1975), is ofwe have failed in severa respects: 1.. We have ignored interesting biological problems and questions. 2. We have not been particularly interested in the consequences of

Introduction

Interindividual variability: an underutilized resource

AL

BE

RT

F.

BE

NN

ET

T

Two principal analytical approaches have been used in studies of organism

al physiology. T

hese are represented by the terms "com

parative physiology" and "physiological ecology." T

he form

er compares functional characters in

two or m

ore populations, species, or higher taxa in an attem

pt to understand m

echanism.

Biological diversity is used to

help understand principles of

physiological design. Often the experim

ental species are chosen specifically because their system

s demonstrate an extrem

e phenomenon o

r because the experim

ental preparation is technically accessible. Th

e selection of a species on

these grounds is known as the K

rogh Principle (Krogh, 1929; K

rebs, 1975), w

hich has been very influential and successful in guiding studies in com

parative physiology for more than fifty years.

The second approach, physiological

ecology or ecological physiology,

examines the physiological attributes of a species and interprets them

in the context of the natural environm

ent or ecological niche of an animal. T

hese studies concentrate on analysis of adaptive pattern, of how

physiology, mor-

phology, and behavior interact to permit survival and reproduction in a given

environment. In this approach, em

phasis is placed on ecological and evolu- tionary aspects of physiological function. M

onitoring the organism in its

natural environment and speculation on selective factors that influenced the

evolution of

characters are the principal

interpretive contexts of

these studies.

These tw

o approaches are by no m

eans exclusive and have often proved com

plementary. T

hey have yielded a substantial understanding of how ani-

mals w

ork and function in the natural world. H

owever, m

y thesis here is that both approaches have overlooked a valuable source of inform

ation. In their concentration on population-, species-, or higher-level phenom

ena, [hey have failed to

analyze and take advantage of biological

differences am

ong individuals. As traditionally practiced, physiological studies neglect

differences among individual anim

als and attempt to

describe the functional response in the average anim

al of the group. I believe that this approach has been very short-sighted and that the study of interindividual differences has

147

much to contribute to both com

parative physiology and physiological ecol- ogy. I w

ill argue that the analysis of the bases and consequences of interin- dividual variability can provide new

tools for both types of physiological

analysis. I believe that it is also capable of building

new and im

portant bridges to other allied fields of biology, especially ecology, ethology, evolu- tion, and genetics.

Th

e tyrann

y of the Go

lden

Mean

Th

e framew

ork of physiological studies implicitly em

phasizes the descrip- tion and analysis of central tendency. D

epending on the data, this involves the calculation of m

ean values or the development of least-squares regression

equations. After these values are determ

ined, they take on a life of their own

and become the only

point of

analysis and com

parison. Th

e complete

breadth of biological variation determined in the investigation then is for-

gotten. Measures of variability (e.g., variance, standard deviation) are calcu-

lated and reported only to stipulate confidence limits about the m

ean or slope of the regression line. G

roups are then compared to determ

ine whether they

are different from one another o

r from hypothesized values. T

he variability inherent in the original data is seen only as "noise,"

through which the

"true" value of the central tendency can be glimpsed w

ith appropriate statis- tical techniques.

This assum

ption of a "true" or "real"

central tendency, which biological

reality only approxim

ates, stems from

P

latonic philosophical

traditions. T

hese maintain that ideal archetypes exist that can be perceived only im

per- fectly through perceptual sensation. T

he concept of an ideal form of a struc-

ture or process w

as central to the thinking of medical physiologists of post-

Renaissance E

urope and heavily influenced the functional biologists of the nineteenth century. T

hese physiologists and morphologists, in their search

for proximate causation, m

aintained a typological approach to experimen-

tation and analysis and were largely unaffected by contem

poraneous devel- opm

ents in evolutionary biology and genetics (cf. Mayr, 1982, for a m

ore detailed discussion). A

nalysis of variability played an important role in these

latter fields, but it w

as ignored by functional biologists at the tim

e and rem

ains largely unexplored by them even today.

To

dispel any doubt that analysis of central tendency and neglect of vari- ability is the dom

inant or exclusive analytical m

ode in organismal physiol-

ogy, I reviewed all papers published during 1985 in the Journal of C

ompar-

ative Physiology, the Journal of E

xperimental B

iology, and Physiological Z

oology. These are som

e of the best and most forw

ard-looking journals in the field. N

early all the articles reported mean values or regression equations

and did statistical analyses. H

owever, less than

5%

of the articles even

reported the range of values of the data obtained, and out of more than 250

BUR

ST SPEED (crnls)

DISTAN

CE C

RAW

LED (m

)

FIG

UR

E 7.1

Frequency distributions of burst speed and total distance craw

led under pursuit by individual newborn garter snakes (T

hamno-

phis radix). Each individual observation is the mean of tw

o trials con- ducted on tw

o successive days; individual repeatability is highly signif- icant (r =

0.60 for burst speed and 0.55 for distance; p < .001).

Distance craw

led is reported on a logarithmic axis. (D

ata from A

rnold and B

ennett, in press.)

articles, only one (Taigen and W

ells, 1985) analytically examined the varia-

bility in the observations. T

he concentration on central tendency has been and will continue to be

very useful in testing certain hypotheses, but it has distracted us from an

examination of

the causes and consequences of biological variability. An

example of this variability is given in Figure 7.1, in this case variability in

locomotor perform

ance capacity of newborn garter snakes. M

aximal burst

speed and the total distance crawled under pursuit w

ere measured in nearly

150 laboratory-born animals shortly after birth (A

rnold and Bennett,

in press). T

hese behaviors are individually repeatable (see below) and represent

the breadth of response of the population at birth, before natural selection by the external environm

ent has had the opportunity to act. B

oth these per- form

ance measures show

strong central tendencies, but they also show enor-

mous interindividual variability. T

he fastest snake has a burst speed ten times

that of the slowest; the endurance of som

e individuals is more than tw

enty tim

es that of others. Assum

ing for a mom

ent that these individual differences are real (see below

), these observations imm

ediately suggest two sorts of ques-

tions. First, what is the functional basis of these individual perform

ance dif- ferences? W

hich physiological or m

orphological factors make a fast snake

fast and which account for the relatively low

stamina of som

e other animals?

Second, w

hat are the ecological and evolutionary consequences of these dif- ferences? Is there differential survivorship or grow

th under natural condi- tions based on locom

otor performance capacities? T

hese questions reflect the som

ewhat artificial dichotom

y raised earlier between com

parative physiol- ogy and physiological ecology, but both o

f them reflect com

pelling questions of general biological interest. T

hey are obscured, however, if one concen-

trates only on central tendency. This is the tyranny of the G

olden Mean: it

restricts our vision of the data and narrows our conceptual fram

ework so

that we cannot take advantage of all the analytical possibilities of biologica

variability. T

he failure to consider interindividual variability is not that of ecologica

or comparative physiology alone. A

lmost identical com

ments and com

pari sons could be m

ade about any other field of organismal biology.

