In a Recent environment two foraminiferal populations can be observed, living and dead. If a Recent environment were preserved at a given instant, the fauna available for observation would be the total population. The dead population is often one or two orders of magnitude larger than the living population, and therefore the dead population is essentially equivalent to the total population. Because most species have certain environmental preferences which often include a characteristic depth range, paleoecological interpretations can be made providing the preserved (fossil) population is equivalent to the total population. Ecological studies of Recent Foraminifera have therefore shown not only the relationship of environmental variables to the distribution and abundance of the living population, but also the relationship of the living population to the total population. The basic data for such studies consist of counts of the numbers of individuals of each species in each sample (a sample usually consists of 10 ml wet sediment). Field and laboratory methods have been described by Phleger (1960). In an area of ecologic interest there are usually many species. While similarities and differences in the distribution of abundances of a single species are often easily recognizable, the assimilation of a large amount of data consisting of the abundance of many species distributed among many samples is very difficult. Greig-Smith (1964) summarized many of the quantitative procedures that have been used to deal with such problems in plant ecology. Most of the procedures calculate from a data matrix a measure for ascertaining the degree of association or correlation between species or between samples. Some measures use presence or absence data while others use abundances. The resulting association or correlation matrix is then subjected to analysis to discover grouping or an underlying structure. Some of the procedures utilize various ramifications of 'factor analysis' for this purpose. Seal (1964, pp. 170-9) reviews the statistical applicability of these procedures. Streeter (1964) has used vector analysis (see Imbrie 1963), a somewhat similar technique, on foraminiferal distributions. Recently, Kaesler (1966) has applied the methods of 'numerical taxonomy' (Sokal & Sneath 1963) in defining Ostracoda and Foraminifera assemblages. All of the above procedures define groupings by consideration of the entire study area as a single unit or statistical universe. The purpose of this study is to investigate the utility of a type of multi-variate analysis called canonical variate analysis which is a method for comparing or discriminating between two or more statistical universes. Rather than trying to define groups or an underlying structure, the method statistically compares groups which the investigator has already defined by some criteria of interest such as bathymetry, sediments, hypothesized faunal boundaries, etc. Our main concern here is not how the researcher constructs his hypothesis of subareas or biofacies but how he can compare them. The data for the analyses of the present study are taken from Phleger's (1956) classic study 'Significance of living foraminiferal populations along the Central Texas Coast'. The data from his