| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Behavioral Ecology Vol. 12 No. 3: 318-324
© 2001 International Society for Behavioral Ecology
Background context and decision making in hoarding gray jays
Department of Evolution, Ecology, and Organismal Biology, and Department of Anthropology, Ohio State University, Columbus, OH 43210-1293, USA
Address correspondence to T.A. Waite. E-mail: waite.1{at}osu.edu .
Received 2 October 1999; revised 22 July 2000; accepted 12 September 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
If decision makers assign stable fitness-related values to options, preference for the most valuable of simultaneously encountered options should be independent of background context (i.e., prior options). The tendency to choose optionx versus y should be unaffected by whether the decision maker has already been given a choice betweenx' and y' or between x'' and y''. Here, food-hoarding gray jays (Perisoreus canadensis) were given an initial choice between x' (one raisin, 0.5 m into a tube) and y' (three raisins, 0.5 m) or between x'' and y'' (both identical to x'). All subjects were then given a choice between x' (one raisin, 0.3 m) and y' (three raisins, 0.7 m). In violation of the principle of irrelevant alternatives, the "market share" ofx depended on prior options. Subjects initially exposed to context {x', y'} showed a stronger preference for x than did subjects initially exposed to {x'', y'}, which implies that the jays did not assign a fixed value to each option. Subjects that initially could obtain a large reward (y') for about the same "price" (perceived danger) as a small reward (x') apparently devalued the large reward (y) in the subsequent choice. This effect may be the joint byproduct of cognitive constraints and an adaptive tendency to use information provided by the context.
Key words: choice, cognitive constraints, context, decision making, hoarding, information processing.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Natural selection is often assumed to favor choice behavior that maximizes fitness. This conventional view suggests, misleadingly (Houston, 1997
;
Ydenberg, 1998), that decision
makers assign a fixed fitness-related value to each option. By extension, an
animal choosing among simultaneously encountered options is expected to choose
the most valuable option. Assuming natural selection favors minimization of
costly errors, partial rather than absolute preference is expected
(McNamara and Houston, 1987;
Waite and Field, 2000), but
the strength of this preference should be independent of the background
context (i.e., options encountered in the past). The tendency to prefer option
x over y in simultaneous encounters should be unaffected by
whether the decision maker has already experienced encounters with options
x' and y' or options x'' and
y''. This prediction is consistent with the principle of
irrelevant alternatives from choice theory
(Tversky and Simonson, 1993)
and with models of simultaneous choice from foraging theory
(Stephens and Krebs, 1986).
Thus, according to standard theories of rational choice and optimal foraging,
the "attractiveness" of an option should be determined by its real
(fitness-related) value, which should be independent of prior options.
Despite the parsimony of this framework, decision-making processes are
known to be subject to background context effects, particularly in human
consumers. For example, Simonson and Tversky
(1992) studied choice of
products (automobile tires) varying on two attributes (warranty and price).
They found that subjects were more likely to choose a low-quality cheap option
over a high-quality expensive option if they had already been given a choice
in which a comparable difference in quality was associated with a smaller
difference in price. Subjects were given an initial choice between options
x' (55,000 miles, $85) and y' (75,000 miles,
$91) or between x'' (30,000 miles, $25) and y''
(35,000 miles, $49). All subjects were then given a choice between x
(40,000 miles, $60) and y (50,000 miles, $75). Subjects exposed to
background B' = {x', y'} showed a
stronger preference for x (57%) than did subjects exposed to
B'' = {x'', y''} (33%). In violation
of the principle of irrelevant alternatives (i.e., contrary to the normative
economic expectation of rational choice theory), the subjective valuations of
options x and y apparently depended on options available in
the past. This seemingly maladaptive effect was interpreted as the byproduct
of a cognitive illusion or bias (Simonson
and Tversky, 1992; Tversky and
Simonson, 1993), a systematic and directional error of judgment in
violation of economic value maximization (for detailed definitions and
extensive reviews of so-called cognitive illusions [including the conjunction
fallacy, base-rate error, Concorde fallacy, and single-event probability
problem] see Anderson, 1998;
Fantino, 1998;
Griffin and Buehler, 1999;
Kahneman and Tversky, 1996;
Nichols, 1999;
Piattelli-Palmarini, 1994).
