Learning to find food: evidence for embryonic sensitization and juvenile social learning in a salamander

Animal Behaviour 142 (2018) 199e206

Contents lists available at ScienceDirect

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

Learning to find food: evidence for embryonic sensitization and juvenile social learning in a salamander Adam L. Crane a, *, Emilee J. Helton b, Maud C. O. Ferrari c, Alicia Mathis d a

Department of Biology, University of Saskatchewan, SK, Canada National Park Service, Moose, WY, U.S.A. c Veterinary Biomedical Sciences, WCVM, University of Saskatchewan, Saskatoon, SK, Canada d Biology Department, Missouri State University, Springfield, MO, U.S.A. b

a r t i c l e i n f o Article history: Received 9 November 2017 Initial acceptance 26 February 2018 Final acceptance 25 May 2018 MS. number: A17-00888R2 Keywords: generalization imprinting local enhancement olfaction social facilitation

For many species, learning is an essential mechanism for dealing with the environment correctly and efficiently. Animals that quickly learn important information, and learn at a young age, can gain a competitive advantage in exploiting resources. Moreover, animals that learn indirectly through social observations can avoid the fitness costs of directly learning about potential dangers. Here we tested such learning capabilities in ringed salamanders, Ambystoma annulatum, a species where adults are primarily solitary and do not provide parental care. Adults lay eggs in ponds where embryos have the opportunity to learn from chemical cues in their environment before hatching, whereupon the high density of larvae provides an opportunity to learn from social information. In this study, we found that these salamanders can learn an attraction to novel food stimuli as embryos and that naïve observer larvae can learn from conspecifics that show attraction to stimuli. Embryonic exposure to a novel food stimulus (shrimp odour) caused attraction to that stimulus posthatching, and this response appeared to be generalized to another potential prey stimulus (mussel odour) but not to a novel plant stimulus. In a test of social learning, only observers that were paired with models corralled near a novel food stimulus were subsequently attracted to the stimulus. This study is the first to report embryonic learning of food or social learning by salamanders, providing more evidence for generalized learning by embryos and social learning by species lacking more complex social behaviours. © 2018 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Correctly responding to environmental stimuli is essential to maximizing fitness, but fluctuations in the environment can pose challenges to decision making (Dall, 2010; Kacelnik & Bateson, 1996; Lima & Dill, 1990). For instance, changes in food availability or predation pressure may lead to a failure to accurately assess such information. Learning is generally viewed as the act of acquiring new information or the modification or reinforcement of existing information based on experience, resulting in behavioural changes (Brown & Chivers, 2005; Papaj & Prokopy, 1989; Stephens, 1991). According to learning theory, stable environments facilitate the evolution of innate responses to stimuli, whereas variable and complex environments can promote learned responses (Stephens, 1991, 1993). In a changing environment, individuals that learn can better exploit resources and have an increased probability of survival and reproductive success (Brown & Chivers, 2005). Such

* Correspondence: A. L. Crane, Department of Biology, University of Saskatchewan, 112 Science Place, Saskatoon, SK S7N 5E2, Canada. E-mail address: [email protected] (A. L. Crane).

learning capabilities may be particularly important during early life periods where mortality rates are generally high (Pianka, 1970; Sogard, 1997), and indeed, young animals often show a high capacity for learning (Bornstein, 1989; Fawcett & Frankenhuis, 2015). In classic research by Lorenz (1935), geese, Anser anser, during a sensitive period in early development would instinctively bond with other moving stimuli in place of their parents. This phenomenon, known as imprinting, allows goslings to quickly learn to recognize their parents. Imprinting (or a learning process akin to imprinting) is not restricted to social stimuli however (Immelmann, 1975). For instance, several studies have explored imprinting of food stimuli during sensitive learning periods (e.g. Burghardt & , Poirel, Houde , & Dickel, 2012; Punzo, 2002). Hess, 1966; Guibe Even as embryos, animals are capable of learning how to maximize their probability of survival in their postnatal environment via a variety of sensory modalities (e.g. Darmaillacq, Lesimple, & Dickel, 2008; Hepper & Waldman, 1992; Lickliter & Hellewell, 1992). Chemosensory cues, for instance, are relatively long lasting, can move around barriers, and are available when visibility is

https://doi.org/10.1016/j.anbehav.2018.06.021 0003-3472/© 2018 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

