ARTICLE

Eurasian perch, Perca fluviatilis, spatial behaviour determinesvulnerability independent of angler skill in a whole-lakereality mining experimentChristopher T. Monk and Robert Arlinghaus

Abstract: To understand the determinants of angling vulnerability arising from the interplay of fish and angler behaviour, wetracked 33 large Eurasian perch, Perca fluviatilis, with fine-scale acoustic telemetry at a whole-lake scale while simultaneouslytracking boats of small groups of experimental anglers (n = 104) who varied by self-reported skill. We report two key findings.First, perch vulnerability was strongly related to a repeatable habitat choice behaviour, but unrelated to swimming activity as apersonality trait; importantly, highly vulnerable perch were captured throughout the lake and not only in their preferredhabitat, suggesting covariance between spatial habitat choice and a behavioural determinant of vulnerability. Second, catch perunit effort of large perch increased with self-reported angling skill, an effect unrelated to skill-dependent lure use or an angler’sability to encounter perch. Importantly, high-skill anglers captured more fish but not different spatial behavioural phenotypes.Our study has implications for designing protected areas by showcasing that angling could systematically alter the habitat useof exploited populations at whole-ecosystem scales, without necessarily changing average swimming activity and home rangeextension.

Résumé : Pour comprendre les déterminants de la vulnérabilité a la pêche sportive a la ligne découlant de l’interaction descomportements des poissons et des pêcheurs, nous avons suivi 33 grandes perches communes (Perca fluviatilis) par télémétrieacoustique de haute résolution a l’échelle d’un lac, en suivant simultanément les bateaux de petits groupes de pêcheursexpérimentaux (n = 104) ayant signalé différents niveaux de compétence. Nous faisons état de deux constatations clés. D’abord,la vulnérabilité des perches est fortement reliée a un comportement de sélection d’habitats qui se répète, mais non reliée al’activité natatoire comme trait de personnalité; fait important, des perches très vulnérables ont été prises a la grandeur du lac,pas seulement dans leur habitat de prédilection, ce qui indiquerait une covariance entre la sélection de l’habitat dans l’espaceet un déterminant comportemental de la vulnérabilité. Deuxièmement, la capture par unité d’effort des grandes perchesaugmente parallèlement au niveau de compétence signalé par le pêcheur, un effet non relié a l’utilisation de leurres dépendantede la compétence ou la capacité d’un pêcheur de rencontrer des perches. Fait important, les pêcheurs plus compétents ontcapturé plus de poissons, mais pas de phénotypes comportements spatiaux différents. L’étude est importante pour la conceptionde zones protégées parce qu’elle démontre que la pêche sportive a la ligne peut systématiquement modifier l’utilisation del’habitat de populations exploitées a l’échelle d’écosystèmes entiers, sans nécessairement modifier l’activité natatoire moyenneou l’étendue de domaines vitaux. [Traduit par la Rédaction]

IntroductionThe intrinsic vulnerability to angling is strongly behaviour-

dependent because fish must ultimately approach and ingest a baitor lure (Uusi-Heikkilä et al. 2008; Løkkeborg et al. 2014; Lennox et al.2017). Recent studies have shown large intrapopulation variationin heritable behavioural types (also known as personality traits) inmany fish species (Conrad et al. 2011; Mittelbach et al. 2014), andthese traits have been found to be repeatable (i.e., consistentlydifferent among individual fish) in the wild (Harrison et al. 2015;Nakayama et al. 2016; Monk and Arlinghaus 2017). If heritable,repeatable behaviours correlate with angling vulnerability, har-vesting could induce selection on behavioural types (Alós et al.2016; Arlinghaus et al. 2017a; Diaz Pauli and Sih 2017). Given rec-reational angling’s global popularity (Arlinghaus et al. 2015), theselective removal of certain behavioural types may be strongly

altering the behavioural composition of fish populations, pheno-typically and genotypically (Tsuboi et al. 2016; Cooke et al. 2017).As a consequence, population-wide behavioural changes couldcreate relevant ecological, managerial, and evolutionary effects(Heino et al. 2015; Arlinghaus et al. 2017a; Diaz Pauli and Sih 2017).Despite substantial conceptual appeal (Arlinghaus et al. 2017a),the degree to which angling-induced selection on behaviour oc-curs in the wild is poorly documented (Heino et al. 2015).

Recently, several empirical studies investigating correlations be-tween behavioural types and vulnerability to fishing gear have beenpublished (reviewed in Arlinghaus et al. 2017a). These studies wereoverwhelmingly conducted under laboratory conditions (Diaz-Pauliet al. 2015; Killen et al. 2015) or with wild-caught fish assayed forpersonality in a laboratory setting or seminatural ponds (Wilsonet al. 2011, 2015; Vainikka et al. 2016). Such approaches could bias

Received 23 January 2017. Accepted 27 April 2017.

C.T. Monk. Department of Biology and Ecology of Fishes, Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Müggelseedamm 310, 12587Berlin, Germany.R. Arlinghaus. Department of Biology and Ecology of Fishes, Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Müggelseedamm 310, 12587Berlin, Germany; Division of Integrative Fisheries Management, Department of Crop and Animal Sciences, Faculty of Life Sciences, Humboldt-Universität zuBerlin, Invalidenstrasse 42, 10115 Berlin, Germany.Corresponding author: Christopher T. Monk (email: monk@igb-berlin.de).Copyright remains with the author(s) or their institution(s). Permission for reuse (free in most cases) can be obtained from RightsLink.

417

Can. J. Fish. Aquat. Sci. 75: 417–428 (2018) dx.doi.org/10.1139/cjfas-2017-0029 Published at www.nrcresearchpress.com/cjfas on 4 May 2017.

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

conclusions (Klefoth et al. 2012; Niemelä and Dingemanse 2014).Indeed, there is conflicting evidence regarding which traits areunder selection by angling in a range of species. For example,swimming activity and general spatial behaviours should corre-late positively with angling vulnerability because both traits in-crease the probability of encountering hooks, which is requiredfor capture (Uusi-Heikkilä et al. 2008; Alós et al. 2012). A modelby Alós et al. (2012) suggests that if vulnerability is entirelyencounter-based, there should be consistent selection on swimmingactivity, while angling-induced selection on home range will dependon the specific style of fishing. However, many studies relying tosome portion on testing behaviours in seminatural or artificialconditions failed to relate individual variation in activity to an-gling vulnerability (Binder et al. 2012; Härkönen et al. 2014, 2016;Wilson et al. 2015; Monk and Arlinghaus 2017). Also in the wild,basal activity and home range size of Atlantic cod, Gadus morhua,were both unrelated to vulnerability to a range of passive fishinggears (Olsen et al. 2012). By contrast, in another study conductedin the wild, Alós et al. (2016) found that more vulnerable adultpearly razorfish, Xyrichthys novacula, had elevated swimming activ-ity and a greater home range size. Selection on the personalitytrait “activity” or on emerging patterns such as home range maythus be fishery- and species-specific, and more studies in situ areneeded.

Under field conditions, the fish behaviours that might influencethe capture process will also interact with angler behaviour (Alóset al. 2012; Matthias et al. 2014; Wiig et al. 2014). Anglers areheterogeneous in specialization, preferred techniques, knowl-edge of fishing sites, general angling skill, and site choice (Bryan1977; Johnston et al. 2010), all of which can affect fish encountersand entice fish to bite. In a simulation of fish and angler move-ment, Alós et al. (2012) found that anglers fishing from fixed po-sitions imposed a stronger negative selection on home range,while anglers searching freely (e.g., by angling from boats) im-posed a stronger negative selection on fish activity. Therefore,angler search and fishing styles jointly affect behavioural selec-tion, and both need to be studied in realistic field settings. Moreover,anglers differing in lure choice might catch fish with different be-havioural types (Wilson et al. 2015), and both searching style andlure choice could vary with angling skill and level of specializa-tion (Bryan 1977). A relationship between angling skill and catchrates has been documented in a range of species (e.g., Dorow et al.2010; Heermann et al. 2013; Ward et al. 2013). However, little re-search is available on the mechanisms causing an angling skilleffect on catch rates, and an important limitation is that there arefew reliable measures of angler skill to classify study participants(Seekell 2011).

Our goal was to understand how vulnerability to lure-basedangling from boats relates to the interplay of angler and Eurasianperch, Perca fluviatilis, behaviour in a natural environment. Eur-asian perch is an appropriate study species because it is highlydemanded by anglers (e.g., Arlinghaus and Mehner 2004) and an-gler experience is known to determine catch rates and capturesize of perch (Heermann et al. 2013). Additionally, perch showpersonality in swimming activity in nature (Nakayama et al. 2016)and in a range of other traits in laboratory contexts (Magnhagen2012; Magnhagen et al. 2012; Kekäläinen et al. 2014). We followedthree objectives and tested three associated hypotheses. The firstobjective was to test for the potential relationship between perchactivity, space use and habitat choice, and vulnerability to an-gling. Following Alós et al. (2012), we hypothesized that moreactive perch (but not necessarily those with a larger home range)are more likely captured. Our second objective was to test for therelative importance of angler skill, skill-dependent lure use, skill-

dependent encounters with perch, and fishing location choice oncatch rates. We hypothesized that anglers who self-report to bemore skilled capture more fish (as per Dorow et al. 2010 in Euro-pean eel, Anguilla anguilla, angling) through improved searchingand lure selection. Our third and final objective was to disentan-gle the relative importance of perch and angler behaviour fordetermining individual vulnerability from the fish perspective. Tothat end, we precisely investigated the encounter process in lightof fish personality traits in situ and hypothesized that higherencounters with skilled anglers will increase angling vulnerabil-ity, beyond an increased intrinsic vulnerability of more activefish.

