Failure to thrive (FTT) is the term widely used to describe poor weight gain in infancy, a condition associated with cognitive deficiency in later childhood. FTT has been investigated in earlier population studies, but little is known about risk factors for FTT or the sequence of events as this requires data to be collected prospectively within the first year of life. Furthermore, several different anthropometric criteria have been used to define FTT, and it is not known whether children identified by the different criteria are comparable. In the present population study we compared risk factors for FTT in a general infant population using different definitions of FTT. Three different criteria of FTT mirroring those used in previous population studies were applied to a birth cohort of 6090 infants. Sociodemographic data and prospectively collected information concerning physical and mental development of the children were obtained from National registries and standardised public health nurse records. Risk factors preceding each of the three 'types' of FTT were compared. The three criteria for FTT identified children with very different profiles and a prevalence of FTT ranging from around 2% to 21% in this affluent population. The criterion of slow weight gain conditional on birthweight (conditional weight gain) was associated with lower birthweight, small-for-gestational-age and deviant overall development. Adding low body mass index did not change this profile. In contrast, the commonly used criterion of downward crossing of centiles on an ordinary weight-for-age chart was associated with factors normally linked with low risk of adverse physical and mental development. Slow conditional weight gain, irrespective of additional thinness, seemed to identify infants with prenatal growth retardation and early developmental delays. In contrast, simple downward crossing of centiles seemed mainly to identify healthy low-risk infants, and thus, seems a poor screening measure of FTT in this affluent infant population. Thus, conditional weight gain appears to be the most sensible measure of FTT at present. However, only longitudinal studies including different anthropometric measures and different outcomes can unravel the discriminating power of the different FTT definitions concerning long-term prognosis.