riticisms of the data that have been presented to demonstrate a decline in sperm counts over time reveal problems with semen analysis methods and a lack of understanding of the genetic and environmental factors that determine a man's sperm count. Potential sources of error in the WHO semen analysis protocol and some areas of ignorance about the biological and environmental factors which can influence sperm counts are briefly discussed. I conclude that there is a need to include semen analyses in a large cohort study to fill the gaps in our knowledge.The paper by Carlsen et al. 1 that is being marked in this issue of Asian Journal of Andrology stimulated widespread concern and considerable research effort. Together with other supportive studies, e.g. Auger et al., 2 strong evidence that testicular cancer was not only becoming more prevalent but was developing at lower ages 3,4 and observations that other disorders of male development, e.g. cryptorchidism and hypospadias, were being more frequently reported, although this might be due to changes in clinical practice (see Thonneau et al. 5 ), have raised serious concerns that something was going amiss with the human male. The proposition that oestrogen exposure in utero might be responsible by an effect on Sertoli cell development gave rise to the 'testicular dysgenesis' hypothesis to explain these effects, 6 although the role of oestrogen has been questioned recently. As most andrologists, I have followed developments in this area with keen interest but other contributors to this issue are far more knowledgeable than me, so I intend to say nothing about the testicular dysgenesis hypothesis or the latest consensus on whether sperm counts have indeed declined. Instead I will review some of the reasons why assessing trends in semen quality over time is difficult, and how these problems reveal gaps, or perhaps chasms, in our knowledge about the biology of human semen.First, although there is no better alternative, the standard semen analysis is inherently imprecise and is prone to many errors. The equipment used has changed over time and with it the nature of the error, making historical comparisons difficult. The errors introduced could be random, leading to imprecision and greater difficulty in demonstrating statistical significance but without effect on the nature of the trend, or systematic, leading to bias in the results. In theory systematic error could be corrected by direct comparison of former and current techniques. However, as discussed below results can potentially be affected by technical errors whose effects are difficult to predict. These can only be detected and eliminated by good quality control systems that have only been in place for at most two decades. Therefore doubt about older data will always remain. Secondly, the environmental factors that might influence a man's sperm count have changed over time and movements of populations have changed the genetic mix in many places. Moreover, social changes might affect the characteristics of the men rec...