. .:
1....
iv.

.

.2: ..
..V. o2 11;

 

.—.;_

A

._....4n-.‘~

And

 

1...};-

4&1.qu-

Aw '.

 

     

,~ ‘ ' ‘lmfi'fi‘r:7i!7?-inz vv--.-~c~a)~ls‘.-|o 9 91x5-/O! ,\,I¢u~-v a - ~0-
T'lo:-._.111f3'y’ ‘. '. ‘

ATE EUNIVERSITY LIBRARIES

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IlllilllllllfllujglllglIlllll
“ ' LIBRARY
1» Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

THE IMPACT OF NATERNAL EFFECTS ON ADAPTIVE EVOLUTION:
COMBINING QUANTITATIVE GENETICS AND PHENOTYPIC SELECTION
IN A NATURAL PLANT POPULATION

presented by
Denise A. Thiede

has been accepted towards fulfillment
of the requirements for

Ph.D.

 

degree in Botany
and Plant Pathology

%%

Major professorfl

 

Date ”W716; l’,é

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

PLACE IN RETURN BOX to remove this chockom from your noord.
To AVOID FINES return on or before date duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l
l

 

 

 

 

MSUloAnAfflnmtivoActiocVEqudOppom lnstltution
nity Wanna-9.1
A A ___.___ ._ . . -._._____‘_ *

 

 

fl

THE IMPACT OF MATERNAL EFFECTS ON ADAPTIVE EVOLUTION:
COMBINING QUANTITATIVE GENETICS AND PHENOTYPIC SELECTION
IN A NATURAL PLANT POPULATION

By

Denise Annette Thiede

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

W. K. Kellogg Biological Station
and
Department of Botany and Plant Pathology

1996

ABSTRACT
THE IMPACT OF MATERNAL EFFECTS ON ADAPTIVE EVOLUTION:
COMBINING QUANTITATIVE GENETICS AND PI-IENOTYPIC SELECTION
IN A NATURAL PLANT POPULATION
By

Denise Annette Thiede

When a mother influences the phenotypic expression of traits in her offspring, the
direction, rate, and duration of adaptive evolution can be modified from standard
Mendelian models. To explore the evolutionary implications of trans-generational maternal
effects, I quantified two aspects of evolutionary response: the quantitative genetic basis of
maternal inheritance and the magnitude of phenotypic selection at the individual and
maternal family level for ten traits expressed at four stages in the life cycle in a winter
annual plant, Collinsia vema. In a hierarchical quantitative genetic analysis of Mendelian
and maternal inheritance, I estimated six additive and environmental causal components of
variance: direct (i. e. Mendelian) additive and environmental, maternal additive and
environmental, and the direct-maternal additive and environmental covariances. The
structure of maternal inheritance changed through the life cycle. Early traits were
influenced more by maternal additive than by direct effects, direct and maternal additive
effects covaried negatively, and direct-matemal environmental covariance was positive. At
subsequent stages, some traits displayed strictly Mendelian inheritance, while others
displayed direct and maternal additive genetic effects of the same magnitude and negative
direct-matemal covariances. Maternal environmental components were negligible beyond

emergence. The negative direct-matemal covariances for all maternally inherited traits

resulted in near zero or negative realized heritabilities indicating no or reversed response
to selection, respectively. In nature, the magnitude of selection on maternally inherited
traits will also determine evohrtionary response. I examined phenotypic selection at two
levels: individual and maternal. An episodic analysis of individual selection across four
stages in the life cycle demonstrated that large fall size and later emergence were directly
favored across all episodes, although the magnitude and direction of selection varied
among episodes. As a result of positive phenotypic correlations among size traits,
selection also indirectly favored heavier seeds and larger initial size. Maternal selection
may also afiect selection response because substantial among maternal family variance in
fitness indicated the opportunity for maternal selection. Maternal efl‘ects are likely to have
dramatic short-term evolutionary consequences by constraining the selection response,
influencing correlated response to selection via the phenotypic variance-covariance matrix,

and affecting offspring fitness directly via maternal selection.

In memory of my mother, Irene Niebuhr Thiede.

ACKNOWLEDGMENTS

I owe a debt a gratitude to Susan Kalisz. Her intellectual involvement greatly
improved the quality of this work and her encouragement helped me complete it. Her
belief in my abilities helped me overcome many obstacles in this process. I am grateful to
have had her as a mentor.

My committee, Tom Getty, Kay Gross, Don HalL Susan Kalisz, Alan Tessier, and
Steve Tonsor, provided constructive criticism at many stages throughout this process. I
thank Steve Tonsor for a SAS bootstrapping program

Ruth Shaw kindly provided the maximum likelihood program for the quantitative
genetic analysis. Frank Shaw modified the program to handle the three generation
pedigree and patiently answered many questions about the program Their help was
invaluable for completing the analysis.

The Kalisz-Tonsor lab group provided critical feedback and technical assistance at
all stages: Brian Black, Fran Hanzawa, Dawn Jenkins-Kins, Paco Moore, Peter Smith, and
Glenda Wardle, our University of Chicago honorary member.

I had the remarkable good fortune of having several dedicated field assistants: Tori
Derr, Shannon Gibb, Jean Tsao, and Pam Woodmfi‘: Their attention to detail and
perseverance enhanced the quality of this work. Many others also provided assistance at
various stages: Pat Frueh, Amy Malone, Barbie Oelslagger, Ann O’Neil, Robin Sitka,

Martha Tomecek, and Loretta Weathers.

KBS stafl‘ provided technical assistance: John Gorentz and Stephan Ozminski
served as computer gurus, Carolyn Hammarsjkold provided excellent library assistance,
and Art Weist helped me in innumerable ways in the greenhouse.

I thank the Balkema family for permission to conduct this work on their property.

Generous financial support was provided by an MSU recruiting fellowship, a
Kellogg Biological Station George Laufl‘ fellowship, the College of Natural Science
Barnett Rosenberg fellowship, and an RTG fellowship fi'om NSF DIR-9113598. I received
financial support for supplies, travel, field assistants, and computing from NSF DEB-
9224046 and DIR-9113598, the College of Natural Science, Kellogg Biological Station,
the Department of Botany and Plant Pathology, the Ecology and Evolutionary Biology
Program

My life at KBS and in Kalamazoo was enriched by many wonderful fiiends and
colleagues: Fran Hanzawa, Dawn Jenkins-Klus, Glenda Wardle, Casey Huckins, Brian
Black, Paco Moore, Andy Turner, Lisa Huberty, Carol Kelly, Martha Tomecek, Michel
Cavigelli, Joanne Dodgson, Diane Walker-Smith, and Deanna Wines.

Lastly, I thank Kim Thompson who witnessed in excruciating detail the ups and
downs of all of this work. She stood by me in the toughest times, collected more data than
she ever should have, and helped me get back on my feet more than once. This work is

dedicated to her.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. x
LIST OF FIGURES ............................................................................................... xiv
INTRODUCTION ................................................................................................. l
Quantifying maternal inheritance ................................................................. 3
Partitioning the phenotypic covariances among relatives .................. 3
Estimating maternal efl‘ect coeflicients ............................................. 5
Estimates of maternal effects in plants ............................................. 5
Quantifying phenotypic selection ................................................................. 6
Evolutionary consequences of maternal efl‘ects ............................................ 8
CHAPTER 1

MATERNAL INI-IERITANCE AND ITS EFFECT ON ADAPTIVE
EVOLUTION: A QUANTITATIVE GENETIC ANALYSIS OF MATERNAL

EFFECTS IN A NATURAL PLANT POPULATION ............................................ 10
Introduction ................................................................................................ 10
Materials and Methods ............................................................................... 20

Study species .................................................................................. 20
Three generations ............................................................................ 21
Traits measured ............................................................................... 26
Estimation of genetic and environmental causal components ............ 27
Analysis .......................................................................................... 28
Within generation genetic correlations ............................................. 30
Results ....................................................................................................... 35
Comparison of estimation models .................................................... 35
Significance tests of specific causal genetic components .................. 39
Relative contribution of components to total phenotypic variance. . .. 39
Direct, maternal, and realized heritabilities ....................................... 43
Maternal effects at difi‘erent stages in ontogeny ............................... 48
Within generation genetic correlations ............................................. 51
Discussion .................................................................................................. 53
Maternal inheritance ........................................................................ 54
Maternal performance ..................................................................... 56
Estimation of maternal efl‘ects ......................................................... 57
Evolutionary consequences of maternal inheritance. ....................... 61
Intergenerational covariances .......................................................... 62

vii

Within generation covariances ......................................................... 64

Multivariate evolution ..................................................................... 65
Conclusions .................................................................................... 65

CHAPTER 2

AN EPISODIC ANALYSIS OF PHENOTYPIC SELECTION ON JUVENILE
TRAITS IN COLLINSIA VERNA: A COMPARISON OF QUANTITATIVE
TRAITS DISPLAYING MENDELIAN AND NON-MENDELIAN

INHERITAN CE .................................................................................................... 67
Introduction ............................................................................................... 67
Materials and Methods ............................................................................... 73

Study site and species ..................................................................... 73
Quantifying phenotypic selection ..................................................... 75
Data collection .................................................................... 75
Natural seedlings ................................................................. 75
Planted seedlings ................................................................. 77
Data analysis ....................................................................... 78
Survivorship analysis ........................................................... 85
Results ....................................................................................................... 85
Opportunity for selection ................................................................ 85
Phenotypic correlations among traits ............................................... 87
Total magnitude and direction of selection ....................................... 92
A. Changes in trait means .................................................... 92
B. Changes in trait variances ................................................ 105
Viability and fecundity selection: episodes of selection ..................... 111
Conditional vs. reconstructed selection analysis ............................... 113
Discussion .................................................................................................. 130
Changes in trait means ..................................................................... 130
Maternal inheritance ........................................................................ 13 1
Direct and indirect effects ................................................................ 133
Episodic analysis ............................................................................ 134
Reconstruction of phenotypic variance-covariance matrix ................ 136
Conclusions ..................................................................................... 137

CHAPTER 3

THE OPPORTUNITY FOR MATERNAL SELECTION IN A NATURAL

POPULATION OF COLLINSIA VERNA (SCROPHULARIACEAE) .................... 139
Introduction ............................................................................................... 139
Materials and Methods ................................................................................ 145

Study site and species ...................................................................... 145
Analysis .......................................................................................... 147
Maternal selection at a global scale ...................................... 147
Maternal selection at a local scale ........................................ 148

Spatial variation in fitness components ..................... 149

Spatial variation in phenotypic traits ......................... 149

Spatial variation in phenotypic selection ................... 149

Local family efl‘ects .................................................. 150

Results ........................................................................................................ 151
Maternal selection at a global scale .................................................. 151
Maternal selection at a local scale .................................................... 152

Spatial variation in fitness components ................................. 165

Spatial variation in phenotypic traits ..................................... 168

Spatial variation in phenotypic selection ............................... 168

Opportunity for maternal selection ....................................... 174

Local family effects .............................................................. 185

Discussion .................................................................................................. 188
Maternal selection on a global scale ................................................. 188
Maternal selection on a local scale ................................................... 190
Inheritance of group traits ............................................................... 193
Conclusions ..................................................................................... 193

LIST OF REFERENCES ....................................................................................... 195

LIST OF TABLES

Table 1. Partitioning the phenotypic (co)variance for each of seven sets of relatives into
causal genetic components of variance and covariance (after Eisen 1967; Thompson
1976). Values represent the coeflicient for each component. A denotes additive genetic, D
denotes dominance, E denotes environmental variance, respectively. The subscript 0
denotes direct effects due to standard Mendelian inheritance, while m refers to maternal
efi‘ects due to maternal inheritance. Individual variance shows which components
contribute to an individual’s phenotypic value. ..................................................... 12

Table 2. Mean, standard error, coefficient of variation , and sample sizes for phenotypic
traits measured at respective life-cycle stages in both F2 and F3 generations. Parents (F2)
were grown in the greenhouse, while offspring (F3) were grown in both the greenhouse
and field. For parents cotyledon diameter at two stages is the average of cotyledon length
and width at emergence and is equal to fall cotyledon length, while in the ofi‘spring
cotyledon diameter was consistently measured by a circular template at each of those
stages. ........................................................................................................................... 22

Table 3. Model 1 restricted maximum likelihood estimates of causal components for
greenhouse and field environments. The log likelihood of the firll model, magnitude of the
fixed generation efl‘ect, direct additive (02A,), direct environmental (025°), and total
phenotypic variance (02p) are presented. See Table 2 for sample sizes. Significance of each
component is noted (*0.1<p<0.05, ** 0.05<p<0.01, ***0.01<p<0.005, ****p<0.005).
...................................................................................................................................... 36

Table 4. Model 2 restricted maximum likelihood estimates of causal components for
greenhouse and field environments. The log likelihood of the full model, magnitude of the
fixed generation effect, the five direct and maternal causal components, and total
phenotypic variance (02;) are presented. See Table 2 for sample sizes. Significance of the
improvement in the log likelihood relative to model 1 (df=3) and the significance of each
component are noted (df=1, *0.1<p<0.05, “ 0.05<p<0.01, ***0.0l<p<0.005,
****p<0.005). Traits not converging are indicated by NC. .......................................... 37

Table 5. Model 3 restricted maximum likelihood estimates of causal components for
greenhouse and field environments. The log likelihood of the firll model, magnitude of the
fixed generation effect, six direct and maternal causal components, and total phenotypic
variance (0%) are presented. See Table 2 for sample sizes. Significance of the
improvement in the log likelihood relative to model 2 (dfil) and the significance of each

component are noted (df=l, *0.1<p<0.05, ** 0.05<p<0.01, *“0.01<p<0.005,
**”p<0.005). Traits not converging are indicated by NC. .......................................... 38

Table 6. Direct, maternal, and realized heritabilities and direct-matemal genetic
correlations for three inheritance models. Significance is based on the significance of the
variance component contributing to the numerator (Tables 3-5), indicating the potential
for genetic response (Shaw and Platenkamp 1993). ..................................................... 47

Table 7. Direct additive genetic correlations between each pair of traits for greenhouse
and field environments. For each bivariate analysis, I estimated genetic correlations from
two estimation models: model 4, a strictly Mendelian model (below), and model 5 with
maternal efl‘ects (above). Model 5 was evaluated only when one or both traits were
influenced by maternal efi‘ects ( ------ indicates this pair of traits was not evaluated).
Estimation models that did not converge are denoted by NC. When significance tests of
the direct additive genetic covariance did not converge, a + follows the value. Bold values
indicate model estimation without maternal environmental variances eliminated to obtain
convergence. Otherwise significance is as follows: * 0.1<p<0.05, *"' 0.05<p<0.01,
***0.01<p<0.005, ‘**"‘p<0.005. ............................................................................ 52

Table 8. Phenotypic selection parameters calculated in each episode(i). ................... 81

Table 9. The phenotypic variance-covariance matrix prior to the original (Po) and final
(P3) episodes of selection for natural and planted seedlings in 1992 and 1993 for three trait
(emergence week (WK), initial size (IS), and fall size (FS) and four trait models (including
seed weight (SD)). The original matrix has been reconstructed from conditional selection
gradients (see Methods). The 95% confidence intervals for each element are also
provided. ............................................................................................................... 90

Table 10. Additive linear selection differentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes and the total across all episodes. Below each vahre are
the 95% confidence intervals based on bootstrap resampling (n=500). linear parameters
reflect the change in the mean in units of standard deviation. Values significant at p<0.05
are bold. ............................................................................................................... 98

Table 11. Additive non-linear selection differentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes and the total across all episodes. Below each value are
the 95% confidence intervals based on bootstrap resampling (n=500). Non-linear
parameters reflect the change in the variance in units of standard deviation squared.
Values significant at p<0.05 are bold. ........................................................................... 106

Table 12. Conditional linear selection difl‘erentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes. Below each value are the 95% confidence intervals

based on bootstrap resampling (n=500). Linear parameters reflect the change in the mean
in units of standard deviation. Values significant at p<0.05 are bold. ............................ 121

Table 13. Conditional non-linear selection differentials (A) and gradients (B) for natural
and planted seedlings in 1992 and 1993 for three trait or four trait models. Values are
given for each of four selection episodes. Below each value are the 95% confidence
intervals based on bootstrap resampling (n=500). Non-linear parameters reflect the change
in the variance in units standard deviation squared. Values significant at p<0.05 are bold.
.................................................................................................................................... 123

Table 14. One-way analysis of variance of family variation in survivorship for globally
planted seedlings in 1993 at TU Ave for four episodes: 1) survival through establishment
(A), 2) survival to the onset of winter (B), 3) survival to spring (C), and 4) final fecundity
(D). Only families with two or more offspring in a given episode were included in this
analysis. Degrees of freedom are for the numerator and denominator, respectively. ...... 153

Table 15. Univariate regression models for globally planted seedlings in 1993 in four
selection episodes: 1) survival through establishment (A), 2) survival to the onset of
winter (B), 3) survival to spring (C), and 4) final fecundity (D). Significant linear selection
coefficients ([3) are bold. ...................................................................................... 156

Table 16. Two AN COVA models for evaluating the potential for maternal selection for
global seedlings in 1993. The full model included the interaction term to evaluate
heterogeneity of slopes (A). Significance was based on Type 1 sum of squares, ie.
sequential sum of squares (SAS Institute, Inc. 1994). Reduced models for all traits not
displaying heterogeneity of slopes among families (B) were tested for significance of
family effect using Type III sum of squares, i.e. partial sum of squares. ................. 157

Table 17. Nested AN OVA of the spatial variation in survivorship and fecundity for
naturally emerging and planted seedlings in 1992 and 1993 at TU Ave for four episodes:
1) survival through establishment, 2) survival to the onset of winter, 3) survival to spring,
and 4) final fecundity. Transect, block nested within transect, and family nested within
block (planted seedlings only) were treated as random effects and appropriate F-tests were
constructed using the Random statement in GLM (SAS Institute, Inc. 1994). Degrees of
freedom for the model are for the error term, other degrees of freedom are for the
numerator. ............................................................................................................. 166

Table 18. Multivariate nested analysis of variance of the spatial variation in phenotypic
traits for naturally emerging and planted seedlings in 1992 and 1993 at TU Ave. The
spatial scale of variation was assessed for transects (n=2), blocks within transects (n=4,3
in 1992 and n=10 in 1993), and families within blocks (planted seedlings only, n=77 and
191 in 1192 and 1993, respectively). All source variables were treated as random effects
with appropriate F-tests specified in the MANOVA statement in GLM (SAS Institute, Inc.
1994). Three traits, emergence week, initial size, and fall size were analyzed as dependent
variables for naturally emerging seedlings. Four traits including seed weight were treated
as dependent variables for planted seedlings. Type [[1 sum of squares were utilized for all

’01

analyses. Univariate probabilities are presented for each trait for comparative pm'poses

only. ......................................................................................................................... 173

Table 19. Nested analysis of covariance to examine maternal selection on individual
Mess while accounting for spatial variation in survivorship for planted seedlings in 1992
and 1993 at TU Ave for four episodes: 1) survival through establishment, 2) survival to
the onset of winter, 3) survival to spring, and 4) final fecrmdity. Type III sum of squares
were utilized for significance test. ........................................................................... 186

LIST OF FIGURES

Figure 1. Path diagrams of Mendelian (A) and maternal inheritance (B and C) (afier
Dickerson 1947; Wilham 1963; Cheverud 1984). Under Mendelian inheritance in model 1
(A), the additive genetic value (A.,) and the environmental value (E0) determine the
phenotypic value (P.,) where the subscript 0 refers to the ofl‘spring trait and w and x refer
to the maternal and ofispring generations, respectively. In Model 2 (B), maternal
inheritance is determined by the phenotypic effects of the maternal performance trait (Pm)
and its additive genetic (A...) and environmental (Em) components, the subscript m
referring to maternal performance. The genetic correlation between direct and maternal
traits (rm), and the square root of the direct (ho) and maternal heritabilites (h...) are
illustrated. The maternal efl‘ect coeflicient (m) indicates the extent to which the maternal
phenotype influences the phenotypic value in the offspring independent of additive genetic
effects. In Model 3 (C) maternal inheritance includes the potential correlation between
direct and maternal environments (Ifiogm). ................................................................. 14

Figure 2. Summary of studies estimating direct-matemal additive genetic correlations
based on three difi‘erent estimation models: 1) the animal model included additive and
environmental components only, 2) the full model also included dominance components,
and 3) cross-fostering models estimated post-natal maternal efi‘ects (Bondari et al. 1978,
Cantet et al. 1988; Southwood and Kennedy 1990; Shi et aL 1993; Van Sanford and
Matzinger 1982; Everett and Magee I965; Young and Legates 1965). ................... 18

Figure 3. Three generation breeding design in which field-collected grandmatemal
families (F 1, GD“, n=lOO) provided seed for the parental generation (F2). Of the twelve
seeds planted from each granddam (F1), one was randomly assigned as a sire (Sn, n=24) or
darn (Du, n=72); while the other eleven grandmatemal full-sibs were considered paternal
(PR..) or maternal relatives (MR..). Greenhouse-raised parents produced up to 40 offspring
(F3) that were divided between greenhouse and field environments. .............................. 23

Figure 4. Path diagrams to estimate genetic correlations between pairs of traits. In Model
4 (A), the genetic correlation between two traits inherited in a Mendelian fashion is
denoted by erlez, where 1 and 2 refer to the two traits in the model All other symbols
are identical to Figure 1. In Model 5 (B) both traits are maternally inherited. Within
generation genetic correlations between direct additive effects (erlez), and between
maternal additive effects (rAm 1m) are depicted. All possible components are depicted,
however, models were simplified based on the best inheritance model determined by the
univariate analysis of each trait. ............................................................................ 32

xiv

Figure 5. The relative contribution of each variance component to the total phenotypic
variance for models 1, 2, and 3 for greenhouse (A) and field (B) environments. Only 5
components, the direct additive (02M), maternal additive (02M) , direct environmental
(6212p), maternal environmental (025...), and the direct-matemal additive covariance (cm)
are included for models 2 and 3 because the direct-matemal environmental covariance
(0505,.) does not contribute to the total phenotypic variance (see Table 1) Significant
components are indicated (*). The model best describing a trait is indicated by an arrow.
...................................................................................................................................... 40