200

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

low (Alcock, 2005; Mathis & Crane, 2017). Animals can use chemical cues to become familiar with the activity patterns of other species, such as those that are predators or prey. Although relatively few species have been studied, embryonic learning of chemical information occurs across a wide range of animal taxa (e.g. in dogs: Wells & Hepper, 2006; chickens: Sneddon, Hadden, & Hepper, 1998; crocodiles: Sneddon, Hepper, & Manolis, 2000; fish:  et al., 2012; Nelson, Alemadi, & Wisenden, 2013; cuttlefish: Guibe mites: Quesada & Schausberger, 2012). For species with aquatic eggs, such chemical information about the environment is widely available. In this context, perhaps amphibians are the most studied group, usually in the context of learning about predation risk (e.g. Ferrari & Chivers, 2009b; Ferrari, Manek, & Chivers, 2010; Garcia, Urbina, Bredeweg, & Ferrari, 2017; Mathis, Ferrari, Windel, Messier, & Chivers, 2008). Hepper and Waldman (1992) were the first to document embryonic learning in amphibians, where exposure to a novel odour (pure orange essence) caused embryonic frogs, Rana temporaria and Lithobates sylvaticus, to prefer that odour after hatching. Such learning could better prepare individuals for locating available food types in their home pond. Learning about novel foods has clear benefits, but sampling different foods can be time consuming and potentially dangerous (e.g. consuming something toxic). However, learning by observing experienced companions (i.e. social learning), allows animals to minimize such costs (Galef, 1993; Galef & Laland, 2005). One social learning mechanism is ‘stimulus enhancement’ where an animal learns to approach certain stimuli after observing the attraction of another individual to the stimuli (Heyes, 1994), but for many species, the opportunities for such learning are limited due to their solitary life history. Most studies on social learning have involved species that provide parental care (e.g. mammals: Whiten, 2000; birds: Lefebvre & Bouchard, 2003) or are highly gregarious throughout their lives (e.g. fishes: reviewed in Brown & Laland, 2003). However, there is a growing body of literature revealing that species that are primarily solitary and are not socially complex can be influenced by the behaviour of conspecifics (Coolen, Dangles, & Casas, 2005; Crane, Mathis, & McGrane, 2012; Wilkinson, Mandl, Bugnyar, & Huber, 2010). A behavioural response in the presence of other individuals that are performing that same behaviour is referred to as ‘social facilitation’ (Clayton, 1978), but to demonstrate that social learning has occurred, observer individuals must display the behavioural change in the absence of others. A few studies have documented social learning in species lacking more complex social behaviours, usually in the context of learning the locations of food (e.g. Brown, Markula, & Laland, 2003; Guttridge et al., 2013; Kis, Huber, & Wilkinson, 2015; Noble, Byrne, & Whiting, 2014; Wilkinson, Kuenstner, Mueller, & Huber, 2010). Here we tested whether ringed salamanders, Ambystoma annulatum, can learn about prey stimuli using two different mechanisms (imprinting and social learning) during two different life stages (embryonic and larval). Ringed salamander embryos are surrounded by a vitelline membrane and outer jelly capsules (Petranka, 1998) that allow environmental cues to diffuse into the immediate vicinity of the embryos, providing an opportunity to become familiar with their future environment prior to hatching (Mathis et al., 2008). These salamanders do not appear to be socially complex; adults spend time alone underground, except when gathering at ponds to breed in the fall (Spotila & Beumer, 1970). Larvae are initially found in close proximity upon hatching (>500 individuals per 1 m2 in one study) (Peterson, Wilkinson, Moll, & Holder, 1991). As is typical for salamander larvae, ringed salamander larvae do not show schooling or shoaling behaviour (A. L. Crane, personal observations), and actively avoid contact with conspecifics, unlike anuran larvae (Wells, 2010). Over the winter, densities

of larval ringed salamanders drop to ~70/m2, and many become cannibalistic (Jefferson et al., 2014). Densities drop to zero as the larvae metamorphosize and leave the pond in May (Peterson et al., 1991). During this period, larval ringed salamanders should have ample opportunities to acquire information from nearby conspecifics (i.e. social information) while foraging on small aquatic invertebrates and being exposed to risk from a diversity of predators (Crane et al., 2012; Mathis, Murray, & Hickman, 2003). We expect that such opportunities can be used to learn about food locations. Here, we exposed salamander embryos to a novel prey stimulus during a conditioning period, predicting increased preference and foraging behaviour towards the stimulus posthatching. In a separate experiment, we tested the prediction that larval ringed salamanders would learn to approach a novel prey stimulus via the social learning mechanism of stimulus enhancement. In this experiment, we expected larvae to first show attraction to a conspecific individual (a ‘model’) that was corralled near a novel stimulus, and then subsequently show attraction to that stimulus in the absence of the model. METHODS Ethical Statement This research was approved by the Institutional Animal Care and Use Committee at Missouri State University (protocol no. 10030). The Missouri Department of Conservation granted permission (permit no. 15193) to collect the salamanders used in this study, and also for their release at their collection site after the completion of this study. The salamanders were collected as eggs (details below) and were transported to the laboratory (a 1 h drive) in buckets with pond water and battery-powered aeration. A total of 316 individuals (sex undetermined) were used in these experiments when larvae were <5 months of age. Throughout the experiments, larvae were maintained in groups (1e30 individuals, depending on their body size to avoid cannibalism) in plastic holding containers (34.5
20
12.5 cm) filled with filtered water (2 litres) and with an airstone attached to an air pump for aeration. A few rocks and artificial plants were also added for enrichment. Larvae were fed daily with Daphnia, and over development were transitioned to a diet of aquatic worms (Lumbriculus variegatus). The experiments involved no potentially harmful, painful or distressful manipulations. Collection, Housing and Maintenance We collected what appeared to be ~40 clutches of eggs (2e31 per clutch at stage 28e31) (Harrison, 1969) in October of two consecutive years (2011e2012) from a pond at Bull Shoals Field Station in southwestern Missouri, U.S.A. Each year, eggs were housed in 24 plastic containers (10 cm3) with approximately 10 eggs per container, sometimes together for large clutches while being combined for smaller clutches. Each container was filled with an equal mixture of pond-water and dechlorinated municipal filtered and dechlorinated water (hereafter, water) and was kept inside an environmental chamber at 14  C and on a 12:12 h light:dark cycle, with weekly water changes. After hatching, larvae from different clutches were mixed and moved into larger plastic holding containers (34.5
20
12.5 cm) with aeration. Larvae were separated by size to prevent cannibalism. Because larvae were mixed across clutches and separated by size, we did not account for clutch variation in our experiments. Each experiment involved 12 or more clutches, with individuals being randomly assigned to treatments (i.e. approximately equal numbers from each clutch in each treatment).

