Objective: Measurements of length at birth, or in the neonatal period, are challenging to obtain and often discounted for lack of validity. Hence, classical 'under-5' stunting rates have been derived from surveys on children from 6 to 59 months of age. Guatemala has a high prevalence of stunting (49·8 %), but the age of onset of growth failure is not clearly defined. The objective of the study was to assess length-for-age within the first 1·5 months of life among Guatemalan infants. Design: As part of a cross-sectional observational study, supine length was measured in young infants. Mothers' height was measured. Length-for-age Z-scores (HAZ) were generated and stunting was defined as HAZ < − 2 using WHO growth standards. Setting: Eight rural, indigenous Mam-Mayan villages (n 200, 100 % of Mayan indigenous origin) and an urban clinic of Quetzaltenango (n 106, 27 % of Mayan indigenous origin), Guatemala. Subjects: Three hundred and six newborns with a median age of 19 d. Results: The median rural HAZ was − 1·56 and prevalence of stunting was 38 %; the respective urban values were − 1·41 and 25 %. Linear regression revealed no relationship between infant age and HAZ (r = 0·101, r 2 = 0·010, P = 0·077). Maternal height explained 3 % of the variability in HAZ (r = 0·171, r 2 = 0·029, P = 0·003). Conclusions: Stunting must be carried over from in utero growth retardation in short-stature Guatemalan mothers. As linear growth failure in this setting begins in utero, its prevention must be linked to maternal care strategies during gestation, or even before. A focus on maternal nutrition and health in an intergenerational dimension is needed to reduce its prevalence.