Methods

Perch taggingWe tracked large piscivorous perch with a calibrated fine-scale,

whole-lake acoustic telemetry system (see Baktoft et al. 2015 forfunctioning and accuracy) in the research lake Kleiner Döllnsee.Kleiner Döllnsee is a 25 ha weakly eutrophic natural lake (totalphosphorus at spring overturn of 38 �g·L−1) located �80 kmnortheast of Berlin (52°59=32.1==N, 13°34=46.5==E). It is closed topublic access and without public fishing since the early 1990s.Reeds, Phragmites australis, surround the lake and provide shelterto fish, as submerged macrophyte coverage is presently minimal(Fig. S11). The maximum lake depth is 7.8 m and the average depthis 4.4 m. The northern section of the lake features a flat, relativelysandy and sparsely vegetated bay area adjacent to a straight shore-line with a relatively steep slope in water depth, while the depthin the southern section increases more gradually and the sub-strate is mainly composed of fine sediments (mud) (Fig. S1). Duringthe study period from 7 September to 19 October 2015, the watertemperature was on average 12.9 ± 2.5 °C and Secchi depth was�2 m. In Kleiner Döllnsee, perch share a role as aquatic top pred-ator with an abundant northern pike, Esox lucius, and less abun-dant European catfish, Silurus glanis, population.

The perch population we studied was never exposed to targetedangling besides some experimental northern pike studies whereperch was a rare bycatch (Klefoth et al. 2008; Kuparinen et al. 2010;Pagel et al. 2015; Laskowski et al. 2016; Arlinghaus et al. 2017b). Amajority (n = 44) of perch we tagged came from Kleiner Döllnsee,but we added several individuals (n = 6) from a nearby (2.3 km),ecologically similar lake, Großer Vätersee, to increase sample size.All perch were collected by gill nets set over 30–60 min to mini-mize fish damage and stress. Acoustic telemetry tags (model MM-M-11-28-TP, transmission rate 27.5 s, wet mass 6.5 g; Lotek WirelessInc., Newmarket, Ontario) were surgically implanted according tomethods described elsewhere (Klefoth et al. 2008) into the bodycavity of perch (total length (TL) 374 ± 20 mm (mean ± SD), wetmass 744 ± 140 g (mean ± SD)) in autumn 2014 (n = 31, watertemperature 10.0 °C) and post-spawning in spring 2015 (n = 19,water temperature 13.5 °C) (see Table S1 for individual taggingdata). Fish were anaesthetized by a 9:1 95% EtOH – clove oil solu-tion (Carl Roth, Karlsruhe, Germany) added at 1 mL·L water−1.Surgical materials and tags were sterilized with 7.5% povidone–iodine (PVP) (Braunol®; B. Braun, Kronberg, Germany) in water,and all efforts were made to minimize handling. Surgeries wereon average 4:16 ± 1:11 min. Perch were also fitted with a 12 mmpassive integrated transponder (PIT) tag (Oregon RFID, Portland,Oregon) under the skin beneath the dorsal fin for later identifica-tion. After recovery from anaesthesia, perch were released intoKleiner Döllnsee.

1Supplementary data are available with the article through the journal Web site at http://nrcresearchpress.com/doi/suppl/10.1139/cjfas-2017-0029.

418 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

Objective 1

Assessment of perch behaviours and personality for studyingangling-induced selection

To assess perch personality and related behavioural outcomes(e.g., space use as an emergent property of systematic variation inactivity, Harrison et al. 2015), a suite of behaviours (distanceswum, distance from the shore, 95% activity space size, meanlatitude, and mean longitude) were scored for each fish daily from adata set of 2 061 249 positions recorded 7 September to 19 October.Each behaviour was chosen because it relates to the encounterprocess (Alós et al. 2012; Matthias et al. 2014; Lennox et al. 2017).We calculated the distance swum from the sum of Euclidean dis-tances between recorded positions for each fish. We excludeddistances below 5 m because these are indistinguishable fromtelemetry error (Baktoft et al. 2015). The distance measurementrepresented only the swimming distance while an individual fishwas in the sublittoral area or open water of the lake because thetelemetry system functions best outside the reed belt in openwater (Baktoft et al. 2015). Each recorded position had a minimumdistance to shore. A perch’s daily tendency to be inshore or offshore,a correlate of angling vulnerability in largemouth bass, Micropterussalmoides (Matthias et al. 2014), was measured as the daily meanminimum shore distance. We estimated the activity space sizefrom the 95% volume contour of the kernel utilization distribu-tion area daily for each fish. Individual activity space size was notcalculated on days with fewer than 30 positions to avoid biasedestimates from small sample sizes (Seaman et al. 1999). The kernelutilization distribution was estimated with the adehabitatHR packagein R (version 3.2.4) on a 200 by 104 cell grid with a cell size of5.78 m and a 10 m smoothing parameter. Lastly, the daily meanlongitude and latitude were calculated to measure the centraltendency of space use, reflecting a measure of habitat choice.Behaviours were also estimated for daytime and nighttime only(defined by civil sunrise and sunset times) to check for diurnalbehavioural patterns.

Repeatability of perch behavioursTo check for systematic among-individual behavioural varia-

tion on which angling selection could act, we calculated eachbehaviour’s repeatability (7 September to 19 October). We parti-tioned within- and between-individual covariances of a behaviouracross days using univariate mixing models fit with Markov ChainMonte Carlo procedures (Dingemanse and Dochtermann 2013).Models were fit with fish identity as a random intercept and anuninformative prior appropriate for the error distribution. Themodels for distance swum and activity space assumed a Poissonerror distribution, and the models of remaining behaviours as-sumed a Gaussian error. Repeatability was calculated according tothe appropriate equation given in Nakagawa and Schielzeth (2010).We ran each model over 500 000 chains with a burn-in of 1000,and every 100 chains were sampled to prevent autocorrelation.Trace plots were examined to verify convergence and assess good-ness of fit (see Supplementary Information 2). Each model was runfive times to ensure consistent estimates (only one model is re-ported).

Understanding the fish perspective — modelling angling selectionon fish behavioural traits

We used a Cox proportional hazards model to test for the rela-tionship between behaviours, TL, and speed to capture and a lo-gistic regression to test for the relationship between the samepredictor variables and a binary fitness (i.e., capture) outcome(1 indicated capture, i.e., theoretical death through harvest, and0 indicated never captured, i.e., survived the fishery). We consid-ered length in addition to behavioural variables because size isfrequently related to angling vulnerability (Lewin et al. 2006) andoften correlates somewhat with encounter-based behaviours (Biroand Post 2008). Therefore, we attempted to disentangle the con-

tribution of size and behavioural variation to vulnerability. TheCox proportional hazards model incorporated units of fishingeffort as a time variable and used counting censoring (Hougaard1999), as some individual fish were recaptured. The logistic modeldid not incorporate information about recaptures, which is irrel-evant from a fitness/selection perspective. We only included vari-ables in the models with variance-inflation factors below 3 toavoid collinearity issues (Zuur et al. 2007). These variables weredistance swum, longitude, latitude, and TL. Activity space size washighly correlated with distance swum (Fig. S2) and may be simi-larly related to vulnerability if distance swum would significantlypredict speed to capture or capture outcome. All predictor vari-ables were z-transformed for standardization of effect sizes. Wecompared a restricted model set based on our hypotheses. To selectthe best model, we compared the conditional Akaike informationcriterion (AICc) and considered models with a delta AICc less than 2to have equal explanatory ability (Burnham and Anderson 2003).

Objective 2

Effects of angler skill on catchesExperimental anglers were recruited via angling shops, clubs,

and internet forums. Voluntary anglers who contacted the au-thors received a questionnaire to pre-classify them according toself-reported skill. Self-reported angling skill (later calibratedagainst actual catching ability) was assessed from three itemstaken from several indices of angler specialization (e.g., Wildeet al. 1998; Beardmore et al. 2013). The first item was worded:“Relative to other anglers whom you know, how do you estimateyour angling skills in terms of catching perch?”, and the seconditem was a modification of the first item, where the word perchwas replaced with pike. The options to answer the first two itemswere “beginner”, ”less good”, ”similarly good”, “rather better”,and “angling expert” and scored from 1 to 5, respectively. The lastitem was worded: “How would you generally rank your anglingskills in comparison to the average angler?”, which was assessedon an 11-point scale with the first point as “a much worse anglerthan the average”, the sixth and middle point as “nearly the sameas the average”, and the highest point as “a much better anglerthan the average”. The answers from each question were standard-ized by z-transformation and transformed scores were summed foreach angler. Anglers scoring in the top third of all scores wereconsidered (self-reported) high-skill, anglers scoring in the middlethird were considered middle-skill, and anglers scoring in thebottom third were considered low-skill.

The experimental anglers (n = 104) were invited for only 1 day ofperch angling at Kleiner Döllnsee, from approximately 10:00 untilsunset (�19:00) with an hour lunch break at 13:00, either alone orin a small group of six anglers over 29 nonconsecutive days. Eachangler had never fished in the study lake. Groups were present on25 of 29 days and were composed of an equal representation ofskill groups to control for potential covariance between skill andday. However, there were days where an invited angler could notattend for personal reasons (e.g., illness) and therefore the actualgroup size ranged from two (4 days) to six (4 days) and 5, 7, and5 days with three, four, and five anglers respectively.

