Anim. Behav., 1977, 25, 1034-1046

DO HENS SUFFER IN BATTERY CAGES?ENVIRONMENTAL PREFERENCES AND WELFARE

BY MARIAN DAWKINSAnimal Behaviour Research Group, Department of Zoology, Oxford

Abstract. The question `Do hens suffer in battery cages T is difficult to answer because of the problemof objectively assessing suffering in animals . It is argued that preference tests may be one way of throw-ing light on this difficult problem . This paper describes some experiments on habitat preference indomestic hens. No preference was observed between a commercial battery cage and a large pen whenhens were given continuous access to the two . A simultaneous choice between a battery cage and anoutside hen-run showed a clear preference for the run, but choice was strongly influenced by priorexperience. The strength of the run preference was investigated by `pitting' the run against food andaccess to companions .

IntroductionTo many scientists the question `Do hens sufferin battery cages?' is invalid, as it is a questionabout an animal's subjective feelings and thesubjective feelings of animals are not directlyaccessible to scientific investigation. We maylook for signs of physical ill-health and we canstudy behaviour, but we can never know forcertain whether these observable symptomsare accompanied by subjective consciousness . Wemay study animals as if they were machines andtry to discover what the behaviour machineryis by looking at the relation between inputs andoutputs. There will be no place in such modelsfor subjective feelings. But just because we studyanimals as if they were machines which merelybehaved and felt nothing, does not mean thatthat is all they are . As Griffin (1976) points out :`It is very easy for scientists to slip into thepassive assumption that phenomena with whichtheir customary methods cannot deal effectivelyare unimportant or even non-existent.' There areat least two reasons for believing that thesubjective feelings of animals are both importantand very far from non-existent .The first reason comes from the way we

react to other people. Exactly the same argu-ments apply to other human beings as to animals .Strictly, we can never know that other peoplehave mental experiences . We cannot do anexperiment to demonstrate that another humanbeing is conscious or has feelings remotely likeour own any more than we can for a hen or achimpanzee . But commonsense and intuitiontell us that other human beings probably do thinkand feel as we ourselves do . Their behaviour andphysiology are sufficiently similar to our ownthat we are quite prepared to accept this without

1034

the need for logical proof. We base our ideas ofmorality, such as not to eat people, kill them ortorture them, on this commonsense view ratherthan on strict logic. There is a'common ground'suffering, pleasure, pain, which most peopleaccept as being universals of human experience .But must this common ground stop at theboundaries of our own species? If we acceptthe evolutionary continuity between man andother animals for physiology, biochemistry andat least some aspects of behaviour, why notmental experiences too (Brophy 1972 ; Griffin1976)? Ryder (1971) and Singer (1976) use theterm 'speciesism' by analogy with racism .Many animals, particularly mammals and birds,seem to have all the basic nervous apparatusfor feeling pain and experiencing emotion .Commonsense suggests that they can suffer . Ofcourse, we cannot be certain about this, butneither can we about other people. If animalsdo suffer from some of our scientific experimentsor from the ways in which we keep them forcommercial profit, then it seems important onmoral grounds to take this into account .

The second reason is a more biological one,and comes from consideration of the possiblesurvival value of subjective feelings . We areaccustomed to asking questions about thefunctional significance of the shape, colouringand behaviour of an animal . It is reasonable toassume that subjective feelings too evolvedbecause animals which possessed them werefitter than those which did not. Exactly why theyshould have been fitter is one of the most pro-found mysteries of biology, and although varioussuggestions have been made, e .g. that they aidedsimulation of the future (Wall 1974), communi-cation or learning, the full reasons are not

understood . But whatever the reasons, each ofus knows that we have subjective feelings andthose feelings must be a product of naturalselection . They are part of biology .

For these two reasons, questions about themental states of animals are important : theyare important on moral grounds in that if weignore an animal's subjective feelings, we maybe guilty of inflicting or at least condoningsuffering in highly sentient and emotional beings ;and they are important on biological groundssince subjective experience is part of an animal'sequipment for survival .

This seems to bring us to an impasse : sub-jective feelings are important and yet we appar-ently cannot study them scientifically . There is,of course, no magic method for solving themajor philosophical issues which still exist andgaining direct access into what animals arefeeling. But I want to suggest that some behavi-ours are likely to be good indicators of mentalstate. The question is : which ones? I wouldlike to discuss this question in relation to oneparticular problem which has recently arouseda lot of public concern-whether hens in batterycages suffer-but I hope that similar lines ofreasoning can also be used for a wider range ofproblems .

Some of the most widely advocated indicatorsof possible suffering are physical ill-health, poorgrowth rate, or in the case of laying hens, asubstandard number of eggs produced . Presum-ably physical health is an essential part of mentalwell-being, and it is often argued that produc-tivity is in itself a sufficient guide on the groundsthat animals which are suffering would notproduce well. But, as pointed out by theBrambell Committee (Report, 1965), it is quitepossible for animals to be growing or layingwell, despite periods of acute but transitoryphysical or mental suffering.