Alternatively, from an evolutionary perspective, such effects are
interpretable as an artifact of imposing unnecessarily narrow norms of
rational choice and presenting subjects with contrived choice tasks
(Gigerenzer, 2000). From this
perspective, background context effects are even potentially representative of
rational (Wernerfelt, 1995) or
adaptive (Houston, 1997) use
of information provided by the context (reviewed by
Anderson, 1998;
Giraldeau, 1997).
Here, I describe an analogous experiment designed to determine whether a
background context effect could be induced in a non-human decision maker, the
gray jay (Perisoreus canadensis). Jays collecting food for storage
were given an initial choice between options x' (small reward)
and y' (large reward) or between x'' and
y'' (both identical to x'). All of these initial
options could be obtained for an intermediate "price." All
subjects were then given a choice between a small cheap reward (x)
and a large expensive reward (y). According to classical theories of
rational choice (Bell et al.,
1988) and optimal foraging
(Stephens and Krebs, 1986),
the tendency to choose x versus y should be unaffected by
any such manipulation of background context. Evidence to the contrary would
challenge the notion that non-human animals are virtually immune to cognitive
illusions because their decision making processes are shaped by natural
selection (see also Arkes and Ayton,
1999).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Subjects and study area
The gray jay (Perisoreus canadensis), a year-round resident of boreal forests of North America, lives in social groups on all-purpose territories (see Strickland and Ouellet, 1993
for a detailed account; see also
Waite and Strickland, 1997).
During summer and autumn, jays routinely make hundreds of food caches per day,
placing each saliva-coated bolus in a separate arboreal site (Waite and Reeve,
1993,
1994). They rely on these
caches throughout the winter and even use them to provision young. Twenty semi-tame, free-ranging gray jays, identifiable by unique color-band combinations, were used as subjects. These subjects comprised seven adult females and eight adult males, two philopatric juvenile females, two philopatric juvenile males, and one immigrant (unrelated) juvenile female.
The experiment was conducted in Algonquin Provincial Park, Ontario, Canada
(45°33' N, 78°38' W; see
Strickland, 1991, for detailed
description) between 22 and 26 October 1998. All tests were conducted between
0750 and 1757. Air temperature varied between -2 and 15°C. No
precipitation occurred during the tests.
Experimental procedure
The experiment was designed to measure the effect of manipulated background
context on binary choice. Options varied along two attributes, quality (number
of raisins) and price (distance into a tube). Presumably, the jays perceived
greater danger with greater distance into the tube, as suggested by their
hesitant sidling into the tubes, especially during performance tests (see
below). By entering farther into a tube, a jay's ability to escape the
surprise attack of a predator would be compromised. Instead of instantly
flying to cover upon detecting the oncoming predator, the jay would have to
hop to the opening of the tube and then flee. The choice task thus measured
the jays' tendency to choose a low-quality, safe option versus a high-quality,
dangerous option as influenced by background context. Defined by previously
encountered options, the background context was manipulated as described
below.
The experimental set-up consisted of two tubes, both containing food
(Figure 1). Subjects were
required to hop into either tube during each visit to obtain the food reward.
The semi-cylindrical (radius = 25 cm) tubes were 1.2-m long, made of 1-cm
welded wire mesh (hardware cloth), and closed at one end. The open ends were
situated equidistant from a standard 25-cm-high perch. From this perch,
subjects could inspect the contents of the two tubes simultaneously. Food was
placed on white plastic discs (diameter = 17.8 cm), positioned at specified
distances into the tubes. Immediately upon each arrival, the subject could
choose between the two options. As soon as the jay collected the food (one or
three raisins) and flew to hoard it in (one or three) nearby arboreal sites,
the food was replenished (as described below) before the jay returned to the
set-up, usually within a minute. Detailed descriptions of hoarding behavior in
this species are published elsewhere
(Waite and Ydenberg, 1996,
references therein).
