Behavioral Ecology Vol. 13 No. 4: 526-530
© 2002 International Society for Behavioral Ecology
A test of the risk allocation hypothesis: tadpole responses to temporal change in predation risk
Institute of Zoology, University of Zürich, CH-8057 Zürich, Switzerland
Address correspondence to J. Van Buskirk. E-mail: jvb{at}zool.unizh.ch .
Received 23 March 2001; revised 1 October 2001; accepted 15 October 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The risk allocation hypothesis predicts that temporal variation in predation risk can influence how animals allocate feeding behavior among situations that differ in danger. We tested the risk allocation model with tadpoles of the frog Rana lessonae, which satisfy the main assumptions of this model because they must feed to reach metamorphosis within a single season, their behavioral defense against predators is costly, and they can respond to changes in risk integrated over time. Our experiment switched tadpoles between artificial ponds with different numbers of caged dragonfly larvae and held them at high and low risk for different portions of their lives. Tadpoles responded strongly to predators, but they did not obey the risk allocation hypothesis: as the high-risk environment became more dangerous, there was no tendency for tadpoles to allocate more feeding to the low-risk environment, and as tadpoles spent more time at risk, they did not increase feeding in both environments. Our results suggest that the model might be more applicable when the time spent under high predation risk is large relative to the time required to collect resources.
Key words: Aeshna, Anura, behavior, predation risk, Rana lessonae, tadpoles.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Lima and Bednekoff (1999
)
recently introduced an intuitively appealing refinement to foraging theory.
Most models of foraging under predation risk assume that animals are exposed
to constant levels of danger or that variation in danger is a feature of the
background environment (Clark and Levy,
1988; McNamara and Houston,
1994; Werner and Anholt,
1993). But animals foraging under temporal variation in risk face
a problem in allocating costly antipredator behavior across situations of
different risk. Lima and Bednekoff examined this problem and arrived at two
key predictions. First, if an animal experiences two environments having high
and low levels of risk, it will allocate increasing amounts of antipredator
behavior to the high-risk environment as the danger in that environment
increases relative to that in the low-risk situation
(Figure 1A). Intuitively, if
one environment is much more dangerous than the other, the animal will choose
to collect most of its food requirements during its time in the less dangerous
environment. The second prediction is that as the environment with higher
danger is encountered more frequently, an animal will exhibit less
antipredator behavior in both environments, but especially when in the
situation with lower risk (Figure
1B). In other words, if high-risk situations predominate, an
animal must take risks to gather enough food to survive, but it will allocate
as much risk taking as possible to the less dangerous environment. Thus, the
pattern of variation dramatically alters the response to predators. Here we
describe an empirical test of this hypothesis.
|
Early foraging models examined how animals should behave to optimize some
measure of resource consumption, without regard to their exposure to predators
(Stephens and Krebs, 1986).
The inclusion of predation risk within foraging theory has been an
improvement, in spite of the more complex theoretical approach that it
requires, because it more accurately predicts the behavior of animals within
natural settings. Behavioral responses to predation risk are diverse and
ubiquitous in foraging animals, and these often entail a change in foraging
rate (Kats and Dill, 1998;
Lima, 1998). Lima and
Bednekoff's study prompts us to ask whether temporal variation in danger is
also a necessary ingredient of foraging models.
We studied the impact of temporal variation in predation risk on tadpoles
foraging in ponds that contain chemical cues signaling whether predators are
present. Amphibian larvae have had ample opportunity to evolve adaptive
responses to temporal variation in predation risk because they encounter
changes in predator density over short and long time scales within natural
ponds (Van Buskirk, 2002).
