lable at ScienceDirect

Animal Behaviour 96 (2014) 177e186

Contents lists avai

Animal Behaviour

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

Environmental factors affecting behavioural responses of an invasivebivalve to conspecific alarm cues

Jarosław Kobak*, Anna Ry�nskaNicolaus Copernicus University, Faculty of Biology and Environmental Protection, Department of Invertebrate Zoology, Toru�n, Poland

a r t i c l e i n f o

Article history:Received 18 April 2014Initial acceptance 28 May 2014Final acceptance 6 August 2014Published onlineMS. number: 14-00322

Keywords:aggregationalarm substanceantipredator defenceDreissena polymorphalightlocomotionsubstratum inclinationzebra mussel

* Correspondence: J. Kobak, Nicolaus Copernicus Uand Environmental Protection, Department of Inverteb100 Toru�n, Poland.

E-mail address: jkob73@umk.pl (J. Kobak).

http://dx.doi.org/10.1016/j.anbehav.2014.08.0140003-3472/© 2014 The Association for the Study of A

Antipredator defences of aquatic animals depend on various environmental parameters. We studiedbehavioural responses of a Ponto-Caspian invasive bivalve, the zebra mussel, Dreissena polymorpha, toconspecific alarm cues. We hypothesized that mussels would change their locomotion and aggregation inresponse to alarm signals. We also hypothesized that body size, light, substratum quality (suitable orunsuitable for attachment) and inclination would affect mussel defences. Changes in horizontal move-ment of mussels exposed to the alarm substance depended on light. In the presence of crushed con-specifics illuminated mussels (all sizes) moved longer distances than control individuals, whereas indarkness their reaction was the opposite. The response of small mussels was the strongest. Furthermore,the alarm substance reduced upward relocations of all size groups on an inclined surface but at the sametime stimulated their downward movement. Large and medium mussels (but not small individuals)exposed to alarm signals formed aggregations more often than control individuals. This effect was onlyexhibited on sand, unsuitable for mussel attachment. Mussels were generally more clumped on sandthan on hard substratum, suggesting that they did not prefer conspecific shells as attachment sites whenalternative substrata were available. All responses of mussels to alarm cues tended to be stronger in light,which is an indirect indication of danger. Our study shows that the responses of the zebra mussel toconspecific alarm cues are not limited to activity reduction, as previously thought. They are adjusted toparticular environmental conditions and may also involve increased locomotion when relocation to asafer site gives a better chance of survival.© 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Predators shape their environment not only by killing prey, butalso by frightening their potential victims (Brown, Laundr�e, &Gurung, 1999). This increases the level of stress in prey pop-ulations, affecting their behaviour and physiology, which in turnexerts prolonged, multidirectional effects on reproduction andsurvival (Clinchy, Sheriff,& Zanette, 2013). Prey organisms exhibit anumber of antipredator strategies, most of them being energeti-cally costly and negatively affecting living conditions and/orfecundity (De Meester, Dawidowicz, Van Gool, & Loose, 1999).

Defences of prey organisms induced by predation cues arecommon in the aquatic environment (Gliwicz, 2005). They includechanges in behaviour (De Meester et al., 1999), morphology(Dzialowski, Lennon, O'Brien, & Smith, 2003) and life histories(�Slusarczyk, Dawidowicz, & Rygielska, 2005) and may substantiallyaffect ecosystem functioning by modifying habitat selection byanimals, their reproductive success or foraging intensity

niversity, Faculty of Biologyrate Zoology, Lwowska 1, 87-

nimal Behaviour. Published by Els

(De Meester et al., 1999; Naddafi, Ekl€ov, & Pettersson, 2007). Asbehavioural defences are very efficient, predation success may begreatly limited, although predators are often able to overcome preydefences (Gliwicz, 2005). None the less, predatoreprey interactionsshape the functioning of communities and evolution of both preyand predator species to a large extent. Responses to predation cuesmay be modified by numerous environmental factors, such as light(De Meester et al., 1999) or substratum type (Baumg€artner, Koch, &Rothhaupt, 2003), which influence the risk level perceived by preyunder particular conditions and provide different possibilities forhiding or escape.

The relationship between zooplankton and planktivorous fish isthe best known model of a system of behavioural antipredatordefences in the aquatic environment (Lass & Spaak, 2003), butstrong evidence for similar mechanisms in bottom-dwelling ani-mals also exists (Baumg€artner et al., 2003; Koperski, 1997; Krist,2002). An interesting case is that of sessile species with limitedmobility, such as bivalves (Cot�e & Jelnikar, 1999; Zardi, Nicastro,McQuaid, Rius, & Porri, 2006). They cannot escape from their en-emies, but decrease the risk of predation by reducing their activity,selecting sheltered sites, adjusting attachment strength (Ishida &

evier Ltd. All rights reserved.

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186178

Iwasaki, 2003; Naddafi & Rudstam, 2013; Reimer & Harms-Ringdahl, 2001; Reimer & Tedengren, 1997) and, over a longerperiod, developing heavier shells (Naddafi & Rudstam, 2014) orstronger adductor muscles (Reimer & Harms-Ringdahl, 2001).

Various types of antipredator defences have been observed inthe freshwater zebra mussel, Dreissena polymorpha (Czarnołeski,Müller, Adamus, Ogorzelska, & Sog, 2010; Kobak & Kakareko,2009; Naddafi et al., 2007). This Ponto-Caspian species is alienin most of its present range in Europe and North America(Karatayev, Burlakova, & Padilla, 2002), but responds to bothsympatric predators and those with a very short coexistencetime, such as crayfish and fish of American origin (Naddafi &Rudstam, 2013). The ability to adjust quickly to new predatorsmay contribute to the impressive invasion success of this species(Karatayev et al., 2002), given the large number of its potentialenemies in novel areas (Molloy, Karatayev, Burlakova, Kurandina,& Laruelle, 1997). Interestingly, responses of the zebra mussel tothe presence of predators and alarm substances produced byinjured conspecifics seem to differ from each other. The pres-ence of predators makes mussels attach more strongly, selectsafer sites, form aggregations and reduce upward climbing(Houghton & Janssen, 2013; Kobak, Kakareko & Pozna�nska,2010; Kobak & Kakareko, 2009; Naddafi & Rudstam, 2013).Moreover, predators induce reduced but more selective feeding(Naddafi et al., 2007), shell thickening and lower growth rate(Naddafi & Rudstam, 2014). In contrast, injured conspecificsinhibit overall activity, including byssogenesis, thus reducingboth adhesion and locomotion (Czarnołeski et al., 2010;Czarnołeski, Müller, Kierat, Gryczkowski, & Chybowski, 2011;Toomey, McCabe, & Marsden, 2002). Despite these findings,the effect of abiotic environmental factors upon the variability ofmussel responses to conspecific alarm cues has not yet beenstudied, although they are likely to change the vulnerability ofmussels to potential predators and affect their defence mecha-nisms (Czarnołeski et al., 2006; Gownaris & Commito, 2008;Kobak & Nowacki, 2007).