Another proposed indicator of suffering hasbeen whether the behaviour patterns of hens inbattery cages differ from those of feral fowl, orBurmese red jungle fowl (Gallus gallus spadiceus),which is thought to be the wild form (Wood-Gush 1971). Thorpe (1965, 1967) argues thatwhere animals are kept in situations that sup-press their natural behaviour patterns, then suffer-ing may well result. Hens in battery cages arephysically prevented from performing manybehaviours such as roosting, dust-bathing andground-scratching. The birds cannot flap theirwings, walk more than a few steps and are frus-trated in their nesting behaviour (Wood-Gush

DAWKINS : DO HENS SUFFER IN BATTERY CAGES?

1035

& Gilbert 1969; Wood-Gush 1972). Such amajor disruption of the `normal' behaviourpatterns should immediately alert us to thepossibility of suffering but does not in itselfconstitute evidence that the animals are suffering(Ewbank 1968) . Battery-kept hens might, forexample, show less anti-predator behaviourthan unrestricted fowl, but it would be implaus-ible to argue from this that the birds weresuffering from lack of anti-predator behaviour .The Brambell Committee (Report, 1965)

argued that the best indicator of suffering inother species was by analogy with the symptomsknown to accompany mental suffering in humanbeings. To draw too close a parallel in thisrespect is, however, most dangerous . The `feargrimace' of the chimpanzee, which to the humaneye looks like a happy smile is in fact given insituations which the animal finds fearful (Jolly1972) . So even in an animal which is very closelyrelated to us there is the possibility of seriouserror if the analogy with our own expressionsis drawn too closely. How much greater is thepossibility of error in animals which are evenless like ourselves and have evolved their ownindependent ways of expressing emotion? Atfirst sight it also seems humane to decry certainhusbandry practices on the grounds that weourselves would not like to be housed in thisway. Whilst we must never forget the `commonground' that may exist between us, and otheranimals, neither should we forget that differentspecies may have genuinely different require-ments. Baby jackdaws relish regurgitated worms,which I would not .

Physiological measures of `stress' (Selye 1952)such as increased activity of the pituitary andadrenal cortex and development of gastic ulcershave also been used to indicate which conditionsanimals find stressful. Duncan (1974), in adiscussion of this in relation to animal welfare,points out that there are difficulties in measuringphysiological changes in living animals . Thereis also still the problem of relating changedphysiological state to the animal's subjectivefeelings of distress . Animals may suffer beforeany physiological disturbances are detected orthere may be physiological changes totallyuncorrelated with suffering.

There is, however, another, possibly moredirect way of inferring what an animal's feelingsare, at least its feelings about a particularenvironment. This is to allow the animal toexpress those feelings by being given the oppor-tunity to move out of that environment if it wants

1 036

ANIMAL BEHAVIOUR, 25, 4

to . For example, if we want to know whetherhens dislike battery cages, the hens could begiven the opportunity to choose between batterycages and some other environment and allowedto `vote with their feet'. If it were to turn outthat hens have a very strong preference forenvironments other than battery cages and ifthey would repeatedly perform some task forthe `reward' of being allowed out of a batterycage, then we could say that they dislikedbattery cages . If, on the other hand, they didnot seem to show a very strong preference oneway or the other, we might be less inclined tosay that they disliked them. Hughes & Black(1973) have shown that hens given the oppor-tunity to stand on different sorts of cage floorspent more time on hexagonal mesh than oncoarse rectangular mesh or than on perforatedsteel sheet. It seems very likely that the hens`like' some floors more than others. The reasonwhy environmental preference is very likely to beclosely correlated with an animals subjectivefeelings is that natural selection will havefavoured such a connection. We should expectco-evolution between an animal's fitness indifferent environments and the capacity of eachenvironment to evoke escape or settling beha-viour (Levins 1968). As Orians (1971) puts it :`birds should evolve to be more "turned on" byhabitats in which their fitness is greater' .

This paper describes a number of experimentson habitat preference in domestic hens . Theexperiments could be taken from a strictly beha-viourist viewpoint with no connotations of sub-jective feelings . However, if the preceding argu-ments are accepted, they may also be seen as a firststep towards developing experimental methodsto deal with the mental experiences of animals .

The first experiment is an attempt to find outwhere battery-kept hens choose to spend theirtime when given the opportunity to be either in abattery cage or in a much bigger pen . The birdswere given continuous access to these twoenvironments over a period of 12 h . Wherewould they go? Would they ever re-enter abattery cage once they had left it? Would theirchoice be influenced by the time of day or bythe lack of familiarity with the pen?

Experiment 1 : Where do Battery Kept Hens Go ifGiven Continuous Access to Battery Cages and a

Larger Pen?Methods

Twelve Sykes Tinted hens which had beenreared commercially on deep litter by Ross

Poultry Ltd, were used for this experiment . Atthe age of 18 weeks they were put into Patchettbattery cages (floor space 0 .38 x 0 .43 m) withtwo birds per cage.

