Does the Macaque Monkey Provide a Good Model for Studying Human Executive Control? A Comparative Behavioral Study of Task Switching

The ability to swiftly and smoothly switch from one task set to another is central to intelligent behavior, because it allows an organism to flexibly adapt to ever changing environmental conditions and internal needs. For this reason, researchers interested in executive control processes have often relied on task-switching paradigms as powerful tools to uncover the underlying cognitive and brain architecture. In order to gather fundamental information at the single-cell level, it would be greatly helpful to demonstrate that non-human primates, especially the macaque monkey, share with us similar behavioral manifestations of task-switching and therefore, in all likelihood, similar underlying brain mechanisms. Unfortunately, prior attempts have provided negative results (e.g., Stoet & Snyder, 2003b), in that it was reported that macaques do not show the typical signature of task-switching operations at the behavioral level, represented by switch costs. If confirmed, this would indicate that the macaque cannot be used as a model approach to explore human executive control mechanisms by means of task-switching paradigms. We have therefore decided to re-explore this issue, by conducting a comparative experiment on a group of human participants and two macaque monkeys, whereby we measured and compared performance costs linked to task switching and resistance to interference across the two species. Contrary to what previously reported, we found that both species display robust task switching costs, thus supporting the claim that macaque monkeys provide an exquisitely suitable model to study the brain mechanisms responsible for maintaining and switching task sets.     

Task Uncertainty Can Account for Mixing and Switch Costs in Task-Switching

Cognitive control is required in situations that involve uncertainty or change, such as when resolving conflict, selecting responses and switching tasks. Recently, it has been suggested that cognitive control can be conceptualised as a mechanism which prioritises goal-relevant information to deal with uncertainty. This hypothesis has been supported using a paradigm that requires conflict resolution. In this study, we examine whether cognitive control during task switching is also consistent with this notion. We used information theory to quantify the level of uncertainty in different trial types during a cued task-switching paradigm. We test the hypothesis that differences in uncertainty between task repeat and task switch trials can account for typical behavioural effects in task-switching. Increasing uncertainty was associated with less efficient performance (i.e., slower and less accurate), particularly on switch trials and trials that afford little opportunity for advance preparation. Interestingly, both mixing and switch costs were associated with a common episodic control process. These results support the notion that cognitive control may be conceptualised as an information processor that serves to resolve uncertainty in the environment.     

The Effects of Foreknowledge and Task-Set Shifting as Mirrored in Cue- and Target-Locked Event-Related Potentials

The present study examined the use of foreknowledge in a task-cueing protocol while manipulating sensory updating and executive control in both, informatively and non-informatively pre-cued trials. Foreknowledge, sensory updating (cue switch effects) and task-switching were orthogonally manipulated in order to address the question of whether, and to which extent, the sensory processing of cue changes can partly or totally explain the final task switch costs. Participants responded faster when they could prepare for the upcoming task and if no task-set updating was necessary. Sensory cue switches influenced cue-locked ERPs only when they contained conceptual information about the upcoming task: frontal P2 amplitudes were modulated by task-relevant cue changes, mid-parietal P3 amplitudes by the anticipatory updating of stimulus-response mappings, and P3 peak latencies were modulated by task switching. Task preparation was advantageous for efficient stimulus-response re-mapping at target-onset as mirrored in target N2 amplitudes. However, N2 peak latencies indicate that this process is faster for all repeat trials. The results provide evidence to support a very fast detection of task-relevance in sensory (cue) changes and argue against the view of task repetition benefits as secondary to purely perceptual repetition priming. Advanced preparation may have a stronger influence on behavioral performance and target-locked brain activity than the local effect of repeating or switching the task-set in the current trial.     

The Role of Task-Related Learned Representations in Explaining Asymmetries in Task Switching

Task switch costs often show an asymmetry, with switch costs being larger when switching from a difficult task to an easier task. This asymmetry has been explained by difficult tasks being represented more strongly and consequently requiring more inhibition prior to switching to the easier task. The present study shows that switch cost asymmetries observed in arithmetic tasks (addition vs. subtraction) do not depend on task difficulty: Switch costs of similar magnitudes were obtained when participants were presented with unsolvable pseudo-equations that did not differ in task difficulty. Further experiments showed that neither task switch costs nor switch cost asymmetries were due to perceptual factors (e.g., perceptual priming effects). These findings suggest that asymmetrical switch costs can be brought about by the association of some tasks with greater difficulty than others. Moreover, the finding that asymmetrical switch costs were observed (1) in the absence of a task switch proper and (2) without differences in task difficulty, suggests that present theories of task switch costs and switch cost asymmetries are in important ways incomplete and need to be modified.     

