ワイヤー方式の電気牧柵の構成要素がヤギの忌避行動に及 ぼす影響
誌名
誌名 Animal behaviour and management ISSN
ISSN 18802133
著者 著者
柿原, 秀俊 石若, 礼子 増田, 泰久 中野, 豊 泉, 清隆 堀江, ちひろ 古澤, 弘敏 下條, 雅敬 巻/号
巻/号 51巻3号
掲載ページ
掲載ページ p. 121-130 発行年月
発行年月 2015年9月
農林水産省 農林水産技術会議事務局筑波産学連携支援センター
Tsukuba Business-Academia Cooperation Support Center, Agriculture, Forestry and Fisheries Research Council Secretariat
OriginarArticl~ レム
The e f f e c
恰o fi n d i v i d u a l componen
恰ofan e l e c t r i f i e d wire fence on avoidance behaviour by goa
恰Hidetoshi KAKIHARA'， Reiko ISHI"砧KA2，YasuhisaMASUDA3， YutakaNAKANO'， Kiyotaka IZUMI'， Chihiro HORIE'， Hirotoshi FURUSAWA'， Masataka SHIMOJO'
'Graduate School ofBioresource and Bioenvironmental Sciences， Kyushu University， Fukuoka， 812・8581，Japan 2Faculty of Agriculture， University ofMiy但 北i，Miy低 北i，889‑2192， Japan
3Kuiu Grassland Ecomuseum， Fukuoka， 810・0033，Japan
*Corresponding au血or.E四mailaddress: mshimojo@agr.勾rushu‑u.acj.p
Abstract
Two experiments were performed to test the hypothesis that fence posts as well as s仕 組dswill affect也e behaviour of goats that have been conditioned to avoid an electrified wire fence. In也efrrst experiment， the reaction of fo町 doesto血efence with an electric s仕 組dand posts drawn across the whole width of the paddock (C condition) was compared with their reaction to the fence with an electric strand terminated at half the width of the paddock (L姐 dR conditions). The does did not cross the fence line under ei血ercondition. In the second experiment， 仕leavoidance behaviour of fo町 doesto two kinds of posts ('familiar wooden post' and standard fibre‑reinforced plastic (FRP) post') with and without an electric s仕 組dwas tested.百lUS，conditions compared were familiar wooden post only (F)， F with an electric strand (W)， standard FRP posts only (P)組 dP with an electric strand (WP). The fence line was crossed in 94， 44， 25， and 6 % oftrials for F， W， P， and WP， respectively. In conclusion， based on the proportion of crossings the fence line，血estandard FRP posts， as with the el即 位icstrand， increased the avoidance behaviour.
Key words: Avoidance behaviour， Crossing behaviour， Electric fence， F巴ncecrossing， Goat
Animal Behaviour and Management， 51 (3): 121・130，2015 (Received 22 December 2014; Accepted for publication 25 May 2015)
Introduction
日ectricfences紅 巳 oftenerected around grazing areas of ruminants such as cattle， sheep and goats (e.g. Kilgour & Dalton 1984; McKillop & Sibly 1988; Jaudas & Mobini 2006; M紅白skainenet al. 2008; Umstatter 2011; Markus et al. 2014). However， such fences are used wi也varyingsuccess for goats because there are often some individuals也atwill exIt the fence (Jaudas & Mobini 2006). To confine goats in an electric fence， basic knowledge of也e fence components白紙improvethe e伍cacyof也efence in stimulating goat avoidance behaviour is required. Researchers and producers have investigated也atthe number of s仕 組dsas well as their spacing affects the
6伍cacy of an elec出c fence in confining various species of 姐 imals (McKillop & Sibly 1988; VerCauteren etα .l2006; Reidy etα. 12008; Goetsch et al. 2012). It is a natural ass田nptlOn也atanimals will avoid an elec出cs仕 組dbecause it administers el即 位lC shocks. and也atan electric fence will be effective because of也eavoidance of the el即 位ic wire by
animals; however， Goetsch et al. (2012) reported that 也ee伍cacyof an electric fence varied considerably according to conditions 0也er也anthose mentioned above. Effects of various conditions (i.e.， experimental designs姐 dtraining procedures) in efficacy have also been reported specifica1ly in goats (Tsukahara et al. 2013). These resu1ts contributed to our interest that goats assess electric fences for components other也 阻 S仕 組dswhen avoiding el即 位icfences. We hypothesised that other components besides the strands affect也e avoidance behaviour by goats of electric fences.τbe objective of也isstudy was to investigate the effects of individual components of an el即 位ifiedwire fence on the avoidance behaviour of goats. In也ecurrent study， except for crossing， all responses of goats to an el即 位ifiedwire fence were considered as avoidance.
