Royal Society Publishing

Are fast explorers slow reactors? Linking personality type and anti-predator behaviour

Katherine A. Jones, Jean-Guy J. Godin


Response delays to predator attack may be adaptive, suggesting that latency to respond does not always reflect predator detection time, but can be a decision based on starvation–predation risk trade-offs. In birds, some anti-predator behaviours have been shown to be correlated with personality traits such as activity level and exploration. Here, we tested for a correlation between exploration behaviour and response latency time to a simulated fish predator attack in a fish species, juvenile convict cichlids (Amatitlania nigrofasciata). Individual focal fish were subjected to a standardized attack by a robotic fish predator while foraging, and separately given two repeated trials of exploration of a novel environment. We found a strong positive correlation between exploration and time taken to respond to the predator model. Fish that were fast to explore the novel environment were slower to respond to the predator. Our study therefore provides some of the first experimental evidence for a link between exploration behaviour and predator-escape behaviour. We suggest that different behavioural types may differ in how they partition their attention between foraging and anti-predator vigilance.

1. Introduction

Avoidance of predation is an important determinant of fitness in many animals (Godin 1997; Ruxton et al. 2004). Individuals must trade-off the risk of predation against the benefits of foraging (Lima & Dill 1990). This ‘starvation–predation risk trade-off’ is central to many foraging decisions individuals make, such as which patch to forage in (Gilliam & Fraser 1987; Butler et al. 2005), when to forage (Dall & Witter 1998; Macleod et al. 2005) and with whom to forage (Giraldeau & Caraco 2000; Krause & Ruxton 2002).

Recent work (e.g. Cresswell et al. 2000, 2009; Fernandez-Juricic et al. 2002; Quinn & Cresswell 2005) has suggested that delays in responding to a predatory attack can actually be adaptive. Individuals must trade-off an immediate response to a potential threat against potential lost-opportunity foraging costs of taking flight (economic hypothesis; Ydenberg & Dill 1986; Quinn & Cresswell 2005). Experiments quantifying the response time of visually guided animals to predator models have shown that there is significant interindividual variation in response times (Krause & Godin 1996; Stankowich & Coss 2006; Jones et al. 2009). However, the mechanisms underlying such variation are not fully understood.

Variation in response delays may result from differences in inherent ability to detect predators (perceptual limit hypothesis; Quinn & Cresswell 2005) or differences in vigilance, with animals that devote more time to vigilance (e.g. via more frequent and/or longer vigilance bouts) being faster to detect an approaching predator (Cresswell et al. 2003). Individuals can vary their vigilance level by allocating their brain's ‘attention’ between searching for and handling food items and searching for predators (Godin & Smith 1988; Clark & Dukas 2003; Fernandez-Juricic et al. 2004).

When behavioural ecologists talk of attention, they refer to the devotion of the brain's computing power to completing a particular task (Dukas & Kamil 2000). Attention has been shown to be limited (Dukas & Kamil 2000, 2001). Therefore, it is hypothesized that the relative division of attention between foraging and anti-predator vigilance will vary both within and between individuals. Birds and fishes are less able to detect approaching predators when attending to a difficult foraging task, most probably owing to their attention being focused on obtaining food (Milinski 1984; Krause & Godin 1996; Kaby & Lind 2003), which can translate into a direct fitness cost owing to an increased chance of capture by a predator (Godin & Smith 1988; Krause & Godin 1996).

Between-individual differences in either inherent ability to detect predators or the relative division of attention towards foraging versus anti-predator vigilance may not be simply ‘noise around a maximally adapted mean’ (Bell 2007). Increasingly, behavioural ecologists are viewing differences in behaviour among individuals in a population as adaptive and representing different behavioural strategies (reviewed in Dall et al. 2004; Bell 2007; Réale et al. 2007; Smith & Blumstein 2008). Depending on environmental and social conditions, different strategies (each associated with different costs and benefits) will have different payoffs. For example, a recent experiment on rainbow trout (Oncorhynchus mykiss) showed that larger, bolder, fast-growing phenotypes were more likely to be captured by fishermen (Biro & Post 2008). Predation has also been proposed as one mechanism underlying the maintenance of behavioural syndromes by selecting against certain combinations of traits (e.g. being bold and non-exploratory, or shy and exploratory; Stamps 2007), which is supported by experimental evidence on three-spined sticklebacks, Gasterosteus aculeatus (Bell & Sih 2007).