Neural mechanisms of transient and sustained cognitive control during task switching

A hybrid blocked and event-related functional magnetic resonance imaging (fMRI) study decomposed brain activity during task switching into sustained and transient components. Contrasting task-switching blocks against single-task blocks revealed sustained activation in right anterior prefrontal cortex (PFC). Contrasting task-switch trials against task-repeat and single-task trials revealed activation in left lateral PFC and left superior parietal cortex. In both sets of regions, activation dynamics were strongly modulated by trial-by-trial fluctuations in response speed. In addition, right anterior PFC activity selectively covaried with the magnitude of mixing cost (i.e., task-repeat versus single-task trial performance), and left superior parietal activity selectively covaried with the magnitude of the switching cost (i.e., task-switch versus task-repeat trial performance). These results indicate a functional double dissociation in brain regions supporting different components of cognitive control during task switching and suggest that both sustained and transient control processes mediate the behavioral performance costs of task switching.     

Foraging for Work and Age-Based Polyethism: The Roles of Age and Previous Experience on Task Choice in Ants

In social insects, colonies commonly show temporal polyethism in worker behavior, such that a worker follows a predictable pattern of changes between tasks as it ages. This pattern usually leads from workers first doing a safe task like brood care, to ending their lives doing the most dangerous tasks like foraging. Two mechanisms could potentially underlie this pattern: (1) age‐based task allocation, where the aging process itself predisposes workers to switch to more dangerous tasks; and (2) foraging for work, where ants switch to tasks that need doing from tasks which have too many associated workers. We tested the relative influence of these mechanisms by establishing nests of Camponotus floridanus with predetermined combinations of workers of known age and previous task specialization. The results supported both mechanisms. Nests composed of entirely brood‐tending workers had the oldest workers preferentially switching to foraging. However, in nests initially composed entirely of foragers, the final distribution of tenders and foragers was not different from random task‐switching and therefore supportive of foraging for work. Thus, it appears that in C. floridanus there is directionality to the mechanisms of task allocation. Switching to more dangerous tasks is age‐influenced, but switching to less dangerous tasks is age‐independent. The results also suggest that older workers are more flexible in their task choice behavior. Younger workers are more biased towards choosing within‐nest tasks. Finally, there are effects of previous experience that tend to keep ants in familiar tasks. Task allocation based on several mechanisms may balance between: (1) concentrating the most worn workers into the most dangerous tasks; (2) increasing task performance levels; and (3) maintaining behavioral flexibility to respond to demographic perturbations. The degree to which behavior is flexible may correlate to the frequency of such perturbations in a species.     

To what extent are task-switching deficits in children with attention-deficit/hyperactivity disorder independent of impaired inhibition?

Executive functions, higher-order cognitive functions needed for goal-directed behavior, have been studied extensively in the search for endophenotypes for ADHD, yet results have been inconclusive. We examine the performance of children with ADHD in task switching as an as yet understudied potential endophenotype. A group of 20 children with ADHD and a group of 23 children without ADHD (ages 7?C12) performed a task-switching paradigm and a Go/No-Go Task. Children with ADHD displayed significantly greater specific switch costs, that is, compared to control children they were especially impaired directly after task switches. There were no group differences with respect to the general switch costs, which are estimated by comparing performance on single task blocks to the block where both tasks are intermixed. Specific switch costs and Go/No-Go error rate were significantly correlated; yet, group differences in the task-switching paradigm remained significant even when inhibition was controlled for. This pattern of results suggests that children with ADHD are neither generally impaired in executive function nor only impaired with respect to inhibition. Instead, they display a highly specific deficit with regard to the flexible suppression and amplification of different task rules according to the context. Our conclusion that task switching has the potential to be added to the list of ADHD endophenotypes is strengthened by the independence of task-switching deficits and inhibition.     

