1. Sherburne F. Cook and Woodrow Borah, Essays in Population History: Mexico and the Caribbean, 2 vols. (Berkeley: University of California Press, 1971; 1974). For an exhaustive bibliography see Nicolás Sanchez-Albornoz, The Population of Latin America: A History (Berkeley: University of California Press, 1974), pp. 263–86.

2. Lombardi inexplicably reports the distribution as “about 20,” “35,” “33,” and “some 15” percent for the periods 1771–1800, 1800–1809, 1810–1819, and 1820–38, respectively (pp. 40–41). The corresponding distributions calculated from the data on page 468 yields 9.1, 52.1, 30.4, and 8.4 percent

3. If sex-specific data had been reported for either single or married adults, all the original data could have been derived by subtraction. With the information published here, one can hardly go beyond the partial data and transformations made available by the author.

4. Lombardi fails to choose a realistic model population, adjust the sex ratio at birth, and translate “párvulos” into the stated chronological age group. The following table permits the comparison of his figures with more likely model population ratios—rate of natural increase of 1 percent instead of zero, sex ratio at birth of 105 males per 100 females, and age groups 0–6 and 0–9 years.

For method of calculation see Ansley J. Coale and Paul Demeny, Regional Model Life Tables and Stable Populations (Princeton: Princeton University Press, 1966), p. 40. Following Lombardi's procedure, I calculated stable population values using Model South, level 3, female life expectancy at birth of twenty-five years (Coale and Demeny, pp. 684, 780). Stable population at ages five and six were approximated by simple linear interpolation. Notwithstanding the unambiguous labels of table A–10 and elsewhere (Lombardi, pp. 77, 78, 137), which indicate that computations are based on ages zero through six years, the stable ratios seem to be calculated with 0–9 years. The term “párvulo” probably had social and religious dimensions as well as a chronological one, reflecting parents' and priests' enthusiasm for religious responsibilities as well as elapsed birthdays. The extraordinary frailty of this kind of test is readily apparent. The model must be selected with little knowledge of the true mortality level, natural growth rate, or departures from stability. Child-woman ratios are particularly ill-suited as a test because tiny variations in the hypothesized growth rate produce large variations in the ratio.

5. Lombardi calculated the population for each parish in the period 1800–1809 by taking a simple average of all returns reported for the parish during that decade; the figures were then summed to produce Bishopric totals. This procedure assigns greater weight to repeated returns.

6. It is impossible to determine the frequency of the sex correspondence pattern for married groups from the published dataset because the information is omitted. Figures 5–4 and 5–8 clearly show that there are two distinct patterns, one before 1810 in which the movement of married husbands is identical to that of married wives and one after 1810 in which changes are proportional but not identical. In figure 5–11, the summary statistics create a more serious distortion. This is a classic example of a statistical model ill-suited to the empirical information. Two outliers have an overwhelming effect on the regression and correlation coefficients. For a discussion of this phenomenon, heteroscedasticity, see John H. Mueller, Karl F. Schuessler, and Herbert L. Costner, Statistical Reasoning in Sociology (Boston: Houghton Miflin, 1970), p. 301. The omission of note six and contradictory labels for figure 5–7 may further confuse the reader.