Materials and Methods
Study location， animals and conditioning for electric fence avoidance
AVOIDANCE OF ELECTRIC FENCES BY GOATS
All of the experiments were carried out at University Farm， K戸lshu University (Fukuoka， K戸lshu，Japan; 33.620N la ，.t 130.460E long.， at 9 m
above sea level). The c町Tentstudy complied with也e policy of the Guidelines for Ethical Treatment of Animals in Applied Animal Behaviour and Welfare Research (ISAE E也csCommittee 2002)阻 dwith that of the Regulations for Animal Experiments at Kyushu University. The goat herd used in the current study comprised fo町 femaleJapanese native goats (Tokara goat， Capra hirc凶).Their estimated ages were 3‑4 years at the begimling of the study， and the height of their withers ranged企om49 to 52 cm. No animals had previously been exposed to an el即 位icfence
Consistent with previous studies (Niven & Jordan 1980; McDonald et al. 1981)， prior to the first experiment， all the does were conditioned in出err resided paddock to become used to an electric fence. The elec出cfence used in this conditioning comprised two fibre‑reinforced plastic (FRP) posts measuring 120 cm in height and 14 m m in diameter (STF‑1414; standard FRP posts')， one stranded 2 m m diameter poly wire (G出 erCord)， two insulators (14 m m Hook) and an energiser (S仕oboDC‑12B). All equipment used was manufactured by Suematsu Electronics Co.， Ltd.， Yatsushiro， Japan. Two standard FRP posts were erected 2 m apぽt，and a s仕 組dwas drawn between the posts 25 cm above也eground. The voltage applied to 也e electric s仕 組ds in our fencing system was maintained at 4.0‑5.5 kV Each conditioning was individually conducted to ens町e independent behavioural reactions by the does. Subsequent to m仕oductionto the paddock， each doe made initial contact with the strand and showed a 甜ongavoidance response (e.g.， jumping away). This exposぽ e to electric str組 dwas repeated on sunny days between 13 September and 24 September 2010. Each doe was exposed to the fence once or twice a day for 20 minutes each between 12.00 hours and 18.00 hours Consequently， each doe was subjected to the conditioning exposure 13 times in total unti1 the does avoided the fence completely. We removed the fence each day at the end of the conditioning， so that the does were on1y exposed to担ydevice relating to ele出lC fence during也econditioning period. In the current study， avoidance behaviour was defined as either組
animal did not cross血efence line or took a detour. The fence line was defined as a line of an electric s仕 組d and/or between experimental posts placed.
Experiment 1
First， we investigated whether the does were exclusively avoiding 配 s仕 組ds when exhibiting avoidance behaviour in response to the presence of an electrified wire fence. If the avoidance of such a fence is solely due to the electric s仕 組ds，the removal of these s仕 組dsshould mean也at也edo巴sare no longer confined
Experimental plot and fence in experiment 1 A 200 m2 paddock with an adjacent pen (wi也 m area of 15 m2) w出 usedas the experimental plot (Fig.
1).百lepen was connected with the paddock by a lockable door. The paddock was enclosed by a non‑el即 位ifiedwire‑mesh fence of 120 cm in height The dominant plant species in the paddock were annual bluegrass (Poa annω)組 drat's‑tai1 fescue (均争lG
myuros)白atvegetated during the winter season;姐 d baluagrass (Pa.平alumnotatum) and yellow foxtailgrass (Setaria glauca)也atvegetated during the SUfililler season.In也isexperiment， three conditions， namely the control (C condition [Fig. 2C])，也eleft‑half fence (L condition [Fig. 2L])姐 d也e right・half fence (R condition [Fig.. 2R])， of experimental fence were designed using an energiser (Strobo DC・12B)，poly
W町es(Getter Cord; diameter 2 mm)， tlrree standard FRP posts (STF・1414)姐 dinsulators (14 m m Hook， all manufactured by Suematsu El即 位onicsCo.， Ltdよall of which were也e s創ne as those used during the conditioning. In the C condition， tlrree standard FRP posts were erected at intervals of 4 m ap紅tand 8 m long fence strand was positioned across the posts at 25 cm above the ground in也eexperimental plot， so也at an 8 m x 16 m rectangul紅 紅 巳awas separated企omthe rest of也epaddock and the pen. Consequently， the fence line in也isexperiment was equal to也eline of 血es佐andin血eC condition because， as mentioned above，也efence line was de白1巴d as a line of組
electric strand and/or between experimental posts placed. However， in the L condition， the s仕 組d terminated at the centre of the fence line so that there was no s仕 組dbut merely standard FRP posts on白e right half of the fence. Under the R condition， the fence was designed by altering the s仕 組dposition， that is，出e left half of the fence was open. If the avoidance of也e fence line by the does was solely due to the s仕 組d，也e does would cross也eline both under the L condition
組 d出eR condition， and avoid the line under the C condition.
