lable at ScienceDirect

Animal Behaviour 100 (2015) 44e50

Contents lists avai

Animal Behaviour

journal homepage: www.elsevier .com/locate/anbehav

Review

Information integration and multiattribute decision making innon-neuronal organisms

Chris R. Reid a, *, Simon Garnier a, Madeleine Beekman b, Tanya Latty b

a Department of Biological Sciences, New Jersey Institute of Technology, Newark, NJ, U.S.A.b Behaviour and Genetics of Social Insects Lab, School of Biological Sciences, The University of Sydney, Sydney, New South Wales, Australia

a r t i c l e i n f o

Article history:Received 22 September 2014Initial acceptance 22 October 2014Final acceptance 31 October 2014Published onlineMS. number: 14-00758

Keywords:brainlesscognitioncompensatory decision makingforagingtrade-off

* Correspondence: C. R. Reid, 439 Boyden Hall, RutgAvenue, Newark, NJ, 07102, U.S.A.

E-mail address: chrisreidresearch@gmail.com (C. R

http://dx.doi.org/10.1016/j.anbehav.2014.11.0100003-3472/© 2014 The Association for the Study of A

Decision making is a necessary process for most organisms, even for the majority of known life forms:those without a brain or neurons. The goal of this review is to highlight research dedicated to under-standing complex decision making in non-neuronal organisms, and to suggest avenues for furthering thiswork. We review research demonstrating key aspects of complex decision making, in particular infor-mation integration and multiattribute decision making, in non-neuronal organisms when (1) utilizingadaptive search strategies when foraging, (2) choosing between resources and environmental conditionsthat have several contradictory attributes and necessitate a trade-off, and (3) incorporating social cuesand environmental factors when living in a group or colony. We discuss potential similarities betweendecision making in non-neuronal organisms and other systems, such as insect colonies and themammalian brain, and we suggest future avenues of research that use appropriate experimental designand that take advantage of emerging imaging technologies.© 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Many organisms, including humans, are continually faced withdecisions about what to eat, with whom to mate and where to live.The enormous literature on decision making spans the disciplinesof psychology, economics, behavioural ecology, ethology andneurobiology. However, given the fact that brainless (or non-neuronal) organisms comprise the vast majority of all known life,there have been a disproportionately small number of studies ontheir decision-making abilities. Although non-neuronal organismslack the complex hardware of brained animals, they may live inenvironments that are no less complex. Hence, they face many ofthe same decision-making challenges as organisms with a brain:they must search for resources, choose between resources ofvarying quality, adapt to changing conditions and choose suitablemicroclimates to inhabit. We use the term ‘non-neuronal’ todescribe those organisms that lack neuron-based information-processing systems. Non-neuronal taxa include the immense va-riety of organisms in the bacterial domain and plant, fungi andprotist kingdoms, but exclude brainless animals such as the Cni-daria which, although lacking a centralized nervous system, stillpossess neuron-based information processing in the form of anerve net.

ers University, 195 University

. Reid).

nimal Behaviour. Published by Els

For the purposes of this review, we define the term ‘decision’ asfollows: the action by an entity (individual organism or group) ofselecting an option from a set of alternatives, based on character-istics of the alternatives that the entity can perceive. This definitiondoes notmake assumptions regarding the nature and complexity ofthe decision-making mechanism at work (i.e. the exact mecha-nisms by which the information on the different characteristics isintegrated by the entity while making a choice). It relies only on theobservable behaviour of the entity in question. A decision isconsidered made when the entity moves or points towards anoptionwith a certainty level and repeatability greater than random.This makes possible the comparison of the decision-making capa-bilities of different entities regardless of their nature or level ofcomplexity. The information relevant to the decision relates to oneor more of each option's ‘attributes’. Attributes are characteristicsthat describe an option, e.g. nutritive value of a food, humidity levelof a shelter, presence/absence of conspecifics. The simplest de-cisions are ‘single attribute’, where the organism considers only oneattribute. For example, cells of the bacterium Escherichia coli, whenpresented with five different concentrations of glucose, preferen-tially migrate towards the most concentrated glucose lure (Kim &Kim, 2010). In these experiments, the choice environment isdeliberately simplified such that each food varies in only a singleattribute: glucose concentration. Examples of non-neuronal or-ganisms making single-attribute decisions include choosing

evier Ltd. All rights reserved.

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e50 45

between light and dark microenvironments (bacteria: Sackett &Armitage, 1997; protists: Iwatsuki, 1992; plants: Thimann &Curry, 1960; fungi: Saranak & Foster, 1997) and choosing betweenenvironments that differ in their concentration of repellent sub-stances (bacteria: Tso & Adler, 1974; protists: Keating & Bonner,1977; fungi: Cameron & Carlile, 1980).