|
To test a particular individual, I positioned myself near the known core of
its territory and whistled to attract it. I then conducted a performance test
in which the prospective subject was given a choice between identical options
(one raisin, same distance into tube). During this 15-visit session, raisins
were positioned 0.3 m into the tubes during the first five visits, 0.5 m
during the next five visits, and 0.7 m during the final five visits. To train
the subjects to treat the task as an exclusive binary choice, I flushed any
jay that attempted to enter both tubes during a given visit. No individual
attempted to do so more than twice. All individuals passed the performance
test, entering one of the tubes during each visit, collecting the raisin and
transporting it to a nearby arboreal site for storage. Throughout this and
subsequent tests, raisins were offered to all accompanying jays to minimize
the subject's tendency to adjust its choice behavior in response to
interference competition (Waite and
Ydenberg, 1996). Following this test, the set-up was moved 
60
m to minimize the subject's tendency to adjust load size in response to
density-dependent risk of cache theft
(Waite and Ydenberg,
1996).
Following a 10-min hiatus, the subject was exposed to background context
B' or B''
(Figure 1) during each of 25
consecutive visits. Half the subjects were exposed to B' and
the other half were exposed to B''. To avoid an order effect,
background context was randomly assigned for the first subject and then
alternated across successive subjects. Under B', subjects were
given a choice between option x' (one raisin, 0.5 m into tube)
and y' (three raisins, 0.5 m). Under B'',
subjects were given a choice between x'' and y''
(both identical to x'). To avoid a side bias, the positioning
of the rewards was randomized initially and then switched after each visit
(even for B''). After the last visit, the set-up was moved
(60 m).
The target-set test (Figure 1) was then conducted, following a 10-min intertrial interval. During this test, all subjects were given a choice between x (one raisin, 0.3 m) and y (three raisins, 0.7 m) during each of 25 consecutive visits. The positioning of x and y was randomized initially and then switched after each visit. Each subject was tested once, following the initial 25-visit exposure to either B' or B''. (Beginning with the onset of the background test, B' subjects completed the experiment [50 hoarding trips] in 44 min [SD = 6.2], on average, while B'' subjects did so in 56 min [SD = 34.3].) I thus used a between-subjects design in which 10 subjects were tested following exposure to context B' and 10 other subjects were tested following exposure to B''.
Data analysis
Analyses used SigmaStat
(1997) routines. The
Mann-Whitney U test was used to test whether the tendency to choose
x rather than y differed between subjects previously exposed
to background B'' versus those previously exposed to
B''. Wilcoxon signed-ranks tests were used to test whether the
tendency to choose x' or x'' during the
background-set test differed from the tendency to choose x in the
target-set test. Wilcoxon tests were also used to evaluate any preferences for
x' versus y', x'' versus
y'', and x versus y. These tests compared the
observed proportion of x', x'', or x
choices with random expectation (0.5). Finally, Spearman's rank correlation
was used to evaluate the degree of autocorrelation (lag = 1) in each subject's
25 consecutive choices during the target-set test. For each series of choices
(where x = 0 and y = 1), C1, 2,..., 25,
the correlation between Ct and
Ct-1 was calculated. For any subject whose
choices were autocorrelated (at the 0.05
-level), logistic regression
was used to evaluate whether the subject's tendency to choose x
versus y increased or decreased across the 25 consecutive choices.
All tests were two-tailed.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The tendency to choose option x over y was stronger among subjects initially exposed to background B' than among those initially exposed to B'' (U = 153.5, p <.001; Figure 2). In fact, nine of the 10 subjects exposed to B' showed a stronger preference for x than did all 10 subjects exposed to B''. Subjects exposed to B' showed a preference for y' over x' during the background-set test (median proportion y' = 0.74; Z = 2.706, p =.007; Table 1), but showed a preference for x over y (median proportion x = 0.92; Z = 2.812, p =.005; Figure 2) during the subsequent target-set test. By contrast, subjects exposed to B'' showed no significant preference during either the background-set test (median proportion x'' = 0.5; Z = -0.36, p =.72; Table 1) or target-set test (median proportion x = 0.62; Z = 1.381, p =.17; Figure 2). Among subjects exposed to B', the tendency to choose x (median = 0.92; Figure 2) differed significantly from the prior tendency to choose x' (median = 0.26; Z = 2.807, p =.005; Table 1). However, among subjects exposed to B'', the tendency to choose x (median = 0.62; Figure 2) did not differ significantly from the prior tendency to choose x'' (median = 0.5; Z = 1.227, p =.22; Table 1).
