Abstract: The simplest information processing theory of STM asserts thatSTM has a capacity for a fixed number of chunks. To gain a deeperunderstanding of STM capacity, it would be useful to compare theimmediate recall span using Chinese characters and phrases with thatusing English words. Here, contral experiments were conducted forcomparison, using three kinds of Chinese materials: one-character words,two-character words and four-character phrases, each with two levelsof familiarity, two manners of expression and two learing methods. Theresults indicated that STM capacity approaches more closely to constancywhen measured in chunks than when measured in other units, but theSTM span decreases slightly with increase in chunk complexity. Relativechunk constancy was also observed in the time required for rotelearning of stimuli. The degree of familiarity of stimulus materials playsonly a moderate role in STM capacity. The visual STM span is largerthan the auditory span. Analysis of the serial position of STM capacityshows that for the first half, STM through vision is better than thatthrough audition. Some speculative explanations were offered.

Key words: NULL