The birds were housed in a light-proof shed,on a `step-up' system of lighting and fed com-mercial mash. At the age of 32 weeks they weretested individually by being placed in the centreof a pen (dimensions shown in Fig. 1). Thefloor of the pen was covered with wood shavings .Two Patchett battery cages with the backsremoved to permit access were placed at one endof the pen. The floor of the backs of the batterycages was flush with the floor of the pen, sothat a hen could walk easily in and out of thebattery cages . The position of the two batterycages was changed from one side of the pento the other for different hens . Food and waterwere available in equal amounts in similartroughs in the pen and the battery cages . Eachbird was tested for 12 h but observations weremade only once in each hour for 5 min at regularintervals . At the beginning of a test, an individulbird was placed in the centre of the large pen .It was left to adjust for 5 min and then the firstobservation session was begun. For subsequentobservations, I entered the hut 5 min before anobservation was due, to allow the bird to adjustto my presence, and then for 5 min recordedwhether the bird was in the pen or one of thebattery cages. I also recorded where and whenany eggs were laid. In order to separate theeffects of time of day from those of familiaritywith the test apparatus, the tests of differentbirds were started at different times of day . Onebird would be tested from 08 .30 until 20 .30hours, another (on its testing day) from 10 .30to 20.30 hours and then again from 08 .30 until10.30 hours the following morning and so on .This means that many of the birds were removedfrom the apparatus overnight and restarted inthe morning. The birds were at the time of testing

Large pen Battery -rcage

LBatterycage

Fig . 1 . Diagram of the large pen and battery cages usedin experiment 1 . The battery cages had the backs removedso that the hens could wander freely in and out . Thefloor of the pen was 2 .6 x 0 . 82 m and was covered withwood shavings . The floor space of each battery cagewas 0 .38 x 0 .43 m, the floor being of plastic coveredrectangular mesh.

on a 14-h light/dark, schedule with the lightscoming on from 07.45 hours daily .

ResultsIn order to ensure that the results represent

independent observations, the preference of thebirds between pen and battery cage were firstcalculated as where the birds were standing atthe beginning of each 5-min observation session(i .e. 12 observations for each of 12 birds) .On this measure, 5 out of the 12 birds preferredthe battery cages in the sense that the majorityof their observation sessions found them stand-ing in one of the battery cages, and 7 preferredthe pen. Overall, there was no significantpreference one way or the other, using aWilcoxon Matched Pairs Test on the sign andmagnitude of the differences (N = 12, P > 0 .05two-tailed (Siegel 1956)) . This is somewhatsurprising in view of the fact that the combinedfloor area of the two battery cages was so muchsmaller than the floor area of the pen (See Fig. 1)so that by chance one might have expected themto be seen more often in the pen . Through the12 h of the test (08 .30 to 20 .30 hours), just underhalf the birds were seen in one of the batterycages (Fig. 2). The data from Fig . 2 can be re-arranged so that the position of the hens is seenat different times since the beginning of theirown individual tests rather than the time of day(made possible by the design of the experimentin which birds started their tests at differenttimes of their day). Figure 3 shows such arearrangement and the birds seemed not tochange in how likely they were to be in pen orbattery cage throughout their tests, even thoughthey might be expected to have become morefamiliar with the pen the longer they had been inthe apparatus .

No. of hens ~in battery 6cage at

5start ofobservationsession

DAWKINS : DO HENS SUFFER IN BATTERY CAGES?

1037

Time of day

Fig . 2. Numbers of hens (out of a total of 12) in one ofthe battery cages at the beginning of each 5-min observa-tion session throughout a 12-h day .

Another measure of preference which makesuse of rather more of the data is how manycomplete minutes of a 5-min observation sessiona hen spent in pen or battery cage . The maxi-mum possible in any one environment would be5, but if a hen repeatedly moved between thetwo, she could score 0 complete minutes foreither environment . In any one 5-min observa-tion session, a hen could be said to prefer oneenvironment over the other depending on howmany complete minutes she spent in each. Outof her 12 observation sessions she could be saidto prefer one environment over the other if themajority of sessions showed a preference forthis environment . Using this measure, four henspreferred the battery cages, seven preferred therun and one scored equally for the two environ-ments, but the difference is not significant(P > 0 .05 Wilcoxon Test, two-tailed) . Nor wasthere any tendency for the hens to spend morecomplete minutes in either environment depend-ing on the time of day (Fig . 4) or how longthey had been in the apparatus (Fig . 5). Nineof the hens laid an egg during the course ofbeing tested, six in the pen and three in one ofthe battery cages but the figures are too smallto draw any conclusions .Discussion

The really rather surprising result of thisexperiment was the lack of demonstrated prefer-ence for either the pen or the battery cages .Of course, such a negative result does not implythat the hens do not have a preference : it couldsimply be that the preference was not demon-strated in these particular experimental con-ditions .

