Anim Cogn (2008) 11:651–659 DOI 10.1007/s10071-008-0155-2

ORIGINAL PAPER

Sign- and goal-tracking in Atlantic cod (Gadus morhua) Jonatan Nilsson · Tore S. Kristiansen · Jan Erik Fosseidengen · Anders Fernö · Ruud van den Bos

Received: 14 September 2007 / Revised: 16 April 2008 / Accepted: 18 April 2008 / Published online: 14 May 2008 © Springer-Verlag 2008

Abstract When animals associate a stimulus with food, they may either direct their response towards the stimulus (sign-tracking) or towards the food (goal-tracking). The direction of the conditioned response of cod was investigated to elucidate how cod read cue signals. Groups of cod were conditioned to associate a blinking light (conditioned stimulus, CS) with a food reward (unconditioned stimulus, US), with the CS and the US located at opposite sides of the tank. Two groups were trained in a delay conditioning procedure (CS = 60 s, interstimulus interval = 30 s) and two groups were trained in a trace conditioning procedure (CS = 12 s, trace interval = 20 s). The response pattern was similar for the delay- and trace-conditioned groups. The initial main response at the onset of the CS was approaching the blinking lights, i.e. sign-tracking. In the early trials, the Wsh did not gather in the feeding area before the arrival of food. In the later trials, the Wsh Wrst approached the blinking lights, but then moved across the tank and gathered below the feeder before the food arrived, i.e. sign-tracking followed by goal-tracking within each trial. These two responses are interpreted as reXecting two learning systems, i.e. one rapid, reXexive response directed at the signal (sign-tracking) and one slower, more Xexible response based on expectations about time and place for arrival of J. Nilsson (&) · T. S. Kristiansen · J. E. Fosseidengen Institute of Marine Research, 5392 Storebø, Norway e-mail: [email protected] A. Fernö Department of Biology, University of Bergen, P.O. Box 7800, 5020 Bergen, Norway R. van den Bos Department of Animals, Science and Society, Utrecht University, Yalelaan 2, 3584 CM Utrecht, The Netherlands

the food (goal-tracking). The ecological signiWcance of these two learning systems in cod is discussed. Keywords Response systems · Pavlovian conditioning · Cognition · Foraging · Fish · Learning

Introduction Interest in the learning capacities of Wsh has increased in recent years, and there is now a large literature on how Wsh can learn to deal with variable ecological challenges in connection with foraging, predator defence, orientation and mating (see Brown et al. 2006 for reviews). Although it is well documented that Wsh can learn to associate stimuli with biologically relevant events (Bull 1928; Overmier and Hollis 1990), it is less well known how Wsh comprehend such predictive stimuli and which characteristics they actually learn about them. Pavlov (1927) suggested that when an association between a conditioned stimulus (CS) and a rewarding unconditioned stimulus (US) is established, the “CS centre” and the “US centre” in the brain become linked, with the result that an activation of the CS centre leads to activation of the US centre. Hence, the CS is perceived as if it was the US and the CS becomes a substitute for the US. In fact, the conditioned response is often directed towards the CS and resembles the response elicited by the US, a phenomenon known as auto-shaping (Brown and Jenkins 1968) or sign-tracking (Hearst and Jenkins 1974). Brown and Jenkins (1968) trained pigeons (Columba livia) to associate a key light with access to grain, and in the course of training pigeons started to peck the lighted key even though pecking was not necessary for access to grain. The behaviour directed at the CS may in fact be determined by