To address the gaps in our knowledge of zebra mussel re-sponses to alarm stimuli, we studied their horizontal and ver-tical relocations as well as aggregation forming in the presenceof crushed conspecifics. So far, the latter two behaviours havenot been investigated in the light of responses to conspecificalarm cues. We expected inhibition of mussel activity in theperceived presence of danger from predation. Furthermore, wehypothesized that the defensive behaviour of mussels would beaffected by their body size and by several environmental factors,such as light, substratum type and gravity. We expected astronger response of mussels in light, an indirect sign of expo-sure to predation risk (Kobak & Nowacki, 2007), and on softsubstratum, unsuitable for attachment. We assumed that en-dangered mussels would more often move downwards and thatsmaller individuals, vulnerable to a wider range of predators(Czarnołeski et al., 2006), would be more responsive than thelargest specimens.

METHODS

Mussels (ca. 10 000e12000 individuals) were collected by adiver from the Włocławek Reservoir, a dam reservoir on the lowerRiver Vistula (Central Poland) from a depth of ca. 2 m. They weretransported to the laboratory (2 h transport time) in 10-litre food-grade plastic containers (ca. 3000 individuals per container) inaerated water. We kept them in two 350-litre stock tanks on hardplastic trays, to which they could attach themselves. Their densitywas ca. 7000e8000 individuals per square metre, which is acommon density at which this species occurs in thewild (Karatayev

et al., 2002). They were fed every second day with ca. 2 g of driedChlorella sp. per 1000 mussels (Kilgour & Baker, 1994). However,mussels were not fed during the experiments. Water was aeratedand filtered, the temperature was 16e18 �C, and the photoperiodwas 9:15 h light:dark. We did not observe any negative effects oftransport and stocking conditions on mussel survival. The musselswere used in the experiments within 3 months of collecting. Afterthe experiments, the mussels were humanely killed by freezing.

We conducted our experiments in tap water, aerated andfiltered for at least 24 h in a 100-litre tank without any animalsbefore use. We checked water quality during the experiments witha multimeter Multi340i (WTWGmbH, Weilheim, Germany). Watertemperature (mean ± SD) was 19.1 ± 1.9 �C, oxygen concentration7.5 ± 1.5 mg/litre (saturation: 78.1 ± 14.8%), pH 8.2 ± 0.3 and con-ductivity 525 ± 100 mS/cm. We conducted the experiments indarkness obtained by using Styrofoam curtains, or under constantfluorescent light (550e650 lx) at the water surface (determinedwith a luxometer L-20A, Sonopan Ltd., Białystok, Poland).

The mussels were divided into three size classes: small (meanlength: 8.7 mm; range 7.3e9.5 mm), medium (14.5; 10.9e18.3 mm)and large (26.4; 22.8e32.5 mm). These groups differ from oneanother in their mobility (Toomey et al., 2002) and antipredatordefences (Kobak & Kakareko, 2009).

We prepared the alarm substance according to Toomey et al.(2002), by crushing living mussels (all sizes mixed) in water(volumetric proportion of mussels to water: 3:7) and adding1e2 ml of this solution to treated experimental dishes, dependingon their size. The number of crushed individuals was kept at theminimum necessary for a given set of trials. Altogether, we used ca.1000 individuals for crushing. The final concentration of thecrushed mussel extract in the treated dishes was 2.2e2.6 ml per1 litre of water. Czarnołeski et al. (2010) and Toomey et al. (2002)have shown that this amount of extract does affect mussel behav-iour. It was necessary to use livingmussels to prepare the extract, aswe needed the active alarm substance, produced by living preyindividuals attacked by a predator (Czarnołeski et al., 2010; Toomeyet al., 2002).

In all experiments, trials with different treatments (with regardto the presence of the alarm substance, mussel size, light andsubstratum type) were carried out in a random sequence.

All experimental procedures applied in our study were carriedout in accordance with the Polish law and ethical guidelines of thePolish National Ethics Committee for Experiments on Animals.

Experiment 1: Horizontal Migration

We put a single mussel in the centre of a glass circular dish witha thin layer of fine sand (ca. 1e2 mm) on the bottom (Fig. 1a).Immediately after placing the mussels in the dishes, we added 1 mlof the alarm substance to the treated dishes. After 24 h of exposurein darkness or under constant illumination, we photographed thetrails left in the sand by crawlingmussels andmeasured them usingImageJ 1.48a (freeware by W. S. Rasband, U.S. National Institutes ofHealth, Bethesda, MD, U.S.A., http://rsb.info.nih.gov/i) according tothe procedure described by Toomey et al. (2002). Unfortunately, itwas impossible to test groups of mussels in this experiment, whichwould be a more natural situation for this gregarious species, asthey would obscure one another's trails in the dish. We replicatedthis experiment 20 times.

Experiment 2: Vertical Migration

We studied mussel behaviour on a glass slope, inclined at 6degrees to the bottom of an experimental tank (Fig. 1b). Thisinclination can be detected bymussels but does not make them slip

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186 179

passively down the slope (Kobak, 2006). We put five small or me-dium mussels or four large individuals (as more of them did not fiton the starting line) along the line dividing the slope in half, withtheir longer axes parallel to the shorter side of the tank (Fig.1b). Thesubstratum in this experiment was bare glass. The mussels couldturn and move up or down the slope. At the beginning of theexperiment, we added 2 ml of the alarm substance to the treatedtanks. After 24 h of exposure in darkness or under constant light,wemeasuredmussel relocations from their initial position in eitherdirection (only linear vertical displacements, ignoring any sidemovements) to the nearest 1 cm. We obtained a net relocation ofmussels in each replicate by averaging their relocations, withpositive values assigned to upward movements, negative values todownward movements and zeros to immobile specimens. Thus, apositive net relocation indicated that mussels selected the upwarddirection, whereas a negative value showed a tendency to movedown. A net relocation close to zero could result from equalmovements of mussels in both directions or from no movement. Todiscriminate between these possibilities, we also calculated abso-lute relocations of mussels, irrespective of the direction of theirmovement. We replicated this experiment 18 times.