Anglers were provided standardized fishing gear (Favorite210 cm VRN-702M rod (Favorite Co., Ukraine), Shimano Exage2500FD reel (Shimano Germany Fishing GmbH, Germany), Power-Pro, 0.13 mm braided yellow-colored line (Shimano Germany Fish-ing GmbH, Germany), and Trilene 0.32 mm fluorocarbon leader(Berkley Fishing, Spirit Lake, Iowa)) and their own boat equippedwith a GPS data logger (i-Blue 747 ProS, Transystem Inc., Taiwan),logging once per second at 3 m precision. Anglers could fish freelyin space, but only with two lure types, a copper color Meppsspinner, size 3.0 (Mepps SNC, France), and an 8.5 cm soft plasticshaker with a Kansas shiner design (Lunker City, Connecticut).Anglers could freely choose among these lure types, but noted in

Monk and Arlinghaus 419

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

an angling diary which lure was used by time stamp. From the GPSdata, we calculated the average latitude and longitude of eachangler while they used each lure as an additional predictor ofcatch per unit effort (CPUE) because the combination of fishinglocation and lure type may be an important predictor of anglingsuccess. To measure effort, the anglers recorded start and stoptimes of their fishing, including small breaks, so only periods ofactive casting were included as effort. Additionally, anglers re-corded all catches and noted the capture time, species, and TL forall captured fish. When anglers captured a perch above a thresh-old size of 28 cm TL, they contacted the first author via walkie-talkie and the research team measured exact sizes and GPSlocation of the catch and checked for a PIT tag. All perch above28 cm TL captured for the first time received a PIT tag in the dorsalmusculature for identification upon recapture and use in a sepa-rate mark–recapture study. In the morning of each angling day,experimental anglers were thoroughly briefed about experimen-tal procedures. We observed high compliance in reporting cap-tures and completing angling diaries. The whole lake was visibleto the research team and constantly monitored. The anglers’ task,given during the morning briefing, was to catch perch over 28 cmTL, and they were told the experiment’s aim was to learn how wellanglers can find and catch perch.

Encounter rate estimationTo estimate encounters between tagged perch and anglers, we

defined an encounter as a minute where any tagged fish waswithin casting range (15 m) of an angler. Our measure assumesthat fish within casting range should sense the angler’s lure. Weassumed that the sample of tagged perch represented the encoun-tering behaviour of all large perch and used tagging-based en-counter data for modelling the catch rate of perch above theminimum acoustically tagged perch size (33.5 cm TL).

From the angler perspective, we considered two measures ofencounter rate: (1) the cumulative number of unique acousticallytagged perch encountered scaled by angling effort (fish per hour)and (2) the total minutes exposed to any acoustically tagged perchindependent of identity scaled by angling effort (fish minutes perhour). We used these two encounter measures because for an angler,encountering a new and potentially vulnerable fish may influencecatch rate more than spending considerable time around a fewpossibly invulnerable individuals. We scaled encounter measuresby effort to account for different encountering potential over lon-ger fishing times.

The angler perspective — determinants of angler success as afunction of self-reported skill

We investigated the skill effect on CPUE of all species, all perchindependent of size, large perch (>28 cm TL), and perch in the sizerange of acoustically tagged individuals (>33.5 cm TL). For modelsof CPUE by skill group, we predicted catch with a generalizedlinear model (glm) offset by angling effort according to Kuparinenet al. (2010). To account for overdispersion, we assumed a negative-binomial error distribution with a log link. In models where skillgroup was a significant effect, we used a post hoc Tukey HSD testfrom the R package “multcomp” (Hothorn et al. 2016) to comparedifferences among the three groups. To test if lure choice differedamong the skill groups, we used a one-way ANOVA to compare theproportion of time the shaker lure was used, which was arcsinesquare-root transformed to conform to normality assumptions.We also compared space use preferences (mean and standard de-viation of latitude and longitude) among skill groups using a one-way ANOVA. Finally, we tested for skill-related differences inperch encounters (both measures) using glms offset by effort,assuming a negative-binomial error distribution and log link.

To understand the relative importance of angler skill, both en-counter measures, lure used, average latitude of the angler, andday as predictors of CPUE, we compared a restricted set of glms

predicting the catch of perch over 33.5 cm TL with a Poisson errordistribution and a log link, offset by effort (Kuparinen et al. 2010).Day was treated as a continuous variable to account for any sys-tematic temporal increase or decrease in CPUE. We further con-sidered several possible interactions, as encounters may onlyincrease catches for high-skilled anglers who can convert an en-counter into a capture, encounters may only increase catches inthe Mepps lure, which requires less technical finesse than theshaker, or encounters may only increase catches in regions of thelake where fish are foraging. Furthermore, the effect of lure maychange temporally as temperatures shift and fish are increasinglyexposed to anglers (Kuparinen et al. 2010), and the effect of en-counter time may strengthen with a higher rate of unique fishencountered. We used the model selection and evaluation ap-proach detailed in objective 1.

Objective 3

Relative influence of fish personality and encounters withheterogeneous anglers on vulnerability

To test our third objective, we compared the joint effects ofencounters and perch behaviours on individual angling vulnera-bility. As a possible covariate of overall vulnerability (i.e., fitness),we first considered the total minutes each fish encountered an-glers and also the number of different anglers encountered foreach angling skill group over the entire study duration. We alsoconsidered a subset of angler encounters by time exposed to eachskill and lure because encounters with the lure types or skill levelsmay not affect vulnerability equally. We further considered en-counters by habitat assuming that vulnerability may vary acrosshabitats (Matthias et al. 2014). Accordingly, we split the lake intothree zones in an ad hoc fashion (Fig. S1): an offshore pelagic zone(>5 m depth) and onshore zones in the north and south halves ofthe lake. We then considered subsets of encounters with anglersin a single skill group in the northern and southern zones toexamine the interaction of skill and habitat for affecting overallvulnerability. We also considered only skilled anglers when fish-ing with the soft-plastic lure, which is a common lure amongmore specialized anglers and has been found to be more effectivein catching pike in the same study lake (Arlinghaus et al. 2017b).All encounter variables were standardized via z-transformationand tested with the most parsimonious logistic regression modelfrom objective one in a restricted model set to compare the rele-vant effects of angler encounters and perch behaviours on vulner-ability. Model selection followed the same methods as describedbefore. All variables in all models had a variance inflation factorbelow 3 (Zuur et al. 2007).

Lastly, we calculated the mean standardized selection gradientaccording to Matsumura et al. (2012) for significant traits in thebest logistic model, which enabled a trait-by-trait fitness elasticitycomparison. To that end, we transformed the regression coeffi-cients to their linear equivalents following Janzen and Stern(1998) before calculating normalized selection gradients on adap-tive traits.

Results

Descriptive informationThe 33 perch swam on average 5.6 ± 2.5 km·day−1 (mean ± SD)

and had a daily 95% activity space size of 5.1 ± 1.5 ha (mean ± SD)(see Fig. S3 for raw data). The 33 tagged perch were typically foundin the sublittoral onshore area and more rarely in the pelagic andthe lake centre (see Fig. S4 for raw data on space use). This patternresulted in an average daily distance of 73 ± 12 m (mean ± SD) fromthe shoreline. Perch displayed strong diurnal behavioural pat-terns. Before sunset, the perch swam on average 4.5 ± 2.4 km andwere on average 87 ± 20 m from the shore, while during the night,they swam significantly less at an average of 1.0 ± 0.6 km (paired ttest, t = 14.63, df = 32, p < 0.001) and were significantly closer to the

420 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

shore, on average 59 ± 18 m away (paired t test, t = 12.56, df = 32,p < 0.001). All behaviours that we measured were significantlyrepeatable (Fig. 1) with 95% credible intervals ranging from 0.27 to0.72. Daily latitude and longitude were the most repeatable behav-iours (R = 0.53 and 0.60, respectively) and activity was least repeat-able (R = 0.40).

Objective 1 — Is there angling-induced selection on perchpersonality in the wild?

There were 20 capture events of acoustically tagged perch con-sisting of 15 unique individuals. Four acoustically tagged perchwere captured twice and one was captured three times resultingin 45% of acoustically tagged perch (n = 33) captured at least onceover 1.5 months. None of the behavioural traits nor TL explainedtime to capture for acoustically tagged perch; the most parsimo-nious model was the null model (Table 1). However, the mostparsimonious model predicting the fitness outcome of being cap-tured during the whole fishing period contained the average lat-itude and longitude (Fig. 2) and TL (Table 1). The effect of latitudehad a considerably higher odds ratio than the effect of longitude,while the 95% confidence interval of the effect size of TL over-lapped 1, indicating that length explained some variance in cap-ture, but the effect was not significant (Table 2). Overall, perchpreferring habitats more north and west in the lake were morelikely to be captured at least once, although these fish were regu-larly captured beyond their core home range (only 60% of captureswere in the lake’s northern onshore zone) (Fig. 2). The differencesin habitat between captured and never captured perch becamegreater in the night after angling (Fig. S5). We thus also tested ourmodels using only variables measured during daylight (becausefishing only took part during daylight) and found that the resultswere unchanged (not shown). Accordingly, the fitness model (i.e.,the logistic regression model examining fisheries selection) docu-mented angling-induced selection pressures in relation to habitatuse as represented by average latitude and to a lesser extent lon-gitude (Table 2). There was, however, no selection on activity as arepeatable personality trait, and by association, there was no evi-dence for selection on the highly correlated (see Fig. S2) trait ofspace use (Table 2).

Objective 2 — Is there a skill effect on catch rate of largeperch?

Collectively, among all 104 experimental anglers fishing over1.5 months, there were 764 perch capture events of fish rangingfrom 6 to 45 cm TL (25 ± 8 cm TL, mean ± SD) over a total 709.6 hof angling effort. There were 95 captures of 79 individual nona-coustically tagged perch larger than 33.5 cm TL, the size range ofacoustically tagged perch. The average CPUE for perch above33.5 cm TL was 0.18 ± 0.15 fish rod·h−1 (mean ± SD, n = 104), while

Fig. 1. Repeatability estimates and their 95% credible intervals forfive Eurasian perch Perca fluviatilis, behaviours.

