Optimal Foraging Theory: How Animals Decide What to Eat and Where to Feed
Topic: Animal Behaviour & Ethology Subject: Diet Choice, Patch Use, and the Economics of Foraging Last verified: 2026-09-23
Every foraging animal faces a series of decisions. Should it eat the prey item in front of it or keep searching for something better? How long should it stay in a patch of food that is slowly running out? Should it choose a reliable food source or a variable one? And how much food is worth the risk of being eaten while collecting it? In 1966, two papers published side by side in The American Naturalist proposed that these decisions could be understood as economic problems that natural selection has shaped animals to solve efficiently [MacArthur & Pianka 1966; Emlen 1966]. The approach that grew from them, optimal foraging theory, became one of the most productive and most argued-over frameworks in behavioural ecology [Pyke 1984; Stephens & Krebs 1986].
This article explains the core models of the theory, the diet (prey) model and the marginal value theorem, and the classic experiments with great tits, bluegill sunfish and starlings that tested them. It then covers two major extensions, risk-sensitive foraging and the trade-off between feeding and predation risk, the critiques the theory attracted, and how modern animal tracking reopened questions about how animals search. It closes with the practical uses of foraging theory in conservation and animal welfare.
Origins: Foraging as an Economic Problem
Robert MacArthur and Eric Pianka asked how a predator should use a "patchy" environment, and in particular which kinds of food it should include in its diet. Their graphical argument treated each food type as offering a gain set against the time spent searching for it and the time spent pursuing and eating it. A forager should widen its diet only while adding another food type increases its overall return, which implies that animals in rich environments should be more specialised than those in poor ones [MacArthur & Pianka 1966]. In the same issue of the journal, J. Merritt Emlen independently argued that food preferences should reflect the time and energy that different foods cost and provide, and should therefore shift with food availability [Emlen 1966].
The underlying logic was made explicit in David Stephens and John Krebs's book Foraging Theory, which synthesised the first two decades of work. Every foraging model, they argued, has three components: the decision being studied (for example, which prey to attack or when to leave a patch), the currency by which alternatives are compared (most often the long-term rate of net energy gain), and the constraints that limit what the animal can do, such as the time needed to handle prey or the animal's ability to recognise prey types [Stephens & Krebs 1986]. The word "optimal" refers to a testable prediction from a clearly stated model, not a claim that animals are perfect [Stephens & Krebs 1986].
The Diet Model and Its Tests
The standard prey (diet) model ranks food types by their profitability: the energy gained divided by the handling time needed to pursue and eat them. The model predicts that the most profitable type should always be taken when encountered. Less profitable types should be added in rank order only when better prey are scarce enough that ignoring poorer items would lower the overall rate of gain. A striking consequence is that whether a low-ranked prey type is included should depend on the abundance of higher-ranked prey, not on its own abundance [Stephens & Krebs 1986].
Earl Werner and Donald Hall tested these ideas with bluegill sunfish (Lepomis macrochirus) feeding on populations of the water flea Daphnia magna made up of different size classes. After correcting for the fact that larger prey are more visible and therefore encountered more often, they found that at low prey abundance the fish ate prey of different sizes as they came across them. As overall abundance rose, the smaller size classes were dropped from the diet in sequence, as the theory predicted. Estimates of search and handling times indicated that these diet shifts maximised the fishes' return for the time spent foraging [Werner & Hall 1974].
John Krebs and colleagues ran a similar test with captive great tits (Parus major). The birds were offered large (profitable) and small (less profitable) pieces of mealworm passing on a moving conveyor belt, so that the experimenters could control how often each prey type was encountered [Krebs et al. 1977]. When encounter rates were low, the birds were unselective. When large prey were encountered often enough, they became selective and ignored the small pieces [Krebs et al. 1977]. The birds did not, however, switch as sharply as the simplest model predicts. Instead of all-or-nothing choices, they continued to take some less profitable prey, a pattern known as partial preference that later models sought to explain [Krebs et al. 1977; Stephens & Krebs 1986].
The Marginal Value Theorem: When to Leave a Patch
Food is rarely spread evenly. More often it occurs in patches, such as a bush of berries or a mudflat of worms, that become depleted as they are exploited, so that the rate of gain falls the longer an animal stays. In 1976 Eric Charnov formalised when a forager should leave. His marginal value theorem states that an animal maximising its long-term rate of gain should leave a patch when its instantaneous rate of gain in that patch falls to the average rate of gain for the habitat as a whole, including the time spent travelling between patches [Charnov 1976]. One prediction follows directly: when patches are further apart, so that travel takes longer, a forager should stay longer in each patch and extract more from it [Charnov 1976; Stephens & Krebs 1986].