Orientation Transfer in Vernier and Stereoacuity Training

Human performance on various visual tasks can be improved substantially via training. However, the enhancements are frequently specific to relatively low-level stimulus dimensions. While such specificity has often been thought to be indicative of a low-level neural locus of learning, recent research suggests that these same effects can be accounted for by changes in higher-level areas–in particular in the way higher-level areas read out information from lower-level areas in the service of highly practiced decisions. Here we contrast the degree of orientation transfer seen after training on two different tasks—vernier acuity and stereoacuity. Importantly, while the decision rule that could improve vernier acuity (i.e. a discriminant in the image plane) would not be transferable across orientations, the simplest rule that could be learned to solve the stereoacuity task (i.e. a discriminant in the depth plane) would be insensitive to changes in orientation. Thus, given a read-out hypothesis, more substantial transfer would be expected as a result of stereoacuity than vernier acuity training. To test this prediction, participants were trained (7500 total trials) on either a stereoacuity (N = 9) or vernier acuity (N = 7) task with the stimuli in either a vertical or horizontal configuration (balanced across participants). Following training, transfer to the untrained orientation was assessed. As predicted, evidence for relatively orientation specific learning was observed in vernier trained participants, while no evidence of specificity was observed in stereo trained participants. These results build upon the emerging view that perceptual learning (even very specific learning effects) may reflect changes in inferences made by high-level areas, rather than necessarily fully reflecting changes in the receptive field properties of low-level areas.     

The Effects of Aging on Conflict Detection

Several cognitive changes characterize normal aging; one change regards inhibitory processing and includes both conflict monitoring and response suppression. We attempted to segregate these two aspects within a Go/No-go task, investigating three age categories. Accuracy, response times and event-related potentials (ERPs) were recorded. The ERP data were analyzed, and the Go and No-go trials were separated; in addition, the trials were organized in repeat trials (in which the subjects repeated the action delivered in the previous trial) and switch trials (in which the subjects produced a response opposite to the previous response). We assumed that the switch trials conveyed more conflict than the repeat trials. In general, the behavioral data and slower P3 latencies confirmed the well-known age-related speed/accuracy trade-off. The novel analyses of the repeat vs. switch trials indicated that the age-related P3 slowing was significant only for the high conflict condition; the switch-P3 amplitude increased only in the two older groups. The ‘aging switch effect’ on the P3 component suggests a failure in the conflict conditions and likely contributes to a generalized dysfunction. The absence of either a switch effect in the young group and the P3 slowing in middle-aged group indicate that switching was not particularly demanding for these participants. The N2 component was less sensitive to the repeat/switch manipulation; however, the subtractive waves also enhanced the age effects in this earlier time window. The topographic maps showed other notable age effects: the frontal No-go N2 was nearly undetectable in the elderly; in the identical time window, a large activity in the posterior and prefrontal scalp regions was observed. Moreover, the prefrontal activity showed a negative correlation with false alarms. These results suggest that the frontal involvement during action suppression becomes progressively dysfunctional with aging, and additional activity was required to reach a good level of accuracy.     

Perceptual switch rates with ambiguous structure-from-motion figures in bipolar disorder

Slowing of the rate at which a rivalrous percept switches from one configuration to another has been suggested as a potential trait marker for bipolar disorder. We measured perceptual alternations for a bistable, rotating, structure-from-motion cylinder in bipolar and control participants. In a control task, binocular depth rendered the direction of cylinder rotation unambiguous to monitor participants' performance and attention during the experimental task. A particular direction of rotation was perceptually stable, on average, for 33.5s in participants without psychiatric diagnosis. Euthymic, bipolar participants showed a slightly slower rate of switching between the two percepts (percept duration 42.3s). Under a parametric analysis of the best-fitting model for individual participants, this difference was statistically significant. However, the variability within groups was high, so this difference in average switch rates was not big enough to serve as a trait marker for bipolar disorder. We also found that low-level visual capacities, such as stereo threshold, influence perceptual switch rates. We suggest that there is no single brain location responsible for perceptual switching in all different ambiguous figures and that perceptual switching is generated by the actions of local cortical circuitry.     