In our concentration

on central

tendency, w

e have

failed in

severa respects:

1.. We have ignored interesting biological problem

s and questions. 2. W

e have not been particularly interested in the consequences of the dati w

e have gathered for survivorship or fitness. 3. W

e have failed to utilize the breadth of our data in assessm

ent of physio logical hypotheses.

4. We have failed to provide sufficient inform

ation in our research report that w

ould permit others to

analyze biological variability.

The reality o

f interindividual variability

I believe rhat part of the difficulty rhat most ecological and com

parative phys iologists have in reporting and utilizing variability is a suspicion of its realit] and inform

ation content. Biological m

easurements are inherently highly var

iable as compared to those m

ade by physicists or chemists. C

oefficients o variation of 20 to 30%

, values that would cause a physical scientist to blanch

are routine in most physiological m

easurements. T

o w

hat extent, however

is this variability real and useful? It seems to m

e that there are three potentia objections to its use:

I. Extrem

e values are atypical or abnorm

al and do

not reflect the trut response of m

ost individuals.

This view

is essentially a restatement of the typological concept: the aver

age is the real. Extrem

e performance certainly is "atypical"

and "abnormal'

in the strict sense of the words, but that does not m

ean that it is not real. A physiologist m

ust be sure that experimental anim

als are in good condition, but it shouId go w

ithout saying that one must have external cause to doubt

any data point. It cannot be questioned only because it happens to lie on the extrem

e of the range.

This view

suggests that the experimenter has m

ore confidence in values that lie closer to

the mean than those at the extrem

es. If this is the case, then not all points should receive equal w

eighting: those closer to the m

ean should be w

eighted more highly. T

he circularity of this logic is apparent. F

urther, norm

al parametric statistics are inappropriate in such a circum

stance. Either

all data points receive equal confidence and equal weight, o

r the analytical m

ethods we norm

ally use are inapplicable; one cannot have it both ways.

2. Observed variability

is due to instrum

entation or procedural error; the

observed range does not result from real biological differences but from

inaccuracies in experim

ental setups or procedures.

According to

the type of m

easurement, this objeccion

may

have some

validity. How

ever, the precision of modern physiological equipm

ent is typi- cally less than

1% and is consequently a doubtful explanation of

much

higher apparent biological variability. Further, if such errors are felt to

be im

portant, their magnitude m

ust be quantified and analyzed (although they alm

ost never are) even in studies that are interested only in central tendency. If the errors are random

, then the mean values w

ill be correct, but the mea-

surements of variance and standard deviation of the m

eans will be inflated.

As statistical com

parisons between groups are dependent on the extent of

intragroup variability, incorrect judgments m

ay be made if experim

ental or

instrumentation error is not analyzed and rem

oved. Consequently, if this type

of error is a problem, it is not a special problem

in the'analysis of variability alone. It also affects any kind of analysis, including that of central tendency.

3. The variation

measured is

real but reflects random

and

unrepeatable responses of individuals;

that is, intraindividual variability is so high that there is n

o significant interindividual com

ponent to total variance.

This is by far the m

ost serious potential objeccion to the analysis of vari-

ability: if the responses are random w

ith respect to individuals, then analyz-

ing the differences among individuals is futile. T

he m

easurements required

to demonstrate w

hether this is an important problem

are a series of repeated observations on the sam

e individuals and analysis of the significance of the individual com

ponent. For instance, if one is interested in oxygen transport

capacity, one might m

easure maxim

al oxygen consumption in each of several

individuals on

sequential days to determ

ine whether som

e individuals have consistently high o

r low capacities.

Given the general lack of interest in interindividual variability, analyses of

intra- versus interindividual variability are relatively few

in ecological or

comparative physiological studies. M

ost of these relate to data on locom

otor perform

ance capacity, and many of the exam

ples in this discussion will be

drawn from

this area. Individual locomotor perform

ance ability has a sig- nificant repeatable interindividual com

ponent in every study in which it has

15

2

AL

BE

KT

F

. B

EN

NE

TT

TABLE 7.1 Studies dem

onstrating significant interindividual variability in locom

otor performance

- -

Gro

up

P

erformance

No

. of species

Lizards burst speed

6" 1 2c

2d

1'

stamina

6" 1' ld

defensive b

eh

avio

r 1"

Snakes

burst speed In 1

stamina

1

defensive behavior 1

' A

nurans stam

ina 2'

"ennett (1 980).

bC

row

ley and P

ietruszka (19

83

). 'H

ue

y an

d H

ertz (1 98

4).

dGarland (1 984, 1985).

'Crow

ley (1 985).

'Joh

n-A

lde

r (1 984). G

arla

nd

and Arn

old

(1983). h

~rn

old

and B

ennett (in press).

'Arn

old

an

d B

en

ne

tt (1984). 'P

utnam and B

ennett (1981).

been examined (T

able 7.1). An exam

ple of individual constancy of day-to-day differences in locom

otor performance is given in Figure 7.2 (B

ennett, 1980). M

aximal burst speed w

as measured in fifteen adult fence lizards on five

sequential days. Rank order of perform

ance was conserved through the re-

petitive trials (p < .001). T

hese individual differences in burst speed capacity w

ere independent of both sex and body mass, Sim

ilarly, individual perfor m

ance rank is stable even when the internal environm

ent of the animals is

grossly altered, as during changes in body temperature. Individual rankings

of burst speed performance of alligator lizards at different body tem

peratures are given in T

able 7.2. Again, individual differences are highly significant (P

< .O

OI): som

e animals are fast and som

e are slow at all body tem

peratures (see also H

uey and Hertz, 1984).

I believe that locomotor m

easurements w

ould a priori be among the least

repeatable of any of the potential spectrum of "physiological"

measurem

ents. T

hey may be influenced by a great m

any motivational and psychological fac-

tors, as well as differences in underlying physiological

or m

orphological

Fl G U

RE

7.2 R

ank order performance of burst speed in fifteen adult

fence lizards (Sceloporus occidentalis) measured on five successive

days. Rank 1 is the fastest anim

al, rank 15

is the slowest. D

ots indicate rank perform

ance on each day; vertical bars, range; horizontal bars, m

ean rank. Individual ranking effects are highly significant (p < .001

by Kendall's coefficient of concordance). (D

ata from B

ennett, 1980,

and unpublished observations.)

capacity. From

that viewpoint, a significant interindividual com

ponent in m

easurements of locom

otor capacity might suggest that m

any other physi- ological variables w

ould also have individual fidelity. Significant interindividual

differences have been

demonstrated in such

diverse systems and m

easures as maxim

al oxygen consumption in am

phibi- ans and lizards (P

ough and Andrew

s, 1984; Wells and T

aigen, 1984; Sullivan and W

alsberg, 19853, enzymatic activities in fruit flies (L

aurie-Ahlberg et al.,

1980), cuticular water loss in cicadas (T

oolson, 19841, muscular m

orphology of birds (B

erman, C

ibischino, Dellaripa, and M

ontren, 19851, skeletal mor-

phology of salamanders (H

anken, 19831, kinematics and m

uscle activity pat- terns during feeding in salam

anders (Shaffer and Lauder, 1985a, 1985b), for-

aging tactics in fish (R

ingler, 1983), food preferences in snakes (Arnold,

1981), and regulated body temperatures of lizards (C

hristian, Tracy, and Por-

ter, 1985).