Information relevant to the decision-making process is availableto an organism both from external sources within the environmentand from internal sources within the organism itself. Externalsources of information can include characteristics such as soilquality, food concentration and the presence/absence of competi-tors, while internal information can include the organism's satia-tion level, memory or mating status. Information can be detectedvia multiple sensory channels and chemical and physical processeswithin the organism. The decisions that require the least processingpower are those that can be reached using information detected viaa single channel, and require only evaluation of a single attribute.Bacteria migrating towards a glucose lure, for example, need onlyprocess information on glucose concentration via a single receptortype (Wadhams & Armitage, 2004). However, many decisions innature are multi-attribute, where each alternative can be charac-terized by an array of attributes, some or all of which the organismmay take into account before making a decision. For example, anorganism may need to simultaneously process information on awide variety of relevant attributes such as the caloric value of afood, toxins in the food (McArthur, Orlando, Banks,& Brown, 2012),the risk of predation (Brown & Kotler, 2004) and suitability of themicroclimate (Webster& Dill, 2006). Normative models of decisionmaking assume that individuals make multiattribute decisions byassigning a value to some or all relevant attributes, calculating the

External info

Decision pr

‘Single-attribute’ when optionsdiffer in only a single attribute(i.e. different concentrations of glucose)

‘Noncompensatory’ when oneattribute overrides all others

Simple

Non-neuronal

Figure 1. Simplified schema for categori

sum of all attributes for each option and then selecting the optionthat has the greatest total value. The idea that an individual shouldchoosewhichever option yields the greatest benefit is encapsulatedin the economics concept of ‘utility’ (Stigler, 1950). Decisionmakingbecomes sophisticated when the utility of available options canonly be determined by combining information relating to multipleattributes. The ability to make multiattribute decisions bycombining information from a variety of sources is called infor-mation integration.

A simplified schema is provided in Fig. 1 that delineates thebasic aspects of the decision-making process, and highlights thesalient steps discussed in this review.

Strategies for dealing with multiattribute problems can beclassified as either compensatory or noncompensatory (Pitz &Sachs, 1984). In noncompensatory decision making, a high valueon one attribute always overrides all other relevant attributes.Noncompensatory decisions are usually highly context dependent.For example, a foraging organism might always choose the bestquality food irrespective of the risk of predation, or it might alwayschoose to avoid risky areas, irrespective of food quality. In thesecases the organism can sense predation risk and food quality, butbases its decisions on only one of these attributes. Non-compensatory strategies have the benefit of being computationallysimple, because the organism need only consider a single attribute,thereby disregarding information from all other attributes. Incontrast, organisms using compensatory strategies must maketrade-offs, so that high values in one attribute can sometimestrump values in the other attributes and vice versa. Compensatorystrategies are thought to be computationally intensive, as theyrequire the organism to compare options based on their relative

rmation

Internalinformation

ocesses

‘Multiattribute’ when multipleattributes per option can be

evaluated independently

‘Compensatory’ when severalattributes may conflict.

Integration of information frommultiple sources returns the

decision

Complex

organism

zing decision making in organisms.

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e5046

differences, rather than whether one attribute simply exceeds adesired threshold.

In many environments, the ability to make trade-offs betweendifferent attributes could lead to improved survivorship andincreased fitness. An organism that always chooses the highestquality food, regardless of the predation risk involved in acquiringthe resource, would seemingly be at a disadvantage compared to anindividual that could weigh up both relative risk and relative gain.However, the computational resources needed to make compen-satory decisions might be lacking in brainless organisms, whichmight therefore be constrained to simple, noncompensatory formsof decision making. In addition, the lack of computationalcomplexity in noncompensatory decision making may allow or-ganisms to make decisions faster, which may be beneficial in someenvironments.

Here we review studies on information integration and multi-attribute decision making in non-neuronal organisms. In particularwe ask: to what extent can non-neuronal organisms use and inte-grate information when making foraging decisions?

DECISION MAKING DURING EXPLORATION

Organisms living in patchy environments are faced with thechallenge of locating new resources. Organisms can search theirenvironment using specialized movement patterns, or by growingexploratory structures such as roots, hyphae or pseudopodia(Fig. 2). ‘Adaptive search strategies’ consist of two or more different

Figure 2. (a) Plasmodium of the slime mould Physarum polycephalum searching for food on(b) Corded mycelium of Phanerochaete velutina after 39 days of foraging (Bebber, Hynes, Ddiagram of the root systems of two neighbouring peach trees showing avoidance behaviodritiformis colony grown on nutrient-limited substrate (Ben-Jacob & Levine, 2006). Reprodu

patterns of movement that can be deployed under different cir-cumstances. Many bacteria achieve chemotaxis by switching from arandom walk to a biased random walk through slight adjustmentsin the frequency of their direction-altering ‘tumbles’. When tum-bling, the bacterium pivots rapidly on the spot, randomly reor-ienting the direction it faces. When swimming towards anattractant, the foraging bacterium responds by decreasing the fre-quency of tumbles, biasing the direction of movement in the di-rection of the attractant (Alon, Surette, Barkai, & Leibler, 1999;Macnab & Koshland, 1972). Chemotaxis is widespread amongbacteria, and at first glance seems a simple process involving aresponse to a single attribute of the local environment (thechemical gradient) at each time step. However, studies using E. colihave shown that the decision about the direction in which to moverequires the integration of internal, external and temporal infor-mation. Owing to their small size, E. coli cells are incapable ofsensing a spatial gradient (Roth, 2013); two receptors spaced atmaximal distance across the cell would still be too close to detectany local change in stimulus concentration. This means that thebacterium cannot discriminate between the chemical concentra-tions of two points in space at the same time, and so cannot rely onspatial information to inform its decisions. Escherichia coli adapts itssearch strategy based on temporal information, which requires aform of memory. The activity of the cell's membrane chemore-ceptors over the past 1 s in time is continually compared with theactivity registered during the previous 3 s (Alejandra Guzm�an,Delgado, & De Carvalho, 2010; Roth, 2013; Segall, Block, & Berg,

