Configural learning: a higher form of learning in Lymnaea Free

Learning, especially associative learning and the subsequent formation of memory, is a vital survival mechanism enabling organisms to generate internal representations of correlations between relevant environmental stimuli. The ability to assign importance to salient stimuli in the environment allows organisms to resolve ambiguity and respond purposefully to their environment (Maren, et al., 2013). The context in which stress is experienced becomes intrinsic to the representation of the experience and influences how and under what circumstances the memory is recalled (de Kloet et al., 1999; Joëls et al., 2006). Here, we tested the hypothesis that Lymnaea are capable of a form of ‘higher-order’ learning; namely, they possess the cognitive capability for configural learning. We define cognitive capability hereafter Dyer (2012) as the capacity to learn and process information in a sophisticated manner not predicted by elemental responses to stimuli. Hence, the meaning of the stimulus depends on what other stimuli are encountered with it (i.e. the relationships between the sensory components). We define this relationship as configural learning, using the term as defined by Sutherland and Rudy in 1989 . That is, the meaning of a stimulus depends on what other stimuli are experienced with it (see also Rudy and Sutherland, 1989; Stout et al., 2018). Configural learning results in the ability to treat stimuli experienced together as different from the simple sum of their elements (Giurfa, 2003, 2007; Mansur et al., 2018). We believe that configural learning is vitally important and essential for survival to Lymnaea as it, along with instinct, occurs in the snails' natural environment, which is both changing and composed of multimodal stimuli, and it is these ‘stimuli compounds’ that animals associate events with rather than a single unimodal element (see Lorenz, 1951).

There are distinct populations or strains of Lymnaea (Dodd et al., 2018) that can be categorized, for example, according to their cognitive or behaviour responses to specific predators (Rothwell et al., 2018). As regards their crayfish predator, there are both predator-experienced and predator-naive strains of Lymnaea (Orr et al., 2009a). Here, we define predator-experienced Lymnaea as those strains that innately respond with anti-predator behaviours when they detect chemical signals (kairomones) that convey that a crayfish predator is present. However, to be predator experienced need not mean that the snail has actually detected a crayfish at any time in its life. Thus, our W-strain snails, which have been in the laboratory since the 1950s, respond to crayfish kairomones even though they have never actually experienced a crayfish in almost 70 years. Hence, we define a strain as predator experienced if snails respond to the chemical cues with anti-predator behaviours (Burks and Lodge, 2002; Orr et al., 2007; Il-Han et al., 2010). For example, when predator-experienced Lymnaea such as our W-strain detect crayfish, a shell-crushing predator, via crayfish effluent (CE), escape behaviours including crawling above the water (Rigby and Jokela, 2000; Dalesman et al., 2006) and increased shadow-elicited withdrawal responses occur (Orr et al., 2007). Importantly, exposing these snails to CE enhances long-term memory (LTM) formation in Lymnaea (Orr and Lukowiak, 2008). Typically, snails will detect CE in a specific stimulus context. We hypothesized that predator-experienced snails will form a connection between an environmental stimulus (e.g. a food odour) and CE, such that subsequent food odour exposure will trigger behaviours typically evoked by CE on its own. These behaviours include suppression of feeding and enhancement of memory formation. A strain is considered to be predator naive if snails of that strain (e.g. TC1 snails) do not respond to CE or other cues transmitted by crayfish with anti-predator behaviours (Orr et al., 2009a).

Here, we used an elemental form of associative learning, operant conditioning of aerial respiratory behaviour. In the inbred predator-experienced W-strain Lymnaea used, LTM formation requires at least two 0.5 h training sessions separated by a 1 h interval (Lukowiak et al., 1996, 1998). In this strain, a single 0.5 h training session does not cause LTM formation (Sangha et al., 2003; Forest et al., 2016). However, if trained in CE, a single 0.5 h training session is sufficient to cause LTM to form (Orr and Lukowiak, 2008; Sunada et al., 2010; Il-Han et al., 2010; Forest et al., 2016). Hence, training in CE promotes enhanced LTM formation. However, if snails are exposed to CE but not immediately trained (i.e. more than a 0.5 h wait), training in normal pond water (PW) does not result in enhanced LTM formation (Orr and Lukowiak, 2008, 2010). Therefore, CE enhancement of LTM formation is an activity-dependent (i.e. training) phenomenon. In the W-strain, specific stressors applied before or during training alter LTM formation in Lymnaea, either enhancing or obstructing it (Martens et al., 2006; Dalesman et al., 2013; Lukowiak et al., 2014; Knezevic et al., 2016; Hughes et al., 2017).