15

4

AL

BE

RT

F

. B

EN

NE

TT

TA

BL

E 7.2 R

ank order of burst speed at different body temperatures in tw

elve individual alligator lizards (C

errhonotus multicarinatus)

Ind

ividu

al

Tem

perature ("a

A

B C

DE

F

GH

I

1 K

L

10

8

7 3

11

4

9.5 1

9.5 2

5 1

2

6 1

5

85

11

13

6

21

07

91

24

2

0

7 6

2 11

3 8.5

5 8.5

1 4

12 10

2 5 9

82

11

3

4 6

7 1

12

1

05

3

0

6 9

2 1

2

7.5 5

4 1

0

1 7.5

11

3

35

8

10

7

11

5.5

2 4

9 1

5.5 1

2

3 37.5

9 8

7 1

1

6 3.5

1 1

0

2 3.5

12 5

Note: R

ank 1 = fastest; p <

,001 by Kendall's coefficient o

f concordance. S

ource: Bennett (1 980).

In my opinion, the large m

ajority of physiological variables that can be sam

pled repeatedly will show

real and significant interindividual variation. T

he question then becom

es how w

e can utilize this variability to our benefit

in asking analytical questions.

Th

e analytical u

tility of in

terind

ividu

al variation

I suggest four different types of studies in which the exploration of interin-

dividual variability might play a crucial role. Som

e represent new sorts of

investigations for ecological or com

parative physiology. Others perm

it a new

approach to both current and classical questions in the fields.

The testing of correlative hypotheses

A com

mon analytical approach in com

parative physiology is to m

easure the correlation betw

een two

or m

ore variables in two o

r more groups (e.g., pop-

ulations, species) and to infer mechanistic relationships if significant corre-

lations exist. For example, if positive associations are found betw

een the length of the loops of H

enle in kidneys of various mam

mals and their ability

to concentrate urine,. one m

ight conclude that these may be functionally

linked. These correlational exam

inations have been central in building the field of com

parative physiology. They have, how

ever, been criticized for their failure to

take into account the phylogenetic history of the experimental ani-

mals involved (G

ould and Lew

ontin, 1979; Felsenstein, 1985; Chapter 4).

A com

panion approach to interspecific analyses is the exam

ination of interindividual correlations of

variables w

ithin a species. This approach

maintains the benefits of com

parative analysis without som

e of the objections associated w

ith using organisms that are distantly related phylogenetically

(see Chapter 4). If tw

o factors are functionally related, they should be signif- icantly correlated am

ong individuals within a species. In fact, if evolutionary

or functional trends are proposed, the argument is strengthened if intraspe-

cific associations can be demonstrated, because selection on traits w

ithin populations m

ust be the ultimate source of adaptation.

The experim

ental protocol required to investigate interindividual variabil- ity is sim

ilar to that of interspecific comparative studies, except that obser-

vations on functional traits of interest must be m

ade on the same individuals

and analyzed on that basis. The researcher then correlates one trait w

ith the other to determ

ine whether they are positively o

r negatively associated. If so, the hypothesis of functional relationships am

ong the traits is supported, and further experim

entation can be planned to explore the nature of the rela-

tionship (see Huey and B

ennett, 1986). If no

significant association is found, then the traits are not functionally linked and the hypothesis is rejected.

One im

portant step in this analysis is the determination of the dependence

of the traits in question on body size (mass) and the elim

ination of such a dependence in the analysis.

So many m

orphological, physiological, and

behavioral traits are dependent on body size (see Calder, 1984; Schm

idt-Niel-

sen, 1984) that it is very easy to obtain positive correlations among otherw

ise unrelated traits because of their m

utual dependence on mass (see A

ppendix for a further discussion and exam

ple). Allom

etric analyses should be per- form

ed (see Chapter 10) and, if m

ass effects are significant, the mass-corrected

residuals'should be analyzed for correlation. A

n illustrative example of the use of interindividual variability in testing

correlative hypotheses may be beneficial here. T

hese data are drawn from

som

e observations on the skeletal muscle physiology and locom

otor perfor- m

ance of tiger salamanders (E

lse and Bennett, 1987, and unpublished obser-

vations). Close (1964,1965) proposed a correlation betw

een the speed of iso- m

etric and isotonic contractions of skeletal muscle: the m

aximal velocity of

shortening (isotonic) is supposed to be positively related to

the rate of tension developm

ent in an isometric tw

itch or tetanus. T

his proposal is a straight- forw

ard mechanistic linkage that is supported by interspecific com

parative studies. W

e can test this hypothetical connection by making observations of

all these factors on individual animals and determ

ining whether they are

associated within individuals. A

further correlation that might also be inves-

tigated is the association between m

uscle contractile speed and locomotor

speed: are the animals that have the greatest intrinsic speed of m

uscle con- traction also the fastest? First, all variables are m

ass-corrected and the resid- uals are then correlated w

ith each other in Table 7.3. C

orrelations are sig- nificant am

ong isometric variables and betw

een isotonic variables, but no associations are significant betw

een any isotonic and isometric variable nor

- betw

een burst speed and any measure of m

uscle contractile performance.

TA

BL

E 7.3

Co

rrela

tion

coe

fficien

ts (r) am

on

g m

ass-co

rrecte

d re

sidu

als o

f loc

om

oto

r pe

rform

an

ce

an

d m

uscle

con

tractile

fa

ctors in

the

sala

ma

nd

er A

mb

ystom

a tig

rinu

rn n

eb

ulo

sum

at 20

"C (n

= 20)

Isometric m

uscle factors Isotonic m

uscle factors

Locomotion

Tetanic

Tw

itch

Maxim

al M

aximal

(burst swim

T

etanic T

witch

contraction

contraction rate o

f pow

er speed)

force force

rate rate

shortening output

Burst ru

n speed

.13 -

Burst sw

im speed

Tetanic force

Tw

itch force

Tetanic contraction

rate T

witch

con

tractio

n

rate M

axim

al rate o

f .71*

shortening

Note: A

sterisks indicate significant correlations (r > 0.56, p

4 0.01).

Source: U

np

ub

lishe

d data of A

. F. Bennett, P

. L. Else, and T. G

arland.

INT

ER

IND

IVID

UA

L V

AR

IAB

ILIT

Y

15

7

These results argue against any necessary m

echanistic association among

these factors. T

his is only one example of an approach that can be utilized in m

any dif- ferent physiological o

r functional studies. For instance, the role of m

aximal

heart rate in limiting m

aximal oxygen consum

ption or that of a particular

muscle in generating force during feeding o

r locomotion could be investi-

gated using an appropriate analysis of interindividual variability.

