The small-scale vertical (10′s of meters) and horizontal (100′s to 1000′s of meters) distribution of zooplankton in the California Current near Guadalupe Island, Baja California, Mexico was studied. Vertical distributions were sampled using a vertically-towed Longhurst-Hardy Plankton Recorder (LHPR) which gave a sequence of samples, each integrated over about 5 m, from 250 m to the surface. Because of the sampling biases of the LHPR, details in vertical structure of less than 15 to 20 m were not considered. The pattern of the vertical tows and the variability of the integrated counts of species were used to infer horizontal distribution. Four series of 8 tows each were taken around noon and midnight over a 2-day period. Four of the 8 tows in each series were randomly positioned within 2000 m of a parachute drogue (first day) or a fixed geographic position (second day); four were replicate tows taken at the drogue or the fixed position. Sixty-seven taxonomic categories were counted. The replicate tows, separated by no more than a few hundred meters, gave more similar vertical profiles for species than did the random tows, with separations of 100′s to 1000′s of meters. The night replicate tows showed less variability in depth distribution than did either the night random or any of the day tows, leading to the hypothesis that the vertical distributions observed were generated by interactions of the organisms' diel behavior with internal waves. Variability of abundance estimates using the integrated counts was the same for both replicate and random tows, indicating that horizontal patches may be smaller than 100 m. No evidence was found for a day-night change in patch size, or for a consistent overlapping of patches of different species. Replicate tows gave more similar estimates of community structure (relative proportions of species) than did random tows. Overall day community structure was more similar between tows than night structure. Similarities in species' proportions of any random tow to the replicate tows or to other random tows of a series decreased with increasing distance between the tows being compared. This decrease was greater for the night samples, suggesting that community structure is more heterogeneous at night.