Optimal Foraging Theory

What optimal foraging theory says

Optimal foraging theory (OFT) treats eating as an economics problem. An animal looking for food has limited time and that time is valuable — every second spent searching for or processing one meal is a second not spent finding another, defending a territory, courting a mate, or watching for predators. So the theory asks a single sharp question: of all the foraging decisions an animal could make, which one returns the most energy for the least time? The prediction is that natural selection has shaped behavior to maximize the net rate of energy intake, written E/T — joules gained, minus joules spent foraging, divided by the total time the foraging takes.

The bet behind the theory is that intake rate is a good proxy for Darwinian fitness. A parent bird that delivers more joules per hour to its brood fledges more chicks. A squirrel that meets its winter caloric budget by October has weeks of margin against a hard freeze. Because the link between energy and survival is so direct, foraging is one of the few behaviors where you can write down a quantitative prediction from first principles and then go watch an animal either obey it or break it. That falsifiability is what made OFT, born in the mid-1960s, one of the founding pillars of behavioral ecology.

OFT splits the foraging problem into two classic sub-problems, each with its own model. The prey-choice (diet-breadth) problem asks: when you encounter a food item, should you eat it or keep searching for something better? The patch-residence problem asks: when food comes in clumps (a flower patch, a leaf litter pile, a mussel bed) that you deplete as you feed, when should you give up and travel to a fresh clump? The first is solved by the prey-choice model; the second by the marginal value theorem.

Profitability: energy divided by handling time

The central quantity in prey choice is profitability — the energy you get from an item divided by the time it takes to handle it once found. Foraging time splits cleanly into two phases. Search time is the time spent looking for the next item; it depends on how abundant the prey is, not on what kind it is. Handling time (h) is everything after encounter — pursuing, subduing, cracking, chewing, swallowing — and it is a fixed property of that prey type. Profitability is energy content E divided by handling time h, measured in joules per second.

This single ratio explains why bigger is not always better. A shore crab can extract more flesh from a large mussel, but a large mussel's shell is thicker and takes disproportionately longer to crack, so handling time climbs faster than energy. Plot E/h against prey size and you get a hump: profitability rises with size, peaks at an intermediate size, then falls as the shell-cracking penalty takes over. The optimal forager should prefer prey near that profitability peak — and shore crabs do exactly that, preferring mussels around 2–2.5 cm rather than the largest available.

Crucially, profitability ranks prey by E/h alone — abundance does not enter the ranking. A rare but highly profitable prey type still tops the list. What abundance controls is whether lower-ranked types make the cut at all, which is the surprising heart of the prey-choice model.

The prey-choice model and the zero-one rule

MacArthur and Pianka (1966) imagined a forager that encounters prey one at a time while searching. Rank the prey types by profitability so type 1 is best. The optimal rule is to always eat type 1 when you find it. Then add type 2 to the diet only if type 2's profitability, E₂/h₂, is greater than the average intake rate you would earn by skipping type 2 and continuing to search for type 1. Add type 3 only if its profitability beats the average rate from feeding on types 1 and 2 together. Each new type is included only while it raises the overall intake rate.

Two consequences are famous and counter-intuitive:

• The zero-one rule. The model predicts no partial preferences. A prey type is either inside the optimal diet (eaten every time it is encountered) or outside it (always ignored). There is no "eat it 40% of the time" — the boundary is razor-sharp.
• Inclusion depends on the abundance of better prey, not the item itself. Whether a low-quality item is worth eating is decided entirely by how often you bump into high-quality prey. If profitable prey are common, search time for them is short, your background intake rate is high, and you should ignore junk food no matter how much of it litters the ground. If profitable prey become rare, the threshold drops and the junk food is suddenly worth taking. The unprofitable type's own density is irrelevant to the decision.

That second prediction is the model's signature test: experimentally flood the environment with low-quality prey and the forager should not broaden its diet, but make high-quality prey scarce and it should. John Krebs and colleagues confirmed this in great tits (Parus major) in 1977 using a conveyor belt that delivered large and small mealworm pieces at controlled rates — when profitable large pieces were frequent, the birds ignored the small ones; drop the rate of large pieces and the birds accepted everything, with the switch happening near the predicted threshold.

The marginal value theorem: when to leave a patch

Food often comes in patches you deplete as you exploit them — a bumblebee draining the florets of one inflorescence, a thrush working over a leaf-litter pile. Inside a patch, the cumulative energy you have gained rises steeply at first (the easy prey go first) then flattens as the patch empties. Eventually the patch is so depleted that you would do better to cut your losses and fly to a fresh one. But travel between patches costs time and yields nothing. So how long should you stay?

Charnov (1976) proved the elegant answer with the marginal value theorem (MVT). The optimal strategy is to leave a patch at the exact moment your marginal (instantaneous) intake rate inside it drops to the average intake rate for the whole habitat — travel time included. Stay any longer and you are earning less than the habitat average; leave any sooner and you abandon prey that were still beating the average. Graphically: mark the average travel time as a point to the left of the patch's start on the time axis, then draw the straight line from that point that just touches (is tangent to) the cumulative gain curve. The point where it touches is the optimal time to leave, and the slope of that tangent line is the maximum achievable habitat-wide intake rate.

The single most testable MVT prediction: when patches are farther apart, stay longer in each one. Longer travel time pushes the tangent point's anchor leftward, which flattens the tangent line and moves the touch-point to the right — a longer residence time and a lower giving-up rate. Cowie (1977) tested this on great tits foraging from sawdust-filled cups (artificial patches) with lids that imposed extra "travel time," and the birds lengthened their stays almost exactly as the MVT predicted.