a petri dish of nutrient-free agar by extending pseudopodia. Photo: Audrey Dussutour.arrah, Boddy, & Fricker, 2007). Reproduced by permission of the Royal Society. (c) Aur (Schenk et al., 1999). Reproduced by permission of Elsevier. (d) Paenibacillus den-ced by permission of the Royal Society and Eshel Ben-Jacob.

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e50 47

1986). If receptor activity (corresponding to an increase in thechemical concentration) has increased in this time, the organismsuppresses tumbling, and meanders towards the food source. Tonavigate away from repellent chemicals detected in the environ-ment, the bacterium uses the same system, but increases tumblingfrequency in response to a climbing gradient (Alon et al., 1999).These studies show that foraging E. coli integrate information fromtheir present environment (current receptor activity) with infor-mation relating to their past environment (receptor activity overthe previous 3 s) to decide which search pattern to employ.

Several non-neuronal taxa are able to use information aboutpreviously searched areas when deciding where to search next.Advancing hyphae of basidiomycetous fungi deposit chemical in-hibitors that prevent hyphae from searching the same area twice(Bottone, Nagarsheth, & Chiu, 1998). Cells of the amoeboid protistPhysarum polycephalum similarly integrate information fromchemical cues deposited in the environment to enhance both theirforaging and navigation. During exploration, this slime mouldleaves behind a thick layer of nonliving extracellular slime, whichacts as a repellent to future local exploration (Reid, Beekman, Latty,& Dussutour, 2013; Reid, Latty, Dussutour, & Beekman, 2012). Thememory of which areas have been searched is thus stored in theexternal environment. Physarum polycephalum is capable ofdiscriminating between conspecific and heterospecific extracel-lular slime deposits, preferring to explore environments alreadyvisited by heterospecifics over those explored by conspecifics (Reidet al., 2013). A possible interpretation of this result is that the slimemould is attracted to the slime trail of heterospecific slime mould,perhaps in order to chase it down for consumption. However, whengiven a choice between virgin, unexplored territory and one coatedin heterospecific deposits, P. polycephalum preferred the former(Reid et al., 2013). This choice hierarchy indicates thatP. polycephalum is capable of eavesdropping on the foraging cues ofother organisms, potentially enabling it to choose the less depletedof two partially exploited environments, as the diet of hetero-specifics is unlikely to overlap completely with P. polycephalum'sdiet. In contrast, an environment explored by conspecifics is likelyto be stripped of useful resources. Such experiments show thatP. polycephalum can integrate information relating to the currentenvironment, as well as its own foraging history and that of otherspecies, in order to make complex decisions in the search for food.

DECISION MAKING DURING EXPLOITATION

Once resources have been located, organisms are often facedwithfurther decisions to make, such as deciding which is the best ofmultiple resources to exploit, and how to trade off conflicting attri-butes such as food availabilityanddanger. The specificmacronutrientcomposition of food items can also have a strong influence on anorganism's choice of food (Raubenheimer & Simpson, 1993). Tomaximize their growth, reproductive output or longevity, organismsneed to balance their intake of two key macronutrients: protein andcarbohydrates. The ratio that optimizes growth is known as the‘intake target’, and varies between organisms. Individuals face thechallenge of balancing their choice of food so that they reach theirintake target. To investigate whether P. polycephalum slime mouldsare able to actively influence their food intake so that they reach theirintake target, Dussutour, Latty, Beekman, and Simpson (2010) pre-sented slime mould amoebae with binary choices between foodblocks that varied in their protein to carbohydrate (P:C) ratio. None ofthe food blocks on their own provided the amoebae with their idealintake target (about 2:1 P:C), so in order to achieve their intake tar-gets, amoebaeneeded to covereach food blockwith a precise amountof biomass through which they digested and absorbed the nutrients.On each of six food pairings, slime mould amoebae exhibited

nutritional balancing by adjusting their biomass allocation in exactlythe amounts needed to ensure uptake of the optimal P:C ratio. Whensimultaneously presented with a set of 11 diets that differed in theirP:C ratios, amoebae selected the combination and coverage of foodblocks that was closest to their intake target. The set-up of thisforaging challenge resembles the combinatorial optimization prob-lemknownas theKnapsackproblem. In theKnapsackproblem, one isgiven a set of items that differ in their mass and ‘value’. The aim is tofill a container (the ‘knapsack’) to less than or equal to itsweight limitwith a combination of these items,while attempting tomaximize thetotal value of the collection (Martello & Toth, 1990). In Dussutouret al.'s experiments (2010), amoebae were clearly able to integrateinformation about both protein and carbohydrate content and makea compensatory decision about how to ideally distribute theirbiomass to reach their preferred intake target.