Experiment 3: Aggregation Forming

We put five mussels into the same circular dishes as in experi-ment 1 (Fig.1c), in the pattern shown in Fig.1d. The substratumwaseither sand (a layer of ca. 20 mm) or the bare glass bottom of thedish. Soft sand was unsuitable for mussel attachment, so they hadto search for another adhesion site (e.g. conspecific shells), whereasthey could easily attach to the glass bottom. We prevented themovement of mussels to the glass wall of the dish by surroundingthem with a plastic mesh cylinder (Fig. 1c). Zebra mussels avoidattaching to mesh objects (Porter &Marsden, 2008), so they had tofind an attachment site on the part of the bottom confined by thecylinder. The initial distances between neighbouring mussels wereequal. They were put halfway between the dish centre and themesh wall of the cylinder, with their anterior ends pointing inside.We conducted this experiment in darkness or under constant lightfor 48 h to allow for cluster formation. We added 1 ml of the alarmsubstance to the treated dishes twice: at the beginning and after24 h of exposure. At the end of the experiment, we determined thenumbers of aggregatedmussels (touching one another's shells). Wereplicated this experiment 10 times.

Statistical Analysis

All the analyses described below were carried out using SPSS 21(IBM Corporation) except t tests, for which we used Microsoft Excel2010.

We analysed horizontal distances moved by mussels (experiment1), as well as net and absolute vertical relocations of mussels (exper-iment 2) using a three-way ANOVAwith Mussel size (large, medium,small), Light conditions (light, darkness) and Alarm cue (present ornot) as factors. Additionally, to check whether one of the verticalmovement directions in experiment 2 was preferred or avoided bymussels, we used sequential Bonferroni-corrected one-sample t testscomparing the actual net relocations in particular treatments withzero values, expected if both directions were equally attended.

We analysed percentages of aggregated mussels (experiment 3)using a four-way ANOVA with Mussel size, Light conditions, Alarmcue and Substratum (soft, hard) as factors.

We log-transformed the data from experiments 1 and 2 (exceptfor the net movement, where a transformation was impossibleowing to the presence of negative values) and arcsine-transformedthe percentage data from experiment 3 to reduce the violations of

normality and homoscedasticity assumptions. We further analysedsignificant ANOVA effects using sequential Bonferroni-correctedFisher LSD tests.

RESULTS

Experiment 1: Horizontal Migration

Almost all small mussels (99%) and 50% of the other individualsmoved from their initial position after 24 h of the experiment. Thelongest path was measured for a small illuminated individualexposed to the alarm substance, which moved a distance of 1.3 m,with many loops and meanders on its way.

Mussel responses to the alarm substance depended on lightconditions, as shown by a significant Alarm cue)Light interaction(Table 1). Illuminated mussels exposed to the alarm signal movedgreater distances than control individuals, whereas the reaction ofmussels kept in darkness was the opposite (Fig. 2). This patternwasconsistent for all size classes, but the strongest responsewas shownby small mussels.

Light inhibited movement of large andmediummussels, but notthat of small individuals, resulting in a significant Light)Sizeinteraction (Table 1). Moreover, small individuals weremostmobileand large specimens least active both in light and in darkness(Appendix Table A1).

Experiment 2: Vertical Migration

Mussels often relocated from their initial position; 79, 61 and64% of small, medium and large individuals, respectively, movedalong the slope in either direction. They did not attach to one an-other's shell, so we can assume that all individuals moved activelyto their final sites (i.e. they were not transported while attached toconspecifics). The alarm substance clearly switched the net relo-cation (Fig. 3a) of all mussels towards more negative values (Alarmcue effect; Table 2). Furthermore, smaller mussel size and darknessincreased the tendency to move upwards (a significant Light)Sizeinteraction; Table 2, Appendix Table A2). The control small in-dividuals exposed in darkness were the only group showing apreference for upward locomotion (Fig. 3a). In contrast, all largemussels and illuminated medium individuals exposed to the alarmcue more often selected the downward direction (Fig. 3a).

The presence of alarm signals affected the absolute relocations(independent of the direction; Fig. 3b) only of large mussels,making them move longer distances (a significant Alarm cue)Sizeinteraction, Table 3). This was a consequence of the downwardmovement stimulation shown above, as largemussels did notmoveupwards at all. Absolute relocations of the other size classes werenot affected by the alarm substance (Appendix Table A3), showingthat the stimulation of downward movement by alarm signals inthese groups was associated with the simultaneous inhibition ofupward movement (Fig. 3). Individual size and light also affectedabsolute relocations of mussels, as shown by a significant Light)Size interaction (Table 3). Illuminated mussels relocated overshorter distances than those kept in darkness, except small in-dividuals, for which this difference was not significant (AppendixTable A3). Also, the larger the mussels, the shorter the distancesthey moved (Fig. 3b, Appendix Table A3).

Experiment 3: Aggregation Forming

Mussels were more aggregated on sand than on glass (Fig. 4).Moreover, any effects of light, mussel size and alarm substancewere significant only on the sandy substratum, as shown by sig-nificant interactions of Substratum with the other variables

Figure 1. Experimental set-up for (a) experiment 1 (horizontal movement), (b) experiment 2 (vertical movement) and (c, d) experiment 3 (aggregation forming). Dimensions aregiven in mm.

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186180

(Table 4). The presence of alarm signals stimulated aggregation ofmedium and large mussels, but not that of small individuals (Ap-pendix Table A4), which resulted in a significant Alarm cue)Sizeinteraction (Table 4). Moreover, mussel aggregationwas affected bylight (a significant Light)Size)Substratum interaction, Table 4),with medium and large individuals on sandy substratum beingmore clustered in darkness than in light (Appendix Table A4).

