AbstractThis study examined how expert abacus operators process imagery. Without imagery instructions a digit series was auditorily presented as one whole number (WHL list) or separate digits (SEP list). RT from offset of the probe to onset of the response was measured. The main findings were as follows: experts showed no difference in RT between the two lists, while significant differences occurred in non‐experts; non‐experts' RT increased with probed position, while experts' RT was flat if the series size was within their image capacity; experts' RT increased with probed position when the series size was longer than their image capacity, but its rate of increase was smaller than that of non‐experts; and the smaller the image capacity, the steeper the slope of the RT function. It was concluded that experts spontaneously encode the digit series into an imaged abacus, while non‐experts encode it verbally; that experts directly access the probed position within their image but serially process the verbally coded overflowed part; and that non‐experts search the digit series