Richard Cowie tested this prediction with great tits foraging in a large indoor aviary. The birds took pieces of mealworm hidden in sawdust-filled cups that acted as depleting patches. Cowie manipulated travel time by changing how difficult it was to remove the lids covering the cups [Cowie 1977]. As predicted, the birds spent longer in each patch when travel time was longer. The observed times agreed well with the predicted ones once the model accounted for the different energetic costs of foraging and travelling, so that net rather than gross energy intake was used as the currency [Cowie 1977].
Central-Place Foraging
Many animals do not eat food where they find it but carry it back to a fixed point such as a nest or burrow. Gordon Orians and Nolan Pearson developed the theory of central-place foraging for such animals. It extends patch-use models to predict how the size and selectivity of the loads an animal brings back should change with the distance travelled [Orians & Pearson 1979].
Alex Kacelnik tested these predictions with wild European starlings (Sturnus vulgaris) feeding their nestlings. Parent birds collected mealworms from an artificial feeder at which the cumulative number of prey obtained rose ever more slowly with time spent there, mimicking a depleting patch. The feeder was moved to distances between 8 and 600 metres from the nest, and 1,364 round trips were recorded [Kacelnik 1984]. As central-place theory predicts, the starlings collected larger loads when the feeder was further from the nest [Kacelnik 1984].
Risk-Sensitive Foraging
Early models assumed that foragers respond only to the average rate of food intake. Thomas Caraco, Steven Martindale and Thomas Whittam showed that animals also respond to its variability. They offered yellow-eyed juncos (Junco phaeonotus) a choice between a feeder that always delivered the same number of seeds and one that delivered a variable number with the same average [Caraco et al. 1980]. Birds on a positive energy budget, expected to gain enough food to meet their needs, preferred the constant feeder and were risk-averse. Birds on a negative energy budget, which would fall short by relying on the average, preferred the variable feeder and were risk-prone [Caraco et al. 1980].
David Stephens explained the logic of this pattern. If an animal must reach an energy requirement to survive, such as enough reserves to last the night, then when the expected intake exceeds the requirement, variability can only increase the chance of falling short. When the expected intake is below the requirement, gambling on a variable option offers the only chance of meeting it [Stephens 1981]. This prediction became known as the energy budget rule, and it moved foraging theory away from maximising average gain towards the probability of survival [Stephens 1981; Stephens & Krebs 1986].
Foraging Under the Risk of Predation
Foraging exposes animals to predators, so the best feeding decision is often not the one that maximises energy intake. In a major review, Steven Lima and Lawrence Dill concluded that animals can assess and behaviourally control their risk of being preyed upon in ecological time, and that this ability strongly influences decisions made by feeding animals [Lima & Dill 1990]. The decisions they reviewed reflected apparent trade-offs between the risk of predation and the benefits of feeding and other activities. They argued that virtually every animal is potential prey for others, so predation risk should be considered throughout the study of behaviour [Lima & Dill 1990].
Joel Brown turned this trade-off into a practical measurement tool. Building on the marginal value theorem, he argued that a forager should leave a patch when its harvest rate no longer exceeds the combined costs of foraging there: the metabolic cost, the predation cost, and the missed-opportunity cost of not doing something else [Brown 1988]. The amount of food left behind when the animal quits, its giving-up density (GUD), therefore measures how costly the forager perceives that patch to be. By placing trays of food mixed into sand or another substrate in different places and weighing what remains, researchers can compare perceived habitat quality, predation risk and competition across sites [Brown 1988]. Higher giving-up densities indicate patches that foragers treat as more costly or dangerous [Brown 1988; Bedoya-Perez et al. 2013].
Critiques of Optimality
Optimal foraging theory attracted sustained criticism. Graham Pyke's review for the Annual Review of Ecology and Systematics surveyed the growing number of tests and assessed the theory's assumptions and the objections raised against it [Pyke 1984]. In a deliberately provocative forum paper, Graham Pierce and John Ollason set out "eight reasons why optimal foraging theory is a complete waste of time" [Pierce & Ollason 1987]. They argued that optimisation theory is inappropriate for investigating the products of evolution, that animals should not be expected to be optimal, and that it is not possible to test whether they are. They also claimed that no optimisation model of foraging behaviour had been supported [Pierce & Ollason 1987]. They stated that their aim was to provoke discussion, because criticisms of the theory were too often dismissed as "well known" without being refuted [Pierce & Ollason 1987].
