Nonrecursive structural equation modeling is applied to cross-sectional data from a survey conducted in Seoul, Korea, in 2009 and from geographic information systems, in an effort to model the reciprocal relationship between attitudes toward travel modes and land use in a neighborhood. Then, this study examines how the direction of the relationship differentiates the effects that the two factors have on trip frequencies. The direction is found to be contingent on trip purpose, that is, the attitudes–land use relationship is recursive for commuting trips, reciprocal for leisure trips, and insignificant for shopping trips. If the attitudes are omitted altogether, the estimated land use effect decreases considerably for commuting trips and slightly for leisure trips, which suggests the degree to which residential self-selection is concerned with these purposes of trips.