Non-neuronal organisms can also take into account the varyingstrengths of different food cues. When a foraging P. polycephalumcell is given a choice between an oat-based food source and an eggyolk-based food source, it shows a strong preference for the eggyolk-based food (Reid et al., 2013). Reid et al. (2013) providedP. polycephalum amoebaewith a choice between a region containingchemical cues of past exploration by conspecifics (which is repellentto the organism) that led to a highly attractive egg yolk-based foodsource and a regionwithout such chemical cues but leading to a lessattractive food source (oats). When the different food cues weredetectable but somewhat weak (by placing them 4 cm from theamoeba), the slime mould avoided areas that had previously beenexplored by conspecifics and chose to consume the low-reward oat-based food. However, if the food cuesweremade stronger byplacingthe food sources closer to the organism (0.5 cm), it shifted itspreference to ignore the extracellular slime and attain the highlyrewarding egg yolk-based food (Reid et al., 2013). The slime mouldthus makes a compensatory decision trading off whether an areahas been explored and the strength of the different food cues.

A few experiments have also shown that P. polycephalum canmake sophisticated trade-offs between overall access to food andexposure to danger. All else being equal, slime mould amoebaeprefer high-calorie foods over low-calorie foods (Latty & Beekman,2010). Slime moulds are also particular about microclimate, andwhenever possible will avoid light exposure, presumably to avoidthe risk of dehydration associated with sunlight (Latty & Beekman,2010, 2011a). Faced with a choice between a high-quality but illu-minated patch, and a lower quality but darkened patch, slimemoulds only select the illuminated high-quality patch when itcontains food that is at least five times more concentrated thanfood present in the ‘safe’ alternative (Latty & Beekman, 2010).

The trade-off between reward and danger extends beyond thechoice between risky food patches, to shape the structure of theorganism itself. Slime mould amoebae connect food resources withtransport tubules built from protoplasm, and these tubules aremost likely to follow the shortest path between food items(Nakagaki, Kobayashi, Nishiura, & Ueda, 2004; Nakagaki, Yamada,& Hara, 2004; Nakagaki, Yamada, & T�oth, 2000; Tero et al., 2010;Reid & Beekman, 2013). When half of the region between twodiagonally placed food sources was illuminated with a strong lightsource, the slime mould faced a trade-off between connecting thefood sources via the shortest path and avoiding the lit region(Nakagaki et al., 2007). To solve this problem, the organism alteredthe course of its connecting tubule to minimize the amount ofbiomass that was exposed to light. Importantly, the slime moulddid not simply avoid the light altogether, as this would haveresulted in an overly elongated tubule. Rather, it calculated a paththat was relatively short, while also keeping light exposure to aminimum: a successful demonstration of compensatory decisionmaking applied to a path selection problem.

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e5048

Like many other foraging organisms, the roots of a plant exploretheir environment in search of resources. Plant roots extendthrough the soil microenvironment, searching for soil regions withthe most favourable combination of nutrients and moisture, whiletrying to avoid competitionwith neighbouring root systems. Hence,plant roots need todecidewhich soil patches to exploit andwhich toavoid. In general, roots avoid foraging in patches that have alreadybeen exploited by a competitor, and they preferentially exploitnutrient-rich soil patches over nutrient-deficient soil patches(Cahill & McNickle, 2011; McNickle, St Clair, & Cahill, 2009). Forinstance, when two Abutilon theophrasti plants were grown besideeach other in soil with homogeneous nutrient availability, their rootsystems strongly avoided one another (Cahill et al., 2010). However,when a nutrient-rich patch was placed between the two plants,each plant's root system overrode its usual avoidance behaviour,and root mixing occurred in the high-quality region (Cahill &McNickle, 2011; Cahill et al., 2010). In other species (Avena sativa),foraging roots employ the opposite strategy, ignoring the roots ofheterospecifics under low-nutrient conditions, but expressingavoidance under conditions of homogeneously high resourceavailability (Litav & Harper, 1967). Foraging plant roots may chooseto avoid neighbouring root systems to avoid searching depletedareas (Schenk, Callaway, & Mahall, 1999), or to minimize competi-tionwhen the neighbouring root system is recognized as belongingto a closely related, nonself plant (Dudley & File, 2007; Murphy &Dudley, 2009). Alternatively, roots may proliferate around areasinhabited bycompeting root systems in order to pre-empt resourcesthat would otherwise be available to the competitor (Bartelheimer,Steinlein, & Beyschlag, 2006), or they may ignore neighbouringroots altogether (Mommer et al., 2010). Multiple combinations ofthese strategies may be employed by the same species of plantdepending on additive cues from the environment such as nutrientconcentration, indicating that compensatory decision making iswidespread in plants (Cahill & McNickle, 2011).

MAKING DECISIONS TOGETHER

When individual non-neuronal organisms group together toform colonies, they gain a new source of data to incorporate intotheir decision-making strategies: signals from their neighbours.This new channel of information can be integrated with that fromtheir own internal physiology and from other external sources intheir surrounding environment, to enhance the individual'sdecision-making abilities, even beyond what they could accom-plish alone.