|
|
No significant autocorrelation in the tendency to choose x versus y was detected for any of the subjects previously exposed to background B'' (all p's >.15), and only one significant autocorrelation was found among the subjects previously exposed to background B'' (rs = -.45, p =.026; all other p's >.07). Logistic regression was used to detect any (increasing or decreasing) trend in this subject's tendency to choose x versus y across the 25-visit target-set test. No strong trend was suggested by the odds ratio (0.59), an estimate of the increase in odds of choosing y in the next visit. The 60% confidence interval (0.47, 0.75) of the odds ratio encompassed 0.5, which implies that the subject's tendency to choose x versus y did not change significantly across the 25 visits in the target-set test.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Contrary to classical theories of rational choice, the jays' choice behavior was influenced by background context. According to the principle of irrelevant alternatives, the market share of an option should be independent of other options, including those encountered in the past (Tversky and Simonson, 1993
).
While minimizing costly errors (Houston,
1997; Waite and Field,
2000), subjects should prefer the most valuable of simultaneously
encountered options, and the strength of this preference should be independent
of past alternatives. In violation of this principle, jays differed in their
tendency to choose option x versus y depending on whether
they had been exposed to background B' or B''
(Figure 2). This finding is
surprising given the body of work showing that gray jays often make decisions
that closely match predictions based on the rate-maximization currency (Waite
and Ydenberg, 1994a,
b,
1996, and references therein).
The observed effect of background context implies that the jays' subjective
valuation of options was not absolute or fixed. Instead, their perception of
an option's value apparently depended on theoretically irrelevant options
encountered in the past. The influence of recent experience on foraging
decisions is well known (reviewed by
Giraldeau, 1997), but unlike
the present study, these studies have focused on the role of information
acquisition via statistical sampling in stochastic environments (see
further discussion below).
The ideas that rewards have relative as well as absolute effects on choice
behavior and that these effects depend on an animal's recent experience have
been long recognized, though largely overlooked, in the psychological
literature (see Flaherty, 1996
for a fascinating historical account). For example, in one experiment in
Crespi's (1942) classic study,
rats were trained to obtain a food reward at the end of a runway. Three groups
of rats were differentially rewarded during the first day of the experiment.
Subjects in one group received a 256-unit reward (5.12 g of dog food), those
in another group received a 64-unit reward, and those in a control group
received a 16-unit reward. Mean performance (running speed) was highest in
subjects receiving the 256-unit reward, intermediate in those receiving the
64-unit reward, and lowest in the control subjects. Beginning on the second
day of the experiment, all subjects received a 16-unit reward. The running
speed of the subjects that had been shifted downward from 256 or 64 units was
compared with that of the control subjects. Paradoxically, the shifted
subjects abruptly reduced their running speed below that of the
control subjects, even though all subjects now received identical 16-unit
rewards. This so-called negative contrast effect persisted throughout the
experiment (i.e., over the subsequent 7 days). Although the experimental
procedures and conceptual framework of this study differed from those of the
present study, the findings are nonetheless reminiscent of the observed effect
of prior options on the jays' behavior and are likewise difficult to explain
from a functional perspective.
More reminiscent of the present results, and perhaps more paradoxical, are
the results of recent work on choice behavior in pigeons (reviewed by
Shettleworth, 1998). In one
experiment (Belke, 1992),
pigeons were trained on two pairs of concurrent variable interval (VI)
schedules. (On a VI schedule, intervals between successive reinforcements vary
randomly around a specified [mean] VI value.) In alternating periods, subjects
preferred option VI:20 s (white key) over VI:40 (red) and option VI:40 (green)
over VI:80 (yellow), where the numerals represent the mean interval between
successive rewards. Not surprisingly, in both pairs of schedules, subjects
preferred the option with the shorter mean delay to reward (i.e., VI:20-white
over VI:40-red, and VI:40-green over VI:80-yellow). However, when offered the
choice between red and green, the subjects strongly preferred green, even
though both colors signaled VI:40 s and hence the principle of value (or rate)
maximization would predict no preference.
Even more paradoxical are the results of a follow-up experiment
(Gibbon, 1995) in which
pigeons, beyond showing a preference where none was expected, showed a
preference opposite that predicted by value maximization.
