Transcript 174-16-Winter_18_3-M..

Lecture 17:
Measuring
Selection
in the Wild
1
Definitions (again):
Natural Selection = individual variation in
Darwinian fitness that is correlated with variation
in one or more phenotypic traits.
Darwinian Fitness (simply) = number of offspring
left to the next generation by a given individual
(measure at same stage, zygote-to-zygote best,
difficult in practice)
Components of Darwinian Fitness, e.g.,
survivorship, fecundity, # of mates
These are often studied because it is too difficult to measure
lifetime reproductive success.
2
We all know what (potential) selection
looks like, but how do we measure it?
3
Remember:
if all of that mortality depicted on the previous
slide was random with respect to all phenotypes,
then no natural selection is occurring.
Mortality alone does not equal selection.
Differential reproduction alone does not equal
selection.
These just give the potential for selection.
4
An influential paper:
Lande, R., and S. J. Arnold. 1983. The measurement of
selection on correlated characters. Evolution 37:1210-1226.
"Natural selection acts on phenotypes, regardless of their genetic
basis, and produces immediate phenotypic effects within a
generation that can be measured without recourse to principles of
heredity or evolution. In contrast, evolutionary response to
selection, the genetic change that occurs from one generation to
the next, does depend on genetic variation. ... Upon making this
critical distinction ... precise methods can be formulated for the
measurement of phenotypic natural selection."
This verbal definition of selection, inheritance, and evolution
is crucial, because it allows clear operational definitions
of the three things consistent with r = h2s.
Many discussions and definitions of natural selection
confound phenotypic selection and inheritance, so be
careful when you are reading!
5
"It cannot be too strongly argued that the problem
of animal evolution is essentially a statistical
problem ..." Weldon, W. F. R. 1893. On certain correlated variations in
Carcinus moenas. Proc. Roy. Soc. London 54:318-329.
Raphael Weldon was
a pioneer in the
application of
statistics to biology
and a founder of the
journal Biometrika.
Worked with Karl
Pearson.
http://www-groups.dcs.stand.ac.uk/~history/Mathematicians/Weld
on.html
"He was by nature a poet, and these give the best
to science, for they give ideas." (K. Pearson, 1906)
6
"... the questions raised by the Darwinian
hypothesis are purely statistical, and the
statistical method is the only one at present
obvious by which that hypothesis can be
experimentally checked." (Weldon, 1894)
Testing for a correlation between phenotype and
lifetime fitness is best, but we can also get
interesting information by correlating phenotype
with some component of fitness, such as survival
from year to year or during a particular episode.
The basic question:
Are survivors a random sample of the original
population before selection?
7
Young, K. V., E. D. Brodie Jr., E. D. Brodie III. 2004. How the horned lizard got its horns. Science 304:65.
8
Bumpus, H. C. 1899. The elimination of the unfit as illustrated by the
introduced sparrow, Passer domesticus.
Biol. Lectures, Marine Biol. Lab., Woods Hole 6:209-226.
In February, 1898, a severe winter storm with rain, sleet, and
snow occurred near Providence, RI. 136 English sparrows
were found freezing and brought to Dr. Bumpus’ laboratory at
Brown University. 72 survived and 64 died. Bumpus took
advantage of the opportunity to study an episode of natural
selection.
He weighed and made 8 linear measurements of the birds and
compared survivors with non-survivors.
However, he did not employ the standard statistical tests that
we would apply today.
But, he did publish all of his data!
Thank God! (yes, I am being ironic)
9
“WE are so in the habit of referring carelessly to the process of
natural selection, and of invoking its aid whenever some pet theory
seems a little feeble, that we forget we are really using a hypothesis
that still remains unproved, and that specific examples of the
destruction of animals of known physical disability are very
infrequent. Even if the theory of natural selection were as firmly
established as Newton's theory of the attraction of gravity, scientific
method would still require frequent examination of its claims, and
scientific honesty should welcome such examination and insist on
its thoroughness.”
Bumpus, H. C. 1899. The elimination of the unfit as illustrated by the
introduced sparrow, Passer domesticus.
Biol. Lectures, Marine Biol. Lab., Woods Hole 6:209-226.