Frontal Lobe Contusion in Mice Chronically Impairs Prefrontal-Dependent Behavior

Traumatic brain injury (TBI) is a major cause of chronic disability in the world. Moderate to severe TBI often results in damage to the frontal lobe region and leads to cognitive, emotional, and social behavioral sequelae that negatively affect quality of life. More specifically, TBI patients often develop persistent deficits in social behavior, anxiety, and executive functions such as attention, mental flexibility, and task switching. These deficits are intrinsically associated with prefrontal cortex (PFC) functionality. Currently, there is a lack of analogous, behaviorally characterized TBI models for investigating frontal lobe injuries despite the prevalence of focal contusions to the frontal lobe in TBI patients. We used the controlled cortical impact (CCI) model in mice to generate a frontal lobe contusion and studied behavioral changes associated with PFC function. We found that unilateral frontal lobe contusion in mice produced long-term impairments to social recognition and reversal learning while having only a minor effect on anxiety and completely sparing rule shifting and hippocampal-dependent behavior.     

Correlating Behavioral Responses to fMRI Signals from Human Prefrontal Cortex: Examining Cognitive Processes Using Task Analysis

The aim of this methods paper is to describe how to implement a neuroimaging technique to examine complementary brain processes engaged by two similar tasks. Participants'' behavior during task performance in an fMRI scanner can then be correlated to the brain activity using the blood-oxygen-level-dependent signal. We measure behavior to be able to sort correct trials, where the subject performed the task correctly and then be able to examine the brain signals related to correct performance. Conversely, if subjects do not perform the task correctly, and these trials are included in the same analysis with the correct trials we would introduce trials that were not only for correct performance. Thus, in many cases these errors can be used themselves to then correlate brain activity to them. We describe two complementary tasks that are used in our lab to examine the brain during suppression of an automatic responses: the stroop¹ and anti-saccade tasks. The emotional stroop paradigm instructs participants to either report the superimposed emotional ''word'' across the affective faces or the facial ''expressions'' of the face stimuli^1,2. When the word and the facial expression refer to different emotions, a conflict between what must be said and what is automatically read occurs. The participant has to resolve the conflict between two simultaneously competing processes of word reading and facial expression. Our urge to read out a word leads to strong ''stimulus-response (SR)'' associations; hence inhibiting these strong SR''s is difficult and participants are prone to making errors. Overcoming this conflict and directing attention away from the face or the word requires the subject to inhibit bottom up processes which typically directs attention to the more salient stimulus. Similarly, in the anti-saccade task^3,4,5,6, where an instruction cue is used to direct only attention to a peripheral stimulus location but then the eye movement is made to the mirror opposite position. Yet again we measure behavior by recording the eye movements of participants which allows for the sorting of the behavioral responses into correct and error trials⁷ which then can be correlated to brain activity. Neuroimaging now allows researchers to measure different behaviors of correct and error trials that are indicative of different cognitive processes and pinpoint the different neural networks involved.     

Stochastic Dynamics Underlying Cognitive Stability and Flexibility

Cognitive stability and flexibility are core functions in the successful pursuit of behavioral goals. While there is evidence for a common frontoparietal network underlying both functions and for a key role of dopamine in the modulation of flexible versus stable behavior, the exact neurocomputational mechanisms underlying those executive functions and their adaptation to environmental demands are still unclear. In this work we study the neurocomputational mechanisms underlying cue based task switching (flexibility) and distractor inhibition (stability) in a paradigm specifically designed to probe both functions. We develop a physiologically plausible, explicit model of neural networks that maintain the currently active task rule in working memory and implement the decision process. We simplify the four-choice decision network to a nonlinear drift-diffusion process that we canonically derive from a generic winner-take-all network model. By fitting our model to the behavioral data of individual subjects, we can reproduce their full behavior in terms of decisions and reaction time distributions in baseline as well as distractor inhibition and switch conditions. Furthermore, we predict the individual hemodynamic response timecourse of the rule-representing network and localize it to a frontoparietal network including the inferior frontal junction area and the intraparietal sulcus, using functional magnetic resonance imaging. This refines the understanding of task-switch-related frontoparietal brain activity as reflecting attractor-like working memory representations of task rules. Finally, we estimate the subject-specific stability of the rule-representing attractor states in terms of the minimal action associated with a transition between different rule states in the phase-space of the fitted models. This stability measure correlates with switching-specific thalamocorticostriatal activation, i.e., with a system associated with flexible working memory updating and dopaminergic modulation of cognitive flexibility. These results show that stochastic dynamical systems can implement the basic computations underlying cognitive stability and flexibility and explain neurobiological bases of individual differences.     