Table 1Three-way ANOVA of horizontal movement of zebra mussels (experiment 1)

Effect df Mean square F P

Alarm cue (A) 1 0.75 0.25 0.618Light (L) 1 58.46 19.47 <0.001Size (S) 2 409.32 136.36 <0.001A�L 1 26.71 8.90 0.003A�S 2 0.19 0.06 0.938L�S 2 25.38 8.46 <0.001A�L�S 2 1.03 0.34 0.710Error 228 3.00

DISCUSSION

In the presence of conspecific alarm cues, mussels changed theirhorizontal locomotion intensity, increaseddownwardmovement andformed aggregations. Even large mussels (>20mm), which did notrespond to the presence of molluscivorous fish in a previous study(Kobak & Kakareko, 2009), were affected by alarm signals. Large in-dividuals experience relatively lower predation risk (Czarnołeskiet al., 2006; Prejs, Lewandowski, & Sta�nczykowska, 1990) and prob-ably only respond to direct danger, indicated by wounded conspe-cifics. Also, a change in horizontal movement, induced by the alarmsubstance in our study, was not observed in the presence of fishpredators (Kobak & Kakareko, 2009). Thus, conspecific alarm signalsseem to be a stronger stimulus than predator kairomones.

Research has shown that zebra mussels in the presence ofconspecific alarm cues reduce all aspects of their activity, includingeven their byssogenesis (Czarnołeski et al., 2010, 2011). Althoughweakly attached mussels are more vulnerable to predation (Kobak& Kakareko, 2011), reduced activity may help them remain unde-tected by a predator (Ishida& Iwasaki, 2003). However, in our study

800

700

500

400

300

200

100

0

600

Dis

tan

ce m

oved

(m

m)

Control Alarm

Light

Large mussels (>20 mm) Medium mussels (10−20 mm) Small mussels (<10 mm)

LightDarkness Darkness Light Darkness

Alarm Alarm Alarm Alarm AlarmControl Control Control Control Control

Figure 2. Distances moved by zebra mussels on horizontal sandy substratum in experiment 1. Error bars indicate SEMs.

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186 181

mussels displayed a more complicated behaviour and, undercertain conditions, responded to alarm cues by increasing theirmobility, showing that their defensive mechanisms were modifiedby a number of other environmental factors.

Horizontal movement, contrary to previous findings(Czarnołeski et al., 2010; Toomey et al., 2002), was greater in thepresence of alarm cues, although only in light. Our experimentslasted for 24 h, whereas the experiments conducted by Czarnołeskiet al. (2010) and Toomey et al. (2002) lasted only 2 h. Thus, it ispossible that mussels, after an initial reduction of movementinduced by the presence of the alarm cue (which was observed inthe shorter studies), changed their behaviour and started to searchfor a refuge (shown in our study as increased locomotion). Thistime-related change in antipredator behaviour was observed inbluemussels,Mytilus sp., which inhibited their byssogenesis duringa short (<10 h) exposure to predators, whereas after a longer time(24 h) the attachment of individuals exposed to predators exceededthat of control mussels (Reimer & Tedengren, 1997).

In the uniform light, an unattached and unsheltered musselwas easy to detect by visual predators regardless of its mobilityand it could only reduce the risk by active searching for anyprotection. Predator-induced increase in activity was also found inplanktonic Daphnia galeata, which increased their swimmingspeed to minimize time spent within the strike volume of apredator (Weber & Van Noordwijk, 2002). In darkness, musselsexposed to alarm cues moved over shorter distances, as expectedfor animals avoiding detection by chemical signals (Ishida &Iwasaki, 2003).

The presence of alarm cues also affected vertical movement ofmussels, increasing their downward crawling and decreasing up-ward relocation. In the field, such behaviour would lead to theoccupation of deeper sites, lower layers of a colony and/or loweredges of rocks and plant bases. Such locations are generally lesssuitable with regard to environmental conditions (food, oxygen,

wastes; Burks, Tuchman, Call, & Marsden, 2002; Tuchman, Burks,Call, & Smarrelli, 2004), but better protected from predation(Houghton & Janssen, 2013). Regardless of the presence of alarmcues, smaller mussels generally climbed upwards more often thanlarger individuals. This behaviour corroborates earlier observations(Burks et al., 2002; Kobak, 2006) and may be explained by smallmussels avoiding competition from larger conspecifics and lesssuitable conditions inside a dense colony (Burks et al., 2002).

A decrease in upward climbing was previously shown by zebramussels exposed to predatory roach, Rutilus rutilus (Kobak &Kakareko, 2009; Kobak, Kakareko, & Pozna�nska, 2010) and bymarine blue mussels in the presence of predator kairomones(Reimer & Tedengren, 1997). Downward migrations or limitation ofupward movements were also observed in numerous species en-dangered by open-water predators, such as zooplankton (DeMeester et al., 1999; Lass & Spaak, 2003) or amphipods(Wudkevich, Wisenden, Chivers, & Smith, 1997). The direction ofvertical movement may be affected by the quality of danger cues, asshown by an intertidal gastropod Planaxis sulcatus, which hides increvices in response to crushed predator cues, but climbs upwardsabove the waterline in the presence of a drilling predator (McKillup& McKillup, 1993).

We have shown that the reduction in upward movement is in-dependent of the overall decrease in activity, which might be aresult of exposure to alarm cues (Ishida & Iwasaki, 2003). In ourexperimental set-up, mussels were potentially able to crawl in bothdirections and avoided only the upward movement, whereas theirdownward relocation was stimulated by the alarm cue. Perhapsthat is why large mussels were the only group exhibiting an in-crease in overall relocation on the slope. They always moveddownwards, so no decrease in upward movement could take placein this group comparedwith the control treatments. Except for this,no significant changes in overall locomotion ofmussels on the slopetook place in response to alarm cues. This result seems somewhat

*

*

*

*

*

*

Control Alarm

Light

Large mussels (>20 mm) Medium mussels (10−20 mm) Small mussels (<10 mm)

LightDarkness Darkness Light Darkness

Alarm Alarm Alarm Alarm AlarmControl Control Control Control Control

Control Alarm

Light

Large mussels (>20 mm) Medium mussels (10−20 mm) Small mussels (<10 mm)

LightDarkness Darkness Light Darkness

Alarm Alarm Alarm Alarm AlarmControl Control Control Control Control

(a)

(b)

30

20

10

10

20

30

40

Dow

nw

ard

s

Net

tra

nsl

ocat

ion

(m

m)

Up

war

ds

70

60

50

40

30

20

10

0

0

40

Abs

olu

te t

ran

sloc

atio

n (

mm

)

Figure 3. Vertical relocation of zebra mussels on an inclined glass surface in experiment 2. (a) Mean net relocation; positive and negative values were assigned to individualsmoving upwards and downwards, respectively. (b) Absolute relocation, calculated for all mussels irrespective of the direction of their movement. Error bars indicate SEMs. Asterisksshow that average mussel displacement in a particular treatment deviated significantly from 0 (one-sample t tests), indicating a directional upward or downward translocation.

