TREE vol. 4, no. 72, December 1989
IF MOST PREY ARE CRYPTIC, then how do predators improve their ability to detect them? The conventional answer, provided by L. Tinbergen in his classic posthumous paper on the exploitation of insect prey by woodland birds, is that predators adopt search(ing) images. Through chance encounters with prey of a given type, they gradually learn to recognize those stimuli that enable them to differentiate these prey from the surroundings. To quote Tinber- genz (p. 3321, this implies that the birds perform a highly selective siev- ing operation on the visual stimuli reaching their retina. This is an amaz- ing ability, especially in the complex environment of needles and twigs where the tits do most of their hunt- ing. Although this is by far Tin- bergens clearest statement that search images can only be acquired for prey that are cryptic, the modern definition of search image implies a perceptual change in the ability of individual predators to detect familiar cryptic prey (with the word familiar inserted to preclude perceptual changes in the acceptability of novel prey13.
Searching for Search Image John A. Allen
they spent scanning the background as they gained experience with cryp- tic artificial prey. Lawrence inter- preted this to mean that the birds were learning to distinguish the prey from their surroundings. The alterna- tive interpretation7 is that the birds started by making many short scans ofthe background, each of which was too brief to detect the cryptic prey; when experienced, their scans were longer but less frequent. These two interpretations could not be dis- tinguished by the data originally pre- sented by Lawrence*pg because he used an indirect method to measure the time spent scanning: from the total time a bird spent at a site he simply subtracted the time spent handling the prey, being vigilant and moving around.
Recent reanalysis of Lawrences original videotapes has proved equivocallO*ll. LawrencelO has now measured directly the scan times of each bird - as defined from the time the head was first stationary after swallowing to the start of the next prey strike -and has shown that this measure decreased significantly dur- ing the course of the first test, and between this test and a fourth repli- cate several days later. In response, Guilford and Dawkins have ques- tioned whether these scan times really were single scans or whether each was composed of a series of consecutive scans, during which the eyes moved to different parts of the background even though the head was stationary. Lawrences record- ings were not sufficiently com- prehensive to distinguish between these possibilities. This detail of methodology is crucial to the design of future tests of search image versus search rate; clearly, whatever species of visual predator is used, it is essen- tial to record eye movements in full.
has sprung up over the past two dec- ades: the hypothesis that predators maintain colour polymorphisms in their prey by concentrating on com- mon morphs 12. No such apostatic (frequency-dependent) selection is possible if predators detect all cryptic morphs within view. Yet, although detailed behavioural experiments have yet to be done, there is evidence that predators do often tend to over- look prey that differ from the ma- jority . l3 It is unclear whether or not search images in the strict sense are involved in this process14, but the presence of frequency dependence argues against the occurrence of search rate modification alone. Of course, as Guilford and Dawkins7 themselves point out, the two mech- anisms could work together. For example, if the perceptual system applies a degree of filtering and there is a decrease in search rate, then apostatic selection could still result.
Most of the experiments on search image and apostatic selection have used predators in abnormal con- ditions feeding on unnatural prey, and there remains the worry that the behaviour of the predators may be correspondingly artificial. Exper- iments with real predators feeding on real prey in natural conditions could yet confirm the search rate hypoth- esis - and if this happens we shall have to look to factors other than predation for the explanation of the maintenance of colour variation in cryptic prey species.
Over the years, there has been fre- quent criticism of Tinbergens in- terpretation of his data4, and of the term search image itself and the way it has been used in the Iiterature3p5,6. For some time it was regarded as an eminently testable hypothesis. Pred- ators that are acquiring search im- ages are expected to increase the rate at which they successfully detect their prey, and several studies - mostly with birds - have apparently demonstrated just this. It now ap- pears that these data can as easily be explained by an alternative hypoth- esis proposed by Guilford and Dawkins. Predators could increase their chances of detecting cryptic prey by simply slowing the rate at which they scan the environment, thus allowing more time for the image of an encountered prey to be registered by the visual system. Ani- mals that have learnt to detect cryptic prey in this way should scan each patch of the environment (irrespec- tive of whether it contains any prey) for longer than when they are failing to detect cryptic prey.
Because one model predicts an in- crease in search rate while the other predicts a decrease, the published data purporting to show search image formation could, at least in theory, be reanalysed for evidence of search rate modification. The work of Lawrenceegg, for example, apparently showed that free-ranging blackbirds (Turdus mew/a) shortened the time
John Allen is in the Dept of Biology, University of Southampton, Southampton SO9 3TU, UK.
0 1989 Elsewer Saence Publfshers Ltd (UKI 0169~534789602 00
Guilford and Dawkins also draw attention to a second method of dis- tinguishing between search image and search rate. Search rate modifi- cation implies that an increased abil- ity to detect one type of cryptic prey should increase the ability to detect other equally cryptic prey but not af- fect the ability to detect conspicuous prey. The search image hypothesis predicts neither of these: an in- creased ability to detect a cryptic prey should decrease the probability of detecting any other sort of prey.
These different predictions have potentially important evolutionary consequences to the prey. If search rate modification proves to be the only way in which cryptic prey are detected then it will jeopardize a whole area of ecological genetics that
References 1 Endler, J.A. (1978) Evol. Biol. 11, 319-363 2 Tinbergen, L. (1960) Arch. NBerl. Zoo/. 13,265-343 3 Dawkins, M. (1970) Anim. Bshav. 19, 566574 4 Royama, T. (1970) J. Anim. Ecol. 39, 619-668 5 Hinde, R.A. (I 970) Animal Behaviour (2nd edn), McGraw-Hill 6 Hollis, K.L. (1989) Anim. Behav. 36, 162-163 7 Guilford, T. and Dawkins, MS (1987) Anim. Behav. 35.1838-1845 8 Lawrence, E.S. (1985) Anim. Behav. 33, 929-937 9 Lawrence, E.S. (1985) Anim. Behav. 33, 1301-1309 10 Lawrence, E.S. (1989) Anim. Behav. 37,157-l 60 11 Guilford, T. and Dawkins, MS. (1989) Anim. Behav. 37,160-162 12 Clarke, B.C. (1962) in Taxonomy and Geography (Nicholas, D., ed.), pp. 47-70, Sysrematics Association (Publ. No. 4) 13 Allen, J.A. (1988) Philos. Trans. R. Sot. London Ser. B 319,486-503 14 Greenwood, J.J.D. (1984) Biol. J. Linn. Sot. 23,177-l 99