Procedure in experiment 1
Each trial was performed in the following sequence. Initially， all the animals of the herd were driven into the pen and one of the conditions of the fence was set up on the fence line. After the standard FRP post that was closest to也epen was temporarily removed， each doe was individually moved企omthe herd to the opposite side of the fence line (Fig. 1). The removed post was replaced inlmediately after the target doe passed 由roughthe fence line. The behaviour of the target doe， which would be constantlyむyingto 印 刷rnto the herd， was recorded with a digital video c組 問ra(DCR‑SR300， Sony Corporation， Tokyo， Japan).百le observation ceased either when any crossing occurred or when 20 minutes had passed， a食er which the doe was returned to the herd. During the obse町atlOn，也e experimenter stayed out of doe's sight at the greatest extent.
KAKIHARA， ISHIWAKA， MASUDA， NAKANO， IZUMI， HORIE， FURUSAWAAND SHIMOJO
Initially， the reaction of does to也efence under the L condition was compared wi也 也e汀 reactionto the fence under the C condition， and subsequently， the reaction of does to the fence under the R condition was compared with their reaction to the fence under the C condition. In bo也compansons，仕le出alwas repeated 12 times for each condition， with也reereplications per individual. To minimize any order effect， two does were observed altemately under the C condition阻d 也e位eatedfence condition (i.e.， L and R conditions)，
Experiment 1
8.0m Entrance
16.0m
SE
I E a E
hほHUAue rfc
・ 時 ︑
o n R h
10.9m
whereas the other two were observed altemately in也c reverse order. Thus， each doe was subjected to a total of six trials in each comparison， 姐d12出alsin白1S
experiment. The comparison between
L
condition and C condition was conducted on four sunny days between 30 September and 21 October 2010. The comparison between R condition and C condition was conducted on four sunny days between 22 October and 9 December 2010. Each doe was tested once or twice a day between 12.00 hours and 18.00 ho町s.Experiment 2 8.0m
12.0m
13.6m
6.0m
Lockable door
Figure 1 The conceptual scheme of experimental plots in experiment 1 (Ieft) and in experiment 2 (right). The wavy lines and the black circles indicate experimental fence lines and positions on which standard fibre‑reinforced plastic (FRP) posts and/or familiar wooden posts were set up， 陪spectively. The solid lines show either non‑electrified wire‑mesh fences enclosing the experimental paddock or walls of the experimental pen. In experiment 1， each doe was separately released from the pen to the opposite side of the fence line by tempora叩 removalof the nearest standard FRP post to the pen (dotted line with arrow); all the remaining does were left in the pen. In experiment 2， after being released from the pen through the lockable door， the target doe was introduced into the paddock through the ent悶nce(do世edline with arrow). Except where the target doe goes around the fence line in experiment 2， the doe should cross the fence line to go back to the rest of the herd in both experiments (dashed line with arrow).
AVOIDANCE OF ELECT即CFENCES BY GOATS
( W )
( F )
(WP)
1そ
Figure 2 Experimental fence conditions. AII the fences were composed of one stranded poly wire 2 m m in diameter (horizontal do仕ed lines) with or without three fibre‑reinforced plastic posts measuring 120 cm in height and 14 m m in diameter (vertical black lines， called 'standard FRP posts') and with or without three wooden posts familiar to the does measuring 30 cm in height and 12 m m in diameter (ve吋icalgrey lines， called 'familiar wooden posts'). (C) Control: This fence consisted of the strand at 25 cm above the ground and three standard FRP posts spaced 4 m apa凡(L)Left half， (R) Right half: These were similar to the control fence design except that the strand terminated at the middle of the three posts. (F) Familiar: This fence consisted of nothing but three familiar wooden posts spaced 2 m apart. (W) Wire: This fence consisted of three familiar wooden posts and a strand at 25 cm above the ground. (P) Pos
: t
This fence consisted of three familiar wooden posts and three standard FRP posts. (WP) Wire and pos: t
This fence consisted of three familiar wooden poぬ， a strand 25 cm above the ground and three standard FRP posts. In combination with the C condition， the L condition and the R condition were used in experiment 1. The other four conditions of fences were used in experiment 2. The wavy lines indicate experimental fence lines， one or both ends of which were terminated to the non‑electrified wire‑mesh fence surrounding the experimental paddock， as shown in Fig. 1.KAKIHARA， ISHIWAKA， MASUDA， NAKANO， IZUMI， HORIE， FURUSAWAAND SHIMOJO
Experiment 2
In也isexperiment， we focused on standard FRP posts as an additional probable factor influencing也e avoidance of the el即 位icfence by也edoes because， like也estr姐 d，血eposts comprise a fund沼 田ntal姐 d prominent component of組 elec位i:fiedwire fence. In addition， some modi:fications were made in experiment 2. The similarities of experimental procedぽesbetween the conditioning and experiment 1 may cause the does to avoid crossing the fence line. The fence line was de:fined across也ewhole width of the paddock in expeI世 間nt 1， so that也e experimenter had to temporarily remove one of the three posts企omthe fence line whenever the target doe was moved to the opposite side of也巴line.From白isaction， a doe may learn to avoid句ringto cross the el即 位icfence and rather wait until the出alis over. To reduce these probable e能cts，in experiment 2， the target doe was introduced into the paddock through a newly made en仕 組ceand was allowed to go around the fence line to get back to the pen， bo也whenit avoided the line and when組 observationceased
Experimental plot and fence in experiment 2 The anirnals and experimental plot in也lS experiment were也es創neas in exp巴riment1. However， instead of也elockable door，血epen and the paddock were connected by a top‑hinged， one‑way swinging door which allowed the does to enter the pen but not exit丘omit. As shown in Fig. 1， a lockable door located at the opposite side (back) of the pen was used to remove the target doe企omthe pen. An entrance to the paddock was constructed at the most remote comer from也epen (Fig. 1). A 4 m long fence line (half the width of the paddock) was de:fined midway between
也enew entrance to the paddock and由epen， one end of which temIinated at the wire‑mesh fence surrounding the paddock (Fig. 1). Given出atthe does behaved in a way s白血l紅 tothat in experiment 1， they could not avoid encountering this fence line at about 4m企om也eposition at which the line had been in the previous experiment. The does could make a detour through the opening next to仕lefence line to re加nto 也eirherd， both when也巴yavoided the line and when the observation ceased (Fig. 1).