Table 2Three-way ANOVA of net vertical relocation of zebra mussels (experiment 2)

Effect df Mean square F P

Alarm cue (A) 1 84.92 14.69 <0.001Light (L) 1 15.47 2.68 0.103Size (S) 2 161.90 28.00 <0.001A�L 1 <0.01 <0.01 0.990A�S 2 4.74 0.82 0.442L�S 2 96.89 16.76 <0.001A�L�S 2 5.94 1.03 0.359Error 222 5.78

Table 3Three-way ANOVA of absolute vertical relocation of zebra mussels (experiment 2)

Effect df Mean square F P

Alarm cue (A) 1 1.66 8.29 0.004Light (L) 1 15.41 76.77 <0.001Size (S) 2 9.90 49.31 <0.001A�L 1 0.46 2.29 0.132A�S 2 0.96 4.76 0.009L�S 2 1.55 7.73 0.001A�L�S 2 0.22 1.11 0.332Error 222 0.20

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Control

Alarm

Light

Dark

Light

Dark

Light

Dark

Light

Dark

Light

Light

Dark

Glass

Sand

Glass

Sand

Glass

Sand

Dark

Largemussels (>20 mm)

Mediummussels (10−20mm)

Smallmussels (<10 mm)

0 40 60 8020

Percentage of aggregated individuals

Figure 4. Aggregation forming by zebra mussels on glass and sand substrata in experiment 3. Error bars indicate SEMs.

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186 183

contradictorywith that of experiment 1, where all mussels changedthe intensity of their movement on a horizontal surface in thepresence of the alarm cue, depending on light conditions. Probablymussels responded to a danger differently when they were able toselect a direction providing them with a higher level of protection.This was possible in experiment 2, in which the downward direc-tion on the slope potentially increased the chances of finding asheltered location, whereas in experiment 1 all directions wereequal andmussels could only respond to the alarm cue by changingthe intensity of their movement.

The effects of the studied factors (light, body size, alarm sub-stance) on mussel aggregation appeared only on the soft sandysubstratum, unsuitable for attachment. Hard substratum wasclearly preferred over conspecific shells. Zebra mussels and otherbivalves are attracted by conspecifics (De Vooys, 2003; Kobak,

Pozna�nska, & Kakareko, 2009; Uryu, Iwasaki, & Hinoue, 1996;Wainman, Hincks, Kaushik, & Mackie, 1996), which may accountfor their clumped distribution. Kavouras and Maki (2003) observedthat zebra mussels avoided attaching to conspecific shells if alter-native, biofilm-covered, hard substratumwas available. Our presentresults suggest that the lack of an alternative hard substratum,rather than mussel preferences, is a major force driving mussels toform aggregations. Mussels may benefit from living outside ag-gregations by avoiding intraspecific competition and deterioratingenvironmental conditions inside a colony (Burks et al., 2002;Tuchman et al., 2004). Solitary individuals have been found togrow faster and have a better physiological condition than clumpedmussels (Chase & Bailey, 1996; Sta�nczykowska, 1964). In contrast,singletons experience greater mortality (Chase & Bailey, 1996) andhave to invest more resources in defences against predation and

Table 4Four-way ANOVA of zebra mussel aggregation (experiment 3)

Effect df Mean square F P

Alarm cue (A) 1 0.96 8.62 0.004Light (L) 1 0.17 1.52 0.220Size (S) 2 0.16 1.45 0.237Substratum (Sb) 1 11.90 106.86 <0.001A�L 1 0.02 0.17 0.679A�S 2 0.39 3.46 0.033A�Sb 1 0.50 4.53 0.034L�S 2 0.19 1.73 0.180L�Sb 1 0.58 5.18 0.024S�Sb 2 0.02 0.13 0.877A�L�S 2 0.02 0.15 0.858A�L�Sb 1 0.01 0.04 0.839A�S�Sb 2 0.23 2.06 0.129L�S�Sb 2 0.52 4.71 0.010A�L�S�Sb 2 0.07 0.60 0.551Error 216 0.11

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186184

hydrodynamic forces (Schneider, Wethey, Helmuth, & Hilbish,2005; Zardi et al., 2006). Therefore, mussels tend to aggregatewith their nearest neighbours, but move away from too crowdedlocations, which shapes the complex, heterogeneous pattern of amussel bed (De Jager, Weissing, Herman, Nolet, & van de Koppel,2011). This is also why mussels increase their aggregation level inresponse to predator scents (Kobak & Kakareko, 2009; Naddafi &Rudstam, 2013).

The alarm substance in our study stimulated medium and largemussels to stay in contact with conspecifics. However, the alarmsubstance did not affect aggregation of small mussels. This size-dependent response may have several causes. First, a group ofsmall mussels may be an insufficient obstacle for a predator, mak-ing aggregations with small conspecifics useless. Small zebramussels do increase their aggregation in the presence of predators(Kobak & Kakareko, 2009; Naddafi & Rudstam, 2013). However,another explanation of the present results should be considered.The control aggregation level of small mussels was relatively highcompared with that of larger individuals, particularly in light(Fig. 4), so perhaps there was no need to increase it further in thepresence of alarm cues. Larger mussels, however, weakly clumpedin the control treatments, were attracted to conspecifics in thepresence of danger.

The antipredator effect of clumping consists of the decreasedprobability of predator attacks on particular aggregation members,particularly those located in the centre (Cheung, Tong, Yip, & Shin,2004), as well as the difficulty of picking a single prey item from agroup (Heller & Milinski, 1979). In byssate bivalves, yet anothermechanism is added to this list: a difficulty in detaching a singlespecimen from a cluster of individuals connected with one anotherand with the substratum. This makes aggregation a particularlyefficient defence for this group of animals (Kobak & Kakareko,2011).