Figure 6. The direct, maternal, and realized heritabilities for each trait in all models in
greenhouse (A) and field (B) environments. Significant heritabilities are indicated (*). The
model best describing a trait is indicated by an arrow. .......................................... 44

Figure 7. The direct (square), maternal (circle), and realized (triangle) heritabilities for
three greenhouse traits measured repeatedly through ontogeny: cotyledon diameter at 3
stages, and leaf length and number of leaves both measured in the late fall and in the spring
prior to flowering. Significant heritabilities are indicated (*). .......................................... 49

Figure. 8. Path diagram depicting the direct efi‘ects of four juvenile traits on four
components of fitness, three episodes of survival and final fecundity. Direct efl‘ects
represented by the single headed arrows allow one to predict the change in the trait mean
in the subsequent generation and are estimated as the partial regression coeflicient (B)
from a multiple regression analysis of all traits on relative fitness. Double-headed arrows
between traits denote the phenotypic correlations among traits that mediate indirect
effects. Maternal inheritance can influence not only the phenotypic value of a trait, but also
the phenotypic correlations between traits. Two traits, seed weight and emergence week,
contribute to survival in the first episode. Three traits including initial size contribute to
the second episode. All four traits contribute to the third and fourth episodes of selection.
...................................................................................................................................... 72

Figure 9. Survival through each episode of viability selection for two years at TU Avenue
for natural and planted seedlings. Standard errors are smaller than the symbols. ....... 86

Figure 10. Average fecundity (2t 1 standard error) of natural and planted seedlings in both
years for all plants, for plants browsed by deer, and plants that were not browsed. ....... 88

Figure 11. The conditional opportunity for selection presented as the variance in relative
fitness for each episode of selection for all individuals surviving the previous episode. The
total opportunity is the variance in relative fitness for all individuals across all episodes.
The patterns for the episodes of selection apply to all subsequent figures. ................... 89

Figure 12. Phenotypic selection through three episodes of viability and one episode of
fecundity selection for natural seedlings in 1992. Viability episodes include survival to
establishment (solid), survival to the onset of winter (vertical), survival to spring

XV

(diagonal), and the final episode of fecrmdity selection (clear). The total magnitude of each
selection parameter (I) across all episodes and the 95% confidence intervals (O) based on
500 bootstrap samples based on reconstructing the original phenotypic variance-
covariance matrix are depicted. Selection parameters include: A) linear selection
difi‘erential, B) linear selection gradient, C) non-linear selection differential, D) non-linear
selection gradient. Trait abbreviations follow Table 9. Significant episodes of selection are
denoted by ‘. Note the difference in scale for non-linear selection parameters. ....... 93

Figure 13. Phenotypic selection through three episodes of viability and one episode of
fecundity selection for natural seedlings in 1993. All symbols follow Figure 12. ....... 95

Figure 14. Phenotypic selection through three episodes of viability and one episode of '
fecundity selection for planted seedlings in 1992. All symbols follow Figure 12. ....... 96

Figure 15. Phenotypic selection through three episodes of viability and one episode of
fecundity selection for planted seedlings in 1993. All symbols follow Figure 12. ....... 97

Figure 16. Indirect selection differential on three traits for all selection episodes for
natural seedlings in 1992 (A) and 1993 (B) and planted seedlings in 1992 (C) and 1993
(D). Viability episodes include survival to establishment (solid), survival to the onset of
winter (vertical), survival to spring (diagonal), and the final episode of fecundity selection
(clear). The total magnitude of each selection parameter (I) for all episodes and the 95%
confidence intervals (0) based on 500 bootstrap samples for the indirect differential are
depicted. Trait abbreviations follow Table 9. Significant episodes of selection are denoted
by *. ......................................................................................................................... 101

Figure 17. Phenotypic selection on four traits including seed weight for planted seedlings
in 1992. All symbols follow Figure 12. ............................................................... 103

Figure 18. Phenotypic selection on four traits including seed weight for planted seedlings
in 1993. All symbols follow Figure 12. ............................................................... 104

Figure 19. Conditional selection parameters on three traits through four episodes of
selection for natural seedlings in 1992. Viability episodes irrchrde survival to establishment
(solid), survival to the onset of winter (vertical), survival to spring (diagonal), and the final
episode of fecundity selection (clear). Selection parameters include A) linear selection
differential, B) linear selection gradient, C) non-linear selection gradient, and D) non-linear
selection gradient. Trait abbreviations follow Table 9. All selection parameters are in units
of standard deviation. Significant episodes of selection are denoted by *. ................. 114

Figure 20. Conditional selection parameters on three traits through four episodes of
selection for natural seedlings in 1993. All symbols follow Figure 19. ................. 116

Figure 21. Conditional selection parameters on three traits through four episodes of
selection for planted seedlings in 1992. All symbols follow Figure 19. ................. 117

xvi

Figure 22. Conditional selection parameters on three traits through four episodes of
selection for planted seedlings in 1993. All symbols follow Figure 19. ................. 118

Figure 23. Conditional selection parameters on four traits through four episodes of
selection for planted seedlings in 1992. All symbols follow Figure 19. ................. 119

Figure 24. Conditional selection parameters on four traits through four episodes of
selection for planted seedlings in 1993. All symbols follow Figure 19. ................. 120

Figure 25. The opportunity for maternal selection portrayed as the among fimily variance
in relative fitness in each episode for locally and globally planted seedlings. Relative

fimily fitness is calculated as mean family fitness standardized by the grand mean of family
fitrresses for the sample population. ........................................................................... 154

Figure 26. The phenotypic distribution of seed weight for all globally planted seedlings
(A), the distribution of family means (I) in standard deviation units with 10, 25, median,
75, and 95 percentiles depicted (B), and the overall linear relationship between seed
weight and survival to the onset of winter (C). ................................................... 159

Figure 27. The phenotypic distribution of emergence week for all globally planted
seedlings (A), the distribution of family means (I) in standard deviation units with 10, 25,
median, 7 5, and 95 percentiles depicted (B), and the overall linear relationship between
emergence week and survival to the onset of winter (C). ........................................ 161

Figure 28. The phenotypic distribution of initial size for all globally planted seedlings (A),
the distribution of fimily means (I) in standard deviation units with 10, 25, median, 7 5,
and 95 percentiles depicted (B), and the overall linear relationship between initial size and
survival to the onset of winter (C). The two lines in (C) are for overall regression and
weighted average of within family regressions, respectively. ........................................ 163

Figure 29. The spatial scale of phenotypic and demographic variation for natural (A, B,
C) and planted seedlings (D,E,F) in 1992. Block means (i 1 standard error ) for
phenotypic variation in seed weight (*), emergence week (I), initial size (O), and fill
size (A) (A, D), survivorship through three episodes, survival to establishment (I),
survival to the onset of winter (OI), survival to spring (A) (B, E), and fecrmdity in the final
episode (8) (C, F). The grand mean across blocks is depicted by a line for each episode of
survival or reproduction. Blocks are located on forest edge and interior. ................. 169

Figure 30. The spatial scale of phenotypic and demographic variation for natural (A, B,
C) and planted seedlings (D,E,F) in 1993. Block means (i 1 standard error ) for
phenotypic variation in seed weight (it), emergence week (I), initial size (O), and fill
size (A) (A, D), survivorship through three episodes, survival to establishment (I),
survival to the onset of winter (0), survival to spring (A) (B, E), and fecundity in the final
episode (36) (C, F). The grand mean across blocks is depicted by ailine for each episode of
survival or reproduction. Blocks are located on forest edge and interior. ................. 171

O.

Figure 31. Spatial variation in phenotypic selection between transects along forest edge
(1) and interior (2) for natural and experimentally planted seedlings in 1992. The linear
selection differentials (A, C) and linear selection gradients (B, D) for three traits,
emergence week, initial size, and fill size, in each viability and fecundity episode. Codes
for each episode as in Figure 25. Total magnitude of selection across all episodes is
depicted by (O) and is based on reconstructed phenotypic variance-covariance matrix (see
Chapter 2). 95% confidence intervals for total values are also depicted by (O) and are
based on 250 resampled data sets for each transect. Natural seedlings along transect 2 in
1992 have confidence intervals that exceed the upper and lower axis values. ..... 175

Figure 32. Spatial variation in phenotypic selection between transects along forest edge
(1) and interior (2) for natural and experimentally planted seedlings in 1993. The linear
selection differentials (A, C) and linear selection gradients (B, D) for three traits,
emergence week, initial size, and fill size, in each viability and fecundity episode. Codes
for each episode as in Figure 25. Total magnitude of selection across all episodes is
depicted by (O) and is based on reconstructed phenotypic variance-covariance matrix (see
Chapter 2). 95% confidence intervals for total values are also depicted by (O) and are
based on 250 resampled data sets for each transect. ................................................... 177

Figure 33. Spatial variability in phenotypic selection among blocks along the edge (1-10)
and interior (1 1-20) transects for natural seedlings in 1993. Linear selection differentials
(A,B,C) and linear selection gradients (D,E,F) in each viability and fecundity episode for
three traits, emergence week, initial size, and fill size are shown. The total magnitude of
selection across all episodes is depicted by (O) and is based on reconstructed phenotypic
variance-covariance matrix (see Chapter 2). 95% confidence intervals for total values are
based on 50 resampled data sets for each block. Total values are connected by a line for
visual clarity. ............................................................................................................. 179

Figure 34. Spatial variability in phenotypic selection among blocks along the edge (1- 10)
and interior (1 1-20) transects for experimental seedlings in 1993. Linear selection
differentials (A,B,C) and linear selection gradients (D,E,F) in each viability and fecundity
episode for three traits, emergence week, initial size, and fill size are shown. The total
magnitude of selection across all episodes is depicted by (O) and is based on reconstructed
phenotypic variance-covariance matrix (see Chapter 2). 95% confidence intervals for total
values are based on 50 resampled data sets for each block. Total values are connected by a
line for visual clarity. Confidence intervals exceed the upper or lower axis values in five
cases in blocks 3 and 12. ...................................................................................... 182

INTRODUCTION

The mechanistic basis of contemporary adaptive evolution requires an
understanding of ecological fictors influencing individual survival and fecundity as well as
the genetic propensity for intergenerational changes in the multivariate phenotype. The
translation of these short-term dynamics ofphenotypic selection and genetic response into
observed patterns of adaptive population differentiation requires assumptions about the
constancy of the genetic variance-covariance matrix and the magnitude and direction of
selection (Lande 1979, 1982). Empirical evidence in natural plant populations shows that
patterns of phenotypic selection vary spatially and temporally (Kalisz 1986; Stewart and
Schoen 1987; Weis et al. 1992; Kelly 1992; Stratton 1992; Bennington and McGraw
1995b). Furthermore, the genetic variance-covariance matrix is not likely to remain
constant due to selection and/or drift (e.g. Shaw et al. 1995). Theoretical models designed
to incorporate both environmental and genetic variability are divided into two schools: 1)
stochastic demography that incorporates the effect of environmental variability on
demographic parameters (Tuljapurkar 1989; Orzack 1993) and 2) evolutionary models
that indicate how spatial and temporal variability in selection maintain genetic variation
(Haldane and Jaykar 1963; Barton and Turelli 1989). These two schools of thought have
yet to be unified in one synthetic theory that incorporates demographic and genetic

variability and their interaction in determining how natural selection produces divergence

2
within and among populations observed in nature. One essential component of such a

unified theory will be a detailed Imderstanding of short-term evolutionary dynamics.

One important, yet unexplored factor afi‘ecting short-term evolutionary dynamics
in natural plant populations is the impact of maternal inheritance. Maternal inheritance is
one type ofnon-Mendelian inheritance in which the resemblance between relatives is
determined not only by the Mendelian inheritance, ie. the transmission of one-half of an
individual’s genes in the fertilization process, but is also influenced by the phenotypic
efi‘ects of the mother on the attn’butes of her ofi‘spring. These maternal effects on 035ng
phenotype can have both a genetic and environmental basis. Maternal effects can increase
or decrease the similarity between mothers and their offspring, likewise they can increase
the similarity between maternal full and half siblings.

Maternal inheritance has a variety of evolutionary consequences. The most striking
efi‘ect is reversed responses to artificial selection (i.e. Falconer 1965) that can result either
from a negative genetic correlation between maternal performance and offspring
phenotype or from a negative environmental effects of maternal performance on offspring
phenotype. In contrast, positive genetic correlations or environmental effects can enhance
selection response. In addition to influencing the direction of selection response, maternal
inheritance introduces time lags in the evolutionary response (Kirkpatrick and Lande 1989,
1992). Unlike standard evolutionary models, maternally inherited traits are influenced by
phenotypic selection in the previous generation. This time lag is the direct result of
phenotypic selection in prior generations altering the distribution of maternal performance
phenotypes which directly impact the expression of phenotypes in the generation currently

subject to selection. The effect of this time lag in response to selection is twofold: 1) the

3
maximal evolutionary rate of response is approached asymptotically, and 2) the response

to selection continues after selection ceases (Kirkpatrick and Lande 1989, 1992). Thus,
maternal inheritance can alter the direction, magnitude, rate, and duration of response to
selection on a short time scale. Given the observed variability in phenotypic selection in
natural populations, these short-term effects may affect the long-term patterns of adaptive
divergence that we observe in natural populations.
MM ' g maternal inheritance

Currently, there are two approaches for estimating the genetic component of
maternal effects. Models developed by animal breeders to improve response to artificial
selection estimate specific causal components of variance by partitioning the phenotypic
covariance on a wide variety of relatives (Dickerson 1947; Wilham 1963; Eisen 1967). In
contrast, the other approach condenses causal components into single parameters
estimating the magnitude of the maternal genetic efi‘ect (Falconer 1965; Kirkpatrick and

Lande 1989).

Partitioning the phenogpic covariances among relatives
Wilham (1963) derived an equation demonstrating how the offspring phenotypic
value (Pox) is a function not only of the direct genotypic (Gox) and environmental values

(Bog), but also of the maternal genotypic (Gm) and environmental values (Em):
P OX=GOX+EOX+GMW+ETHW

Subsequent derivations of the causal components of genetic variance and covariance for

different types of relatives hinges on this construction of maternal efi‘ects.

4
The estimation of the causal genetic components that contribute to phenotypic

covariances of different types of relatives when maternal effects are present necessitate
phenotypic measurements on a wide variety of different types of relatives (Wilham 1963;
Eisen 1967). Because direct additive and maternal additive efi‘ects both contribute to the
phenotypic covariance when maternal efi‘ects are present, their statistical separation is
compromised. Crossefostering ofl‘spring post-partum is one technique that has been used
to disentangle direct and maternal genetic effects (Riska et al 1985). This experimental
approach allows the estimation of post-natal maternal effects in species exhibiting parental
care. Embryo transfer can also be used to explore pre-natal maternal efl‘ects (Cowley
1991). In angiosperms, either of these experimental approaches is not currently feasible.
Alternatively, Wilham (1980) suggests that perhaps the best types of relatives to use in
estimation are different types of first cousins where the correlation between direct and
maternal additive effects and their covariance is not as great.

Statistical techniques for estimating these causal (co)variance components fill into
two categories: 1) least-squares (Eisen 1967; Cantet et al. 1988; Cantet 1990; Cantet et al.
1992 a,b) and 2) restricted maximum likelihood (Thompson 1976; Meyer 1991, 1992).
Because sampling correlations between the various components are large as a
consequence of direct and maternal effects being confounded in maternal lineages (Eisen
1967; Thompson 1976; Wilham 1980; Meyer 1992), variance component estimates and
heritabilities and genetic correlations derived fiom them have large variances associated
with them (Thompson 1976; Meyer 1992). Even when some of the variance components

are eliminated such that models include only the direct and maternal additive and

5
environmental components, large sample sizes are required to overcome this problem

(Meyer 1992).

Estimating mtemal efi‘ect coefiicients

Falconer (1965) derived a simplified approach to estimating maternal efi‘ects in
which maternal effects are not decomposed into specific genetic components, but rather
are lumped into one term, m, the maternal effect coeficient. This generalized approach is
analogous to the estimates of non-additive maternal efi‘ects fi'om reciprocal crosses
(Topham 1966; Cockerham and Weir 1977). In this model, Falconer ( 1965) considered
the single case of a maternal trait influencing its expression in the offspring. Kirkpatrick
and Lande (1989) have extended this approach to multiple maternal and ofi‘spring traits in
which specific maternal traits can influence their own expression or the expression of other
traits in the subsequent generation. Lande and Price (1989) and Schluter and Gustafi‘son
(1993) have applied this approach to natural populations. This approach requires the
assumption that all maternal traits are measured and included in the analysis. It does not
provide information about the specific nature of the genetics of maternal efi‘ects. Lynch
and Walsh (1996) derive the quantitative genetic relationship between this model and the
Wilham (1963) approach and demonstrate that m, the maternal efi‘ect coefficient, is a
fimction of voAm, 0000..., 0505,“, 62A". and 025m. The advantage of this simplified approach
is that it can be integrated with estimates of phenotypic selection to predict multivariate
response to selection (Kirkpatrick and Lande 1989). The disadvantage is that it does not
identify the specific genetic basis of maternal effects.

Estimates of maternal effects in plants

6
In natural plant populations, diallel breeding designs allow the estimation of

maternal efl‘ects (Cockerham and Weir 197 7 ). This maternal variance component
compares maternal to paternal half-sib offspring and thus includes maternal additive,
maternal dominance, maternal environment, direct maternal additive covariance, and direct
maternal environmental covariance as well as any cytoplasmic inheritance. A number of
studies have found that seed weight, emergence time, and early seedling size display
significant maternal variance in artificial or natural environments (Biere 1991a; Mitchell-
Olds and Bergelson 1990a; Montalvo 1994; Schmid and Dolt 1994). Most of these
studies demonstrate that this general maternal effect in angiosperms is short-lived and does
not persist beyond seedling stages. Two of these studies were actually able to compare the
relative magnitude of maternal genetic vs. maternal environmental effects by using clonal
replicates in the same (Biere 1991a) or multiple environments (Schmid and Dolt 1994),
but both found that maternal genetic efiects were larger in magnitude than maternal
environmental effects.

While these studies suggest that maternal genetic effects are likely to affect the
evolutionary potential of juvenile and maternal traits in plants, no study has estimated the
contribution of specific causal conrponents to juvenile traits in plants. Thus the importance
of the direct maternal additive genetic covariance in enhancing or constraining the
evolution of juvenile and maternal traits is unknown.

ant' ' Phen ic Selection

To evaluate the evolutionary consequences of maternal inheritance, we need

information on the nature ofphenotypic selection on traits influenced by maternal effects.

The methodology for estimating selection on a multivariate phenotype is well developed

7
(Lande and Arnold 1983; Arnold and Wade 1984 a,b; Phillips and Arnold 1989; reviewed

by Brodie et al 1995). This multiple regression approach can be utilized to estimate the
total magnitude of selection on traits even when mortality eliminates individuals before
they express all phenotypic traits of interest (Lynch and Arnold 1988). In that respect it
difl‘ers from the path analytic models of Crespi and Booksteirr (1989) in which causal
relationships between traits are constructed for each selection episode. In path analytic
models, neither the total magnitude of selection nor the total nature of direct and indirect
effects of traits can be determined. An alternative approach to estimating selection that
allows more complex fitness firnctions was initially limited by considering only a single
trait (Schluter 1988). However, recently this nonparametric approach has been extended
to multivariate descriptions of selection (Schluter and Nychka 1994).

In natural plant populations, juvenile traits are likely to display maternal inheritance
(see above). Second, juvenile traits like seed size and emergence time can influence early
biotic and abiotic interactions and produce a very skewed distribution of fitness (Stanton
1985; Waller 1985). As a result of these early acting selective events manifested both by
mortality and differences in individual size, the total magnitude of selection acting on
juvenile traits can only be determined by an episodic selection analysis (Arnold and Wade
1984 a,b) modified to account for changes in the phenotypic variance-covariance matrix
due to early mortality (Lynch and Arnold 1988). The magnitude of phenotypic selection
on traits determines the importance of maternal inheritance either by direct selection on a
trait displaying maternal inheritance or by indirect efl'ects of selection on phenotypically

correlated traits.

8
In addition to phenotypic selection on juvenile traits that are likely to be influenced

by maternal inheritance, it is possible for maternal traits to directly influence the survival
and fecundity of their oflspring. This type of maternal effect on offspring fitness has been
termed maternal selection. The inclusion of maternal attributes in a selection analysis
allows the estimation of maternal selection, a form of group selection (Heisler and Damuth
1987). Like maternal inheritance, maternal selection influences the evolutionary dynamic.
Kirkpatrick and Lande (1989) illustrate that maternal selection can result in maladaptive
evolution. It is also different from other forms of selection because the magnitude depends
on the resemblance between mothers and their offspring. Maternal selection represents the
possibility for different levels of selection to influence the evolutionary response for traits
that display either standard Mendelian inheritance or maternal inheritance.
Evolutionary consequences of maternal effects

In nature, the phenotypic expression of a trait can be influenced by the phenotype
of the individual’s mother, maternal inheritance. Furthermore, the survival and fecundity of
an individual relative to other individuals in the population can be directly influenced its
mother, maternal selection. In this dissertation I explore the evolutionary consequences of
these two types of maternal effects. In Chapter 1, I quantify the magnitude of maternal
inheritance by carefirlly partitioning phenotypic covariances among relatives into explicit
causal components of genetic variance and covariance. This estimate of the additive
genetic variance-covariance matrix for each trait suggests how maternal inheritance
influences univariate response to selection. In Chapter 2, I document the nature of
phenotypic selection on juvenile traits likely to display maternal inheritance in a natural

population. In addition to estimating the total magnitude of selection, I also describe the

9
nature of the indirect efl‘ects of selection. How does selection acting on one trait produce

responses in phenotypically correlated traits? This multivariate selection analysis indicates
how various juvenile traits directly influence survival and fecundity and suggests
hypotheses about causal agents of selection. In Chapter 3, I explore the extent to which
selection may discriminate not only among individuals ofi‘spring, but also among their
mothers. Maternal selection, a form of group selection, indicates that mothers can directly
influence the fitness of their offspring. I also explore the extent to which spatial variation
in selection can influence among fimily differences in fitness.