To detemIine也eeffects of the strand alone and the effects of the standard FRP posts， we constructed wooden posts with a diameter of 12 m m (similar to the standard FRP posts)， which were 30 cm in height. These posts had been in position in the resided paddock for a month. Consequently， by the beginning of experiment 2， it appe紅 巳das though all也edoes had become farniliar with these posts (hereinafter， referred to as 'farniliar wooden posts'). In也isexperiment， four conditions were established using three of these farniliar wood巴nposts， which wer巴setup on the fence line (Fig. 2). To enhance也epresence of posts， the intervals between 叫jacentposts were shortened to 2 m， which were half of those in experiment 1. Under the
fanriliar condition (F condition)， only the farniliar wood巴nposts were set on the fence line. In addition to
白reefarniliar wooden posts， the experimental fences composed of a s仕 組dof wire under the wire condition (W condition)， three standard FRP posts under the post condition (P condition) and a s仕 組dand白reestandard FRP posts under也.ewire and post condition (WP condition) were set. In the W and WP conditions，也e strand was drawn across all of血efanriliar wooden
抑 制at25 cm above the ground， and in也eP and WP conditions，也巴 standardFRP posts were erected in proximity to the farniliar wooden posts. Ideally， the F condition should act as the negative COll'住01，with the proportion of observations in which the does crossed the line approaching 100%. In contrast， the WP condition should ideally act as the po叫lvecorr仕01， where the proportion of observations in which the does crossed the line approaching 0%. The proportions of observations in which the does crossed the line under
也eW condition and the P condition would depend on
白 日 能ctof the presence of the s仕 組dand出es.tandard FRP posts. These proportions would be not lower than that under the WP condition， and not higher出 組 曲at under the F condition. The energiser， poly wire， standard FRP posts， and the insulators were the same as those used in the conditioning.
Procedure in experiment 2
Each出alinvolved two steps. Initially，也edoes were exposed to the same experimental setup as in the F condition to住ainthem to cross the fence line and consequently to reduce the effect of the previous仕ial (Step 1)， and subsequently，也eirreactions to any one of 也巴fourdifferent fence conditions were observed (Step 2). Step 1 was performed in the following sequence. A target doe was released企om血epen through也e lockable door and in柱。ducedinto the paddock wi也 也e F condition由roughthe err仕 組ceat仕lefar side of血e paddock， 組dthen the experimenter shut the entr叩 ce and remained out of view of the doe at the greatest extent.τms step ceased when the doe retumed to也e pen白rou出血eone‑way door (Fig. 1).
Step 2 was the observation step and performed in the following sequence. Before the release of the target doe企om也epen， one of the conditions of the fence was set up on the fence line. In the same way as in the :first step， the target doe was in仕oducedinto the paddock， with the experimenter staying out of view of the doe.百lebehaviour of the target doe， which would be constantly句ringto re加mto也eherd， was recorded with a digital video camera (DCR‑SR300， Sony Corporation) .百世sobservation ceased either when the doe went back to the pen through the one‑way door or when 10 minutes had passed. In the latter case，白edoe was brought back to the pen佐賀oughthe opening next to the fence line.
This位ialwas repeated 16 times for each condition， with four replications per individual. To mininuze any order effect， the order of conditions Was randomly
AVOIDANCE OF ELECτ'RIC FENCES BY GOATS
detennined for each doe by means of a 10悦 η.Thus， each doe was subjected to a total of 16出alsin也lS experinIent. This experinIent was conducted on也企teen sunny days between 1 April 2011 and 13 May 2012 Each doe was tested once or twice a day between 12.00 hours and 18.00 hoぽ s.