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

201

Embryonic Learning: Conditioning Exposure

Embryonic Learning: Testing Arenas

Over both 2011 and 2012, half of the containers holding eggs were exposed to a novel food stimulus (N ¼ 135) while the other half received a water control (N ¼ 137) (Fig. 1a and b). For the novel food stimulus, we used commercial frozen Artemia (brine shrimp, hereafter, shrimp). These ‘shrimp’ do not co-occur with ringed salamanders (or exist near their range), and thus the stimulus was novel in an individual sense. From an evolutionarily perspective of novelty, the closest relatives of Artemia sp. with which larval ringed salamanders have coevolved are Daphnia sp., and hence there was potential for innate attraction to the stimulus. Each day of exposure, we freshly prepared the stimulus by grinding 50 g of shrimp in 500 ml of water and straining the product to remove solid particles from the liquid solution. Then, 3 ml of either the shrimp stimulus or water was added to each container. We allowed 48 h for the stimuli to diffuse into the eggs and for learning to occur, and then we conducted a 100% water change. In total, three exposures occurred over the course of 1 week before hatching.

After the larvae hatched and were actively swimming, we assessed their attraction to either the stimulus (shrimp) that was experienced during their embryonic development or to a novel stimulus. At the time of testing, the larvae were 2e3 weeks old and at posthatching stage 1e2 (Watson & Russell, 2000). First, larvae were moved individually via a pipette into an opaque acclimation cylinder in the centre of the testing arena e a PVC trough (13
6 cm area) with caps on each end and filled with water (150 ml) and a sand substrate (50 ml) (Fig. 2a). A cotton ball (~2.5 cm diameter) was dipped in the stimulus (either the conditioned shrimp stimulus or a novel stimulus) and placed at a randomly selected end of the arena, while another cotton ball was dipped in control water and placed at the opposite end. A test with dye indicated that the chemical cue did not disperse into the opposite half of the arena within 10 min. Lines on the walls of the arena demarked five zones, with the stimulus cotton ball always designated as zone 5 and water as zone 1 (Fig. 2a).

(a) Embryo experiment 1: shrimp vs mussel Background embryonic (conditioning) phase

Larval learning (testing) phase

Embryonic Learning: Testing Trials

Shrimp

Water

Shrimp

Mussel

Shrimp

Mussel

vs

vs

vs

vs

Water

Water

Water

Water

(b) Embryo experiment 2: shrimp vs melon Background embryonic (conditioning) phase

Larval learning (testing) phase

Shrimp

Water

Shrimp

Melon

Shrimp

Melon

vs

vs

vs

vs

Water

Water

Water

Water

We conducted two experiments to test whether embryonic learning occurred. In the first experiment, in 2011, we used mussel (commercial frozen Mytilidae sp.) as the novel control stimulus to test against shrimp (Fig. 1a). Like the shrimp stimulus, the mussel stimulus provided an allopatric animal tissue odour, but from a species distantly related to shrimp. Larvae from both conditioning treatments were randomly assigned to the test cue treatments (N ¼ 20e29 per group) and given 30 s to acclimate before the central acclimation cylinder was manually removed and a 5 min trial began. We recorded the zone occupied by the larva every 20 s, where the total represented a score for overall attraction. We also recorded the number of times that the larva made contacts with the stimulus cotton ball (hereafter, contacts). The second experiment, in 2012, followed this same design, but the novel stimulus was even more distinct from the shrimp stimulus. Here, we used an allopatric fruit (honeydew, Cucumis melo) as the novel stimulus (Fig. 1b), and hence the experiments allowed us to assess the possibility that a learned response would be generalized to novel animal or plant stimuli. After removing the mussel shell and the melon rind, preparation of each novel stimulus occurred as described above for the shrimp stimulus (e.g. same g/ml concentration). Sample sizes were 42e47 per treatment group. Social Learning: Conditioning with Models

(c) Social experiment Background larval (conditioning) phase

Larval learning (testing) phase

Shrimp model Observer

vs Water

Shrimp

Water model vs

Observer

Shrimp

Shrimp

Figure 1. Phases and treatments of experiments testing embryonic (a, b) and social (c) learning. In embryonic learning experiments, salamander embryos were exposed to either shrimp stimuli or a water control. Then, after hatching, larvae were tested for attraction to the shrimp stimulus compared to novel control stimuli, which were mussel (a) or melon (b). In the social experiment (c), observer larvae were paired with social models that were corralled near either a shrimp stimulus or a water control. Subsequently, observers were tested alone to assess their attraction to the shrimp stimulus.

The social learning experiment involved a conditioning period where observer larvae were individually paired with models that were located near either the shrimp stimulus (prepared as above) or a water control (N ¼ 21 per treatment) (Fig. 1c). The observers and models were about 4 months old at posthatching stage 8e9 (Watson & Russell, 2000) and were not used previously in the embryonic learning experiment. Observers were conditioned only once with one of five demonstrator individuals. The conditioning arena (34.5
20
12.5 cm) contained 1200 ml of water and 250 ml of sand substrate. Inside the arena, observers and models were separated by a central dividing wall that was transparent and perforated, and hence allowed passage of visual and chemical information (Fig. 2b). In each section, a cotton ball was placed at each end (4 total), one soaked in the shrimp stimulus and the other in water. We intended for models to provide information about attraction to the stimulus, so we added a small barrier (transparent and perforated) at one end of the model's section of the arena, perpendicular to the central divider (Fig. 2b). This ‘corral’ ensured

202

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

(a) Testing arena for embryonic learning Zone 5

Stim

(b) Arena for social conditioning

Zone 6

Stim

(c) Arena for tests of social learning

Stim Model inside corral

Zone 5

Zone 4

Zone 4

Observer

Zone 6

Stim

Zone 5

Zone 4

Observer

Zone 3

Zone 2

Zone 1

Con

Zone 3

Zone 3

Zone 2

Zone 2

Zone 1

Con

Con

Zone 1

Figure 2. Diagrams of the embryonic learning testing arena (a), the social learning conditioning arena (b) and the social learning testing arena (c). Filled grey circles represent cotton balls soaked in either the experimental stimulus (stim: shrimp, mussel or melon) or control water (con). In social learning trials, the stimulus was always shrimp (b and c). Black circles represent the holding cylinders that were removed after the acclimation period. Long dashes (grey) represent the different zones, increasing in number with proximity to the stimulus. For social conditioning trials (b), short dashes (black) represent the transparent dividing wall with holes allowing chemical diffusion, through which the observer had the opportunity to acquire information from a model corralled near the stimulus.