These chapters deal explicitly with the two components of the dynamic equations
for evolutionary change: inheritance (Chapter 1) and selection (Chapters 2 and 3). In each
chapter I explicitly address the evolutionary implications for the component of interest.
The synthesis of these components as they are influenced by maternal inheritance and
maternal selection remains a challenge. My long-term goal is to unify these estimates into
a single evolutionary model to make explicit predictions about response to selection when
both maternal inheritance and maternal selection influence offspring phenotype and Mess,
respectively. The impact of maternal effects on short-term evolutionary dynamics are
likely to be central to our understanding of the adaptive divergence of both juvenile and

maternal traits within and among natural populations.

Chapter 1
MATERNAL INHERITANCE AND ITS EFFECT ON ADAPTIVE EVOLUTION:

A QUANTITATIVE GENETIC ANALYSIS OF MATERNAL EFFECTS IN A
NATURAL PLANT POPULATION.

INTRODUCTION

Phenotypic similarity between mothers and offspring resulting from standard
Mendelian inheritance can be altered by phenotypic effects of a mother on traits in her
young, termed maternal inheritance (Kirkpatrick and Lande 1989, 1992; Lande and
Kirkpatrick 1990). In a classic study of maternal effects in mice, Falconer (1955, 1965)
found that large mothers had many small young. At maturity the females from these large
litters had only a few large young. The lack of resemblance between mothers and their
daughters in litter size and ofl‘spring size was mediated by the phenotypic efi‘ect of
maternal size and its relation to maternal provisioning ability. Interestingly, this maternal
efi‘ect had short-term evolutionary consequences. Artificial selection on litter size showed
a reversed response to selection in a single generation (Falconer 1965). More recently in
the collared flycatcher, Schluter and Gustafison (1993) experimentally quantified how
maternal condition and clutch size influenced the resemblance between mothers and
daughters in clutch size. In their study, resemblance in clutch size was moderated by the

negative effect of maternal clutch size on a daughter’s condition and by a positive effect of

10

l l
maternal condition on a daughter’s condition. The balance between negative and positive

maternal efi‘ects led Schluter and Gustafi’son (1993) to predict a positive response to
selection on clutch size in the first generation. These two examples illustrate how maternal
inheritance can impact the evolutionary process. The underlying genetic architecture of
maternal inheritance can be described by the genetic basis of traits such as maternal size,
condition, or provisioning ability, and their genetic correlation with traits expressed in the
juvenile stages such as birth weight or offspring size. It is the underlying genetic
architecture of maternal inheritance that determines how trans-generational effects will
impact the process of adaptive evolution.

The best model to date for exploring the underlying genetic architecture of
maternal inheritance was developed by Dickerson (1947) (hereafter Dickerson’s model).
Dickerson’s model considers two traits, the individual trait of interest and the maternal
trait affecting its expression. The goal is to partition the covariance (i.e. resemblance)
between mothers and their offspring into explicit Mendelian and maternal components
(Table 1; Dickerson 1947; Wilham 1963, 1972; Eisen 1967; Cheverud 1984; Lynch 1987;
see Cheverud and Moore 1994; Lynch and Walsh 1996 for current reviews). The nine
possible causal genetic and environmental components from Dickerson’s model (Table 1)
are obtained by the statistical partitioning of phenotypic covariances of the individual trait
among difl‘erent types of relatives generated in a complex breeding design. The maternal
trait is a composite trait termed maternal performance and is unobserved. While
Dickerson’s path analytic approach assumes a causal model of maternal inheritance, the
estimates of causal components are correlational in nature because they are derived from

covariances.

12

 

 

 

 

 

 

 

 

o _ _ _ _ 0 ~ m
o o o o o c we _ o
o o o o o o 3 ~ o
o o o c Q _ 3 ~ Em C
_ o o o N: _ Em C
0 fl o o o o 0 Em
C o _ ~ ~ o _ 3 —
o o o o o o o o
Emomb emub Swab EQNO Same amen—O E<e<b eQNO

 

 

 

_

E _
Q ~
v:
Q _
Q ~
Q _
3 _
e<~b

 

 

88580 Em 33> ,8 3:08.580 38:0

€2.38:—
wfiambebzaa—o: 3:825
wfiefibebhm
wfinmbebzafio: 13.882
wfiunmbeéan

3.5 :3 £55,

3.5 =:m

3.5.32. 1:38.“

o. cam—«fl e. .0. glam—Maegan...

.039» 2.38:3: Riggs :a 8 8:95:00 3:08:88 83>» 385 35E? 33.3%:— .oo:afi2_:_ 18:85:
3 26 30ch 3:38.: 3 30.8: E can? .oogfioafi 5:28:22 2.85% 8 26 33.8 82% 38:0: 8 «am—8:8
2F 302883: 60:35, 5:088:33 38:8 m 66:88:: 38:0: O 63:3 0258: 38:0: < 80:38.8
83 8m 808808 on: 80382 32.3 .352 8882:. ”$2 :85 8:5 85580 :5 8:333:
8:08:88 £83m .328 SE 3352 .8 $8 :38 .8 :03 :8 355183 2.58:3: 2: 8838:.“ A 03;

13
In this paper I utilize Dickerson’s model to estimate the underlying architecture of

maternal inheritance, ie. the specific causal genetic and environmental components of
variance, in a winter annual plant, Collinsia verna Nutt. (Scrophulariaceae). My goal is to
describe how maternal inheritance affects the magnitude and direction of predicted
response to selection for a number of traits expressed at different stages in the life cycle.
The Maternal Inheri_t_ance Model

With simple Mendelian inheritance (Figure 1A), the phenotypic value of a trait
(P0) is determined by additive genetic (A0) and environmental (E0) components where the
subscript 0 refers to the individual trait of interest. In this two generation path diagram, an
ofl‘spring in the second generation (x) receives 1/2 of its genes from its mother in the
previous generation (w). The translation of additive effects into phenotypic value for this
trait is denoted by the direct (i.e. Mendelian) heritability (h).

In contrast, in Dickerson’s model of maternal inheritance (Figure 1B), the
phenotypic value of the trait of interest (P0,) is influenced not only by Mendelian
inheritance (described above), but also by the unobserved maternal performance
phenotype (PM), subscripts m and w referring to the maternal performance trait and the
maternal generation, respectively. The phenotypic value (PM) for maternal performance is
determined both by additive genetic (A...) and environmental (Em) conrponents. Direct and
maternal additive effects can be genetically correlated (rm). The resemblance between a
mother and her offspring (Figure 1B) can be influenced by four components: 1) maternal
additive genetic variance (02A...) and it translation into maternal performance denoted by

the heritability (hm), 2) direct-matemal additive genetic covariance (0mm. ) standardized

14

Figure 1. Path diagrams of Mendelian (A) and maternal inheritance (B and C) (after
Dickerson 1947; Wilham 1963; Cheverud 1984). Under Mendelian inheritance in model 1
(A), the additive genetic value (A) and the environmental value (E0) determine the
phenotypic value (P..) where the subscript 0 refers to the offspring trait and w and x refer
to the maternal and ofispring generations, respectively. In Model 2 (B), maternal
inheritance is determined by the phenotypic effects of the maternal performance trait (P...)
and its additive genetic (A...) and environmental (E...) components, the subscript m
referring to maternal performance. The genetic correlation between direct and maternal
traits (erAm), and the square root of the direct (h..) and maternal heritabilites (h...) are
illustrated. The maternal effect coefficient (m) indicates the extent to which the maternal
phenotype influences the phenotypic value in the offspring independent of additive genetic
effects. In Model 3 (C) maternal inheritance includes the potential correlation between
direct and maternal environments (r505...)

15

 

 

 

 

 

 

 

 

A. MODEL 1
GENERATION: :
MATERNAL 3 OFFSPRING
GENOTYPE+PIIENOTYPE § GENOTYPE—>PHENOTYPE
AW m F A°\
Eo ’ Pox
B. MODEL 2
A... “2 e A.
erAm ierAm \
E0 $ P0X
Amw 1/2 ’ Am
B... r me3
C. MODEL 3
A... “2 ; ~ A.)
g be
1’ AoArn
i * Pox
Amw

 

 

Figure 1.

16
by the additive genetic variances of both the individual and maternal traits as a genetic

correlation (rm), 3) maternal environmental variance (025m), and 4) the purely
phenotypic efl‘ects of the mother on her ofl’spring (m), termed the maternal efl‘ect
coeflicient. In a second version of Dickerson’s model (Figure 1C) a fifth component, the
environmental covariance between generations (6505,.) standardized as direct-maternal
environmental correlation (r505, ), can also contribute to the resemblance between a
mother and her offspring. Thus, relative to standard quantitative genetic models of
Mendelian inheritance (Figure 1A), the decomposition of the trait, (Pox), into genetic and
environmental components is complicated by the additional paths of maternal inheritance.
The response to selection on the maternally inherited individual trait (Pox) will be
determined by the realized heritability (112,) (Dickerson 1947; Wilham 1963; Van Vleck

197oy
11.2 ——-(e’AO+3/2eA,Am+1/2e2,,m) /e% (1)

The realized heritability is a function of the direct additive genetic variance (62A,), the
maternal additive genetic variance (02M), and the direct-matemal additive genetic
covariance (UAW) relative to the total phenotypic variance (62p). When the direct-
matemal genetic covariance (cm) is negative and >|2/302A0+1/302Am|, the response will
be in the opposite direction to selection. Similarly, positive maternal additive genetic
variance (02M) and direct-matemal genetic covariance (cm) can accelerate response to
selection. Thus, the underlying genetic architecture of maternal inheritance influences the

direction and rate of adaptive evohrtion. The time lag in the maternal inheritance can also

l7
efi‘ect the rate, direction, and duration of the selection response (Kirkpatrick and Lande

1989, 1992; Lande and Kirkpatrick 1990).

Empirical estimates of the causal variance components affecting evolutionary
response in domesticated and experimental laboratory species for traits such as litter size,
birth weight, and weaning weight show that maternal additive genetic effects can be
substantial (e. g. Bondari 1978; Cantet et a1. 1988; Shi et al 1993), can increase from birth
to weaning (Shi et al 1993), and generally display significant negative direct-matemal
additive genetic covariances (Figure 2). Maternal efl‘ects on a single trait through
ontogeny decline after weaning (Atchley 1984; Cheverud et aL 1983 ).

In contrast in natural populations, empirical estimates of causal variance
components determining maternal inheritance are lacking. In plants, the magnitude of
maternal effects estimated by less detailed methods also shows a decline through
ontogeny. In general, traits expressed early in the life cycle such as seed weight,
emergence time, or seedling size are influenced more strongly by maternal genetic effects
than direct (i.e. Mendelian) genetic eflects (Biere 1991a; Platenkamp and Shaw 1993;
Montalvo and Shaw 1994; Schmid and Dolt 1994). The duration of maternal genetic
effects beyond the seedling stage is rare (Schmid and Dolt 1994). Maternal genetic efi‘ects
tend to persist longer in competitive environments (Schnrid and Dolt 1994), a pattern
analogous to the persistence of initial size difi‘erences in more competitive environments
(Gross 1984; Stanton 1985; Waller 1985; Weiner 1985, 1990; Stratton 1989; Gross and
Smith 1991). While maternal environmental efi‘ects are well documented (reviewed by

Roach and Wulfl‘ 1987) and can persist for multiple generations (Miao et al 1991; Lacey

18

Figure 2. Summary of studies estimating direct-maternal additive genetic correlations
based on three difiermt estimation models: 1) the animal model included additive and
environmental components only, 2) the full model also included dominance components,
and 3) cross-fostering models estimated post-natal maternal effects (Bondari et al. 1978;
Cantet et a1. 1988; Southwood and Kennedy 1990; Shi et a1. 1993; Van Sanford and
Matzinger 1982; Everett and Magee 1965; Young and Iegates 1965).

19

 

 

ii

bumowémEU

 
 

w” .M . ”a”. a"
fi. .730”. .
w‘ . .

+13.” ........

fl

 

/
7

 

 

 

 

 

.N 8%;

Eon—e: ZOE—.«JgOU UPmZm—O A<ZME<§FU§D
ad 5o md Md _.o .6. m6- nd 50. ad- 7

_

 

C‘OOFOVWV'MN—

O

ADNEIIIOEIHJ

20
1991), studies examining the magnitude of maternal genetic and maternal environmental

efl‘ects have found that maternal genetic efl“ects predominate over maternal environmental
effects for traits expressed early in the life cycle (Biere 1991a; Schmid and Dolt 1994).
In this paper I present the first quantitative genetic analysis of maternal efl‘ects,
estimating the causal variance components relevant to maternal inheritance in a natural
plant population. Causal components are critical for predicting the dynamic role that
maternal inheritance plays in adaptive, multivariate evohrtion. By examining a large
number of traits expressed both early and late in the life cycle of the winter annual, C.
verna, I explore how maternal inheritance changes through ontogeny. In addition, I
comp are how genetic correlations among traits within a generation difl‘er from the
between generation genetic correlations associated with maternal inheritance. The goal is
to describe the underlying genetic architecture of maternal inheritance and its implication

for evolution by natural selection.

MATERIALS AND METHODS
W
C. verna (Scrophulariaceae), a winter annual plant, germinates in the fall in
response to diurnal temperature fluctuations (Baskin and Baskin 1983 ), overwinters under
the leaf litter and snow as a small rosette, and bolts and flowers in mid to late May in the
mesic floodplain forests throughout the midwest. Seed and seedling traits vary
significantly among maternal families (Thiede, tmpublished data) suggesting the likelihood

of maternal inheritance early in the life cycle. Seed and seedling traits also strongly

2 1
influence individual survival and fecundity (Kalisz 1986; Chapter 2). Thus, maternal

inheritance is also likely to afl‘ect adaptive evolution. In this study, I consider traits
expressed at four stages in the life cycle: seed, seedling, overwintering fall rosette, and
pre-flowering spring rosette (Table 2). By considering the same trait at multiple stages, I
evaluate the magnitude of maternal inheritance through ontogeny.

In Dickerson’s model, estimates of causal components of variance and covariance
are obtained by partitioning the phenotypic covariances among relatives (Table l) for
traits hypothesized to be influenced by maternal efl‘ects. In this approach a single trait is
measured in various relatives, while the maternal trait exerting the efl‘ect is not quantified.
The magnitude of the maternal components is estimated solely by partitioning the
covariance of the trait of interest into causal components assuming different models of
inheritance. In essence, this design treats the maternal effect as a composite of all maternal
traits that influence a particular trait in the ofl‘spring (Cheverud 1984; Cheverud and
Moore 1994).

Three Generations

I obtained the seven types of relatives in Table 1 fiom a three generation breeding
design (Figure 3). In the first generation, 100 wild grandmatemal (F1) individuals bearing
naturally pollinated seeds were collected every 2 m along a 200 m transect fi'om a natural
population of C. verna in Kalamazoo County, MI in May, 1991. Twelve seeds from each
grandmatemal plant founded the second generation (F2) that was grown to maturity in the

greenhouse. The F2 individuals contributed to the estimates in one of two ways: 1) a

22

 

Nun
Ge
30
oo:

vow
oon
oov
Eh
Eh
now
own
”no
own
I.»

a

8.: 2.0
3.3 coo
2.2 3.0
$60 3.0
210d
33% 8.0
33 2 .o
3.: 2.0
on: cod
03:. 2 .o
2.2 :0
2.2 3.0
32 2.0
:30 8.0
2.3 3.0
>|o “WM
gauge:
3

ood
mod
2 .m
end

soda
2.6m
3.3
owN
nmfi
and—
mm.m
Swan
no;V
omw
x

N?
Ev
mwv
go

Now.
8N
boa
wmv
wmv
bmv
wmv
mmv
53‘
mvo

wmdm m— .o
om A m boo
wvfim boo
bmdN woo
356:805
Nude ovo
no.3V oNo
mofiN w— .o
\rmoo ooo
ovdo :o
deN 3 .o
no. _ m boo
Zoom mvo
oodN moo
hm .wN vo.o
Nu wilt“
Own—oases
3

bed
oo.m
and
ooé

om.o~
bvo~
2&3
vmd
no.5
mod
mod
mwom
wed
0o;V

ASHE 60080:. move—boo
Aaao 365% 8338
0.003 008325—

35 awe; e80

0026— .«e 8:852

:55 amazes .20
Aaao 505 8335
0030— .«e hoe—52

Aaao amazes .35
9:5 80080:. goo—boo
GEE 6885 goo—boo
000w 00§w~08m

A05 :33 cream
35 £0?» 8%

0... al.—.0. 390182:—

0586“ am

3680

0:00am mam—am

03003— :0..—

E880

30m

:wll0 3m

 

.m0w30 0025.3 :80 «0 0:29:00 330:0 a .3

808008 b88350 003 588.0% deco—boo wfiambo 05 a 0—23 .5w:0_ none—boo Ea 3 :33
mm 28 0083080 00 523 2.0 awefl £6053 me own—020 05 mm momma 95 «a 50080:. coco—boo
3:0an 8m 20¢ 28 834.50% 05 £3 E :3on 803 Ammo mam—ammo 0—53 .00seneoohw 05
a :3on 0.53 3.5 $5.3m .maouabgw mm 28 mm flop E 00?.» 0.03.0.5 0382—00.. 00 3.50008
3:5 29580.:— Sm 0% 0—988 one . gaggmo E06808 .85 23:30 .502 .N 030,—.

23

Figure 3. Three generation breeding design in which field-collected grandmatemal
families (F1, GD“, n=lOO) provided seed for the parental generation (F2). Of the twelve
seeds planted from each granddam (Fl), one was randomly assigned as a sire (Sn, n=24) or
dam (Do, n=72); while the other eleven grandmatemal filll-sibs were considered paternal
(PR..) or maternal relatives (MR,.). Greenhouse-raised parents produced up to 40 offspring
(F 3) that were divided between greenhouse and field environments.

GD;

GD1

F1

01...O4o 01...O4o

O]...O40

F3

24

25
subset of individuals served as parents in the nested breeding design to generate the third

generation (see below), 2) the remaining individuals were classified as parental relatives
(Figure 3). To determine the coeficients of causal components for parental relatives
(Table 1), I assumed that F2 individuals within an F1 grandmatemal family were full-sibs
produced by natural outcrossing. This assumption is justified because the outcrossing rate
in this population was consistently greater than 0.85 for three years (including 1991).
Furthermore, a high estimate of correlated matings suggests that these outcrossed
individuals share the same father (Holtsford et al in prep).

To produce the third generation, one individual (F2) from each grandmatemal
family was randomly assigned to serve as a sire or dam in a nested breeding design in May,
1992. Twenty-four sires were crossed to three dams per sire in a standard nested design to
generate 24 paternal half-sib and 72 matemal full-sib families (Figure 3). Flowers were
emasculated in bud and pollinated within 5 days post-emasculation. Pollinations were
performed on all floral whorls to control for position efi‘ects. Fruits were harvested as they
matured. An accident in the lab eliminated 10 maternal full sib families resulting in a total
of 62 maternal full sib families.

The third generation (F3) was planted in a randomized block design in two
locations: greenhouse (n=87 l offspring from 24 sires and 58 dams) and field (n=1212
offspring from 24 sires and 62 dams). Seeds were planted to a depth of 1 cm in Sunshine
seedling mix either in 96 well trays (F2 and F3 in the greenhouse) or in 2 cm long plastic
tubes 16 mm in diameter (F3 in the field). In both locations one individual fiom each
maternal full-sib family was planted into each of 20 blocks. In the greenhouse each block

consisted of a 96 well tray, all 20 on a single bench in the greenhouse. In the field each

26
block of F3 individuals was divided into three sets of 24 and each set was then randomly

assigned to one of three 0.5 m2 quadrats at one of 20 locations. This planting design was
utilized to maintain natural seed/seedling densities in a given quadrat. The 20 blocks
spanned the natural habitat and included forest edge and interior.

T331_1t' 5 Measured

To estimate maternal inheritance, I measured the same trait in the F2 and F3
generations: ten traits at four stages in the life cycle in the greenhouse or four traits at
three stages in the field (Table 2). Prior to planting, seeds were weighed to the nearest 0.]
microgram Afier seedling emergence, seed coats were carefully excavated from the soil,
air dried, and weighed (F2 and F3 in greenhouse only). Embryo weight was calculated as
the difi‘erence between seed weight and seed coat weight. Thus, embryo weight more
accurately reflected the diploid genetic composition when compared to seed weight which
contained both the diploid embryo, a small amount of residual endosperm, and the diploid
maternal seed coat.

Seedling emergence date was scored weekly in the field (F3) and every 3-4 days in
the greenhouse (F2 and F3) from September to the beginning of December. Emergence
date was defined as the first date when cotyledons were expanded. At emergence, I
quantified seedling size by measuring cotyledon diameter using a template of circles of
increasing diameter in increments of 0.5 mm (F3 ). In the F2 generation, cotyledon
diameter at emergence was the average of cotyledon length and width.