Exam

ining the functional bases of organismal or physiological variables

Another use that can be m

ade of interindividual variability is the determi-

nation of which of a potential suite of characters m

ight influence perfor- m

ance at a higher level of biological organization. This is a m

ultivariate sta- tistical approach based

on an array of characters m

easured in

identified individuals of a species. T

he researcher measures a perform

ance variable, such as burst speed o

r lower critical tem

perature, and a number of m

orpho- logical an

d/o

r physiological predictor variables that might reasonably be

associated with it (e.g., lim

b length and maxim

al velocity of muscle short-

ening, or fur density and body tem

perature, respectively). All these m

easure- m

ents are made on the sam

e series of individuals. Mass dependence of any

of the factors is analyzed and removed, as discussed previously. T

hen step- w

ise multiple regression analysis (or another appropriate technique, such as

canonical correlation) is used to determine w

hich, if any, of the predictor variables are associated w

ith the performance variables.

An exam

ple of this approach is provided by the study of Garland (1984)

on locomotor perform

ance by a lizard, Ctenosaura sim

ilis. Endurance, burst

speed, and maxim

al distance run under pursuit were m

easured in a series of individuals, along w

ith a variety of physiological and morphological vari-

ables, including body mass and length; standard and m

aximal rates of oxygen

consumption and carbon dioxide production; m

ass of thigh muscle, heart,

and liver; hematocrit and hem

oglobin concentration of the blood; myofibril-

lar AT

Pase activity of thigh m

uscle; and activities of three selected metabolic

enzymes in heart, liver, and skeletal m

uscle tissue. Body m

ass effects were

removed by regressing all variables o

n m

ass and analyzing only mass-cor-

rected residuals. Each m

easure of locomotor perform

ance was then regressed

as a dependent variable on the suite of morphological and physiological char-

acters as independent variables. Th

e results of these analyses are given in T

able 7.4. N

early 90% of the m

ass-corrected interindividual variation in endurance could be attributed to

four predictive factors, including maxim

al oxygen consum

ption, skeletal muscle and heart m

ass, and hepatic aerobic enzym

e activity. This is a rem

arkable amount of predictive pow

er. More than

half the variation in maxim

al distance run is correlated with m

aximal carbon

dioxide production and anaerobic enzyme activity of the skeletal m

uscle. N

one of the variables measured in this study w

as significantly associated with

15

8

AL

BE

RT

F

. B

EN

NE

TT

TA

BL

E 7.4 Stepw

ise multiple regression analysis of locom

otor performance of

the lizard Ctenosaura sim

ilis

Perform

ance V

ariable P

artial R2

Endurance

Th

igh

muscle m

ass 0.540

Ma

xima

l oxygen consumption

0.187 H

ea

rl mass

0.0

86

Liver aerobic enzym

e ac

tiv~

~y

O

.OUO

Total

0.893 (p

< .0001)

Distance ru

n

Ma

xima

l carbon dio

xide

pro

du

ction

0.405

Thigh anaerobic enzym

e activity 0.1 77

Total

0.582 (p

= .0022)

Burst speed

No

ne

N

.5.

Source: G

arland (1 984).

burst speed. Thus, a m

ultivariate statistical approach does not necessarily find a significant association am

ong any set of variables. It may uncover strong

correlations (as in the case of endurance) or no correlation (as with burst

speed). A subsequent investigation on another species of lizard found signif-

icant interindividual correlations between burst speed and glycolytic enzy-

matic activity of skeletal m

uscle and an inverse relationship between burst

speed and muscle fiber diam

eter (T. G

leeson, unpublished data). A

m

ultivariate statistical approach can be

particularly powerful w

hen num

erous underlying variables might be expected to

influence higher-level perform

ance. It can help to single out the most significant factors from

an entire array and allow

a researcher to concentrate further on those. T

he result of the analysis m

ay serve to confirm a priori associations or m

ay sug- gest entirely unexpected linkages that can be explored further. T

his multi-

variate analysis should be regarded as a first-stage approach, to be followed

by more detailed com

parative and experimental research o

n the factors iden-

tified with this technique. T

hese further studies hay

also take advantage of interindividual variability.

Measurem

ent of selective importance of traits under field or

experimental conditions

Physiological ecologists and comparative physiologists usually assum

e that the traits that they study are of

adaptive significance, that is, that they enhance survivorship and reproductive potential. T

his assumption is, how

- ever, alm

ost never tested directly (Arnold, 1983; E

ndler, 1986). Using inter-

individual variability, one can evaluate whether perform

ance of any given

physiological or organism

al trait is in fact correlated with differential sur-

vivorship under natural conditions. Th

e observations required involve scor- ing a trait on a large num

ber of individual animals, releasing them

into their natural environm

ent, and recapturing the survivors after exposure to this

environment. T

he survivors are then exam

ined to determ

ine whether they

are drawn from

any subset of the original distribution. Selection m

ight operate in a number of w

ays to favor different portions of the original distribution of the trait (Sim

pson, 1953; Lande and A

rnold, 1983; E

ndler, 1986). It might be directional and favor individuals at one end of the

range of variability. For example, d

o birds w

ith greater insulation survive better during the w

inter or do caterpillars that eat m

ore metam

orphose more

rapidly and successfully? Do

newborn snakes that are very fast or have a high

endurance (see Figure 7.1) accrue an advantage under natural conditions such that they are m

ore likely to survive to reproductive age? Selection m

ay also be stabilizing, favoring anim

als with m

odal values for a given trait, thereby reducing variability and reinforcing central tendency in the population. In these cases, both very w

ell and poorly insulated birds, caterpillars with both

large and small appetites, and very fast and very slow

snakes would be

selected against. Selection might also be disruptive, favoring anim

als at both extrem

es of the distribution and tending to increase overall variability. T

he

null hypothesis against which the presence of selection m

ust be tested is the absence of any detectable effect of the variable on such indices of fitness as survivorship, grow

th, or reproduction. In the exam

ples above, variability in plum

age quality, feeding capacity, or speed would have no detectable influ-

ence on fitness under field conditions. T

his correspondence betw

een physiological or perform

ance characters

and survivorship or fitness under field conditions is termed the "fitness gra-

dient" (Arnold, 1983). Its determ

ination is judged to be essential for the char- acterization of the ecological and evolutionary im

plications of any physio- logical variable. H

owever, com

prehensive studies of the fitness gradient have rarely been attem

pted for any variable. The effects of natural selection on

physiological variation generally are unknown (E

ndler, 1986). A lack of cor-

respondence between m

aximal oxygen consum

ption and some m

easures of reproductive perform

ance has been reported in adult male toads (W

ells and T

aigen, 1984; Sullivan and Walsberg, 1985), but its effect on differential sur-

vivorship up to adulthood has not been measured. Studies on the effects of

locomotor perform

ance on postnatal survivorship are currently underway on

fence lizards (R. H

uey, University of W

ashington) and garter snakes (my lab-

oratory). Th

e direct measurem

ent of the impact of a character on perfor-

mance under natural conditions, in spite of its obvious im

portance to field

ecology and evolutionary biology, is almost unexplored. It m

ay be opera- tionally difficult o

r even impossible on som

e types of organisms, but I believe

it is in fact feasible for many different types of anim

als in many different

environments.