that the model remained near the stimulus. As a control, half of the conditioning trials involved models corralled near the water end of the arena, with each model being randomly assigned to either treatment (shrimp stimulus or water control). Lines on the arena's walls demarked six zones, with the stimulus end designated as zone 6, and water as zone 1. After adding the model, each observer was placed inside the central acclimation cylinder for 30 s before the cylinder was lifted. We again recorded attraction score and contacts. Social Learning: Testing Observer larvae from both social conditioning treatments (N ¼ 21 per treatment) were tested alone 24 h later to determine whether social learning had occurred. If so, larvae that had previously been conditioned with models near the shrimp stimulus would show more attraction to the shrimp stimulus compared to larvae that had been conditioned with models near the control stimulus. One cotton ball containing the shrimp stimulus was randomly assigned to one end of the arena and designated as zone 6 (Fig. 2c). We again recorded attraction to the stimulus and contacts, as in the embryonic learning experiments but with a longer duration (8 min). Statistical Analyses We analysed response variables (attraction score and contacts) separately because they had different patterns (see Results). For the embryonic learning experiments, data failed the normality assumption for parametric testing, so before proceeding with ANOVA, we transformed the data using the aligned rank transformation, where data are aligned and ranked independently for the assessment of each term in the model (Higgins & Tashtoush, 1994; Wobbrock, Findlater, Gergle, & Higgins, 2011). This allowed us to test for an interaction between the conditioning treatment (shrimp versus water) and the testing treatment (either shrimp versus mussel, or shrimp versus melon, depending on the experiment) in a 2
2 design. Post hoc testing to interpret interactions

involved splitting the data by the conditioning treatment and conducting ManneWhitney tests. For social learning data, each individual was tested twice (during conditioning and again during testing), and hence we used repeated measures ANOVAs with the experimental phase as the within-subjects factor (assumptions were met). For attraction scores from each experiment, we further assessed each treatment group versus the null expectation for no preference. The null expectation score differed between tests of embryonic learning (15 scoring periods in the middle of the arena (zone 3) ¼ a score of 45) and tests of social learning (24 scoring periods in the middle of the arena (zone 3e4 or ‘3.5’) ¼ a score of 84) due to differences in trial length and the number of scoring zones. We used one sample t tests (or Wilcoxon tests) to test for stimulus preferences versus the null expectations, and we reduced alpha for multiple comparisons with Bonferroni corrections (a/ 4 ¼ 0.0125). RESULTS Embryonic Learning: Shrimp versus Mussel Testing Shrimp versus mussel embryonic learning tests revealed no effect of the testing stimulus (F1,90 ¼ 0.04, P ¼ 0.85) and no interaction between the testing and conditioning stimuli (F1,90 ¼ 0.01, P ¼ 0.92; Fig. 3a). Instead, there was a significant main effect of the conditioning stimulus where embryos exposed to the shrimp stimulus were significantly more attracted to both testing stimuli (shrimp and mussel) as larvae (F1,90 ¼ 5.28, P ¼ 0.024; Fig. 3a). Moreover, our tests of preference matched these results (a ¼ 0.0125; shrimpeshrimp: t28 ¼ 3.19, P ¼ 0.004; shrimpemussel: t21 ¼ 2.74, P ¼ 0.012; watereshrimp: t19 ¼ 0.17, P ¼ 0.87; wateremussel: t22 ¼ 0.39, P ¼ 0.70; Fig. 3a), and indicated no innate attraction to either stimulus (i.e. in the water conditioning group). In terms of contacts, larvae that had embryonic experience with shrimp were again more attracted to both stimuli (F1,90 ¼ 6.37, P ¼ 0.013; Fig. 3b). However, larvae from both background treatments showed increased contact with the shrimp stimulus during testing, compared to the mussel stimulus

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

60

*

(a)

*

*

55

50

Attraction score

Attraction score

60

(a)

55

203

45 40

50 45 40

35 29

22

20

35

23

30

42 (b)

3

(b)

47

47

42

47

47

*

3 2

1

29 0

22

Shrimp Mussel test test Shrimp background

20

23

Shrimp Mussel test test Water background

Figure 3. Mean (±SE) attraction scores (a) and number of contacts (b) for larvae tested with either the shrimp or mussel stimulus following embryonic exposure to either the shrimp stimulus or the water control. A higher score indicates stronger attraction to the stimulus, and the dashed line represents no preference. Asterisks indicate statistical significance from null expectation. Numbers on bars represent sample sizes for each group.

(F1,90 ¼ 5.84, P ¼ 0.018). This result appeared mainly driven by larvae from the shrimp conditioning, but there was no significant interaction (F1,90 ¼ 0.85, P ¼ 0.36; Fig. 3b). Embryonic Learning: Shrimp versus Melon Testing When using melon as the novel control stimulus, there again was a significant main effect of the embryonic exposure on attraction to both stimuli (F1,174 ¼ 5.06, P ¼ 0.026; Fig. 4a). There was no effect of the testing stimulus (F1,174 ¼ 2.43, P ¼ 0.12), and no significant interaction (F1,174 < 0.01, P ¼ 0.95). However, specific tests for preference revealed that only larvae exposed to shrimp as embryos showed a preference that was significantly different from the null expectation (a ¼ 0.0125, t41 ¼ 3.36, P ¼ 0.002), unlike the other groups (all P > 0.10; Fig. 4a), and thus again there was no innate attraction. For contacts, a significant interaction occurred between the embryonic treatment and testing stimulus (F1,174 ¼ 4.23, P ¼ 0.041; Fig. 4b), revealing that only larvae exposed to the shrimp stimulus as embryos contacted it more frequently (shrimp background: U ¼ 1098.0, P ¼ 0.044; water background: U ¼ 1108.5, P ¼ 0.97; Fig. 4b).