Electrophysiological Evidence for Domain-General Inhibitory Control during Bilingual Language Switching

This paper presents an experiment that explored the role of domain–general inhibitory control on language switching. Reaction times (RTs) and event–related brain potentials (ERPs) were recorded when low–proficient bilinguals with high and low inhibitory control (IC) switched between overt picture naming in both their L1 and L2. Results showed that the language switch costs of bilinguals with high–IC were symmetrical, while that of bilinguals with low–IC were not. The N2 component failed to show a significant interaction between group, language and task, indicating that inhibition may not comes into play during the language task schema competition phase. The late positive component (LPC), however, showed larger amplitudes for L2 repeat and switch trials than for L1 trials in the high–IC group, indicating that inhibition may play a key role during the lexical response selection phase. These findings suggest that domain–general inhibitory control plays an important role in modulating language switch costs and its influence can be specified in lexical selection phase.     

Neural Mechanisms Underlying the Cost of Task Switching: An ERP Study

Background

When switching from one task to a new one, reaction times are prolonged. This phenomenon is called switch cost (SC). Researchers have recently used several kinds of task-switching paradigms to uncover neural mechanisms underlying the SC. Task-set reconfiguration and passive dissipation of a previously relevant task-set have been reported to contribute to the cost of task switching.

Methodology/Principal Findings

An unpredictable cued task-switching paradigm was used, during which subjects were instructed to switch between a color and an orientation discrimination task. Electroencephalography (EEG) and behavioral measures were recorded in 14 subjects. Response-stimulus interval (RSI) and cue-stimulus interval (CSI) were manipulated with short and long intervals, respectively. Switch trials delayed reaction times (RTs) and increased error rates compared with repeat trials. The SC of RTs was smaller in the long CSI condition. For cue-locked waveforms, switch trials generated a larger parietal positive event-related potential (ERP), and a larger slow parietal positivity compared with repeat trials in the short and long CSI condition. Neural SC of cue-related ERP positivity was smaller in the long RSI condition. For stimulus-locked waveforms, a larger switch-related central negative ERP component was observed, and the neural SC of the ERP negativity was smaller in the long CSI. Results of standardized low resolution electromagnetic tomography (sLORETA) for both ERP positivity and negativity showed that switch trials evoked larger activation than repeat trials in dorsolateral prefrontal cortex (DLPFC) and posterior parietal cortex (PPC).

Conclusions/Significance

The results provide evidence that both RSI and CSI modulate the neural activities in the process of task-switching, but that these have a differential role during task-set reconfiguration and passive dissipation of a previously relevant task-set.     

Consequences of behavioral dynamics for the population dynamics of predator-prey systems with switching

This article explores how different mechanisms governing the rate of change of the predators preference alter the dynamics of predator-prey systems in which the predator exhibits positive frequency-dependent predation. The models assume that individuals of the predator species adaptively adjust a trait that determines their relative capture rates of each of two prey species. The resulting switching behavior does not instantaneously attain the optimum for current prey densities, but instead lags behind it. Several mechanisms producing such lags are discussed and modeled. In all cases examined, our question is whether a realistic behavioral lag can significantly change the dynamics of the system relative to an analogous case in which the predators switching is effectively instantaneous. We also explore whether increasing the rate parameters of dynamic models of behavior results in convergence to the population dynamics of analogous models with instantaneous switching, and whether different behavioral models produce similar population dynamics. The analysis concentrates on systems that undergo endogenously generated predator-prey cycles in the absence of switching behavior. The average densities and the nature of indirect interactions are often sensitive to the rate of behavioral change, and are often qualitatively different for different classes of behavioral models. Dynamics and average densities can be very sensitive to small changes in parameters of either the prey growth or predator switching functions. These differences suggest that an understanding of switching in natural systems will require research into the behavioral mechanisms that govern lags in the response of predator preference to changes in prey density.     

Feature Integration and Task Switching: Diminished Switch Costs after Controlling for Stimulus,Response, and Cue Repetitions

This report presents data from two versions of the task switching procedure in which the separate influence of stimulus repetitions, response key repetitions, conceptual response repetitions, cue repetitions, task repetitions, and congruency are considered. Experiment 1 used a simple alternating runs procedure with parity judgments of digits and consonant/vowel decisions of letters as the two tasks. Results revealed sizable effects of stimulus and response repetitions, and controlling for these effects reduced the switch cost. Experiment 2 was a cued version of the task switch paradigm with parity and magnitude judgments of digits as the two tasks. Results again revealed large effects of stimulus and response repetitions, in addition to cue repetition effects. Controlling for these effects again reduced the switch cost. Congruency did not interact with our novel “unbiased” measure of switch costs. We discuss how the task switch paradigm might be thought of as a more complex version of the feature integration paradigm and propose an episodic learning account of the effect. We further consider to what extent appeals to higher-order control processes might be unnecessary and propose that controls for feature integration biases should be standard practice in task switching experiments.     