Further, tadpoles are particularly well suited for testing the risk allocation
hypothesis because they often fulfill the conditions under which risk
allocation is most likely to operate (Lima
and Bednekoff, 1999). For example, the model assumes that an
animal must meet its energetic requirements within a fixed time period. Many
pond-breeding tadpoles satisfy this assumption because they must attain a
minimum size for metamorphosis within the limited time available before their
larval habitat degenerates (Wilbur,
1980). The model assumes further that an animal may defend itself
against predation but that successful defense can be achieved only at the cost
of reduced energy intake. The hiding behavior and reduced feeding activity
exhibited by many tadpoles when exposed to predators fulfills this assumption
because it protects tadpoles from predation
(Skelly, 1994) and yet carries
a cost of reduced rates of growth and development (Van Buskirk,
2000,
2002). Finally, the model
assumes that an animal can assess current predation risk and remember changes
in risk for a meaningful period of time. Tadpoles also meet this assumption:
they use chemical cues as a sensitive gauge of short-term changes in predation
risk (Stauffer and Semlitsch,
1993), and they behave as if they can "remember" a
previously experienced predation environment for several days
(Semlitsch and Reyer, 1992;
Van Buskirk, 2002). In these
respects, then, anuran larvae offer an appropriate system for testing the risk
allocation hypothesis.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The two main predictions of the risk allocation hypothesis (Figure 1) can be tested by experimentally manipulating two features of the predation environment perceived by an animal: the relative predation risks experienced within two or more environments, and the amount of time spent in the more risky environment. We tested the two predictions by manipulating both features simultaneously in an experiment with tadpoles of the pool frog, Rana lessonae.
Experimental design and implementation
The experiment was conducted in plastic tubs holding 80 l of water (0.27
m2) placed outdoors in a field on the campus of the University of
Zürich. The large size of these tubs and their exposure to prevailing
outdoor conditions provided a favorable environment for tadpoles, which seemed
to grow and behave normally in the experiment. We added 2 g rabbit food to
each tub on the first day of the experiment; thereafter, the tadpoles consumed
this food and the algae growing naturally on the walls of the tubs. To
facilitate behavioral observations, we left tubs free of litter or structures
under which tadpoles might take refuge. The conditions in our experimental
tubs did not simulate those prevailing in natural ponds, but this does not
compromise a test of Lima and Bednekoff's hypothesis. Risk allocation behavior
emerges as the solution of an optimality model and therefore presumably is
maintained by optimizing selection under conditions specified by the model. As
long as these conditions and assumptions are satisfied by our experimental
system, as we assert that they are, then tadpoles should obey the model if it
incorporates all biologically important processes.
The experiment had the five treatments depicted in Figure 2, plus a control treatment not shown in the figure. For five treatments, tadpoles were captured and transferred between tubs every day, so that they were exposed over time to two distinct environments with potentially different predation risks. We manipulated the perceived danger by confining different numbers of predatory dragonfly larvae (Aeshna cyanea) within four cages floating in every tub. Cages were constructed of a 10-cm length of plastic tube (11 cm diam, covered at the ends with fiberglass window screen), and depending on the treatment they were either left empty or they contained a single Aeshna. We fed each dragonfly 300 mg R. lessonae tadpoles on every third day, to ensure that chemical signals of predation were present in the pools.
|
Three treatments included different ratios of risk between the high- and low-risk situations (vertical column in Figure 2). In the first treatment, tadpoles were transferred regularly between two tubs that had no dragonflies. This represents the case in which the risk in the two environments is identical. In the second and third treatments, tadpoles spent half the time in tubs that had no predators and half the time in tubs that had either one or four caged Aeshna larvae. These treatments represented increasing ratios of risk between the high- and low-risk situations. In all three cases, tadpoles were switched between tubs every 24 h, at 1100 h each day, so they spent 50% of the time in each environment over every 48-h period.
These treatments allowed us to test the prediction shown in
Figure 1A. The risk allocation
hypothesis will be supported if tadpoles respond to increasing ratio of risk
by becoming increasingly inactive when at high risk and increasingly active at
low risk. An underlying premise of our design is that a tub with no predators
was not perceived by tadpoles as a completely danger-free situation. Although
we cannot gauge the risk assigned by tadpoles to a tub having no dragonflies,
we do know that a tub with four Aeshna is perceived as riskier than a
tub with one Aeshna, which in turn is more dangerous than a
zero-predator tub. A separate experiment confirmed that R. lessonae
tadpoles respond with increasing strength along a gradient in risk from tubs
with 0-4 caged Aeshna (Van
Buskirk and Arioli, 2002).