Number of contacts

Number of contacts

42

30

2

1

42 0

Shrimp Melon test test Shrimp background

Shrimp Melon test test Water background

Figure 4. Mean (±SE) attraction scores (a) and number of contacts (b) for larvae tested with either the shrimp or melon stimulus following embryonic exposure to either the shrimp stimulus or the water control. A higher score indicates stronger attraction to the stimulus, and the dashed line represents no preference. Asterisks indicate statistical significance from the null expectation (a) or from the corresponding treatment (b). Numbers on bars represent sample sizes for each group.

models near the control water, those conditioned with models near the shrimp stimulus were more attracted to the stimulus during both the conditioning and testing phases (F1,40 ¼ 4.74, P ¼ 0.035; Fig. 5a), and there was no significant interaction (phase
conditioning treatment: F1,40 ¼ 0.47, P ¼ 0.50). Other terms were also nonsignificant (phase: F1,40 ¼ 0.15, P ¼ 0.70), as were the tests for preference (or avoidance) of the shrimp stimulus (all P > 0.10; Fig. 5a). When assessing the contacts made by larvae, we again found a significant main effect of the conditioning stimulus, where larvae that were conditioned with models near the shrimp stimulus made more contact with the shrimp stimulus during both the conditioning and testing phases (F1,40 ¼ 5.36, P ¼ 0.026). We also found significant differences among subjects (F40,40 ¼ 1.79, P ¼ 0.034), whereas other terms were nonsignificant (phase: F1,40 ¼ 2.35, P ¼ 0.13; phase
conditioning treatment: F1,40 ¼ 0.32, P ¼ 0.57; Fig. 5a). DISCUSSION

Social Learning In the social learning experiment, the conditioning treatment had a significant main effect on attraction. Compared to larvae with

Our results provide evidence for both embryonic and social learning in ringed salamanders. First, we found that embryonic exposure to a novel food stimulus (shrimp) led to increased

204

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

100 (a) Attaction (to shrimp) score

95 90 85 80 75 70 65

Number of contacts with shrimp

(b) 3.5 3 2.5 2 1.5 1 Conditioning: With shrimp models

Testing: All with shrimp

With water models Figure 5. Mean (±SE) attraction scores (a) and number of contacts (b) for larvae tested with the shrimp stimulus following social conditioning with a conspecific model corralled near either the shrimp stimulus or a water stimulus. A higher score indicates stronger attraction to the stimulus, and the dashed line represents no preference. Sample sizes were 21 per treatment group.

exploration of that stimulus later in life (as larvae), compared to control individuals. However, whether larvae became attracted to the stimulus because they recognized it specifically as food, or as a more general habitat cue is unknown. Learning while still in the egg provides the earliest possible acquisition of information, and without the higher predation and competition risk that will be present in the postnatal environment (Mathis et al., 2008). For juveniles of many species, quickly locating food can be critically important in allowing individuals to out-grow the gape of some of their major predators (Urban, 2007) (e.g. adult newts and larval conspecifics for ringed salamanders). The benefits of learning should also be high for species that occupy habitats that experience substantial seasonal and annual variation in prey availability (e.g. the ephemeral ponds used by ringed salamanders: Brooks, 2000). Here, we also found evidence that this learned attraction was generalized to another food stimulus (mussel), presumably due to overlap in the chemical composition of their odours, whereas attraction was not generalized to a plant stimulus (melon). However, melon and mussel share little chemical composition, so an alternative possibility is that the sensory system of these salamanders is not sensitive to plant odours such as