Manipulating the context in which a snail undergoes associative learning alters the snail's ability to retrieve memory (Haney and Lukowiak, 2001; Lukowiak et al., 2007). We define context as the surrounding environmental conditions in which a learning task occurs, including stimuli such as food odours. Haney and Lukowiak (2001) demonstrated that carrot odour (CO) can be used to create context-specific memories, and this technique has been implemented successfully in other publications from our laboratory (e.g. Lukowiak et al., 2007; Hughes et al., 2016, 2017). Context learning has been demonstrated in invertebrate model systems including Aplysia, honeybees, Drosophila and ants, and in mammalian preparations (Waddell and Quinn, 2001; Giurfa, 2003; Leonard and Edstrom, 2004; Vianna et al., 2004; Bos et al., 2010). Typically, animals trained in one context only showed memory if the animal was tested for memory in the same context. However, context-specific memory cannot be applied across all conditions. For example, predator-experienced Lymnaea trained in CE demonstrate context generalization. To wit, following training in CE, memory is recalled in multiple contexts (e.g. PW or CO) despite the animal being capable of discriminating between the stimuli (Orr and Lukowiak, 2008). This stimulus generalization provides an animal with an adaptive advantage because it allows the animal to generalize the context of a predator encounter to different environmental conditions.

Elemental learning or ‘lower-order’ forms of associative learning (i.e. classical and operant conditioning) have been shown in Lymnaea (Ito et al., 2013). Additionally, Lymnaea consolidate these different forms of associative learning into LTM (Benjamin et al., 2000; Lukowiak et al., 2003; Fulton et al., 2005; McComb et al., 2005a,b; Parvez et al., 2005, 2006; Sakakibara et al., 2005). Forms of higher-order conditioning are conditional learning, sensory pre-conditioning, configural learning and second-order conditioning (Giurfa, 2007). However, we have yet to perform experiments to determine whether forms of higher-order associative learning such as second-order conditioning or configural learning occur with operant conditioning in Lymnaea. We have previously shown second-order classical conditioning using our taste aversion procedure in Lymnaea (Sugai et al., 2006); thus, we were optimistic that configural learning was possible. Higher­-order conditioning is considered to be more ‘cognitive’, as the animal's behaviour depends on its ability to make meaningful comparisons between current sensory stimuli and its representation of previous sensory experiences (Sahley et al., 1981; Hawkins et al., 1998; Giurfa, 2003, 2007; Devaud et al., 2015; Onuma and Sakai, 2016).

We hypothesized that configural learning in Lymnaea will occur when predator-experienced Lymnaea simultaneously encounter a food odour (e.g. CO) and CE. Here, we show that when predator-experienced Lymnaea are exposed to a stimulus that does not elicit anti-predator behaviours, such as CO, at the same time as they experience CE (i.e. CO+CE), CO now elicits behaviours previously elicited by the predator odour. These snails trained in CO show enhanced LTM formation and decreased rasping in response to CO.

The animals used here are an inbred laboratory strain (W-strain) of Lymnaea stagnalis (Linnaeus 1758) maintained at the University of Calgary Biology Department. W-strain snails respond innately and appropriately to CE; thus, they are classified as predator experienced even though they have never experienced a crayfish predator. The W-strain L. stagnalis used here originated from an inbred stock maintained at the Vrije University of Amsterdam and were originally bred from animals collected in the 1950s in polders near Utrecht, The Netherlands. The W-strain snails are housed in artificial PW (0.25 g l⁻¹ of Instant Ocean in deionized water, Spectrum Brands, Madison, WI, USA) supplemented with CaCO₃ to ensure calcium concentrations remain above 50 mg (Dalesman and Lukowiak, 2010). Snails were fed romaine lettuce ad libitum and maintained at 20±1°C on a 16 h:8 h light:dark cycle. A total of 163 snails were used, and a snail was not used more than once in any experiment.