Three treatments manipulated the proportion of time spent in situations with high and low risk (horizontal row in Figure 2). In the first treatment, tadpoles spent 4 h out of every 24-h period in a tub with one caged Aeshna and spent the other 20 h in a tub with no predators. In the second treatment, tadpoles spent 24 h in the Aeshna tub and 24 h in the predator-free tub. (The data from this same treatment were also used to assess tadpole responses to changes in the ratio of risk, as explained above.) In the third treatment, tadpoles were exposed to one Aeshna for 20 h before being transferred to a predator-free tub for the remaining 4 h. Thus, the proportions of time spent at higher risk, over every 48-h period, were 0.17, 0.5, and 0.83 in the three treatments. All tadpoles were captured twice per day: in the 0.17 and 0.83 treatments they were switched between caged Aeshna and predator-free tubs; in the 0.5 treatment they were switched between tubs at 1100 h, but 4 h later they were simply recaptured and returned immediately to the same tub.
These three treatments allowed us to test the prediction shown in Figure 1B. Support for the risk allocation hypothesis will be obtained if, within both the caged Aeshna tubs and the predator-free tubs, activity increases with increasing time spent at high risk, and the increase should be more obvious in tadpoles within the low-risk environment.
In a sixth "control" treatment, there were no predators and the tadpoles were never captured or transferred between tubs. Comparison between the control and the treatment with a risk ratio of 0:0 estimates the behavioral consequences of transferring tadpoles.
The experiment began on 26 May 2000, when we added to every tub 10 tadpoles
(37/m2), drawn equally from two clutches produced from crosses
between adult frogs collected at a pond that naturally contains
Aeshna, 22 km SE of Zürich. When we started the experiment, the
tadpoles were 6 days old (stage 25; Gosner,
1960) with an average weight of 21 mg (SE = 2.8). We replaced dead
tadpoles in a number of tubs on two occasions early in the experiment, after 4
and 6 days. The behavioral observations began on 13 June, after tadpoles had
been exposed to the predation treatments for 18 days. By the end of the
experiment, the tadpoles averaged 500 mg and were at about Gosner stage
34.
The numbers of predators and tadpoles used in the experiment were realistic
in comparison with natural densities observed in a sample of 40 ponds near
Zürich between 1997 and 2000 (Van Buskirk, unpublished data). The
experimental treatment with four caged predators was similar to the ponds with
highest aeshnid densities in nature (14/m2), while the one-predator
treatment had a lower Aeshna density than 21% of the ponds. Fourteen
percent of all ponds supported higher densities of R. lessonae
tadpoles than we used in the experiment, 61% of ponds had lower densities of
tadpoles, and 25% contained no R. lessonae. The pond sampling data
also illustrate that large temporal changes in predator density, occuring over
a period of weeks, are very common in natural ponds
(Van Buskirk, 2002). Sudden
changes in predator composition, such as we imposed here, probably occur only
when tadpoles move through different microhabitats within ponds.
Behavioral observations
Activity data were collected on 13-14 June and 16-17 June. Over these 4
days we made 18 observations of every group of tadpoles in both high- and
low-risk tubs. The control tadpoles were never transferred between tubs and
therefore were observed 36 times within the same tubs. Observations were
spaced at least 10 min apart and consisted of a count of the number of
tadpoles active (swimming or feeding) and resting inactively. We began
collecting data at least 90 min after tadpoles were transferred into a new
tub, and for each group of tadpoles we averaged the observations within each
of the two tubs separately to obtain measures of behavior in the high-and
low-risk environments.
We interpret the proportion of tadpoles that are inactive as a measure of
cautiousness, corresponding to vigilance in Lima and Bednekoff's
(1999) formulation (see
Figure 1). This interpretation
is justified by two features of the tadpole activity response. First, reduced
activity protects tadpoles from dragonfly predation
(Skelly, 1994) in the same way
that vigilance acts to decrease vulnerability in the Lima and Bednekoff model.
