Historical reversals highlight a basic methodological problem: is it possible to treat two successive periods both as independent cases to compare for causal analysis and as parts of a single historical sequence? I argue that one strategy for doing so, using models of path dependency, imposes serious limits on explanation. An alternative model which treats successive periods as contrasting solutions for recurrent problems offers two advantages. First, it more effectively combines analytical comparisons of different periods with narratives of causal sequences spanning two or more periods. Second, it better integrates scholarly accounts of historical reversals with actors' own narratives of the past.Between the late 19th century and World War I, San Francisco's bourgeoisie moved in succession along two opposing paths. In contrast to most U.S. cities, the trend in San Francisco from the 1880s through the 1900s was for businessmen to be divided among themselves and for a majority of employers to recognize and bargain with labor unions. This pattern reflected the cumulative effects of racial cleavages in industry (smaller employers allied with craft labor against the Chinese), the balance of class power (a powerful labor movement made collective bargaining prudent for employers insulated from national competition), and the character of business organization (most manufacturers organized by trade rather than in more inclusive civic organizations). Between the early 1910s and the 1920s, however, the pattern switched. Employers increasingly rallied together, particularly in opposition to unions. In the early 1920s, they finally purged unions from such strongholds as metal working and construction, replacing collective bargaining with the open shop (Haydu 2008).