Table 1. Candidate models for predicting time to capture (survivalmodel) and overall capture outcome (logistic model) based on distanceswum (DST), latitude (LAT), longitude (LON), distance from the shore(SHDST), and total length (TL) and their residual deviance (Dev), numberof parameters (P), and AICc.

Model typeModelNo. Model structure Dev P AICc

Survival 1 DST × LAT + LON + SHDST + TL 7 118.62 DST × LON + LAT + SHDST + TL 7 118.73 DST × SHDST + LAT + LON + TL 7 119.94 DST + LAT + LON + SHDST + TL 6 116.95 DST × SHDST 4 100.46 DST + SHDST + TL 4 100.57 LAT + LON + TL 3 126.48 TL 2 97.19 NULL 1 95.1

Logistic 1 DST × LAT + LON + SHDST + TL 24.60 6 43.12 DST × LON + LAT + SHDST + TL 24.86 6 43.33 DST × SHDST + LAT + LON + TL 22.56 6 41.04 DST + LAT + LON + SHDST + TL 25.11 5 40.35 DST × SHDST 43.38 3 52.86 DST + SHDST + TL 44.13 3 53.67 LAT + LON + TL 26.71 3 36.18 TL 45.40 1 49.89 NULL 45.47 0 47.6

Note: The most parsimonious models according to AICc model selection isbold.

Fig. 2. Mean latitude and longitude positions of captured (redpoints) and never captured (black) acoustically tagged Eurasianperch Perca fluviatilis, and their standard deviations. Blue trianglesindicate capture location for these acoustically tagged perch. Thebackground of the lake is the kernel utilization distribution of thefishing locations of all anglers. The white background denotesregions outside the 95% volume contour of the kernel utilizationdensity of the angler’s space use. Note that although averagepositions tend to be near the centre of the lake, the perch tend tostay closer to the shoreline (see Fig. S4). [Colour online.]

Monk and Arlinghaus 421

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

the average CPUE for all perch was 1.03 ± 0.98 fish rod·h−1 (mean ±SD, n = 104).

Anglers of all skill levels concentrated most of their effort alongthe lake’s shoreline (Fig. 3), but the high-skill group spent margin-ally more time fishing in the open water (Fig. S6). Yet, there wereno significant differences among the three groups (low-skill, n =42; middle-skill, n = 35; high-skill n = 27) in terms of latitude use(ANOVA, F[2,101] = 0.48, p = 0.62), longitude use (ANOVA, F[2,101] =0.16, p = 0.86), and the standard deviation in latitude (ANOVA,F[2,101] = 2.33, p = 0.10) and longitude (ANOVA, F[2,101] = 0.17, p = 0.84)(Fig. 3), indicating no skill-related differences in the choice ofangling habitats in Kleiner Döllnsee.

There was a significant effect of skill group on the CPUE of largeperch in the same size range as acoustically tagged perch (glm,null deviance = 125.4, residual deviance = 111.0, df = 2, 101, p < 0.001;Tukey HSD comparisons, high-skill to middle-skill, p = 0.053; high-skill to low-skill, p < 0.001; middle-skill to low-skill, p = 0.28), butself-reported skill did not predict performance in terms of catch-ing perch over 28 cm TL (glm, null deviance = 149.8, residualdeviance = 146.1, df = 2, 101, p = 0.152), perch in general (glm, nulldeviance = 143.6, residual deviance = 142.7, df = 2, 101, p = 0.63), orall species (glm, null deviance = 142.6, residual deviance = 140.4,df = 2, 101, p = 0.34) (Fig. 4).

An ANOVA indicated that there were no significant differencesin lure use among the skill groups (F[2,101] = 1.83, p = 0.17), as thesoft-plastic lure was used 57% ± 30% (mean ± SD, n = 42), 61% ± 34%(mean ± SD, n = 35), and 70% ± 36% (mean ± SD, n = 27) of the timeby the low-skill, middle–skill, and high-skill anglers, respectively.However, individual variation in lure use was substantial, proba-bly contributing to the lack of significant differences despite aclear trend that the use intensity of the soft-plastic lure increasedwith skill level. In terms of encounters, which were estimatedseparately according to each lure used by each angler, the rate(per unit angling effort) of encountering new individual perch(fish per hour) was not significantly different among the threeskill groups (glm, null deviance = 199.3, residual deviance = 199.2,df = 2,187, p = 0.93) and also the encounter time per unit effortwith perch independent of fish identity (fish minutes per hour)was not significantly different among the three skill groups (glm,null deviance = 222.5, residual deviance = 221.7, df = 2, 187, p = 0.68)(Fig. S7).

Of models predicting CPUE of large perch in the size range ofacoustically tagged individuals, the most parsimonious model in-cluded an interaction between the rate of encountering new indi-vidual perch and skill; however, models that included only skill orskill and day both had a delta AICc less than 2 and thus equalexplanatory power (Table 3). The simpler model excluding en-counters was then considered the best model. Accordingly, CPUEincreased for those in the high-skill group relative to the low-skillgroup and also increased to a lesser extent for the middle-skillgroup relative to the low-skill group (Table 4). An equally parsi-monious model indicated that CPUE increased with higher en-counters of new individual perch; however, this effect was small,particularly in comparison to the effect of skill (shown by the

strongly different intercept at zero encounters in Fig. 5). The 95%confidence intervals for the effect size of day overlapped 1 and theeffect was overall small, indicating that day did not have a signif-icant effect on angler CPUE. Encounter time per unit effort withany perch (fish per minute per hour) was of no importance in anyof the best models (Table 3). Similarly, lure type, encounters whileusing particular lure types, and habitat choice were not retainedin the best models (Table 3). Overall, individual CPUE for largeperch was largely independent of encounters, lure use, and anglerhabitat choice (as indexed by latitude) and was instead mainlydriven by a plain skill effect (Table 4).

Objective 3 — Understanding the encountering process offish and anglers

The total encounter duration over 29 angling days of individualtagged perch with anglers ranged from 7 to 142 min, with a meanof only 58 ± 26 min per tagged perch. When total encounter timeswere subset by angler skill level, individual perch encounteredlow-skilled anglers on average 17 ± 10 min, middle-skilled anglerson average 23 ± 13 min, and high-skilled anglers on average 18 ±12 min. When total encounter times were subset by lure, individ-ual fish encountered the spinner on average 16 ± 8 min comparedto on average 42 ± 22 min with the shaker over 29 fishing days.Individual tagged perch encountered anglers on average 8 ± 6 minin the southern onshore area of the lake, while in the northernonshore area, fish encountered anglers on average 24 ± 21 minover 29 fishing days. Finally, individual perch encountered onaverage 53 ± 7 of the 104 unique anglers (range = 33–65).

We found four models comparing the joint effects of encoun-ters and perch behaviours on angling vulnerability with equalexplanatory power. In these, the selection of angling on habitatchoice (indexed by latitude and longitude of the fish) remained inall four models (Table 5). Importantly, angling-induced selectionon perch habitat choice remained significant despite accountingfor differential exposure to anglers. Hence, after controlling forencounters with different angler types, there was still evidencefor selection on traits characterizing individual perch in terms ofhabitat choice. Additionally, the most parsimonious model alsoincluded an effect of minutes encountering anglers in the lake’snorthern onshore zone. The other three equally parsimoniousmodels included the encounters in the lake’s northern and south-ern onshore zones, the encounters in the lake’s northern andsouthern onshore zones when subset by middle-skill group an-glers only, and finally encounters with all anglers (Table 5). Thedirection of encounter effects was as expected in that more en-counters with anglers resulted in a higher probability of beingcaptured, with the exception of encounters in the lake’s southernonshore zone, which had a negative coefficient (Table S2). How-ever, the lower estimate of the 95% confidence interval of theeffect size for encounters in the northern sublittoral region of thelake was near 1, indicating that the effect was small compared tothe habitat choice variables characterizing the individual perch(Table 6). Similarly, the 95% confidence intervals of the effect sizesfor all other encounter variables included in the most parsimoniousmodels overlapped 1 and were therefore not significant (Table S2).The mean standardized selection gradients ± their standard errorwere −1.49 ± 0.57, 0.84 ± 0.33, and −0.17 ± 0.10 for the latitude(scaled as the distance in metres from the southernmost point inKleiner Döllnsee), longitude (scaled as the distance in metres fromthe westernmost point in the Kleiner Döllnsee), and encounters inthe northern onshore zone, respectively. Consequently, angling-induced selection operated mainly on habitat choice variables inperch and here most strongly on latitude. Vulnerability was addi-tionally affected slightly by the duration of encounters in thelake’s northern onshore zone.

Table 2. Standardized parameter estimates, their standard error (SE),odds ratios, 95% confidence interval (CI) estimations of the odds ratios,and deviances for the most parsimonious model (logistic regression)predicting vulnerability of acoustically tagged Eurasian perch, Percafluviatilis (shown in Table 1).

Covariate Estimate SEOddsratio

95% CI ofodds ratio Deviance

Intercept −0.01 0.49 0.99 0.37 2.73Latitude 2.81 1.00 16.66 3.38 204.89 8.02Longitude −1.99 0.81 0.14 0.02 0.50 10.26Total length −0.34 0.49 0.71 0.24 1.84 0.49

422 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

DiscussionWe found partial support for the three hypotheses that guided

our analysis. First, our study revealed that angling indeed inducesselection on behaviours of large piscivorous perch in the wild, butin disagreement with our hypothesis, selection operated neitheron activity nor space use (with the lack of effects only expected forspace use in mobile anglers, Alós et al. 2012). Instead, selectionwas directed at systematic habitat choice variation of large perch.Second, our hypothesis relating to self-reported angling skill wasstrongly supported, as high-skilled anglers outperformed low-skilled anglers in catching large perch. However, in contrast withour hypothesis, the success of high-skilled anglers was derivedfrom their ability to use artificial lures or remain attentive duringfishing rather than superior lure choice or fish finding ability.Lastly, we rejected our third hypothesis because perch vulnerabil-ity to capture was largely independent of the angler type encoun-tered and was instead driven by habitat choice and more modestlyby overall encounter duration in the lake’s northern area. Overall,while perch capture efficiency was a function of skill, the specificbehavioural type of perch captured by anglers was independent ofangler skill.