Supporters responded that optimality models are tools for generating testable hypotheses about the decisions animals face, not proofs that natural selection produces perfection [Stephens & Krebs 1986]. When predictions fail, as with the partial preferences of Krebs's great tits, the failure points to a missing constraint or a different currency, such as the time needed to recognise prey [Krebs et al. 1977; Stephens & Krebs 1986]. Risk-sensitivity and predation trade-offs are examples of this process: each began as a departure from simple rate-maximising predictions and became a model in its own right [Caraco et al. 1980; Lima & Dill 1990]. Other findings still challenge simple rate maximisation. Many animals will work for food even when identical free food is available, a phenomenon called contrafreeloading. Ian Inglis and colleagues reviewed this behaviour and argued that it may help animals gather information about food sources that could become more valuable in the future [Inglis et al. 1997].
Movement Ecology and the Lévy-Flight Debate
Miniaturised tracking devices have allowed researchers to ask how animals search when they do not know where food is. In 1996, Gandhimohan Viswanathan and colleagues analysed data from activity loggers on wandering albatrosses (Diomedea exulans) and reported that the birds' flight durations followed a power-law distribution. That pattern is characteristic of a Lévy flight, a random search made up of many short moves and rare very long ones, which in theory can be efficient when prey is sparse and patchily distributed [Viswanathan et al. 1996].
In 2007, Andrew Edwards and colleagues, including several authors of the original study, revisited the result. Using a new, high-resolution data set of wandering albatross flights, they found no evidence for Lévy flight behaviour [Edwards et al. 2007]. Re-examining the original data with additional information, they showed that the very long "flights" essential to the Lévy pattern were spurious: the loggers were dry, and therefore recorded as flying, while birds were on land at their breeding colony. The same paper also found that published claims of Lévy flights in bumblebees and deer were not supported on reanalysis [Edwards et al. 2007].
The debate did not end there. Nicolas Humphries and colleagues analysed movement data from 14 species of open-ocean predatory fish, including sharks, tuna and marlin, using statistical methods designed to avoid earlier errors [Humphries et al. 2010]. Lévy-like movement was associated with less productive waters, where prey is sparse, while Brownian (simple random) movement was associated with productive habitats such as shelf waters and oceanic fronts. Some individuals switched between the two as they moved between habitat types [Humphries et al. 2010]. The albatross correction shows how data errors can shape an influential idea. The marine predator results suggest that search patterns depend on the environment, as foraging theory would predict, rather than being fixed properties of a species [Edwards et al. 2007; Humphries et al. 2010].
Conservation and Welfare Relevance
Foraging theory gives conservation managers ways to measure how animals experience their environment, not just where they occur. William Sutherland argued that behavioural studies, including studies of how animals forage and respond to one another, can help predict how populations will respond to environmental change such as habitat loss, and that behavioural ecology deserves a larger role in conservation biology [Sutherland 1998]. Because foraging decisions respond quickly to changes in food supply and danger, they can reveal problems before population numbers change.
Giving-up densities are one of the most widely used tools of this kind. Because a GUD reflects a forager's perceived costs of feeding in a patch, comparing GUDs across sites or treatments can show which habitats animals treat as safe and profitable [Brown 1988]. A practical guide by Miguel Bedoya-Perez and colleagues reviewed how the method has been applied and set out the assumptions and design choices, such as the food and substrate used and the identity of the forager, that must be considered for results to be interpreted reliably [Bedoya-Perez et al. 2013]. The predation-risk framework also reminds managers that food availability alone does not determine habitat quality: animals may avoid food-rich areas that they perceive as dangerous [Lima & Dill 1990; Brown 1988].
Foraging research also informs the care of captive animals. In the wild, animals spend much of their time searching for and processing food, whereas captive animals are often fed from a dish. The widespread tendency of animals to work for food even when free food is available suggests that foraging activity itself can matter to them, independently of the food obtained [Inglis et al. 1997]. This finding supports feeding practices in zoos, sanctuaries and laboratories that give animals opportunities to search for and extract food. Since contrafreeloading declines as hunger increases and as the effort required rises, such feeding challenges should be set at an appropriate level of difficulty [Inglis et al. 1997].