10
“… it is the purpose of this lecture to show that the birds which
perished, perished not through accident, but because they were
physically disqualified, and that the birds which survived, survived
because they possessed certain physical characters. These
characters enabled them to withstand the intensity of this particular
phase of selective elimination, and distinguish them from their more
unfortunate companions.”
Bumpus, H. C. 1899. The elimination of the unfit as illustrated by the
introduced sparrow, Passer domesticus.
Biol. Lectures, Marine Biol. Lab., Woods Hole 6:209-226.
11
"The birds which perished … are longer than those which endured,
and we are justified in concluding that when nature selects, through
the agency of winter storms of this particular kind of severity,
those sparrows which are relatively short stand a better chance
of surviving." "… the birds which survived invariably average less
in weight than those which perished, and that among the males
this difference amounts to more than a gram" Directional
"The process of selective elimination is most severe with extremely
variable individuals, no matter in what direction the variations may
occur. It is quite as dangerous to be conspicuously above a certain
standard of organic excellence as it is to be conspicuously below the
standard. It is the type that nature favors."
Stabilizing
Bumpus, H. C. 1899. The elimination of the unfit as illustrated by the
introduced sparrow, Passer domesticus.
Biol. Lectures, Marine Biol. Lab., Woods Hole 6:209-226.
12
All Traits log Transformed. P values from 2-tailed tests.
Total
Length
Alar
Extent
Cube root
of Weight
Beak &
Head
Length
Humerus
length
Femur
Length
TibioTarsus
Length
Skull
Width
Sternum
Keel
Length
t-test
comparing
Means
= evidence for Directional Selection
Females
0.320
0.704
0.076
0.853
0.674
0.446
0.236
0.765
0.978
Males
0.00003
0.895
0.010
0.329
0.090
0.151
0.309
0.501
0.158
Levene's
test
comparing
Variances
… but not for Stabilizing Selection
Females
0.175
0.236
0.091
0.410
0.058
0.146
0.079
0.206
0.160
Males
0.940
0.382
0.918
0.870
0.213
0.580
0.948
0.761
0.336
These t-tests do not assume equal variances
13
The correlations among traits make it difficult to know
what selection was really acting upon.
Females, all traits
log transformed
(cube root of weight)
Alar
Extent
Weight
Total Length
0.736
0.589
0.662
0.613
Alar Extent
Weight
Beak & Head Length
Humerus Length
Femur Length
Tibio-Tarsus Length
Beak &
Head Humerus
Length
Length
Femur
Length
TibioTarsus
Length
Sternum
Skull
Keel
Width
Length
0.647
0.478
0.490
0.499
0.590
0.677
0.772
0.586
0.684
0.517
0.533
0.635
0.638
0.460
0.502
0.608
0.554
0.762
0.742
0.703
0.691
0.508
0.810
0.822
0.592
0.592
0.829
0.638
0.477
0.584
0.454
Skull Width
0.558
Pairwise Pearson product-moment correlations. N = 49.
All correlations are significant at P < 0.001 (2-tailed).
14
ln Beak & Head Length (mm)
For example, if only those birds with humeri > -0.3 on the
natural log scale survived, then the survivors would also
have larger beaks & heads.
t-test
2-tailed P < 0.001
Survivor Mean = 3.4656
Overall Mean
= 3.4434
ln Humerus Length (inches)
15
Remember the breeder's equation:
r = h2 s
r = response to selection
h2 = narrow-sense heritability
s = directional selection differential
= difference in mean phenotype between the
original whole population before selection and
the mean of the individuals who actually breed
to produce the next generation
16
Multivariate Selection Theory
DZ
multivariate
response to
selection
a vector
indicating
the crossgenerational
change of
character
means or
traits Z1, Z2,
Z3, etc.
=
G
additive
genetic
variance
covariance
matrix
P-1
s
inverse of
phenotypic
variance
covariance
matrix
vector of
directional
selection
differentials
b
directional
selection
gradient
Lande, R. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain:body size allometry. Evolution 33:402-416.
Lande, R., and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210-1226.
17
b
can be calculated as the vector
of partial regression coefficients
in a multiple regression to predict
fitness.
Fitness = bo + b1Z1 + b2Z2 + b3Z3 + residuals
If all relevant traits have been included in the
analysis, then these describe selection that is
acting directly on the individual traits.
18
Hypothetical Example:
Imagine a population of birds in which we measure clutch
size at fledging, which is a major component of Darwinian
fitness, for 50 nests.