At two subsequent stages, in late fall prior to overwintering and in early spring
prior to flowering, I quantified individual size by measuring three traits: cotyledon

diameter, leaf length of the most basal leaf (mm), and number of leaves. Fall rosettes were

27
measured in November (F2) or early December (F3 ). In December, greenhouse grown

plants (F 2 and F3) were transferred to a sheltered area outdoors and covered with a thick
layer of leaf litter to mimic natural field conditions. Overwintering survival was greater
than 90%. In April plants were returned to the greenhouse and I transplanted a random
subset of 2-3 individuals per maternal granddam (F2) or all seedlings (F3) into 15 cm2
pots filled with a 2: 1:1 mix of Sunshine seedling mix, perlite, and turface. Size traits were
measured on pre-flowering spring rosettes (F2 and F 3) afier transplanting.
Estimation of Genetic and Environmental Causal Components

I estimated six of the nine causal components relevant to maternal inheritance
(Table 1), additive (62.40, 02A,“, CAM) and environmental components (0ng, 025m, 0505.“),
by considering three sequential models of inheritance in a hierarchical approach (Figure 1).
The simplest model of inheritance was a purely additive Mendelian model (Figure 1A,
hereafter model 1) in which 0on and 025., were estimated. In model 2, maternal inheritance
was incorporated by estimating three additional components, 02A,“, CAM, and 025...
(Figure 1B). In model 3, all possible additive and environmental covariances were
considered by including a sixth component (6505,.) (Figure 1C). The hierarchical approach
allowed me to ask: 1) Did the more complex estimation model for maternal inheritance
better describe the data? 2) Which causal components were significant in each estimation
model?

The estimation of additive and environmental components only is often necessary
(e. g. Bondari 1978; Meyer 1992; Shi et al. 1993) because obtaining a sufficient number of
relatives to estimate all nine components is very diflicult (see Cantet et. al 1988 for an

example of the design required for the full model). In addition to standard quantitative

28
genetic assumptions of random mating, linkage equih’brium, and the absence of epistasis

and of genotype by environment interactions (Wilham 1963; Eisen 1967; Thompson
1976), the three models, therefore, required the assumption that direct and maternal
dominance variances and their covariance (0200, 029.“, ODoDm) were zero. To test the
assumption of zero dominance variances and covariance, I included them in some
preliminary analyses and discuss these results when relevant.

The nature of transmission of maternal inheritance dictates that direct and maternal
components are correlated in maternal lineages (Table 1). This biological reality results in
a statistical limitation in estimation because causal components are correlated even when
numerous types of relatives are considered (Eisen 1967; Thompson 1976; Wilham 1980;
Meyer 1992). In this design, correlations among components based on the coeficients in

Table 1 showed that 021)... and 025,“ were perfectly correlated (p<0.001). Therefore, only
021)... or 025,“ or their sum was estimable (Thompson 1976). Maternal component, 02A,“,
was positively correlated with 0AM , 029m, and 025m (r=0.85, 0.80, 0.80, respectively,
p<0.05 for all) and direct components, 025,, and 02130, were also positively correlated
(r=0.94, p<0.001). However, even in more complicated designs involving 10-13 types of
relatives, Eisen (1967) found similar correlations among causal components. Thus, the
inclusion of more types of relatives did not necessarily decrease sampling correlations
among causal components. The inability to estimate all nine causal components and the
high sampling correlation between components are both issues that affect the
interpretation of the following analysis and are limitations of this approach.

Anabzsis

29
To estimate the causal components of variance and covariance, I employed a

modified version of a six component restricted maximum likelihood (REML) program
(Shaw and Shaw 1992; Shaw 1987). RENE provides unbiased estimates, is not sensitive
to lack of balance in the data, is flexible in handling non-standard designs, and assumes
multivariate normality (Shaw 1987 ; Thompson and Shaw 1992; Meyer 1992).

Each normally distributed trait was analyzed separately to estimate the causal
components related to maternal inheritance. A fixed efl‘ect for generation was inchrded in
each model because trait means differed between F2 and F3 generations (Table 2) and
including a fixed generation effect in the model resulted in smaller likelihoods. The
convergence criteria determining the termination of iterations was set at 0.001. Non-
negativity constraints on causal component estimates were not imposed because of their
adverse effect on significance tests (Shaw 1987).

The log-likelihood ratio test was utilized to evaluate significance in two contexts.
First, I evaluated the significance of the models by calculating twice the difference in log-
likelihoods for sequential models (1-3). This statistic has a chi-square distribution with
degrees of fi'eedom determined by the difference in the number of components estimated
in the two models (Shaw and Shaw 1992; Shaw 1987). Second, I utilized this test to
evaluate the significance of all components (except E0) within a given model To test the
significance of each component, I constrained the component of interest to be zero,
obtained the log-likelihood of the constrained model, and compared twice the difi‘erence in
log-likelihood's between the constrained and full models to a chi-square distribution with

one degree of fieedom

30
The estimates of variance components were used to calculate direct and maternal

heritabilities and direct-maternal genetic correlations. Resampling methods required to
determine the standard errors around these heritabilities and genetic correlations would
require inordinate CPU time. Here I indicate the significance of heritabilities and genetic
correlations based on the significance of the variance components in the numerator of each
respective ratio. Shaw and Platenkamp (1993) used the same approach suggesting that
significance in this case reflects the potential for evolutionary response, but not the rate of
evolutionary response. The calculation of realized heritability has several components in
the numerator (equation 1) and, therefore, no significance is indicated.

An important assumption of this REML analysis is the independence of error
terms, ie. that the contribution of random environmental effects contributing to each
individual’s phenotype is uncorrelated among individuals and therefore, does not affect
their phenotypic covariance. This study was specifically designed to estimate maternal
effects which if not included in an analysis can lead to the violation of this assumption. The
presence of 0290, 029..., and 0909..., or other factors such as uniparental or cytoplasmic
inheritance could have inflated some phenotypic covariances and thus violate the
assumption of independent and random error terms. A second bias resulted from not
estimating m, the maternal efi‘ect coeficient, a scaling factor for maternal phenotypic
effects that afl‘ected the dam-offspring covariance. A bias in some phenotypic covariances
would necessarily result in errors in the estimation of all components because they are

estimated simultaneously.

Within-@eration Genetic Correlations

3 1
To estimate genetic correlations among traits, I considered each pairwise

combination of traits in two hierarchical models, Mendelian inheritance in model 4 (Figure
4A) and maternal inheritance in model 5 (Figure 4B). In model 4, I included only the
direct additive (02A,) and environmental components (025,) for each trait as well as their
respective covariances (Ommoz, 0501502) to estimate direct genetic correlations
(erWzXFigure 4A). In model 5, I incorporated the components relevant to maternal
inheritance to estimate genetic correlations for direct (erlez) and maternal (mum)
efi‘ects (Figure 4B). However, the structure of the bivariate model depended on the results
of the separate analysis of each trait. For example, in figure 4B I show all possrble
components that would be estimated if both traits were best described separately by model
3 (Figure 1C). If both traits were described separately by model 2 (Figure 18), then the
direct-maternal environmental covariances (0505,“) would not be estimated. Thus, the
structure of model 5 varied depending on the traits included. I estimated direct and
maternal genetic correlations when both traits displayed maternal inheritance or only direct
genetic correlations when only one trait displayed maternal inheritance. All covariances
were unconstrained (i e. 0A0 1A02, 0501502, 6A,“ 1M2, O'Emlanz) except the covariances between
traits for direct-matemal additive covariance (voAmleAmz) and direct-matemal

environmental covariance (O'Eofinmosmz) components that were constrained to zero.

32

Figure 4. Path diagrams to estimate genetic correlations between pairs of traits. In Model
4 (A), the genetic correlation between two traits inherited in a Mendelian fashion is
denoted by erlez, where l and 2 refer to the two traits in the model. All other symbols
are identical to Figure 1. In Model 5 (B) both traits are maternally inherited. Within
generation genetic correlations between direct additive effects (erlez), and between
maternal additive efi‘ects(rAm1Am2) are depicted. All possrble components are depicted,
however, models were simplified based on the best inheritance model determined by the
univariate analysis of each trait.

33

.S. as»:

 

% N53,”:

 

 

 

x l I 3
No< N: No<
50¢ A Sm
N235 9:05
o m
_ «A W
5 h 3—
o< I m N: o<
mm>SZNIQ+mm>SZm0 WEEZ§Q+MQ>SZMO

02:35th w A<Z¢E<§

34

.mV 2&5

 

 

 

 

 

 

 

 

 

Nam
xmonm
3No< «:2 :9:
==m
3 _E< ~o< _o<.~
cmomh
X—Om
: A 3.
o< M Q _ c<
ESmeEdlmm>SZmO W ESmen—‘lgszmo

0255950 A<deh<2

35
RESULTS

Comparison of estimation models

I detected significant (5on in the simplest inheritance model (1) for 11 out of 14
traits (Table 3). This estimate of 02A,, was based on all seven relatives in Table 1 and,
therefore, may be inflated by maternal genetic efi‘ects in 6 of the 7 phenotypic covariances.
Thus, these significant 02A,, indicated the potential for 11 traits to be influenced by
maternal effects. Lack of significant 02,“, for three traits, fall leaf length and number of
leaves in the greenhouse and fall cotyledon diameter in the field, suggests that these fall
size traits were not influenced either by direct additive or by maternal efi‘ects. Two of
these traits, fall leaf length and number of leaves in the greenhouse, also had higher
coefficients of variation relative to all other traits displaying significant 02A,, (Table 2).

In model 2, the addition of three maternal components of variance (02m, 025...,
cm) significantly improved the likelihood of the estimation model for 8 of the 11 traits
that displayed significant 0on in model 1 (likelihood ratio test, df=3, p<0.05; Table 4).
Thus, the five component maternal inheritance model better described the data at hand for
six traits in the greenhouse, seed weight, embryo weight, cotyledon diameter at emergence
and in fall and spring, and spring leaf length and two traits in the field, seed weight and
emergence week. Model 2 did not significantly improve the likelihood for three traits
(emergence date, spring number of leaves, and field cotyledon diameter at emergence) and
did not converge for fall cotyledon diameter in the field.

In model 3 the addition of 0501;... again significantly improved the likelihood of the

estimation model for two traits in the greenhouse, seed weight, and. embryo weight (Table

- uI-tl\‘¢ N. n\~a-w.~t

36

 

 

$0 0.8 2.0 8.0- 2.807 3385 so :00
3.0 03 1.2.0 00.0 :00? 5.08.3 80
a: NS .3040 $0- ~33- 083 380580
«3 :3 Ease 00.. 3.002- 2003 080
3m;
«30 8.2 3.2 .0 2...: 00.00:- 823.8 50:52 0550
00.2 3.2 it: .0 2.0 3.82. 60.3080 0350
03 00.0 :38; 03. $.02 7 50:3 8858 0350
~00 Ea 8.0 23. 3.32. 03.300 28:2 3
00.2 2.2 :0 :8 38:. 50:308.— .3
as 3.0 1:30 00.0 8.00:- spasm 80 as
Ea on. ............2.0 and 0205- 8.08380
5:8 3.0» £132 No.2 8.3%. 35 880880
>2 0; :330 42 3.000- 5003 cream
43 NS :taqo :2 3:02 7 £003 800
m0aomzmm¢o

“8&0 002500:—
afle asp 3% 02$ 03 as;

 

.fioo.ovn......_..... woodvav—odt... .deavgo

a... .modval .99 008: 0_ 80:00:80 4000 we Ragga .00nm0 03800 he.“ N

050,—. 00m 0080005 0.3 103 0050.53 0:50:21 :33 28 A3003 35808300

82:. .33 30002 85. .800 802080 use 22.. 32:02.
40008 :3 05 mo coca—00E we. 2:. .0auofiuoh>=0 20m 23 asoauoohw Bu

030000800 .0030 we 00:08:00 002500.: 8582:: 008.5000 fl 3002 .m 030,—.

37

 

02 .5088 .00 000

 

80 :0 00.0 ~00 - 000 00.0 - 00.0 00.000 - 080805 80
~00 000 0~0 -0 - 0~0 0~0 000 - 000.000 - 083 880.200
000 00.0 - :00 00.0 000 :0 000 03.00.0000- 0.00003 080
808.0
00.00 000 - 00.0 - 00.0 00.00 80 0000 0~.000~- 038000 80802 0800
N— .m— iii—0.? . Nb.— .. team‘s cod SIS—N6 ~00 ****w~ .MVMT £33.03 mam—am
00.0 ............0 0 .0 - 000 - ......~0.0 00.0 ............00.0 00.0 ............00.000 0- 5080 000000000800
000 00 .0 - 00 .0 - 00 .0 00.~ 00.0 00.0 - 00 .0000- 0300000 80802 000
0000 00.0 00.0 ~0 .0 - 00.00 ~00 - 00.0 000~00- 5080080 000
0~.0 2.000 - 00.0 - .000 00.0 ............000 00.0 :00 .~000- 50285 .00 :00
~00 ............00.0 - .000 - 2.000 000 3.000 00.0 3.00.000 - 8085 .00
0000 00.0 - 00.0 - 00.00 00.00 00.0 ~0.00 00.0000- 85 880880
000 00 .0 - 0~0 0~0 00.0 00.0 000 ......._....00.000 .. £003 00580
000 00.0 0000 00.0 000 S .0 ~00 ..._........0~000 0- 0.00003 0000
0090:2055

Soho
006 .230 500 2006 0000 200 00me 002500:— w3 008,—.

 

.02 00 0880000 20 00000308 80 0020 0000.80.03...

.0000V0v000t... 000V0v000 t. .000V0v0 .0... 003 0200 20 00000088 008 00 808000000 20 000 0003 0 .0008 8
0230.. 0002500:— w£ 05 8 808006098 05 .00 00803035 .0080 03800 8.0 N 030B 00m 00080008 0000 103008000000 00me080:.—
33 .80 08080083 30:00 0080008 080 800% 0>u 08 .Bobo aetfioaow 0003 05.00 003308 000008 :8 05.00 008500:

we- 00:. 080888380 20¢ 080 00088008 Bu 080800800 .0030 .00 00:08:00 0008500:— 808808 0000008000 N 00002 .0 030,—.

 

38

 

 

 

02 0228000 000 000.0
00.~ 00.0- ~0 .0 00.0 000- 00.0 000- 00.0 00.000- 0308.000 80
~00 ~00 00.0 0~0 00 .0- 0~.0 00.0 00.0- ~0.000- 083
00.0 00.0 00.0- 00 .0 000 00.0 00.0 0~0 00.0000- 0000003 080
0000000
00.00 00.0 0 00.00- 00.0- 00 .- 00.00 00.0 0~.~0 .00 .000- 038.000 0008070 000000
00 .00 00.0 ....00.0- ..~0.- ..~0.0 00.0 ....00.0 -0 S007 00008-0080 0800
00.0 00.0 2.80. 00.0- 00.0 ~00 ....00.0 00.0 00.000 0- 00080 000000000 000000
00.~ ~00 00 .0- 00.0- 0~0 00.~ 00 .0 000- 00 .0000- 038-000 8.0802 0.0.0
0000 2 .0- 000 000 00.0- 0000 00.0- 00.0 000000- 00080080 00.0-.0
0~.0 00.0 .000- 00.0- 00.~ 00.0 ....00.0 00.0 00 .~000- 8008.000 000 00.000
00.0 00.0 .000- 0~0- .000 00.0 ..000 00.0 0000.000- 0828000 000
00 .~0 000 00.00- 00.0- 00.0~ 00.00 ~00 0 00 00 00 .~000- 25
000 .000 ..000- 00 .0- .80 00.0 ..000 00.0 ..00000- 0.00003 0.0.0800
00.0 ...000 ..000- ~00- ..0~.0 00.0 00.0 00.0 ...00.~00 0- 0000003 0000
00000000me000

woo-mo vac—ESE
.00 00000 00000 file 8.0 0000.0 00mm 0.900 mfl 00.0040

 

.02 3 00000000000 000 w£w00>0000 0000 0008-0-

.0000.0v0.... .000.0V0v000... 00.0V0V000 .. .000V0V0 .0. 0.000 0300 20 0800088 000800 8080000000 200 0..00 07000

N 00008 8 0.50000 00000500000 we. 05 000 0000800600080 0000 .00 000003000me .0080 000800 00.0 N 050,—- 00m 0000000008 000 1030000000000
0090000000000 088 0000 00000020800 0000000 0080008 00000 0000000 0000 .000-mo 08000-00000w 0003 0000.00 0003000808 00.58 0000.0 0000.00 00000000005
w00 00:- .08080080200 200.0 00000 00080000000» 00.0 0800000088 00000000 .00 000080000 0000000000000 8008000008 00000-5000 0 0000002 .0 030.0-

 

39
5). Thus, the inclusion of 05013,.- inrproved the description of the data for traits manifested

early in the life cycle.
Signfi cance tests of specific causal genetic compongts

In model 2, four traits in the greenhouse, cotyledon diameter at three stages
(emergence, fall and spring), and spring leaf length, displayed significant 02-0, 02A,.- , and
negative 0AM". (Table 4). Maternal environmental variance, 025,,- , was significantly
positive for greenhouse and field seed weight, significantly negative for greenhouse
cotyledon diameter at emergence, and not significantly different from zero for all other
traits. A negative 025,.- is outside the range of possible values.

In model 3, the significance of causal components for two greenhouse traits, seed
weight and embryo weight, changed substantially from model 2 as expected from the
change in likelihood of the estimation model (Table 5). Seed weight displayed significant
02A,“, 01:,an , and negative CAM-n. These components and 020,-, were also significant for
embryo weight. Although the likelihood for all other traits did not improve in model 3, the
significance of causal components changed slightly. Maternal additive genetic variance
(020,.) was no longer significant for cotyledon diameter in fall or spring. Maternal
environmental variance (620,-) had a significantly negative value for spring leaf length, but

was not significantly different fiom zero for all other traits.

Relative contribution of components to total phenoggpic vafl' ce

The relative contribution of the variance components to the total phenotypic
variance differed among the models (Figure 5). The inclusion of maternal components in

model 2 decreased the contribution of 02A, to the phenotypic variance in all field traits and

40

Figure 5. The relative contribution of each variance component to the total phenotypic
variance for models 1, 2, and 3 for greenhouse (A) and field (B) environments. Only 5
components, the direct additive (02M), maternal additive (02M) , direct environmental
(0250), maternal environmental (0250), and the direct-maternal additive covariance (cm)
are included for models 2 and 3 because the direct-maternal environmental covariance

(swam) does not contribute to the total phenotypic variance (see Table 1) The model best
describing a trait is indicated by an arrow.

41

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m

.0. mmmm00m

D 0

0 0 \Im

l§ 0E

.0. l' .0. .0. _**m
fiw.‘ _..0 ..
\\H\ . _
\_\\ E

 

 

 

 

 

 

 

 

 

 

 

 

 

 

rim

 

A

 

 

qqqqqqqqqqqq

 

 

— d

1
fi
-«
0

 

 

 

 

 

 

 

 

 

414 A—d

dfidq «nu—uqq
1 2 5. 1 5.
.1 0

 

NUZ<2<> UEEOme-E 0.0 ZOE-MOn—OME

«‘«q

 

 

yam—>52 m<ma~

0.5-0qu “Fa:

ME§<0Q HOD

00.0.0qu .035.—

.mmmuZDZ m<mq

MPH-Ema HOD

amp-0.200 HOD

E<D MUZmCMm—Em

Emu—m3 O>Mm2m

Emu—m3 Qmmm

ROSETTE

SEED-D SDLG—DFALL -—‘DSPRING

ROSETTE

Figure 5A.

42

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m
A
1 m o
L U M % A 0m A H
E E E
m m \\ I m
000002000 0.00
\\ - 000-003 00200000020
IV 0000000300000
.1. :1: 1...: :1 3034.:
5. 1 5. 5. r.- 2 5. l 5. 0 5. 1.. 2 5. 1 5. 0 5. 1.1
1 0 0 1|. 0 O 1 O 0.

NUZ<E<> UEEOZMIQ m0 ZOE-1000000.?”

SEED —D SDLG

Figure 5B.

43
in greenhouse embryo weight and seed weight. In contrast, the contribution of 02A,,

appeared to increase for cotyledon diameter at emergence, in fall , and in spring, and
spring leaf length relative to model 1. The addition of 0505-,- in model 3 produced little
change in the relative contribution of components to the total phenotypic variance
between models 2 and 3 for six traits in the greenhouse (emergence date, cotyledon
diameter at three stages, fall number of leaves, fall leaf length) and two traits in the field
(cotyledon diameter at emergence and in the fall). However, this additional component did
change the relative contribution of components for seed weight in both greenhouse and
field, embryo weight, spring leaf length, and spring number of leaves. For these traits three
components, 02A,, 02A,“, and mom, increased in their absolute magnitude and in their
contribution to the total phenotypic variance (Tables 4 and 5, Figure 5).
Direct, matemaL and realized heritabilities

Changes in the absolute magnitude and relative contribution of 62.0, 02,0“, and
cam to the total phenotypic variance among models affected direct, maternal, and
realized heritabilities. When maternal effects biased the estimation of 0'on (model 1), a
number of traits displayed substantial heritabilities (Figure 6, Table 6). Traits best
described by model 2 of maternal inheritance (cotyledon diameter at three stages, spring
leaf length, field seed weight, and field emergence week) had significant direct and
maternal heritabilities of similar magnitude (Figure 5, Table 6). In contrast, for two traits
best described by model 3 (seed weight and embryo weight in the greenhouse), significant

maternal heritabilities appeared substantially larger than direct heritabilities. This increase

44

Figure 6. The direct, maternal, and realized heritabilities for each trait in all models in
greenhouse (A) and field (B) environments. The model best describing a trait is indicated
by an arrow.

45

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.m .w
m m m M R
I
I; &
I E
*
.1 I'M
10'-
L:—
J Ivar
.1 l7. 0
*
in a.
.1 h
.i 0
1. 202-2002. 12... 2
18.647002418642024
30-00020-0000000 00.0-00002020000

MODEL 3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mam—>52 090m:

3.0-0th0 0100mm:—

MmE-g—D HOD

£02m.— "3&3

Mme/52 m<mt0

mag—Q HOD

Mag—D HOD

E000 WUZm—mvmmmzm

Emu—m3 O>mm=2m

E053 Qmmm

SEED—D SDLG-D FALL —D SPRING

ROSETTE '

ROSETTE

Figure 6A.