One reason why a preference was not demon-strated may have been that the chickens did notregard the pen as sufficiently different from the

No. of hens 6in battery 5cage atstart ofobservationsession

I

II

No. of observation sessions sincestart of test

Fig . 3 . Numbers of hens (out of a total of 12) in one of thebattery cages at the beginning of each 5-min observationperiod throughout a 12-h test .

1 038

ANIMAL BEHAVIOUR, 25, 4

battery cages to exhibit a preference betweenthem (both pen and battery cage were inside,and the main differences were that the pen waslarger and had the floor covered with woodshavings). To meet this possibility, all the experi-ments described below involve testing a hen'sresponses to two environments which differedin very many more ways . The two environmentswere a battery cage and a hen-run outside in thegarden.

The second reason for the apparent lack ofpreference may have been the particular methodof testing. It is known that a hen's preferencemay be strongly affected by how the prefer-ence is measured. Hughes (1976) tested thepreference of domestic hens for wire or litterfloors and found that the birds chose differentlydepending on whether they were given constantaccess to both floor types or whether they hadto make a choice between separate cages, wherethe choice was irrevocable for several hours .Because of the possibility that the method oftesting described above may have given a mis-leading impression of the kinds of environ-ments that chickens like and dislike, I haveadopted two other methods of assessing therelative attractiveness of environments to hens .

The first of these methods involved releasinga hen from a starting box and measuring howquickly she moved into either a battery cage or ahen-run. These experiments are described indetail elsewhere (Dawkins 1976) . The resultsshowed that using this particular measure of

5-

Minutes

4

3

2

8.30

10.30

1830

14 .30

18.30

18.30

HR5

Time of dayFig. 4. Mean number of complete minutes spent by 12hens in a battery cage (- - - -) and pen (-) in each of12 5-min observation sessions throughout the day .There is no significant tendency for the time spent ineither environment to be different at different times of theday (Friedman two-way analysis of variance, P > 0 .5).

preference, battery-kept hens initially preferredbattery cages to a run in the garden in the sensethat they moved more quickly into a batterycage than into an outside run. However, thispreference was found to be highly dependent onthe environment in which the hens had beenliving before being tested. Hens which had beenliving outside in a garden hen-run preferred therun, while caged birds preferred the batterycage. So the initial reluctance of a battery-kepthen to enter a run was probably due to itsstrangeness rather than to its undesirability .It seemed very important, therefore, to investi-gate precisely how much of the hen's behaviourwas due to the unfamiliarity of the environ-ments and to plot the time course of adjustmentto the new environment. This was done in theexperiment to be described next by repeatedlyoffering a hen a simultaneous choice between arun and a battery cage . When she had made herfirst choice, she was then confined for 5 min inwhichever environment she had chosen and thenreplaced in the starting box and offered the samechoice again . As the trials progressed, therefore,the hen would gradually accumulate experienceof being confined in each of the two environ-ments and by making her choose repeatedly, anychanges in her preference could be monitoredas a function of this experience .

5

4

3Minutes

2

EN

''BATTERYCAGE

12 3456789101112No. of gbservation sessions sincestart of test

Fig. 5 . Mean number of complete minutes spent by 12hens in a battery cage (- - - -) and pen (-) in each of 125-min observation sessions throughout the 12-h test .There is no significant tendency for the time spent ineither environment to vary depending on how long thebirds have been in the apparatus (Friedman two-wayanalysis of variance, P > 0 .90).

Since the conditions in which the hens hadbeen living before being tested is known to havea profound effect on their performance in a test,(Dawkins 1976), two groups of hens were used,one accustomed to living in battery cages andone group used to living outside in the garden .

Experiment 2: Do Hees Choose a Battery Cageor an Outside Run and Does the Choice Change

with Experience?Methods

Fourteen Sykes Tinted hens, aged 48 weeksat the time of testing, were used for this experi-ment. They had been reared commercially ondeep litter by Ross Poultry Ltd, until the age of20 weeks. On arrival in Oxford they wererandomly divided into two groups : half were putoutside into a hen-house and run (house 1 .83 x1 .14 m ; run area 7 .34 m2) ; half were keptindoors in Patchett battery cages (floor space0 .38 x 0 .43 m) with two birds in each cage.

The birds were individually given a choicebetween the two environments presented as thealternatives in a T-maze (see Fig . 6): an outsidehen-run (1 .53 x 0 .76 x 0 .9 m high) made ofwood and chicken wire and supplied by ParkLines Ltd, and a Patchett battery cage with theback removed to allow a hen to enter . Fortechnical reasons it is very difficult to present achoice between a battery hen-house and a gardenrun, so a compromise in the form of a smallhut on wheels that could be moved to anydesired position, was adopted . The hut was litand heated as much as possible like the housein which the hens had been living, and in orderto simulate a `battery' of cages, the hut con-tained three battery cages, the middle one ofwhich was the test cage .

Each hen was offered this choice 24 timessuccessively, between the hours of 07 .30 to 11 .00and 16.30 to 18.30 daily, extending over several

hut ori wheels with 3~ I battery cages

I~ slid;ng doors

~J

DAWKINS : DO HENS SUFFER IN BATTERY CAGES?

outside run

L,taming box

Fig. 6. Diagram of the choice test apparatus . The runmeasured 1 .83 x 0 .76 x 0 .9 m high. The floor space ofeach battery cage was 0 .38 x 0 .43 m.