In this species, males help to take care of the young, both
feeding them and protecting the nest.
So, for each nest, we also measure male body mass and
take a blood sample to determine his circulating
testosterone level.
C:\174\PPT 2007 Win\Multiple Regression Example.SPS
19
Clutch Size
Body Mass (grams)
Clutch Size = 0.824 + 0.107 * Body Mass, r2 = 0.187
Standardized Regression Coefficient = 0.433
20
Clutch Size
Testosterone (ng/ml)
Clutch Size = 8.594 - 3.458 * Testosterone, r2 = 0.304
Standardized Regression Coefficient = -0.551
21
Testosterone (ng/ml)
Body Mass (grams)
r = 0.263, 2-tailed P = 0.065
22
Clutch Size = 0.824 + 0.107 * Body Mass, r2 = 0.187
Standardized Regression Coefficient = 0.433
Clutch Size = 8.594 - 3.458 * Testosterone, r2 = 0.304
Standardized Regression Coefficient = -0.551
Multiple Regression
(a 3-dimensional plane):
Clutch Size = 3.429 + 0.154 * Body Mass
- 4.482 * Testosterone
Multiple r2 = 0.662
Standardized Partial Regression Coefficients
Body Mass = 0.621
Testosterone = -0.714
23
Clutch Size = 3.429 + 0.154 * Body Mass - 4.482 * Testosterone
10
8
6
CLTCHRND
4
2
60
50
40
BODYMASS 30
.6
.8
1.0
1.2
1.4
1.6
TESTOST
C:\174\PPT 2007 Win\Multiple Regression Example.SPS
24
Females
B
S.E.
F1,39
P
ZLTOTLEN
-0.21
0.28
0.55
0.4633
ZLALAREX
-0.12
0.32
0.14
0.7125
ZLWEIGHT3
-0.53
0.26
4.21
0.0470
ZLSKULLL
0.04
0.33
0.02
0.9001
ZLHUMER
0.24
0.41
0.34
0.5657
ZLFEMUR
-0.15
0.37
0.17
0.6836
ZLTIBTAR
0.45
0.35
1.66
0.2056
-0.01
0.26
0.00
0.9766
0.19
0.23
0.65
0.4249
B
S.E.
F1,39
P
ZLTOTLEN
-0.52
0.10
26.76
2e-06
ZLALAREX
0.10
0.13
0.64
0.4271
ZLWEIGHT3
-0.27
0.09
8.42
0.0048
ZLSKULLL
0.05
0.09
0.28
0.5994
ZLHUMER
0.16
0.17
0.94
0.3350
ZLFEMUR
0.11
0.17
0.39
0.5326
ZLTIBTAR
-0.01
0.13
0.01
0.9304
ZLSKULLW
0.14
0.09
2.64
0.1085
ZLKEEL
0.16
0.09
3.40
0.0689
ZLSKULLW
ZLKEEL
Males
In Bumpus' sparrow sample,
selection favored shorter
males.
Selection also favored
lighter males.
The P values are lower than
in the univariate t-tests.
The significant negative
selection on female weight
was not detected by the ttests (see next slide).
Bumpus, H. C. 1899. The elimination of the unfit as illustrated
by the introduced sparrow, Passer domesticus.
Biol. Lectures, Marine Biol. Lab., Woods Hole 6:209-226. 25
All Traits log Transformed. 2-tailed tests.
Total
Length
Alar
Extent
Cube root
of Weight
Beak &
Head
Length
Humerus
length
Femur
Length
TibioTarsus
Length
Skull
Width
Sternum
Keel
Length
Females
0.320
0.704
0.076
0.853
0.674
0.446
0.236
0.765
0.978
Males
0.00003
0.895
0.010
0.329
0.090
0.151
0.309
0.501
0.158
Females
0.175
0.236
0.091
0.410
0.058
0.146
0.079
0.206
0.160
Males
0.940
0.382
0.918
0.870
0.213
0.580
0.948
0.761
0.336
t-test
comparing
Means
Levene's
test
comparing
Variances
These t-tests do not assume equal variances
26
Y = Fitness
Stabilizing or disruptive selection can be detected
by including the quadratic terms for traits in the
multiple regression model: Trait A * Trait A
45
40
35
30
25
20
15
10
5
0
2
y = -0.0277x + 1.7674x + 0.834
2
R = 0.3079
0
10
20
30
40
50
X = Trait A
Significant downward curvature indicates
stabilizing selection.