In search of food on resource-limited substrates, colonies of thebacterium Paenibacillus dendritiformis take on a dendriticmorphology reminiscent of a slime mould plasmodium or basid-iomycetous fungus (Fig. 2d). This shape arises as the colony self-organizes to overcome a complex trade-off. The individual bacte-ria require a high density of neighbours to collectively generate asurface layer of lubricant through which they can swim; however,the resource-limited substrate contains insufficient food to sustaina dense population (Ben-Jacob & Levine, 2006). The bacteria solvethis multiconstraint optimization problem by accurately adjustingthe number of individuals within each branch such that a sufficientlocalized density for locomotion is reached, while not exceedingthe carrying capacity of the substrate (Ben-Jacob & Levine, 2006;Harshey, 2003). This group-level search strategy is decided by in-dividual cells each integrating information on local nutrient andlubricant levels. Individuals must make a compensatory decisionabout whether to join a branch, in which they balance the choicebetween increasing group size for maximal lubrication and limitinggroup size to prevent starvation and subsequent death of thebranch.

Groups of Myxobacteria, such as Myxococcus xanthus, integrateinformation about individual and group-level starvation to makecollective decisions. Myxobacteria use a surface-exposed proteincalled the C-factor to externally display information about theirown internal physiological status (Ben-Jacob, Becker, Shapira, &Levine, 2004). The bacteria use this information to performdifferent cooperative group behaviours depending on the groupstarvation consensus. At an advanced level of starvation, the cellsaggregate and move in circular motions, both clockwise and anti-clockwise, around a centre of rotation that will become the fruitingbody, the site of sporulation (Koch, 1998). Under low levels ofstarvation, the bacteria move as roving groups of cells. The numberof cells within the group must be high enough for their secretedlytic enzymes to digest prey bacteria in the surrounding environ-ment (Koch, 1998). Deciding which group behaviour to undertakerequires integration of internal and external information relating tostarvation signals, a consensus decision based on the differentstarvation levels of all members of the group, and recognition of thegroup number exceeding a critical threshold.

SIMILARITIES WITH OTHER DECISION-MAKING SYSTEMS

Despite lacking the cognitive architecture typically associatedwith information processing, non-neuronal taxa are still capable ofintegrating information from multiple, disparate sources fromwithin both their external environment and their own internalphysiology. We have presented several examples of compensatorydecisionmaking by non-neuronal taxa and of unicellular organismsincorporating information from their neighbours when makingtheir choices. Taken together, the body of literature strongly sug-gests that brains or even simpler neuronal networks are not pre-requisites for complex decision making.

The mechanisms underlying the decision-making strategies ofnon-neuronal taxa may be strikingly similar to those observed inbrained organisms. Evidence from slime moulds suggests that non-neuronal organisms may even be subject to the same cognitiveconstraints as humans. One study indicates that slime moulds aresusceptible to the same kind of ‘irrational’ decision making previ-ously only seen in humans and other animals (Latty & Beekman,2011b). When given a choice between two options that varied intwo competing attributes, in this case food concentration and lightintensity, slime moulds did not show a strong preference for oneoption over the other. However, in the presence of a third optioninferior to both original options (a decoy), slime moulds changedtheir preference (Latty & Beekman, 2011b). Such behaviour is ir-rational, as the presence of an inferior option should not affect theorganism's original preference. A related study found that slimemoulds, like humans, are subject to speed / accuracy trade-offs, butonly when presented with a difficult task set (Latty & Beekman,2011a). These similarities raise the tantalizing possibility that de-cision making in non-neuronal and neuronal organisms is based onthe same underlying principles.

Decision making in brains has been modelled using a variety ofapproaches, most of which are based on the idea that evidence infavour of each alternative accumulates over time until some deci-sion threshold is reached. In ‘tug-of-war’ models, decisions arebased on the relative weights given to alternative options. Oncesufficient information has accumulated, one option is chosen.Because such choice models require some sort of comparison be-tween the different options, decision making becomes more diffi-cult themore options there are or the closer the options are in value(Kacelnik, Vasconcelos, Monteiro, & Aw, 2010). An alternativedecision-making model, coined the sequential choice model byKacelnik et al. (2010), assumes no comparisons among differentoptions. Instead each option is considered sequentially, and the first

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e50 49

that reaches the decision threshold is chosen (Bogacz, Brown,Moehlis, Holmes, & Cohen, 2006). The fundamental differencebetween the two decision-making models is the existence of atrade-off between decision accuracy and cost of evaluation in tug-of-war models, but not in the sequential choice model. Thus, whenthe number of choices increases, the tug-of-war model predictsthat decision time will increase, whereas the opposite is predictedunder the sequential choice model (Kacelnik et al., 2010).

In slime moulds, biomass is transported through tubules thatwiden when in contact with resources. It has been argued thatcompetition between biomass flowsmay be analogous to the build-up of ‘evidence’ thought to occur in human brains (Latty &Beekman, 2011a). Although the specific mechanism is different(neuron firing rate versus biomass flow rate), the underlyingprinciples appear to be the same. A similar idea was proposed byMarshall et al. (2009), who showed that positive feedback couldunderlie decision making in systems as dissimilar as ant coloniesand human brains. Similarly, Nicolis, Zabzina, Latty, and Sumpter(2011) argued that positive feedback could account for the irratio-nal behaviour observed in both slime moulds (Latty & Beekman,2011b) and ant colonies (Sasaki & Pratt, 2011). We suggest thatdecision-making processes in brains (Usher & McClelland, 2001),slime moulds (Nakagaki et al., 2000) and social insects (Pratt,Mallon, Sumpter, & Franks, 2002; Seeley, 2010) are all based onthe accumulation of information and positive feedback. Whetherthese similarities are purely superficial or whether they point to adeeper principle underlying decision making and informationprocessing in biological systems remains to be tested.