This approach has great potential to m

easure the importance of selection

on traits in natural populations in natural environments. It also can be used

in situations in which the environm

ent has been experimentally altered. In

this case, the response of the trait in the population can be compared to

a priori

expectations about the effect of

such alteration. F

or instance, one m

ight remove predators and determ

ine whether burst speed o

r endurance declines in a population in the absence of this particular selective agent. O

ne m

ight supplement anim

als living in saline ponds or deserts w

ith fresh water

and investigate whether osm

otic tolerance or fluid-concentrating capacity

changes as a result of altered environmental circum

stances. An

excellent exam

ple of this experimental approach is provided by the study of F

erguson and F

ox (1984). A com

bination of studies, examining responses of popula-

tions in both natural and experimentally m

anipulated environments, has a

great deal of potential to help us understand the im

portance of various phys- iological processes to

total fitness of organisms. T

his approach presents a protocol for testing assum

ptions about adaptation, not simply asserting them

axiom

atically. I believe this is one of the most exciting new

developments

and directions for physiological ecology as a field.

De

term

ina

tion

of heritabilities o

f organismal o

r physiological characters F

or adaptation and evolution of a trait to occur, it m

ust have a genetic basis. W

ithout a heritable basis, selection on a trait within each generation w

ill not influence the variability o

r distribution of the trait in ensuing ger,erations. It is necessary, for exam

ple, for fast parents to have fast offspring if the popu-

lation is to respond to

a new agent that selects against slow

er individuals. S

tudies of the heritnbility of physiologicnl traits are n valunble supplement to

ecological studies because they permit the determ

ination of both the poten- tial of the trait to

evolve and the rapidity with w

hich the response can occur. S

ome progress has been m

ade in particular systems in identifying effects

of individual loci on organism

al physiology and perform

ance (e.g., Watt,

1977, 1983; DiM

ichele and Pow

ers, 1982; Chappell and S

nyder, 1984; Barnes

and Laurie-A

hlberg, 1986; Chapters 5 and 8). W

hile individual loci may have

identifiable effects, many of the traits of interest to

a physiological ecologist w

ill be under multilocus control. C

onsequently, the techniques of quantita- tive genetics w

ill be the most appropriate for exam

ining the inheritance of these characters (see F

alconer, 1981, and Chapter 9 for a general discussion

of the field and appropriate methodology). T

echniques involve examining the

similarities of traits in parents and offspring an

dlo

r among the offspring of

given parents. They require that the organism

s in question can be bred suc-

cessfully in the laboratory or that gravid fem

ales can be obtained that will

deliver offspring in the laboratory. Few

studies have examined the heritability of physiological

or perfor-

mance characters. M

ost of them have dealt w

ith the inheritance of locomo-

tor performance. Significant heritabilities have been found for speed in race

horses (Langlois, 1980; T

olley, Notter, and M

arlowe, 1983), speed in hum

ans (B

ouchard and Malina, 1983a, 1983b), burst speed and stam

ina in lizards (van B

erkum and T

suji, in press; R. B

. Huey, unpublished data) and snakes (S. J.

ArnoId and A

. F. Bennett, unpublished data; T

. Garland, unpublished data).

Defensive behaviors in snakes are also heritable (A

rnold and Bennett, 1984).

In these locomotor studies, a m

inimum

of 30 to 5

0%

of the variability among

individuals is genetic. Other types of

physiological traits have also been

found to be heritable: for exam

ple, growth rate and efficiency in pigs (Sm

ith, K

ing, and Gilbert, 1962), reproductive output of chickens (E

msley, D

icker- son, and K

ashyap, 1977), and thermoregulatory behaviors of m

ice (Lacy and

Lynch, 1979). O

bservations are so few at this point that a case m

ay be made

for a general investigation of t.he topic of heritability per se of physiological system

s in different types of animals. Future studies m

ay concentrate on more

specific genetic issues concerning this inheritance, but, given the multilocus

nature of these traits, these are bound to be m

ore difficult.

Conclusions

Interindividual physiological variability is rarely studied. How

ever, this vari- ation is real and repeatable in m

any physiological traits. I believe that the analysis of the causes and consequences of interindividual variability has m

ajor promise as an analytical tool in physiological studies. E

cological and com

parative physiology have often been characterized as major branches of

organismal biology, but their view

of the organism has been ideal o

r typo- logical. It has been that of the nonexistent anim

al that possesses the average value of all physiological, m

orphological, and behavioral attributes of the population. Such anim

als do

not exist. Real individuals are unique com

bi- nations of traits, som

e above and some below

average. It is time to

recognize the uniqueness of the individual and to

turn it to our advantage as biologists.

The analysis of variation can be useful in studies on physiological corre-

lation and mechanism

, on the importance of the variable to

fitness under natural conditions, and on the potential for inheritance of the trait, w

ith the consequent possibility of its adaptation and evolution. I d

o not suggest that

the study of variation should supplant other approaches nor that it is even feasible for all physiological variables. B

ut where such study is applicable, it

can be a powerful analytical tool, for analysis of both m

echanism and adap-

tation. Its particular advantage is that it can pull together so many different

aspects of biology, not only physiology and ecology, but also behavior, mor-

phology, population biology, and evolution. Biologists often treat these as

different areas, but of course individual organisms d

o not m

ake these arbi- trary divisions and distinctions. T

hey react to problem

s and opportunities as integrated organism

s. Appreciating and studying individual differences can

be a synthetic approach that puts the individual organism back into organ-

ismal biology and gives us a m

uch broader understanding of animals and

their evolution.

Ap

pen

dix

Th

e problem of m

ass-dependent correlation is so ubiquitous in correlation analysis that I w

ill provide an illustrative example of the utility of the analysis

of residuals. In Figure 7.3a, two physiological traits are found to be positively

associated when one is plotted as a function of the other. T

hese might, for

example, be length of a M

alphigian tubule and secretion rate in an insect, or tidal volum

e and anatomical dead space during ventilation in a m

amm

al. S

uch a result might lead one to conclude that the traits are positjvely linked

functionally. If, however, both traits are plotted as a function of body m

ass (F

igures 7.3b and 7.3

4 each is also found to

be strongly and positively mass

dependent. Are the traits truly linked to each other o

r is their apparent asso- ciation due to

their mutual but independent relationship to

body mass? T

his size influence m

ay be removed by exam

ining the deviation of each data point from

the mass regression line (i.e., the residuals of the regression). If these

mass-corrected residuals are then plotted against each other (F

igure 7.3d), their relationship can be exam

ined without the interfering effects of body

size. In the case illustrated, the traits are found to be negatively related to

each other, which is exactly the opposite of the original conclusion based on

Figure 7.3a. T

heir apparent positive association was due only to

their mutual

correlation with body m

ass. This w

as, of course, a contrived example: the

residuals might also have been positively associated o

r not significantly cor- related w

ith each other. Th

e point is that an examination of the original,

uncorrected data in Figure 7.3a w

ould not have permitted this assessm

ent.