Early work on personality in animals has focused on laboratory paradigms, such as exploration, neophobia and risk-taking (reviewed in Sih et al. 2004; Réale et al. 2007). There is thus a need to investigate how these paradigms translate to more natural contexts such as foraging (Wilson & McLaughlin 2007), mate choice (Godin & Dugatkin 1996) and parental care (Budaev et al. 1999). In particular, there are few studies investigating the relationship between personality traits and anti-predator behaviour (reviewed in Réale et al. 2007), most likely because predation events are infrequently observed in nature compared with other behavioural phenomena. In birds, van Oers et al. (2004) found that great tits (Parus major), which were more willing to explore more of a novel environment, were also more likely to return to forage quicker after being startled. Quinn & Cresswell (2005) is one of the few studies to consider the effects of prey animal personality at the moment of a predatory attack; they reported that chaffinches (Fringilla coelebs), which were less active at the beginning of a trial, were more likely to freeze than flee in response to a hawk attack and took longer to resume activity following the attack. However, this study did not quantify exploration in a non-predatory context; that is, their anti-predator behaviour and personality measurements were conducted in the same, non-independent trials.

In the current study, we investigated experimentally whether individual juvenile convict cichlid fish (Amatitlania nigrofasciata) differ in their response latency time to an approaching predatory threat and whether such differences correlate with individual differences in behaviour in other contexts (i.e. exploration of a novel environment, and response to a novel object). We tested two plausible alternative hypotheses regarding the nature of this relationship: (i) that the relationship is a positive one; if risky behaviours such as exploration and foraging reduce (limited) available attention for other behaviours, then anti-predator vigilance might be compromised, such that good explorers would be expected to exhibit longer delays (latencies) in their responses to predatory threats than poorer explorers or (ii) that the relationship is a negative one: if exploratory behaviour is risky, then more exploratory individuals might be selected to compensate with increased anti-predator vigilance, such that good explorers would be expected to exhibit shorter delays in their responses to predatory threats than poorer explorers. Our study thus aimed to characterize the relationship between individual behavioural types or ‘personalities’ and anti-predation behaviour at the moment of a predatory attack. In doing so, we provide further explanation for the variation in response times to predatory attack observed in several species and a potential mechanism for the observed differential survival of personality types (Bell & Sih 2007; Smith & Blumstein 2008).

2. Material and methods

(a) Subjects

We used two- to three-month-old juvenile convict cichlids (mean ± s.d. = 19.5 ± 3.0 mm, standard body length) originating from a laboratory stock population derived from wild Costa-Rican adults as experimental subjects. Broods were maintained in 80 l stock aquaria containing aged filtered tap water (24–25°C), a gravel substratum and a plastic plant refuge.

Each individual test fish (n = 45) was exposed to a predator-attack test, two novel-object tests, and two novel-environment tests over an 11-day period (table 1). We used the novel-object and novel-environment tests to assess individual personality. To control for order of presentation, individual fish were exposed to the predator-attack test either before (n = 29) or after (n = 16) the aforementioned two sets of personality tests (table 1). Because the results were not dependent on the order of presentation of the predator-attack test (§3), we pooled the before–after data for statistical analyses.

View this table:
Table 1.

Experimental time schedule. Fish were put through these tests in batches of five (total fish tested = 45) between January and May 2008. Individual fish were exposed to the predator-attack test only once, either before (Day 0) or after (Day 11) the two sets of ‘personality’ tests.

(b) Predator-attack trials

Following Krause & Godin (1996), we standardized the threat of predation presented to individual test cichlids using a model (19 cm length, constructed to scale and painted realistically) of an approaching wolf cichlid (Parachromis dovii), a major predator of juvenile convict cichlids in nature (J.-G. J. Godin 2008, personal observations). The model was suspended in the water column (5 cm above the gravel substratum) at one end of the experimental tank (110 cm × 40 cm × 34 cm, L × W × H; water depth = 15 cm; figure 1a) by two pieces of monofilament nylon line attached to an overhead clear Plexiglas track. When not moving, the model was hidden from the test fish in a simulated refuge constructed of a piece of opaque plastic overhead, rocks and plastic plants.

Figure 1.

