonwords than for words. The search process apparently self-terminated for nonword displays, as the negative slope was about twice the positive slope. The positive and negative slope differences were less for words and were about equal for words in intact form. Case mixtures resulted in slower responses for words only, and this effect was apparently limited to encoding rather than to search or comparison processes. The relevance of these findings for similar results in same-different judgment tasks is discussed.A common finding in the word recognition literature is that words are recognized more rapidly and accurately than strings of unrelated letters (e.g., Manelis, 1974;Reicher, 1969) . This perceptual word-superiority effect could be due to differences in the way the letters in the strings are processed , since the orthographic redundancy in words could facilitate letter recognition processes (Massaro, 1975). Alternatively, the word advantage could be due to the use of higher order units, such as spelling patterns or even whole words, that are recognized more efficiently than are strings of independent letters.One way to differentiate among these alternatives is to use displays that are presented in their normal visual form (either all uppercase or all lowercase letters) vs. displays that are altered in some way that preserves their orthographic structure (e.g., by mixing upperand lowercase letters). If the use of orthographic redundancy is the key to word-superiority effects, word advantages over nonwords in perception should be the same for intact and mixed-case displays. In fact, same-different judgment times for simultaneously presented letter strings are slowed more for words than for nonwords when their letters are printed in mixed case relative to intact upper-or lowercase form (Bruder, 1978; Pollatsek , Well,& Schindler,1975 ;Taylor, Miller, & Juola , 1977). Apparently, the mixing of letter cases in words disrupts the visual familiarity of multiletter units and increases the likelihood that they must be processed letter