1039

days. Only when one hen had completed her 24choices was the next hen started . For eachchoice, an individual hen was placed in thestarting box with the door closed for 1 min.The door was then raised and the hen releasedinto a corridor from which she could seesimultaneously the run and the battery cage .The time from the raising of the starting boxdoor to when the hen entered one of the twotest environments was measured and called thelatency of choice. A hen was considered to havemade a choice only when she had put both feetin one or other environment. If a hen failed tomove out of the starting box within 15 min, shewas placed gently in the centre of the corridor(this happened in about 10 % of trials) . In orderto ensure that the hen had a standard andmeasured amount of time in the environmentof her choice, once she had entered it, the doorleading to it was closed behind her and she wasshut in for 5 min . After this time, she wasreturned to the starting box and the next trialbegun. In this way, a hen gradually accumulated5-min blocks of experience of the environ-ments she chose. A hen which had been livinginside a battery cage was given its 24 trialsfollowing the series of a hen that had beenliving in the garden so that test order could notbe a confounding variable in the comparisonbetween the two groups. The position of thetwo environments with respect to the choicecorridor was changed daily to avoid positioneffects . Food was available in similar troughs inboth environments .

ResultsAs expected from previous work with latencies

to the two environments presented successively(Dawkins 1976), the first choice of the hens wasrelated to the environment in which they hadbeen living. All seven of the hens which hadbeen living in the garden chose the run on theirfirst trial whereas only three of the battery-keptbirds did so (P = 0 .05, Fisher Exact ProbabilityTest). However, the battery-kept hens veryquickly began to change their pattern of choiceand became more and more likely to choosethe run as their trials proceeded (Fig . 7). Thisresult is interesting as it would seem to suggestthat only a relatively small amount of experienceof the run (a matter of minutes) is sufficient tomake even battery-kept hens prefer the run .

In addition to which environment was chosen,the latency of that choice also provides valuableinformation about the hens' preferences. These

1040

No. of birds(out of 7)choosingrun

Na of birds(out of7)choosingrun

latencies were, however, so variable both betweenindividual birds and from trial to trial even ofthe same bird to the same environment thatgraphs for the 14 birds are presented separatelyin Fig . 8. Each graph incorporates two kinds ofinformation : information about which environ-ment the bird chose and information about thelatency of choice, plotted cumulatively as thetrials progressed . The bird could allocate itstesting time between three separate places : therun (5 min at a time), the battery cage (also amaximum of 5 min/trial), and the choicecorridor (this time being measured as the latencyof response) . After each trial, a certain amountof time would have elapsed since the beginningof the series for that bird and that time couldhave been spent in some or all of those threeenvironments. Each graph plots cumulativelythe proportion of the total test time spent inrun and battery cage after different numbers oftrials. The direction and slope of the graph showsthe effect of previous experience, for example, a

ANIMAL BEHAVIOUR, 25, 4

a . BATTERY CAGED HENS

Trial no.

b. OUTSIDE HENS

Trial no.

Fig. 7. Numbers of (a) battery-caged birds and (b) birds which had beenliving outside choosing the run in each of 24 trials . The difference betweenthe two groups is significant only for the first three trials (P = 0 . 05, FisherExact Probability test) . The battery caged birds become more likely tochoose the run as the trials proceed (Spearman rank correlation : rs = 0 . 35 ;N = 24, P < 0.05).

positively accelerating line would mean that themore time a bird spent in an environment, themore time she would spend in the future in thatsame place . Fig. 8(a) shows that for six of thehens kept in battery cages the proportion of timespent in the run gradually increased over 24trials . Only one hen consistently spent more timein the battery cage and even this hen began tospend more time in the run towards the end ofthe series . On the other hand, the amount oftime spent in the battery cage remained roughlyconstant or dropped off. This implies that where-as the effect of being in the run was to increasethe future amount of time the bird spent inrun, this was not true of the battery cage. Thiswould seem to be a fairly objective way of sayingthat hens `liked' the experience of being outsidein the run more than they `liked' being in abattery cage. Future experiments might profitablyinvolve recording details of the behaviour of thehens during the 5-min period after each choice .

Hens which had been living outside (Fig . 8(b))tended to show consistently high levels of timein the run and certainly no tendency to increasetime spent in the battery cage as a result ofentering the battery cage . Most of them enteredonce or twice at the most and never again .Discussion

These results show that battery-kept hensprefer an outside run to a battery cage, onceinitial unfamiliarity with the run has beenovercome. These hens take a very short time toget used to the run (a matter of minutes) andto come to prefer it to their familiar cages. Themore experience of the run they have, the morelikely they are to choose to spend time in the

proportion

of total

test

time

DAWKINS: DO HENS SUFFER IN BATTERY CAGES?