On theoretical grounds, the accumulation of a fish’s encounterswith angling gear is thought to strongly predict the probability ofcapture (Lennox et al. 2017) and was a key reason for expectingangling-induced selection on activity or possibly home range be-haviours under certain fishing styles (Alós et al. 2012; Arlinghauset al. 2017a). Encounters with angling gear are certainly a neces-sary condition for capture (Monk and Arlinghaus 2017), and weindeed revealed that the total time a perch was in close proximityto anglers when in the northern onshore area of the lake signifi-cantly elevated its chances to be captured. Therefore, our resultsshowed for an individual perch that it mattered where theangler was encountered. However, compared to the intrinsicand highly repeatable habitat choice variables characterizingvariation among individual perch, the encounter rate with an-glers, independent of angling skill, was a poor predictor of indi-vidual capture probability. Similarly, the encounter rate withperch was a poor predictor of angler CPUE of large perch. There-fore, we conclude that encounter rates are at most weakly relatedto capture in perch angling with lures, a finding that agrees witha recent telemetry study with common carp, Cyprinus carpio, andtench, Tinca tinca, in the same study lake (Monk and Arlinghaus2017). By contrast, another recent study in pearly razorfish fished

Fig. 3. Boxplots of catch of all species, catch of all Eurasian perch, Perca fluviatilis, catch of perch over 28 cm total length, and large perch over33.5 cm total length per hour of fishing according to the self-reported three angler skill levels. The bars inside the box represent the median,the box represents the range between the 25th and 75th percentiles, and whiskers extend to 1.5 times the interquartile range. Outliers areindicated by points.

Fig. 4. Mean fishing position of anglers separated by skill level. Vertical and horizontal error bars represent standard deviation in the latitudeand longitude, respectively.

Monk and Arlinghaus 423

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

with natural bait found strong evidence for boat-angling-basedselection on both activity and home range (Alós et al. 2016), sug-gesting strong species- and fishery-specific effects on determi-nants of vulnerability. It is possible that our findings would bedifferent when fishing for perch with natural baits or in other lifestages (Ballew et al. 2017). Fish react differently to the presence ofnatural or artificial baits (Payer et al. 1989; Arlinghaus et al. 2008)as shown previously in northern pike (Beukema 1970), whichlearned to avoid artificial lures but not natural bait. Encounterslikely have a greater influence on vulnerability when fishing withnatural bait because natural baits should attract all fish behaviouraltypes whereas artificial lures are expected to attract more aggressiveand bold individuals (Wilson et al. 2015), thereby reducing the pre-dictive power of encounters.

The reasoning for our finding that encounters between fish andanglers were less important than predicted may be found in vul-nerable pool dynamics conceptualized in foraging arena theory.When applied to angling, foraging arena theory states that fishexchange rapidly between vulnerable and invulnerable states attime scales of minutes or hours (Cox and Walters 2002; Camp et al.2015). Under natural predation, spatial vulnerability is important,but in angling, the fish must be in a “reactive” state to attack a lurein addition to being spatially encountered (Camp et al. 2015). Ac-cordingly, many fish encountered will be spatially vulnerable(Matthias et al. 2014) but effectively invulnerable, particularly toartificial lures, diluting the effect of encounter duration on cap-ture probability (Lennox et al. 2017). From a fish perspective, theirbehavioural characteristics may far outweigh the effects of anglerencounters towards vulnerability with artificial lures (Sutter et al.2012). Thus, encounters are a necessary but insufficient conditionof angling vulnerability (Monk and Arlinghaus 2017).

Contrary to our initial expectations, perch swimming activitywas unrelated to vulnerability despite its significant repeatability(the latter confirming earlier studies in the study lake, Nakayamaet al. 2016). The poor relationship between encounters and vulner-ability of perch that we documented already explains why activitywas unrelated to vulnerability. Our findings also agree with re-sults from a range of species in laboratory contexts (Wilson et al.2011, 2015; Härkönen et al. 2016), seminatural ponds (Binder et al.

2012), and the wild (Olsen et al. 2012). By contrast, Biro and Post(2008) found that more active rainbow trout, Oncorhynchus mykiss,genotypes were more easily captured in gill nets. This indicatesthat selective properties on given personality traits depend ongear type (Diaz Pauli et al. 2015; Andersen et al. 2018), species andfishery (Arlinghaus et al. 2017a; Diaz Pauli and Sih 2017; Lennoxet al. 2017).

We found a strong relationship between habitat choice andvulnerability, similar to Matthias et al. (2014) in largemouth bass.We accounted for direct encounters in different lake basins andangling locations, and space use was independent of skill level;hence, our finding is not a result of captured perch dwelling in alocation where anglers spent more effort. Capture by angling isultimately dependent on fish behaviour, and therefore, the rela-tionship between vulnerability and habitat choice more likelyresults from some fine-scale habitat-specific behaviour. We spec-ulate about several possibilities. First, habitat-specific prey prefer-ences may be present. Stable isotope analysis has shown thatdifferences in diet composition among large perch in KleinerDöllnsee are related to life-history prototype and behavioural typealong a risk gradient (Nakayama et al. 2017), indicating intrinsicdifferences in foraging behaviour among Kleiner Döllnsee’s largeperch. Lure choice and natural foraging have been found to relateto vulnerability to angling in several species (Nannini et al. 2011;Sutter et al. 2012; Wilson et al. 2015), although Kekäläinen et al.(2014) failed to find a relationship of capture method (naturalversus artificial bait) and boldness in smaller perch. As perch arepredominantly visual foragers, lure appearance could be a keyfactor determining if the lure is considered a prey item. Therefore,perch preferring the lake’s northwest region may have visuallyassociated our lures with preferred prey, increasing the attack like-lihood. Second, perhaps habitat-specific aggression exists among theperch. Aggression is genetically related to largemouth bass vul-nerability (Sutter et al. 2012) and therefore remains a good candi-date for selective capture in perch. Additionally, Härkönen et al.(2016) found that initial exploration predicted relative perchcatchability in laboratory contexts. However, nearly all perchtracked in our study were familiar with the lake before tagging, andtherefore, we could not measure exploration in a novel environ-ment. It is possible that initial exploration measured in Härkönenet al. (2016) was a proxy for aggressive behaviour (Adriaenssens andJohnsson 2013). We thus suggest that habitat choice is indicativeof underlying behavioural patterns related to selecting a high-quality habitat and is presumably to be correlated with variationin aggression and (or) foraging mode.

We detected a clear signature of self-reported skill in the catchrates, in agreement with previous findings that angling experiencestrongly predicts perch catch rates and capture size (Heermann et al.2013). By contrast, Seekell (2011) suggested that exceptional anglingcatches are often the result of chance or simply caused by timeinvestment into fishing, but he could not rule out that skill alsoaffected catches. Angling skill is difficult to rapidly approximate insurveys, as sophisticated assessment methods are lacking (Seekell2011). Our work suggests that it may be easy to classify anglingskill in a new lake a priori based on self-assessment from threeitems, thereby contributing to calls to develop skill indices forinclusion in human dimensions surveys (Seekell 2011). Our three-question index must still be validated in other fisheries and culturesbecause responses and angler behaviours may vary regionally (Wardet al. 2013).

We have not only found a clear effect of skill on catch rates butalso gained a better understanding of how skilled anglers catchmore perch. As catch rates were largely independent of total en-counters, encounters with a particular lure type, and encounterswith anglers of particular skill types (who also used similar habi-tats), the skill effect appears to be largely the result of finesse orconcentration with the lure. Collectively, our findings indicatethat the ability to either take advantage of an encountered fish in

Table 3. Candidate models for predicting angler catch per unit effort(CPUE) based on encounter time with telemetry Eurasian perch, Percafluviatilis, scaled by fishing time (EPUE), number of unique telemetryfish encountered scaled by fishing time (EFish), skill level (GRP),latitude fished (LAT), and day (DAY) and their residual deviance (Dev),number of parameters (P), and AICc.

ModelNo. Model structure Dev P AICc

1 EPUE × GRP + EFish + LAT + DAY 236.56 9 401.02 EPUE × LR + GRP + DAY 228.66 8 394.53 EPUE × LAT + GRP + DAY 236.02 8 399.04 EPUE + GRP + LR + LAT + DAY 237.30 7 396.95 EPUE × GRP 239.26 6 396.06 EFish × GRP 234.12 6 390.97 GRP × LR 238.02 6 394.88 GRP + LR + LAT 242.21 5 396.89 LR × DAY 251.80 4 404.3

10 GRP + DAY 238.61 4 391.111 EPUE × EFish 251.83 4 404.312 EPUE + DAY 250.96 3 401.813 EFish 252.78 2 401.114 GRP 242.24 3 392.715 LR 254.91 2 403.316 NULL 255.06 1 401.4

Note: The most parsimonious model according to AICc model selection isbold. Models are a generalized linear model assuming a poisson error with a loglink. All models are predicting catch of perch >33.5 cm total length and areoffset by effort, thus effectively predicting CPUE.