46

 

MODEL 1

 

.0
oo

 

9
O\
t

HERITABILITY
o 9
b) £8
4 111 11L1111111 .11 1..
h. %

 

 

 

 

 

 

 

9.0 .
Hume
H J

 

 

1 1 L1

 

9.0
o~oo

 

 

.9
A

 

 

 

 

 

 

HERITABILITY

vb
IE

 

CD .
l

 

 

C)
N
111;1111J 111 1

 

 

 

 

 

l I
9.0
$3 b)

 

 

fl

MODEL 3

 

9.0
000

HERITABILITY
.0 .0
C) b) #8
1 IIJIIIAIJIJI 111 111 111

 

 

 

 

 

 

 

 

 

 

-O.2
-O.4
:4 a:
E g E
g a E
Q _.
a 5 E
00 E5 E3
LL]
SEED—DSDLG

Figure 6B.

-- (kn: C 17:».5

47

 

 

5.0:...5 80

 

 

 

...... ...... ...... ...... ...... ...... ...... ...... .2...
...... ...... ...... ...... ...... ...... ...... 2... .3... .83 aawsam
2.. - ......- R... ... ... .....- 8... ...... S... ...m... ...»..3 8%
5m...
..N..- ......- 8... ...... ......- 3... ...... 8... .8... .38.... 5.8.... 3:5
.8..- 3...- .8... .2... .2...- ..... .2... ...-... ... ... ems-:8. .....am
.3..- 8... t... .3... ......- ....... .3... ...... .... ... emu-.... 8......ch mafia
...... - .....- 2.... ...... m... . - .....- ...... 8... 8... .38.. ... 3.8.2 ......
...... ...... ...... ...... ...... ...... ...... ...... ...... swan-:3. ......
.8..- ...... ..N... ...-N... .3..- ...... .2... ...-N... .... ... 3.08.... .8 =5.
.8... 8...- .8... ...-.... ...-....- ....... an... ...N... .8... 3.08.... ...u
9.....- .....- ...... N. ... S..- 8... ... ... ...... .... ... 8.5 88mg...
...: ..- .....- .3... ...... ......- N. ... : ... ..N... .5... ...th ......am.
..N..- a. ...- .E... ...... mm... N. ... ...... 8... .8... ...th ...-3
w mmsogmmmc
.22. L... ...... .... 52. .... ...... om. .... ......

m ......2 N .282 . .25.).

 

A mam.— ..Eamssa ...... 32m. 8.882
owouow ....“ 32.8...— ufi wfiaomvfi .GA 33:. 3.308.... 2... 3 wéafifiaoo 25.89.50 3:553 05.... oouuofimm 05 no .62... mm
350%5 20.5... 35.38:... con... 8.. mucus—2...... 0:25» Gaga-82% ...... 83—535.. comm—3.. ...... ....Eo...8 .825 d 03:.

48
in maternal heritabilites was also observed for greenhouse spring leaf length and field seed

weight in model 3.

For maternally inherited traits, realized response to selection is a function of 02,-”,
62M, and 6A.-Am (equation 1). Despite substantial and significant 6on and 02A,.- for a
number of traits in both models of maternal inheritance, realized heritabilites were near
zero or negative (Table 6, Figure 6) because 0AM,“ tended to be negative (Tables 4 and 5).
Negative covariances resulted in negative genetic correlations for most traits (Table 6)
indicating that only alleles that differed in their effects on individual phenotype and
maternal performance were maintained. The prediction from the realized heritabilities is
that phenotypic selection on a single trait would produce no response. There was an
interesting difference in predicted selection response between models 2 and 3 for four
greenhouse traits (seed weight, embryo weight, spring leaf length, and spring leaf number,
Tables 4 and 5). Because of the changes in the magnitude of the additive components
(0on- ozAm, and CAM-...) between models, the predicted response to selection is in the same
direction as selection in model 2 and in the opposite direction to selection in model 3. For
two of these traits, seed weight and embryo weight, mode13 best described the data
(Table 5). As a result, seed weight and embryo weight would be expected to show
reversed responses to selection in the first generation of selection.
Matemg effects at difi‘erent stages in ontogeny

For the three size traits quantified in the greenhouse at multiple stages in the life
cycle, maternal heritabilites did not decrease through ontogeny (Figure 7). For cotyledon
diameter, best described by model 2, significant maternal and direct heritabilities were

similar in magnitude from emergence to spring. Leaflength, best described by model 2,

49

Figure 7. The direct (square), maternal (circle), and realized (triangle) heritabilities for
three greenhouse traits measured repeatedly through ontogeny: cotyledon diameter at 3

stages, and leaf length and number of leaves both measured in the late fall and in the
spring prior to flowering.

0.7
0.6
0.5
0.4
0.3
0.2
0.1

HERITABILITY

-0.l

Figure 7.

MODELZ

50

 

 

K

 

 

311/

*

 

 

 

 

1J11

H

==*"

 

 

 

 

 

INITIAL

FALL

SPRING

C OTYLEDON DIAMETER

FALL

 

SPRING
LEAF LENGTH

 

 

 

 

 

 

 

 

 

FALL

SPRING

LEAF NUMBER

 

 

 

 

 

 

 

 

 

 

 

 

MODEL 3

0.5 . * -

.4
0.4 3 \\ *
0.3-:—.*"%
0.23
0.1

2 ...—A
-0.I

INITIAL FALL SPRING

COTYLEDON DIAMETER

t

0.7

 

 

 

 

 

 

 

 

 

FALL SPRING
LEAF LENGTH
0.3 _
0.2 : //
0.1 . /
j I——--—‘"""' '
o
-o.1 ‘
FALL SPRING
LEAF NUMBER

5 l
displayed no heritable variation in the fall and significant direct and maternal heritabilities

in the spring demonstrating an increase in maternal inheritance through ontogeny. Number
of leaves displayed no heritable variation under any model. Again, the realized heritabilites
remained at low values for all three traits at different stages because of the negative 0AM..-
at all stages.

W_rth_m' ' Generation Genetic Correlations

In contrast to intergenerational genetic correlations (Table 6), within generation
genetic correlations calculated from a strictly Mendelian model (1) were positive among a
number of size related traits (Table 7). Seed weight and embryo weight were both
positively correlated with cotyledon diameter at all three stages in the greenhouse, and
seed weight and cotyledon diameter at emergence were also positively correlated in the
field. Other traits in the greenhouse showed the following pattem. Cotyledon diameter at
emergence was positively correlated with cotyledon diameter at the two subsequent stages
with a value close to one. Embryo weight and cotyledon diameter at emergence were also
positively correlated with spring leaf length. Emergence date was positively correlated
with seed weight, embryo weight, and spring cotyledon diameter and negatively correlated
with fall number of leaves, the only significant, negative correlation.

When maternal effects were included in the estimation of genetic correlations
(Figure 4b), 31 of 34 estimation models that converged showed improvement in log
likelihood. Direct additive genetic correlations were smaller in magnitude and difl‘ered in
significance fi'om those estimated in a strictly Mendelian model (Table 7 ). Seed weight

was positively correlated with cotyledon diameter at emergence in the field and with fall

 

 

 

02 oz 02 .005 508.... .8 ....

........ N...- ....0... .00. 5.20.... 8.5.08

........ ...-... ... ... .00. .33 880.00....

........ ....N... m... .30. 0.0.03 00.0

0......

N..... 8...- S... 3.... m.....- 0.... ...-...- .....- .....- ...-5.5.0.0203. 00000

0N... +0.... .....- ... ... +3... .....0... .N... ...m... ..N... ...-.0. 0.08.0.3 00.00

......- 0z .0...- ......- +0.... ......... ...... ....00... ......0... 800.035.980.000
-----... 0...... .....o- ...... MN..- .....- :00...- ......- ......- .z-.....§.a=z..o-. ......
+N..... NN...- ....... ...... ....o... N...- o... 0.... ...-......080080 ......
02 +00... 02 ...-...- 0.... ....0... NN... ......N... .....N... 800.50.80.98 ......

...... +0.... 02 ...-.... .8... 02 m... .......... ......... 80.50800 80208

........ 0.... ...N... m. ... ...... ...... ...... .00. 2.... 880.08...
0.....- 0.... +0.... a..- ...... +0.... oz N... 02 30.50.03 ......am
2...... 8.... ...03 .N...- 0.... N... 02 ...... 02 30.50.03 080
00000200....

2.0 ...0 000 20. .0. ...... .... ..m. 30 30 ......

 

089...... ......0v..v........

.....0V0v3... .. .mo.0v..v. .0 . ”030:0. 00 0. 000000.0w.0 00.30050 .0000w.0>000 0.0.00 0. 0000.800 00000...0> .0.00.000..>00
30.0.00. 0.05.3 00500500 .000... 0.0000. 000.0.» 0.0.. 00.0.. 05 030:0. + 0 690000 .00 0.0 00005000 05000w 05.000

.00.... 05 ..0 0.00. 000005.90 006$ .02 .... 00.0000 0.0 090000 .00 0.0 00.03 0.000... 005.053 ..00.00_0>0 .00 003 000....0 .....—
0.5 00.00.00. -..-..-. 0.00...0 30.0.00. .... 00000.50. 0.03 000.. 50.. .0 000 0003 ...00 00.00.03 003 n .0002 .0500. 0.00.50 0.00.00.
5.3 n .000... 000 A320... .000... 00200002 .5050 0 ... .0000. ”0.0000. 00500500 03. .00.. 00050—0000 05000» 00.00500 . 0.03000
0.0.0.3... 0000 00.. .0.000.00..>00 0.0.. 000 00000000.& .0.. 0...... ..0 00.. ..000 0003.00 00050—0000 05000» 05000 80.5 ... 030...

53
and spring cotyledon diameter in the greenhouse. Emergence date was also positively

correlated with spring cotyledon diameter in the greenhouse. Fall number of leaves and
spring cotyledon diameter were negatively correlated in the greenhouse. A number of trait
pairs that showed genetic correlations close to a value of one in the simpler Mendelian
model did not converge under maternal inheritance (Table 7 ). Maternal additive genetic
correlations were not significant for any estimation model in which they were included.
Only one trait pair displayed a large positive value (rm1m=0.73), spring leaf length and
fall cotyledon diameter, however significance tests for this component did not converge.
Therefore, maternal performance appeared to be genetically uncorrelated in its efl‘ects on
traits in the subsequent generation.
DISCUSSION

The most significant result in this study is the effect of maternal inheritance on
predicted response to selection. Negative genetic correlations between the direct additive
and maternal additive effects (erAm) result in realized heritabilities near zero for traits
expressed at all stages in the life cycle. These negative correlations are so large early in life
that traits in the seed stage exhibit negative realized heritabilities. For seed weight and
embryo weight, the predicted selection response is in the opposite direction to selection.
Thus, the structure of maternal inheritance in C. vema is such that trans-generational
efl‘ects of a mother on her young dramatically constrain the evolutionary response of traits
expressed both early and late in the life cycle. It is also interesting that maternal
inheritance persists throughout the life cycle in this annual plant. Below I summarize the

pattern of maternal inheritance and its consequence for adaptive evolution.

54
Maternal inheritance

The causal components contributing to maternal inheritance and their magnitude
change over the course of development (Figure 1). Four components contribute to
phenotypic variation in the seed traits in the greenhouse, seed weight and embryo weight:
62.0, 62A,“, CAM, and 050E.“ (Table 5). Maternal additive efl‘ects are 2-3 times as large as
direct additive effects. Maternal environmental effects are small, presumably as a result of
relatively uniform environmental conditions in the greenhouse. The positive covariance in
environmental effects results from the temporal overlap of environmental conditions in the
mother and her young at this stage. For all traits expressed beyond the seed stage, the
covariance in environmental effects does not appear to contribute to the resemblance
between mothers and offspring, most likely because both parents and offspring were
randomized across environmental conditions. The magnitude of these four components is
similar for seed weight in the field (Table 5), however, the trait is best described by model
2 (Table 4) in which only the maternal environmental component is significant.

In the seedling stage, three components contribute to the phenotypic value for
cotyledon diameter at emergence in the greenhouse: 02.0, 62.0, negative om (Table 4).
Direct and maternal additive effects are more similar in magnitude when compared to seed
traits. In contrast in the field, cotyledon diameter is best described by the strictly
Mendelian model (Table 3). Emergence time is best described by Mendelian inheritance in
the greenhouse and by a marginally significant maternal inheritance model (2) in the field
(Table 4). In the latter model , however, none of the components are significant.

The pattern of maternal inheritance for cotyledon diameter in the greenhouse

remains the same throughout subsequent stages in the life cycle (Table 4, Figure 6) with

55
three additive genetic components contributing to an individual’s phenotypic value. The

number of leaves and leaf length show no additive genetic variation when traits are
expressed in fall rosettes in the greenhouse (Table 3), however, prior to flowering in the
spring, leaf length displays the same pattern of maternal inheritance as cotyledon diameter
at all three stages (Table 4). In contrast, the number of leaves displays simple Mendelian
inheritance in the spring (Table 3).

This hierarchical analysis clearly shows that the magnitude and structure of
maternal inheritance changes throughout the life cycle. Early in life, both genetic and
environmental components of maternal performance contribute to the offspring phenotype.
At emergence, however, maternal genetic effects predominate. These maternal genetic
effects persist throughout the life cycle for cotyledon diameter, while other size related
traits show more variation in the model of inheritance. lF or example, leaf length shows no
heritable variation in the fall. In contrast, leaf length in the spring is influenced by maternal
genetic effects.

The structure of maternal inheritance appears to differ between the field and
greenhouse environments. Seed weight is maternally inherited in both environments, but
the best model differs between environments. Emergence week is best described by
Mendelian inheritance in the greenhouse and maternal inheritance (model 2) in the field. In
the field cotyledon diameter at emergence does not display maternal inheritance, however
it does at all three stages in the greenhouse. It is not unusual to obtain different estimates

of causal components when ofl‘spring are reared in difi‘erent environments (e. g. Mazer and

56
Schick 1991; Schmitt et al 1992; Schmid and Dolt 1994;P1atenkamp and Shaw 1993;

Montalvo and Shaw 1994). In this study, it is difficult to compare the structure of
maternal inheritance between the field and greenhouse environments because the census
interval differed between the two environments, emergence time and size at emergence
were measured weekly in the field and twice per week in the greenhouse. As a result, I
expected and observed larger variances associated with these two field traits (Table 2).
More importantly, estimates from the field are compromised by the possibility of genotype
by environment interaction. An analysis of paternal half-sib means in the two environments
showed little evidence for genotype by environment interactions in the final generation
(Thiede, unpublished data). However, other studies have documented that maternal
genotypic effects can depend on the environment in which the ofl‘spring are raised
(Schmitt et al. 1992; Schmid and Dolt 1994). This type of maternal genotype by offspring
environment interaction would compromise this quantitative genetic analysis of maternal
inheritance. Therefore, the field estimates of maternal inheritance should be viewed with
caution.
Maternal performance

What phenotypic traits are likely to contrrbute to the composite maternal
performance phenotype (PM)? Maternal size (Platenkamp and Shaw 1993), maternal
nutritional status (Parrish and Bazaaz 1985; Miao et aL 1991), maternal phenology (Lacey
1991), and maternal source-sink relations (Rocha and Stephenson 1990) (see review in
Roach and Wulfl‘ 1987) could all contribute to maternal performance. The position of
seeds in a fi'uit during development determines source-sink relationships that can affect a

number of seed traits, especially seed size (e. g. Rocha and Stephenson 1990). When

57
position effects were included as a fixed efl‘ect in the analysis of seed weight, they

significantly improved the likelihood of the estimation model indicating that position
effects account for a significant amotmt of the observed phenotypic variation. The within
maternal family variation in seed weight accounted for by the fixed efi‘ect may be
determined by the architecture of the mother. Ifthe architectural traits that determine
position efl‘ects are genetically based, they may allow the variance in seed weight to evolve
as well as the mean (Bull 1987; Carriere 1994). Biere (1991a) suggested similar reasoning
for selection on the variance in emergence time in Lychnisflos-cuculi. Response of the
phenotypic variance to selection may result not only from non-linear components of the
selection gradient (Brodie et al. 1995), but also fiom higher levels of selection such as
maternal selection (Thiede, 1996).
Estimation of maternal effects

In natural plant populations, the magnitude of maternal genetic efl‘ects have been
estimated by three different approaches. First, the “bio-model” from a diallel design
(Cockerham and Weir 1977) permits the estimation of maternal and paternal extranuclear
effects (e.g Antonovics and Schmitt 1986; Mazer 1987; Biere 1991a; Kelly 1992;
Platenkamp and Shaw 1993; Montalvo and Shaw 1994). The estimate of maternal
extranuclear efl‘ects in the above studies contains a number of specific maternal genetic
and environmental causal components, but does not require assumptions about an
underlying model of maternal inheritance. The second approach is a nested breeding
design in which maternal effects are confounded by dominance, therefore, limiting
conclusions about the magnitude of these effects (Mitchell-Olds 1986; Mitchell-Olds and

Bergelson 1990a; Schwaegerle and Levin 1991). The final approach uses clonal replicates

58
to experimentally separate maternal genetic, maternal environment, and their interaction as

sources of phenotypic variation in offspring traits (Biere 1991a; Schmitt et a1 1992;
Platenkamp and Shaw 1993; Schmid and Dolt 1994). Like the diallel, this approach does
not provide estimates of specific causal components related to maternal inheritance, but
does allow one to compare the magnitude of genetic vs. environmental effects in artificial
environments as well as explore the possibility of genotype by environment interactions.
The multi-generation approach that I present here is novel in its detailed
partitioning of the phenotypic variance into specific causal components allowing more
ermlicit predictions about evolutionary responses to selection (see below). The general
pattern of maternal effects documented in this study is consistent with previous findings.
Seed weight and emergence date exhibit low direct heritabilites and substantial maternal
effects (Biere 1991a; Platenkamp and Shaw 1993; Montalvo and Shaw 1994; Schmid and
Dolt 1994). Subsequent size related traits exhibit moderate direct heritabilities and
maternal genetic effects in some studies (Biere 1991a; Schmid and Dolt 1994), but not in
others (Montalvo and Shaw 1994). In other studies maternal genetic effects generally
decline through the life cycle (Biere 1991a; Schmid and Dolt 1994; Montalvo and Shaw
1994). In contrast in this study, maternal genetic eflects continue to contribute
significantly to phenotypic variation all the way through the life cycle for two of three
traits (Figure 7; for another exception see Schmid and Dolt 1994). The larger magnitude
of maternal genetic efi‘ects relative to smaller maternal environmental effects in this study
is consistent with other studies (Biere 1991a; Schmid and Dolt 1994; Platenkamp and
Shaw 1993 ). However, there is ample evidence that maternal genotype by environment

interactions may eliminate the direct maternal genetic effect when maternal genotypes are

59
replicated across contrasting environments (Schmitt et a1 1992; Platenkamp and Shaw

1993; Schmid and Dolt 1994). These genotype by environment interactions for maternal
efl‘ects should not obscure maternal genetic effects in this study because all mothers were
raised under relatively uniform greenhouse conditions. However, the impact of maternal
genotype by environment interactions on the evolution of maternally inherited traits awaits
the development of theoretical models that incorporate these higher order interactions in
the response to selection.

While advantageous for a mechanistic understanding of the evolutionary process,
this biometrical approach for estimating maternal effects by partitioning the phenotypic
covariances among numerous relatives has limitations (Eisen 1967; Foulley and Lefort
1978; Wilham 1980). The primary limitation is the confounding of direct Mendelian
inheritance and maternal efiects in maternal lineages that results in large sampling
correlations among causal components. In designs such as the one used here, sampling
correlations can cause substantial bias in estimation of variance components when not all
components are estimable. Experimental approaches that decouple direct and maternal
transmission provide an alternative approach. Cross-fostering offspring after birth provides
estimates of post-natal maternal effects by separating the maternal effect from the direct
effect by using nurse mothers (Riska et aL 1985). Embryo transplantation is another
approach that provides estimates of both pre-natal and post-natal maternal effects by
decoupling direct and maternal effects (Cowley 1991). Experimental manipulation of
maternal attributes such as maternal provisioning can also be utilized to estimate the
magnitude of the maternal phenotypic effect separately fiom genetic contributions

(Sinervo 1991). In the absence of similar experimental approaches for detangling maternal

60
and direct efl‘ects in plants, the best solution may be to include numerous types of relatives

in biometrical analyses. For example, Cantet et al. (1988) were able to estimate all nine
variance components in a maternal effects model by utilizing 17 types of relatives.
Alternatively, utilizing relatives like second cousins in which direct and maternal effects
are less confounded may provide a better approach (Wilham 1980).

A second limitation of this approach is the potential bias that may result from not
estimating additional components that may be influencing phenotypic covariances: l)
dominance components, 2) cytoplasmic inheritance, and 3) the maternal effect coeflicient.
Meyer (1992) indicates that the magnitude of the excluded efl‘ect must be quite large (i.e.
30%) to affect the estimates of variance components. To what extent might dominance
variance bias variance component estimates in this study? When I included direct
dominance in a five component estimation model (02A,, 0250, 62130, 62”., cm), the
estimates of the additive components did not change. Furthermore, for seven out often
traits in the greenhouse, 0290 was negative, indicating a vahre not different from zero.
Montalvo and Shaw (1994) also detected no significant dominance variance in similar
traits. Therefore, in this study direct dominance variance is unlikely to change the
estimates of the direct and maternal additive genetic variance and their covariance.
Maternal dominance and maternal environmental variances are perfectly correlated in this
design. Ifone views the estimates of maternal environmental variance as the sum of these
two components (suggested by Thompson 1976), it is clear that maternal dominance is
also not significantly influencing phenotypic covariances because the maternal
environmental variance was not different from zero in most cases (Tables 4 and 5).

Therefore, the estimation of variance conrponents in a reduced animal models appears

61
robust to the assumptions of no direct or maternal dominance variances or their

covariance in this study.

