Some have argued that failure to find evidence of a link between psychosocial stress factors and birth outcomes in humans may be attributed to a restriction in the range of measures used to characterize stress severity. This study describes the process of building and testing a measurement model from real life stressful experiences. To create a conceptual model, seven life events identified as stressful by successive samples of 3205 and 2090 Hartford inner city women between 1987 and 1992 and ten other experiences identified as potentially stressful were submitted to 10 academic colleagues and 4 health professionals associated with a preterm birth prevention project. Each rater attempted to identify a meaningful conceptual structure for clustering these psychosocial stressors. Seven of the thirteen proposed models were suitable for testing by means of structural equation modeling. Three models emerged as conceptually close and statistically acceptable with the number of information laden stressors dropping from 17 to 11 and clustering to form three latent variables representing economic, social support, and family stressors. The root mean square error of approximation was 0.03, where a value as low as 0.08 represents an adequate fit, and the generalized fit index was 0.99. The final model was cross-validated on the second sample where these and other fit indices were similar, indicating that a structurally meaningful measurement model fit the data well and could be used to investigate further the relationship of these theoretical variables to birth outcomes.