123

652

the characteristics of the US (Jenkins and Moore 1973). Pigeons pecked the key light with open beaks and closed eyes (eating behaviour) when the US was grain, but with closed beaks and open eyes (drinking behaviour) when the US was water, i.e. they responded to the CS as if it was the US. Similar results have been found in other animals such as rats (Peterson et al. 1972), cuttleWsh (Purdy et al. 1999; Cole and Adamo 2005) and Wsh (Woodard and Bitterman 1974; Waxman and McCleave 1978). Even when the CS and the US are spatially separated the approach response is commonly directed at the CS (signtracking, Hearst and Jenkins 1974; Burns and Domjan 1996; Purdy et al. 1999). However, under certain conditions the conditioned response is directed at the US rather than the CS, a behaviour referred to as goal-tracking (Boakes 1977). In this case, the CS is believed to announce and bring about a cognitive expectation of the forthcoming US, but has no reinforcing properties in itself. The existence of the two diVerent orienting responses to the CS, i.e. stimulus substitution leading to sign-tracking (Hearst and Jenkins 1974) and expectations leading to goal-tracking (Boakes 1977) may be explained by the existence of two distinct learning systems, i.e. an associative system in which conditioned responses are reXexive and unconsciously elicited, and a cognitive system based on explicit expectations (Toates 1998; Lieberman 2000; see also Squire 1992; Clark et al. 2002). Interestingly, with diVerent temporal and/or spatial relationships the same CS and US may produce either sign-tracking or goal-tracking, or a switch from signtracking to goal-tracking within a trial (Brown et al. 1993; Peden et al. 1977; Holland 1980; Silva et al. 1992). The functional signiWcance of the diVerent learning systems has seldom been addressed. Sign-tracking studies have mainly focused on mammals and birds, and in particular pigeons. Fish have been little studied, and to the best of our knowledge there are no published studies of how Wsh respond during CS presentation when the CS is spatially separated from the site of the US. In a previous conditioning experiment (Nilsson et al. 2008), we demonstrated that Atlantic cod (Gadus morhua) could associate a CS (blinking lights) in the feeding area with a food US when the stimuli were temporarily separated for up to 2 min. The cod approached the CS/US area immediately at CS onset. Groups trained with a 60 or 120 s trace interval between the CS oVset and the US onset initially left the area during the trace interval, and returned on the arrival of food. However, in the later part of the experiment the 60 s trace groups remained crowded in the area throughout the trace interval, while the 120 s trace groups never did so. We interpreted the crowding in the CS/US area during the trace interval as reXecting that the Wsh had some expectation of the forthcoming US. If this was the case, Wsh went from a stage where they responded

123

Anim Cogn (2008) 11:651–659

reXexively and implicitly (approaching and leaving the CS) in the Wrst trials to a stage in which they responded according to expectation (stay at the CS waiting for food to arrive). With a 120 s trace interval, they did not seem to reach the expectation stage during the 64 trials of the experiment. In the present study, we conditioned groups of cod to the same CS and US as in our previous study, but the CS and the US were located on opposite sides of the experimental tank. We used groups rather than individual subjects, as social interactions permitting social learning represent a more usual situation for cod than social isolation (Methven et al. 2003; Anderson et al. 2007). In order to determine whether the pattern of sign versus goal-tracking diVered when the CS remained on throughout the interstimulus interval as against when there was a time gap between the CS oVset and US onset, we performed one delay and one trace procedure. According to the hypothesis that conditioning involves two response stages (Toates 1998; Lieberman 2000), i.e. reXex and expectation, the response should be primarily sign-tracking during the Wrst phase of learning, while goal-tracking should dominate in a later phase.

Methods Experimental set-up Indoor circular tanks of 3 m diameter and 1 m depth were Wlled with about 70 cm seawater. The water had a temperature of 8.2–8.8°C and O2 saturation above 90%, and the Xow was 75 l min¡1. The illumination period was 24 h with a mean intensity of 1.14
E (approximately 60 Lux) midwater. A 50 cm ring of tubing (US ring) was placed below an automatic feeder at one side of the tank, in order to make the site of food release distinct. On the opposite side of the tank a ring of similar size, containing 40 light bulbs of 2.4 W (light ring, total eVect 96 W) was positioned. A video camera covering the whole tank hung about 2.5 m above the water. The set-up is illustrated in Fig. 1. Subject Wsh The cod were hatched from naturally fertilized eggs from wild-caught broodstock from a mixed population of coastal cod (Dahle et al. 2006). After hatching in March 2004 the larvae were reared semi-intensively in Xoating bags on natural zooplankton, and were later weaned to dry food. In November the same year the Wsh were transferred to a sea-cage and fed marine dry food. In June 2005, 60 individuals were transferred from the sea-cage to the experimental tanks. For practical reasons, the group were initially divided over two tanks with 30 individuals in each tank. Seven days later the Wsh were assigned to four diVerent groups of 15 individuals each,

Anim Cogn (2008) 11:651–659

653

onset of the US) = 30 s]. For the other two groups the CS had a duration of 12 s, and was temporally separated from the US with a 20 s trace interval (ISI = 32 s). In both the procedures the Wsh received a total of 56 trials. Each trial was recorded from 3.5 min before CS onset to 3 min after feeding and stored on DVD for later analysis. Analysis