(a) Experimental apparatus for quantifying the anti-predator response of test fish to an approaching predator model. The fish predator model was hidden prior to its ‘attack’ in a dimly lit refuge of rocks and a plastic plant at one end of the experimental aquarium (for simplicity, not shown in the figure). The arrow denotes the direction of movement of the fish predator model during the ‘attack’ phase of the protocol. (b) Experimental apparatus for quantifying the exploratory behaviour of test fish in a novel environment.

At the opposite end of the experimental tank, we placed a small glass aquarium (30 cm × 12.5 cm × 20 cm, L × W × H; water depth = 15 cm) to house the test fish (figure 1a). One-half of the small tank was covered overhead and on the sides with opaque plastic and contained a plastic plant, providing a ‘refuge’ for the test fish. The open half of the housing tank contained a food ‘patch’ (figure 1a), which consisted of a glass microscope slide that was thinly coated with a 2 × 2 cm layer of gelatin, onto which crushed TetraMin cichlid pellets (Tetra Werke, Melle, Germany) were evenly sprinkled, to provide an excess of food, simulating foraging on algae-covered rocks in nature (J.-G. J. Godin 2008, personal observations). The slide was placed at an angle against the end wall of the tank facing the refuge. When feeding on this food patch, test fish were typically oriented horizontally or slightly obliquely downward and with their right side facing the direction of the predator model. Fish were given prior experience feeding on such glass slides in their home aquaria prior to testing. Test fish were deprived of food for 24 h prior to experimentation to standardize their hunger level.

For each predator-attack trial, we haphazardly dip-netted a test fish from a brood and placed it in the small housing tank within the experimental apparatus and allowed it 30 min to acclimatize. To control for capture bias, for a given trial, we captured five fish from a brood and then used a random number between 1 and 5 to determine which one of the individuals captured would be used as the test fish in the trial. Following the acclimatization period, the food patch (described above) was gently lowered into the housing tank (see figure 1a for position). Once the test fish had been foraging on the food patch for 30 s, the model was moved remotely towards the test fish at 15 cm s−1 (simulating an ‘attacking’ wolf cichlid) by activating a stepping motor with a computerized controller. Once activated, the predator model travelled forwards towards the test fish for 6 s (covering a distance of 90 cm), then waited at the end of its ‘attack’ just in front of the test fish housing tank for 5 s before slowly moving backwards over 45 s into its refuge. Test fish typically responded to the approaching predator model by stopping foraging and either freezing near the substratum, fleeing from the patch and freezing within the open zone of the housing tank or fleeing into the nearby refuge.

The behaviour of test fish was filmed with a digital HD video camcorder (Sony model HDR-HC7) mounted overhead. From the video tapes, we recorded (i) the time when the predator model first started moving and (ii) the time when the test fish first visibly reacted to the approaching model, as indicated by an abrupt change in behaviour (e.g. interruption of foraging, fleeing). We calculated the test fish's response latency time to the approaching predator model by subtracting (i) from (ii).

(c) Personality trials: novel-object test

Prior to a novel-object trial, the test fish was measured (standard body length) and transferred with two size-matched ‘companion’ conspecific fish (each marked with a small elastomer tag, Northwest Marine Technology Inc., Shaw Island, WA, USA) into a holding tank (40 cm × 20 cm × 25 cm, L × W × H). Companion fish were used to minimize stress to the focal fish (Barlow 1968), which commonly occur in social groups as juveniles. Fish were given two days to acclimatize to their new home tank before the first novel-object trial was conducted. The novel object consisted of a block of small green or brown Lego (no. 6162; LEGO Canada Inc., Richmond Hill, Canada) attached to a small metal sinker. The test fish was presented with the brown Lego in one trial and the green Lego in the other (repeated) trial, with the order of presentation and trial time randomized. A test fish was given 10 min to settle after setting up a video camcorder in front of the tank, following which the novel object was placed in the center of the tank using long forceps and fish behaviour recorded for 10 min. The novel object was removed at the end of the observation trial. From our analysis of the video tapes, we quantified the test fish's (i) latency to first approach the object, (ii) total number of approaches (to within 1 cm), and (iii) number of bites on the object. Aggressive interactions between the companion fish and the focal test fish were also recorded. One week after the first novel-object test trial (table 1), a second novel-object trial was conducted using the same protocol to ascertain repeatability of behaviour.