Specifically, the pigeons strongly preferred the equivalent of VI:40-green
over VI:20-white. Thus, the VI:40 option from the VI:40,VI:80 background
context was more "attractive" than the VI:20 option from the
VI:20,VI:40 background context, even though the delay to reward was twice as
long for the preferred option. While it is conceivable that the jays' choice
behavior is representative of the usually adaptive use of information provided
by recent experience (see below), the pigeons' choice behavior in Gibbon's
experiment seems to have no reasonable functional explanation. Because the
pigeons' preference directly contradicts value maximization, only a blithe
adaptationist would be quick to interpret it as being somehow adaptive. After
all, how could it be adaptive to respond to recent experience by strongly and
persistently preferring an option that is only half as profitable as an
alternative? Instead, this finding suggests that background context may
produce a cognitive illusion or bias, a directional error of judgment in
violation of value maximization (for reviews see
Anderson, 1998;
Fantino, 1998;
Griffin and Buehler, 1999;
Kahneman and Tversky, 1996;
Nichols, 1999;
Piattelli-Palmarini, 1994; but
see Gigerenzer, 1996).
The effect of background context observed in the gray jays, though less
dramatic, is likewise inconsistent with value maximization. It is consistent,
though, with a descriptive (nonfunctional) hypothesis of choice behavior in
human consumers. The tradeoff contrast hypothesis
(Tversky and Simonson, 1993)
predicts an increased tendency to choose a low-quality cheap option over a
high-quality expensive option if the decision maker has already been given a
choice in which a comparable difference in quality was associated with a
smaller difference in price. In agreement with this prediction, gray jays
showed a strong preference for a low-quality cheap option
(Figure 2) if they had already
been given a choice in which the same difference in quality was associated
with no difference in price (Figure
1). Subjects initially offered a three-for-the-price-of-one deal
showed a strong subsequent preference for the small reward. Thus, the large
but now expensive reward was less "attractive" on this background.
A higher-than-background difference between the price of two options, as in
the present experiment, may prompt the decision maker to undervalue the
higher-quality option. If fixed values were assigned to each option, decision
makers would be immune to such effects, which occur apparently because the
background influences the subjective valuation of options.
To explore more formally the potential influence of background context on
subjective valuation, I describe here a model developed to explain context
effects in human consumers (Tversky and
Simonson, 1993). The purpose is not to test the model, but to
explore whether conditions like those in the present experiment could lead to
paradoxical preferences and to generate predictions to be tested in future
experiments. This modeling exercise reinforces the notion that models of
decision making should recognize that an option's fitness value is context
dependent rather than fixed (Houston,
1997) and that cognitive biases can lead to paradoxical departures
from value maximization. Whatever the underlying mechanisms, the model shows
how comparative evaluation (Shafir,
1994) of options varying along multiple dimensions can lead to
paradoxical effects of background context on choice. By contrast, standard
theories of choice cannot account for such effects.
For a given background B, the decision rule is to chose x
from option set S = {x, y} if and only if the subjective
value VB(x, S) exceeds
VB(y, S) (see
Table 2 for definitions of
symbols). As done elsewhere (Waite and
Field, 2000), it would be straightforward to incorporate the
notion that decision-making processes are inevitably subject to error
(McNamara and Houston, 1987),
particularly when the fitness consequences of erroneous choice are trivial
(Houston, 1997). Future work,
both theoretical and experimental, should explore the possibility that
paradoxical effects of context are more common when the fitness consequences
of such erroneous choice are smaller. Animals may be especially vulnerable to
background effects when the fitness consequences are trivial. However, to
simplify the present analysis, the decision maker is expected to
choose—repeatedly and consistently—the option with higher
subjective value:
 | (1) |
). With
no apparent adaptive basis (and in violation of the condition termed
regularity; Huber et al.,
1982; Tversky and Simonson,
1993), the jays responded to the introduction of an
"unattractive" option by increasing their absolute preference for
the more similar (less preferred) of the two original alternatives. This local
context effect strengthens the inference from the present findings that the
jays do not assign fixed values to options. Instead, they apparently use a
comparative method of evaluation, assigning flexible values to options based
on comparisons with current (Hurly and
Oseen, 1999; Shafir,
1994; Shafir et al.,
1989; Tversky,
1969; Waite, 2001)
and past (this study) alternatives.