Fig. 1 The experimental tank. Dotted lines indicate the division of the screen image during analysis

with seven or eight Wsh from each of the original groups transferred to each of four tanks. The Wsh were then allowed to acclimate for at least 2 days before the experiments started. The light ring was installed but not switched on. After the end of the experiment the Wsh were killed by an overdose of the anaesthetic benzocaine. Benzocaine blocks the signal transmission in all nerves (Ross and Ross 1999), and the Wsh lost equilibrium and stopped respiration within a minute. In addition to measures of length (38.9 § 2.9 cm) and weight (743.69 § 178.5 g, mean § SD, n = 60), the Wsh were examined for sex and maturity. Forty individuals were females and 20 males, with 4 or 6 males in each tank. All females were immature, while nine males had developing testes. No Wsh were in spawning condition. Procedure The CS was a series of light blinks (about 0.8 s on and 2 s oV) from the ring. We have previously shown that light blinks are a neutral stimulus for cod, which is not approached unless it is paired with a reward (Nilsson et al. 2008). The unconditioned stimulus (US) consisted of 7 mm sinking marine Wsh pellets (Skretting, Europa). The subjects were not fed at other times or with other food types than the US pellets during the experiment. In order to avoid satiation and hence loss of learning motivation, the amount of food given per day was 0.5% of the estimated biomass, or about 2/3 of the satiation ratio, corresponding to an average of 33 pellets per trial. The Wsh were fed at 4-h intervals (intertrial interval = 4 h) throughout the 24-h cycle to avoid that hunger level or other internal factors diVered between trials. Each feeding session lasted 1 min. Two of the groups were trained in a delay procedure, with the duration of the CS being 60 s, of which 30 s overlapped with the US [interstimulus interval (ISI, i.e. the time from onset of the CS to

The image of the tank on the video screen was divided into four equal 90º sectors, with the feeder and the US ring in the centre of one sector (US sector) and the light ring in the centre of the opposite sector (CS sector, Fig. 1). The number of Wsh in the CS and US sectors were recorded on frozen images 1 s before (T¡1, the baseline distribution of the Wsh), 10 s after (T10) and 30 s after (T30, immediately before food arrival) CS onset in all trials. An individual was considered to be in the sector in which its snout was located. A more detailed analysis was carried out in 12 early trials (trials 10– 21) and in 12 late trials (trials 45–56), in order to investigate whether the behaviour of Wsh diVered when they had recently established the association between the CS and the US and when they were more experienced. Here, the number of Wsh in the CS and US sectors was recorded more frequently in order to obtain a higher resolution of the response pattern. Statistics A generalized linear model (McCullagh and Nelder 1989) was used to test whether the time from onset of the CS (T¡1, T10 and T30) had an eVect on the number of Wsh in the CS and US sectors. To compensate for overdispersion a quasipoisson distribution was assumed (Myers et al. 2002). The predictor variable was set as factorial, with N = 36 (3 groups of 12 values) for all tests. All tests were within groups. The level of signiWcance was set at 0.05. In post hoc t tests, T10 and T30 were compared with T¡1, and hence the Bonferroni corrected level of signiWcance was 0.025.

Results Delay procedure When the cod had acquired response to the CS, the immediate reaction to the CS was an approach towards the light ring at CS onset, and the number of Wsh in the CS sector at T10 was higher than at T¡1 levels in most trials (Fig. 2). The number of Wsh in the CS sector was higher at T10 than at T¡1 in six subsequent trials (which has a probability of <0.05 occurring randomly and hence indicating learning) from trials 4 and 6 onwards for replicates 1 and 2, respectively. At T10, the number of Wsh in the US sector was generally lower than T¡1 lev-

123

654

Anim Cogn (2008) 11:651–659

Detailed analysis

there was a trend in replicate 2 [F(2, 33) = 2.602, P = 0.089] (Fig. 4a). Post hoc t tests revealed that the number of Wsh in the CS sector was higher than T¡1 levels at T10 in both the groups, and at T30 in one group (Table 1). In replicate 1, the number of Wsh in the US sector was lower than T¡1 levels at T10 but not at T30 (Table 1).

Early trials (10–21)

Late trials (45–56)

Time from onset of the CS had a signiWcant eVect on the number of Wsh in the CS sector in both the groups [replicate 1: F(2, 33) = 15.085, P < 0.001; replicate 2: F(2, 33) = 6.731, P = 0.0035], and on the number of Wsh in the US section in replicate 1 [F(2, 33) = 5.052, P = 0.012], while

There was an eVect of time from onset of the CS on the number of Wsh in the CS sector in replicate 2 [F(2, 33) = 17.725, P < 0.001], with more Wsh at T10 and fewer Wsh at T30 than at T¡1 (Table 1, Fig. 4a). In replicate 1, there was no signiWcant eVect [F(2, 33) = 2.476, P = 0.10]. However,

els throughout the experiment (Fig. 2). Between T10 and T30, i.e. immediately before food delivery, the number of Wsh fell in the CS sector and rose in the US sector. This was especially clear from around trial 20 onwards (Fig. 3).

