Optimal Foraging — How Animals Decide What Is Worth Pursuing

An animal foraging for food faces a series of implicit calculations. Which prey to pursue. How long to stay in a patch before moving on. When a smaller, easier meal is better than waiting for a larger one. These decisions shape survival, reproduction, and the structure of entire ecosystems.

Optimal foraging theory provides a mathematical framework for understanding these choices. It emerged in the 1970s as behavioral ecologists began treating animal foraging as an economic problem: maximize net energy intake per unit time, subject to the constraints of the environment and the body. The predictions turned out to be surprisingly precise — and their limitations just as instructive.

The currency of foraging

The theory starts with a definitional choice: what is the forager trying to optimize? Most models use net energy gain per unit time as the currency. The intuition is straightforward. An organism that acquires more energy per minute can allocate the surplus to growth, reproduction, or storage. Natural selection should favor individuals whose foraging behavior approaches the optimum.

But the currency matters. Different environments select for different objectives. Animals in energy-poor habitats may maximize intake rate. Animals facing predation risk may trade off intake rate against safety. Animals with limited digestive capacity may prioritize food quality over quantity. The choice of currency determines the shape of the prediction.

The diet breadth model

The simplest and most influential model is the optimal diet model, also called the prey choice or diet breadth model. It was developed by Eric Charnov in 1976, building on earlier work by MacArthur and Pianka.

The model considers a forager encountering prey types ranked by profitability — defined as energy content divided by handling time. Handling time includes capture, processing, and consumption. The central prediction is a threshold rule: when high-profitability prey are abundant enough, the forager should ignore lower-profitability types entirely. It should only broaden its diet when encounter rates with preferred prey drop below a calculable threshold.

Formally, a forager should accept prey type $i$ only if its profitability $E_i / h_i$ exceeds the expected intake rate from searching for better prey alone. The result is a sharp prediction: diet breadth should expand as the abundance of preferred prey declines. The forager switches from specialist to generalist at a specific encounter rate, not gradually or based on prey quality alone.

Field tests with great tits (Parus major) by John Krebs and colleagues in the 1970s found broad support. Birds presented with beads of different sizes and colors — simulating prey with different energy contents and handling times — selectively ignored smaller beads when large ones were abundant, and accepted them when large beads became scarce. The switching point matched the model’s prediction.

Similar patterns appeared in oystercatchers foraging on mussels. Researchers Meire and Ervynck found that these birds prefer mussels 30 to 45 millimeters in diameter — not the largest available, but the ones that maximize energy per handling time. Initially, a simple profitability model predicted a preference for 50 to 55 millimeter mussels. Adding prey density to the calculation corrected the prediction, showing that encounter rate matters as much as individual profitability.

The marginal value theorem

The diet model assumes the forager moves between discrete encounters. Many real foragers work in patches — a flower cluster, a clump of insects, a berry bush — where resources are concentrated but deplete over time. Each additional minute spent in a patch yields less than the previous one. At some point, leaving and traveling to a new patch becomes more profitable.

Eric Charnov’s 1976 marginal value theorem formalized this trade-off. The prediction is elegant: a forager should leave a patch when its instantaneous rate of gain drops to the average rate available across the entire habitat, including travel time between patches.

Graphically, the gain from a patch follows a diminishing-returns curve — steep at first, then flattening. The optimal departure point is where a tangent line from the travel-time intercept touches that curve. Longer travel times push the tangent point further along the curve, predicting longer stays. Shorter travel times predict earlier departures.

The theorem has been tested across diverse taxa with consistent results:

• European starlings (Sturnus vulgaris) foraging for mealworms in experimental patches left at rates matching the model’s prediction. When Kacelnik and colleagues increased the travel time between patches, the birds stayed longer and carried larger loads — exactly as the marginal value theorem predicted.

• Great tits foraging in artificial patches with depleting food supplies departed closer to the optimal time than would be expected by random variation.

• Even dung flies choosing copulation duration — a non-foraging context — followed the pattern, leaving a mate when the marginal gain dropped below the habitat average.