Second, tadpoles often suffer a growth cost of their response to predators
(Van Buskirk, 2000), just as
vigilance is assumed by Lima and Bednekoff to carry the cost of reduced food
consumption. Our analysis therefore asked whether tadpole inactivity responded
to the experimental treatments in the manner predicted in
Figure 1.
Analyses
The treatments were replicated eight times, but our transferring procedure
destroyed independence between some experimental units. For the three
treatments used to test the response to increasing risk ratio, the eight
separate groups of tadpoles were arranged into four pairs that were switched
back and forth between matched pairs of tubs, one of which contained no
predator and the other of which had a caged Aeshna. The independent
experimental units for this comparison were the paris of tubs, so we began by
calculating the mean response for each paired group of tadpoles. The tadpoles
used to assess the impact of increasing proportion of time at high risk were
also switched between matched pairs of tubs, but here we did not combine
results from the two sets of non-independent tadpoles because they belonged to
different treatments. We therefore performed analyses and prepared figures as
if the eight replicates were independent for these two treatments, which
increased the apparent power of the experiment. The extent to which our data
fit the pattern depicted in Figure
1B does not depend on the replication of these two treatments, so
we suspect that this was not a cause for concern. The eight replicates of the
control treatment were completely independent, as were the four replicates of
the manipulation of risk ratio.
In the absence of information about all its parameters, the risk allocation
hypothesis cannot make quantitative predictions. We therefore assessed the
degree of support offered by the data for statistical outcomes predicted by
the model, rather than testing for significant departure from specific null
hypotheses. For example, when the ratio of risk between high-and low-risk
situations is manipulated, we expect to observe a main effect of predation
risk and a predation-by-risk ratio interaction
(Figure 1A). When the
proportion of time at risk is manipulated, the risk allocation hypothesis
predicts effects of both predation risk and time at risk, as well as an
interaction between the two (Figure
1B). We used model selection techniques to ask how much support
these predicted outcomes obtain from the data, relative to alternative
outcomes in which other combinations of the manipulated factors are important.
Our approach used a small-sample version of Akaike's Information Criterion
(AICc) to identify the statistical model that best represents the data using
the fewest parameters (lowest AICc value). We evaluated the overall weight of
evidence favoring each model according to its Akaike weight. Burnham and
Anderson (1998) provide a
complete description of model selection methods.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
When faced with increasing ratios of danger between high-and low-risk situations, R. lessonae tadpoles spent more time inactive in the high-risk environment (Figure 3A). There was good support for the statistical model predicted by the risk allocation hypothesis, containing the predator effect and its interaction with risk ratio (Table 1), but the observed pattern clearly conflicted with the predicted outcome because tadpoles did not show the expected increase in activity within the low-risk environment (Figure 3A).
|
|
We observed no change in cautious behavior as the proportion of time spent in the high-risk environment increased (Figure 3B). Here there was little support for the outcome predicted by the risk allocation hypothesis, and most support went instead to the two models that did not include an effect of the proportion of time at risk (Table 1).
Across the entire experiment, tadpoles responded appropriately to the presence of predators (21% decline in activity in tubs with at least one predator; contrast between predator presence and absence: F1,22 = 41.6, p <.001). However, the patterns of response to change in risk ratio and proportion of time at risk did not agree with those predicted by the risk allocation hypothesis (Figure 1).
Behavior in the control treatment was not very different from that in the treatment involving daily transfers between predator-free tubs (contrast between the two treatments: F1,10 = 3.29, p =.10; Figure 3A). This suggests that repeatedly capturing the tadpoles increased their time spent inactive by only a small amount.