Data analyses
A11 data were collected企omthe video recordings. A11仕切lsin experinIent 1 were classified according to whether the t紅get doe crossed the fence line. Furthermore， it was detennined in each trial whether 也.edoe approached the half of the fence line without s仕 組ds. Taking the visual acuity (Hughes &
京市i仕 切dge1973) and也ebody size of the does into account， it was estimated也atthe does were able to perceIVe也e s仕andwhen their企onthooves were located less也 組 2m from the fence line. Therefore， when the doe had approached close enough to the fence so that the right and le食 合onthooves were placed Wl白血2m of the line segment，出iswas counted as也e doe having approached'也efence.
A11仕ialsin experinIent 2 were classified according to whether the target doe crossed the fence line. The
紅白lsin which the doe did not cross the fence line were subdivided according to whether the doe detoured around the line to get back inImediately to the pen or the doe did not go back to the pen until出etrial ceased， bo也ofwhich were considered as avoidance behaviour. The number of仕ialsin which the doe crossed the fence line under each condition was counted for each doe. The proportion of trials in which the fence line was crossed was calculated by dividing the crossing count by the number of trials required under each condition.
In experinIent 2， the data were analyzed using the glm紅 白nction of the lme4 package version 1.1・7 (Bates et a .l2014)血 也eR statistical package version 3.10.0 (R Core Team 2014). To exarnine whether the existence of the s回ndand/or the standard FRP posts decreased the probability of the fence line being crossed， we自社eda generalized linear mixed model (GLMM) with a binomial distribution and a logit link function.百lepresence or absence of the s仕 組dand the standard FRP posts and 也 巴 町 interaction were incorporated into 0ぽ modelas a fixed effect， and each doe was fi枕edinto the model as a random effect because individuals were repeatedly measured. The白H model is given by
logit(句ωqω)=s此0+s
where q i岱st也h巴p戸>roba油bi出li句勿r白tha瓜ta doe crosses t也hefe切nce lin眠e伐;W， P and W:P are variables indicating也.epresence or absence of the s仕 組d，the presence or absence of the standard FRP posts and their interaction， respectively (presence = 1; absence = 0); so is the intercept; s" s2 and s3 are 白巴 coefficients of W， P and W:P， respectively; and r is the random effect.
In general，也efit of any model can be improved by increasing the number of parameters; however， the ideal model wou1d be appropriately simple， based on
concepts of parsimony (BUl吐1創n& Anderson 2002). In addition， when there are m佃ycandidate models， model selection in a context of nu11 hypothesis testing is inappropriate because of the multiple四testing problem. For these reasons， the best model was selected on the basis of minimum Akaike information criterion (AIC)， a smaller value of which indicates a better fit and a more parsimonious model. In addition， using the lsmeans function of the lsmeans package version 2.14 (Len也 &Herve 2015)， we calcu1ated least squares (LS) means for each出 a回 ent釦ddetennined 也esignificant differences between these LS means. Then， the values of LS means on the logit scale were back仕 組sformedto批 original(probability) scale.
Results Experiment 1
In the comparison between L condition and C condition， none of the does came wi白血 2m of the right half of the fence line (where the s仕 組dwas absent) or passed across白efence line under either the L condition or也eC condition. In出ecomparlson between R condition and C condition， two does came Wl也in2 m of the left half of the fence line (where也e s仕 組dwas absent)血 two and one of曲目etrials， respectively， under the C condition. Sirnil紅ly，two does approached the fence line in two trials and the other two did in oneむialunder the R condition. However， no crossing was observed under any condition in the comparison R condition and C condition， as was observed in the comparison L condition and C condition.
Experiment 2
In most of血etrials， the does ran towards the pen inImediately after their introduction into the paddock. During the住ialsunder the W condition， the does sometimes紅白dcrossing the fence line and clearly stumbled over也estrand. The behavioural responses of the does in experinIent 2 were classified into three categories， as shown in Table 1. In all except one trial， the does crossed the fence line under the F condition. Under the W condition. three does crossed也efence line in two of four trials， and the remaining doe crossed only in one出al. Under the P condition， one doe crossed the line in two of four出als，two does did so in one trial and the remaining doe one did not cross出e line in four trials. 0nly one crossing of the fence line was observed under the WP condition. However， the doe白atcrossed the line in出istrial appeared to a抗emptto stop right in企ont of the line but the momerr加m carried it over the s仕 組das was seen in some trials under the W condition. In descending order in terms of condition， does passed across the fence line in 94%， 44%， 25%組 d6% ofthe佐ialsunder the F， W， P姐 dWP conditions， respectively (Table 1).
KAKIHARA， ISH:rn戊KA，MASUDA， NAKANo， IZUMI， HORIE， FURUSAWAAND SHIMOJO
T'able 1. Behavioural陪sponsesof does to experiment創fencelines in experiment 2
Condition F W P W P
Trial 1 2 3 4 1 2 3 4 2 3 4 Goat1 + + + + +
一 一
+ + DGoat2 + + + + ー D + ー D D D ー Goat3 + + + + + + +
一
ー +Goat4 + ー + + + D + ー D +
Proportion of
94 44 25
crossing (%)
1 2 3 4 D D D ‑ D D D ‑
+ ー
6 The plus symbol (+) indicates that the doe crossed the fence line; the minus symbol (一)indicates
that the doe did not go back to the pen until the trial ceased; the 'D' symbol indicates that the doe detoured to go back immediately to the pen.
