The essence of fractal analysis is seeking for a pattern that is indep
endent of scale. This paper examines the existence of long-term memory
in nine Asian stock markets together with US and UK indices using the
modified rescaled-ranged (R/S) statistic. The modified R/S statistic
is robust not only with respect to the normality assumption, but also
to short-term autocorrelation. The data in the sample range from 1 Jan
uary 1988 to 30 June 1992 and are arranged in daily, weekly and monthl
y returns. In most cases, the phenomenon of long-term memory is not fo
und; hence the random walk hypothesis cannot be rejected. The UK marke
t, however, exhibits some long-term memory for various data frequencie
s and lags. The result of this paper provides directions for future re
search.