The literature on the protection imparted by conventional whole-cell pertussis vaccines was reviewed, and the extent to which the great variation in estimates of vaccine efficacy is attributable to methodologic problems in study design and analysis or to biologic features of the natural history of pertussis was explored. The protection against disease imparted by pertussis vaccines may be greater than that against infection. Estimates of vaccine efficacy from case-control studies are higher than those from studies of household secondary-attack rates; likewise, estimates of efficacy are higher when based on clinically severe or bacteriologically positive cases rather than simply on notified cases. Some of the reported differences in protection by different vaccines may be attributable to relations between the antigenic composition of the vaccine used and that of the circulating strain of Bordetella pertussis. Failure to consider age trends has sometimes led to spuriously high estimates of efficacy. Many biases can affect efficacy studies, and it is usually difficult to assess whether the net effect has been to underestimate or to overestimate "true" efficacy. The immunity imparted by conventional pertussis vaccines may be neither as solid nor as stable as that imparted by many live-virus vaccines. These issues must be considered during the evaluation of acellular pertussis vaccines.