1. Log values are perhaps best used descriptively to flesh out common patterns across variables whose values differ or change dramatically. In a more recent examination of the Cleveland data, Professor Gary C. Jacobson and I employ log values in this way. See “Congress and the Presidency as News in the Nineteenth Century,” Journal of Politics (1987), forthcoming.

2. For several reasons, the number of political articles measured in this way is significantly larger than those reported in Figure la of the original article. First, the definition of political news is more inclusive than the one we used. Any news involving a government agency tended to be included under this rubric in the index. Second, because the synopses were so brief as to make coding unreliable, we excluded from analysis all entries of one column inch or less. Frequently, these consisted of announcements and brief anecdotes to fill in the interstices of articles. I failed to note this in the appendix; for a fuller discussion of the methodology employed here, see Kernell and Jacobson (1987). Third, I drew samples for the 1870s, where here the entire population of articles is included.