|
Here, I explore the flexible valuation of options as influenced by
background context alone. To isolate the effect of background context, I
eliminate the local-context component of the model:
 | (2) |
 | (3) |
Table 3 summarizes two numerical examples. In the first example, the real values of option x (= vq(xq) + vp(xp) = 1 + 3) and y (= vq(yq) + vp(yp) = 3 + 1) are identical. For ß = 0, the subjective value of an option equals its real value (Equation 3 reduces to value maximization). Therefore, the subjective value of the options is identical in the two contexts and no preference is predicted. However, for ß = 10, the subjective value of x exceeds that of y on background B', reflecting the greater weighting of price (bp = 0.75) under that context. The model predicts a preference for x under B', but no preference under B''. In the second example, the real value of option y (= 3 + 1) exceeds that of x (= 1 + 2). Thus, for ß = 0, the subjective value of y exceeds x (by the same degree in both contexts). However, for ß = 10, the model predicts a preference for x under B' and for y under B''. This preference reversal reflects the interaction between the asymmetry in true value and the greater relative weighting of price under B'. Figure 3 illustrates the effects of variation in ß and bp where the true values of options x and y are identical. As ß approaches zero (Figure 3a), or bp approaches 0.5 (Figure 3b), the difference in the subjective values of x and y declines linearly to zero. Thus, the model generates predictions (e.g., Figure 3b) that are qualitatively consistent with the measured effect of background context in our experiment (Figure 2). For animals faced with the choice between options varying along two dimensions (e.g., quality and price), as in our experiment, background context has the potential of inducing dramatic shifts and even reversals in preference.
|
|
Although it may be tempting to view such paradoxical effects strictly as
byproducts of intrinsic cognitive biases
(Tversky and Simonson, 1993),
these apparent deficits could have an adaptive basis
(Houston, 1997), especially
when decisions are made under uncertainty (e.g.,
Anderson, 1998;
Bateson and Kacelnik, 1996;
Dall et al., 1999;
Kamil et al., 1993;
Sih, 1992;
Stephens, 1989; reviewed by
Dukas, 1998;
Giraldeau, 1997). Animals may
routinely cope with uncertainty about tradeoffs between attributes by using
information provided by the background. If so, background effects may be
especially dramatic during early stages of exposure to novel tasks, as in the
present experiment. If an animal in a stochastic environment uses information
to update its valuation of options
(Giraldeau, 1997), the
influence of background may become less dramatic as the animal gains
experience with an array of option sets (equivalently, ß may become
smaller). (From this perspective, because the jays were exposed to a single
background context and so were not able to acquire information from a
stochastic array of option sets, it is not surprising that the jays'
preferences were steady throughout the target-set test.) Similarly, as an
animal gains experience with a choice involving a quality-price tradeoff, the
relative weighting of price (danger) may decline (bp may
become smaller) as the animal gains experience, especially if the animal
habituates to the danger. (The jays' perception of danger may have stabilized
during the performance and background-set tests.) If so, an animal may become
increasingly willing to pay a high price for a given valuable reward
(Figure 3) (cf.
Lima and Bednekoff, 1999).
Thus, background context effects may be common if animals making tradeoffs
between attributes initially tend to overweight the background (see above) or
overestimate the predation risk (as theorized by
Abrams, 1995;
Bouskila et al., 1995, and
references therein).
This speculation prompts a testable prediction: if animals faced with
simultaneous choice under uncertainty and predation risk behave as if they
weight background or price less heavily as they acquire information, the
effect of background should decline with experience. The background effect
should be weakened by increased exposure to an array of background option
sets, and choice behavior should converge on value maximization as assigned
values and hence preferences become more stable. Moreover, closer convergence
on value maximization should be expected in situations where the fitness costs
of erroneous choice are higher (see
Houston, 1997;
Waite and Field, 2000).