424 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

a vulnerable state or to promote an encountered fish to switch toa vulnerable state through angling technique is most influentialto the probability of capturing large perch compared to all otherfactors that we tested and is thus the most plausible explanation

for the skill effect that we documented. Importantly for our studyobjectives, differently skilled anglers caught more fish but cap-tured the same behavioural types. Hence, selection by anglersvarying in skill will only be altered by possible differences in

Table 4. Parameter estimates from two of the most parsimonious models predicting catch perunit effort (CPUE) of >33.5 cm total length Eurasian perch, Perca fluviatilis (shown in Table 3),their standard errors (SE), effect sizes, 95% confidence interval (CI) estimations of the effectsize, and deviances.

ModelNo. Covariate Estimate SE

Effectsize

95% CI ofeffect size Deviance

10 Intercept −6.71 0.23 0.001 0.0007 0.002Group (middle-skill) 0.12 0.27 1.12 0.67 1.90 12.82Group (high-skill) 0.74 0.24 2.10 1.34 3.39Day 0.01 0.007 1.01 1.00 1.02 3.63

6 Intercept −6.08 0.31 0.002 0.001 0.004EFish −1.19 0.81 0.30 0.05 1.33 2.28Group (middle-skill) −0.72 0.43 0.49 0.21 1.15 12.32Group (high-skill) 0.25 0.36 1.28 0.66 2.67Efish: group (middle-skill) 2.09 0.90 8.05 1.51 51.89 6.34Efish: group (high-skill) 1.40 0.85 4.07 0.84 23.67

Note: The model No. refers to models listed in Table 3.

Fig. 5. Angler catch of large Eurasian perch Perca fluviatilis, over 33.5 cm total length per hour fishing as a function of the rate of encounterwith new unique acoustically tagged perch and separated by three skill groups: (A) low-skill group, (B) medium-skill group, and (C) high-skillgroup. Point size increases with effort, point colour represents the lure used (grey, Mepps spinner; black, soft-plastic shaker). Predictions fromthe most parsimonious model (solid line) of angler catch per unit effort based on an interaction between the rate of encountering new perchand skill and the 95% confidence intervals of that model (broken line) are also shown. Predictions are based on fishing at 75% of themaximum effort observed during the experiment.

Monk and Arlinghaus 425

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

fishing mortality, not by different angler types catching differentphenotypes.

An asset of the present study is that it was conducted in anatural system with real fish–angler dynamics, thereby reducingexperimental artefacts from a laboratory environment. However,there are still a number of limitations to our approach. First, wewere unable to measure theoretically important fish behaviourssuch as aggression and boldness that operate at scales finer thanthe resolution of the telemetry system, which has an averageprecision of 5 m. Second, we were restricted to study a single lake.It remains to be seen how generalizable our findings are becauseof the temporal and lake-specific population dynamics of perch.The population dynamics of perch for example depend on thetrophic state of the lake, lake size, and its clarity (Persson et al.2004; Svanbäck and Persson 2009; Jacobsen et al. 2015). Third,anglers may behave differently when they are familiar with awaterbody and are more easily able to find fish with an echo-sounder. However, we expect that with enhanced ability to findfish, the separation in catch rates according to skill would growwider. Nevertheless, more studies in other lakes are needed be-fore our results can be generalized.

We conclude that perch angling induces selection on spatialbehavioural traits, and assuming that some repeatable trait vari-ance has a genetic origin, fisheries-induced evolution of behav-iours is possible as previously argued (Arlinghaus et al. 2017a;Diaz Pauli and Sih 2017). We further conclude that angler skillstrongly affects catch outcomes, which is an intrinsic effect unre-lated to searching ability or lure choice. Finally, the ultimate catch

outcome from a fish perspective is mainly driven by fish behav-iour and less by angler behaviour. Different angler types exertdifferential overall mortality but do not seem to catch differentfish personalities, at least not under the conditions we examined.To understand the potential for evolutionary consequences fromangling, future studies are required to establish the precise traitunder selection and its heritability and plasticity. With respect toplasticity, it is currently unclear how the remaining perch mightrespond to freed niches once vulnerable perch are harvested. Beforesuch research becomes available, we suggest fisheries managers con-sider the potential for habitat-specific behavioural selection in thedesign of protected areas and the spatial management of fisheries(Villegas-Ríos et al. 2017). Moreover, our finding that angler skillplays a substantial role in catch rates of target fish via lure abilityindicates that skill ought to be considered in creel surveys, partic-ularly as skill might be reliably measured through a short index.The skill effect that we found may have important managerialimplications because anglers are known to sort their effort amongregional sites according to specialization (Ward et al. 2013), whichis correlated with angling skill (Dorow et al. 2010). Hence, overallfishing mortality on a given stock will depend on where a givenangler type preferentially fishes as previously documented (Johnstonet al. 2010; Ward et al. 2013), although these effects will depend onthe differential tendency of a given angler type to practice volun-tary catch-and-release as opposed to catch-and-kill.

AcknowledgementsThis study was financially supported by the project B-Types

(SAW-2013-IGB-2; www.b-types.igb-berlin.de) to R.A. from the Leib-niz Community. The invasive animal procedures were approved bythe Landesamt für Umwelt, Gesundheit und Verbraucherschutz ofBrandenburg Germany under the project “Fangbarkeit, project No.2347-21-2014.” We thank Daniel Hühn, Jonathan Nickl, and ThiloDamerau for their assistance in the field and operating the exper-iment. We would also like to thank Alexander Türck, Jan Haller-mann, and Andreas Mühlbradt for excellent technical assistanceand reviewers for good feedback. Lastly, we thank the 104 anglerswho volunteered to fish for science for 1 day and the many indi-viduals from tackle shops, fishing clubs, and internet forums whohelped to advertise the study and recruit participants.

ReferencesAdriaenssens, B., and Johnsson, J.I. 2013. Natural selection, plasticity and the

emergence of a behavioural syndrome in the wild. Ecol. Lett. 16(1): 47–55.doi:10.1111/ele.12011. PMID:23034098.

Table 5. Candidate model set for understanding the interplay between fish vulnerability and anglerbehaviour, where E is an encounter variable subset by skill (Low, Mid, and High), lure (Mepps orPlastic), lake region (Northern or Southern), and the unique number of anglers (Boats).

ModelNo. Model structure Dev P AICc

1 LAT + LON + TL + ELow + EMid + EHigh 22.68 7 41.162 LAT + LON + TL + EHighBoats + EMidBoats + ELowBoats 24.59 7 43.073 LAT + LON + TL + EMepps + EPlastic 22.81 6 38.044 LAT + LON + TL + ENorthern + ESouthern 9.69 6 34.935 LAT + LON + TL + EHighNorthern + EHighSouthern 23.20 6 38.436 LAT + LON + TL + EMidNorthern + EMidSouthern 18.85 6 34.087 LAT + LON + TL + ELowNorthern + ELowSouthern 21.00 6 36.238 LAT + LON + TL + EHigh 24.77 5 36.999 LAT + LON + TL + ENorthern 21.35 5 33.58

10 LAT + LON + TL + EHighPlastic 24.71 5 36.9311 LAT + LON + TL + EHighNorthern 24.64 5 36.8612 LAT + LON + TL + EBoats 25.50 5 37.7213 LAT + LON + TL + E 23.14 5 35.3714 LAT + LON + TL 26.71 4 36.1415 NULL 45.47 1 47.60

Note: Residual deviances (Dev), parameter numbers (P), and AICc are also presented, where the most parsimo-nious model is in bold.

Table 6. Standardized parameter estimates from the most parsimo-nious mechanistic model predicting vulnerability of acousticallytagged Eurasian perch, Perca fluviatilis (shown in Table 5), their stan-dard errors (SE), odds ratios, 95% confidence intervals (CI) of the oddsratios, and deviances.

Covariate Estimate SEOddsratio

95% CI ofodds ratio Deviance

Intercept 0.48 0.70 1.16 0.47 8.71Latitude 2.97 1.14 19.57 3.04 327.59 8.02Longitude −2.61 1.02 0.07 0.006 0.37 10.26Total length −0.10 0.59 0.90 0.26 2.75 0.49Encounters

(northernshoreline)

1.77 1.00 5.86 1.26 76.81 5.35

426 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

Alós, J., Palmer, M., and Arlinghaus, R. 2012. Consistent selection towards lowactivity phenotypes when catchability depends on encounters among hu-man predators and fish. PLoS ONE, 7(10): e48030. doi:10.1371/journal.pone.0048030. PMID:23110164.

Alós, J., Palmer, M., Rosselló, R., and Arlinghaus, R. 2016. Fast and behavior-selective exploitation of a marine fish targeted by anglers. Sci. Rep. 6: 38093.doi:10.1038/srep38093. PMID:27922022.

Andersen, K.H., Marty, L., and Arlinghaus, R. 2018. Evolution of boldness and lifehistory in response to selective harvesting. Can. J. Fish. Aquat. Sci. 75. Inpress. [Published online ahead of print.] doi:10.1139/cjfas-2016-0350.

Arlinghaus, R., and Mehner, T. 2004. A management-orientated comparative anal-ysis of urban and rural anglers living in a metropolis (Berlin, Germany). Environ.Manage. 33(3): 331–344. doi:10.1007/s00267-004-0025-x. PMID:15037955.

Arlinghaus, R., Klefoth, T., Kobler, A., and Cooke, S.J. 2008. Size selectivity,injury, handling time, and determinants of initial hooking mortality in rec-reational angling for northern pike: the influence of type and size of bait.N. Am. J. Fish. Manage. 28(1): 123–134. doi:10.1577/M06-263.1.

Arlinghaus, R., Tillner, R., and Bork, M. 2015. Explaining participation rates inrecreational fishing across industrialised countries. Fish. Manage. Ecol. 22(1):45–55. doi:10.1111/fme.12075.