Ackn

ow

ledg

men

ts 1

I thank S. J. Arnold, M

. E. Feder, T. G

arland, R. B. H

uey, G. V

. Lauder, H

. B. Shaffer, and C

. R. T

aylor for helpful discussions andlor comm

ents on the manuscript. Support

for the workshop w

as provided by NSF G

rant BSR 86-07794. Support for the author's research cited herein is from

NSF G

rants BSR 86-00066, DC

B 85-02218, D

EB

81- 14656, and P

CM

81-02331.

- 16

0 '

260 '

3b0 4

b0

40

TRA

IT 2 M

ASS

Fi GU

RE

7.3 A hypothetical exam

ple of the effect of mass-correlated

effects on

the apparent association between tw

o traits. Data are

reported for two traits and body m

ass in arbitrary units for ten individ- uals (A

through J) along with least-squares regression lines. (a) T

he two

traits are positively and significantly correlated when they are related

to each other directly. (b) and (c) Each trait is positively m

ass-corre- lated. (d) M

ass effects are removed by calculating residuals, that is, the

difference between the observed value for each individual and the

value predicted by the m

ass regression. These are plotted against each

other, demonstrating a significant negative association betw

een the

traits after the confounding effects of m

ass are eliminated. N

ote that opposite conclusions about th

e relationship between the traits w

ould b

e derived from (a) and (d).

References

Arnold, S. J. (1981) B

ehavioral variation in natural populations. I. Phenotypic, genetic and environm

ental correlations between chem

oreceptive responses to prey in the garter snake, T

hamnophis elegans. E

volution 35489-509. A

rnold, 5. J. (1983) Morphology, perform

ance and fitness. Am

. Zool. 23:347-361.

Arnold, 5. J., and B

ennett, A. F. (1984) B

ehavioural variation in natura1 populations. 111. A

ntipredator displays in the garter snake Tham

nophis radix. Anim

. Behav.

32:llO8-lll8.

Arnold, 5. J., and B

ennett, A. F. (in press) B

ehavioural variation in natural popula- tions. V

. Morphological correlates of locom

otion in the garter snake Tham

no- phis radix. B

iol. 1. Linn. Soc.

Barnes, P. T

., and Laurie-A

hlberg, C. C

. (1986) Genetic variability of flight m

etabo- lism

in Drosophila rnelanogaster. 111. E

ffects of Gpdh allozym

es and environ- m

ental temperature on pow

er output. Genetics 112:267-294.

Bennett, A

. F. (1980) The therm

al dependence of lizard behaviour. Anim

. Behav.

28:752-762. B

erman, S. L., C

ibischino, M., D

ellaripa, P., and Montren, L. (1985) Intraspecific

variation in the hindlimb m

usculature of the house sparrow. A

m. Z

ool. 25:21A.

Bouchard, C

., and Malina, R. M

. (1983a) Genetics for the sport scientist: selected

methodological considerations. E

xerc. Sport Sci. R

ev. 11:274-305. B

ouchard, C., and M

alina, R. M

. (1983b) Genetics of physiological fitness and

motor perform

ance. Exerc. S

port Sci. Rev. 11:306-339.

Calder, W

. A., 111. (1984) Size, F

unction, and Life H

istory. Cam

bridge, Mass.: H

ar- vard U

niversity Press. C

happell, M. A

., and Snyder, L. R. G

. (1984) Biochem

ical and physiological corre- lates of deer m

ouse a-chain hemoglobin polym

orphisms. Proc. N

atl. Acad. Sci.

U.S.A

. 815484-5488.

Christian, K

. A., T

racy, C. R

., and Porter, W

. P. (1985) Inter-individual and intra- individual variation in body tem

peratures of the Galapagos land iguana (C

ono- lophus pallidus). 1. T

herm. B

iol. 10:47-50. C

lose, R. (1964) Dynam

ic properties of fast and slow skeletal m

uscles of the rat dur- ing developm

ent.]. Physiol. (L

ond.) 173:74-95 C

lose, R. (1965) T

he relation between intrinsic speed of shortening and duration of

:he active state of m

uscle. 1. Physiol. (Lond.) 180542-559.

Crow

ley, S. R. (1985) Insensitivity to desiccation of sprint running perform

ance in the lizard S

celoporus undulatus.]. H

erpetol. 19:171-174. C

rowley, S. R

., and Pietruszka, R. D

. (1983) Aggressiveness and vocalization in the

leopard lizard (Gam

belia wislizennii [sic]): the influence of tem

perature. Anim

. B

ehav. 31:1055-1060. D

iMichele, L., and Pow

ers, D. A

. (1982) Physiological basis for swim

ming endur-

ance differences between L

DH

-B genotypes of F

undulus heteroclitus. Science 216:1014-1016.

Else, P. L., and B

ennett, A. F. (1987) T

he thermal dependence of locom

otor perfor- m

ance and muscle contractile function in the salam

ander Am

bystoma

tigrinum nebulosum

.]. E

xp. Biol., 128:219-233.

Em

sley, A., D

ickerson, G. E., and K

ashyap, T. S. (1977) G

enetic parameters in prog-

eny-test selection for field performance of strain-cross layers. Poult. Sci.

56:121-146. E

ndler, J. (1986) Natural Selection in the W

ild. Princeton, N

.J.: Princeton U

niversity Press.

Falconer, D. S. (1981) introduction to

Quantitative G

enetics, 2nd ed. New

York:

Longm

an. Felsenstein, J. (1985) Phylogenies and the com

parative method. A

m. N

atur. 125:l- 15.

Ferguson, G. W

., and Fox, S. F. (1984) Annual variation of survival advantage of

large juvenile side-blotched lizards, Uta stansburiana: its causes and evolution-

ary significance. Evolution 38:342-349.

Garland, T

., Jr. (1984) Physiological correlates of locomotory perform

ance in a liz- ard: an allom

erric approach. Am

.]. Physiol. 247:R

806-R815.

Garland, T

., Jr. (1985) Ontogenetic and individual variarion in size, shape and speed

in rhe Australian agam

id lizard Am

phibolurus nuchalis. 1. Zool. (L

ond.) 207:425-440.