CHALLENGES AND FUTURE

The Principal challenge facing future behavioural studies of non-neuronal organisms lies in appropriate experimental design. Thelong history and extensive literature relating to information pro-cessing, memory, learning and problem solving in neuronal animalsshould provide the scaffolding for the design of experimentsinvestigating these phenomena in non-neuronal organisms. Thisapproach has proven fruitful with the labyrinth maze, first used toexplore the mental processes of rats (Small, 1901), and recentlysuccessfully applied to slime moulds (Nakagaki et al., 2000; Reid &Beekman, 2013) and fungal hyphae (Hanson, Nicolau, Filipponi,Wang, & Lee, 2006). Similarly, the mechanisms of problem solv-ing and decision making in slime moulds have been explored usingthe classical Y-maze choice set-up (Reid et al., 2013, 2012), the U-shaped trap maze used in robotics for testing autonomous navi-gation (Reid et al., 2012), and the protocols used to study choice andirrationality in animals and humans (Latty & Beekman, 2009,2011b). The difficulty and importance of appropriate experi-mental design are illustrated by the history of research seeking todemonstrate learning in the protozoan Paramecium caudatum. Ahost of studies stretching back to 1911 have claimed to demonstratediscrimination learning and the learning of tube escape behaviourin P. caudatum. For each of these studies, however, there exists aslew of counter-studies demonstrating that the original evidenceresulted from artefacts (recounted in Mingee & Armus, 2009). Ar-tefacts are difficult to avoid when studying the behaviour of uni-cellular organisms, because the environmental scale at which theyfunction is so removed from ours. Researchers studying thebehaviour of brained animals take care to minimize familiar con-founding factors such as distracting visual, auditory and olfactorystimuli, seasonal changes and the animals' past experience. It isintrinsically more difficult to account a priori for such minute andunfamiliar factors as surface ion accumulation or homogeneousculture mixing: environmental factors pivotal to unicellular func-tioning, yet overlooked in past experiments purporting to

demonstrate unicellular learning (Mingee & Armus, 2009). Onesolution to this problem is to work with non-neuronal organismsthat function at a scale closer to our own, which perhaps explainsthe disproportional success of behavioural experiments usingmacroscopic slime mould rather than microscopic unicellularmodel systems.

Innovative design of experimental apparatus and techniquesoffers another avenue for successful research. Where the challengeis a question of spatial scale, live cell imaging technologies, whichhave progressed dramatically over the last decade, can be of greatuse. Tools such as atomic force microscopy, traction force micro-scopy and total internal reflection microscopy have found wideapplication in cellular and subcellular research, especially wherethe molecular dynamics and mechanical properties of individualcell movements are investigated. These tools have the potential touncover the underlying functional mechanics of unicellulardecision-making behaviour, in the same way that neurobiologyinforms animal behaviour. We suggest that these established andpowerful tools designed for detailed observation of cell activityshould be incorporated into future unicellular behaviourexperiments.

The studies reviewed here clearly demonstrate the capacity ofnon-neuronal organisms to make complex decisions. The mostimportant area of future study is to prove the potential adaptivesignificance of this decision-making behaviour. The study by Lattyand Beekman (2009) is one of the few to attempt this, bydemonstrating that P. polycephalum amoebae had greater foragingsuccess (measured by weight gain) when grown in environmentswith patchy food distribution, and gained less weight in moreuniformly distributed foraging environments. Since the number ofspores that can be produced by a plasmodium is directly related toplasmodium mass, weight gain is an excellent proxy for repro-ductive fitness in plasmodia. Future studies should follow thisexample, and be specifically designed to quantify adaptive out-comes of certain decision-making behaviours in non-neuronalorganisms.

Acknowledgments

This work was funded by Australian Research Council grants toT.L. (DP140103643 and DP110102998) and M.B. (FT120100120 andDP140100560).

References

Alejandra Guzm�an, M., Delgado, A., & De Carvalho, J. (2010). A novel multiobjectiveoptimization algorithm based on bacterial chemotaxis. Engineering Applicationsof Artificial Intelligence, 23, 292e301.

Alon, U., Surette, M. G., Barkai, N., & Leibler, S. (1999). Robustness in bacterialchemotaxis. Nature, 397, 168e171.

Bartelheimer, M., Steinlein, T., & Beyschlag, W. (2006). Aggregative root placement:a feature during interspecific competition in inland sand-dune habitats. Plantand Soil, 280, 101e114.

Bebber, D. P., Hynes, J., Darrah, P. R., Boddy, L., & Fricker, M. D. (2007). Biologicalsolutions to transport network design. Proceedings of the Royal Society B: Bio-logical Sciences, 274, 2307e2315.