(d) Personality trials: exploration of novel-environment test

The day after the first and second novel-object test trials, we conducted novel-environment tests (table 1). The novel environment consisted of a tank (75 cm × 30 cm × 28 cm, L × W × H; water depth = 25 cm) divided into two compartments (a small holding compartment and a larger novel compartment/environment) with an opaque partition, equipped with a sliding door, and illuminated from above (figure 1b). The novel environment contained a gravel substratum and three plastic plants, which were either red or green (plant colour was alternated between the two repeated trials on each test fish, with the order of presentation randomized, to mitigate potential habituation to the environment between trials). The holding compartment contained the two tagged companion fish, in a weighted clear container, for the test fish to socially associate with. Test fish were individually transferred using a dip-net to the holding compartment of the tank and left undisturbed to acclimatize for 1 h. Following this period, the opaque sliding door of the compartment was raised remotely (using a string and pulley system) and the test fish was allowed up to 20 min to enter and swim freely in the open novel compartment. Once the test fish entered the novel environment, its behaviour was filmed for 10 min with a video camcorder 1 m from the front of the tank. A grid of 18 equal-sized squares (three rows of six) was taped to the back wall of the tank (figure 1b) to facilitate the recording of area explored by the test fish.

From our analysis of the video tapes, we quantified (i) latency of the test fish to leave the holding refuge area, (ii) proportion of trial time it spent in the refuge area, (iii) number of squares on the grid it traversed, and (iv) its activity level (number of darts forward and turns).

(e) Statistical analyses

All statistics were performed in R v. 2.6.2 (The R foundation for statistical computing,

(i) Personality measures

Behavioural measures from the two tests of personality were analysed separately as some of the variables were binary, therefore making it difficult to robustly apply a principal components analysis (Quinn & Keough 2002). We only used behavioural measures that were repeatable across trials and therefore represented a robust measure of individual exploration of a novel environment and response to novelty in a familiar environment.

To ascertain whether individual test fish were consistent in their behaviour (and therefore to ensure that our behavioural measures could be used reliably to classify individuals into different behavioural types), we calculated the repeatability of each behavioural measure across the two repeated novel-object and novel-environment trials. Only variables which were significantly repeatable (table 2) were used in subsequent analyses for comparison with fish response latency time to the approaching predator model.

View this table:
Table 2.

Repeatability estimates (and 95% confidence intervals) of fish behavioural measures obtained from novel-object and novel-environment tests. Estimates that are statistically significant are indicated in boldface. The actual p-values for the number of squares explored and chasing companion measures were 0.039 and 0.023, respectively.

Repeatability (R) was calculated by performing a general linear mixed model with individual fish identification (ID) inserted as a random factor in the model. From the resulting ANOVA results, the ratio of variance explained by fish ID to total variance represents a measure of repeatability of behaviour (following Nakagawa et al. 2007; Minderman et al. 2009). Confidence intervals around each repeatability estimate were calculated using a Fisher r-to-z transformation (Cohen et al. 2003), and z values were then divided by the standard error and compared against the z distribution. By way of comparison, our calculated R values were very similar to ones we alternatively obtained using the intraclass correlation coefficient method described in Lessells & Boag (1987).

(ii) Correlation between personality measures and response time to a fish predator model

The relationship between each repeatable personality measure and response latency time to the fish predator model was determined using the Pearson correlation test if both variables were continuous, Fisher's exact test if both were binary and linear models if one variable was continuous and the other binary. For linear models, we arbitrarily used the continuous variable as the dependent variable, but verified that the results were similar if the binary variable had been specified as the dependent variable (using a general linear model with a binomial link function); we therefore did not infer a priori the direction of the relationship. To control for potential confounding effects of brood, fish body size and order of predator-model presentation, we also constructed two mixed models (using the lmer function in lme4 v. 0.999375-31 package (Bates & Maechler 2009) in R v. 2.6.2) with brood specified as a random effect, standard length and personality measure specified as fixed effects and response time as the dependent variable. Separate models were constructed for each of the three personality measures because small samples sizes and missing data meant a very reduced sample size with a full model. p-values were calculated using the pvals.fnc function in the languageR package v. 0.955 (Baayen 2009).