Garland, T

., Jr., and Arnold, S. J. (1983) E

ffects of a full stomach on locom

otory perform

ance of juvenile garrer snakes (Tham

nophis elegans). Copeia

1983:1092-1096. G

ould, S. J., and Lew

ontin, R. C

. (1979) The spandrels of San M

arco and rhe Pan- glossian paradigm

: a critique of the adaptationist programm

e. Proc. R. Soc.

Lond. [B] 205581-598.

Hanken, J. (1983) H

igh incidence of limb skeletal varianrs in a peripheral popula-

tion of the red-backed salamander, Plethodon cinereus (A

mphibia: Plethodonri-

dae), from N

ova Scoria. Can. 1. 2001. 61:1925-1931.

Huey, R

. B., and B

ennett, A. F. (1986) A

comparative approach to

field and labora- tory studies in evolutionary ecology. In P

redator-Prey R

elationships: Perspec- tives an

d A

pproaches from the S

tudy of Low

er Vertebrates, ed. M

. E. Feder and G

. V. L

auder, pp. 82-98. Chicago: U

niversity of Chicago Press.

Huey, R

. B., and H

ertz, P. E. (1984) Is a jack-of-all-temperatures a m

aster of none? E

volution 38:441-444. John-A

lder, H. B. (1984) Seasonal variations in activity, aerobic energetic capacities,

and plasma thyroid horm

ones (T, and T

,) in an iguanid lizard.]. C

omp. Phys-

iol. 154:409-419. K

rebs, H. A

. (1975) The A

ugust Krogh Principle: "For m

any problems there is an

animal on w

hich it can be mosr conveniently studied."].

Exp. Z

ool. 194:221- 226.

Krogh, A

. (1929) Progress of physiology. Am

. 1. Physiol. 90:243-251. Lacy, F. C

., and Lynch, C

. B. (1979) Quantitative genetic analysis of tem

perature regularion in M

trs musculus. I. Partirioning of variance. G

enetics 91:743-753. Lande, R

., and Arnold, S. J. (1983) T

he measurem

enr of selection on correlared characters. E

volution 37:1210-1226. L

anglois, B. (1980) Heritability of racing ability in thoroughbreds-A

review

. Live-

stock Prod. Sci. 7591- 605.

Laurie-A

hlberg, C. C

., Maroni, G

., Bew

ley, G. C

., Lucchesi, J. C

., and Weir, B. S.

(1980) Quantitative genetic variarion of enzym

e activities in natural populations of D

rosophila rnelanogaster. Proc. Natl. A

cad. Sci. U.S.A

. 77:1073-1077. M

ayr, E. (1982) The G

rowth of B

iological Thought: D

iversity, Evolution, and

Inheritance. Cam

bridge, Mass.: B

elknap. Pough, F. H

., and Andrew

s, R. M

. (1984) Individual and sibling-group variation in m

etabolism of lizards: the aerobic capacity m

odel for the evolution of endo- therm

y. Com

p. Biochem

. Physiol. 79A:415-419.

Putnam, R

. W., and B

ennett, A. F. (1981) T

hermal dependence of behavioural per-

formance of anuran am

phibians. Anim

. Behav. 29502-509.

Ringler, N

. H. (1983) V

ariation in foraging ractics in fishes. In Predators and Prey

in Fishes, ed. D. L. G

. Noakes, D

. G. L

indquist, G. S. H

elfman, and J. A

. Ward,

pp. 159-171. The H

ague: Dr. W

. Junk.

Schmidt-N

ielsen, K. (1984) Scaling: W

hy Is Anim

al Size So Im

portant? Cam

bridge U

niversity Press. Shaffer, H

. B., and Lauder, G

. V. (1985a) Patterns of variation in aquatic am

bysto- m

atid salamanders: kinem

atics of the feeding mechanism

. Evolution 39:83-92.

Shaffer, H. B

., and Lauder, G

. V. (1985b) A

quatic prey capture in ambystom

atid sal- am

anders: patterns of variation in muscle activity.].

Morphol. 183:273-284.

Simpson, G

. G. (1953) T

he Major Features of E

uolution. New

York: C

olumbia U

ni- versity Press.

Smith, D

., King, J. W

. B., and Gilbert, N

. (1962) Genetic param

eters of British L

arge W

hite pigs. Anim

. Prod. 4:128-143. Sullivan, B. K., and W

alsberg, G. E. (1985) C

all rate and aerobic capacity in Wood-

house's toad (Bufo w

oodhousei). Herpetologica 41:404-407.

Taigen, T

. L., and Wells, K

. D. (1985) E

nergetics of vocalization by an anuran am

phibian (Hyla versicolor). 1. C

omp. Ph ysiol. 155:163-170.

Tolley, E. A., N

otter, D. R

., and Marlow

e, T. J. (1983) H

eritability and repeatability of speed for tw

o-and three-year-old standard bred racehorses. 1. Anim

. Sci. 56:1294-1305.

Toolson, E. C

. (19841 Interindividual variation in epicuticular hydrocarbon composi-

tion and water loss rates of the cicada, T

ibicen dealbatus (Hom

optera: Cicadi-

dae). Physiol. Zool. 57:550-556.

van Berkum

, F. H., and T

suji, J. S. (in press) Am

ong-family differences inisprint

speed of hatchling Sceloporus occidentalis. 1. Zool. (L

ond.) W

att, W. B. (1977) A

daptation at specific loci. 1. Natural selection on phosphoglu-

cose isomerase in C

olias butterflies: biochemical and population aspects. G

enet- ics 87:177-194.

Watt, W

. B. (1983) Adaptation at specific loci. 11. D

emographic and biochem

ical ele- m

ents in the maintenance of the C

olias PGI polym

orphism. G

enetics 103:691- 724.

Wells, D

. K., and T

aigen, T. L. (1984) R

eproductive behavior and aerobic capacities of m

ale Am

erican toads (Bufo am

ericanus): is behavior constrained by physiol- ogy? H

erpetologica 40:292-298.

Discussion

LIN

DS

TE

DT

: I agree w

ith B

ennett that there are identifiable

individuals w

hich are low perform

ers - they have a lo

w m

axim

um

oxygen consumption,

a low m

axim

um

running speed, etc. - and there are oth

er individuals that are high perform

ers. But o

n any given day a low

-performing anim

al may

outperform th

e high-performing anim

als: the ranges of their perform

ances overlap, even if their m

eans are repeatedly different. Th

e likelihood of find-

ing mechanistic differences to

account for those mean differences m

ay be rather low

. We

still need to have th

e broader overview betw

een species,

wh

ere we have a higher signal-to-noise ratio.

BE

NN

ET

T: I a

m n

ot advocating th

at we abandon all o

ther approaches for

the study of variation, n

or th

at we should ignore th

e means. B

ut we

should

take interindividual variations into account in addition. Sometim

es the least and m

ost able individuals overlap, but you can still find that the individual differences are statistically repeatable.

