Fig. 4. Energy budget for knots feeding on poor-quality cockles as a function of gizzard size. For the experimental birds, gizzard mass = 8.13±0.98 g (mean ± S.E.M.), we plotted the cumulative cost levels for the three components that we measured (rest, foraging and heat increment of feeding, HIF; filled circles; figures as in Table 2B).Values are means ± S.E.M. Theoretically, each of these cost levels would increase with gizzard mass (with concomitant increases in intestine mass, Table 3), since larger gizzards require larger maintenance costs (Piersma et al., 1996), larger gizzards increase the cost of walking (Bruinzeel et al., 1999), and larger gizzards quadratically increase HIF (since intake increases quadratically with gizzard size; see van Gils et al., 2003a). Two digestive constraints are presented: (1) the gizzard-size independent rate at which flesh can be digested (horizontal grey bar; based on Kirkwood, 1983 and Kvist and Lindström, in press), and (2) the rate at which shell material is processed, a quadratic function of gizzard size (solid curve; van Gils et al., 2003a). Intake rates measured in the two most `natural' trials (means of trials IF-C<sub>1</sub> and IF-C<sub>2</sub>; open square) appear to be set by this `gizzard-size constraint'. The hatched area below the two digestive constraints but above the cumulative cost levels gives the scope for a positive energy budget and is maximal at the arrow.