melon, although this seems unlikely given the sophistication of amphibian olfaction (Mason, Chivers, Mathis, & Blaustein, 1998; Mathis & Crane, 2017). Embryonic exposure to olfactory stimuli may shape the neuroanatomy of the developing olfactory system, resulting in enhancement of detection of those stimuli (Todrank, Heth, & Restrepo, 2010). However, a modification of behaviour following embryonic exposure suggests a memory of the stimulus. Our results revealed that larval salamanders modified their behaviour following embryonic exposure (~3 weeks prior), which can be explained by either olfactory imprinting or a mechanism known as sensitization e where a preference is learned and lasts for only a short period of time, irrespective of when the learning occurs (Pinsker, Hening, Carew, & Kandel, 1973). Olfactory imprinting involves learning to recognize a stimulus during an early sensitive period and then showing a specific long-term preference for that stimulus in the future (Lorenz, 1935, 1970). Such imprinting has been described as a homing mechanism for several species (e.g. Grassman, 1993; Hasler & Scholz, 2012; Ogurtsov, 2004). However, in our view, imprinting seems unlikely for larval salamanders because their diet appears to be highly plastic through development (Hutcherson, Peterson, & Wilkinson, 1989), and thus, the learned attraction should be only temporary. Hence, a learning mechanism of generalized embryonic sensitization may explain our results in the context of the plastic diet of larval ringed salamanders. Such a mechanism is consistent with other studies that have reported generalization of stimuli that are learned during embryonic development, although in different ecological contexts (e.g. Ferrari & Chivers, 2009b; Todrank & Heth, 2003). Embryonic learning by larval ringed salamanders has also been reported in the context of predation risk (Mathis et al., 2008), where embryonic exposure to chemical stimuli from cannibalistic larvae caused decreased levels of activity and increased use of habitat refuge in posthatching larvae. Hence, the embryos had learned that their habitat would be dangerous upon hatching. However, in a similar study on spotted salamanders, Ambystoma maculatum, no evidence for such learning was found when using embryonic exposures to chemical stimuli from a heterospecific predator (larval marbled salamanders, Ambystoma opacum) (Sehr, Beasley, Wilson, & Gall, 2016). Predators in both studies (Mathis et al., 2008; Sehr et al., 2016) were fed a diet of aquatic worms, L. variegatus, so perhaps the different outcomes resulted from a difference in innate recognition of the predator stimuli as a threat during the embryonic period. Neither study used injured conspecific cues as an unconditioned stimulus, as has been used in studies on embryonic learning of predator recognition in wood frogs, L. sylvaticus (e.g. Ferrari & Chivers, 2009a; Ferrari et al., 2010). Larval salamanders also appeared to learn about food from social information in this study, showing a subsequent attraction to the food stimulus after observing a conspecific model near that stimulus. We can rule out a simple overall attraction to shrimp because larvae exposed to water conditioning did not show such an attraction pattern. Behaviour during the conditioning phase was socially facilitated, with observers in both treatments being located closer to the model's end of the arena. However, the testing phase indicated that the difference between the social treatments persisted in the absence of models and hence was not simply socially facilitated but was indeed learned. However, the effect size for socially learned attraction was smaller than the stimulus attraction through social facilitation (Hedges' g: 1.9 versus 2.9) (Fig. 5a), although not so for contacts (Hedges' g: 0.9 versus 0.8) (Fig. 5b). Often, socially learned information does weaken through social transmission (Curio, 1988), and multiple trainings are often required to enhance social learning of food locations (e.g. Crane & Mathis, 2011; Reader, Kendal, & Laland, 2003).

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

The learning mechanism underlying our social learning results is likely ‘stimulus enhancement’. As described in the Introduction, this type of learning occurs when an individual changes behaviour following exposure to a ‘stimulus’ paired with a model (Heyes, Ray, Mitchell, & Nokes, 2000; Hoppitt & Laland, 2013). Local enhancement is a similar learning mechanism where the presence of others is used to learn a ‘location’ (Thorpe, 1956), but in our study, the location of the stimulus was randomized across trials, revealing that salamanders learned the stimulus rather than the location. Chapman, Holcomb, Spivey, Sehr, and Gall (2015) found that spotted salamander larvae, unlike wood frog tadpoles, did not become attracted to locations previously occupied by a group of conspecifics. This finding, coupled with our results, indicates that larval Ambystoma may not be attracted to conspecifics unless a potential food source is present. Such a strategy appears beneficial due to the risk of cannibalism for these larvae. Our study is the first documentation of a social learning outcome by a salamander, and provides further evidence that species with little social complexity may still be capable of learning from socially available information. Acknowledgments We thank the Department of Biology and the Graduate College at Missouri State University for funding and support. References Alcock, J. (2005). Animal behavior: An evolutionary approach. Sunderland, MA: Sinauer. Bornstein, M. H. (1989). Sensitive periods in development: Structural characteristics and causal interpretations. Psychological Bulletin, 105, 179e197. Brooks, R. T. (2000). Annual and seasonal variation and the effects of hydroperiod on benthic macroinvertebrates of seasonal forest (‘vernal’) ponds in central Massachusetts, USA. Wetlands, 20, 707e715. Brown, G. E., & Chivers, D. P. (2005). Learning as an adaptive response to predation. In P. Barbosa, & I. Castellanos (Eds.), Ecology of predatoreprey interactions (pp. 34e54). Oxford, U.K: Oxford University Press. Brown, C., & Laland, K. N. (2003). Social learning in fishes: A review. Fish and Fisheries, 4, 280e288. Brown, C., Markula, A., & Laland, K. (2003). Social learning of prey location in hatchery-reared Atlantic salmon. Journal of Fish Biology, 63, 738e745. Burghardt, G. M., & Hess, E. H. (1966). Food imprinting in the snapping turtle, Chelydra serpentina. Science, 151, 108e109. Chapman, T. L., Holcomb, M. P., Spivey, K. L., Sehr, E. K., & Gall, B. G. (2015). A test of local enhancement in amphibians. Ethology, 121, 308e314. Clayton, D. A. (1978). Socially facilitated behavior. Quarterly Review of Biology, 53, 373e392. Coolen, I., Dangles, O., & Casas, J. (2005). Social learning in noncolonial insects? Current Biology, 15, 1931e1935. Crane, A. L., & Mathis, A. (2011). Landmark learning by the Ozark zigzag salamander Plethodon angusticlavius. Current Zoology, 57, 485e490. Crane, A., Mathis, A., & McGrane, C. (2012). Socially facilitated antipredator behavior by ringed salamanders (Ambystoma annulatum). Behavioral Ecology and Sociobiology, 66, 811e817. Curio, E. (1988). Cultural transmission of enemy recognition by birds. In C. M. Heyes, & B. G. Galef (Eds.), Social learning in animals: The roots of culture (pp. 75e97). San Diego, CA: Academic Press. Dall, S. R. X. (2010). Managing risk: The perils of uncertainty. In D. F. Westneat, & C. W. Fox (Eds.), Evolutionary behavioral ecology (pp. 194e206). New York, NY: Oxford University Press. Darmaillacq, A.-S., Lesimple, C., & Dickel, L. (2008). Embryonic visual learning in the cuttlefish, Sepia officinalis. Animal Behaviour, 76, 131e134. Fawcett, T. W., & Frankenhuis, W. E. (2015). Adaptive explanations for sensitive windows in development. Frontiers in Zoology, 12(Suppl. 1), S3. https://doi.org/ 10.1186/1742-9994-12-S1-S3. Ferrari, M. C. O., & Chivers, D. P. (2009a). Latent inhibition of predator recognition by embryonic amphibians. Biology Letters, 5, 160e162. Ferrari, M. C. O., & Chivers, D. P. (2009b). Sophisticated early life lessons: Threatsensitive generalization of predator recognition by embryonic amphibians. Behavioral Ecology, 20, 1295e1298. Ferrari, M. C. O., Manek, A. K., & Chivers, D. P. (2010). Temporal learning of predation risk by embryonic amphibians. Biology Letters, 6, 308e310. Galef, B. G. (1993). Functions of social learning about food: A causal analysis of effects of diet novelty on preference transmission. Animal Behaviour, 46, 257e265.