Table 2. Estimated parameters of the generalized linear mixed model for the proportions of trials in which does crossed the fence line in experiment 2
t
Estimate (Std. error)
s 1 s 2
Model
戸
。
s 3
a u Aus・h
quAU e R e FM
n H
a
ve
DAIC
s o
+払W+ s 2
P+ s 3
W:P+
r 2.71 ‑2.96 ‑3.81 1.35 26.69 11 36.69 ( 1.03 ) ( 1.15 ( 1.18 ) 1.65戸。+
s 1
W+ s 2
P+
r 2.29 情2.42 耐3.24 27.34 12 35.34 ( 0.76 ) ( 0.83 ( 0.85s o + s 1
W+
r 0.38 ぺ.48 51.02 13 57.02 ( 0.36 ) ( 0.54 )s O + s 2
P+
r 0.79 ‑2.47 39.29 13 45.29( 0.38 ) ( 0.62 )
s o
+ r ‑0.32 白 幽 幽 58.96 14 62.960.25
↑: Counts of crossing range from a minimum of 0 (a doe did not cross in any of the trials) to a maximum of 4 (a doe crossed in all trials).
W， the existence of the strand (presence
=
1; absence=
0); P， the existence of the standard FRP posts (presence = 1; absence = 0);s o
， intercept;恥‑ s 3
，the paramete陪 ofthe fixed effects; r， the random effect for individuals.As shown in Table 2，出eGLMM with the lowest AIC (the best model) included the variables for th巴
presence or absence of the s仕 組d(W) and the standard FRP posts (P). In也ismodel， the intercept and the coe伍cientsfor the variables W and P were estimated to be 2.29. ‑2.42 and ‑3.24 with a stand紅derror of 0.76 0.83組 d0.85， respectively. The devi姐 cewas 27.34 on 12 degrees of企eedom. The LS me佃 s were significantly different for the strand仕 切 回ent(0.67 with a standard e汀orof 0.50 for the absence of也e s仕 組dversus ‑1.75 wi也 astandard eηor of 0.57 for
the presence; t = 2.91， Pく 0.05)and for也epost
仕ea加lent(1.08 wi也 astandard error of 0.48 for the absence ofthe posts versus ‑2.16 with a standard error of 0.60 for the presence; t = 3.83， Pく0.05).The values of LS means on the original (probability) scale were estimated at 0.66 (68% confidence interval (CI) 0.54・
0.76) and 0.15 (68% CI 0.09・0.23)for the absence and presence of the strand， 組d0.75 (68% CI 0.65幽 0.83)阻 d0.10 (68% CI 0.06・0.17)for也.eabsence
組 dpresence of the posts， respectively.
AVOIDANCE OF ELECTRIC FENCES BY GOATS
Discussion
h 血e current study， two experiments were performed to test our hypothesis that when goats avoid
組 electrifiedwire fence， posts of the fence have an e能cton也eiravoidance in addition to the strand. First， in experiment 1， we investigated whether the avoidance of such a fence is solely due to the s仕and.In the comparison between L condition and C condition， none of the does passed across the fence line under either也巴 L or也eC condition. This result may indicate that the doe avoided something other than the strand. However， it may just have been that也.edoes had not discemed 也巴disappe紅anceof the strand企omthe experimental fence. This is because， in all of the trials under the L condition，也.eydid not come within 2 m of也eright half of the line where there was no s佐and. In也e comparison between R condition and C condition， all foぽ of出edoes approached the line segment at least once， indicating白 紙 也ey would have had an oppo比四百句rto perceive the absence of也e s仕 組d. Never出eless，the does did not cross the fence line under any condition in the comparison between R condition and C condition. The absence of crossing of the fence line in experiment 1 may support也e expectation based on our hypothesis. On the other hand， in addition，也eresults may suggest也atthe does could have been refusing to cross the fence line regardless of the condition in any of these comparisons， simply because the experimen臼1procedures in experiment 1 were similar to that in the conditioning. In experiment 1， each doe was taken to也epaddock through也.esame lockable door as used as in the conditioning. Another possibility is that after the first trial， they had leamed to wait until the trial was over
Experiment 2 was designed to avoid the aforementioned uncertainties， as well as to determine the components of the electrified wire fence也attrigger avoidance by the do巴s. To prevent the does企om simply waiting for the trial to end，也eywere allowed to
仕avelonly in one direction (to the pen)， and to make a detour around也efence line to也epen. Thus， the doe had a choice of crossing or going around the fence line on every出al;when the doe did not do so after a
10‑minute period of observation， an experimenter也en brought the doe back to也epen through the opening next to也efence line. If也edoes had waited for the trial to end or had made a detour around the fenc巴line regardless of the condition， the proportion of trials in which the does had crossed the line would have uniformly approached 0% under any condition. In fact， this proportion varied with the conditions， which is supported by the best model也atincluded the presence or absence of the strand and presence or absence of the standard FRP posts目 Additionally，the proportion of crossrngs血eline was very high under F condition and it was very low under the WP condition， as we expected.百leseresults clearly indicate that the does passed across the fence line when they were able to
cross也eline without di伍c叫tyand not when they experienced di伍cultyin doing so.