Light reduced distances travelled by mussels (experiments 1and 2), except for small individuals exposed to alarm cues. Thegreater level of aggregation of control mussels in darkness inexperiment 3 can be explained by their more intense relocationsand, in consequence, higher probability of encounters than in light.Greater mobility in darkness generally corroborates the results ofother studies (Kobak & Nowacki, 2007; Toomey et al., 2002),although an opposite behaviour was observed in the blue mussel,Mytilus edulis, which clumps more often in light (Gownaris &Commito, 2008).

In all the experiments, light, apart from being a stand-alonefactor affecting mussel behaviour, clearly modified mussel

responses to alarm cues bymaking them stronger than in darkness.Light may be regarded as an indirect signal of danger (Koperski,1997), informing an animal that it is at an exposed, unshelteredsite, potentially vulnerable to detection and predator attack.Therefore, light reduces activity, as shown in our experiments (inthe absence of alarm cues) and in other studies (Kobak & Nowacki,2007). However, we have shown that when the alarm substance isadditionally present, light may induce a stronger antipredatorresponse, sometimes associated with more intensemovements in aparticular direction.

We have demonstrated that the responses of zebra mussels toconspecific alarm cues depend on several variables, such as bodysize, light, substratum type and its inclination. Mussel responsesare not always limited to activity reduction: in certain conditionsthey can even increase their movement in search of a betterattachment site. Thus, mussels are able to adjust their defensivemechanisms to complex situations, which they can experience inthe field. Antipredator defences of the zebra mussel may contributeto its invasion success and have been postulated as a potentialadvantage of this species over its congener, the quagga mussel,Dreissena rostriformis bugensis, generally regarded as a more effi-cient competitor, but exhibiting lower attachment (Peyer,McCarthy, & Lee, 2009) and weaker antipredator responses(Naddafi & Rudstam, 2013). This fine tuning of antipredator de-fences has been observed in other organisms with regard topredator species with different feeding modes (Freeman, 2007;Weber, 2003; Wudkevich et al., 1997), predator density(Czarnołeski et al., 2010) and prey abundance (Gliwicz,Dawidowicz, & Maszczyk, 2006). The commonness and quickevolution of antipredator defensive mechanisms (Naddafi &Rudstam, 2013) indicate a great importance of predatoreprey in-teractions for shaping animal behaviour and the functioning of preyorganisms in the wild.

Acknowledgments

We are grateful to Łukasz Jermacz for collecting mussels for ourstudy and to Mrs Hazel Pearson for language corrections. We alsothank two anonymous referees for their valuable comments thathelped greatly improve our manuscript.

References

Baumg€artner, D., Koch, U., & Rothhaupt, K. O. (2003). Alteration of kairomone-induced antipredator response of the freshwater amphipod Gammarus roeseliby sediment type. Journal of Chemical Ecology, 29, 1391e1401.

Brown, J. S., Laundr�e, J. W., & Gurung, M. (1999). The ecology of fear: optimalforaging, game theory, and trophic interactions. Journal of Mammalogy, 80,385e399.

Burks, R. L., Tuchman, N. C., Call, C. A., & Marsden, J. E. (2002). Colonial aggregates:effects of spatial position on zebra mussel responses to vertical gradients ininterstitial water quality. Journal of the North American Benthological Society, 21,64e75.

Chase, M. E., & Bailey, R. C. (1996). Recruitment of Dreissena polymorpha: does thepresence and density of conspecifics determine the recruitment density andpattern in a population? Malacologia, 38, 19e31.

Cheung, S. G., Tong, P. Y., Yip, K. M., & Shin, P. K. S. (2004). Chemical cues frompredators and damaged conspecifics affect byssus production in the green-lipped mussel Perna viridis. Marine and Freshwater Behaviour and Physiology,37, 127e135.

Clinchy, M., Sheriff, M. J., & Zanette, L. Y. (2013). Predator-induced stress and theecology of fear. Functional Ecology, 27, 56e65.

Cot�e, I. M., & Jelnikar, E. (1999). Predator-induced clumping behaviour in mussels(Mytilus edulis Linnaeus). Journal of Experimental Marine Biology and Ecology,235, 201e211.

Czarnołeski, M., Kozłowski, J., Kubajak, P., Lewandowski, K., Müller, T.,Sta�nczykowska, A., et al. (2006). Cross-habitat differences in crush resistanceand growth pattern of zebra mussels (Dreissena polymorpha): effects of cal-cium availability and predator pressure. Archiv für Hydrobiology, 165,191e208.

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186 185

Czarnołeski, M., Müller, T., Adamus, K., Ogorzelska, G., & Sog, M. (2010).Injured conspecifics alter mobility and byssus production in zebramussels Dreissena polymorpha. Fundamental and Applied Limnology, 176,269e278.

Czarnołeski, M., Müller, T., Kierat, J., Gryczkowski, L., & Chybowski, Ł. (2011). Anchordown or hunker down: an experimental study on zebra mussels' response topredation risk from crayfish. Animal Behaviour, 82, 543e548.

De Jager, M., Weissing, F. J., Herman, P. M. J., Nolet, B. A., & van de Koppel, J. (2011).L�evy walks evolve through interaction between movement and environmentalcomplexity. Science, 332, 1551e1553.

De Meester, L., Dawidowicz, P., Van Gool, E., & Loose, C. J. (1999). Ecology andevolution of predator-induced behavior of zooplankton: depth selectionbehavior and diel vertical migration. In R. Tollrian, & C. D. Harvel (Eds.), Theecology and evolution of inducible defenses (pp. 160e176). Princeton, NJ:Princeton University Press.

De Vooys, C. G. N. (2003). Effect of a tripeptide on the aggregation behaviour of theblue mussel Mytilus edulis. Marine Biology, 142, 1119e1123.

Dzialowski, A. R., Lennon, J. T., O'Brien, W. J., & Smith, V. H. (2003). Predator-inducedphenotypic plasticity in the exotic cladoceran Daphnia lumholtzi. FreshwaterBiology, 48, 1593e1602.

Freeman, A. S. (2007). Specificity of induced defenses in Mytilus edulis andasymmetrical predator deterrence. Marine Ecology Progress Series, 334,145e153.

Gliwicz, Z. M. (2005). Food web interactions: why are they reluctant to bemanipulated? Plenary Lecture. Verhandlungen Internationale Vereinigung fürtheoretische und angewandte Limnologie, 29, 73e88.

