Underfeeding of gestating dairy goats can occur for a number of reasons. If underfeeding is severe it can provoke pregnancy toxaemia which can lead to welfare problems and even mortality. Behaviour may also be modified in the context of underfeeding and it is important to integrate this information in management practices. The objectives of this experiment were to study the effect of restricted feeding during the last third of gestation on metabolism and behaviour of goats. Special attention was paid to the feed restricted group of goats to define characteristics which could be used to identify animals at risk before they develop pregnancy toxaemia. A total of 60 Alpine and Saanen dairy goats were used. Starting from ?8 weeks before parturition the goats were allocated to one of two dietary treatments, control (C) or restricted (R). The C group was fed ad libitum (5% feed refusals) a TMR (C, n = 30) and the R group (R, n = 30) was given the same TMR but the quantity corresponded to 50% of the amount given to the C group between ?8 and ?4 weeks, 60% between ?5 and ?4 weeks, 70% between ?4 and ?3 weeks and 80% from ?2 weeks to parturition. The R goats were also given free access to straw. Both metabolism and behaviour were modified by food restriction. The R goats put on less weight and their body condition score (BCS) decreased more steeply than the C goats during late gestation. Plasma glucose and ?-hydroxybutyrate (BHB) were lower while non-esterified fatty acids (NEFA) were higher in the R goats compared to the C goats during the period of feed restriction. The R goats spent more time in activities involved with feeding and moving than the C goats. Four of the R goats developed pregnancy toxaemia as a result of dietary restriction. A principal component analysis (PCA) was performed on variables analysed before the start of dietary restriction: liveweight (LW), BCS, fasting plasma glucose, urea, NEFA and BHB. The four R animals that developed toxaemia were characterised by high NEFA and low plasma glucose together with low BCS and high urea before dietary restriction. This particular profile may therefore be used to determine which animals are at risk of developing pregnancy toxaemia.