Resemblance among relatives sharing a common maternal lineage can also be
influenced by cytoplasmic inheritance of chloroplast and mitochondrial genomes (reviewed
by Gillham 1994; but see Chin and Sears 1993; Sewell et al 1993 for exceptions). Lynch
and Walsh (1996) suggest how these models could also be extended to include uniparental
cytoplasmic and mitochondrial transmission. In the present study, full-sib, dam-offspring,
and maternal relative-ofl‘spring covariances could include effects due to cytoplasmic
inheritance which would inflate estimates of maternal additive, maternal environmental
variances, and direct-maternal environmental covariance. Similarly, not estimating the
maternal efl‘ect coefficient also has the potential to inflate specific variance components
(see Cantet et al. 1988).

Evolutionag consequences of maternal inheritance

Previous studies of maternal effects have often suggested that response to
selection on juvenile traits such as seed mass or emergence time will be slower (i.e.
Antonovics and Schmitt 1986, Roach and Wulfl‘ 1987; Biere 1991a) because maternal
genetic efl"ects mask the small amount of zygotic genetic variation. Several authors have
suggested that selection may act solely on the maternal genetic variation for juvenile traits
lacking direct additive genetic variation (Biere 1991a; Platenkamp and Shaw 1993;
Montalvo and Shaw 1994; Schmid and Dolt 1994). It is, of course, possible for selection
to differentiate among ofl‘spring and also among mothers. The resulting response to
multiple levels of selection will depend critically on the genetic variance for both ofl‘spring

phenotype and maternal performance. The strength of the approach presented here is that

62
it allows one to evaluate the response to selection, not only based on direct and maternal

additive genetic variance, but also based on their covariance which all other studies in
natural populations have not estimated. Accurate predictions about evolutionary responses
to selection hinge on this detailed partitioning. This study clearly demonstrates that these
direct-maternal genetic covariances will constrain selection response (Figure 6, Table 6).

Intergenerational covariances

Direct-matemal additive genetic covariances between maternal performance and
ofl‘spring phenotype are consistently negative for 7 of 8 traits displaying maternal
inheritance (Tables 4 and 5). Furthermore, the magnitude of this direct-maternal
covariance is large enough to result in a predicted reversed response to selection for two
traits, seed weight and embryo weight. For all other traits in both models, the negative
direct-maternal covariance reduces the predicted response to selection to near zero
(Figure 6). Thus, despite substantial direct and maternal additive effects, the evolutionary
potential of these traits is limited by the underlying direct-maternal genetic covariances.

Since Dickerson’s (1947) seminal paper documenting the evolutionary
consequences of maternal effects in domestic hogs, a number of animal breeders and
evolutionary biologists have demonstrated negative direct-maternal additive genetic
covariances (Figure 2). Others utilizing Falconer’s (1965) simplified approach have
demonstrated negative maternal efl‘ect coeflicients. Negative m’s have been found for litter
size in mice (Falconer 1955,1965), age to maturity in springtails (Janssen at al. 1988) and
clutch size and condition in flycatchers (Schluter and Gustafl‘son 1993 ). In some cases, the
magnitude of these direct-maternal covariances or maternal efl‘ects coeflicients are large

enough to produce reversed responses to selection in the short-term In theory, long-term

63
responses to consistent selection should asymptotically approach the expected rate in the

absence of maternal effects (Kirkpatrick and Lande 1989). In nature, however, spatial and
terrrporal variation in selection (e. g. Kalisz 1986; Kelly 1992; Stratton 1992a) in
conjunction with maternal inheritance can be expected to produce complex evolutionary
dynamics.

Trade-offs between life history traits have been central in the theory of life-history
evolution (e.g. Williams 1957; Lande 1982). In his review of life-history tradeofl‘s, Stearns
(1992) points out that most of the theoretical and empirical literature on life history have
dealt with tradeofl‘s within an individual such as allocation to current vs. future
reproduction or current reproduction vs. sub sequent survival. However, tradeofl‘s between
generations have received less attention. This analysis of maternal effects in C. vema
suggests that there is a fundamental, genetically based intergenerational trade-ofl‘ between
maternal performance and offspring phenotype for 7 of 14 traits examined (Table 6).
Perhaps the simplest explanation for the existence of antagonistic pleiotropy is that
directional selection on maternal performance and/or offspring phenotype has led to the
maintenance of alleles that difl‘er in their effect on the phenotype (Falconer 1981). In
theory, mutation could supply suflicient variation to prevent the fixation of these differing
alleles via selection (Charlesworth 1990), so the explanation for the existence of these
negative direct-maternal genetic correlations may require a more complicated model of
functional genetic architecture involving pleiotropic efl‘ects on allocation and acquisition
(Houle 1991). Whatever the mechanistic explanations for these negative genetic

correlations, the consequence is that joint evolution of maternal performance and oflspring

64
phenotype will be constrained for a number of traits at difl‘erent stages in the life cycle of

C. verna.

Within generation covariances

In contrast to intergenerational covariances described above, most of the
significant additive genetic covariances between traits within a generation are positive.
Under Mendelian inheritance in model 4, these positive additive genetic correlations show
substantial pleiotropic effects for traits related to size early in the life cycle. Seed weight is
genetically correlated with cotyledon diameter at emergence, but the magnitude of this
correlation declines in subsequent measures of this trait (Table 7). Cotyledon diameter is
correlated across the three censuses. The only significant negative correlation is between
emergence date and fall leaf number. Therefore, in the absence of maternal effects, these
estimates of within generation genetic correlations indicate substantial positive pleiotropy
among size related traits.

When maternal inheritance is included in the estimation of these genetic
correlations, however, the magnitude and significance of direct genetic correlations
changes substantially (Table 7). Most correlations remain positive, but many are no longer
significant. The inclusion of maternal inheritance in the estimation model reveals decreased
pleiotropy. It is common to observe positive correlations among size traits in plants (e. g.
Montalvo and Shaw 1994). In general morphological traits tend to show positive genetic
correlations (Rofl‘ 1996), however, many of these estimates may be inflated by maternal
effects. While morphological traits show some pleiotropy, there is no evidence for

significant genetic correlations among the unobserved maternal performance traits.

65
MM' ma tg golution

Equations for predicting multivariate evolution require estimates of the additive
genetic variance-covariance matrix (G) for all traits as well as estimates of the selection
gradient (Lande 1982; Lande and Arnold 1983 ). However, it is not clear how univariate
estimates of direct and maternal additive components and bivariate estimates of genetic
correlations between traits such as those estimated in this study translate into a
rrrultivariate G. Currently, evolutionary biologists are technically constrained fiom
obtaining these multivariate estimates with Dickerson’s genetic model for estimating
maternal efl‘ects. An alternative approach for considering the evolutionary consequences
of maternal effects in a multivariate framework describes the structure of maternal
inheritance by a single term, the mother-daughter covariance (Kirkpatrick and Lande
1989, 1992; Lande and Kirkpatrick 1990; Riska 1991). In a subsequent manuscript I
explore the multivariate evolutionary dynamics of maternal inheritance using this simplified
covariance approach.

Conclusions

This quantitative genetic analysis demonstrates that maternal inheritance will
influence the evolutionary dynamics for a number of traits in this natural plant population.
Traits reflecting individual size at the seed, seedling, and adult stages in the life cycle were
significantly influenced both by direct and maternal additive genetic variances and their
covariance. The persistence of maternal inheritance to later stages in the life cycle is
unusual in plants. Perhaps the most significant contribution of this study is the negative
estimates of direct-maternal additive genetic covariances, the first demonstration of this

evolutionary constraint in a natural plant population. In conjunction with direct and

66
maternal additive genetic variances, this direct-maternal additive covariance clearly results

in predicted reversed response to selection for two traits, seed weight and embryo weight,
and minimal responses to selection in traits later in the life cycle. In contrast, within
generation genetic covariances among size traits are likely to enhance selection response
such that direct selection for increased seed or seedling size will result size increases in
prior or subsequent traits. The incorporation of within and between generation
covariances in a multivariate framework for predicting response to selection remains a
challenge. While most authors have suggested that maternal effects may slow the
evolutionary response by masking the zygotic genotype, this study illustrates that maternal
effects have the potential to enhance or constrain the selection response depending on the
sign and magnitude of the direct-maternal additive genetic covariance. In the study
population, the joint evolution of maternal performance and individual phenotype is
constrained for all traits displaying significant maternal effects suggesting an underlying

fundamental trade—off between mothers and their offspring.

Chapter 2
AN EPISODIC ANALYSIS OF PHENOTYPIC SELECTION
ON JUVENILE TRAITS IN COLLINSIA VERNA:
A COMPARISON OF QUANTITATIVE TRAITS DISPLAYING MENDELIAN
AND NON-MENDELIAN INHERTTANCE.
INTRODUCTION
Studies of evolution in natural populations consider two phases in the evolutionary

process: phenotypic selection and Mendelian inheritance. These separate estimates of
within generation selection ([3 and y) and between generation response to selection based
on inheritance (G) can be combined in the standard multivariate equation of evolution to

predict the change in the trait mean, A z-=GB, or the trait variance or covariance,

AG=G(7-,BflT)G (afier Phillips and Arnold 1989; Lande and Arnold 1983). However,
when an individual’s phenotypic value is a function not only of its genotypic value in the
environment, but is also influenced by its mother’s phenotypic vahre then evohrtionary
responses will difl‘er from expectation based on the standard equations.

Kirkpatrick and Lande (1989, 1992) have demonstrated that when traits display
maternal or non-Mendelian inheritance, the evohrtionary change in a trait mean is a
fimction not only of current phenotypic selection and Mendelian and non-Mendelian
inheritance, but also is a function of phenotypic selection in previous generations. Thus,

maternal inheritance introduces time lags in the evolutionary process. These time lags

67

68
influence the rate of evolutionary response such that the maximal rate is approached

asymptotically under a constant selection. Furthermore, the response to selection
continues after selection ceases and its direction can vary depending on the sigr and
magnitude of the maternal efl‘ect coeflicient. Kirkpatrick and Lande (1989) call this
evolutionary momentum. In addition to time lags, maternal inheritance can afl‘ect the
direction of response depending on the sign and magnitude of the direct-maternal additive
genetic covariance (Wilham 1963 ). Thus, predicting the direction and magnitude of
evolutionary responses for maternally inherited traits is complicated.

Animal breeders have demonstrated how maternal inheritance can alter predicted
responses to artificial selection (Dickerson 1947; Wilham 1963). Both negative direct-
matemal genetic covariances (Riska et a1. 1985) and negative maternal phenotypic effects
(Falconer 1965) can produce reversed responses to selection. In addition to the influence
of maternal inheritance on artificial selection, quantitative geneticists have demonstrated
that maternal genetic effects on traits like body size decrease through ontogeny
(Cheverud et al. 1983; Atchley 1984). While maternal inheritance may decline through the
life cycle, it can still influence multivariate evolution in natural populations. If selection
acts directly on maternally inherited traits or traits influencing maternal performance, then
genetic correlations between these traits will influence their joint evolution.

Like animal breeders, plant population biologists have documented the persistence
of maternal effects through ontogeny. In nearly all cases, these studies focus on how
maternal environmental conditions influence ofl‘spring phenotype. Maternal

environmental effects have the potential to influence an extensive number of plant traits

69
including seed weight and early size (reviewed by Roach and Wulfi‘ 1987). Matemally

influenced traits like seed weight, emergence time, and early relative growth rate can
determine early size differences and affect irrtraspecific competitive interactions (Gross
1984; Gross and Smith 1991). These differences in seedling size tend to persist through
the life cycle in competitive situations (Fenner 1983; Gross 1984), therefore, 'matemal
efl’ects can be long-lasting in these situations. If juvenile traits influence the outcome of
competitive interactions that generate size and consequently fitness hierarchies in plant
populations (Waller 1985; Stanton 1985; Weiner 1985, 1990), then maternal effects can
directly impact fitness. Thus, maternal environmental efl‘ects influence a number of plant
traits and their effects can persist to late in life. Lacey (1991) has shown that maternal
environmental effects can persist through two generations and influence phenological traits
like flowering time.

The demonstration of maternal genetic effects on plant traits is less common. A
number of studies have demonstrated a significant maternal genetic component to seed
weight (Platenkamp and Shaw 1993; Montalvo and Shaw 1994; Schmid and Dolt 1994;
Biere 1991a; Mitchell-Olds and Bergelson 1990a), germination date (Montalvo and Shaw
1994; Schmid and Dolt 1994; Biere 1991a; Mitchell-Olds and Bergelson 1990a), and
seedling size (Schmid and Dolt 1994; Biere 1991a). These studies demonstrate that
maternal genetic efl‘ects decrease through ontogeny with effects being strongest on seed
weight, and smaller or non-significant on seedling size. In two studies, maternal genetic
effects were larger than maternal environmental efl‘ects (Biere 1991a; Schmid and Dolt

1994). Schmitt et al. (1992) demonstrated that maternal genotypes difl‘er in their response

70
to maternal environmental conditions such that maternal genetic efl‘ects on offspring are

influenced by the maternal genotype by maternal environment interaction. The maternal
genetic efl‘ects estimated in these studies can not be incorporated into evolutionary models
predicting the response of maternally inherited traits because the genetic parameters do
not estimate either the maternal additive variance (except Platenkamp and Shaw 1993) or
the covariance between direct additive and maternal additive values. Thus, straightforward
predictions about the evolutionary role of these maternal genetic effects in natural plant
populations are not possible.

In contrast, the nature of phenotypic selection in natural populations has been well
documented for a number of these maternally influenced traits. Univariate studies of seed
weight and emergence date have documented the efl‘ects of these traits on individual
survival and fecundity (e.g. Kalisz 1986; Winn 1988; Biere 199 lb). In multivariate studies
the direct contribution of traits to components of fitness can be separated from indirect
effects on phenotypically correlated traits. Multivariate studies including a number of
juvenile plant traits have shown that direct selection acts primarily on early size, while seed
weight and emergence date contribute mostly indirectly to fitness components via their
effect on size (Bennington and McGraw 1995b; Stratton 1992a; Mitchell-Olds and
Bergelson 1990b). Thus, traits likely to display maternal inheritance like seed weight,
emergence date, and early seedling size can directly or indirectly influence components of
fitness in a number of species.

My motivation in this study is to quantify the extent to which maternally inherited

traits impact the rate and direction of multivariate evolution by examining the relationship

7 1
between a number of maternally inherited juvenile traits and fitness. I have formd that a

number of juvenile traits in the winter annual, Collinsia vema, display maternal
inheritance. Specifically, three traits, seed weight, cotyledon diameter at emergence, and
cotyledon diameter in late fall displayed both significant additive genetic and maternal
additive genetic variance as well as negative direct by maternal genetic covariance. A
fourth trait, emergence date, displayed only significant additive genetic variance and no
maternal efl"ects (Chapter 1). Understanding and predicting the evohrtionary response of
these maternally inherited traits hinges not only on the nature of phenotypic selection, but
also on the observed maternal inheritance.

Here I quantify the magnitude of direct selection on each of these four traits in
four episodes of selection (Figure 8). The analysis of selection for sequential episodes in
the life cycle is required because early mortality can eliminate individuals before they
express all four of the phenotypic traits. Individuals not expressing all traits can not be
included in a rrnrltiple regression analysis of a single episode spanning the entire life cycle.
Therefore, I partitioned the life-cycle into three episodes of viability selection and one
episode of fecundity selection (Arnold and Wade 1984a, b). This episodic approach allows
me to identify how traits displaying maternal inheritance afl‘ect sequential viability and
fecundity components of fitness as well as estimate the total magnitude of phenotypic

selection on these traits across all episodes (Lynch and Arnold 1988).

72

.8838 we 88¢an 5.8m can Eu: Bu 2 335:8 3%: Sea =< 28850 9.88 05 8 «395:8 use. REE wave—ea
38.: 8:: 0.8880 65 05 E 3328 8 235:8 £83 88828 can 3&8? 80m .38: 95. .385 83.3 825288
293088 05 8% 83 .8: a me 029 290058 05 bee «on 883:5 8 888:5 R852 .mwoobo 85:: 08:88
5.53 385 mecca 82.3288 ofibofinq 05 88% 36: 8253 mBotm 882.0589 88:5 0252 no 3R: =a we
mix—25 868.38 033:8 a 89a av 2.20508 :28“on 3:3 05 mm 8888 08 25 888mg “8:838 05 E 88
an: on: E omfifi 05 8:85 8 08 26% $68 880: £95 05 3 8:58.53 38% 8s Snug—zoom Rem EB 3233
.3 89880 8:: 865$ MO 38888 See 8 flat 0:82... Sow mo $0ch 82% 05 mqtofiou Baum—«.6 5mm .w gamm—

>H—QZDOmm

      
    

Cay—mm OH 4.3/55 4

/ mam .55 .
~5th ..5 820 ch .2238 r

I
//::§: ,
x/M ...... 8

H3053 Qmmm a

HZme—msmthmm ZODOMIH A<>SMDm a

73
In this study I quantify the nature (linear and non-linear), magnitude, and direction

of phenotypic selection on four juvenile traits: seed weight, emergence week, cotyledon
diameter at emergence, and cotyledon diameter prior to winter to address the following
questions: 1) what is the total magnitude of linear and non-linear selection on these four
traits, 2) which episodes are most critical in contributing to the total magnitude of linear
and non-linear selection, therefore, suggesting possible hypotheses for the causal agents of
selection, 3) what is the relative contribution of direct and indirect effects to the response
of particular traits, i.e. do maternally inherited juvenile traits influence survivorship and
fecundity directly or indirectly by influencing subsequent traits that then impact fitness?
MATERIALS AND METHODS

Study Site and Species

C ollinsia verna Nutt. (Scrophulariaceae) is a winter annual that inhabits mesic
forests of the eastern United States (Femald 1970). Autumn diurnal temperature
fluctuations cue germination (Baskin and Baskin 1983; Kalisz 1986) which begins in late
September and continues into late November. Seedlings consist of a pair of cotyledons
that expand in diameter throughout the fall. In southern Michigan, the first pair of leaves
begin to develop in late November or early December, however, most plants overwinter
with only cotyledons. These seedlings persist until early spring under a cover of leaf litter
and snow. Rapid spring growth leads to rosettes with two to many pairs of true leaves. In
May these rosettes initiate flowering which lasts two to three weeks. Fruits mature at the
beginning of June and primary dispersal takes place as the plants senesce. While primary

dispersal is limited in this species that lacks any specialized dispersal morphology (Thiede,

74
unpublished data), secondary dispersal by surface flow of water is likely to influence seed

dispersion because these seeds tend to float.

This study was conducted in a small (is 10 hectare) privately owned woodlot on
TU Avenue in Kalamazoo County, Michigan. The tree canopy of this mature forest
consisted of Prunus serotina, Acer saccharum, and T ilia americana. C. verna and
F loerkea proserpinacoides were the predominant understory herbs in the spring. Other
species in the herbaceous comnrrmity included Phlox divaricata, Laportea canadensis,
Trillium grandiflorum, Arisaema triphyllum.

The biotic and abiotic environment experienced by C. verna at TU Avenue varied
spatially and temporally. C. verna occurred both in the center and along the edge of the
woodlot, reaching highest densities along the edge. Agricultural fields created a sharp
boundary at the edge of the woodlot. I observed moderate to severe wilting in early
germinating seedlings in some locations along the edge, a sign of drought stress in that
location, while wilting was only observed in a few plants in the interior. Therefore, light
levels and soil moisture differed between the edge and center of the woodlot. Two
herbivores, slugs and deer, consumed C. vema at two different times in its life cycle. In
the fall primarily afier leaf drop, slugs would consume both cotyledons and the apical
meristem of seedlings. While the stem and root persisted after slug browsing, the seedling
never recovered. In the spring deer browsed the apical meristem of 15-20% of the rosettes
each year. As a result of deer browsing, axillary nodes were released from apical
dominance and developed branches. Deer browsed seedlings were able to produce flowers

and sometimes produced seeds, but their fecundity was very low when compared to

75
rmbrowsed plants. Therefore, drought and slug and deer herbivory may be potentially

important selective agents in this population. However, the effects of these biotic and
abiotic factors varied spatially in the population and temporally in the timing of their
efl'ects in the life cycle of C. vema.

t' ' Pheno ic Selection
Data Collection

To quantify patterns of phenotypic selection on traits displaying non-Mendelian

inheritance, I monitored survival and reproduction of seedlings at TU Ave fi'om 1992-
1994. Along each of two 100 m transects, one on the edge and one in the interior of the

woodlot, I marked ten blocks at 10 m intervals for a total of 20 blocks. Within each block

I marked eight or three 0.5 m2 quadrats in 1992 and 1993, respectively. Halfm wide aisles
were retained between adjacent quadrats. The blocks originated at the same distance along
each of the transects in both years. In 1992 the blocks were placed on the north side of the
transect and in 1993 the blocks were placed on the south side of the transect, one meter
away flom the 1992 blocks. In 1992 the quadrats were arrayed in four rows of two
columns per row, so the block occupied a 8 by 1.5 meter rectangular area along the
transect. In 1993 the quadrats occupied a 0.5 by 2.5 meter area along the transect.
Natural Seedlings

Each fall on a weekly basis I tagged naturally occurring seedlings as they emerged
with numbered poultry leg bands (N=l3,568 in 1992, N=4,522 in 1993). During each
emergence week, I measured cotyledon diameter on a subset of newly emerging seedlings

(hereafter referred to as initial size) using a template of circles ranging from 1 to 9 mm in

76
diameter in 0.5 mm increments. Between 1,000 and 1,600 randomly selected seedlings

were measured in each census week. All seedlings in early and late censuses were
measured for cotyledon diameter because the total number of seedlings emerging in those
censuses was less than 1,000. Because the cotyledons grow during the fall, seedlings
measured at emergence and sru'viving to the onset of winter were measured again for
cotyledon diameter in early December 1992 and late November 1993 (hereafter referred to
as fall size). Most seedlings had not yet begun to initiate true leaves by early December, so
cotyledon diameter reflects seedling size. Cotyledon diameter at emergence explained 88%
or 55% of the variation in photosynthetic area and total seedling weight, respectively,
(photosynthetic area =-2.74+0.891t(diameter/2)2, $309, p=0.0001 and total seedling
weight =1.28+0.021t(diameter/2)2, df=309, p=0.0001). At the onset of winter cotyledon
diameter explained 85% and 75% of the variation in these two traits (Photosynthetic area
=15.22+0.821t(diameter/2)2, df=95, p=0.0001 and total seedling weight =-
0.75+0.061r(diameter/2)2, df=95, p=0.0001). In the spring of 1993 and 1994 prior to seed
dispersal, I collected all surviving plants in the quadrats and counted flower, fiuit, and
seed number for each individual, noting removal of the apical meristem by deer.