References
[Bedoya-Perez et al. 2013] Bedoya-Perez, M.A., Carthey, A.J.R., Mella, V.S.A., McArthur, C. & Banks, P.B. (2013). A practical guide to avoid giving up on giving-up densities. Behavioral Ecology and Sociobiology, 67(10), 1541–1553. https://doi.org/10.1007/s00265-013-1609-3
[Brown 1988] Brown, J.S. (1988). Patch use as an indicator of habitat preference, predation risk, and competition. Behavioral Ecology and Sociobiology, 22(1), 37–47. https://doi.org/10.1007/BF00395696
[Caraco et al. 1980] Caraco, T., Martindale, S. & Whittam, T.S. (1980). An empirical demonstration of risk-sensitive foraging preferences. Animal Behaviour, 28(3), 820–830. https://doi.org/10.1016/S0003-3472(80)80142-4
[Charnov 1976] Charnov, E.L. (1976). Optimal foraging, the marginal value theorem. Theoretical Population Biology, 9(2), 129–136. https://doi.org/10.1016/0040-5809(76)90040-X
[Cowie 1977] Cowie, R.J. (1977). Optimal foraging in great tits (Parus major). Nature, 268(5616), 137–139. https://doi.org/10.1038/268137a0
[Edwards et al. 2007] Edwards, A.M., Phillips, R.A., Watkins, N.W., Freeman, M.P., Murphy, E.J., Afanasyev, V., Buldyrev, S.V., da Luz, M.G.E., Raposo, E.P., Stanley, H.E. & Viswanathan, G.M. (2007). Revisiting Lévy flight search patterns of wandering albatrosses, bumblebees and deer. Nature, 449(7165), 1044–1048. https://doi.org/10.1038/nature06199
[Emlen 1966] Emlen, J.M. (1966). The role of time and energy in food preference. The American Naturalist, 100(916), 611–617. https://doi.org/10.1086/282455
[Humphries et al. 2010] Humphries, N.E., Queiroz, N., Dyer, J.R.M., Pade, N.G., Musyl, M.K., Schaefer, K.M., Fuller, D.W., Brunnschweiler, J.M., Doyle, T.K., Houghton, J.D.R., Hays, G.C., Jones, C.S., Noble, L.R., Wearmouth, V.J., Southall, E.J. & Sims, D.W. (2010). Environmental context explains Lévy and Brownian movement patterns of marine predators. Nature, 465(7301), 1066–1069. https://doi.org/10.1038/nature09116
[Inglis et al. 1997] Inglis, I.R., Forkman, B. & Lazarus, J. (1997). Free food or earned food? A review and fuzzy model of contrafreeloading. Animal Behaviour, 53(6), 1171–1191. https://doi.org/10.1006/anbe.1996.0320
[Kacelnik 1984] Kacelnik, A. (1984). Central place foraging in starlings (Sturnus vulgaris). I. Patch residence time. Journal of Animal Ecology, 53(1), 283–299. https://doi.org/10.2307/4357
[Krebs et al. 1977] Krebs, J.R., Erichsen, J.T., Webber, M.I. & Charnov, E.L. (1977). Optimal prey selection in the great tit (Parus major). Animal Behaviour, 25, 30–38. https://doi.org/10.1016/0003-3472(77)90064-1
[Lima & Dill 1990] Lima, S.L. & Dill, L.M. (1990). Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology, 68(4), 619–640. https://doi.org/10.1139/z90-092
[MacArthur & Pianka 1966] MacArthur, R.H. & Pianka, E.R. (1966). On optimal use of a patchy environment. The American Naturalist, 100(916), 603–609. https://doi.org/10.1086/282454
[Orians & Pearson 1979] Orians, G.H. & Pearson, N.E. (1979). On the theory of central place foraging. In D.J. Horn, G.R. Stairs & R.D. Mitchell (Eds.), Analysis of Ecological Systems (pp. 155–177). Columbus: Ohio State University Press. (Book chapter; no DOI.)
[Pierce & Ollason 1987] Pierce, G.J. & Ollason, J.G. (1987). Eight reasons why optimal foraging theory is a complete waste of time. Oikos, 49(1), 111–117. https://doi.org/10.2307/3565560
[Pyke 1984] Pyke, G.H. (1984). Optimal foraging theory: a critical review. Annual Review of Ecology and Systematics, 15, 523–575. https://doi.org/10.1146/annurev.es.15.110184.002515
[Stephens 1981] Stephens, D.W. (1981). The logic of risk-sensitive foraging preferences. Animal Behaviour, 29(2), 628–629. https://doi.org/10.1016/S0003-3472(81)80128-5
[Stephens & Krebs 1986] Stephens, D.W. & Krebs, J.R. (1986). Foraging Theory. Princeton, NJ: Princeton University Press. (Monographs in Behavior and Ecology; book, cited without DOI.)
[Sutherland 1998] Sutherland, W.J. (1998). The importance of behavioural studies in conservation biology. Animal Behaviour, 56(4), 801–809. https://doi.org/10.1006/anbe.1998.0896
[Viswanathan et al. 1996] Viswanathan, G.M., Afanasyev, V., Buldyrev, S.V., Murphy, E.J., Prince, P.A. & Stanley, H.E. (1996). Lévy flight search patterns of wandering albatrosses. Nature, 381(6581), 413–415. https://doi.org/10.1038/381413a0
[Werner & Hall 1974] Werner, E.E. & Hall, D.J. (1974). Optimal foraging and the size selection of prey by the bluegill sunfish (Lepomis macrochirus). Ecology, 55(5), 1042–1052. https://doi.org/10.2307/1940354