Tadpole growth during the experiment declined with increasing predation risk, although the effect was significant only for the proportion of time spent at risk (effect of time at risk: F1,26 = 5.48, p =.27; effect of risk ratio: F1,18 = 2.64, p =.12). Final mass was 568 ± 32 mg (mean ± SE) in the control treatment, where tadpoles were never exposed to predators, and declined to 434 ± 43 mg in the treatment with the highest proportion of time at risk and 438 ± 62 mg in the treatment with four Aeshna in the high-risk environment. These results, in combination with the behavioral data in Figure 3, suggest that inactivity was associated with a cost of reduced growth rate, as assumed by the risk allocation hypothesis.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Our findings do not support the predation risk allocation hypothesis (compare Figure 3 with Figure 1). R. lessonae responded to predators, as expected, but the pattern of temporal variation in risk had no influence on antipredator behavior. Disagreements such as this between predictions of theory and results of experiments may originate from three general causes: the biological system may not fulfill assumptions underlying the theory, the experiment may not include appropriate conditions or treatments, or the theory may not incorporate processes that have biological importance. We believe that the tadpole system and our experimental design were appropriate for testing Lima and Bednekoff's (1999
) model and therefore
suggest that the hypothesis can be rejected for this particular situation. Of
course, no model can be globally discounted on the basis of a single
experimental test, and our results suggest conditions under which variation in
predation risk is likely to influence antipredator behavior in animals.
We do not believe that the results arose from a violation of assumptions of
the Lima and Bednekoff model. For R. lessonae and many other anurans,
sufficient food must be gathered to reach metamorphosis within a limited
period of time, with fitness declining precipitously for individuals that fail
to escape a drying pond or delay metamorphosis until late in the season
(Altwegg, 2001;
Smith, 1987;
Wilbur, 1980). Thus, tadpoles
probably meet the assumption of a fixed time horizon better than most other
animals. Fitness costs of avoiding predators, involving reduced growth and
development rates, are well established for many anurans
(Van Buskirk, 2000). Finally,
R. lessonae and other tadpoles behave as if they can integrate
predation risk over time, although this may simply involve monitoring
instantaneous hunger level or physiological conditions that result from
previous environments rather than an active memory of earlier predation levels
(Semlitsch and Reyer, 1992;
Van Buskirk, 2002). The three
main assumptions described in the Introduction are therefore fulfilled by this
system.
We also believe that our experimental design produced a valid test of the risk allocation hypothesis because we directly manipulated the two main parameters explored by Lima and Bednekoff. The outcome was not a consequence of using predator-free tubs to represent the low-risk environment, although that decision makes it impossible to calculate an exact ratio of predation risk between the high- and low-risk situations. The expected outcomes are not altered by recalculating optimal vigilance values under the condition that the threat of predation in the low-risk environment approaches zero. For example, as the number of predators in the high-risk environment increases (Figure 1A), the model always predicts symmetrical and compensatory behavioral responses in the two environments, even when one environment is nearly risk free. Thus, the differences between our results and those predicted under the risk allocation hypothesis cannot be blamed on this aspect of our design.
We are thus forced to reject the risk allocation hypothesis for the tadpole
system and to conclude that this model is not universally upheld in real
animals. Are animals other than tadpoles more likely to support predictions of
the Lima and Bednekoff model? Allocation of risky behavior to the more
dangerous environment should be especially clear when the time spent within
the dangerous environment is high relative to the time needed to collect
resources. Our experiment showed that tadpoles can maintain high individual
growth rates even when exposed to high risk for 83% of their lives, suggesting
that they acquire sufficient food without having to increase their allocation
of feeding behavior to the high-risk periods. However, a few other
well-studied animals may be unable to avoid risk allocation. Some desert
rodents and insectivorous birds, for example, must devote large fractions of
their time to gathering food (Ekman,
1987; Kotler et al.,
1994; Mitchell et al.,
1990; Van Buskirk and Smith,
1989), and high-risk predation environments for these animals may
be fairly frequent in nature (Bouskila,
1995; Kotler et al.,
1993). If we assume that foraging time is related to the minimum
required rate of energy intake and that animals must gather resources within a
limited period of time, then these observations suggest that the risk
allocation hypothesis can be relevant in natural settings. For both desert
rodents and insectivorous birds, the proportion of time at high risk may be
large relative to the food supply in nature, and as a consequence changes in
time at risk are expected to impinge upon time necessary for gathering food.
