No consensus has emerged on how to estimate adult mortality in countries lacking complete vital registration of deaths and accurate periodic censuses. This paper applies a range of methods to census, registration and survey data for Guatemala for the period from 1981 to 1994. The findings are less than conclusive because of marked errors in the census populations. Methods using intercensal survival perform very poorly, giving rise to results that are hard to interpret. Methods using the distribution of deaths by age and rates of change of the population by age appear to work better, but still give rise to substantially different results. Simulations suggest that a combination of two methods appears to work well. In the Guatemala case, survival of mother appears to over-estimate female adult mortality, whereas survival of siblings appears to underestimate adult mortality. A new method for analyzing intercensal changes in cohort proportions with surviving mother, presented in the paper, gives results broadly consistent with estimates based on adjusted registered deaths.