Gliwicz, Z. M., Dawidowicz, P., & Maszczyk, P. (2006). Low-density anti-predation refuge in Daphnia and Chaoborus? Archiv für Hydrobiologie, 167,101e114.

Gownaris, N., & Commito, J. A. (2008, April). Are mussel beds selfish herds? Musselsself-organize into fractal aggregations. In Benthic ecology meeting. Providence,RI, U.S.A.: Brown University.

Heller, R., & Milinski, M. (1979). Optimal foraging of sticklebacks on swarming prey.Animal Behaviour, 27, 1127e1141.

Houghton, J. S., & Janssen, J. (2013). Variation in predator-prey interactions betweenroud gobies and dreissenids mussels. In T. F. Nalepa, & D. W. Schloesser (Eds.),Quagga and zebra mussels biology, impacts, and control (2nd ed.). (pp. 359e367).Boca Raton, FL: CRC Press, Taylor & Francis.

Ishida, S., & Iwasaki, K. (2003). Reduced byssal thread production and movement bythe intertidal mussel Hormomya mutabilis in response to effluent from preda-tors. Journal of Ethology, 21, 117e122.

Karatayev, A. Y., Burlakova, L. E., & Padilla, D. K. (2002). Impacts of zebra musselson aquatic communities and their role as ecosystem engineers. InE. Leppakoski, S. Gollasch, & S. Olenin (Eds.), Invasive aquatic species ofEurope: Distribution, impacts and management (pp. 433e446). Boston, MA:Kluwer Academic.

Kavouras, J. H., & Maki, J. S. (2003). The effects of natural biofilms on the reat-tachment of young adult zebra mussels to artificial substrata. Biofouling, 19,247e256.

Kilgour, B. W., & Baker, M. A. (1994). Effects of season, stock, and laboratory pro-tocols on survival of zebra mussels (Dreissena polymorpha) in bioassays. Ar-chives of Environmental Contamination and Toxicology, 27, 29e35.

Kobak, J. (2006). Geotactic behaviour of Dreissena polymorpha (Bivalvia). Mala-cologia, 48, 305e308.

Kobak, J., & Kakareko, T. (2009). Attachment strength, aggregation and movementof the zebra mussel (Dreissena polymorpha, Bivalvia) in the presence of potentialpredators. Fundamental and Applied Limnology, 174, 193e204.

Kobak, J., & Kakareko, T. (2011). The effectiveness of the induced anti-predatorbehaviour of zebra mussel Dreissena polymorpha in the presence of mollus-civorous roach Rutilus rutilus. Aquatic Ecology, 45, 357e366.

Kobak, J., & Nowacki, P. (2007). Light-related behaviour of zebra mussel (Dreissenapolymorpha, Bivalvia). Fundamental and Applied Limnology, 169, 341e352.

Kobak, J., Pozna�nska, M., & Kakareko, T. (2009). Effect of attachment status andaggregation on behaviour of the zebra mussel, Dreissena polymorpha, Bivalvia.Journal of Molluscan Studies, 75, 109e117.

Kobak, J., Kakareko, T., & Pozna�nska, M. (2010). Changes in attachment strength andaggregation of zebra mussel, Dreissena polymorpha in the presence of potentialfish predators of various species and size. Hydrobiologia, 644, 195e206.

Koperski, P. (1997). Changes in feeding behaviour of the larvae of the damselflyEnallagma cyathigerum in response to stimuli from predators. Ecological Ento-mology, 22, 167e175.

Krist, K. C. (2002). Crayfish induce a defensive shell shape in a freshwater snail.Invertebrate Biology, 121, 235e242.

Lass, S., & Spaak, P. (2003). Chemically induced anti-predator defences in plankton:a review. Hydrobiologia, 491, 221e239.

McKillup, S. C., & McKillup, R. V. (1993). Behavior of the intertidal gastropod Pla-naxis sulcatus (Cerithiacea: Planaxidae) in Fiji: are responses to damaged con-specifics and predators more pronounced on tropical versus temperate shores?Pacific Science, 47, 401e407.

Molloy, D. P., Karatayev, A. Y., Burlakova, L. E., Kurandina, D. P., & Laruelle, F. (1997).Natural enemies of zebra mussels: predators, parasites, and ecological com-petitors. Reviews in Fisheries Sience, 5, 27e97.

Naddafi, R., Ekl€ov, P., & Pettersson, K. (2007). Non-lethal predator effects on thefeeding rate and prey selection of the exotic zebra mussel Dreissena polymorpha.Oikos, 116, 1289e1298.

Naddafi, R., & Rudstam, L. G. (2013). Predator induced behavioural defences in twocompetitive invasive species: the zebra mussel and the quagga mussel. AnimalBehaviour, 86, 1275e1284.

Naddafi, R., & Rudstam, L. G. (2014). Predator-induced morphological defences intwo invasive dreissenid mussels: implications for species replacement. Fresh-water Biology, 59, 703e713.

Peyer, S. M., McCarthy, A. J., & Lee, C. E. (2009). Zebra mussels anchor byssal threadsfaster and tighter than quagga mussels in flow. Journal of Experimental Biology,212, 2027e2036.

Porter, A. E., & Marsden, J. E. (2008). Adult zebra mussels (Dreissena poly-morpha) avoid attachment to mesh materials. Northeastern Naturalist, 15,589e594.

Prejs, A., Lewandowski, K., & Sta�nczykowska, A. (1990). Size selective predation byroach (Rutilus rutilus) on zebra mussel (Dreissena polymorpha): field studies.Oecologia, 83, 378e384.

Reimer, O., & Harms-Ringdahl, S. (2001). Predator-inducible changes in blue mus-sels from the predator-free Baltic Sea. Marine Biology, 139, 959e965.

Reimer, O., & Tedengren, M. (1997). Predator-induced changes in byssal attachment,aggregation and migration in the blue mussel, Mytilus edulis. Marine andFreshwater Behaviour and Physiology, 30, 251e266.

Schneider, K. R., Wethey, D. S., Helmuth, B. S. T., & Hilbish, T. J. (2005). Implicationsof movement behavior on mussel dislodgement: exogenous selection in aMytilus spp. hybrid zone. Marine Biology, 146, 333e343.

