Research to date has treated the problem of interchange spacing from an operations perspective only. Almost no safety analysis is used to justify interchange spacing recommendations. In this study, we endeavor to look at the interchange spacing problem from a safety perspective. Data were used from three freeways in California - Interstates 5, 10, and 15, to extract crashes and freeway characteristics. Regression models were developed to relate crash frequencies to highway characteristics and specifically interchange spacing. Predictive efficacy of the best models was tested using a new set of validation data. Related interesting problems for future research were discovered and listed in the paper.