Study question Can a standardized assessment of abnormal human sperm morphology provide additional useful information by identifying men with more severe disturbances in different types of abnormalities? Summary answer Definition-based categorization of sperm head, midpiece and tail defects has shown how differently these abnormalities are distributed in fertile men and other groups of men, thus providing high and low thresholds, a starting point for diagnosis or research purposes. What is known already Several recent studies have reported indisputable genetic origins for various sperm defects. A few studies have also identified associations between environmental factors and low percentages of morphologically normal spermatozoa. Nevertheless, with the exception of rare situations in which the vast majority of spermatozoa have specific, easily characterized defects, such as 'globozoospermia', little attention has been paid to the description and precise quantification of human sperm abnormalities. The lack of standardization in the phenotyping of sperm morphological defects by conventional microscopy is a limiting factor for diagnosis and for intra- or inter-observer or centre consistency in studies investigating the causal factors and possible functional consequences of the abnormalities detected. There are currently no baseline data for abnormalities of sperm morphology based on a standardized classification, in the general population, among fertile or other groups of men. Study design, size, duration This study is based on detailed sperm abnormality datasets acquired by a standardized classification method, from several groups of men, over the same 5-year period. Participants/materials, setting, methods We studied cross-sectional data from fertile men (n = 926), male partners from infertile couples (n = 1747) and testicular cancer patients (n = 239). We used a standardized classification to analyse Shorr-stained slides, taking into account all the abnormalities encountered. Main results and the role of chance Most sperm defects were significantly more frequent in infertile than in fertile men, with 20-30% of infertile men having frequencies of abnormalities above the 95th percentile in fertile men for 9 out of the 15 categories of abnormalities. Interestingly, several head abnormalities were significantly more frequent in patients with testicular cancer than in infertile men, highlighting the particular impact of this condition on sperm morphogenesis. We used the 95th percentile in fertile men as the lower threshold and the 99th percentile in infertile men as an extreme upper threshold, for the classification of morphological abnormality frequencies into three levels: low, intermediate and high. The assessment of several semen samples, with or without a genetic background, for abnormal sperm morphology, based on the percentage of normal spermatozoa, a teratozoospermia index, and the detailed profile of abnormalities categorized according to the three levels proposed, has highlighted the value of detailed phenotyping for diagnosis and research purposes. Limitations, reasons for caution The thresholds proposed for the various categories of sperm abnormality should be considered relative rather than absolute, owing to the known sampling error related to the limited number of spermatozoa assessed per sample, or when studying the general population or populations from regions other than Western Europe. The standardized assessment of abnormal sperm morphology requires time and experience. We therefore suggest that this assessment is carried out during a first andrological check-up or for epidemiological or research studies, rather than in the routine management of infertile couples for assisted reproductive technologies. Wider implications of the findings The study design used for the fertile group of men was similar to that previously used for the WHO reference values, providing a rationale for considering the 95th percentile in fertile men as the level below which abnormalities may be considered to occur at a frequency representing random background variations of a normal spermiogenesis process. The crude frequencies obtained, and the three levels of abnormality frequency proposed for each standardized category of sperm defect, provide baseline data useful for diagnosis and a starting point for future studies aiming to identify associations with genetic or environmental factors. Study funding/competing interests Part of this study was supported by contract BMH4-CT96-0314 from the European Union. The authors have no competing interests to declare.