Arlinghaus, R., Laskowski, K.L., Alós, J., Klefoth, T., Monk, C.T., Nakayama, S.,and Schröder, A. 2017a. Passive gear-induced timidity syndrome in wild fishpopulations and its potential ecological and managerial implications. FishFish. 18(2): 360–373. doi:10.1111/faf.12176.

Arlinghaus, R., Alós, J., Pieterek, T., and Klefoth, T. 2017b. Determinants ofangling catch of northern pike (Esox lucius) as revealed by a controlled whole-lake catch-and-release angling experiment — The role of abiotic and bioticfactors, spatial encounters and lure type. Fish. Res. 186: 648–657. doi:10.1016/j.fishres.2016.09.009.

Baktoft, H., Zajicek, P., Klefoth, T., Svendsen, J.C., Jacobsen, L., Pedersen, M.W.,March Morla, D., Skov, C., Nakayama, S., and Arlinghaus, R. 2015. Perfor-mance assessment of two whole-lake acoustic positional telemetry systems —is reality mining of free-ranging aquatic animals technologically possible?PLoS ONE, 10(5): e0126534. doi:10.1371/journal.pone.0126534. PMID:26000459.

Ballew, N.G., Mittelbach, G.G., and Scribner, K.T. 2017. Fitness consequences ofboldness in juvenile and adult largemouth bass. Am. Nat. 189(4): 396–406.doi:10.1086/690909. PMID:28350493.

Beardmore, B., Haider, W., Hunt, L.M., and Arlinghaus, R. 2013. Evaluating theability of specialization indicators to explain fishing preferences. Leis. Sci.35(3): 273–292. doi:10.1080/01490400.2013.780539.

Beukema, J.J. 1970. Acquired hook-avoidance in the pike Esox lucius L. fished withartificial and natural baits. J. Fish Biol. 2(2): 155–160. doi:10.1111/j.1095-8649.1970.tb03268.x.

Binder, T.R., Nannini, M.A., Wahl, D.H., Arlinghaus, R., Klefoth, T., Philipp, D.P.,and Cooke, S.J. 2012. Largemouth bass selected for differential vulnerabilityto angling exhibit similar routine locomotory activity in experimentalponds. Trans. Am. Fish. Soc. 141(5): 1252–1259. doi:10.1080/00028487.2012.688919.

Biro, P.A., and Post, J.R. 2008. Rapid depletion of genotypes with fast growth andbold personality traits from harvested fish populations. Proc. Natl. Acad. Sci.U.S.A. 105(8): 2919–2922. doi:10.1073/pnas.0708159105. PMID:18299567.

Bryan, H. 1977. Leisure value systems and recreational specialization: the case oftrout fishermen. J. Leis. Res. 9(3): 174.

Burnham, K.P., and Anderson, D.R. 2003. Model selection and multimodelinference: a practical information-theoretic approach. Springer, New York.

Camp, E.V., van Poorten, B.T., and Walters, C.J. 2015. Evaluating short openingsas a management tool to maximize catch-related utility in catch-and-releasefisheries. N. Am. J. Fish. Manage. 35(6): 1106–1120. doi:10.1080/02755947.2015.1083495.

Conrad, J.L., Weinersmith, K.L., Brodin, T., Saltz, J.B., and Sih, A. 2011. Behav-ioural syndromes in fishes: a review with implications for ecology and fish-eries management. J. Fish Biol. 78(2): 395–435. doi:10.1111/j.1095-8649.2010.02874.x. PMID:21284626.

Cooke, E.L.L., Wilson, A.D.M., Elvidge, C.K., and Cooke, S.J. 2017. Does capturemethod or the presence of aquatic protected areas influence the selectiveharvest of behavioural types in largemouth bass? Can. J. Fish. Aquat. Sci.74(8): 1151–1157. doi:10.1139/cjfas-2016-0192.

Cox, S., and Walters, C. 2002. Modeling exploitation in recreational fisheries andimplications for effort management on British Columbia rainbow trout lakes.N. Am. J. Fish. Manage. 22(1): 21–34. doi:10.1577/1548-8675(2002)022<0021:MEIRFA>2.0.CO;2.

Diaz Pauli, B., and Sih, A. 2017. Behavioural responses to human-inducedchange: why fishing should not be ignored. Evol. Appl. 10(3): 231–240. doi:10.1111/eva.12456. PMID:28250808.

Diaz Pauli, B., Wiech, M., Heino, M., and Utne-Palm, A.C. 2015. Opposite selectionon behavioural types by active and passive fishing gears in a simulated guppyPoecilia reticulata fishery. J. Fish Biol. 86(3): 1030–1045. doi:10.1111/jfb.12620.PMID:25619538.

Dingemanse, N.J., and Dochtermann, N.A. 2013. Quantifying individual varia-tion in behaviour: mixed-effect modelling approaches. J. Anim. Ecol. 82(1):39–54. doi:10.1111/1365-2656.12013. PMID:23171297.

Dorow, M., Beardmore, B., Haider, W., and Arlinghaus, R. 2010. Winners andlosers of conservation policies for European eel, Anguilla anguilla: an eco-

nomic welfare analysis for differently specialised eel anglers. Fish. Manage.Ecol. 17(2): 106–125. doi:10.1111/j.1365-2400.2009.00674.x.

Härkönen, L., Hyvärinen, P., Paappanen, J., Vainikka, A., and Tierney, K. 2014.Explorative behavior increases vulnerability to angling in hatchery-rearedbrown trout (Salmo trutta). Can. J. Fish. Aquat. Sci. 71(12): 1900–1909. doi:10.1139/cjfas-2014-0221.

Härkönen, L., Hyvärinen, P., Niemelä, P.T., and Vainikka, A. 2016. Behaviouralvariation in Eurasian perch populations with respect to relative catchability.Acta Ethol. 19: 21–31. doi:10.1007/s10211-015-0219-7.

Harrison, P.M., Gutowsky, L.F.G., Martins, E.G., Patterson, D.A., Cooke, S.J., andPower, M. 2015. Personality-dependent spatial ecology occurs independentlyfrom dispersal in wild burbot (Lota lota). Behav. Ecol. 26(2): 483–492. doi:10.1093/beheco/aru216.

Heermann, L., Emmrich, M., Heynen, M., Dorow, M., König, U., Borcherding, J.,and Arlinghaus, R. 2013. Explaining recreational angling catch rates of Eur-asian perch, Perca fluviatilis: the role of natural and fishing-related environ-mental factors. Fish. Manage. Ecol. 20(2–3): 187–200. doi:10.1111/fme.12000.

Heino, M., Díaz Pauli, B., and Dieckmann, U. 2015. Fisheries-induced evolution.Annu. Rev. Ecol. Evol. Syst. 46(1): 461–480. doi:10.1146/annurev-ecolsys-112414-054339.

Hothorn, T., Bretz, F., Westfall, P., Heiberger, R.M., Schuetzenmeister, A., Scheibe, S.,and Hothorn, M.T. 2016. Package “multcomp.” http://cran.stat.sfu.ca/web/packages/multcomp/multcomp.pdf.

Hougaard, P. 1999. Fundamentals of survival data. Biometrics, 55(1): 13–22. doi:10.1111/j.0006-341X.1999.00013.x. PMID:11318147.

Jacobsen, L., Berg, S., Baktoft, H., and Skov, C. 2015. Behavioural strategy of largeperch Perca fluviatilis varies between a mesotrophic and a hypereutrophiclake. J. Fish Biol. 86(3): 1016–1029. doi:10.1111/jfb.12613. PMID:25644471.

Janzen, F.J., and Hal, S. 1998. Logistic regression for empirical studies of multi-variate selection. Evolution, 52(6): 1564–1571. doi:10.2307/2411330.

Johnston, F.D., Arlinghaus, R., and Dieckmann, U. 2010. Diversity and complex-ity of angler behaviour drive socially optimal input and output regulations ina bioeconomic recreational-fisheries model. Can. J. Fish. Aquat. Sci. 67(9):1507–1531. doi:10.1139/F10-046.

Kekäläinen, J., Podgorniak, T., Puolakka, T., Hyvärinen, P., and Vainikka, A. 2014.Individually assessed boldness predicts Perca fluviatilis behaviour in shoals,but is not associated with the capture order or angling method. J. Fish. Biol.85(5): 1603–1616. doi:10.1111/jfb.12516. PMID:25270290.

Killen, S.S., Nati, J.J.H., and Suski, C.D. 2015. Vulnerability of individual fish tocapture by trawling is influenced by capacity for anaerobic metabolism.Proc. R. Soc. B Biol. Soc. 282(1813): 20150603. doi:10.1098/rspb.2015.0603.

Klefoth, T., Kobler, A., and Arlinghaus, R. 2008. The impact of catch-and-releaseangling on short-term behaviour and habitat choice of northern pike (Esoxlucius L.). Hydrobiologia, 601: 99–110. doi:10.1007/s10750-007-9257-0.

Klefoth, T., Skov, C., Krause, J., and Arlinghaus, R. 2012. The role of ecologicalcontext and predation risk-stimuli in revealing the true picture about thegenetic basis of boldness evolution in fish. Behav. Ecol. Sociobiol. 66(4):547–559. doi:10.1007/s00265-011-1303-2.

Kuparinen, A., Klefoth, T., and Arlinghaus, R. 2010. Abiotic and fishing-relatedcorrelates of angling catch rates in pike (Esox lucius). Fish. Res. 105(2): 111–117.doi:10.1016/j.fishres.2010.03.011.

Laskowski, K.L., Monk, C.T., Polverino, G., Alós, J., Nakayama, S., Staaks, G.,Mehner, T., and Arlinghaus, R. 2016. Behaviour in a standardized assay, butnot metabolic or growth rate, predicts behavioural variation in an adultaquatic top predator Esox lucius in the wild. J. Fish Biol. 88(4): 1544–1563.doi:10.1111/jfb.12933. PMID:26947935.

