Considerable research has focused on the identification of predictors of suicidal ideation. This research has led to the identification of numerous demographic variables and problems in psychosocial functioning that appear to be related to suicidal thinking. Typically these variables have been combined in linear models to predict suicidal thinking, and while the resulting models have been statistically significant, the proportions of explained variance in these studies have consistently been relatively low. These consistent findings have led researchers to suggest that models of suicidal thinking should make use of nonlinear terms. To date, however, relatively few studies have used nonlinear methods to predict suicidal ideation. In this study a convenience sample of 526 responses to selected subscales of Hudson's MPSI was obtained in two separate studies. The first sample was obtained from persons seeking services in any of several family service agencies, while the second was obtained from women seeking to terminate a pregnancy in a women's health clinic. Data were obtained on demographic variables and on a number of measures of problems in psychosocial functioning: depression, problems with self-esteem, aggressive behavior, suicidal thinking, personal stress, and family relationships problems. These data were then used to investigate the relationship between the demographic variables, the measures of problems in psychosocial functioning, and the severity of suicidal thinking. Nonlinear regression methods were used to build an intrinsically nonlinear model for predicting severity of suicidal ideation. The results suggested that a nonlinear model including only three variables - depression, problems with aggressive behavior, and family relationships problems - predicted severity of suicidal thinking, and that this nonlinear model explained about 52% of the total variation in suicidal thinking in the combined samples. Statistical tests suggested that this nonlinear model was invariant for both samples of respondents. Results further suggested that some variables previously found to be predictive of suicidal ideation, such as level of personal stress, may not be related to suicidal thinking in nonlinear models. This raises the possibility that some variables currently considered to be predictive of suicidal thinking may be based upon evidence that is an artifact of analytic approach. These results imply that researchers need to continue investigating nonlinear models for predicting suicidal thinking, and that this particular model warrants further investigation. The results also have a number of implications for how practitioners think about, and conduct, suicide risk assessments. First, social workers and other practitioners in health and mental health need to pay close attention not only to the level of depression being experienced by clients but also to relationships problems within their families and the extent to which they engage in aggressive behavior. Practitioners also need to think about the relationship between problems in psychosocial functioning and severity of suicidal ideation in terms of what might be called “interactive co-morbidity.” This may be especially so for the manner in which the co-morbidity of depression, family problems, and aggressive behavior combine to elevate a client's level of suicidal thinking.