1041

1D-

.5 .s

25

1S 18 So 24

Trial no,Fig. 8 (a)

run in the future . Being in a battery cage, on theother hand, does not seem to increase the likeli-hood that the hens will choose to spend time in abattery cage in the future, either for battery-kept birds or for hens which have been livingoutside in a garden run. Birds which have beenliving outside consistently prefer the run overover the battery cage to a very marked extent .

Although these results give a preliminaryindication that hens `like' (as revealed by theirbehaviour) an outside run more than they `like' abattery cage, there is a long way to go beforeany conclusions can be drawn about the extentto which hens suffer in battery cages. If a manwere to be given a choice between a £1 note anda £10 note, for example, and he consistently

1 .a

.7b

BATTERY KEPTHENS

4

8 1S 18 f30 d4

Trial na

1 042

test

time

chose the £10 note, we could hardly be entitledto conclude that he suffered in the presence ofsingle pound notes. Preference per se cannotbe used to infer suffering or distress . It can onlybe used in this way if we can establish how greatthe preference is in terms of some yardstick,such as how much of some other commodityan animal is prepared to give up in order toobtain something . For example, hens are knownto be very social animals and the sight of other .

ANIMAL BEHAVIOUR, 25, 4

.5 .

Cumulative.25

4 8 12 18 20 24

OUTSIDEHENS

4

.7b

.2C

8 12 "1$ 20 24

4 8 12 18 20 84

7W

.5

.20

8 12 1t 20 24

4 8 12 8 E _ 94

Trial no.

Trial no.Fig. 8 (b)

Fig . 8 . (a) cumulative proportion of total test time which seven individualbattery-kept hens spent in the run (-) and the battery cage ()during the course of their 24 trials (for details see text) . (b) Cumulativeproportion of total test time which seven individual hens which had beenliving outside in a garden run spent in the run (-) and the battery cage() during the course of their 24 trials .

hens appears to increase the attractiveness of anenvironment, since hens moved more quicklyinto a place where there were other hens(Dawkins 1976). Would hens still choose anoutside run if it meant isolation from their flockmates? If they did still prefer an outside rununder these circumstances, we could concludethat they `liked' being in the run more than beingwith other hens . If the opposite result were found,and they chose to be with the other birds,

then the conclusion might be that although theyprefer a run, other things being equal, thispreference is not as strong as their attraction toother birds. In this way it should be possibleto get the hens to rank their preferences fordifferent commodities. In the next experimentreported below, hens were offered such a choicebetween an outside run in which they would bealone, and a battery cage near other birds .

Experiment 3: Do Hens Prefer an Outside Runon Their Own or a Battery Cage Near Other

Hens?Methods

Ten Sykes Tinted hens which had been usedfor the previous experiment were used for thistest. Half the birds were from each group, andsince being tested previously, had been returnedto their original living conditions, i .e. thebattery-kept birds had been living in cages andthe outside group in the garden. This experimenttook place immediately after the previousone, but since this had taken 5 weeks to complete,the time lag between the two was different fordifferent individuals. The choice apparatuswas the same as that used in experiment 2except that the two battery cages on either sideof the test cage in the hut (see Fig . 6) were notempty as in the previous experiment, but eachcontained a single bird unfamiliar to the testbird. This meant that the choice presented totest the bird was a battery cage with other birdsin cages on either side of it and an outside runin the garden without other hens . In order tomake sure that the birds realized that there wereother birds in the battery hut but not in the run,the experimental procedure was rather differentfrom the earlier experiment . Before the actualchoice was presented, each bird had a series of`training' trials. in which entry to one environ-ment was prevented by closing the sliding doorleading to it and the bird had, therefore, tostay in the choice corridor or go into the otherenvironment. All the birds eventually went intothe single available environment upon which theywere shut in for 5 min . Three of these trainingtrials were given to each environment so thatby the end of the six trials, the bird had had atleast 15 min experience of both environments .The seventh test was the choice test, with the birdhaving access to both environments simul-taneously. Only one such choice test wasgiven to each bird. One bird received its sixtraining trials and it choice test in quick suc-cession, the entire test usually taking about 1 h .

DAWKINS: DO HENS SUFFER IN BATTERY CAGES? 1043

The latency of responses both during the, trainingtrials and the choice test was recorded . Theposition of the two environments was changedbetween individuals and each hen received thethree training trials for each environment inalternation . Battery-kept and outside hens weretested equally often before and after each other .Another difference from the previous experi-ment was that there was no food available ineither the run or the battery cage (the reason forthis will become apparent from the next experi-ment) .Results

In the final choice test, 9 out of the 10 birdschose the run by themselves over the batterycage near other birds . The one bird that chosethe battery cage was a bird that had been livingoutside and was showing pre-egg-laying activity .She laid an egg soon after her test so her choicebehaviour may have been temporarily upset .The latency data from the training trials alsoprovides information about the hens' preferencesfor the two environments . If the latency ofresponse to the battery cage in each of the`battery alone' training trials is compared withthe latencies in the `run alone' training trials(Fig. 9) it will be seen that the birds took verymuch longer to enter the battery cage than the

2-

Latency

(minutes) I

`,battery cagenear hens

run ,

I

2

3Training trial

Fig. 9 . Median response latency to battery cage presentedalone and run presented alone in the training trials beforea choice test. There is no significant tendency for theselatencies to change over the three trials (P > 0 . 1,Friedman test, N = 10) .