Fig. 2 Change in the number of Wsh in the CS sector (Wlled circles) and US sector (open circles) between 1 s before and 10 s after CS onset for all trials in the delay (upper) and trace (lower) procedures. A positive number on the y-axis means an increased number of Wsh from T¡1 to T10, a negative number a decreased number in the respective sector (CS or US). Left and right Wgures are replicate groups 1 and 2, respectively. In the latter half of the experiment, the delay replicate 1 group tended to initially gather in the CS sector and then to move within 10 s towards the US sector, and hence the CS and US points intertwine around trial 30

Fig. 3 Change in the number of Wsh in the CS sector (Wlled circles) and US sector (open circles) between 10 s after and 30 s after CS onset, i.e. immediately before feeding, for all trials in the delay (upper) and trace (lower) procedures. A positive number on the y-axis means an increased number of Wsh from T10 to T30, a negative number a decreased number of Wsh in the respective sector (CS or US). Left and right Wgures are replicate groups 1 and 2, respectively

123

Anim Cogn (2008) 11:651–659

655

Fig. 4 Mean number of Wsh in the CS sector (Wlled circles) and US sector (open circles) before CS, during the ISI and during feeding in the early trials (trials 10–21) and late trials (trials 45– 56) of the a delay procedure, b trace procedure. Whiskers indicate standard error of mean. Cross-hatch indicates CS (low left–high right) and US (high left–low right) periods. Left and right Wgures are replicate groups 1 and 2, respectively

this group left the CS sector within 5 s (see Fig. 4a, low left panel), and when T10 was replaced with T5 in the model, there was a highly signiWcant eVect of time from onset of the CS on the number of Wsh in the CS sector [F(2, 33) = 13.765, P < 0.001], with more Wsh in the CS sector at T5 than at T¡1 (t = 3.376, P = 0.0019). There was no diVerence between T¡1 and T30 (t = ¡1.620, P = 0.11). Time from onset of the CS also had an eVect on the number of Wsh in the US sector [replicate 1: F(2, 33) = 18.764, P < 0.001; replicate 2: F(2, 33) = 41.320, P < 0.001], and opposite to the early trials there were more Wsh at T30 than at T¡1 in both the groups (Table 1, Fig. 4a).

Trace procedure As in the delay procedure, when the response to the CS had been acquired the immediate reaction to the CS was an approach towards the light ring at CS onset, and the number of Wsh in the CS sector at T10 was higher and the number of Wsh in the US sector lower than T¡1 levels in most of the trials of the experiment (Fig. 2). The number of Wsh in the CS sector was higher at T10 than at T¡1 in six subsequent from trials 14 and 15 for replicate 1 and 2, respectively. At T30, i.e. shortly before food arrival, the number of Wsh had fallen in the CS sector and risen in the US sector in most

123

656

Anim Cogn (2008) 11:651–659

Table 1 Results of the post hoc t tests from the general linear model analysis of the eVect of time from CS onset on the number of Wsh in the CS sector and the US sector. Asterisk indicates signiWcance at the Bonferroni corrected level of signiWcance (adjusted
= 0.025) Procedure (replicate)

Trials

CS sector H0:T10 = T¡1

US sector H0:T30 = T¡1

H0:T10 = T¡1

H0:T30 = T¡1 t = ¡0.832, P = 0.411

Delay (1)

10–21

t = 5.294, P < 0.001*

t = 3.440, P = 0.0016*

t = ¡2.911, P = 0.0064*

Delay (2)

10–21

t = 3.515, P = 0.0013*

t = 1.277, P = 0.210

No main eVect of time

Delay (1)

45–56

No main eVect of timea

t = 1.445, P = 0.158

t = 5.231, P < 0.001*

Delay (2)

45–56

t = 2.677, P = 0.012*

t = ¡2.915, P = 0.0063*

t = ¡2.493, P = 0.018*

t = 5.519, P < 0.001*

Trace (1)

10–21

t = 4.198, P < 0.001*

t t 0, P t 1

t = ¡4.297, P < 0.001*

t = ¡0.541, P = 0.592

Trace (2)