Phenotypic lability and the evolution of predator-induced plasticity in tadpoles

The hypothesis that predator-induced defenses in anuran larvae are maintained by divergent selection across multiple predation environments has not been fully supported by empirical results. One reason may be that traits that respond slowly to environmental variation experience a fitness cost not incorporated in the standard adaptive model, due to a time lag between detecting the state of the environment and expressing the phenotypic response. I measured the rate at which behavior and morphology of Rana temporaria tadpoles change when confronted with a switch in the predation environment at two points in development. Hatchling tadpoles that had been exposed during the egg stage to Aeshna dragonfly larvae were not phenotypically different from those exposed as eggs to predator-free conditions, and both responded similarly to post-hatching predator treatments. When 25-day-old tadpoles from treatments with and without dragonflies were subjected to a switch in the environment, their activity budgets reversed completely within 24-36 h, and their body and tail shape began changing significantly within 4 days. The behavioral response was conservative: Tadpoles switched from high-risk to predator-free treatments were slower to adjust their activity. The study confirmed that behavioral traits are relatively labile and exhibit strong plasticity, but it did not reveal such a pattern at the level of individual traits: Morphological traits that developed slowly did not show the least plasticity. Thus, I found that differences in lability of traits were useful for predicting the magnitude of plasticity only for fundamentally different kinds of characters.     

Viability Costs of Reproduction and Behavioral Compensation in Western Mosquitofish (Gambusia affinis)

The cost of reproduction hypothesis suggests that current reproduction has inherent tradeoffs with future reproduction. These tradeoffs can be both in the form of energy allocated to current offspring as opposed to somatic maintenance and future reproduction (allocation costs), or as an increase in mortality as a result of morphological or physiological changes related to reproduction (viability costs). Individuals may be able to decrease viability costs by altering behavior. Female western mosquitofish, Gambusia affinis experience a reduction in swimming ability as a consequence of pregnancy. We test for a viability cost of reproduction, and for behavioral compensation in pregnant female G. affinis by measuring survival of females in early and later stages of pregnancy when exposed to predation. Late-stage pregnant females experience a 70% greater probability of mortality compared to early-stage pregnant females. The presence of a refuge roughly doubled the odds of survival of both early and late-stage pregnant females. However, there was no interaction between refuge availability and stage of pregnancy. These data do not provide evidence for behavioral compensation by female G. affinis for elevated viability costs incurred during later stages of pregnancy. Behavioral compensation may be constrained by other aspects of the cost of reproduction.     

Go with the flow: water velocity regulates herbivore foraging decisions in river catchments

Foragers typically attempt to consume food resources that offer the greatest energy gain for the least cost, switching between habitats as the most profitable food resource changes over time. Optimal foraging models require accurate data on the gains and costs associated with each food resource to successfully predict temporal shifts. Whilst previous studies have shown that seasonal changes in food quantity and quality can drive habitat shifts, few studies have shown the effects on habitat choice of seasonal changes in metabolic foraging costs. In this study we combined field and literature data to construct an optimal foraging model to examine the effect of seasonal changes in food quantity, food quality and foraging costs on the timing of a switch from terrestrial to aquatic habitat by non‐breeding mute swans Cygnus olor in a shallow river catchment. Feeding experiments were used to quantify the functional response of swans to changes in aquatic plant biomasses. By sequentially testing alternative models with fixed or variable values for food quantity, food quality and foraging cost, we found that we needed to include seasonal variance in foraging costs in the model to accurately predict the observed habitat switch date. However, we did not need to include seasonal variance in food quantity and food quality, as accurate predictions could be obtained with fixed values for these two parameters. Therefore, the seasonal changes in foraging costs were the key factor influencing the behavioural decision to switch feeding habitats. These seasonal changes in foraging costs were driven by changes in water velocity; the profitability of aquatic foraging was negatively related to water velocity, as faster water required more energy to be expended in swimming. Our results demonstrate the importance of incorporating seasonal variation in foraging costs into our understanding of the foraging decisions of animals.