Crayfish are a natural predator of Lymnaea. Predator-experienced Lymnaea innately detect and respond to the scent of a crayfish (CE) with several anti-predator behaviours (Alexander and Covich, 1991; Dalesman et al., 2006; Orr et al., 2007; Rigby and Jokela, 2000). The crayfish used here were housed in a 70 l aquarium containing artificial pond water and were fed a diet of lettuce and snails. The water in the crayfish tank is termed crayfish effluent (CE; Orr et al., 2007). Previously, we have demonstrated that exposing predator-experienced Lymnaea to CE during training enhances LTM formation (Orr and Lukowiak, 2008, 2010; Lukowiak et al., 2014). It should be noted here that some strains of Lymnaea (e.g. TC1 snails collected from a pond in Alberta) are predator naive as regards crayfish (Orr et al., 2009a,b). That is, they do not respond with anti-predator behaviours to crayfish. Here, we examined whether simultaneously exposing snails to CE plus a food odorant (CO, see below; CO+CE) causes configural learning, such that the CO will elicit anti-predator behaviours in predator-experienced Lymnaea.

Lymnaea exhibit context-specific learning, memory formation and recall, where snails operantly conditioned in CO during training behave as if they are naive (i.e. no LTM) when tested in PW (i.e. no CO; Haney and Lukowiak, 2001; Lukowiak et al., 2007). The CO slurry was made by combining a single large commercially obtained carrot in a blender along with approximately 350 ml of PW. Following blending and straining of the mixture, we obtained a carrot–PW slurry without any observable pieces of carrot (i.e. CO). Thus, we used CO as a chemosensory cue in our experiments. We exposed snails to CO by setting up an apparatus that bubbles either eumoxic air or hypoxic N₂ gas through a sealed flask containing the carrot slurry. The gas bubbled into the flask smelling of carrot was then diverted from the flask into a beaker containing PW. Snails were exposed to CO either under eumoxic conditions prior to training or under hypoxic conditions during conditioning of aerial respiration. It needs to be explicitly stated that multiple batches of fresh carrots purchased from different grocery stores were used over a 6 week period for Figs 1–6 and a 4 month period for Fig. 7.

Operant conditioning training occurred in three hypoxic contexts: (1) CO; (2) CE; or (3) PW. Individually labelled snails were transferred from their home aquarium into a 1 l beaker filled with 500 ml of one of the three contexts. The specific training environment was made hypoxic (≤0.1 ml O₂ l⁻¹) by vigorously bubbling N₂ gas through the water for 20 min prior to the operant conditioning session. This is the procedure we have used since our initial study in 1996 (Lukowiak et al., 1996). The hypoxic environment caused snails to attempt to open their pneumostome more frequently. Following this 20 min period, the intensity of bubbling was reduced and the snails to be trained were placed in the beaker for a 10 min acclimation period before training began. The reduced bubbling maintained the established hypoxia without disturbing the snails. The operant conditioning procedure consisted of applying a tactile stimulus (i.e. a poke) to the edge of the pneumostome with a wooden applicator every time a snail attempted to perform aerial respiration. The stimulus was strong enough to cause the pneumostome to close, but was gentle enough that the snails did not complete a full body withdrawal response. The total number of pokes per snail was recorded. Between sessions, snails were returned to their eumoxic home aquarium. In all cases, the memory test (MT) session was performed in hypoxic PW. As in the training session, snails received a tactile stimulus each time they attempted to open their pneumostome in the MT. LTM has been operantly defined as significantly fewer attempted pneumostome openings in the MT session than in the training session. In the W-strain, the standard operant conditioning procedure requires two 0.5 h training sessions (TS) separated by 1 h to cause formation of a 24 h LTM; however, enhanced LTM formation occurs with a single 0.5 h TS in this strain when trained in the presence of certain stressors (e.g. CE-predator detection), as well as when trained in food stuffs containing epicatechin (Lukowiak et al., 2000, 2014; Knezevic et al., 2016; Swinton et al., 2018).

