Recent researchers have suggested that the combination of horizontal curves or “bendiness” of a length of road contributes to traffic crash occurrence. A previous study of New Zealand fatal crashes using an aggregated approach found no significant correlation between crash occurrence and road bendiness for rural roads but a minor correlation for urban roads. This thesis further explores the effect of road bendiness on traffic crash occurrence in New Zealand by developing a method more suited to traffic engineering. The method involves Geographical Information Systems (GIS) firstly to process data and secondly to calculate bendiness values. The following bendiness measures: bend density, detour ratio, cumulative angle, mean angle and standard deviation of angles; are applied to “influence areas” surrounding crash and comparison sites. The method then dictates that some form of statistical analysis should be performed to distinguish between the bendiness of crash and comparison sites, while accounting for other influencing factors. Binary logistic regression is recommended. The method was applied in a case study of New Zealand fatal crashes, with two main analysis techniques employed. Firstly, binary logistic regression models were developed. It was found that, for rural roads, sections with consistent and frequent curves were safer than completely straight sections or those with isolated curves. The urban model was less conclusive, which suggests that the method was not appropriate in the urban situation. The second analysis method involved comparing bendiness values of a site’s “immediate area” with those of its influence area. It was found that, although the spreads of the comparison sites’ distributions were smaller than those of the crash sites, the mean values were generally very similar and no appropriate bendiness ratios could be specified to reduce crash risk. Overall it appears that, if design consistency is maintained, bendiness is a protective quality for rural roads.