LIN

DS

TE

DT

: Y

es, we do try to do that. By the sam

e token, I think we have

to be cautious about throwing out sim

pler statistics because they are simple,

especially if we risk losing som

e biological insight w

ith greater statistical sophistication. For exam

ple, in using stepwise m

ultiple regression, it can be hard to intuit w

hat the result means. A

lso, we have found that as w

e increase the sam

ple size, the total proportion of explained variation changes very lit- tle, yet the relative contributions of various independent variables to

the total explained variation changes a great deal. A

gain, that leads me to

gain less insight.

BENNETT: I agree that m

any times the sim

ple statistics are adequate, but w

here they are inadequate, we should not continue using the old m

odels and old w

ays of doing things. Th

e stepwise m

ultiple regression approach seems

to me to be a tool to

suggest further directions for study. If you find that no

factors are correlated w

ith performance (m

easured as burst speeds), that may

tell you that you should be looking at other factors. When you do have sig-

nificant correlations, then you have a basis for further experimental analysis.

It is a first pass in looking for important variables.

SCH

EID: I agree that interindividual variability is really im

portant. Nature

intends to tell us something that w

e have mostly neglected so far. N

ow, is it

not true that if you want to

address the interindividual variability, then you have to look at the intraindividual variability first? In fact, the only thing that rem

ains beyond intraindividual variability is true interindividual variability.

BENNETT: T

hat's right. You have to be able to m

ake repeated measures on

individuals. This is feasible for som

e factors and unfeasible for other factors. If you are looking at w

hole-body lactate content, for instance, you can do

that only once. B

ut there are a large number of physiological characters, such

as blood flow param

eters, that we can now

sample nondestructively because

of improved instrum

entation.

SCHEID: W

e now have im

proved techniques to w

ork on

uninstrumented,

nonanesthetized animals, w

hich is mandatory if you w

ant to ask questions

about variability. I think that the techniques were not form

erly available to

address this variability in a meaningful w

ay.

I HU

EY: T

he standard statistical m

ethod of m

easuring repeatability is the

intraclass correlation coefficient, which m

easures the proportion of the vari- ation that is due to

difference among individuals versus w

ithin individuals. By that

measure, som

etimes the types of

measurem

ents that Bennett w

as

referring to are highly repeatable: m

ost of the variation is among individuals

and not within individuals. F

or example, A

rt Dunham

and I looked at sprint speeds of lizards in natural populations over a w

hole year, and the repeat- abilities are on the order of 0.5 to 0.6, w

hich is higher than in thoroughbred horses.

That is probably

high enough that w

e, can begin to analyze the

mechanistic basis of individual variation and also look at the adaptive sig-

nificance of that variation.

PO

WE

RS

: O

ne problem of reproducing the sam

e experiment on the sam

e individual is that som

e organisms becom

e trained. In addition, we found that

we cannot put m

ore than one individual in a track at a time because of behav-

ioral interactions between them

.

BE

NN

ET

T: For running speed, in about half of the species that w

e observe, w

e see what w

e assume is a conditioning effect, from

day one to day tw

o, but after that the m

eans stay exactly the same, the order of the individuals

stays the same. S

ome species show

this initial effect, others don't.

PO

WE

RS

: One thing we have to

do

with fish is to

acclimate them

to w

ater that is m

oving at a constant speed, for thirty to sixty days. Everything from

then on is very reproducible. I am

sure that a lot of the variation in the lit- erature is a function of this training phenom

enon and where the organism

s cam

e from.

FL

OR

AN

T:

I think that developmental effects can be extrem

ely important,

and I wondered w

hether you were rearing these anim

als in the lab or being

careful about the developmental processes that w

ere going on prior to, dur- ing, and after birth.

BE

NN

ET

T: All the anim

als that we have been dealing w

ith are adult animals,

taken directly from the field and tested w

ithin a matter of days. In the breed-

ing studies, gravid animals are collected and young anim

als are born under constant conditions in the laboratory. T

he whole issue of developm

ental effects and constancy of rank-ordered perform

ance over time has not even

begun to be explored.

FU

TU

YM

A:

Suppose you are interested in very short term

acclimation effects,

the capacity of the individual

to change its phenotype from

m

oment to

mom

ent, which is the opposite of repeatability. H

ow

do

you deal with that?

There are interesting questions there as w

ell.

BE

NN

ET

T: Y

OU

begin by im

mediately asking questions about your equip-

ment and techniques, and get that out of the w

ay first. Then perhaps you can

begin building correlations from m

oment to m

oment by m

easuring the vari- ables sequentially, to

see whether you are getting tracking of one variable by

the other.

INT

ER

IND

IVID

UA

L V

AR

IAB

ILIT

Y

169

AR

NO

LD

: W

e can examine the capacity to

change performance over short

or long periods of time as traits, and that's a virtually unexplored area. B

ut it is not th

e opposite of repeatability. Suppose w

e define a new variable that

represents the capacity to change perform

ance as a function of an elevation in tem

perature from 10 OC

to 20 O

C. We can m

easure its repeatability, we

can ask whether it is inherited, w

hether it is genetically correlated with other

traits, and so forth. The statistical field for dealing w

ith such traits is some-

times called profile analysis of variance.

FE

DE

R: I w

ant to shift the focus of this discussion to

the point that Bennett

made about the prospect for perform

ing natural experiments using natural

populations. I am very excited about this prospect. U

sing the variation in populations as a substrate, altering an environm

ental variable for individuals in a population or adding individuals to a population and looking at the effects could potentially be a very pow

erful technique. Dennis P

owers said

that it may soon be possible to take individual genes and m

ove them into or

out of individual organisms, w

hich could offer us a lot of insight.

PO

WE

RS

: It is already possible for som

e species. In lower vertebrates,.it w

ill probably take another year.

DA

WS

ON

: T

here are detraining or conditioning effects that go with captiv-

ity. When w

e studied cold resistance in small birds, w

e found that the ani- m

als maintained in outdoor flight cages, given the seeds of the type that they

were using naturally, abandoned their w

inter fattening, perhaps because they had assured m

eals and more com

plicated cues. They also had m

uch lower

cold resistance than freshly captured animals. If

one is dealing with badly

distorted responses, which is som

etimes a risk w

ith wild anim

als, that ought to be determ

ined. A good deal of w

hat one may be dealing w

ith in animals

long standing in captivity may not be relevant to

the natural situation.

FLOR

AN

T: In keeping hibernators for many years in the lab, the hibernators

begin to free-run, and it is as if certain physiological responses occur at "non-

adaptive" times of year. T

his obscures the optimal tim

e that the animal per-

forms a particular response under natural circum

stances.

BENN

ETT: These are valid concerns. O

ne way of keeping track of them

is to run appropriate controls, so that w

e can place boundaries on the magni-

tude of the captivity responses.

DA

WS

ON

: By attem

pting to determ

ine repeatability,

if you

start early enough, -you can discern if there are any effects of that type. T

hat is not done a lot. T

his is a caveat about use of material from

animal dealers, w

hich may

have a very fuzzy history indeed.