The breadth of these matches is what made the marginal value theorem one of the most cited results in behavioral ecology. A single geometric argument predicted departure timing across birds, insects, and mating behavior.

What the models get wrong

Optimal foraging theory has faced sustained criticism, and the criticisms illuminate the gap between mathematical optimality and biological reality.

Perfect design assumption. The models assume natural selection produces behaviors that match the mathematical optimum. Real organisms inherit constraints — morphological, cognitive, and developmental — that prevent perfect optimization. A bird’s beak shape limits which prey it can handle. A fish’s visual system constrains what it can detect. These are not design choices; they are inherited starting conditions.

Predation risk. The classic models optimize energy intake without accounting for danger. Animals foraging in open terrain face higher predation risk than those in cover. Many species trade off intake rate against safety, accepting lower energy gains to reduce exposure. Researchers later extended the models to include risk, but the extensions multiplied the number of free parameters, weakening predictive power.

Nutrient balancing. Energy is not the only currency. Animals need proteins, fats, minerals, and vitamins in specific proportions. A high-energy food that lacks essential nutrients is not optimal. The two-dimensional optimal foraging model by Cooper and Lumey (1980) added protein constraints, showing that nutrient needs can override pure energy maximization. More recent work using geometric frameworks has shown that animals often balance multiple nutritional axes simultaneously.

Cognitive limits. The models assume perfect information about encounter rates, handling times, and patch quality. Real foragers estimate these quantities from experience, with error. Some use simple rules of thumb — like leaving a patch after a fixed number of empty encounters — that approximate the optimum without requiring explicit calculation.

Lifetime vs. instantaneous optimization. Kacelnik’s work with honeybees revealed a surprising deviation. Unlike starlings, which maximize intake rate, bees appear to maximize energy efficiency — energy gained per energy spent. Heavy nectar loads shorten individual lifespan, but the colony benefits from efficient foraging. The unit of selection (individual vs. colony) changes the optimal strategy.

Why the framework endures

Despite its limitations, optimal foraging theory remains a foundational tool in behavioral ecology. Not because the models are perfectly accurate, but because they establish a null hypothesis. When observed behavior deviates from the prediction, the deviation points to something the model missed — predation risk, nutrient constraints, cognitive limits, or conflicting selective pressures.

The marginal value theorem has found applications beyond foraging. Researchers have used it to model human decision-making in foraging contexts, animal dispersal, mate choice, and even the timing of medical interventions. The core structure — leave when the marginal gain drops below the alternative — is a general principle about resource allocation under diminishing returns.

The theory’s lasting value lies in its clarity. It takes a messy biological problem and reduces it to testable predictions with specific numerical thresholds. When those predictions fail, the failure is informative. It tells us what forces the simple model omitted, and points toward a more complete account of how organisms navigate a world of limited resources and competing demands.

Sources

• Charnov, E. L. (1976). “Optimal foraging: the marginal value theorem.” Theoretical Population Biology, 9(2), 129–136. DOI: 10.1016/0040-5809(76)90040-X

• MacArthur, R. H., & Pianka, E. R. (1966). “On optimal use of a patchy environment.” The American Naturalist, 100(915), 377–383. DOI: 10.1086/282450

• Krebs, J. R., Davies, N. B., & Taylor, P. M. (1978). “Optimal foraging: cursorial birds in patchy distributions.” Animal Behaviour, 26(3), 842–853. DOI: 10.1016/0003-3472(78)90069-9

• Kacelnik, A., Krebs, J. R., & Taylor, P. J. (1986). “Travel costs and optimal foraging in European starlings.” The American Naturalist, 127(6), 865–877. DOI: 10.1086/284493

• Meire, P., & Ervynck, T. (1985). “Optimal foraging in the oystercatcher (Haematopus ostralegus): prey-size selection and profitability.” Journal of Animal Ecology, 54(2), 339–350. DOI: 10.2307/4698

• Cooper, W. E., & Lumey, L. H. (1980). “Optimal foraging on two foods.” Theoretical Population Biology, 18(2), 203–221. DOI: 10.1016/0040-5809(80)90017-3