3. Results

(a) Personality measures

For the novel-object test, two behavioural measures were repeatable (table 2); whether an individual bit the novel object or not (a response to novelty) and whether an individual chased either of the companion fish or not (a measure of aggression). For the novel-environment test, the number of grid squares explored (i.e. the amount of the environment explored in 10 min) was repeatable. Consequently, all subsequent analyses use the following personality measures: (i) biting/inspecting a novel object (binary two-level factor), (ii) chasing companion fish (binary two-level factor), and (iii) and number of grid squares explored (continuous covariate).

(b) Correlation between personality measures and response time to a fish predator model

There was no significant association between aggression (chase or not chase) and tendency to inspect a novel object (Fisher exact test, p = 0.72) nor between inspection of a novel object and exploration of a novel environment (F1,34 = 0.89, p = 0.35). However, there was a significant relationship between aggression and exploration of a novel environment (F1,31 = 9.61, p = 0.004). Fish that chased companion fish in the novel-object trial explored less of the novel environment (coefficient chase versus not chase = −4.64 squares).

There was a significant positive correlation between exploration of a novel environment during the first trial, where an individual fish entered the novel tank compartment (when this environment was most ‘novel’), and the response latency time to the simulated predator attack while foraging (Pearson's correlation, r22 = 0.52, p = 0.010, figure 2). Controlling for the potential confounding influences of brood origin (variance close to zero), size of individual test fish (p = 0.70) and order of predator-model presentation (p = 0.35) did not remove the significant positive relationship between exploration of the novel environment and response latency time to a predator attack while foraging (p = 0.0035, estimate of slope = 0.286). To remove aggression as a potential confound, we repeated the analysis on chasing fish only (i.e. removing the five non-chasing fish, and five fish for which there were no aggression data, from the dataset). A positive relationship of similar magnitude remained, although marginally non-significant, between exploration and response latency time to the fish predator model (r12 = 0.47, p = 0.089) using this reduced dataset, which seems reasonable given the smaller sample size (n = 14).

Figure 2.

Relationship between the exploratory behaviour of individual test fish in the novel-environment test and their response latency time to an approaching fish predator model while foraging. Data presented are raw values. The regression line was obtained from our statistical model, which controlled for order of presentation and brood.

There was no significant relationship between whether an individual bit the novel object or chased the companion fish in the novel-object trial and response latency time to the fish predator model (coefficient bite versus not bite = −0.92 s, p = 0.28, n = 26; coefficient chase versus not chase = −0.60 s, p = 0.59, n = 21, both models controlling for brood, order of presentation and fish standard length).

4. Discussion

The major novel finding of our study is that juvenile convict cichlids that were more exploratory in a novel environment were slower to react to a simulated fish predator attack. This salient result supports the first of our two alternative hypotheses under test, namely, that the level of exploratory behaviour of an individual is negatively correlated with its response time to an immediate predatory threat (in other words, fast explorers are slow reactors).

This observed relationship could be driven by two possible mechanisms. Firstly, consistent with the economic hypothesis of fleeing from predators (Ydenberg & Dill 1986; Fernandez-Juricic et al. 2002; Quinn & Cresswell 2005), bolder, more exploratory fish could devote more attention to foraging than to anti-predator vigilance, thereby delaying their response to a detected predator threat and reducing potential lost-opportunity foraging costs (Lima & Dill 1990). Devoting less attention to anti-predator vigilance may allow for more efficient foraging, but at the potential cost of increased risk of mortality to predation (Godin & Smith 1988; Krause & Godin 1996). Secondly, consistent with the perceptual limits hypothesis (Quinn & Cresswell 2005), fish differing in personality may be ‘hard-wired’ differently, having a different neurological capacity to detect predator attacks. In another study on convict cichlids (on adults rather than juveniles), Reddon & Hurd (2008) found that the extent of cerebral lateralization differed between aggressive and non-aggressive fish. Neurological architecture, such as cerebral lateralization, may also differ along other behavioural dimensions and has already been shown that it can influence vigilance behaviour in dark-eyed, juncos Junco hyemalis (Franklin & Lima 2001). Further research is required to elucidate potential relationships between predator detection, personality and brain structure.

Contrary to previous studies reporting that aggressive animals also tend to be more exploratory (Dingemanse & de Goede 2004; Bell & Sih 2007), we found the opposite in juvenile convict cichlids. Fish that chased conspecifics were less, not more, exploratory. However, this does not detract from our main result that fish that are more willing to leave a companion shoal and explore a novel environment are slower to respond to a predator attack, as the magnitude of this positive correlation remained relatively unchanged for those fish showing aggression in the novel-object trial.

