Twenty-nine compounds, including amino and nitro derivatives of aromatic hydrocarbons as well as azaarenes and indole analogs, were assayed. The biological activity was monitored as population growth to axenic cultures of the ciliate Tetrahymena pyriformis. Structure-activity relationships were examined by linear regression analysis of log biological activity ( Y) and log 1-octanol/water partition coefficient ( X). Partition coefficients were determined by linear regression analysis of the retention time index of HPLC data. Attempts to generate linear predictive correlations for all the test compounds ( n = 29; r 2 = 0.325; P > 0.0013) and groups, subdivided based on the following chemical classes: aromatic hydrocarbons ( n = 13; r 2 = 0.122; P > 0.2417), azaarenes ( n = 9; r 2 = 0.348; P > 0.0946) and indole analogs ( n = 7; r 2 = 0.743; P > 0.0126), met with little success. Similarly, subdivisions based on the presence of a single ( n = 14; r 2 = 0.514; P > 0.0039) or multiple H-polar groups, i.e., NO 2, NH 2 ( n = 15; r 2 = 0.195; P > 0.0996), met with little success also. However, if one segregates the compounds which have para oriented H-polar substituents one gets a strong linear predictive correlation ( n = 11; r 2 = 0.889; P > 0.0001). A subdivision of the remaining data set leads to two models ( n = 11; r 2 = 0.845; P > 0.0001) and ( n = 7; r 2 = 0.965; P > 0.0001). An examination of the toxicity partitioning product values suggests the three-line predictive correlation model of these data is valid.