Herzog and Miller (1998) reported that people judged alleys with sharper curves as less dangerous than straighter alleys. The authors investigated the role of perceived alley length as a possible confounding influence. Raters judged a large sample of urban alleys for two target variables (preference and danger) and six predictor variables (setting care, mystery, shadow, curvature, length, and width). The partial correlation of danger and curvature, with length partialed out, was not significant. However, in regression analyses controlling for the entire set of predictors, curvature had a modest negative partial relation with preference and a tendency toward a modest positive partial relation with danger. Thus, when controlling for all predictors, the counter-intuitive relation between danger and curvature disappears. In the regression analyses, the strongest predictors were setting care (positive for preference, negative for danger) and shadow (the opposite pattern). Mystery tended to be positively related to both target variables.
Herzog, Thomas R. and Flynn-Smith, Jennifer A., "Preference and Perceived Danger as a Function of the Perceived Curvature, Length, and Width of Urban Alleys" (2001). Peer Reviewed Articles. 21.