The sample consisted of about 1,500 blocks in Kansas City, Mo., which were stratified by composite indexes of economic status, family disintegration, racial status, and a preliminary assessment of street lighting. The 1970 police offense records were merged with the lighting and socioeconomic data as well as with Dun and Bradstreet's DMI commercial data. Findings indicate that only a slight decrease in the proportion of night crimes to total street crimes is associated with Mercury and Lucalux lighting types. However, various types of street offenses, including rape, murder, and assault, are strongly deterred by Mercury and Lucalux lighting. A basic model is developed that postulates that the proportion of offenses at night is an additive but possibly nonlinear function of neighborhood effects. Each offense is considered as an independent Bernouilli trial resulting in night or day. The model specifies the probability of night, varying with certain block consequences. The model is validated by testing the assumption that the probability that an offense occurs at night is constant among all similar blocks. Although the block effects are significant at the .5-percent level, data indicate that a better regression model can be formulated. Additionally, estimates of proportion in cluster sampling are used to evaluate the sampling effects on measures of significance. Eight tables and 10 figures are included.