The results of our experiment, although not supporting Lima and Bednekoff's
(1999) model, nevertheless
point toward conditions under which the hypothesis is more likely to
apply.
|
ACKNOWLEDGEMENTS |
|---|
We are grateful to R. Altwegg for providing tadpoles for the experiment; to M. Arioli and B.R. Schmidt for help and ideas; to P. Bednekoff, A. Bourke, B. Hellriegel, S. Peacor, B.R. Schmidt, C. Vorburger, and two anonymous reviewers for constructive comments on the manuscript; and to the Swiss National Science Foundation for financial support (31-50525.97). Authorization for the experiment was granted by the Kantonales Veterinäramt Zürich (73/2000).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Altwegg R, 2001. Complex life cycles and predation risk: plastic strategies for growth and life history in waterfrogs (PhD dissertation). Zürich: University of Zürich.
Bouskila A, 1995. Interactions between predation risk and competition: a field study of kangaroo rats and snakes. Ecology 76: 165-178.
Burnham KP, Anderson DR, 1998. Model selection and inference. New York: Springer.
Clark CW, Levy DA, 1988. Diel vertical migrations by juvenile sockeye salmon and the antipredator window. Am Nat 131: 271-290.
Ekman J, 1987. Exposure and time use in willow tit flocks: the cost of subordination. Anim Behav 35: 445-452.
Gosner KL, 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica 16: 183-190.
Kats LB, Dill LM, 1998. The scent of death: chemosensory assessment of predation risk by prey animals. Ecoscience 5: 361-394.
Kotler BP, Ayal Y, Subach A, 1994. Effects of predatory risk and resource renewal on the timing of foraging activity in a gerbil community. Oecologia 100: 391-396.
Kotler BP, Brown JS, Slotow RH, Goodfriend WL, Strauss M, 1993. The influence of snakes on the foraging behavior of gerbils. Oikos 67: 309-316.
Lima SL, 1998. Nonlethal effects in the ecology of predator-prey interactions. Bioscience 48: 25-34.[ISI]
Lima SL, Bednekoff PA, 1999. Temporal variation in danger drives antipredator behavior: the predation risk allocation hypothesis. Am Nat 153: 649-659.
McNamara NB, Houston AI, 1994. The effect of a change in foraging options on intake rate and predation rate. Am Nat 144: 978-1000.
Mitchell WA, Abramsky Z, Kotler BP, Pinshaw P, Brown JS, 1990. The effect of competition on foraging activity in desert rodents: theory and experiments. Ecology 71: 844-854.
Semlitsch RD, Reyer H-U, 1992. Modification of anti-predator behaviour in tadpoles by environmental conditions. J Anim Ecol 61: 353-360.
Skelly DK, 1994. Activity level and the susceptibility of anuran larvae to predation. Anim Behav 47: 465-468.
Smith DC, 1987. Adult recruitment in chorus frogs: effects of size and date of metamorphosis. Ecology 68: 344-350.
Stauffer H-P, Semlitsch RD, 1993. Effects of visual, chemical, and tactile cues of fish on the behavioural responses of tadpoles. Anim Behav 46: 355-364.
Stephens DW, Krebs JR, 1986. Foraging theory. Princeton, New Jersey: Princeton University Press.
Van Buskirk J, 2000. The costs of an inducible defense in anuran larvae. Ecology 81: 2813-2821.
Van Buskirk J, 2002. Phenotypic lability and the evolution of predator-induced plasticity in tadpoles. Evolution 56: 361-370.[Medline]
Van Buskirk J, Arioli M, 2002. Dosage response of an induced defense: how sensitive are tadpoles to predation risk? Ecology (in press).
Van Buskirk J, Smith DC, 1989. Individual variation in winter foraging of black-capped chickadees. Behav Ecol Sociobiol 24: 257-263.
Werner EE, Anholt BR, 1993. Ecological consequences of the tradeoff between growth and mortality rates mediated by foraging activity. Am Nat 142: 242-272.
Wilbur HM, 1980. Complex life cycles. Annu Rev Ecol Syst 11: 67-93.[ISI]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  P. C. Griffin, S. C. Griffin, C. Waroquiers, and L. S. Mills Mortality by moonlight: predation risk and the snowshoe hare Behav. Ecol., September 1, 2005; 16(5): 938 - 944. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||