Consistent with the theme of the emerging field of cognitive ecology
(Dukas, 1998;
Giraldeau, 1997;
Pepperberg et al., 1998; Real,
1991,
1993;
Shettleworth, 1998),
experimental tests of these predictions could help resolve the conflict
between the view that context effects are cognitive illusions
(Tversky and Simonson, 1993)
versus the view that such effects are artifactual
(Gigerenzer, 2000; see also
Anderson, 1998) or may even
represent rational (Wernerfelt,
1995) or adaptive (Houston,
1997) use of information provided by the context.
|
ACKNOWLEDGEMENTS |
|---|
This article is dedicated to Dan Strickland—naturalist and corvidophile extraodinaire—on the occasion of his (nominal) retirement. I thank Dan for three decades' worth of background information, Sharoni Shafir for background context (inspiration and prior collaboration), John Vucetich for discussion, and two anonymous reviewers for valuable comments on the manuscript.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abrams PA, 1995. Overestimation versus underestimation of predation risk: a reply to Bouskila et al. Am Nat 145: 1020-1024.[Web of Science]
Anderson JL, 1998. Embracing uncertainty: the interface between Bayesian statistics and cognitive psychology. Cons Ecol [online] 2: 2. URL: http://www.consecol.org/vol2/iss1/art2.
Arkes HR, Ayton P, 1999. The sunk cost and Concorde effects: are humans less rational than lower animals? Psych Bull 125: 591-600.[Web of Science]
Bateson M, Kacelnik A, 1996. Rate currencies and the
foraging starling: the fallacy of the averages revisited. Behav
Ecol 7:
341-352.
Belke TW, 1992. Stimulus preference and the transitivity of preference. Anim Learn Behav 20: 401-406.[Web of Science]
Bell DE, Raiffa H, Tversky A (eds), 1988. Decision making. Cambridge: Cambridge University Press.
Bouskila A, Blumstein DT, Mangel M, 1995. Prey under stochastic conditions should probably overestimate predation risk: a reply to Abrams. Am Nat 145: 1015-1019.[Web of Science]
Crespi LP, 1942. Quantitative variation in incentive and performance in the white rat. Am J Psych 40: 467-517.
Dall SRX, McNamara JM, Cuthill IC, 1999. Interruptions to foraging and learning in a changing environment. Anim Behav 57: 233-241.[Web of Science][Medline]
Dukas R (ed), 1998. Cognitive ecology: the evolutionary ecology of information processing and decision making. Chicago: University of Chicago Press.
Fantino E, 1998. Judgment and decision making: behavioral approaches. Behav Anal 21: 203-218.
Flaherty CF, 1996. Incentive relativity. Cambridge: University of Cambridge Press.
Gibbon J, 1995. Dynamics of time matching: arousal makes better seem worse. Psycho Bull Review 2: 208-215.
Gigerenzer G, 2000. Adaptive thinking: rationality in the real world. Oxford: Oxford University Press.
Giraldeau L-A, 1997. The ecology of information use. In: Behavioural ecology: an evolutionary approach, 4th ed. (Krebs JR, Davies NB, eds). Oxford: Blackwell Science; 42-68.
Griffin D, Buehler R, 1999. Frequency, probability, and prediction: easy solutions to cognitive illusions? Cogn Psych 38: 48-78.
Houston AI, 1997. Natural selection and
context-dependent values. Proc R Soc Lond B
264: 1539-1541.
Huber J, Payne JW, Puto C, 1982. Adding asymmetrically dominated alternatives: violations of regularity and the similarity hypothesis. J Consumer Res 9: 90-98.[Web of Science]
Hurly TA, Oseen MD, 1999. Context-dependent risk-sensitive foraging preferences in wild rufous hummingbirds. Anim Behav 58: 59-66.[Web of Science][Medline]
Kahneman D, Tversky A, 1996. On the reality of cognitive illusions. Psych Rev 103: 582-591.[Web of Science][Medline]
Kamil AC, Misthal RL, Stephens DW, 1993. Failure of
simple optimal foraging models to predict residence time when patch quality is
uncertain. Behav Ecol 4:
350-365.
Lima SL, Bednekoff PA, 1999. Temporal variation in danger drives antipredator behavior: the predation risk allocation hypothesis. Am Nat 153: 649-659.[Web of Science]
McNamara JM, Houston AI, 1987. Partial preferences and foraging. Anim Behav 35: 1084-1099.
Nichols N, 1999. Cognitive illusions, heuristics, and climate prediction. Bull Am Meteorol Soc 80: 1385-1397.