In experiment 2， if the avoidance of an electrified wire fence by does had been solely due to也ep.resence of a s仕 組d，the proportion of仕ialsin which they crossed血巴 fenceline under也eW condition should have been nearly the same as也atunder the WP condition， which was the positive contro. lHowever， the proportion of crossings under the W condition (44%) was much higher也an that under the WP condition (6%). This result indicates也at也es仕 組dis not the only factor affecting the avoidance behaviour of does. If the standard FRP posts had no effect on白e avoidance of an electrified wire fence by does， the proportion of crossings under the P condition should have been nearly也esame as that under the F condition， which was the negative control. However， the proportion of crossings under the P condition (25%) was much lower than也atunder the F condition (94%). This result indicates that， like the s仕 組d，the standard FRP posts also have an effect on the avoidance behaviour of does. The GLMM analysis also confirmed 白atthe presence of standard FRP posts is a factor affecting the avoidance， in addition to也es仕 組ds.百le best model included the variables for the presence or absence of the s仕 組dand血estandard FRP posts but not their interaction. The negative coe伍cientsfor these variables in this model indicate也at也eprobabili勿of the does crossing the fence line is decreased by the presence of也estrand and the standard FRP posts， independently of each 0也er.
As described in也ern仕oduction，an elec出cfence appears to work effectively by means of the avoidance of也es仕 組dby animals. In fact， at least during the early stage after initial conditioning to an electric fence， the avoidance behaviour is only partly due to the s仕 組d as shown in experiment 2. Posts， which are as fundamental and prominent components of the fence as the strand， also have an effect on也巴avoidanceof the fence by the goats， even though they fail to administer an ele位icshock. These results imply that the does focus on possible shocks given by both el即 位lCS仕 組ds and posts. Lee et al. (2007) reported也atca抗lecould be conditioned to associate an aversive stimulus (elec出cshocks given by shock collar) wi也 atrough. Similarly， Schilder and Van del Borg (2004) inferred 出ata dog trained using a shock collar would expect an aversive event (electric shocks) to occur whenever its handler was around. Al白oughthere are differences between species and kinds of electric shock device used，血 our experiments， the does also expected an aversive event to occur whenever there were standard FRP posts on也efence line. Adding posts or a sirnilar component， such as battens， to an el即 位ifiedwire fence could increase也eavoidance behaviour to也efence， as well as improve the physical strength of the fence Although only s仕 組dsand posts were investigated in 出sstudy， oth巴rcompone凶s，such as insulator and energiser could be investigated. Therefore， without
KAKIHARA， ISHIWAKA， MASUDA， NAKANO， IZUM!， HORIE， FURUSAWAAND SHIMOJO
increasing the number of s甘andsor their spacing， the e伍cacyof such a fence could be improved by adding some prom立lentcomponents associated with electric shocks by animals even if those components fail to administer an electric shock.
Conclusions
We concluded也atwhen goats avoid an electrified wire fence， their avoidance is not solely due to也e electrified s仕 組ds. Posts， at least among 白e components of such a fence， also have some effect on the avoidance behaviour.
Acknowledgements
This paper is based on a poster presentation at the 45th Congress of the Intemational Socie勿forApplied Ethology. The presentation was supported in part by a gr組 t企om也巴 JapaneseSociety for Applied Aninlal Behaviour.
References
Bates D， Maechler M， Bolker B， Walker S. 2014. lme4: Linear mixed‑effects models using Eigen and S4 [homepage on也e Intemet]. R Foundation for Statistical Computing， Vienna， Austria; [ cited 4 February 2015]. Available 企om URL:
htや://CRAN.R‑project.org/package=lme4.
Burnham KP， Anderson DR. 2002. Model selection and multimodel iゆrence:αpractical information‑ theoretic approαch. 2nd ed. Springer， New York. Goetsch A， Detweiler G; Puchala R， Sal由 T，Gipson T.
2012. Conditions to test electric fence additions to cattle barb wire fence for goat contaInment. Journal of Applied Animal Research 40
，
43・55.Hughes A ， Whitteridge D. 1973. The receptive fields and topographical organization of goat retinal ganglion cells. Vision Research 13
，
1101・1114. ISAE Ethics Committee. 2002. E出icalTreatment ofAninlals in Applied Animal Behaviour Research [homepage on the Intemet]. The Intemational Society for Applied Ethology， Edinburgh; [cited 2 April 2014]. Available 企om URL:
坤://www.applied‑抽 ology.org/ ethical ̲guidelines. h1:I吐
Jaudas U， Mobini S. 2006. The Goαt Handbook Barron's Educational Series Inc， New York.