We placed a naive cohort of 10–12 W-strain snails in a beaker containing 500 ml of CE, made eumoxic by bubbling CO atmospheric air into the beaker. Thus, snails simultaneously experienced CO+CE. The snails were left in this beaker undisturbed for 45 min and subsequently transferred back to their home aquarium (i.e. neither CO nor CE present), where they remained for 2 h. The snails were then trained with a single 0.5 h training session in hypoxic CO using the operant conditioning of aerial respiration procedure as described above (Fig. 1). In all experimental and control groups (see below), MT was performed 24 h later in PW (i.e. neither CO nor CE was present in the MT session).

We first exposed a naive cohort of snails to CO alone for a 45 min period. That is, they did not experience CE, only CO. These snails were then trained as described above in hypoxic CO (Fig. 2). Second, we trained a naive cohort of snails that had been exposed to only CE for 45 min. That is, they did not experience CO. These snails were then trained as described above in hypoxic CO (Fig. 3). Third, we paired CO with boiled CE and exposed a naive cohort of snails to this for 45 min. Boiling alters the composition of the CE (e.g. metabolites in the urine; see Poulin et al., 2018) such that snails no longer respond to it with anti-predator behaviours (Orr et al., 2007). These snails were then trained as described above in hypoxic CO (Fig. 4). Finally, we exposed a naive cohort of snails to CO+CE for 45 min (Fig. 5) or 10 min (Fig. 6). However, these snails were trained in hypoxic PW (i.e. neither CO nor CE present).

Feeding behaviour or rasping in Lymnaea is a rhythmic motor behaviour in which repeated movements of the radulae scrape the surface of a substrate, leading to the ingestion of food (Ito et al., 2013). In this experiment, snails were placed into a 14 cm diameter Petri dish with enough normal eumoxic PW or eumoxic CO-PW for the snails to be fully submerged. The snails were given a 10 min acclimation period in each session (i.e. PW and CO-PW) before their rasping behaviour was monitored. Each snail was monitored for 2 min and the number of rasps counted; the average number of rasps per minute was then calculated. In Fig. 7A, snails were exposed to CO only, on two occasions 24 h apart; during this period, snails were not fed. In Fig. 7B, snails were first monitored in PW, returned to their home aquarium for 3 h, then tested in CO-PW. Snails were returned to their home aquarium for 24 h, and were not fed during this interval. Then, rasping behaviour was again determined in these snails in PW. Three hours later they were exposed to CO+CE for 45 min. Following this pairing they were returned to their home aquarium for 3 h before rasping behaviour was again determined in CO-PW.

Using Prism 7 for Mac, a two-way repeated-measures analysis of variance (RM-ANOVA) determined whether there was a statistical difference in the number of attempted pneumostome openings in each training and memory test session for the data in Figs 1–5. A post hoc test (Sidak's multiple comparisons test) was then performed to determine whether there was a statistical difference between TS and MT for each cohort of snails subjected to the various conditions (i.e. paired CO+CE, etc.). We chose Sidak's test over the more popular Bonferroni test because Sidak's test has more power than the Bonferroni post hoc test, as recommended by Prism. A one-way RM-ANOVA followed by Tukey's post hoc tests was employed to compare whether there were significant differences in rasping behaviour. In Fig. 7A, a paired t-test was used. Differences were considered significant if P-values were smaller than 0.05.

We hypothesized that Lymnaea are capable of a form of higher-order learning known as configural learning. If our hypothesis is correct, W-strain snails will learn to associate the simultaneous of a food smell (CO) with the ‘smell’ of a predator (i.e. CE). The statistical analyses of data (a two-way ANOVA followed by Sidak's test) from the experiments performed to test this hypothesis are presented here and in Table 1. We report the following for the data in Figs 1–6: (1) there was a significant interaction effect (F_5,258=2.669; P=0.0226); (2) there was also a significant ‘pairing effect’ (i.e. comparing the data from CO+CE with those from CO only, etc.: F_5,258=10.70; P<0.0001); and (3) there was no significant TS and MT effect (i.e. TS in CO+CE versus CO only, etc.: F_1,258=0.8450; P=0.3588). The post hoc Sidak's multiple comparisons tests are presented in Table 1 and show that the only significant difference between TS and MT was in the CO+CE pairing group when snails were then trained in CO. This was the only cohort to exhibit LTM.