Pepperberg I, Balda R, Kamil AC (eds), 1998. Animal cognition in nature. San Diego: Academic Press.
Piattelli-Palmarini M, 1994. Inevitable illusions. New York: John Wiley & Sons.
Real L, 1991. Animal choice behavior and the evolution
of cognitive architecture. Science 253:
980-986.
Real L, 1993. Toward a cognitive ecology. Trends Ecol Evol 8: 413-417.
Real L, 1996. Paradox, performance, and the architecture of decisionmaking in animals. Am Zool 36: 518-529.
Shafir EB, Osherson DN, Smith EE, 1989. An advantage model of choice. J Behav Dec Making 2: 1-23.
Shafir S, 1994. Intransitivity of preferences in honey-bees: support for "comparative" evaluation of foraging options. Anim Behav 48: 55-67.
Shettleworth SJ, 1998. Cognition, evolution, and behavior. New York: Oxford University Press.
SigmaStat, version 2.0, 1997. Chicago: SPSS Inc.
Sih A, 1992. Prey uncertainty and the balancing of antipredator and feeding needs. Am Nat 139: 1052-1069.[Web of Science]
Simonson I, Tversky A, 1992. Choice in context: tradeoff contrast and extremeness aversion. J Marketing Res 29: 281-295.
Stephens DW, 1989. Variance and the value of information. Am Nat 134: 128-140.[Web of Science]
Stephens DW, Krebs JR, 1986. Foraging theory. Princeton, New Jersey: Princeton University Press.
Strickland RD, 1991. Juvenile dispersal in gray jays: dominant brood member expels siblings from natal territory. Can J Zool 69: 2935-2945.
Strickland D, Ouellet H, 1993. Gray jay. In: The birds of North America, No. 40 (Poole A, Stettenheim P, Gill F, eds). Philadelphia: Philadelphia Academy of Natural Sciences; Washington DC: American Ornithologists' Union.
Tversky A, 1969. Intransitivity of preferences. Psych Rev 76: 31-48.[Web of Science]
Tversky A, Simonson I, 1993. Context-dependent preferences. Manage Sci 39: 1179-1189.
Waite TA, 2001. Intransitive preferences by hoarding gray jays. Behav Ecol Sociobiol (in press).
Waite TA, Field KL, 2000. Erroneous choice and foregone gains in hoarding gray jays. Anim Cog 3: 127-134.
Waite TA, Reeve JD, 1993. Food storage in gray jays: source type and cache dispersion. Ethology 93: 326-336.[Web of Science]
Waite TA, Reeve JD, 1994. Cache dispersion by grey jays sequentially exploiting adjacent sources. Anim Behav 47: 1232-1234.
Waite TA, Strickland RD, 1997. Cooperative breeding in gray jays: philopatric offspring provision juvenile siblings. Condor 99: 523-525.[Web of Science]
Waite TA, Ydenberg RC, 1994a. Shift towards efficiency maximizing by grey jays hoarding in winter. Anim Behav 48: 1466-1468.
Waite TA, Ydenberg RC, 1994b. What currency do scatter-hoarding gray jays maximize? Behav Ecol Sociobiol 34: 43-49.[Web of Science]
Waite TA, Ydenberg RC, 1996. Foraging currencies and the load-size decision of scatter-hoarding grey jays. Anim Behav 51: 903-916.[Web of Science]
Wernerfelt B, 1995. A rational reconstruction of the compromise effect: using market data to infer utilities. J Consumer Res 21: 627-633.[Web of Science]
Ydenberg RC, 1998. Behavioral decisions about foraging and predator avoidance. In: Cognitive ecology: the evolutionary ecology of information processing and decision making (Dukas R, ed). Chicago: University of Chicago Press; 343-378.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N. J. Royle, J. Lindstrom, and N. B. Metcalfe Context-dependent mate choice in relation to social composition in green swordtails Xiphophorus helleri Behav. Ecol., September 1, 2008; 19(5): 998 - 1005. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. I Houston, J. M McNamara, and M. D Steer Do we expect natural selection to produce rational behaviour? Phil Trans R Soc B, September 29, 2007; 362(1485): 1531 - 1543. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Marsh and A. Kacelnik Framing effects and risky decisions in starlings PNAS, March 5, 2002; 99(5): 3352 - 3355. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||