Colin R. Tosh

The Confusion Effect in Predatory Neural Networks

A simple artificial neural network model of image reconstruction in sensory maps is presented to explain the difficulty predators experience in targeting prey in large groups (the confusion effect). Networks are trained to reconstruct multiple randomly conformed “retinal” images of prey groups in an internal spatial map of their immediate environment. They are then used to simulate prey targeting by predators on groups of specific conformation. Networks trained with the biologically plausible associative reward‐penalty method produce a more realistic model of the confusion effect than those trained with the popular but biologically implausible backpropagation method. The associative reward‐penalty model makes the novel prediction that the accuracy–group size relationship is U shaped, and this prediction is confirmed by empirical data gathered from interactive computer simulation experiments with humans as “predators.” The model further predicts all factors known from previous empirical work (and most factors suspected) to alleviate the confusion effect: increased relative intensity of the target object, heterogeneity of group composition, and isolation of the target. Interestingly, group compaction per se is not predicted to worsen predator confusion. This study indicates that the relatively simple, nonattentional mechanism of information degradation in the sensory mapping process is potentially important in generating the confusion effect.

Behavioural consequences of sensory plasticity in guppies

Sensory plasticity, whereby individuals compensate for sensory deprivation in one sense by an improvement in the performance of an alternative sense, is a well-documented phenomenon in nature. Despite this, the behavioural and ecological consequences of sensory plasticity have not been addressed. Here we show experimentally that some components (vision and chemoreception) of the sensory system of guppies are developmentally plastic, and that this plasticity has important consequences for foraging behaviour. Guppies reared under low light conditions had a significantly stronger response to chemical food cues encountered in isolation than fish reared at higher light levels. Conversely, they exhibited a weaker response to visual-only cues. When visual and olfactory/gustatory cues were presented together, no difference between the strength of response for fish reared at different light intensities was evident. Our data suggest that guppies can compensate for experience of a visually poor, low light environment via a sensory switch from vision to olfaction/gustation. This switch from sight to chemoreception may allow individuals to carry out the foraging behaviour that is essential to their survival in a visually poor environment. These considerations are especially important given the increasing frequency of anthropogenic changes to ecosystems. Compensatory phenotypic plasticity as demonstrated by our study may provide a hitherto unconsidered buffer that could allow animals to perform fundamental behaviours in the face of considerable change to the sensory environment.

The significance of a facultative bacterium to natural populations of the pea aphid Acyrthosiphon pisum

Abstract. 1.  Laboratory studies have implicated various accessory bacteria of aphids as important determinants of aphid performance, especially on certain plant species and under certain thermal regimes. One of these accessory bacteria is PABS (also known as T‐type), which is distributed widely but is not universal in natural populations of the pea aphid Acyrthosiphon pisum in the U.K.

Theoretical predictions strongly support decision accuracy as a major driver of ecological specialization

We examine the proposal that the high levels of ecological specialization seen in many animals has been driven by benefits in decision accuracy that accrue from this resource-use strategy. Using artificial analogs of real neural processing (artificial neural networks), we examine the relationship between decision accuracy, level of ecological specialization/generalization, and the punishment/reward for selecting non-host resources. We demonstrate that specialists make more accurate resource-use decisions than generalists when the consequences of using a non-host are neutral or positive but not very positive. Pronounced unsuitability of non-host resources in fact promotes higher decision accuracy in generalists. These unusual predictions can be explained by the special properties of neural processing systems and are entirely consistent with patterns of performance of many specialists in nature, where non-used resources are, curiously, often quite suitable for growth and reproduction. They potentially reconcile the long-observed discrepancy between the presence of high levels of ecological specialization in many animal groups and the absence of strong negative fitness correlations across resources. The strong theoretical support obtained here, and the equally good support in experimental studies elsewhere, should bring the “neural limitations” hypothesis to the forefront of research on the evolutionary determinants of ecological range.

The ultrasonic mating signal of the male lesser wax moth

Abstract. Male lesser wax moths, Achroia grisella (Fabricius) (Lepidoptera: Pyralidae: Galleriinae), produce both a pheromone and an ultrasonic acoustic signal that function in mate attraction. We describe the structure of the acoustic signal, in particular the interpulse intervals and the spectral properties of the pulses. The song consists of a train of ultrasonic pulses. The interpulse interval is usually bimodally distributed, but can sometimes be unimodal. This reflects variation in the duration of the up and down wing strokes. The pulses are also usually paired which can produce multimodality of the interpulse intervals. These paired pulses probably reflect wingbeat asynchrony because they are not found in males in which the signalling capability of one wings sound producing structure is abolished.