10–21

t = 2.638, P = 0.013*

t t 0, P t 1

t = ¡2.726, P = 0.010*

t = 0.107, P = 0.915

Trace (1)

45–56

t = 9.286, P < 0.001*

t = ¡2.833, P = 0.0078*

t = ¡4.996, P < 0.001*

t = 4.918, P < 0.001*

Trace (2)

45–56

t = 3.594, P = 0.0011*

t = ¡3.826, P < 0.001*

t = ¡1.586, P = 0.122

t = 4.906, P < 0.001*

In trials 45–56 the delay replicate 1 group left the CS sector within 10 s, but number of Wsh in the CS sector was higher 5 s after the onset of the CS. See text for more details a

trials. As in the delay procedure, this was especially clear in the latter half of the experiment, i.e. from around trial 30 onwards (Fig. 3). Detailed analysis Early trials (10–21) Time from onset of the CS had an eVect on the number of Wsh in the CS sector [replicate 1: F(2, 33) = 15.303, P < 0.001; replicate 2: F(2, 33) = 4.766, P = 0.015], with more Wsh at T10 than T¡1, while T30 did not diVer from T¡1 (Table 1, Fig 4b). Similarly, time from onset of the CS had an eVect on the number of Wsh in the US sector [replicate 1: F(2, 33) = 11.247, P < 0.001, replicate 2: F(2, 33) = 6.983, P = 0.0030], with fewer Wsh at T10 than at T¡1, while T30 did not diVer from T¡1 (Table 1, Fig. 4b). Late trials (45–56) As in the early trials, time from onset of the CS had an eVect on the number of Wsh in the CS sector [replicate 1: F(2, 33) = 115.010, P < 0.001, replicate 2: F(2, 33) = 30.201, P < 0.001], with more Wsh at T10 than at T¡1,while there were fewer Wsh at T30 than at T¡1 (Table 1, Fig. 4b). Time from onset of the CS also had an eVect on the number of Wsh in the US sector [replicate 1: F(2, 33) = 54.504, P < 0.001, replicate 2: F(2, 33) = 29.928, P < 0.001]. Opposite to the early trials, there were more Wsh in the US sector at T30 than at T¡1 (Table 1, Fig. 4b). Group behaviour The behaviour of other Wsh seemed to inXuence whether individuals tracked the CS or the US. Although the main conditioned response in all groups was always an approach to the light ring, individuals occasionally approached the

123

feeding area at CS onset. For replicates 1 and 2 of the delay groups, one to Wve individuals approached the feeding area in the course of the Wrst two light blinks without a preceding approach towards the light ring in 2 and 3 of 12 of the early trials (trial 10–21), and in Wve and four of the late trials (45–56). For the replicates 1 and 2 of the trace groups, individuals initially approached the feeding area in 4 and 1 of the early trials, and 7 and 7 of the late trials. With few exceptions, Wsh that initially approached the feeding area left this area within a few seconds and joined the main group at the light ring. Individuals were also frequently observed to swim back and forth between the CS and the feeding area during the ISI.

Discussion The delay-conditioned groups displayed a reliable response, i.e. an increase in number of Wsh in the CS sector at CS onset in six subsequent trials, after somewhat fewer trials than the trace-conditioned groups. The main response to the CS early in the experiment was to approach the CS (sign track) at CS onset, regardless of whether the cod were trained in a delay or trace procedure. However, at the end of the experiments in both procedures cod sign-tracked in the earlier part and goal-tracked in the later part of the interstimulus interval, and gathered in the US area prior to the arrival of food. In the delay procedure, in the early trials (10–21) the number of Wsh in the CS sector was signiWcantly higher 30 s after onset of the CS than before CS onset in one group with a similar pattern in the other group (see Fig. 4a, upper panel), while in the trace groups there was no diVerence. During this interval the CS was on for the delay groups but oV for the trace groups, so that the trace groups had no stimulus to attend to. This may explain why the trace groups left the CS area sooner. The overall response pattern was thus essentially the same in the two