Exploration of a novel environment by our experimental subjects was repeatable, although the large confidence intervals around the repeatability estimate suggest significant individual variation in behaviour, perhaps owing to plasticity in the amount of habituation to the experimental apparatus (Martin & Réale 2008). In repeated trials, an environment becomes less novel, and individuals may even habituate to novelty itself (Réale et al. 2007). Here, we attempted to mitigate habituation by altering the colour of landmarks (plastic plants) in the environment between repeated trials. Given our sample size and significant repeatability estimate for exploration, we are confident that the number of grid squares explored by the test fish is a reliable measure of an individual's exploratory behaviour. The observed positive correlation between exploration and response latency time to a predator attack suggests that individuals consistently differ in their propensity to take risks across two different ecological contexts (i.e. exploration of novel habitats and response to a predation threat).

One limitation of our study is that we were unable to measure repeatability of response-delay behaviour. Very few studies have demonstrated repeatability of response delays (notably Fuiman & Cowan 2003). This is most likely not due to behaviour being unrepeatable (as animal behaviour is commonly repeatable; Bell et al. 2009), but rather owing to behavioural habituation effects (described above) being particularly strong when associated with negative stimuli such as a predator attack (thus making true repeatability of the behaviour difficult to quantify). As we were interested in how animals may partition their attention between foraging and anti-predator vigilance, our experimental set-up required individuals to forage in order to be subjected to an attack. However, with larger sample sizes (subject to ethical considerations) and for certain behaviours (cf. Bell et al. 2009), we believe measurement of repeatability of response delays to be a fruitful area for future research.

Our study does not demonstrate cause and effect. A behavioural syndrome is by definition a ‘suite of correlated behaviors reflecting between-individual consistency in behavior across multiple (two or more) situations’ (Sih et al. 2004). Therefore, it can be difficult to ascertain whether observed correlations are due to direct linkage of traits or due to so-called ‘third variable effects’. A more robust test of the linkage between exploratory and predator-escape behavioural traits would be to selectively breed fish lines based on their exploratory behaviour as juveniles and, after several generations, test whether fast and slow explorer groups differ in their response times to a predator attack. Speed of exploration here is taken to represent amount of a novel environment explored in a standardized time period, and not speed of movement per se, following van Oers et al. (2004). Working with the great tit, the latter researchers selectively bred lines of fast and slow explorers and then tested for differences in other unselected traits. They found that selectively bred fast explorers were more willing to take risks (such as returning to forage after being ‘startled’) than slow explorers. Nonetheless, our study does provide unique and novel evidence for a link between exploratory behaviour and predator-escape behaviour, which ultimately could have fitness consequences.

Recent work in behavioural ecology has shown a shift of emphasis from ‘optimality’ models to understanding why individuals within populations consistently behave differently and what maintains this variance in behaviour (reviewed in Dall et al. 2004; Bell 2007; Réale et al. 2007). Although we have a relatively good understanding of the various trade-offs between foraging and predation an individual animal can face, we are yet to fully explain why different individuals and different species adopt different anti-predator strategies. We suggest that being slow to flee from a predator is not always maladaptive but may simply represent an alternative behavioural strategy, operating much like alternative reproductive strategies and tactics (Gross 1996). In particular, the proximate mechanisms underlying variance in responses to predator attack (such as hormonal pathways and neurological pathways) need greater understanding. Further work on the fitness consequences of fleeing from predators is required to test whether the benefits of faster exploration outweigh the costs of being a slow reactor.


Our study was approved by the institutional Animal Care Committee at Carleton University (protocol B07-12), and adheres to the guidelines for the care and use of animals in research of the Canadian Council on Animal Care and of the Animal Behaviour Society.

We thank Stacey Lee-Jenkins for breeding convict cichlids in our laboratory, Shinichi Nakagawa and Jeroen Minderman for statistical advice and Alex Wilson for advice on the experimental apparatus and comments on the manuscript. We are also grateful to two anonymous reviewers for their constructive comments on the manuscript. K.A.J. was funded by a Canadian Commonwealth Post-Doctoral Research Fellowship and J.-G.J.G. by a Discovery research grant from NSERC of Canada.


    • Received September 4, 2009.
    • Accepted October 9, 2009.


View Abstract