205

Galef, B. G., & Laland, K. N. (2005). Social learning in animals: Empirical studies and theoretical models. BioScience, 55, 489e499. Garcia, T. S., Urbina, J. C., Bredeweg, E. M., & Ferrari, M. C. (2017). Embryonic learning and developmental carry-over effects in an invasive anuran. Oecologia, 184, 623e631. Grassman, M. (1993). Chemosensory orientation behavior in juvenile sea turtles. Brain, Behavior and Evolution, 41, 224e228. , M., Poirel, N., Houde , O., & Dickel, L. (2012). Food imprinting and visual Guibe generalization in embryos and newly hatched cuttlefish, Sepia officinalis. Animal Behaviour, 84, 213e217. Guttridge, T. L., van Dijk, S., Stamhuis, E. J., Krause, J., Gruber, S. H., & Brown, C. (2013). Social learning in juvenile lemon sharks, Negaprion brevirostris. Animal Cognition, 16, 55e64. Harrison, R. (1969). Harrison stages and description of the normal development of the spotted salamander, Amblystoma punctatum (Linn.). In S. Wilens (Ed.), Organization and development of the embryo (pp. 44e66). New Haven, CT: Yale University Press. Hasler, A. D., & Scholz, A. T. (2012). Olfactory imprinting and homing in salmon: Investigations into the mechanism of the imprinting process. Berlin, Germany: Springer Science & Business Media. Hepper, P. G., & Waldman, B. (1992). Embryonic olfactory learning in frogs. Quarterly Journal of Experimental Psychology: Section B, 44, 179e197. Heyes, C. M. (1994). Social learning in animals: Categories and mechanisms. Biological Reviews, 69, 207e231. Heyes, C. M., Ray, E., Mitchell, C., & Nokes, T. (2000). Stimulus enhancement: Controls for social facilitation and local enhancement. Learning and Motivation, 31, 83e98. Higgins, J. J., & Tashtoush, S. (1994). An aligned rank transform test for interaction. Nonlinear World, 1, 201e211. Hoppitt, W., & Laland, K. N. (2013). Social learning: An introduction to mechanisms, methods, and models. Princeton, NJ: Princeton University Press. Hutcherson, J. E., Peterson, C. L., & Wilkinson, R. F. (1989). Reproductive and larval biology of Ambystoma annulatum. Journal of Herpetology, 23, 181e183. Immelmann, K. (1975). Ecological significance of imprinting and early learning. Annual Review of Ecology and Systematics, 6, 15e37. Jefferson, D. M., Ferrari, M. C., Mathis, A., Hobson, K. A., Britzke, E. R., Crane, A. L., et al. (2014). Shifty salamanders: Transient trophic polymorphism and cannibalism within natural populations of larval ambystomatid salamanders. Frontiers in Zoology, 11, 76. https://doi.org/10.1186/s12983-014-0076-7. Kacelnik, A., & Bateson, M. (1996). Risky theories: The effects of variance on foraging decisions. American Zoologist, 36, 402e434. Kis, A., Huber, L., & Wilkinson, A. (2015). Social learning by imitation in a reptile (Pogona vitticeps). Animal Cognition, 18, 325e331. Lefebvre, L., & Bouchard, J. (2003). Social learning about food in birds. In D. M. Fragaszy, & S. Perry (Eds.), The biology of traditions: Models and evidence (pp. 94e126). Cambridge, U.K: Cambridge University Press. Lickliter, R., & Hellewell, T. B. (1992). Contextual determinants of auditory learning in bobwhite quail embryos and hatchlings. Developmental Psychobiology, 25, 17e31. Lima, S. L., & Dill, L. M. (1990). Behavioral decisions made under the risk of predation: A review and prospectus. Canadian Journal of Zoology, 68, 619e640. Lorenz, K. (1935). Der Kumpan in der Umwelt des Vogels. Journal of Ornithology, 83, 137e213, 289e413. Lorenz, K. (1970). Companions as factors in the bird's environment. Studies in Animal and Human Behavior, 1, 101e258. Mason, R. T., Chivers, D. P., Mathis, A., & Blaustein, A. R. (1998). Bioassay methods for amphibians and reptiles. In K. F. Haynes, & J. G. Millar (Eds.), Methods in chemical ecology (Vol. 2, pp. 271e325). Norwell, MA: Kluwer Academic. Mathis, A., & Crane, A. L. (2017). Chemoreception. In J. Call (Ed.), APA handbook of comparative psychology (pp. 69e87). Washington, D.C: APA Books. Mathis, A., Ferrari, M. C. O., Windel, N., Messier, F., & Chivers, D. P. (2008). Learning by embryos and the ghost of predation future. Proceedings of the Royal Society B: Biological Sciences, 275, 2603e2607. Mathis, A., Murray, K. L., & Hickman, C. R. (2003). Do experience and body size play a role in responses of larval ringed salamanders, Ambystoma annulatum, to predator kairomones? Laboratory and field assays. Ethology, 109, 159e170. Nelson, A. B., Alemadi, S. D., & Wisenden, B. D. (2013). Learned recognition of novel predator odour by convict cichlid embryos. Behavioral Ecology and Sociobiology, 67, 1269e1273. Noble, D. W., Byrne, R. W., & Whiting, M. J. (2014). Age-dependent social learning in a lizard. Biology Letters, 10, 20140430. https://doi.org/10.1098/rsbl.2014.0430. Ogurtsov, S. (2004). Olfactory orientation in anuran amphibians. Russian Journal of Herpetology, 11, 35e40. Papaj, D. R., & Prokopy, R. J. (1989). Ecological and evolutionary aspects of learning in phytophagous insects. Annual Review of Entomology, 34, 315e350. Peterson, C. L., Wilkinson, R. F., Moll, D., & Holder, T. (1991). Premetamorphic survival of Ambystoma annulatum. Herpetologica, 47, 96e100. Petranka, J. W. (1998). Salamanders of the United States and Canada. Washington, D.C.: Smithsonian Institution Press. Pianka, E. R. (1970). On r- and K-selection. American Naturalist, 104, 592e597. Pinsker, H. M., Hening, W. A., Carew, T. J., & Kandel, E. R. (1973). Long-term sensitization of a defensive withdrawal reflex in Aplysia. Science, 182, 1039e1042. Punzo, F. (2002). Food imprinting and subsequent prey preference in the lynx spider, Oxyopes salticus (Araneae: Oxyopidae). Behavioural Processes, 58, 177e181.