Anim Cogn (2008) 11:651–659

procedures, with attention only to the CS early in the experiment to a switch in attention from the CS to the feeding area when the cod were more experienced. The learning process appeared to involve at least two stages. First, Wsh learned an association between the CS and the US, resulting in a sign-track response according to Pavlov’s (1927) stimulus substitution and classical stimulusresponse theory (Lieberman 2000). This is an implicit, procedural learning process and the response is regarded as reXexive (Lieberman 2000). However, during the course of the experiment the Wsh learnt where and when the food would be delivered, and eventually gathered below the feeder before food arrival, i.e. they goal-tracked. This behaviour does not Wt with the stimulus substitution hypothesis, but rather with the cognitive learning hypothesis, i.e. the response is based on explicit expectations about the US (Lieberman 2000). It has been suggested that responses to stimuli are usually guided by a combination of stimulus-response and cognitive processes, and which of these processes dominates the response may change with experience (Toates 1998, 2006). The results of our study support this hypothesis. In our previous study of associative learning in cod (Nilsson et al. 2008), we found that with trace intervals of 60 s Wsh left the feeding area (the location of both the CS and the US) during the trace interval early in training. However, in the later training sessions, i.e. after 20–35 trials, the Wsh remained crowded into the feeding area throughout the trace interval, a behaviour that we interpreted as an expectation of the forthcoming food reward. The results of the present study suggest that the approach to the CS/US area in our previous study represented sign-tracking, and crowding during the trace interval goal-tracking. One might expect that when the cod had learnt the temporal and spatial relationship between the CS and the US, i.e. that the CS predicted that food would soon be delivered at the other side of the tank, they would directly approach the US rather than the CS area at CS onset. Although individual Wsh occasionally did so (see below), most Wsh approached the CS Wrst, and thereafter moved to the US area. There are at least two not mutually exclusive explanations for this. First, reXexive stimulus-response processes and cognitive processes may operate together (Toates 1998). Stimulus-response processes are faster than cognitive processes (Toates 1998), and the immediate response at CS onset, namely sign-tracking, may have been dominated by stimulus-response processes, with the slower cognitive processes becoming dominant seconds later. Secondly, after the Wrst occasion of sign-tracking further learning will to some extent be inXuenced by operant learning (Brown and Jenkins 1968). As behaviour directed at the CS was always followed by the US, subjects may have learnt that approaching the CS was rewarded.

657

The behaviour of individual cod appeared to be strongly inXuenced by the behaviour of the rest of the group. On some occasions individuals approached the US at CS onset without a preceding approach to the CS. In the majority of such occasions, these goal-tracking individuals left the US area or terminated their approach within a few seconds and joined the main group in the CS area or swam back and forth between the CS and US areas, eventually approaching the US area together with the rest of the group. These observations suggest that at least some individuals are able to dissociate the CS from the US immediately at CS onset. In individually trained rats (Rattus norvegicus), Flagel et al. (2007) found that within the same delay conditioning paradigm, some individuals were sign-trackers while others were goal-trackers. It is possible that, in fact, such individual diVerences were also present in cod, but that individual decisions were overridden by social decisions. Even when an individual made a “correct decision”, it may choose a suboptimal behaviour, i.e. approaching the CS, when the rest of the group did so. Social animals such as shoaling Wsh must make trade-oVs between the costs and beneWts of behaving individually or following the group (Pitcher and Parrish 1993), and the cost of deviant behaviour can be high. Individual decision overridden by social constraints can be released in the absence of “leader” individuals, suggesting that in group situations Wsh may not make use of all the knowledge they have gained (Brown and Laland 2002). Laland and Williams (1998) demonstrated that in the guppy, Poecilia reticulata, maladaptive information can be acquired through social learning and may inhibit learning of optimal behaviour. To always approach a CS that is spatially separated from the US, even when expecting the US to arrive elsewhere, may seem maladaptive. However, spatial separation of the cue of a forthcoming food reward and the location of the food reward hardly represents a common situation in nature. Rather, approaching the cue makes responses to cues more rapid, increasing capture success rates. For instance, sign-tracking may be advantageous for cruising predators, such as cod, whose foraging behaviour involves orientation and approach to, and eventually chase of, the prey (Purdy et al. 1999). Such behaviour can be used by cod hunting for free-swimming Wsh (Steingrund and Fernö 1997). Here, the Wrst perception of stimuli emitted by the prey may be the cue of the potential “dinner”, i.e. “CS”, while the capture and ingestion of the prey is the reward or “US”. Another situation in which sign-tracking may be beneWcial is foraging on prey that are sheltering under recognizable structures such as stones and vegetation that may be associated with food and act as a CS (Purdy et al. 1999), as when cod forage on crabs or bottom-dwelling Wsh [see Hollis (1982) for a further discussion of biological advantages of signal-centred response actions].