Mortality was scored at three stages in the life cycle that reflected difl‘erent
selective episodes. Mortality due primarily to slug herbivory was observed during
establishment (1). Slug herbivory resulted in seedlings that lacked cotyledons or an apical
meristem and was easily scored. Mortality was also scored at the onset of winter (2) and
in the spring (3). As a consequence of mortality during these three episodes of viability

selection, not all seedlings were scored for all traits. For example, seedlings that emerged,

77
but were eaten by slugs could only be scored for the trait emergence week. Seedlings that

were not eaten by slugs were scored for initial size, and seedlings surviving to the onset of
winter were scored for fall size. The final episode of fecundity selection (4) included only
those individuals that survived to spring and thus had expressed all three traits. In order to
include all seedlings in the multivariate analysis described below, I partitioned the analysis
of the magnitude and direction of phenotypic selection into four biologically relevant
episodes (Figure 8). This episodic analysis is biologically relevant because the episodes
relate to the difl‘erent postulated selective agents. In the first two episodes, slug herbivory,
drought and intraspecific competition were likely sources of mortality. From fall to spring,
mortality agents included intraspecific competition, physiological stress, and deer
herbivory. Deer herbivory also influenced fecundity in the final episode of fecundity
selection. This episodic approach allows me to estimate the total magnitude of phenotypic
selection on three traits, emergence week, initial size, and fall size.
Planted Seedlings

Seed weight is another maternally inherited trait that is genetically and
phenotypically correlated with emergence week, initial size and fall size (Chapter 1). Seed
weight can influence the outcome of competitive interactions (Gross 1984; Gross and
Smith 1991) and the genesis of size and fecundity hierarchies in plant populations (Waller
1985; Stanton 1985). To remove the efl'ects of selection on seed weight from other
maternally inherited and correlated traits included in the multivariate selection analysis and
to determine the extent to which seed weight influences either viability or fecundity

components of fitness, I conducted a field experiment with seeds of known weight.

78
I monitored emergence, survival and fecundity of individually weighed seeds that I

had planted back into the field. These seeds originated fiom natural fiuiting maternal
plants collected in early June. Seeds were planted in July into moist Sunshine seedling mix
to a uniform depth of 1 cm into either 2 cm sections of 15 mm diameter clear plastic
tubing in 1992 or 3 cm sections of 7 mm diameter plastic straws in 1993. These swds
were maintained in the greenhouse until August when they were tranplanted into the field
prior to natural gerrrrination cues and with minimal soil disturbance. In addition to
naturally produced seed, I also planted greenhouse produced seeds from the breeding
design described in Chapter 1. In 1992 a total of 3 180 seeds from field and greenhouse
mothers were planted in the population. In 1993 a total 2495 seeds fiom field collected
mothers were planted.

Seeds from maternal families were planted at two spatial scales to address how
spatial variation in selection influenced maternal family fitness when seeds from a family
were planted locally (i.e. experienced only one selective environment) or when they were
planted in numerous blocks across the population (i.e. experienced many selective
environments) (see Chapter 3). In addition, to address whether families were better
adapted to the location in which they were produced, seeds that were planted locally
consisted of two types. The first type of maternal family originated in the block in which it
was planted, while the second type was a maternal farrrily that was randomly assigned to
that block from the population at large. In this chapter I combine all planted seedlings in
one analysis to describe the overall pattern of phenotypic selection in each year.

Data Ana sis

79
I quantified phenotypic selection with two models that difl‘ered only in the traits

included in the analysis. For natural and planted seedlings, I examined a three trait model
that included : l) emergence week, 2) initial size (cotyledon diameter at emergence), 3)
fall size (cotyledon diameter in November). For the planted seedlings, I considered a four
trait model that inchrded seed weight. These two models allowed me to evaluate how the
inclusion of seed weight affected the estimates of direct selection on other traits in the
planted seedlings.

This multivariate selection analysis which quantifies the magnitude and direction of
selection acting directly on phenotypic traits by removing the effects of changes in
correlated traits can only include observations in which all phenotypic traits have been
measured for each individual (Lande and Arnold 1983; see recent review Brodie et al.
1995). When mortality eliminates some individuals, traits expressed later in ontogeny are
missing and those individuals must be excluded from the analysis. Arnold and Wade (1984
a, b) developed an episodic approach to selection analysis such that one can estimate the
direct efl‘ects of particular traits on components of fitness by considering episodes of
viability, fecundity, or sexual selection. This analysis by episodes, therefore, allows one to
include individuals who die before expressing all phenotypic traits of interest. The
estimates of selection resulting from this episodic analysis are conditional because they
only provide an estimate of the magnitude of selection if the individual survived to the
beginning of the episode being considered. To quantify the total magnitude of selection on
a set of traits throughout the life cycle, these conditional measures of selection must be

additive. Ifthe phenotypic variance-covariance matrix (P) does not change across all

80
episodes, then conditional selection gradients sum to the total selection gradient (Arnold

and Wade 1984 a, b). When P does change across episodes, selection gradients are made
additive by weighting conditional gradients by the cunnrlative change in P to that point in
the life cycle (Wade and Kalisz 1989; Kalisz 1986). This approach to additive partitioning
of the selection gradient requires that the original P be known at birth, i.e. all traits are
measured before selection occurs. When all traits of interest are not expressed at birth, the
additive partitioning of the selection gradient requires that the original P be reconstructed
(Lynch and Arnold 1988). Reconstruction of the original P requires the assumption that
changes in P are due solely to selection and that traits distributions are not changed by
selection prior to the time that they are manifested. Bennington and McGraw (1995a)
provide an empirical demonstration that this reconstruction can account for changes in P
due to selection.

Because mortality eliminated individuals at establishment and during the fall, I
errrployed an episodic analysis to estimate selection for three episodes of viability selection
and one episode of fecundity selection (Figure 8). I reconstructed the original P according
to Lynch and Arnold (1988) to make conditional selection parameters additive.

Phenotypic selection can produce changes both in the mean and variance of
phenotypic traits (Table 8). The conditional selection differential, 8;, measured as the
covariance between a trait and relative fitness, describes that change in the trait mean as a
result of selection in a given episode. This change may be due to direct selection on the
trait as well as changes due to selection on phenotypically correlated traits. The

conditional selection gradient, [3,, describes the change in the trait mean due only to direct

81
Table 8. Phenotypic selection parameters calculated in each episode(i).

 

 

 

Linear Non-linear
Parameter Response Symbol Response Symbol
Selection Change in trait mean 8, Change in (co)variance of C;
difl'erential trait
Selection Change in trait mean due [3, Change in (co)variance due 7,
gradient only to direct selection only to direct selection

 

effects and is calculated as the partial regression coeflicient for a given trait on relative
fitness in that episode given all other traits expressed in that episode. In order to quantify
changes in the variance, traits values must be expressed as squared deviations from the
mean (Lande and Arnold 1983; Brodie et al. 1995). The covariance between these squared
deviations and relative fitness in a given episode is the non-linear selection differential, C;,
that describes the change in trait variances due to selection. Changes in the variance due
only to direct effects of selection, the non-linear selection gradient, 7;, is calculated as the
partial regression coeflicient of the squared deviation trait values when the linear terms are
included in the model. Thus, the linear and non-linear selection gradients are determined
by two separate multiple regression models, 1) the linear model includes only trait values
and 2) the non-linear model includes trait values and their squared deviations. Therefore,
non-linear models account for changes in the mean when estimating changes in the
variance. In each episode I calculated selection differentials and selection gradients for
linear and non-linear components of selection. An analysis of variance inflation factors
indicated that these regression models were not compromised by multicollinearity (N eter,

Wasserman, and Kutner 1985).

82
Phenotypic traits were standardized to a mean of zero and a variance of one prior

to all analyses so that all differentials and gradients were expressed in units of standard
deviation and were comparable among traits and episodes. The covariances describing
selection differentials were calculated with a denominator of 11 rather than n-l (see Arnold
and Wade 1984 b p.726). In each viability episode fitness was either zero or one
depending on whether an individual died or survived, respectively. Each fitness was
standardized to the mean in that episode to calculate relative fitness. In the final episode of
fecundity selection, fitness was the number of seeds produced. Relative fitness was
expressed as seed number relative to the population mean in that episode. Relative fitness
was not transformed (Lande and Arnold 1983).

Additive linear (B) and non-linear gradients (y) are calculated by weighting the
changes in the phenotypic variance-covariance matrix over all episodes (i) to the original

phenotypic variance-covariance matrix (Po) according to the equations:

l3(i)=Po'l S(i) (1)

and

7(i)= P0-1C(i)P0.l (2)

where the linear selection differential, S(i), and the non-linear selection difl‘erential C(i) are
weighted by the fraction surviving to that episode (Lynch and Arnold 1988). Ifone
assumes that changes in the phenotypic variance-covariance matrix from episode to

episode are due only to selection quantified by the linear and non-linear conditional

83
gradients (i. e. not development), then Po can be reconstructed by sequentially back

calculating variances and covariances for traits not observed in a particular episode

according to the following equation:
Pi=Pi—1+Pi-IYPi-1'Pi-lBi-l[Pi-IBi-1]T (3)

(Lynch and Arnold 1988, equation. 2) where B and y are the conditional linear and non-
linear selection gradients in episode i, respectively.

In this study, all four phenotypic traits were measured by the third episode (i=2).
So only P1 and Po needed to be reconstructed. In the second episode (I), fall cotyledon
diameter was unobserved, so the reconstruction involved solving three simultaneous
equations for the variance in fall cotyledon diameter and its covariance with emergence
week and initial size based on the observed conditional selection gradients in that episode.
Likewise, in the first episode (0) the variance of fall cotyledon diameter and initial
cotyledon diameter and covariances of these traits with emergence week were based on
the simultaneous solution of five equations. When seed weight was included in the
analysis the number of unknowns in each episode increased, so that there were four
equations for P1 and seven equations for P0. As the number of unknowns and the number
of episodes involved in reconstruction increases, error associated with estimation can
increase. However, the compounding of errors in reconstruction is likely to be minor in
this study because reconstruction involved only two traits in two episodes. In addition,

reconstructed estimates of the phenotypic variances and covariances were tested for

84
significance by constructing 95% confidence intervals obtained from the bootstrapping

procedure described below.

Significance tests of selection parameters were based on bootstrap resampling
methods (Efion 1982; Dixon 1987, Dixon et al 1993). This approach was required
because I) regression analysis of viability selection is likely to be compromised by non-
normality of residuals (Mitchell-Olds and Shaw 1987), and 2) the additive partitioning of
the selection differentials and gradients according to Lynch and Arnold (1988) involves
the transformation of these regression parameters. Once conditional estimates are
transformed into additive estimates, they are no longer associated with significance tests
from the regression analysis. My protocol for resampling with replacement was as follows:
I) calculate the covariances between traits (including squared deviations) and relative
fitness to estimate linear and non-linear conditional selection differentials in each episode
2) estimate linear and non-linear conditional selection gradients in each episode via
multiple regression analysis, 3) use conditional gradients and the phenotypic variance-
covariance matrix to reconstruct Po, 4) transform linear and non-linear conditional
difl‘erentials into additive difl‘erentials using conditional gradients and weighting by the
fiaction that survived to that episode, 5) transform the linear and non-linear conditional
gradients into additive gradients using equations 1 and 2 above. Thus, the 95%
confidence intervals of both conditional and additive parameters as well as the original P
were obtained by the shifl distribution method in which the bootstrapped parameter means

are centered on the real value before the confidence intervals are calculated (Noreen

85
1989). Each resampled data set contained the number of observations originally observed

in that data set and 500 bootstrapped estimates were obtained for each parameter.
Survivogip An sis

Statistical comparisons of the survivorship of natural and planted seedlings through
the three episodes of viability selection for two years were based on a failure time analysis
using the log-rank chi-square statistic from the lifetable method of the lifetest procedure in
SAS (Fox 1993 ). First, I examined temporal difl‘erences across years by combining both
natural and planted seedlings within a year. Subsequently, I examined differences between
natural and planted seedlings within a given year. I examined temporal variation in
fecundity with a nested AN OVA in which treatment (natural vs. planted) was nested
within year.

RESULTS

mum—unity for selegion

The proportion surviving across episodes did not differ between years (log-rank
chi-square, x2 =3.13, df=1, p=0.0768) or between natural and planted seedlings in 1992
(log-rank chi-square, x2 =3.43, df=l, p=0.0639), but did difl‘er between natural and
planted seedlings in 1993 (log-rank chi-square, x2 =25.95, df=1, p=0.0001) (Figure 9).
Most mortality occurred between the fall and spring censuses. For both natural and
planted seedlings in 1992 and 1993, on average 10.4% died during establishment, 28.5%
died prior to winter, 39.2% died prior to fruit maturation in late May and early June, only

21.9% survived to the flowering/fiuiting stage.

86

.2858 05 85 8:88 08 888 “889me 35:88 8253 v8 i=8: 8m
o=8>< DH 8 88.» 95 Ba 8828 358.5 me 8850 88 awn—25 R335 .0 mama

mQOmEm
022mm 154‘.» 592221;?»th 139:7:

 

Se
a 25

fi
% m one w
an m 23 m
J / N 35 m
7 m 8.0 N

 

 

 

 

 

 

 

82 855m + / m 23 m

82 855.— + // m Snow
82 455p; rel ///l one 9

H cad
82 45.8% ill

woo;

 

 

 

 

 

 

 

 

 

 

 

 

 

87
Female fecundity, the number of seeds produced, was highly variable ranging from

O to 70, with an overall average of 8. 12 seeds per individual Plants browsed by deer had
lower fecundity than unbrowsed plants (Figure 10). Average fecundity difi‘ered
significantly among years and among treatments (natural vs. planted) within years (Nested
ANOVA, df=3,]529, p=0.0001) with lower seed production in 1992 (Figure 10). This
variance both in survival and seed production resulted in the greatest opportunity for
selection in the spring episodes of viability and fecundity (Figure 11).
Phenoggpic correlations among traits

The phenotypic variance-covariance matrix prior to the first and final episodes of
selection and their 95% bootstrapped confidence intervals are presented in Table 9. The
original matrix, Po, has been reconstructed based on the conditional selection gradients in
the first episode of viability selection according to Lynch and Arnold (1988), while the
final matrix, P3, is based only on individuals that survived to the spring. Emergence week
and fall size display a significantly negative covariance through all episodes, initial size and
fall size display a significantly positive covariance through all episodes, while the
covariance between emergence week and initial size displays positive, negative, and non-
significant values in the original matrix depending on the year and treatment (natural or
planted). When seed weight was included in the analysis of planted seedlings in both years,
seed weight displayed a significant positive covariance with initial size, fall size, however

the covariance with emergence week varied fiom negative in 1992 to positive in 1993.

88

,com3o5 8: 80>» 85 3.83 can goon 3 «.8305 3E3 8m 551:? 8a memo»
Ben E mmflfioa U253 35 5‘5an C28 235% _ Hy bags—zoom ommb>< .2 oBmE

 

 

 

 

 

 

 

‘ V § mo
am K 7,,
\ s H a
ammaomm D m K 10 m
ammaommza § \ t w
\v ,2
fi< I not m ‘

 

 

 

 

 

 

89

8ch “5:338 =a 8 bags cocoa—om mo caveman 2: 5m 3.5th 2.5.

«3523 =~ 890m m§u$€E =m 5m 36:5 02:22 E oogcg 05 m. buBtonEo :38
2S. .3880 3050.5 2: szSSm min—USE: .6 5m cocoa—om $0 $830 :28 Sm 368m
0238 E 885? 05 mm @2585 cacao—om 5m barrage Ecocficoo 2F ._ _ oBmE

QmPZ<Am A<§<Z
moo. N09 mom: Nag

N-q—‘tn
.—. d

 

 

Egmfimfimm I

fig SE
022% ”A.

Eczema D

“’1
N

SSEINJJJ HALLV'IEIH NI HDNVDIV A

vnm
m

 

 

V‘.
v

 

90

Table 9. The phenotypic variance-covariance matrix prior to the original (Po) and final
(P3) episodes of selection for natural and planted seedlings in 1992 and 1993 for three
traits (emergence week (WK), initial size (IS), and fall size (FS) and four trait models
(inchrding seed weight (SD)). The original matrix has been reconstructed from conditional
selection gradients (see Methods). The 95% confidence intervals for each element are also
provided.

 

TRAIT P0 95% CI P3 95% CI

 

LOWER UPPER LOWER UPPER

 

1. NATURAL SEEDLINGS 1992

WEEK 0.999 0.999 0.999 0.738 0.548 0.934

WK-IS 0.089 0.025 0.158 0.313 0.154 0.497

WK-FS -0.701 -0.759 -0.636 -0.278 -0.383 -0.174
INITIAL 0.999 0.999 0.999 1.065 0.865 1.332

IS—FS 0.488 0.432 0. 542 0.321 0.194 0.446

FALL 0.996 0.970 1.023 0.691 0.586 0.812

2. NATURAL SEEDLINGS 1993

WEEK 1.000 1.000 1.000 0.875 0.801 0.953

WK-IS -0.020 -0.059 0.023 0.022 -0.056 0.104

WK-FS -0.411 -0.450 -0.369 -0.353 -0.428 -0.287
INITIAL 1.000 1.000 1.000 0.961 0.858 1.066

IS-FS 0.600 0.563 0.636 0.519 0.432 0.600

FALL 1.024 1.006 1.042 0.852 0.761 0.934

3. PLANTED SEEDLINGS 1992

WEEK 0.999 0.999 0.999 0.698 0.610 0.795

WK-IS -0.106 -0.177 -0.034 -0.075 -0. 165 0.010

WK-FS -0.417 -0.489 -0.343 -0.295 -0.376 -0.218
INITIAL 0.999 0.999 1.000 0.891 0.771 1.010

IS-FS 0.587 0.525 0.647 0.414 0.297 0.523

FALL 1.039 1.011 1.068 0.735 0.636 0.845

4. PLANTED SEEDLINGS 1992 WITH SEED WEIGHT

WEIGHT 0.999 0.999 0.999 0.841 0.729 0.953

SD-WK -0.113 -0. 179 -0.047 -0.017 -0. 109 0.070

SD-IS 0.403 0.342 0.460 0.323 0.234 0.429
SD—F S 0.555 0.479 0.630 0.458 0.366 0.548
WEEK 0.999 0.999 0.999 0.698 0.608 0.798
WK-IS -0. 122 -0.198 -0.058 -0.075 -0. 161 0.009
WK-FS -0.454 -0.526 -0.379 -0.295 -0.364 -0.220
INITIAL 1.003 0.998 1.011 0.891 0.772 1.008
IS-FS 0.601 0.529 0.664 0.414 0.314 0.531

FALL 1.076 1.032 1.125 0.735 0.634 0.840

 

91

 

 

 

Table 9 (cont’d).
TRAIT P0 95% CI P3 95% CI
LOWER UPPER LOWER UPPER
5. PLANTED SEEDLINGS 1993
WEEK 0.999 0.999 0.999 0.746 0.626 0.885
WK-IS 0.043 -0.009 0.098 0.143 0.041 0.242
WK-FS -0.393 -0.441 -0.347 -0.274 -0.357 -0.198
INITIAL 0.999 0.999 1.000 1.022 0.903 1.149
IS-FS 0.569 0.531 0.611 0.436 0.345 0.540
FALL 1.000 0.987 1.013 0.720 0.638 0.807
6. PLANTED SEEDLINGS 1993 WITH SEED WEIGHT
WEIGHT 0.999 0.999 0.999 1.018 0.873 1.173
SD—WK 0.154 0.111 0.200 0.135 0.044 0.225
SD-IS 0.475 0.434 0.515 0.504 0.394 0.604
SD-F S 0.463 0.410 0.509 0.464 0.373 0.556
WEEK 0.999 0.999 0.999 0.746 0.618 0.878
WK-IS 0.051 0.001 0.106 0.143 0.042 0.236
WK-F S -0.376 -0.424 -0.324 -0.274 -0.348 -0.188
INITIAL 0.999 0.997 1.001 1.022 0.896 1.145
IS-FS 0.562 0.522 0.601 0.436 0.343 0.529
FALL 0.986 0.966 1.010 0.720 0.632 0.809

 

92
Total magm de and direction of selection

A Changes in trait means

Three trait models

The pattern of selection is remarkably consistent among years and between natural
and planted seedlings within years: seedlings that are large at establishment and large at
the onset of winter have enhanced survival and reproduction. Selection on emergence
week is more variable. The total linear selection difierential estimating the change in a trait
mean due both to direct selection on that trait and indirect selection on phenotypically
correlated traits (Table 9) summed across all episodes shows that for natural and planted
seedlings in both years, selection favors large initial size and large fall size (Figures 12-
15A). In addition to direction, the magnitude of the selection differentials for initial and
fall size are similar for natural and planted seedlings in both years (Table 10). In contrast,
the magnitude and direction of the total selection differential for emergence week varies
among years: in 1992-3 emergence week does not change (Figures 12, 14A), while in
1993-4 late emergence is favored in the natural seedlings (Figure 13A) and early
emergence is favored in the planted seedlings (Figure 15A). Total selection difl‘erentials on
emergence week are of small magnitude in both years, but only significant in 1993-4.