Ben-Jacob, E., Becker, I., Shapira, Y., & Levine, H. (2004). Bacterial linguisticcommunication and social intelligence. Trends in Microbiology, 12, 366e372.

Ben-Jacob, E., & Levine, H. (2006). Self-engineering capabilities of bacteria. Journalof the Royal Society Interface, 3, 197e214.

Bogacz, R., Brown, E., Moehlis, J., Holmes, P., & Cohen, J. D. (2006). The physics ofoptimal decision-making: a formal analysis of models of performance in two-alternative forced-choice tasks. Psychological Review, 113, 700e765.

Bottone, E. J., Nagarsheth, N., & Chiu, K. (1998). Evidence of self-inhibition by fila-mentous fungi accounts for unidirectional hyphal growth in colonies. CanadianJournal of Microbiology, 44, 390e393.

Brown, J. S., & Kotler, B. P. (2004). Hazardous duty pay and the foraging cost ofpredation. Ecology Letters, 7, 999e1014.

Cahill, J. F., & McNickle, G. G. (2011). The behavioral ecology of nutrient foraging byplants. Annual Review of Ecology, Evolution, and Systematics, 42, 289e311.

C. R. Reid et al. / Animal Behaviour 100 (2015) 44e5050

Cahill, J. F., McNickle, G. G., Haag, J. J., Lamb, E. G., Nyanumba, S. M., & St Clair, C. C.(2010). Plants integrate information about nutrients and neighbors. Science,328, 1657.

Cameron, J. N., & Carlile, M. J. (1980). Negative chemotaxis of zoospores of thefungus Phytophthora palmivora. Microbiology, 120, 347e353.

Dudley, S. A., & File, A. L. (2007). Kin recognition in an annual plant. Biology Letters,3, 435e438.

Dussutour, A., Latty, T., Beekman, M., & Simpson, S. J. (2010). Amoeboid organismsolves complex nutritional challenges. Proceedings of the National Academy ofSciences of the United States of America, 107, 4607e4611.

Hanson, K. L., Nicolau, D. V., Filipponi, L., Wang, L., & Lee, A. P. (2006). Fungi useefficient algorithms for the exploration of microfluidic networks. Small, 2,1212e1220.

Harshey, R. M. (2003). Bacterial motility on a surface: many ways to a common goal.Annual Review of Microbiology, 57, 249e273.

Iwatsuki, K. (1992). Stentor coeruleus shows positive photokinesis. Photochemistryand Photobiology, 3, 469e471.

Kacelnik, A., Vasconcelos, M., Monteiro, T., & Aw, J. (2010). Darwin's “tug-of-war” vs.starlings' “horse racing”: how adaptations for sequential encounters drivesimultaneous choice. Behavioral Ecology and Sociobiology, 65(3), 547e558.

Keating, M., & Bonner, J. (1977). Negative chemotaxis in cellular slime molds. Journalof Bacteriology, 130, 1e5.

Kim, M., & Kim, T. (2010). Diffusion-based and long-range concentration gradientsof multiple chemicals for bacterial chemotaxis assays. Analytical Chemistry, 82,9401e9409.

Koch, A. L. (1998). The strategy of Myxococcus xanthus for group cooperativebehavior. Antonie van Leeuwenhoek, 73, 299e313.

Latty, T., & Beekman, M. (2009). Food quality affects search strategy in the acellularslime mould, Physarum polycephalum. Behavioral Ecology, 20, 1160e1167.

Latty, T., & Beekman,M. (2010). Food quality and the risk of light exposure affect patch-choice decisions in the slime mold Physarum polycephalum. Ecology, 91, 22e27.

Latty, T., & Beekman, M. (2011a). Speed-accuracy trade-offs during foraging de-cisions in the acellular slime mould Physarum polycephalum. Proceedings of theRoyal Society B: Biological Sciences, 278, 539e545.

Latty, T., & Beekman, M. (2011b). Irrational decision-making in an amoeboid or-ganism: transitivity and context-dependent preferences. Proceedings of theRoyal Society B: Biological Sciences, 278, 307e312.

Litav, M., & Harper, J. L. (1967). A method for studying spatial relationships betweenthe root systems of two neighbouring plants. Plant and Soil, 26, 389e392.

Macnab, R. M., & Koshland, D. E. (1972). The gradient-sensing mechanism in bac-terial chemotaxis. Proceedings of the National Academy of Sciences of the UnitedStates of America, 69, 2509e2512.

Marshall, J. A. R., Bogacz, R., Dornhaus, A., Planqu�e, R., Kovacs, T., & Franks, N. R.(2009). On optimal decision-making in brains and social insect colonies. Journalof the Royal Society, Interface, 6, 1065e1074.

Martello, S., & Toth, P. (1990). Knapsack problems. New York, NY: Wiley.McArthur, C., Orlando, P., Banks, P. B., & Brown, J. S. (2012). The foraging tightrope

between predation risk and plant toxins: a matter of concentration. FunctionalEcology, 26, 74e83.

McNickle, G. G., St Clair, C. C., & Cahill, J. F. (2009). Focusing the metaphor: plant rootforaging behaviour. Trends in Ecology & Evolution, 24, 419e426.

