Human diseases in adulthood are increasingly associated with growth patterns in early life, implicating early-life nutrition as the underlying mechanism. The thrifty phenotype hypothesis proposed that early-life metabolic adaptations promote survival, with the developing organism responding to cues of environmental quality by selecting an appropriate trajectory of growth. Recently, some authors have proposed that the thrifty phenotype is also adaptive in the longer-term, by preparing the organism for its likely adult environment. However, windows of plasticity close early during human development, and subsequent environmental changes may result in the selected trajectory becoming inappropriate, leading to adverse effects on health. This paradox generates uncertainty as to whether the thrifty phenotype is indeed adaptive for the offspring in humans. The thrifty phenotype should not be considered a dichotomous concept, rather it refers to the capacity of all offspring to respond to environmental information during early ontogenetic development. This article argues that the thrifty phenotype is the consequence of three different adaptive processes - niche construction, maternal effects, and developmental plasticity - all of which in humans are influenced by our large brains. While developmental plasticity represents an adaptation by the offspring, both niche construction and parental effects are subject to selection on parental rather than offspring fitness. The three processes also operate at different paces. Human offspring do not become net calories-producers until around 18 years of age, such that the high energy costs of the human brain are paid primarily by the mother, even after weaning. The evolutionary expansion of human brain volume occurred in environments characterised by high volatility, inducing strong selective pressure on maternal capacity to provision multiple offspring simultaneously. The thrifty phenotype is therefore best considered as a manipulation of offspring phenotype for the benefit of maternal fitness. The information that enters offspring phenotype during early development does not predict the likely future environment of the offspring, but rather reflects the mother's own developmental experience and the quality of the environment during her own maturation. Offspring growth trajectory thus becomes aligned with long-term maternal capacity to provision. In contemporary populations, the sensitivity of offspring development to maternal phenotype exposes the offspring to adverse effects, through four distinct pathways. The offspring may be exposed to (1) poor maternal metabolic control (e.g. gestational diabetes), (2) maternally derived toxins (e.g. maternal smoking), or (3) low maternal social status (e.g. small size). Adverse consequences of these effects may then be exacerbated by (4) exposure either to the "toxic" western environment in postnatal life, in which diet and physical activity levels are mismatched with metabolic experience in utero, or at the other extreme to famine. The rapid emergence of the epidemic of the metabolic syndrome in the 20th Century reflects the rapid acceleration in the pace of niche construction relative to the slower physiological combination of developmental plasticity and parental effects.