SIMULTANEOUS BETTER THAN SEQUENTIAL FOR BRIEF PRESENTATIONS

During perceptually intensive tasks such as reading, there is a bottle neck in the information transfer between the large number of alphanume ric characters available and the acquiring of these characters. This i s due mainly to the limited number of characters that one can report a t a glance (also known as the ''magic number 7 +/- 2'') [Psychol. Rev. 63, 81 (1956)]. To examine where in the perceptual pathway this bottl eneck occurred, several investigators tested and compared performance with simultaneous and with sequential target presentations [J. Exp. Ps ychol. 79, 1 (1969); 93, 72 (1972); Percept. Psychophys. 14, 231 (1973 )]. They found that performance was nearly equal in the two cases and concluded that the bottleneck must be due to the limitation of short-t erm memory. However, these studies were limited either by a long stimu lus-onset asynchrony (SOA), or time interval between onsets of icon pr esentations, or by a lack of poststimulus masking. We report on experi ments designed to overcome these limitations. We used shorter SOA's th an did previous investigators, and we removed persistence effects by p oststimulus masking. Our stimuli were presented either sequentially or simultaneously. For the sequential presentation a numeral ranging fro m 0 to 9 was displayed at any one of eight positions 1.5 deg from a ce ntral fixation cross. The appearance of the next numeral in another pa rt of the display coincided with the masking of the previous numeral. This was done for trials of one to four numerals and SOA's of 16.7, 33 .3, and 50.0 ms. For the simultaneous presentation, similar numbers of numerals and SOA's were used, but the numerals were presented at the same time, all followed by masks. We found that, in general, performan ce declined with reduced SOA. Significantly, in contrast to previous f indings, performance at all three SOA's was better for simultaneous th an for sequential presentation, especially for larger numbers of numer als. In addition, control experiments with either one or two sequentia l masks followed by two simultaneous numerals showed that forward mask ing and apparent motion cannot be the main factors responsible for the poorer sequential performance. Since the manner in which the numerals were presented (either simultaneous or sequential) influenced perform ance, the bottleneck is most likely due to the properties of the proce ss for transferring information into short-term memory rather than to the limitation of short-term memory itself. In addition, we confirmed the previous finding that it is still possible to report four sequenti ally presented numerals with acceptable accuracy if temporal order is not required, and we extended the result to show that it can be done a t SOA as brief as 16.7 ms/numeral. Further, we found that eccentricity had relatively little influence on performance.