relationship. Further discussion of the subtle and potentially confusing issues involved in useful. We welcome the opportunity to clarify our definitions. Three time dimensions are germane for time-variance analysis-two to the theoretical formulation, and two to the empirical study, with the length of the interval applying to both: 1. The T. The focal time dimension is the the length of time over which each return (expressed as the annualized log of price relatives) is measured. The time-variance relationship refers to the way in which variance of the interval returns changes as the interval is altered. 2. The short period, n. In the theoretical analysis, the relationship between the length of the interval and returns variance was obtained by introducing a second time dimension which we refer to as the short or the short period. Decomposing the interval enables us to show how the time-variance relationship would be affected by patterns of first and higher order autocorrelation in the short period returns [see our eq (3), p. 42] 3. The Calendar Span. Because intertemporal correlation patterns in short period returns can affect the time-variance relationship, the purpose of our empirical tests was to determine how the variance of the interval returns varies with the length of the interval itself. To do this, a series of returns had to be obtained for different values of the interval (we used eight different values of T). Because returns distributions for any given value of T might not actually be stationary over calendar time, we have used just one calendar span (360 weeks) to measure the returns series. The calendar span plays no role in the theoretical analysis. Schneller's first claim is that there can be confusion between differencing intervals and intervals in regard to the behavior of security returns. In our first paper, SW [6], we never mention the term interval, and in our second paper, SW [7], we clearly define it to be the interval. Schneller takes T to be the interval (as do we), but n to be the measurement interval. We feel this is misleading; the short period plays no role in the empirical study. Schneller states that . . the reader is posed with a paradox: if the one period return is stationary, how can its variance and R' change with a change in T? (p.