27
Stabilizing or disruptive selection can be detected
by including the quadratic terms for traits in the
multiple regression model:
Trait A * Trait A
Correlational selection can be detected by
including cross-products terms:
Trait A * Trait B
Example:
Brodie, E. D., III. 1992. Correlational selection for color
pattern and antipredator behavior in the garter snake
Thamnophis ordinoides. Evolution 46:1284-1298.
28
Explain garter snakes, measures of performance, reversals, etc.
http://www.californiaherps.com/snakes/pages/t.ordinoides.html
Active in the daytime. Mostly terrestrial,
escaping into vegetation not water when
threatened, but capable of swimming. When
first handled, often releases cloacal contents
and musk, but rarely bites. Escape behavior
of this snake is related to pattern: striped
snakes will escape by crawling away, since
the stripes make it difficult to determine the
snake's speed, while spotted or plain snakes
will crawl, suddenly change direction, then
hold still, as their pattern tends to blend in
with the background. (E. D. Brodie III)
29
30
Correlational selection can be detected by
including cross-products terms:
Trait A * Trait B
31
32
Brodie, E. D., III. 1992. Correlational selection for color
pattern and antipredator behavior in the garter snake
Thamnophis ordinoides. Evolution 46:1284-1298.
"This prediction precisely fits the pattern of
correlational selection observed in the Tenmile
population of T. ordinoides, where the snakes with
the highest probability of survival perform
uninterrupted flight if striped but flee evasively
if spotted or unstriped."
33
This pattern of selection within a population is
consistent with the pattern of differences seen
among species of snakes.
Jackson, J. F., W. Ingram, III, and H. W. Campbell. 1976. The dorsal pigmentation pattern of snakes
as an antipredator strategy: a multivariate approach. American Naturalist 110:1020-1053.
"Irregularly banded and blotched-spotted patterns
are associated with an antipredator strategy of
defense rather than flight, and these patterns likely
function disruptively to minimize initial detection by
predators.
Striped and unicolored-speckled patterns are
associated with antipredator strategies emphasizing
flight more than defense."
34
Husak, J. F. 2006b. Does survival depend on how fast you can run or
how fast you do run? Functional Ecology 20:1080-1086.
"ecological performance" = % of maximal performance exhibited in nature
The Collared Lizard, Crotaphytus collaris
Studied yearling and adult lizards.
Measured:
1. Maximum sprint speed in the laboratory
2. "Foraging" speed in field:
attacking a fishing fly
3. "Escape" speed in field:
approach of human "predator"
35
Husak, J. F. 2006b. Does survival depend on how fast you can run or
how fast you do run? Functional Ecology 20:1080-1086.
Before Selection
After Selection
(survived to
following year)
36
McGlothlin, J. W., J. M. Jawor, and E. D. Ketterson. 2007. Natural
variation in a testosterone-mediated trade-off between mating effort
and parental effort. American Naturalist 170:864-875.
Abstract: Male birds frequently face a
trade-off between acquiring mates and
caring for offspring. Hormone
manipulation studies indicate that
testosterone often mediates this trade-off,
increasing mating effort while decreasing
parental effort. …
These relationships suggest that natural
variation in testosterone, specifically the
production of short-term increases, may
underlie individual variation in the mating
effort/parental effort trade-off. …
Male dark-eyed junco
(Junco hyemalis)
37
McGlothlin, J. W., J. M. Jawor, and E. D. Ketterson. 2007. Natural
variation in a testosterone-mediated trade-off between mating effort
and parental effort. American Naturalist 170:864-875.
Because we were interested in generalized territorial aggression, and
because territorial behaviors were intercorrelated, we extracted a single
principal component to describe response to simulated territorial intrusions.
The first principal component, which described 47% of variance, was loaded
as in table 1 and was used as our measurement of aggression in the
statistical analyses.
38
McGlothlin, J. W., J. M. Jawor, and E. D. Ketterson. 2007. Natural variation in a testosteronemediated trade-off between mating effort
and parental effort.
American Naturalist 170:864-875.
39
Some General Reviews of
Selection in the Wild
40
Endler, J. A. 1986. Natural selection in the wild. Princeton
University Press, Princeton, New Jersey. 336 pp.