123

658

Although procedural actions directed at the predictive cue, the “CS”, have the advantage of being rapid and eVective when the cue-outcome, or “CS-US” relationship, is invariable, they suVer from the disadvantage of low Xexibility. In changing environments, cognitively guided Xexibility in the response to cues may be highly beneWcial (Toates 1995, 1998), in spite of being slower and more costly due to the involvement of specialized processes and brain structures such as the hippocampus (mammals and birds, Eichenbaum et al. 1992) or lateral pallium (teleosts, Broglio et al. 2005). This study demonstrates that cod, which forage on a wide range of prey (Nordeide and Fosså 1992; Fjøsne and Gjøsæter 1996) and may switch between main prey types as prey composition changes (Hanson and Chouinard 2002), enjoy the rapid sign-track response but also have the ability to produce more Xexible responses, which should facilitate foraging on a variable menu. Acknowledgments We thank Victoria Braithwaite and three anonymous referees for their comments on earlier versions of the manuscript, and Einar Heegaard for advices on the statistical analysis. This study was funded by the Research Council of Norway, and complies with Norwegian regulations on animal experimentation.

References Anderson JL, Laurel BJ, Brown JA (2007) Diel changes in behaviour and habitat use by age-0 Atlantic cod (Gadus morhua L.) in the laboratory and Weld. J Exp Mar Biol Ecol 351:267–275. doi:10.1016/j.jembe.2007.07.001 Boakes RA (1977) Performance on learning to associate a stimulus with positive reinforcement. In: Davis H, Hurwitz HM (eds) Operant-Pavlovian interactions. Erlbaum, Hillsdale, pp 67–97 Broglio C, Gomez A, Duran E, Ocana FM, Jimenez-Moya F, Rodriguez F, Salas C (2005) Hallmarks of a common forebrain vertebrate plan: specialized pallial areas for spatial, temporal and emotional memory in actinopterygian Wsh. Brain Res Bull 66:277–281. doi:10.1016/j.brainresbull.2005.03.021 Brown BL, Hemmes NS, Devaca SC, Pagano C (1993) Sign and goal tracking during delay and trace autoshaping in pigeons. Anim Learn Behav 21:360–368 Brown PL, Jenkins HM (1968) Auto-shaping of pigeon’s key-peck. J Exp Anal Behav 11:1–8. doi:10.1901/jeab.1968.11-1 Brown C, Laland KN (2002) Social learning of a novel avoidance task in the guppy: conformity and social release. Anim Behav 64:41– 47. doi:10.1006/anbe.2002.3021 Brown C, Laland KN, Krause J (2006) Fish cognition and behavior. Blackwell, Oxford Bull HO (1928) Studies on conditioned responses in Wshes. J Mar Biol Assoc UK 15:485–533 Burns M, Domjan M (1996) Sign tracking versus goal tracking in the sexual conditioning of male Japanese quail (Coturnix japonica). J Exp Psychol Anim Behav Process 22:297–306 Clark RE, Manns JR, Squire LR (2002) Classical conditioning, awareness, and brain systems. Trends Cogn Sci 6:524–531. doi:10.1016/S1364-6613(02) 02041-7 Cole PD, Adamo SA (2005) CuttleWsh (Sepia oYcinalis: Cephalopoda) hunting behavior and associative learning. Anim Cogn 8:27–30. doi:10.1007/s10071-004-0228-9