Prey-choice model vs marginal value theorem

Property	Prey-choice (diet-breadth) model	Marginal value theorem (patch model)
Question answered	Eat this item or keep searching?	Stay in this patch or travel to a new one?
Originators	MacArthur & Pianka, 1966	Eric Charnov, 1976
Key variable	Profitability E ÷ h (joules per second)	Marginal intake rate vs habitat average
Decision rule	Zero-one: include type if E/h > current average rate	Leave when marginal rate = habitat average rate
Role of travel time	Enters via search time for ranked prey	Sets the tangent anchor; longer travel → longer stays
Signature prediction	Diet breadth set by abundance of good prey only	Farther patches → stay longer, lower giving-up rate
Classic test	Great tits on a mealworm conveyor (Krebs 1977)	Great tits in sawdust patches (Cowie 1977)
Surprising corollary	Junk food's own abundance is irrelevant	Optimal forager never fully empties a patch

The numbers: how the math plays out

Concrete figures make the logic land. Take a forager with two prey types. Type 1 yields E₁ = 20 J and takes h₁ = 2 s to handle, so its profitability is 10 J/s. Type 2 yields E₂ = 8 J and takes h₂ = 4 s, so its profitability is 2 J/s. Suppose searching turns up a type-1 item on average every 5 s. Then ignoring type 2 entirely gives an intake rate of 20 J ÷ (5 s search + 2 s handle) ≈ 2.86 J/s. Because type 2's profitability (2 J/s) is below that 2.86 J/s threshold, the forager should ignore type 2 — even if type 2 carpets the ground. Now make type 1 scarce so it takes 30 s to find: the rate from specializing falls to 20 J ÷ (30 + 2) ≈ 0.63 J/s, now below type 2's 2 J/s, so the forager should switch to taking both. The diet expands not because type 2 changed, but because type 1 got rare.

Real-world magnitudes:

• Shore crabs (Elner & Hughes 1978). Preferred mussel size ≈ 2.0–2.5 cm, matching the predicted profitability peak; the largest mussels were taken far less than their availability would predict because shell-cracking time scaled up faster than flesh content.
• Starlings provisioning chicks (Kacelnik 1984). Birds carried more crane-fly larvae per trip as the feeder was placed farther from the nest — load size climbed with travel distance along the curve the central-place model predicts.
• Bluegill sunfish (Werner & Hall 1974). At low prey density fish ate all Daphnia sizes; at high density they became selective for the largest, most profitable size class — diet breadth contracting as good prey grew abundant, exactly the prey-choice prediction.
• Oystercatchers and worm-handling. Birds adjust prey size to balance energy against handling time and the risk of having prey stolen (kleptoparasitism), a real-world wrinkle the bare model leaves out.
• Energy budgets. A small bird like a great tit may need to consume on the order of 30% of its body mass in food per day in winter, which is exactly why a few percent improvement in foraging efficiency is under strong selection.

Where it shows up: from bees to humans

• Pollinators and patch-leaving. Bumblebees deplete an inflorescence's nectar as they probe each floret, then leave for a fresh plant — their giving-up behavior tracks the marginal value theorem, and they leave sooner when the next plant is close. This shapes plant–pollinator coevolution and flower spacing.
• Risk-sensitive foraging. When animals face starvation, they shift between energy-maximizing and variance-sensitive strategies. A bird below its nightly survival threshold will gamble on a risky high-variance patch; one above threshold plays it safe. This extends OFT from mean-rate maximization to whole reward distributions.
• Predation risk trade-offs. The richest patch is often the most dangerous. Bluegill and many small mammals trade some intake rate for safety, foraging in poorer but safer cover — the "ecology of fear" layered on top of pure energy maximization.
• Human behavioral ecology. The diet-breadth model has been applied to hunter-gatherer subsistence: optimal-foraging analysis predicts which prey foragers pursue versus ignore, and explains diet shifts (the "broad-spectrum revolution") when high-ranked large game became scarce in the late Pleistocene.
• Information foraging in cognition. Computer scientists Pirolli and Card adapted patch-leaving theory to how people search the web: a reader "forages" through a page, then "leaves the patch" for a new link when the marginal information rate drops — the MVT, repurposed for clicks instead of calories.
• Conservation and management. Predicting how foragers redistribute when habitat is lost (giving-up densities, the level of remaining food at which animals abandon a patch) is a practical tool for assessing habitat quality and the impact of disturbance.

Common misconceptions

• "Animals maximize total energy eaten." No — they are predicted to maximize the rate, energy per unit time. An animal that gorged on the largest, slowest-to-handle prey could eat more total calories per meal yet earn fewer joules per hour, which is what counts when time is the scarce resource.
• "More abundant junk food means a broader diet." The prey-choice model says the opposite can hold: whether to include a low-ranked item depends only on the abundance of higher-ranked prey. Flooding the world with poor prey does not, by itself, make a forager start eating it.
• "The optimal forager strips each patch bare." The marginal value theorem predicts you should leave a patch while it still holds food — at the point its marginal rate has merely dropped to the habitat average. A perfectly emptied patch means you stayed too long.
• "Energy is always the right currency." Nectar-feeders may be nitrogen-limited, herbivores protein- or toxin-limited, and parents provisioning young maximize delivery rate to the nest, not self-feeding rate. Picking the wrong currency is the most common way OFT predictions fail — and the most common criticism of the theory.
• "Deviations disprove the theory." Real animals show partial preferences because they sample to track changing densities, face uncertainty, and trade food against predation risk. These are refinements, not refutations — the qualitative predictions (prefer peak-profitability prey, stay longer when patches are far apart) hold across taxa from bees to crabs to birds.
• "Optimal means the animal is consciously calculating." Nothing is being computed deliberately. The "optimization" is what natural selection produces over generations; the equations describe the outcome of selection on simple behavioral rules, not a forager doing calculus in its head.