Lennox, R., Alós, J., Arlinghaus, R., Horodysky, A., Klefoth, T., Monk, C., andCooke, S. 2017. What makes fish vulnerable to capture by hooks? A concep-tual framework and a review of key determinants. Fish Fish. 18(5): 986–1010.doi:10.1111/faf.12219.

Lewin, W.-C., Arlinghaus, R., and Mehner, T. 2006. Documented and potentialbiological impacts of recreational fishing: insights for management and con-servation. Rev. Fish. Sci. 14(4): 305–367. doi:10.1080/10641260600886455.

Løkkeborg, S., Siikavuopio, S.I., Humborstad, O.-B., Utne-Palm, A.C., andFerter, K. 2014. Towards more efficient longline fisheries: fish feeding behav-iour, bait characteristics and development of alternative baits. Rev. Fish Biol.Fish. 24(4): 985–1003. doi:10.1007/s11160-014-9360-z.

Magnhagen, C. 2012. Personalities in a crowd: what shapes the behaviour ofEurasian perch and other shoaling fishes? Curr. Zool. 58(1): 35–44. doi:10.1093/czoolo/58.1.35.

Magnhagen, C., Hellström, G., Borcherding, J., Heynen, M., and Genner, M. 2012.Boldness in two perch populations — long-term differences and the effect ofpredation pressure. J. Anim. Ecol. 81(6): 1311–1318. doi:10.1111/j.1365-2656.2012.02007.x. PMID:22694656.

Matsumura, S., Arlinghaus, R., and Dieckmann, U. 2012. Standardizing selectionstrengths to study selection in the wild: a critical comparison and suggestionsfor the future. Bioscience, 62(12): 1039–1054. doi:10.1525/bio.2012.62.12.6.

Matthias, B.G., Allen, M.S., Ahrens, R.N.M., Beard, T.D., and Kerns, J.A. 2014. Hideand seek: interplay of fish and anglers influences spatial fisheries manage-ment. Fisheries, 39(6): 261–269. doi:10.1080/03632415.2014.903836.

Mittelbach, G.G., Ballew, N.G., Kjelvik, M.K., and Fraser, D. 2014. Fish behavioraltypes and their ecological consequences. Can. J. Fish. Aquat. Sci. 71(6): 927–944. doi:10.1139/cjfas-2013-0558.

Monk and Arlinghaus 427

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.

Monk, C.T., and Arlinghaus, R. 2017. Encountering a bait is necessary but insuf-ficient to explain individual variability in vulnerability to angling in twofreshwater benthivorous fish in the wild. PLoS ONE, 12(3): e0173989. doi:10.1371/journal.pone.0173989. PMID:28301558.

Nakagawa, S., and Schielzeth, H. 2010. Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists. Biol. Rev. 85(4): 935–956.PMID:20569253.

Nakayama, S., Laskowski, K.L., Klefoth, T., and Arlinghaus, R. 2016. Between- andwithin-individual variation in activity increases with water temperature inwild perch. Behav. Ecol. 27(6): 1676–1683. doi:10.1093/beheco/arw090.

Nakayama, S., Rapp, T., and Arlinghaus, R. 2017. Fast–slow life history is corre-lated with individual differences in movements and prey selection in anaquatic predator in the wild. J. Anim. Ecol. 86(2): 192–201. doi:10.1111/1365-2656.12603. PMID:27748952.

Nannini, M.A., Wahl, D.H., Philipp, D.P., and Cooke, S.J. 2011. The influence ofselection for vulnerability to angling on foraging ecology in largemouth bassMicropterus salmoides. J. Fish Biol. 79(4): 1017–1028. doi:10.1111/j.1095-8649.2011.03079.x. PMID:21967587.

Niemelä, P.T., and Dingemanse, N.J. 2014. Artificial environments and the studyof “adaptive” personalities. Trends Ecol. Evol. 29(5): 245–247. doi:10.1016/j.tree.2014.02.007. PMID:24629720.

Olsen, E.M., Heupel, M.R., Simpfendorfer, C.A., and Moland, E. 2012. Harvestselection on Atlantic cod behavioral traits: implications for spatial manage-ment. Ecol. Evol. 2(7): 1549–1562. doi:10.1002/ece3.244. PMID:22957161.

Pagel, T., Bekkevold, D., Pohlmeier, S., Wolter, C., and Arlinghaus, R. 2015.Thermal and maternal environments shape the value of early hatching in anatural population of a strongly cannibalistic freshwater fish. Oecologia,178(4): 951–965. doi:10.1007/s00442-015-3301-y. PMID:25894093.

Payer, R.D., Pierce, R.B., and Pereira, D.L. 1989. Hooking mortality of walleyescaught on live and artificial baits. N. Am. J. Fish. Manage. 9(2): 188–192.doi:10.1577/1548-8675(1989)009<0188:HMOWCO>2.3.CO;2.

Persson, L., Claessen, D., Roos, A.M., De Byström, P., Persson, L., Claessen, D.,De Roos, A.M., Bystrom, P.A.R., Sjogren, S., Svanback, R., Wahlstrom, E.V.A.,and Westman, E. 2004. Cannibalism in a size-structured population: energyextraction and control. Ecol. Monogr. 74(1): 135–157. doi:10.1890/02-4092.

Seaman, D.E., Millspaugh, J.J., Kernohan, B.J., Brundige, G.C., Raedeke, K.J., andGitzen, R.A. 1999. Effects of sample size on kernel home range estimates.J. Wildl. Manage. 63(2): 739–747. doi:10.2307/3802664.

Seekell, D.A. 2011. Recreational freshwater angler success is not significantlydifferent from a random catch model. N. Am. J. Fish. Manage. 31(2): 203–208.doi:10.1080/02755947.2011.572788.

Sutter, D.A., Suski, C.D., Philipp, D.P., Klefoth, T., Wahl, D.H., Kersten, P.,

Cooke, S.J., and Arlinghaus, R. 2012. Recreational fishing selectively capturesindividuals with the highest fitness potential. Proc. Natl. Acad. Sci. U.S.A.109(51): 20960–20965. doi:10.1073/pnas.1212536109. PMID:23213220.

Svanbäck, R., and Persson, L. 2009. Population density fluctuations change theselection gradient in Eurasian perch. Am. Nat. 173(4): 507–516. doi:10.1086/597223. PMID:19226234.

Tsuboi, J.-i., Morita, K., Klefoth, T., Endou, S., and Arlinghaus, R. 2016. Behaviour-mediated alteration of positively size-dependent vulnerability to angling inresponse to historical fishing pressure in a freshwater salmonid. Can. J. Fish.Aquat. Sci. 73(3): 461–468. doi:10.1139/cjfas-2014-0571.

Uusi-Heikkilä, S., Wolter, C., Klefoth, T., and Arlinghaus, R. 2008. A behavioralperspective on fishing-induced evolution. Trends Ecol. Evol. 23(8): 419–421.doi:10.1016/j.tree.2008.04.006. PMID:18582988.

Vainikka, A., Tammela, I., and Hyvärinen, P. 2016. Does boldness explain vulner-ability to angling in Eurasian perch Perca fluviatilis? Curr. Zool. 62(2): 109–115.doi:10.1093/cz/zow003.

Villegas-Ríos, D., Réale, D., Freitas, C., Moland, E., and Olsen, E.M. 2017. Individ-ual level consistency and correlations of fish spatial behaviour assessed fromaquatic animal telemetry. Anim. Behav. 124: 83–94. doi:10.1016/j.anbehav.2016.12.002.

Ward, H.G.M., Askey, P.J., and Post, J.R. 2013. A mechanistic understanding ofhyperstability in catch per unit effort and density-dependent catchability ina multistock recreational fishery. Can. J. Fish. Aquat. Sci. 70(10): 1542–1550.doi:10.1139/cjfas-2013-0264.

Wiig, J.R., Moland, E., Haugen, T.O., and Moland Olsen, E. 2014. Spatially struc-tured interactions between lobsters and lobster fishers in a coastal habitat:fine-scale behaviour and survival estimated from acoustic telemetry. Can. J.Fish. Aquat. Sci. 70(10): 1468–1476. doi:10.1139/cjfas-2013-0209.

Wilde, G., Riechers, R., and Ditton, R. 1998. Differences in attitudes, fishingmotives, and demographic characteristics between tournament and non-tournament black bass anglers in Texas. N. Am. J. Fish. Manage. 18: 422–431.doi:10.1577/1548-8675(1998)018<0422:DIAFMA>2.0.CO;2.

Wilson, A.D.M., Binder, T.R., McGrath, K.P., Cooke, S.J., and Godin, J.-G.J. 2011.Capture technique and fish personality: angling targets timid bluegill sun-fish, Lepomis macrochirus. Can. J. Fish. Aquat. Sci. 68(5): 749–757. doi:10.1139/f2011-019.

Wilson, A.D.M., Brownscombe, J.W., Sullivan, B., Jain-Schlaepfer, S., andCooke, S.J. 2015. Does angling technique selectively target fishes based ontheir behavioural type? PLoS ONE, 10(8): e135848. doi:10.1371/journal.pone.0135848.

Zuur, A.F., Ieno, E.N., and Smith, G.M. 2007. Analysing ecological data. series: statis-tics for biology and health. Edited by M. Gail, K. Krickeb, J.M. Samet, A. Tsiatis, andW. Wong. Springer-Verlag, New York. doi:10.1007/978-0-387-45972-1.

428 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y L

EIB

NIT

Z-I

NST

ITU

T F

UR

on

02/2

3/18

For

pers

onal

use

onl

y.