123

Anim Cogn (2008) 11:651–659 Dahle G, Jørstad KE, Rusaas HE, Otterå H (2006) Genetic characteristics of broodstock collected from four Norwegian coastal cod (Gadus morhua) populations. ICES J Mar Sci 63:209–215. doi:10.1016/j.icesjms.2005.10.015 Eichenbaum H, Otto T, Cohen NJ (1992) The hippocampus—what does it do? Behav Neural Biol 57:2–36. doi:10.1016/01631047(92) 90724-I Fjøsne K, Gjøsæter J (1996) Dietary composition and the potential of food competition between 0-group cod (Gadus morhua L) and some other Wsh species in the littoral zone. ICES J Mar Sci 53:757–770. doi:10.1006/jmsc.1996.0097 Flagel SB, Watson SJ, Robinson TE, Akil H (2007) Individual diVerences in the propensity to approach signals vs goals promote diVerent adaptations in the dopamine system of rats. Psychopharmacology 191:599–607. doi:10.1007/s00213-006-0535-8 Hanson JM, Chouinard GA (2002) Diet of Atlantic cod in the southern Gulf of St Lawrence as an index of ecosystem change, 1959– 2000. J Fish Biol 60:902–922. doi:10.1111/j.10958649.2002.tb02417.x Hearst E, Jenkins HM (1974) Sign-tracking: the stimulus-reinforcer relation and directed actions. Psychonomic Society, Austin Holland PC (1980) CS-US interval as a determinant of the form of Pavlovian appetitive conditioned-responses. J Exp Psychol Anim Behav Process 6:155–174 Hollis KL (1982) Pavlovian conditioning of signal-centered action patterns and autonomic behavior—a biological analysis of function. Adv Stud Behav 12:1–64 Jenkins HM, Moore BR (1973) Form of auto-shaped response with food or water reinforcers. J Exp Anal Behav 20:163–181. doi:10.1901/jeab.1973.20-163 Laland KN, Williams K (1998) Social transmission of maladaptive information in the guppy. Behav Ecol 9:493–499 Lieberman DA (2000) Learning: behavior and cognition. Wadsworth, Belmont McCullagh P, Nelder JA (1989) Generalized linear models. Chapman & Hall, London Methven DA, Schnieder DC, Rose GA (2003) Spatial pattern and patchiness during ontogeny: post-settled Gadus morhua from coastal Newfoundland. ICES J Mar Sci 60:38–51. doi:10.1006/ jmsc.2002.1315 Myers RH, Montgomery DC, Vining GG (2002) Generalized linear models: with applications in engineering and the sciences. John Wiley & Sons, New York Nilsson J, Kristiansen TS, Fosseidengen JE, Fernö A, van den Bos R (2008) Learning in cod (Gadus morhua): long trace interval retention. Anim Cogn 11:215–222. doi:10.1007/s10071-007-0103-6 Nordeide JT, Fosså JH (1992) Diet overlap between 2 subsequent yearclasses of juvenile coastal cod (Gadus morhua) and wild and reared cod. Sarsia 77:111–117 Overmier J, Hollis K (1990) Fish in the think tank: learning, memory, and integrated behavior. In: Kesner R, Olton D (eds) Neurobiology of comparative cognition. Lawrence Erlbaum, Hillsdale, pp 205–236 Pavlov IP (1927) Conditioned reXexes. Oxford University Press, Oxford Peden BF, Browne MP, Hearst E (1977) Persistent approached to a signal for food despite food omission for approaching. J Exp Psychol Anim Behav Process 3:377–399 Peterson GB, Frommer GP, Ackil JE, Hearst ES (1972) Conditioned approach and contact behavior toward signals for food or brainstimulation reinforcement. Science 177:1009–1011. doi:10.1126/ science.177.4053.1009 Pitcher TJ, Parrish JK (1993) Functions of shoaling behaviour in teleosts. In: Pitcher TJ (ed) Behaviour of teleost Wshes. Chapman & Hall, London, pp 363–439

Anim Cogn (2008) 11:651–659 Purdy JE, Roberts AC, Garcia CA (1999) Sign tracking in cuttleWsh (Sepia oYcinalis). J Comp Psychol 113:443–449 Ross LG, Ross B (1999) Anaesthetic and sedative techniques for aquatic animals. Blackwell, London Silva FJ, Silva KM, Pear JJ (1992) Sign-tracking versus goal-tracking—eVects of conditioned-stimulus-to-unconditioned-stimulus distance. J Exp Anal Behav 57:17–31. doi:10.1901/jeab.1992.5717 Squire LR (1992) Memory and the hippocampus: a synthesis from Wndings with rats, monkeys, and humans. Psychol Rev 99:195– 231 Steingrund P, Fernö A (1997) Feeding behaviour of reared and wild cod and the eVect of learning: two strategics of feeding on the two-spotted goby. J Fish Biol 51:334–348. doi:10.1111/j.10958649.1997.tb01670.x

659 Toates F (1995) Cognition and evolution—an organization of action perspective. Behav Process 35:239–250. doi:10.1016/03766357(95)00058-5 Toates F (1998) The interaction of cognitive and stimulus-response processes in the control of behaviour. Neurosci Biobehav Rev 22:59–83. doi:10.1016/S0149-7634(97)00022-5 Toates F (2006) A model of the hierarchy of behaviour, cognition, and consciousness. Conscious Cogn 15:75–118. doi:10.1016/j.concog.2005.04.008 Waxman HM, McCleave JD (1978) Auto-shaping in archer Wsh (Toxotes chatareus). Behav Biol 22:541–544. doi:10.1016/S00916773(78) 92727-X Woodard WT, Bitterman ME (1974) Autoshaping in the goldWsh. Behav Res Meth Instr 6:409–410

123