This paper investigates variations in both space and time of total column ozone through the Nimbus 7 Version 7 Total Ozone Mapping Spectrometer (TOMS) data. While the gridded daily TOMS dataset has high spatial and temporal resolution, the asynoptic nature of the data and artifacts in the observations make it difficult to use at the available level of resolution. In the past, only subsamples or averages over both space and/or time of the TOMS data have been subjected to statistical analysis. This work discusses three ways of modeling the space-time structure of the TOMS data. We conclude that a representation in which the data points at a fixed latitude are treated as if they were on a line in the longitude-time plane is the best. At each grid point, the total column ozone field is partitioned into a 31 day moving average and a remainder or residual ozone field. For a fixed latitude and a fixed season, the residual field can be reasonably modeled as a stationary random process with mean zero. The sample correlation functions are calculated for 52 seasons over 13 years and 7 latitudes. Comparing these correlation functions and their associated periodograms across seasons, years, and latitudes provides a useful means for discerning patterns in the TOMS data. For example, the correlation functions show that at latitudes around the equator, the apparent movement of ozone is from east to west, and at higher latitudes, the movement is from west to east. In addition, the movement is more strongly from west to east in the southern hemisphere than in the northern hemisphere. Periodograms for latitudes near the equator show clear evidence of a scan-angle effect noted by others. However, our analysis demonstrates that the scan-angle effect shows up at frequencies other than the orbital frequencies and its harmonics. We suggest a simple filtering method for removing most of this artifactual effect. ¿ 1998 American Geophysical Union