Predictive models were developed that relate lake trophic state to diatom assemblages in surface sediments. These models can be used to assess quantitatively the historical impact of cultural activities on lake productivity. One set of models relates trophic state index (TSI) values to a new diatom index based on autecological nutrient categories. A second set uses nutrient categories in multiple regression equations to predict TSI. Both types of models explain ~80% of the variance in total phosphorus TSI and averaged TSI values. Diatom‐pH transfer functions were also developed to assess the influence of eutrophication on lake pH.Because trophic state and pH are positively correlated in Florida lakes, statistical confounding might result when predictive models for pH and trophic state are calibrated. Long‐term changes in pH might lead to errors when inferring historical TSI values, while changes in trophic state could similarly influence historical pH inferences. Partial correlations showed that two multiple regression models were statistically confounded by the nontarget variable, though predictive models based on diatom indices were free from such effects.