Division of labour and ‘foraging for work’: simulating reality versus the reality of simulations

Division of labour by worker age is one of the most prominent and widespread features of colony organization in the social insects. Tofts & Franks (1992) and Tofts (1993) proposed through simulations that a simple algorithm (‘foraging for work’) serves as a general model for understanding task allocation in eusocial animals ranging from ants and bees to naked mole-rats. They claimed that a correlation between worker age and tasks can arise from a set of simple behavioural rules in combination with a spatial organization of tasks within the nest, without any mechanism that causally links worker age and behaviour. Temporal polyethism emerges as a by-product of individual rules; ultimately they question whether age polyethism is itself ‘a myth’ (Franks 1994, page 238). The ‘foraging for work’ model represents a currently popular type of simulation used to investigate decentralized control in the organization of complex systems. The goal is to demonstrate how group-level behaviour may arise as an emergent property of the collective actions of individuals, by specifying the rules of only individual interactions (Deneubourg & Goss 1989; Kawata & Toquenaga 1994; Taylor & Jefferson 1994; Theraulez et al. 1995). Such simulations can play a critical role in inquiry, but their use and interpretation remain equivocal (Pattee 1989; Maddox 1995; Traniello & Robson 1995). Prior criticism of ‘foraging for work’ has been directed at the simulation’s unrealistic assumptions, contradiction by experiments designed to test it, and confusion and misrepresentation over caste theory (Robinson et al. 1994; Calderone 1995; Calderone & Page 1996; Traniello & Rosengaus 1997). Our critique focuses on the application and evaluation of simulations per se, because simulations present their own unique problems that are not always apparent. Our goal is to provide a constructive dialogue on the use of such approaches in the study of biological systems, using ‘foraging for work’ as an example. What exactly is ‘foraging for work’, exactly how has its utility been evaluated, and what does it illustrate about the use of models and simulations? Tofts & Franks (1992) argued that the tasks faced by a social insect worker can be viewed as being spatially ordered and sequential, starting at the brood pile where individuals are born into the first task and extending out to foraging, the last task. Individuals actively seek work, and will remain in a certain task zone as long as work is forthcoming. If an individual fails to find work too frequently, it will move randomly in either direction. ‘Foraging for work’ is an algorithm that incorporates these features. Tofts and Franks think that this algorithm can explain temporal polyethism: ‘Recent experiments seem to conclude that this is indeed the process through which the observed temporal polyethism [in social insects] arises’ (Tofts & Franks 1992, page 347). Unfortunately, this model cannot be evaluated on the basis of information provided in the summary article by Tofts & Franks (1992). The critical information detailing its underlying organization is provided by Tofts (1993), where the model appears more artificial than the simplified version presented by Tofts & Franks (1992). As described in more detail by Tofts (1993), ‘foraging for work’ actually comprises two components, each with its own rationale and justification. The first component involved deriving an algorithm to allocate individual workers to tasks, without any reference to worker age. The algorithm draws an analogy between an industrial Correspondence: S. N. Beshers, Department of Entomology, University of Illinois, Urbana, IL 61801, U.S.A. (email: [email protected]). S. K. Robson is now at the Department of Zoology, James Cook University, Townsville, Queensland 4811, Australia.