Mingee, C. M., & Armus, H. L. (2009). Unsuccessful reinforcement of a discrete ac-tion in paramecia, P. caudatum. Psychological Reports, 105, 533e538.

Mommer, L., Van Ruijven, J., De Caluwe, H., Smit-Tiekstra, A. E., Wagemaker, C. A. M.,Joop Ouborg, N., et al. (2010). Unveiling below-ground species abundance in abiodiversity experiment: a test of vertical niche differentiation among grasslandspecies. Journal of Ecology, 98, 1117e1127.

Murphy, G. P., & Dudley, S. A. (2009). Kin recognition: competition and cooperationin impatiens (Balsaminaceae). American Journal of Botany, 96, 1990e1996.

Nakagaki, T., Iima, M., Ueda, T., Nishiura, Y., Saigusa, T., Tero, A., et al. (2007).Minimum-risk path finding by an adaptive amoebal network. Physical ReviewLetters, 99, 068104e2e4.

Nakagaki, T., Kobayashi, R., Nishiura, Y., & Ueda, T. (2004). Obtaining multipleseparate food sources: behavioural intelligence in the Physarum plasmodium.Proceedings of the Royal Society B: Biological Sciences, 271, 2305e2310.

Nakagaki, T., Yamada, H., & Hara, M. (2004). Smart network solutions in anamoeboid organism. Biophysical Chemistry, 107, 1e5.

Nakagaki, T., Yamada, H., & T�oth, A. (2000). Maze-solving by an amoeboid organism.Nature, 407, 470.

Nicolis, S. C., Zabzina, N., Latty, T., & Sumpter, D. J. T. (2011). Collective irrationalityand positive feedback. PLoS One, 6, e18901.

Pitz, G. F., & Sachs, N. J. (1984). Judgment and decision: theory and application.Annual Review of Psychology, 35, 139e164.

Pratt, S., Mallon, E., Sumpter, D., & Franks, N. (2002). Quorum sensing, recruitment,and collective decision-making during colony emigration by the ant Leptothoraxalbipennis. Behavioral Ecology and Sociobiology, 52, 117e127.

Raubenheimer, D., & Simpson, S. (1993). The geometry of compensatory feeding inthe locust. Animal Behaviour, 45, 953e964.

Reid, C. R., & Beekman, M. (2013). Solving the Towers of Hanoi e how an amoeboidorganism efficiently constructs transport networks. The Journal of ExperimentalBiology, 216, 1546e1551.

Reid, C. R., Beekman, M., Latty, T., & Dussutour, A. (2013). Amoeboid organism usesextracellular secretions to make smart foraging decisions. Behavioral Ecology,24, 812e818.

Reid, C. R., Latty, T., Dussutour, A., & Beekman, M. (2012). Slime mold uses anexternalized spatial “memory” to navigate in complex environments. Pro-ceedings of the National Academy of Sciences of the United States of America, 109,17490e17494.

Roth, G. (2013). The long evolution of brains and minds. Berlin, Germany: Springer.Sackett, M., & Armitage, J. (1997). Photoresponses of the purple nonsulfur bacteria

Rhodospirillum centenum and Rhodobacter sphaeroides. Journal of Bacteriology,179, 6764e6768.

Saranak, J., & Foster, K. (1997). Rhodopsin guides fungal phototaxis. Nature, 387,465e466.

Sasaki, T., & Pratt, S. C. (2011). Emergence of group rationality from irrational in-dividuals. Behavioral Ecology, 22, 276e281.

Schenk, H., Callaway, R., & Mahall, B. (1999). Spatial root segregation: are plantsterritorial? Advances in Ecological Research, 28, 145e180.

Seeley, T. D. (2010). Honeybee democracy. Princeton, NJ: Princeton University Press.Segall, J. E., Block, S. M., & Berg, H. C. (1986). Temporal comparisons in bacterial

chemotaxis. Proceedings of the National Academy of Sciences of the United Statesof America, 83, 8987e8991.

Small, W. (1901). Experimental study of the mental processes of the rat, II. TheAmerican Journal of Psychology, 12, 206e239.

Stigler, G. J. (1950). The development of utility theory. I. Journal of Political Economy,58(4), 307e327.

Tero, A., Takagi, S., Saigusa, T., Ito, K., Bebber, D. P., Fricker, M. D., et al. (2010). Rulesfor biologically inspired adaptive network design. Science, 327, 439e442.

Thimann, K. V., & Curry, G. M. (1960). Phototropism and phototaxis. ComparativeBiochemistry, 1, 243e309.

Tso, W., & Adler, J. (1974). Negative chemotaxis in Escherichia coli. Journal ofBacteriology, 118, 560e576.

Usher, M., & McClelland, J. L. (2001). The time course of perceptual choice: the leaky,competing accumulator model. Psychological Review, 108, 550e592.

Wadhams, G. H., & Armitage, J. P. (2004). Making sense of it all: bacterial chemo-taxis. Nature Reviews. Molecular Cell Biology, 5, 1024e1037.

Webster, S. J., & Dill, L. M. (2006). The energetic equivalence of changing salinity andtemperature to juvenile salmon. Functional Ecology, 20, 621e629.