206

A. L. Crane et al. / Animal Behaviour 142 (2018) 199e206

Quesada, P. C. P., & Schausberger, P. (2012). Prenatal chemosensory learning by the predatory mite Neoseiulus californicus. PLoS One, 7, e53229. https://doi.org/10. 1371/journal.pone.0053229. Reader, S. M., Kendal, J. R., & Laland, K. N. (2003). Social learning of foraging sites and escape routes in wild Trinidadian guppies. Animal Behaviour, 66, 729e739. Sehr, E. K., Beasley, L. N., Wilson, K. W., & Gall, B. G. (2016). When can embryos learn? A test of the timing of learning in embryonic amphibians. Ecology and Evolution, 6, 2486e2493. Sneddon, H., Hadden, R., & Hepper, P. (1998). Chemosensory learning in the chicken embryo. Physiology & Behavior, 64, 133e139. Sneddon, H., Hepper, P. G., & Manolis, C. (2000). Embryonic chemosensory learning in the saltwater crocodile, Crocodylus porosus. In G. C. Grigg, F. Seebacher, & C. E. Franklin (Eds.), Crocodilian biology and evolution (pp. 378e382). Chipping Norton, NSW: Surrey Beatty. Sogard, S. M. (1997). Size-selective mortality in the juvenile stage of teleost fishes: A review. Bulletin of Marine Science, 60, 1129e1157. Spotila, J. R., & Beumer, R. J. (1970). The breeding habits of the ringed salamander, Ambystoma annulatum (Cope), in northwestern Arkansas. American Midland Naturalist, 84, 77e89. Stephens, D. (1991). Change, regularity, and value in the evolution of animal learning. Behavioral Ecology, 2, 77e89. Stephens, D. W. (1993). Learning and behavioral ecology: Incomplete information and environmental predictability. In D. Papaj, & A. Lewis (Eds.), Insect learning: Ecological and evolutionary perspectives (pp. 195e218). New York, NY: Chapman & Hall. Thorpe, W. (1956). Learning and instinct in animals. London, U.K: Methuen.

Todrank, J., & Heth, G. (2003). Odoregenes covariance and genetic relatedness assessments: Rethinking odor-based ‘recognition’ mechanisms in rodents. Advances in the Study of Behavior, 32, 77e130. Todrank, J., Heth, G., & Restrepo, D. (2010). Effects of in utero odorant exposure on neuroanatomical development of the olfactory bulb and odour preferences. Proceedings of the Royal Society B: Biological Sciences, 278(1714). https://doi.org/ 10.1098/rspb.2010.2314, 20102314. Urban, M. C. (2007). The growthepredation risk trade-off under a growing gapelimited predation threat. Ecology, 88, 2587e2597. Watson, S., & Russell, A. P. (2000). A posthatching developmental staging table for the long-toed salamander, Ambystoma macrodactylum krausei. Amphibia-Reptilia, 21, 143e154. Wells, K. D. (2010). The ecology and behavior of amphibians. Chicago, IL: University of Chicago Press. Wells, D. L., & Hepper, P. G. (2006). Prenatal olfactory learning in the domestic dog. Animal Behaviour, 72, 681e686. Whiten, A. (2000). Primate culture and social learning. Cognitive Science, 24, 477e508. Wilkinson, A., Kuenstner, K., Mueller, J., & Huber, L. (2010). Social learning in a nonsocial reptile (Geochelone carbonaria). Biology Letters, 6, 614e616. Wilkinson, A., Mandl, I., Bugnyar, T., & Huber, L. (2010). Gaze following in the redfooted tortoise (Geochelone carbonaria). Animal Cognition, 13, 765e769. Wobbrock, J. O., Findlater, L., Gergle, D., & Higgins, J. J. (2011). The aligned rank transform for nonparametric factorial analyses using only anova procedures. In Proceedings of the SIGCHI conference on human factors in computing systems, Vancouver, BC, Canada, 7e12 May 2011 (pp. 143e146). New York, NY: ACM Press. https://doi.org/10.1145/1978942.1978963.