We first placed a cohort of untrained W-strain snails (Fig. 1, n=21) into a beaker containing eumoxic CE-PW (CE) and CO. Snails experienced this procedure for 45 min (i.e. paired CO+CE) and were then returned to their home eumoxic aquaria (important to note, neither CE nor CO was present there). Two hours later, this cohort of snails was trained with a single 0.5 h TS in hypoxic CO and then tested 24 h later (MT) in hypoxic PW (CO not present). There were significantly fewer attempted pneumostome openings in the MT than in the TS (P<0.01, determined by Sidak's post hoc test). Thus, after the simultaneous exposure to CO+CE for 45 min, when trained in CO, snails exhibited an enhanced ability to form LTM in PW. Importantly, note that the TS occurred in CO 2 h after the simultaneous exposure to CO+CE, whilst the MT session 24 h later occurred in PW.

We next performed a number of experiments to determine whether it was both necessary and sufficient to experience the CO+CE pairing for CO to cause enhancement of LTM formation. First, we exposed an untrained cohort of snails (n=29) to CO alone for 45 min and then, following a 2 h rest, trained them in hypoxic CO (Fig. 2). These snails did not form LTM in PW. That is, there was no significant difference (determined by Sidak's post hoc test) between the number of attempted pneumostome openings in the MT and TS. We then exposed a cohort of untrained snails (n=21) to CE only, for 45 min, and then trained them 2 h later in hypoxic CO (Fig. 3). As can be seen, LTM was not formed. That is, the number of attempted pneumostome openings in the MT was not significantly lower than the number in the TS (determined by a Sidak's post hoc test). Thus, in these two control experiments, training in hypoxic CO was not sufficient to lead to LTM formation when the CO+CE pairing did not occur.

It has previously been shown (Orr et al., 2007, 2009b) that the crayfish kairomones that signal predator presence and which elicit anti-predator behaviours in W-strain Lymnaea are inactivated by boiling the CE. Therefore, we first boiled the CE, let it cool to room temperature, and then added the CO mixture. We then exposed a cohort of untrained snails (n=22) to a pairing of CO+boiled CE for 45 min. When these snails were trained 2 h later in hypoxic CO, they did not exhibit LTM (Fig. 4), as there was no significant difference between the number of attempted openings in the MT and TS (determined by Sidak's post-hoc test).

We next determined whether training snails in CO subsequent to the CO+CE pairing is necessary to cause enhanced LTM formation. We simultaneously exposed a cohort of untrained snails (n=22) to CO+CE for 45 min and then 2 h later we trained these snails (Fig. 5) in PW (not CO as we did in Fig. 1). We tested memory 24 h later in PW and found that LTM was not present (determined by Sidak's post hoc test).

The final experiment examining LTM formation that we performed was to simultaneously expose a cohort of untrained snails (n=22) to CO+CE for 10 min rather than 45 min (Fig. 6). We then trained these snails in CO as in Fig. 1 and tested memory 24 h later in PW. We found that LTM was not present in the MT, as there was no significant reduction (determined by Sidak's post hoc test) in the number of attempted pneumostome openings from the TS to the MT. That is, a 10 min simultaneous exposure to CO+CE was not sufficient for CO to cause enhancement of LTM.