Conclusions of Endler's book:
1. Selection intensities in nature often are as strong
as those implemented by animal breeders.
2. Differences in fitness of > 10% are not uncommon
for polymorphic traits.
3. Selection related to survival is generally less strong than
selection related to mating ability, fertility or fecundity.
4. As of yet, few if any cases have quantified selection
acting over the entire lifecycle of an organism (e.g.,
zygote to zygote).
5. If quantitative traits, such as size, shape or metabolic rate
are affected by many genes, then even relatively strong
selection on them may still allow genetic drift to
determine the fate of most mutations at such loci.
41
Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S.
N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli. 2001.
The strength of phenotypic selection in natural populations.
American Naturalist 157:245-261.
Abstract:
1. We tabulated 63 published studies of 62 species that
reported over 2,500 estimates of linear or quadratic
selection. More than 80% of the estimates were for
morphological traits; there is very little data for
behavioral or physiological traits.
2. Most published selection studies were unreplicated
and had sample sizes below 135 individuals, resulting
in low statistical power to detect selection of the
magnitude typically reported for natural populations.
3. The absolute values of linear selection gradients |b |
were exponentially distributed with an overall median
of 0.16, suggesting that strong directional selection
was uncommon.
42
Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S.
N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli. 2001.
The strength of phenotypic selection in natural populations.
American Naturalist 157:245-261.
Abstract:
4. Comparisons of estimated linear selection
gradients and differentials suggest that indirect
components of phenotypic selection were usually
modest relative to direct components.
5. The absolute values of quadratic selection
gradients |g| were exponentially distributed with an
overall median of only 0.10, suggesting that
quadratic selection is typically quite weak.
6. The distribution of g values was symmetric about 0,
providing no evidence that stabilizing selection is
stronger or more common than disruptive selection
in nature.
"gradients" account for trait correlations
43
Kingsolver, J. G., and S. E. Diamond. 2011. Phenotypic selection
in natural populations: what limits directional selection?
American Naturalist 177:346-357.
Abstract:
Our analyses provide little evidence that fitness
trade-offs, correlated selection, or stabilizing
selection strongly constrains the directional
selection reported for most quantitative traits.
44
Extra Slides Follow
This was about 15 min short in 2011, but Gabe handed back the exams.
For 2012, Ted added some Husak slides, and it was still about 10 minutes short.
Husak, J. F. 2006b. Does survival depend on how fast you can run or how fast you do run? Functional Ecology 20:1080-1086.
For 2013, I added three slides from:
McGlothlin, J. W., J. M. Jawor, and E. D. Ketterson. 2007. Natural variation in a testosterone-mediated trade-off between mating effort and
parental effort. American Naturalist 170:864-875.
This was about 10 min short in 2013
This was a couple minutes too long in 2014
This was ~8 minutes SHORT in 2014
Need to add graphs explaining PCA
Could still add some from this, but is complicated story:
Koteja_Natural_Sel_Voles_UCR_27_April_2012_v1.pptx
Calsbeek, R. 2008. An ecological twist on the morphology–performance–fitness axis. Evolutionary Ecology Research 10:197-212. [path analysis
via CALIS procedure in SAS v8]
Calsbeek, R., and D. J. Irschick. 2007. The quick and the dead: Locomotor performance and natural selection in island lizards. Evolution
61:2493-2503. [Correlational selection on limb length, speed, and habitat preference -- check. Endurance is on circular track, and sort of
miscites Garland 1999.]
Hereford, J., Hansen, T. F. & Houle, D. 2004. Comparing strengths of directional selection: How strong is strong? Evolution: 58: 2133-2143.
Irschick, D. J., J. J. Meyers, J. F. Husak, and J.-F. Le Galliard. 2008. How does selection operate on whole-organism functional performance
capacities? A review and synthesis. Evolutionary Ecology Research 10:177-196.
Kingsolver, J. G., and S. E. Diamond. 2011. Phenotypic selection in natural populations: what limits directional selection? American Naturalist
177:346-357.