1044

run. There was no significant trend for theselatencies to change during the training period .

DiscussionBoth by the criterion of hens choosing in a

simultaneous choice test and by that of thelatency of response when the environments werepresented singly during `training', it would seemthat these hens preferred to be in an outside run,even though it meant being by themselves tobeing in a battery cage near other hens . This wastrue even though the previous experience of thetest birds was rather varied . Half had beenliving in battery cages, and half outside and allhad experienced various degrees of time in bothenvironments during a previous test . It could beobjected to this experiment that the period of`training' was not long enough and that if therehad been more training trials, the birds wouldhave had more chance to `take stock' of thetwo environments and might then have chosendifferently. This may indeed be the case, but tojudge by the previous experiment, even veryshort periods of time seem to be enough for a hento learn about two environments . It wouldcertainly be important to repeat this experimentwith a longer period of training before we candraw any definite conclusions about how hensrank the opportunity to go into an outsiderun compared with the opportunity to be in acage near other hens.

Another objection is that the fact that otherbirds are attractive to hens was inferred fromprevious work rather than demonstrated in thepresent experiment, so that, strictly speaking,all that can be concluded from this experiment isthat hens prefer a run to a battery cage and thispreference still exists even if there are otherhens near the battery cage and not in the run .Nevertheless, taking this as a preliminary result,it would certainly seem to suggest that, subjectto the objections just mentioned, the preferencefor the run over the battery cage would seem tobe sufficiently strong to override the probableattraction of other hens. Further experimentsalong these lines are planned .

Another way in which it may be possible toassess the strength of hens' preferences for therun is by finding out whether they still choosethe run if they can obtain food only in the batterycage. Will the attraction of the run outweighthat of food?

ANIMAL BEHAVIOUR, 25, 4

Experiment 4: Do Hens Prefer an Outside RunWith No Food or a Battery Cage with Food?Methods

The ten hens which were used for experiment 3were also used for this experiment . The choice-testing apparatus and procedure were the sameas that described for experiment 3, except thatthe battery hut did not contain the two hens inbattery cages . The test battery cage this timehad a trough of food and the outside run con-tained no food. The hens were not deprived offood before being tested, but the daily renewal oftheir food in their living cages was not carriedout until after they had been tested. The henswere given six training trials, three to the batterycage alone and three to the run alone. As before,the seventh trial was the choice test .Results

In the final choice test 3 out of the 10 henschose the battery cage with food, 7 chose therun without food. This result is not significantlydifferent from chance (P > 0 .1 Binomial Test),but the fact that three hens (two battery-keptand one outside hen) chose the battery cage withfood rather than the run means that at the veryleast, the hens' preferences are not absolutelyclear cut, even though they were not deprivedof food. Furthermore, the latencies of theresponse to the two environments presented sing-ly (during the training trials) showed that therun was entered more quickly (Fig . 10). Thereseemed to be some tendency for the latencyof response to the battery cage to go down overthe three trials, perhaps as the birds learned thatthere was food in the cage, but this is notsignificant. However, it would certainly suggestthat it would be worth repeating this experimentwith a longer period of training and to moreaccurately `titrate' access to a run againstaccess to food.Discussion

It was argued in the introduction that althoughit is very difficult, it may not be totally impossibleto develop a methodology for dealing with thesubjective feelings of animals . The long-termhope is to be able to assess animal sufferingin a reasonably objective way . Different methodscan all contribute something to our picture, buta particularly important source of evidence maybe the kinds of environmental conditions thatthe animals themselves choose when given theopportunity to do so . The experiments describedhere, together with others such as those ofHughes (1976) are a first step in developing

7

6.

5.

Latency 4 -(minutes)

a

2-

I .

DAWKINS : DO HENS SUFFER IN BATTERY CAGES?

1045

%,battery cagewith food

run

1.

I ' 2 ' 3Training trial

Fig. 10. Median response latency to battery cage pre-sented alone and run presented alone in the training trialsbefore a choice test . There is no significant tendency forthese latencies to change over the three trials (P > 0 . 1,Friedman test, N = 10) .

such a methodology . The experiments do not,and could not, demonstrate the existence ofmental events in hens. It would be possible toview these preference tests simply as overtbehaviour, with no overtones of mental experi-ences. However, given that mental experienceswill have evolved by natural selection, we wouldexpect that behavioural choice and subjectivefeelings about the environments chosen wouldnot be completely uncorrelated. If we had aclear idea of the kinds of situations that hensprefer and the extent to which they avoid lesspreferred environments, then we would be in amuch better position to assess the mental statesof hens in battery cages. It has been shown thathens prefer an outside run to a battery cage,

once initial unfamiliarity with a strange environ-ment has worn off. On the other hand, it is as yetimpossible to say how strong this preference is .Hens given continuous access to a battery cageand a larger pen did not show evidence of prefer-ence one way or the other and the experimentsin which access to a run is possible only if thehen chooses to forgo some other commoditysuch as food or other hens (experiments 3 and4), do not enable firm conclusions to be drawnwithout further work along the same lines .