Next, we asked whether configural learning (i.e. the simultaneous exposure for 45 min to CO+CE) also alters the feeding response of the snails. As part of the suite of anti-predator behaviours elicited in predator-experienced Lymnaea by predator detection (e.g. CE), feeding behaviour (rasping) is significantly reduced (Dalesman et al., 2006; Orr et al., 2007). We first determined that the feeding response elicited by CO was not significantly reduced with two repeated presentations of CO. Thus, an untrained cohort of snails (n=19) was exposed to CO and then 24 h later exposed to CO a second time. These data are presented in Fig. 7A. It is important to note here that in the 24 h interval between the first and second exposure to CO, snails were not fed. As can be seen, rasping behaviour was not significantly decreased on the second exposure to CO; in fact, it significantly increased (paired t-test; t=6.801, d.f.=18; P<0.0001). We then performed a similar experiment using another cohort of untrained snails (n=26). These data are presented in Fig. 7B. The number of rasps in these snails was observed in PW and CO before and after CO+CE pairing. An ANOVA (F_3,100=22.80; P<0.0001) followed by Tukey's post hoc test indicated that there were significant differences in the feeding responses in this experiment. The rate of rasping seen in PW is typical for Lymnaea (Sugai et al., 2006; Ito et al., 2013). Following the initial observation of rasping in PW, we returned the snails to their home aquarium for 3 h. These snails were subsequently exposed to CO and their rasping behaviour was observed. As can be seen, the number of rasps increased significantly compared with the response in PW (P<0.001). Following this, the snails were returned to their home aquarium for 24 h and, as in Fig. 7A, were not fed. They were again observed in PW. The number of rasps in their second exposure to PW was not different from the number of rasps observed in their first exposure, 24 h previously. Immediately after their second exposure to PW, the snails experienced the CO+CE pairing for 45 min and were then returned to their home aquarium. We then tested their response to CO 3 h later. The number of rasps elicited by this exposure to CO decreased significantly (P<0.001) compared with their initial response to CO (i.e. the response before the CO+CE pairing). The number of rasps elicited by CO was now not significantly different from the rasping behaviour seen in PW. Thus, following the CO+CE pairing, CO no longer elicited a significant increase in the feeding response.

We show here that predator-experienced W-strain Lymnaea possess features of ‘higher-order’ learning as they have the ability to associate a food odour (CO), which inherently elicits a feeding response, with the scent of a predator (i.e. CE) when they simultaneously experience CO+CE. During this simultaneous exposure, the snails freely behave and no other external stimuli are applied. The experiencing of the two stimuli together leads to CO, which previously did not elicit anti-predator behaviours, to acquire the properties of the predator kairomone (i.e. CE) in eliciting anti-predator behaviours (i.e. enhanced memory-forming ability and suppression of feeding behaviour). We classify this new association between CO and CE as configural learning as defined by Sutherland and Rudy (1989). That is, snails have the ability to assign a new ‘meaning’ to the odour stimulus (CO) as a result of experiencing CO simultaneously in the presence of kairomones from crayfish (CE) that evoke anti-predator responses. We believe that configural learning is of great importance to the snail as it may be an essential form of learning in the snails' natural environment. This new learning (i.e. CO signalling anti-predator behaviour) persists for at least 2 h. We hypothesize snails come to associate the CO with a ‘motivational state’ that is instinctively elicited by CE in W-strain snails. Consequently, CO after pairing now elicits a ‘fear state’ as a result of the association made by the snails between CO and CE. A component of this fear state, the suppression of the feeding response, becomes elicited by CO (Fig. 7B). CO now signals ‘fear’ rather than eliciting an appetitive feeding response. CO, therefore, has acquired a new motivational state (i.e. fear) as opposed to its intrinsic motivational state (i.e. enhanced rasping). This change in stimulus signalling did not occur when CO was paired with boiled CE, as boiling CE inactivates its predator-signalling properties (Orr et al., 2007). Configural learning also did not occur with a 10 min simultaneous exposure to CO+CE. Finally, enhanced LTM formation did not happen if, after the simultaneous 45 min pairing of CO+CE, the snails were trained in PW and not CO.