Sta�nczykowska, A. (1964). On the relationship between abundance, aggregationsand “condition” of Dreissena polymorpha Pall. in 36 Mazurian Lakes. EkologiaPolska A, 34, 653e690.

�Slusarczyk, M., Dawidowicz, P., & Rygielska, E. (2005). Hide, rest or die: a light-mediated diapause response in Daphnia magna to the threat of fish predation.Freshwater Biology, 50, 141e146.

Toomey, M. B., McCabe, D., & Marsden, J. E. (2002). Factors affecting the movementof adult zebra mussels (Dreissena polymorpha). Journal of the North AmericanBenthological Society, 21, 468e475.

Tuchman, N. C., Burks, R. L., Call, C. A., & Smarrelli, J. (2004). Flow rate and verticalposition influence ingestion rates of colonial zebra mussels (Dreissena poly-morpha). Freshwater Biology, 49, 191e198.

Uryu, Y., Iwasaki, K., & Hinoue, M. (1996). Laboratory experiments on behavior andmovement of a freshwater mussel, Limnoperna fortunei (Dunker). Journal ofMolluscan Studies, 62, 327e341.

Wainman, B. C., Hincks, S. S., Kaushik, N. K., & Mackie, G. L. (1996). Biofilm andsubstrate preference in the dreissenid larvae of Lake Erie. Canadian Journal ofFisheries and Aquatic Sciences, 53, 134e140.

Weber, A. (2003). More than one ‘fish kairomone’? Perch and stickleback kair-omones affect Daphnia life history traits differently. Hydrobiologia, 498,143e150.

Weber, A., & Van Noordwijk, A. (2002). Swimming behaviour of Daphnia clones:differentiation through predator infochemicals. Journal of Plankton Research, 24,1335e1348.

Wudkevich, K., Wisenden, B. D., Chivers, D. P., & Smith, R. J. F. (1997). Reactions ofGammarus lacustris to chemical stimuli from natural predators and injuredconspecifics. Journal of Chemical Ecology, 23, 1163e1173.

Zardi, G. I., Nicastro, K. R., McQuaid, C. D., Rius, M., & Porri, F. (2006). Hydrodynamicstress and habitat partitioning between indigenous (Perna perna) and invasive(Mytilus galloprovincialis) mussels: constraints of an evolutionary strategy.Marine Biology, 150, 79e88.

Table A4

J. Kobak, A. Ry�nska / Animal Behaviour 96 (2014) 177e186186

APPENDIX

Sequential Bonferroni-corrected Fisher LSD tests used as a post hoc procedure forthe ANOVA on mussel aggregation (experiment 3, Table 4)

Compared groups Other variable levels P

Alarm cue£Mussel size interactionAlarm vs Control Small mussels 0.705

Medium mussels 0.001*Large mussels 0.005*

Small vs Medium mussels Alarm 0.161Control 0.764

Small vs Large mussels Alarm 0.896Control 0.013

Medium vs Large mussels Alarm 0.203Control 0.030

Alarm cue£Substratum interactionAlarm vs Control Sand <0.001*

Glass 0.569Sand vs Glass Alarm <0.001*

Control <0.001*

Light£Mussel size£Substratum interactionLight vs Darkness Small mussels on sand 0.162

Small mussels on glass 0.780Medium mussels on sand 0.002*Medium mussels on glass 0.488Large mussels on sand 0.002*Large mussels on glass 0.388

Small vs Medium mussels Light on sand 0.021Light on glass 0.961Darkness on sand 0.038Darkness on glass 0.356

Small vs Large mussels Light on sand 0.003Light on glass 0.865Darkness on sand 0.263Darkness on glass 0.190

Medium vs Large mussels Light on sand 0.511Light on glass 0.826Darkness on sand 0.336Darkness on glass 0.697

Sand vs Glass Small mussels in darkness 0.001*Small mussels in light <0.001*Medium mussels in darkness <0.001*Medium mussels in light 0.009Large mussels in darkness <0.001*Large mussels in light 0.028

Asterisks indicate significant results (P < 0.05 after the Bonferroni correction).

Table A1Sequential Bonferroni-corrected Fisher LSD tests used as a post hoc procedure forthe ANOVA on the horizontal movement of mussels (experiment 1, Table 1)

Compared groups Other variable levels P

Alarm cue£Light interactionAlarm vs Control Light 0.015*

Darkness 0.020*Light vs Darkness Alarm 0.313

Control <0.001*

Light£Mussel size interactionLight vs Darkness Small mussels 0.420

Medium mussels 0.001*Large mussels 0.001*

Small vs Medium mussels Light <0.001*Darkness <0.001*

Small vs Large mussels Light <0.001*Darkness <0.001*

Medium vs Large mussels Light 0.013*Darkness 0.008*

Asterisks indicate significant results (P < 0.05 after the Bonferroni correction).

Table A2Sequential Bonferroni-corrected Fisher LSD tests used as a post hoc procedure forthe ANOVA on the net vertical relocation of mussels (experiment 2, Table 2)

Compared groups Other variable levels P

Light£Mussel size interactionLight vs Darkness Small mussels 0.0001*

Medium mussels 0.039Large mussels <0.001*

Small vs Medium mussels Light 0.045Darkness <0.001*

Small vs Large mussels Light 0.218Darkness <0.001*

Medium vs Large mussels Light 0.434Darkness <0.001*

Asterisks indicate significant results (P < 0.05 after the Bonferroni correction).

Table A3Sequential Bonferroni-corrected Fisher LSD tests used as a post hoc procedure forthe ANOVA on absolute vertical relocation of mussels (experiment 2, Table 3)

Compared groups Other variable levels P

Alarm cue£Mussel size interactionAlarm vs Control Small mussels 0.614

Medium mussels 0.220Large mussels <0.001*

Small vs Medium mussels Alarm 0.001*Control <0.001*

Small vs Large mussels Alarm <0.001*Control <0.001*

Medium vs Large mussels Alarm 0.103Control <0.001*

Light£Mussel size interactionLight vs Darkness Small mussels 0.069

Medium mussels <0.001*Large mussels <0.001*

Small vs Medium mussels Light <0.001*Darkness 0.023

Small vs Large mussels Light <0.001*Darkness <0.001*

Medium vs Large mussels Light 0.072Darkness 0.004*

Asterisks indicate significant results (P < 0.05 after the Bonferroni correction).