Before these laboratory experiments can beused to say anything one way or the other aboutthe mental state of hens kept under commercialbattery cage conditions, it is very important toestablish the generality of these results. Is thepreference the same whatever the method oftesting used? Do different genetic strains ofhens show different environmental preferences?Is the nature or extent of the preference affectedby the type of battery cage used, for examplea four bird cage, or by the stocking density?These questions can only be answered by furtherresearch.I feel strongly that these are preliminary resultsand should not at the moment be used either tocriticize or to support the battery-cage system forkeeping laying hens . Anyone who uses theseresults for either of these purposes is in my viewgoing beyond what our present knowledgewarrants. However, I think it is important tocontinue to research into the question of whetherhens suffer in battery cages along the lines thathave been started, and I hope that one day suchwork will enable us to take an objective andhumane judgment on life in a battery cage .

AcknowledgmentsI am very grateful to the Royal Society forPrevention of Cruelty to Animals for financialassistance and to Dr R . Dawkins for readingthe manuscript . I would also like to thank MissJ. Lomer for valuable technical assistance.

REFERENCESBrambell, F. R. (Chairman) 1965 . Report of the Technical

Committee to Enquire into the Welfare of Animalskept under Intensive Livestock Systems . CommandPaper 2896, H.M.S.O ., London .

Brophy, B. 1972. The ethical argument against the use ofanimals in biomedical research . In : The RationalUse of Living Systems in Biomedical Research .U.F.A.W., Potters Bar, Hertfordshire.

Dawkins, M. 1976. Towards an objective method ofassessing welfare in domestic fowl . Appl. Anim .Ethol., 2, 245-254 .

Duncan, I . J . H . 1974. A scientific assessment of welfare.Proc. Br . Soc. Anim. Prod., 3, 9-19 .

1046

ANIMAL BEHAVIOUR, 25, 4

Ewbank, R. 1968 . The behavior of animals in restraint .In : Abnormal Behavior in Animals (Ed. by M. W.Fox), pp. 159-178. Philadelphia : W. B . Saunders.

Gril£n, D. R . 1976 . The Question of Animal Awareness.New York : Rockefeller University Press.

Hughes, B. O. 1976. Preference decisions of domestichens for wire or litter floors. Appl. Anim . Ethol., 2,155-165 .

Hughes, B. O. & Black, A . J. 1973 . The preference ofdomestic hens for different types of battery cagefloor. Br. Poult. Sci.,14, 615-619 .

Jolly, A. 1972. The Evolution of Primate Behavior.New York: Macmillan Co.

Levins, R. 1968 . Evolution in Changing Environments.Princeton, N.J . : Princeton University Press .

Orians, G. 1971 . Ecological aspects of behavior. In :Avian Biology. Vol. I (Ed. by D. S. Farner &J. R . King), pp . 513-546 . London and New York :Academic Press.

Ryder, R. 1971 . Experiments on animals . In : Animals,Men and Morals (Ed. by S . & R. Godlovitch &J. Harris), pp . 41-82. London : Victor GollanczLtd .

Selye, H . 1952. The Story of the Adaptation Syndrome .Montreal : Acta, Inc.

Siegel, S. 1956. Nonparametric Statistics for the BehavioralSciences. New York: McGraw-Hill .

Singer, P . 1976. Animal Liberation . London: JonathanCape.

Thorpe, W. H. 1965 . The assessment of pain and distressin animals. Appendix III. Brambell CommitteeReport, Command Paper 2896, H.M.S.O ., Londonpp. 71-79 .

Thorpe, W. H. 1967. Discussion to Part II. In : Environ-mental Control in Poultry Production (Ed. byT. C. Carter). British Egg Marketing BoardSymposium No . 4. Edinburgh : Oliver & Boyd.

Wall, P. D. 1974. 'My foot hurts me' : an analysis of asentence . In : Essays on the Nervous System. AFestschrift for Professor J. Z. Young (Ed. by R.Bellairs & E. G. Gray), pp. 391-406. Oxford :Clarendon Press .

Wood-Gush, D. G. M. 1971 . The Behaviour of theDomestic Fowl. London: Heinemann.

Wood-Gush, D. G. M. 1972 . Strain difference in responseto sub-optimal stimuli in the fowl . Anim. Behav .,20, 72-76.

Wood-Gush, D . G. M. & Gilbert, A. B. 1969 . Observa-tions on the laying behaviour of hens in batterycages. Br. Poult. Sci., 10, 29-36.

(Received 17 November 1976 ; revised 26 January 1977 ;MS. number: 1585)