To our knowledge in an invertebrate model system, configural learning has only previously been shown in bees (Giurfa, 2003, 2007, Dyer, 2012; Devaud et al., 2015). For example, bees possess the ability to acquire olfactory patterning discrimination. Interestingly, this ability is blocked when neural activity in the mushroom bodies is blocked, even though elemental learning still occurs. Other ‘higher forms’ of learning such as second-order conditioning and contingency learning have been shown in a number of molluscs including Limax (Sahley et al., 1981, 1990), Hermissenda (Farley, 1987), Aplysia (Hawkins et al., 1998) and Lymnaea (Sugai et al., 2006). The second-order conditioning seen in Sugai et al.’s (2006) Lymnaea study utilized a conditioned taste aversion (CTA) procedure (i.e. classical conditioning), which is different from the operant conditioning procedure used here. We direct interested readers to the paper by Pearce (2002) on configural versus more ‘elemental’ forms of learning. Thus, our finding of higher-order learning here is not too surprising. What is different in our study is that we showed that Lymnaea are capable of configural learning using an operant conditioning procedure. That is, snails acquired a new association by experiencing a carrot odour, which typically only elicits feeding behaviour (Sugai et al., 2006), simultaneously with a crayfish kairomone (CE), which elicits a suite of anti-predator behaviours, including enhanced memory formation and suppression of feeding (Orr et al., 2009a,b). The simultaneous encountering of CO+CE leads to CO subsequently acquiring an ability on its own to elicit anti-predator behaviours rather than eliciting an increase in feeding. We believe this is the initial instance of configural learning in a mollusc.

One of the stress factors that animals experience in their natural environment leading to alterations in behaviour, physiology and even anatomy (e.g. shell thickness or shape) is predator detection (Rundle and Brönmark, 2001; Orr et al., 2007; Herberholz et al., 2004; Stokes et al., 2004). When animals are confronted with predation risk, they often exhibit purposive behavioural responses that may help them survive (Dalesman et al., 2006; Orr et al., 2007; Rigby and Jokela, 2000). Learning about predators and under what circumstances a predator threat may occur is an important adaptation and will produce alterations in behaviour during times of predation risk (Coolen et al., 2005). Additionally, the ability to make associations to the circumstances under which predation risks or actual events are experienced plays an important role in eliciting appropriate responses to similar circumstances in the future.

Stress, in the biological context, is defined as any circumstance or condition that disrupts the physiological or psychological homeostasis of an organism (Kim and Diamond, 2002). Stress itself is a physiological mechanism that helps organisms take on challenges. Animals are exposed to many stressful survival challenges in their natural environments. Our results show that simultaneously experiencing a predator threat with an odour changes the behavioural response and probably the neuronal response to that odour (Funk and Amir, 2000). Our results also allow us to hypothesize that the neurons responsible for LTM formation (e.g. RPeD1; Scheibenstock et al., 2002) will become responsive to an odour that has now gained emotional or ecological significance as a result of being experienced with a kairomone that signals danger. Experiments will be performed in the future to test this hypothesis.

Our results are consistent with the hypothesis that stimulus–stimulus learning is an important adaptive learning mechanism that helps animals decipher the meaning of important stimuli in their environment. From previous experiments, we know that exposure to CE results in short­-lasting changes in the electrophysiological properties of RPeD1, and that combining CE exposure with training is necessary for long-lasting changes to occur (Orr and Lukowiak, 2008; Orr et al., 2009b). The perception of CE primes the molecular mechanisms required for LTM formation (Parvez, et al., 2006) and we believe that after configural learning used here, CO does the same.

We believe that the context in which predator detection, or for that matter any significant stressful event, is processed in the nervous system will enable the organism to predictively respond to the context as if it were the stressor. Whether this is adaptive in all cases – for example, it may be maladaptive in post-traumatic stress disorder (PTSD) – remains to be determined. However, in the example we tested here, experiencing predator threat with a non-threatening odour (i.e. CO+CE) shows that the context in which a stressor is presented will signal a response that is phenotypically similar to that following presentation of the stressor alone. Further investigation is required to determine whether the causal neuronal mechanisms underlying enhanced memory formation are similar to those demonstrated when the animal is trained with the CE stressor alone.

In summary, predator-experienced W-strain Lymnaea associate the chemosensory CO with the scent of a predator (CE) by just being simultaneously exposed to the two stimuli. This simultaneous exposure results in CO acquiring a new ability to elicit anti-predator behaviours. Our results show that configural learning and memory formation occur in Lymnaea, indicating that the snails not only possess the ability to consider the individual components of stimuli (i.e. CO and CE) but also can form a relationship between the two stimuli. Additionally, the two stimuli are treated by the snail as different from the simple sum of the stimuli alone.