On the mechanistic basis of plant affiliation in the black bean aphid (Aphis fabae) species complex

A number of aphid species comprise multiple taxa (biotypes, races, subspecies etc.) with distinct but often overlapping plant ranges. These species complexes are of considerable evolutionary interest, with a long-standing debate as to whether they may have evolved sympatrically through disruptive selection arising from trade offs in the capacity to utilise different plants (e.g., Futuyma & Mayer, 1980; Diehl & Bush, 1989; Via, 1991; Bush, 1994). However, little is known about the behavioural or physiological processes that may underpin the negative correlations between the capacity of closely-related aphids to utilise different plant species [but see Wilkinson & Douglas (1998) and Caillaud (1999)]. This study concerns the black-bean aphid Aphis fabae Scop., a heterecious aphid with ca. 80 summer (secondary) hosts and 1–3 winter (primary) hosts. The differential capacity of the four recognised subspecies (A. f. cirsicanthoides, A. f. fabae, A. f. mordwilkoi and A. f. solani) to utilise secondary hosts may have selected for assortative mating of the sexual generation on the primary hosts (Mackenzie & Guldemond, 1994), and a negative correlation has been demonstrated in aphid utilization of Vicia faba and Tropaeolum majus, specific host plants of A. f. fabae and A. f. mordwilkoi respectively (Mackenzie, 1996). However, this trade off is not symmetrical (Douglas, 1997): A. f. fabae perform more poorly on T. majus than A. f. mordwilkoi on V. faba, and it was suggested that T. majus may be unsuitable for feeding by A. f. fabae. The purpose of this study was to use the electrical monitoring technique (EPG) of Tjallingii (1988) to test the prediction that the capacity of aphids to penetrate to the sieve elements and exhibit sustained feeding on phloem sap is reduced for A. f. fabae on T. majus, but not for A. f. mordwilkoi on V. faba.

Which conditions promote negative density dependent selection on prey aggregations

Negative density dependent selection on individuals in prey aggregations (negative DDS, the preferential selection by predators of spatially isolated prey) is assumed to contribute in many cases to the evolution and maintenance of aggregation. Both positive and negative DDS on prey groups have been documented in nature but there is no existing framework to predict when each of these forms of natural selection is most likely. By exploiting the tendency of artificial neural networks to exhibit consumer-like emergent behaviours, I isolate at least two environmental factors impinging on the consumer organism that may determine which form of density dependent natural selection is shown: the distribution of prey group size attacked by the predator and the spatial conformation (dispersed or compacted) of the prey group. Numerous forms of DDS on artificial prey (positive, negative, and non-DDS) are displayed through different combinations of these factors. I discuss in detail how the predictions of the model may be tested by empiricists in order to assess the usefulness of the framework presented. I stress the importance of understanding DDS on prey groups given the recent emergence of these systems as test beds for ideas on biological self-organisation.

Basic features, conjunctive searches, and the confusion effect in predator–prey interactions

The confusion effect describes the observed decrease in the likelihood that a predator will successfully catch any prey when attacking larger groups of moving prey. We introduce readers to the work of cognitive psychologists interested in human visual attention who have been studying their own version of the confusion effect for many years, developing methods and concepts that may be of fundamental utility to behavioral ecologists. In psychology, ‘basic features’ are characteristics unique to a target object in the visual field that no distracter objects share. Images containing targets with basic features are often less likely to induce the confusion effect in human subjects. Target objects with conjunctions of features, on the other hand, have no individual characteristics unique from distracters, but unique characteristics in combination. Such targets more often induce the confusion effect in humans. We propose the ‘basic feature’ (vs. conjunctions of features) as a new organizing concept for studies on the occurrence of the confusion effect in nature, potentially allowing predictions about which types of prey groups are likely to induce the confusion effect in predators.

Individuals from different-looking animal species may group together to confuse shared predators: simulations with artificial neural networks

Individuals of many quite distantly related animal species find each other attractive and stay together for long periods in groups. We present a mechanism for mixed-species grouping in which individuals from different-looking prey species come together because the appearance of the mixed-species group is visually confusing to shared predators. Using an artificial neural network model of retinotopic mapping in predators, we train networks on random projections of single- and mixed-species prey groups and then test the ability of networks to reconstruct individual prey items from mixed-species groups in a retinotopic map. Over the majority of parameter space, cryptic prey items benefit from association with conspicuous prey because this particular visual combination worsens predator targeting of cryptic individuals. However, this benefit is not mutual as conspicuous prey tends to be targeted most poorly when in same-species groups. Many real mixed-species groups show the asymmetry in willingness to initiate and maintain the relationship predicted by our study. The agreement of model predictions with published empirical work, the efficacy of our modelling approach in previous studies, and the taxonomic ubiquity of retinotopic maps indicate that we may have uncovered an important, generic selective agent in the evolution of mixed-species grouping.

The need for stochastic replication of ecological neural networks

Artificial neural networks are becoming increasingly popular as predictive statistical tools in ecosystem ecology and as models of signal processing in behavioural and evolutionary ecology. We demonstrate here that a commonly used network in ecology, the three-layer feed-forward network, trained with the backpropagation algorithm, can be extremely sensitive to the stochastic variation in training data that results from random sampling of the same underlying statistical distribution, with networks converging to several distinct predictive states. Using a random walk procedure to sample error–weight space, and Sammon dimensional reduction of weight arrays, we demonstrate that these different predictive states are not artefactual, due to local minima, but lie at the base of major error troughs in the error–weight surface. We further demonstrate that various gross weight compositions can produce the same predictive state, suggesting the analogy of weight space as a ‘patchwork’ of multiple predictive states. Our results argue for increased inclusion of stochastic training replication and analysis into ecological and behavioural applications of artificial neural networks.