Abstract: Studies of phenotypic selection document directional selection in many natural populations. What factors reduce total directional
selection and the cumulative evolutionary responses to selection? We combine two data sets for phenotypic selection, representing more than
4,600 distinct estimates of selection from 143 studies, to evaluate the potential roles of fitness trade-offs, indirect (correlated) selection,
temporally varying selection, and stabilizing selection for reducing net directional selection and cumulative responses to selection.We detected
little evidence that trade-offs among different fitness components reduced total directional selection in most study systems. Comparisons of
45
http://en.wikipedia.org/wiki/Natural_selection#mediavie
wer/File:Life_cycle_of_a_sexually_reproducing_organis
46
Miles, D. B. 2004. The race goes to the swift: fitness consequences of variation in
sprint performance in juvenile lizards. Evolutionary Ecology Research 6:63-75.
From November to June
P
< 0.03
= 0.14
47
Miles, D. B. 2004. The race goes to the swift: fitness consequences of variation in
sprint performance in juvenile lizards. Evolutionary Ecology Research 6:63-75.
● Survivors
r = 0.56, P < 0.0001
o Non-survivors
Multiple logistic regression indicates speed, not size, predicts survivorship
48
Miles, D. B. 2004. The race goes to the swift: fitness consequences of variation in
sprint performance in juvenile lizards. Evolutionary Ecology Research 6:63-75.
However, multiple logistic regression with all traits indicated
that stride length was the best predictor of survivorship.
49
Miles, D. B. 2004. The race goes to the swift: fitness consequences of variation in
sprint performance in juvenile lizards. Evolutionary Ecology Research 6:63-75.
"Two key patterns emerged from the analysis of differential
survivorship in juvenile U. ornatus. First, larger individuals
were more likely to survive than smaller individuals.
However, inclusion of the performance data also indicated
that size alone did not completely explain the patterns of
survival. Including body size and burst velocity
simultaneously in a regression analysis resulted in only burst
velocity significantly predicting survivorship. This result
indicates that performance is a better predictor of
survivorship in juvenile lizards than body size. Second, the
selection analyses revealed a complex pattern of linear,
quadratic and correlational selection on locomotor
performance. There was evidence for strong directional
selection on stride length. However, the quadratic selection
analyses revealed a disparate pattern of selection acting on
locomotor performance."
50
545-99.DOC Lecture 10 = "Measuring natural and sexual
selection in the field" -- has Lande-Arnold multivariate
selection theory, phenotypic manipulations, etc.
See also 410-99 Lecture 11 pages 120-122 (is in
41095LEC.DOC)
Use Brodie et al. TREE review paper with figure …
Use Hayes and O'Conner (1999)
51
545-99.DOC Lecture 10
55. TRANSPARENCY = univariate and multivariate equations
(repeat)
Multivariate equation is from Lande (1979):
DZ = G P-1 s = G b
It would be most useful if we could obtain measures of selection
that corresponded directly to s or to b, because this would allow us
to model and predict adaptive evolution, if we could also get
information on inheritance, i.e., the h2 or G.
However, if we just want to demonstrate that selection is occurring,
then all we need to do is demonstrate statistically a correlation
between some aspect of the phenotype and fitness.
We can control for effects of other traits by computing the ...
3. Selection Gradient = b = standardized partial regression
coefficient from a multiple regression of fitness on the various
phenotypes measured
… and much more stuff to add as of Feb. 2007
52
Brodie, E. D., III. 1989. Behavioral modification as a means of reducing the cost of
reproduction. Am. Nat. 134:225-238.
Brodie, E. D., III. 1989. Genetic correlations between morphology and antipredator
behaviour in natural populations of the garter snake Thamnophis ordinoides.
Nature 342:542-543.
Brodie, E. D., III. 1991. Functional and genetic integration of color pattern and
antipredator behavior in the garter snake Thamnophis ordinoides. Unpubl. Ph.D.
Dissertation, Univ. of Chicago. 123 pp.
Brodie, E. D., III. 1992. Correlational selection for color pattern and antipredator
behavior in the garter snake Thamnophis ordinoides. Evolution 46:1284-1298.
Brodie, E. D., III. 1993. Consistency of individual differences in anti-predator
behaviour and colour pattern in the garter snake, Thamnophis ordinoides. Anim.
Behav. 45:851-861.
Brodie, E. D., III. 1993. Homogeneity of the genetic variance-covariance matrix for
antipredator traits in two natural populations of the garter snake Thamnophis
ordinoides. Evolution 44:844-854.
53