Kilgour R， Dalton C. 1984. Livestock Behaviour: A Practical Guide. Granada， London.
Lee C， Prayaga K， Reed M， Henshall 1. 2007. Methods of仕ainingcattle to avoid a location using el即 位ical cues. Applied Animal Behαviour Science 108
，
229‑238
Lenth ，R He四e M. 2015. lsmeans: Least‑Squares Means [homepage on the Intemet]. R Foundation for Statistical Computing， Vienna， Austria; [cited 4 February 2015]. Available 合om URL htや//CRAN.R ‑project.org/package=lsme創18.
Markus SB， Bailey D W， Jensen D. 2014. Comparison of elec住icfence and a simulated fenceless control system on cattle movements. Livestock Science 170
，
203‑209.
Martiskainen P， Tuomisto L， Huuskonen A， Mononen J. 2008. Training dairy b叫1calves to stay within light‑built electric fences. Agricultural and Food Science 17
，
121・126.McDonald CL， Beilliarz RG; McCutchan JC. 1981 Training cattle to control by elec出cfences. Applied Animal Ethology 7
，
113・12l.McKillop IG; Sibly RM. 1988. Aninlal behavior at electric fences and the implications for management. Mammal Review 18
，
91・103.Niven DR， Jordan DJ. 1980. Electric fencing for feral goats. QueensZαnd Agricultur，α1 Journal 106
，
331‑332.
R Core Team. 2014. R: A Language and environment for statistical computing [homepage on the Intemet] R Foundation for Statistical Computing， Vienna， Aus出a;[cited 5 July 2014]. Available企omURL:
h句://www.r‑project.org/.
Reidy M M， Campbell TA， Hewitt DG 2008. Evaluation of el即 位icfencing to inhibit feral pig movements. The Journal ofWildlife Management 72
，
1012‑1018.
Schilder M， Van der Borg J. 2004. Training dogs wi也 he恥 ofthe shock coll訂 short and long term behavioural effects. Applied Animal Behaviour Science 85
，
319・334.Tsukah釘aY， Detweiler GD， Sahlu T， Gipson TA， Goetsch AL. 2013. E紅白ts of conditions of an experimental model to evaluate methods of electric‑fence strand addition to barbed‑wire fence to contain goats. Journal of Animal Science 91
，
4476‑4485.
Umstatter C. 2011. The evolution of virtual fences: a review. Computers and Elecff切1icsin Agriculture 75
，
10・22.VerCauteren KC， Lavelle MJ， Hygns仕omSE. 2006 Fences and deer‑damage management: A review of designs and巴伍cacy.Wildl~作品ciety Bulletin 34
，
191‑200
AVOIDANCE OF ELECTRIC 回NCESBY GOATS
ワイヤ一方式の電気牧柵の構成要素がヤギの思避行動に及ぼす影響
柿原秀俊1・石若礼子2・増田泰久3.中野豊1・泉清隆1.堀江ちひろl・古津弘敏 1.下僚雅敬1 1九州大学大学院生物資源環境科学府、福岡市 812‑8581
2宮崎大学農学部、宮崎市 889‑2192
3久住牧野の博物館、福岡市 810‑0033
*Corresponding author. E‑mail address: mshimojo@agr.匂rushu‑u.ac.jp
要 約
ヤギがワイヤ一方式の電気牧柵に対する忌避学習をする際に、架線に加えて支柱が彼らの忌避 に影響を及ぼしているという仮説を検証するため、 2つの実験を行った。まず始めに、ヤギの電 気牧柵に対する忌避が架線のみに起因するのかどうかを確認するために、パドックの両側柵を結 んで立てた全ての支柱に一本の架線が取り付けられた牧柵に対するヤギの反応 (C区)を全長の 半分にのみ架線が取り付け、架線を避けて通過できるようにした牧柵に対するそれと比較した (L 区および R 区)。しかし、いずれの処理区においてもヤギは牧柵を通過しなかった。次に、ヤギが ワイヤ一方式の電機牧柵のどの構成要素を忌避しているかを特定するために、ヤギが警戒を示さ なくなった支柱(思11致支柱)および絶縁された支柱(電牧支柱)、架線を組み合わせて4つの処理 区を設け、それらに対するヤギの反応を比較した。それぞれの処理区の牧柵は、馴致支柱のみ (F 区)、馴致支柱および架線(W区)、馴致支柱および電牧支柱 (p区)、思11致支柱および架線・電牧支 柱 (WP区)を用いて作成した。その結果、各処理区の全観察のうち、ヤギはF区では94%、W 区では44%、P区では25%、WP区では 6%の頻度割合で牧柵を通過した。この結果は、架線と 同様に、電牧支柱もヤギのワイヤ一方式の電気牧柵に対する忌避学習に影響を及ぼしていること を示している。
キーワード:脱柵，電気牧柵，忌避行動，ヤギ
Animal Behaviour and Management， 51 (3): 121‑130，2015 (2014. 12. 22.受付; 2015. 5. 25.受理)
