Abstract
The World Health Organization (WHO) has extensively revised its manual for semen analysis, and seeks to provide reference limits for semen quality parameters. This raises the question of what is meant by 'normal' (excluding the use of this term to denote a Gaussian distribution). It could be taken in a purely statistical sense, using a biologically arbitrary cut-off point to denote an abnormal level, typically the extreme 5 percent of the population. Alternatively, 'normal' could be defined according to the biological concept of normality and abnormality, in terms of the point at which biological function becomes impaired. Either of these can be used in descriptive epidemiology, for example, to study trends, but in the case of fertility, both semen quality and functional fertility (time to pregnancy) are continuous variables with no clear threshold. The WHO manual uses the biological meaning of normal, in that it provides the semen parameter distributions for men who have recently fathered pregnancies that took 12 months or less to conceive. However, what is really needed is the same information the other way around: given a particular semen test result, what should be expected in terms of ability to conceive, and how long it is likely to take. In considering epidemiological research, the focus has been mainly on internal comparisons, rather than reference limits, but it would be beneficial if more attention were paid to the absolute levels and to what these mean in terms of function-in other words, if the data were better calibrated biologically.