The total selection gradient that reflects changes in trait means due only to the
direct effects of selection showed a difi‘erent pattern from the total selection difl‘erential
(Figures 12-15B). Later emergence and large fall size were favored for natural and planted

seedlings in both years, while mean initial size did not change as a result of direct selection

93

Figure 12. Phenotypic selection through three episodes of viability and one episode of
fecundity selection for natural seedlings in 1992. Viability episodes include survival to
establishment (solid), survival to the onset of winter (vertical), survival to spring
(diagonal), and the final episode of fecundity selection (clear). The total magnitude of each
selection parameter (I) across all episodes and the 95% confidence intervals (O) based on
500 bootstrap samples based on reconstructing the original phenotypic variance-
covariance matrix are depicted. Selection parameters include: A) linear selection
differential, B) linear selection gradient, C) non-linear selection differential, D) non-linear
selection gradient. Trait abbreviations follow Table 9. Significant episodes of selection are
denoted by *. Note the difference in scale for non-linear selection parameters.

94

m

S:I'SI

412m

SJ'XM

ASH—7:

SI'XM
)IEIEIM

 

v5.95

mnNn~nonwnw
N —- ° C: —.-
lNHICIV‘dD WEINI'I'NON

 

m

S:1'SI

j<m

'IVLLINJ
S:I')IA\

Adah—7:

SI'XAA
)IEIEIM

 

vim?

.2 8:3

 

95

.2 28¢ 26:92 $38.3 =< .82 a
3:588 fig: Sm mauve—om 3658M me £6856 28 v8 bang me 8389 3:: swag: cage—om 0:56:23 .m. 2.35

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m ...... .m m m m ...... m .m m
1 M «a. m X 1 W... M 9. H
No- N
25AM
3- T I.
8.9m 6
o w , .... H l
85 9 In M *fl
3 m . 1
m; m
m
No
.32 45:7: vim? dé 45:7: Mum?
.3- "3.
W 3- 1 m Sm
.. 2 .. mo 8
.V .. = .. W; m .. ...=______ a m;
... \ W3 o .. ”1 mg m
ILIII Mme N ”no a
D ME “to m.
. 2 3 m3
mg m (M ed
m .3 who

96

.2 05mm 325 285% .2 «g s
mmfivoow coin—n 3m 8:028 bfigoom Mo 2523 one can tapas me «$523 ooh: amazon: 82628 "vibes—E .2 onE

m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m m m M m m W .... m m.
1 w «on Q X 1. was E Q X
”No- N .3- m
) H2.0.0 O Had-N
I m N m 1
0 0 Ed. m o o .. ”.3. m
1 P u u 1. ”.89 u ”18.?
E .1. E 1 S *flwo W “E *flllfil * * we m
o g 0 ENE Waco I E... wnoem
. . u... m . . n u... m
0 W26 M 0 Mn;
ammo 0.3
.SE 45:22 2mm? .35 .3222 mums,
,No- “3-
0 2.0-1 8.2%
Ho .. mo
1% .a H 3.0m l\\ . .. .films m
1H \ ... W8 w * \\ ... Mao 0
IR I m . 9 I... I . O m . fl
1* “no N I . mo E
III . .
0 ac m g u
.3 m ,3
Moo Moo
m”; (The

 

 

 

97

.2 05mm 30:3 23:5 =< .82 s 2568...
c853 8m canoe—om 3:858.“ me £529 28 was big? mo 8vo2ao 035 5:88 8:028 29365.3 .2 oar—

W W

m

..d
S

SJ'XM
)IM
)IEIEIM
'I'IVJ
SI'XM

SJ'SI

N.
C?

m m
1 ......
O

m _ .o-
_.o-
86.

If} .—
_, .
O

lNEIIGVHD WHNI'I'NON

 
 

NWF—V‘l
o"".c>'°.
o o

413‘..— A<HE7= Mum—3 413‘..— ASEZfl vim},
. Nd-

m S-
L‘ a m
2&1 we
*
... _.o
H No

 

 

 

 

 

 

 

H to
no
, so
H 6o

 

 

"1
o
.INEIICIVHD WEINI'I

Cl

 

 

 

 

 

 

 

98

Table 10. Additive linear selection difi‘erentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes and the total across all episodes. Below each value are
the 95% confidence intervals based on bootstrap resampling (n=500). Linear parameters
reflect the change in the mean in units of standard deviation. Values significant at p<0.05
are bold

 

 

 

 

 

A LINEAR DIFFERWIAL
TRAIT EPISODE
0 1 2 3 TOTAL
1. NATURAL SEEDLINGS 1992-3
WEEK 0.022- 0.181- 4.1.186“ 0.007 0.010
0.011 0.032 0.151 0.209 -0.228 0.143 -0.026 0.012 0.052 0.061
mm 0.002 0.144- 0.109- 0.021 0.276“
0.000 0.004 0.110 0.181 0.064 0.162 0.005 0.047 0.217 0.338
FALL .0015- .0051- 0.260“ 0.046- 0239‘-

-0.023 -0.008 -0.085 -0.017 0.219 0.300 0.023 0.069 0.183 0.299
2. NATURAL SEEDLINGS 1993-4

WEEK -0.034“ 0. 105“ -0.004 -0.018“ 0.049“
-0.049 -0.019 0.080 0.133 -0.041 0.030 -0.034 -0.003 0.003 0.094

INITIAL 0.001 0.096“ 0.094“ 0.044“ 0.236“
-0.001 0.002 0.070 0.125 0.060 0.131 0.025 0.064 0.184 0.286

FALL 0.014“ 0.014 0.182“ 0.067“ 0.276“

0.008 0.020 -0.006 0.034 0.143 0.218 0.050 0.084 0.227 0.318
3. SEEDLINGS 1992-3

WEEK -0.005 0.046“ -0.041 0.001 0.002
-0.039 0.030 0.017 0.077 -0.089 0.006 -0.025 0.025 -0.066 0.068

INITIAL 0.000 0.035“ 0.113“ 0.032“ 0.180“
-0.003 0.004 0.003 0.066 0.062 0.163 0.001 0.064 0.117 0.247

FALL 0. 002 0.002 0.220“ 0.071 “ 0.295“

-0.013 0.016 -0.018 0.024 0.176 0.268 0.044 0.098 0.239 0.346
4. PLANTED SEEDLINGS 1992-3 WITH SEED WEIGHT

SEED -0.045“ 0.033“ 0.070“ 0.056“ 0.113“
-0.077 -0.012 0.002 0.062 0.019 0.117 0.031 0.082 0.041 0.181
WEEK -0.005 0.046“ -0.041 0.001 0.002
-0.039 0.031 0.018 0.077 -0.088 0.009 -0.023 0.026 -0.063 0.064
INITIAL -0.017“ 0.035“ 0.1 13“ 0.032“ 0.163“
-0.032 -0.004 0.003 0.068 0.062 0.163 0.002 0.064 0.092 0.231
FALL -0.021 0.009 0.220“ 0.071 “ 0.278“

-0.046 0.001 -0.016 0.039 0.178 0.265 0.046 0.099 0.221 0.335
WDLINGS 1993-4

WEEK -0.060“ 0.085“ -0.065“ -0.025“ 0.065“
-0.079 -0.044 0.063 0.109 -0.117 -0.008 -0.046 -0.006 -0. 130 -0.004

INITIAL -0.003 0.046“ 0.110“ 0.031“ 0.185“
-0.006 0.001 0.022 0.071 0.047 0.171 0.010 0.053 0.114 0.257

FALL 0.024“ -0.01 1 0.275“ 0.057“ 0.345“

0.017 0.031 -0.026 0.006 0.222 0.332 0.036 0.077 0.285 0.406
6. PIANI'ED SEEDLINGS 1993-4 WITH SEED WEIGHT

SEED -0.000 0.084“ 0.077“ 0.041 “ 0.203“
-0.013 0.014 0.060 0.109 0.017 0.138 0.019 0.064 0.133 0.274
WEEK -0.060“ 0.085“ -0.065“ -0.025“ -0.065
-0.081 -0.042 0.062 0.109 -0.121 «0.008 -0.045 -0.004 -0.129 0.006
INITIAL 0.001 0.046“ 0.110“ 0.031 “ 0.189“
-0.006 0.009 0.021 0.069 0.044 0.170 0.006 0.052 0.116 0.254
FALL 0.028“ 0.007 0.275“ 0.057“ 0.367“

0.018 0.039 -0.012 0.025 0.221 0.332 0.039 0.076 0.308 0.427

99
Table 10 (cont’d).

 

B. LINEAR GRADIENT

 

 

 

 

 

 

 

TRAIT EPISODE
0 I 2 TOTAL
1. NATURAL SEEDLINGS 1992-3
WEEK 0.022“ 0.189“ 0.029 0.342“
00]] 0.032 0.152 0.228 -0.090 0.171 0.201 0.524
INITIAL 0.000 0.114“ -0.040 0.014
0.000 0.000 0.081 0.152 -0.l39 0.065 -0.1 17 0.132
FALL 0.000 0.026“ 0.301“ 0.475“
0.000 0.000 0.013 0.043 0.152 0.469 0.291 0.668
2. NATURAL SEEDLINGS 1993-4
WEEK -0.034“ 0.106“ 0.096“ 0.177“
-0.049 -0.019 0.082 0.133 0.059 0.135 0.131 0.227
INITIAL 0.000 0.100“ -0.051“ 0.054
0.000 0.000 0.075 0.128 «0.096 -0.005 -0.001 0.112
FALL 0.000 -0.003“ 0.246“ 0.310“
0.000 0.000 -0.005 -0.001 0.194 0.300 0.254 0.367
3. PLANTED SflDWGS 1992-3
WEEK -0.005 0.050“ 0.061“ 0.146“
-0.039 0.030 0.021 0.081 0.011 0.11 1 0.071 0.214
INITIAL 0.000 0.040“ -0.029 -0.008
0.000 0.000 0.008 0.071 -0.098 0.035 -0.091 0.079
FALL 0.000 -0.001 0.252“ 0.346“
0.000 0.000 0.002 0000 0.184 0.320 0.264 0.430
4. PLANTED SEEDLINGS 1992-3 WITH SEED WEIGHT
SEED -0.046“ 0.027 -0.072“ -0.068
0.079 -0.013 -0.007 0.060 -0.133 -0.013 -0.148 0.010
WEEK -0.010 0.053“ 0.078“ 0. 160“
-0.045 0.027 0.024 0.083 0.020 0.133 0.088 0.232
INITIAL 0.000 0.030 -0.020 «0.01 1
0.000 0.000 -0.004 0.066 -0.089 0.049 -0.097 0.072
FALL 0.000 -0.000 0.286“ 0.368“
0.000 0.000 -0.002 0.001 0.216 0.364 0.292 0.451
5. PLANTED SEEDLINGS 1993-4
WEEK -0.060“ 0.081“ 0.084“ 0.102“
-0.079 -0.044 0.059 0.104 0.024 0.141 0.033 0.171
INITIAL 0.000 0.046“ -0.102“ 0.056
0.000 0.000 0.021 0.069 -0.186 -0.024 -0.l48 0.034
FALL 0000 -0.005“ 0.366“ 0.417“
0.000 0.000 -0.008 -0.002 0.291 0.452 0.340 0.510
6. PLANTED SEEDLINGS 1993-4 WITH SEED WEIGHT
SEED 0.010 0.069“ -0.090“ 0.014
-0.004 0.023 0.042 0.097 -0.l6l -0.015 -0.073 0.092
WEEK -0.062“ 0.072“ 0.107“ 0.105“
-0.083 -0.043 0.050 0.095 0.044 0.170 0.031 0.188
INITIAL 0.000 0.013 -0.083 -0.074
0.000 0.000 -0.012 0.037 0171 0.005 -0.l74 0.027
FALL 0.000 -0.006“ 0.409“ 0.447“

0.000 0.000 -0.010 -0.003 0.322 0.501

0.354 0.535

100
(Figures 12-15B). The magnitude of the selection gradient was similar for fall size across

years, while the magnitude for emergence week showed weaker selection for the planted
seedlings (Figures 14- 1 SB) when compared to natural seedlings (Figures 12-13B) in both
years (Table 10).

The difl‘erence between selection differentials and selection gradients indicates that
indirect efi‘ects of selection on phenotypically correlated traits will influence the overall
change in the trait mean. One can partition the selection difl‘erential into direct and indirect
components. The direct component is analogous to the selection gradient. The indirect
component of the difierential (Figure 16) shows how selection on phenotypically
correlated traits will produce a change in the mean of a given trait. The pattern of indirect
selection was consistent: indirect selection favored early emergence and large initial size
via direct selection on fall size and decreased fall size (not significant in all cases) via direct
selection on emergence week (Table 9).

Four trait models with seed weigm

When seed weight is included in this episodic selection analysis, it does not change
the sign or magnitude of the total linear selection difl‘erentials or gradients observed in the
three trait analysis (Figures 17-18AB, Table 10). However, in 1993 the total selection
differential on emergence week becomes non- significant. Furthermore, the significance
and magnitude of certain episodes contributing to the total gradients does change. In both
years the inclusion of seed weight eliminates significant direct selection on initial size in
any episode (Figures 17-18B).

As with the other three traits, the total selection difl‘erential for seed weight is

consistent in sign and magnitude across years favoring an increase in seed weight

101

Figure 16. Indirect selection difi‘erential on three traits for all selection episodes for
natural seedlings in 1992 (A) and 1993 (B) and planted seedlings in 1992 (C) and 1993
(D). Viability episodes include survival to establishment (solid), survival to the onset of
winter (vertical), survival to spring (diagonal), and the final episode of fecundity selection
(clear). The total magnitude of each selection parameter (I) for all episodes and the 95%
confidence intervals (0) based on 500 bootstrap samples for the indirect differential are
depicted. Trait abbreviations follow Table 9. Significant episodes of selection are denoted
by *

102

gm

«'1.
C?

 

T—r

 

 

$7M] —-OI

w No-

 

o

 

 

 

 

 

 

 

 

 

 

 

WYQN":
ooooo

 

d<m

ASE

Mmm?

 

 

 

I
"1 N. "'2
cl: c3 c|>

 

V

'11 III! In. .1

 

 

 

 

 

N—
GO

 

 

 

 

 

 

WY").
coo

 

'IVILNEIHHMG .LDEDIICINI

'IVLLNHHEHJICI lDHIIICINI

417$

Ares—.57:

Emma

.2 03E

M.
C?

 

No-

 

 

1%..

 

*-

k

*3
m ....lu
(D v-I.

0.

v v
'—

 

 

 

 

 

 

 

 

 

 

 

'IVIlNEIHHJflIG .IQEDIICINI

 

 

417E

179:7:

Mmm?

 

“Evin”.
ooooo

M.
‘?

Nd-

r-1
0
I

'IVLLNEIHEIEIJIG .IQEDIICINI

“I‘VE“."f
oocoo

103

”UV!

 

.2 08mm 32.8 ...—388 =< .82 a 38:88... 383 8.. E03 89. wages 8%: Be 8 8829. 2.5.5.: .: 3:3

mm
mm

min

SJ'SI
SJ’CIS
SI‘CIS

Nd-
n _ .o-

'—
O
I

3.?

mod

.—
O

o
.LNEIICIVHD WEINI'I‘NON

2d

O Q

N.
o

d<m ASE vim—3 FIG—m3

N.
o.

 

TTTT

 

I I
O —-
°.

 

 

“fig-l
TI IIY'Y

 

 

 

 

 

MS

 

:E.
h ...K
*
O

 

 

«101.
CO

to

 

 

'0.
o

.LNEIICIVHD HVEINI’I

 

 

 

 

 

 

M 2.
t he

 

..dI
mxU
ma
S

j<m

m

ASE

m

SJ‘CIS
SI‘CIS
XM'CIS
.LHDIEIM

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.2 05mm 3&8 828% =< .82 a 38:38. BEE é €83 88 wages as: 58.. 8 8828 2988.5 .2 855

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

S M S M
mammmmmmflm mammwmmmflm
1w mmxmmmm 1m ngummm

n3- 2.o-N
do- w 3- m
2.0-.N 2.0-1.
8. 1 S- m
mco.m no.0-W
o W o m
3o m 36 a
m S m S m.
2am flew
No No
.92 45.57: vim? E985 3% 45:5 Mum? $9.95
a ”do- ”No-
.. ... .131 1.9%
L\\ g ... ... \ mo N «5:: ... mo W
\ .. low ..1\ m3
.\ A _, ,
m . D O b m .
k mom mmoy
_.+. A”... .... Em
C . m n
mmod <mo.ow
”Z ”Z

105
However, the selection gradient showed no significant change in mean seed weight due to

direct efl‘ects of selection on that trait. Therefore, indirect efl‘ects mediated via significant
positive covariance with fall size (Table 8) must generate the positive selection differential
on seed weight. Indirect efl‘ects on seed weight due to its covariance with emergence week
opposed the indirect effects due to direct selection on fall size in 1992 and complemented
them in 1993 (Table 9).

B. Changes in trait variances and covariances

Three trait models

In contrast to changes in trait means that were relatively consistent in sign and
magnitude across years and between natural and planted seedlings, changes in trait
variances and covariances due to direct and indirect selection were not consistent (Figures
lZ-lSCD, Table 11). In general the total non-linear selection difi‘erentials and gradients
only displayed significant values in 1992. While the planted seedlings in that year displayed
a few significant total non-linear difi”erentials and gradients, the traits of natural seedlings
showed the strongest non-linear effects.

The total non-linear selection differential showed a decreased variance in
emergence week, and initial and fall size, a decreased covariance between initial size and
fall size and an increased covariance between week and fall size in the natural seedlings in
1992 (Figure 12C). In that year planted seedlings also exhibited decreased variance in
emergence week and increased covariance between emergence week and fall size (Figure

14C). The magnitude of these differentials differed between natural and planted seedlings

106

Table 11. Additive non-linear selection differentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes and the total across all episodes. Below each value are
the 95% confidence intervals based on bootstrap resampling (n=500). Non-linear
parameters reflect the change in the variance in units of standard deviation squared.

Values significant at p<0.05 are bold.

 

 

 

 

 

A. NON-LINEAR DIFFERENTIAL
TRAIT EPISODE
0 l 2 3 TOTAL
1. NATURAL SEEDLINGS 1992-3
WEEK 0.009“ 0.041“ -0.453“ 0.002 -0.401“
0.002 0.016 0.020 0.065 -0.681 -0.241 -0.036 0.044 -0.628 -0.l91
WK-IS 0.001 0.008 0.004 0.013 0.026
0.000 0.002 -0.016 0.034 -0.058 0.065 -0.020 0.045 -0.041 0.099
WK-FS -0.007“ -0.025“ 0.392“ -0.000 0.361“
-0.012 -0.002 -0.045 -0.005 0.211 0.593 -0.033 0.031 0.189 0.553
INITIAL 0.000 -0.042“ -0.060 -0.009 -0.111“
-0.000 0.000 -0.061 -0.027 -0.120 0.004 0.047 0.025 -0.178 -0.045
IS-FS -0.001 -0.029“ -0.052 -0.014 -0.095“
-0.001 -0.000 -0.050 -0.005 -0.126 0.027 -0.048 0.013 -0.l71 -0.018
FALL 0.005“ 0.002 -0.320“ 0.002 -0.312“
0.001 0.008 -0.021 0.026 -0.488 -0.l45 -0.028 0.031 -0.474 -0.l4l
2. NAT 9% SEEDLINGS 1993-4
WEEK -0.038“ 0.01 1 -0.024 0.019“ -0.033
-0.052 -0.024 -0.013 0.034 -0.062 0.020 0.003 0.036 -0.083 0.024
WK-IS 0.001 0.005 0.021 0.010 0.036
-0.001 0.002 -0.021 0.027 -0.019 0.058 -0.009 0.026 -0.017 0.087
WK—FS 0.016“ -0.001 0.055“ 0.013 0.057
0.010 0.022 -0.021 0.018 0.009 0.102 -0.032 0.005 «0.000 0.109
INITIAL -0.000 -0.043“ 0.040“ 0.014 0.01 1
-0.000 0.000 -0.063 -0.026 0.002 0.082 -0.015 0.043 -0.043 0.072
IS-FS -0.000 -0.028‘I 0.023 0.008 0.003
-0.001 0.000 -0.045 -0.012 -0.025 0.077 -0.022 0.039 -0.058 0.078
FALL -0.006“ «0.016 -0.031 0.014 -0.039
-0.010 -0.004 -0.032 0.000 -0.083 0.028 -0.014 0.044 -0.103 0.040
3. PLANTED SEEDLINGS 1992-3
WEEK -0.041“ -0.033“ -0.041“ -0.012 -0.127“
-0.070 -0.013 -0.059 -0.007 -0.077 -0.007 -0.033 0.010 -0.172 -0.075
WK-IS 0.004“ 0.046“ -0.025 0.019 0.043
0.001 0.009 0.011 0.083 -0.075 0.017 -0.011 0.048 -0.019 0.102
WK-FS 0.017“ 0.037“ 0.021 0.016 0.091“
0.005 0.029 0.014 0.063 -0.021 0.064 -0.015 0.045 0.027 0.148
INITIAL -0.000 «0.014 0.018 0.005 0.008
-0.001 0.000 -0.042 0.013 -0.037 0.074 -0.029 0.038 -0.063 0.080
IS-FS -0.002 -0.024“ 0.044 -0.009 0.009
-0.004 -0 000 -0.049 -0.002 -0.031 0.113 -0.047 0.030 0.0% 0.091
FALL -0.007“ -0.027“ 0.010 -0.012 -0.035
-0 013 -0.002 -0.050 -0.006 -0.057 0.073 -0.045 0.023 -0.117 0.040

 

Table 11 (cont’d).

107

 

 

 

 

 

 

ANON-LINEAR DIFFERENTIAL
TRAII EPISODE
0 l 2 3 TOTAL
4. MIEQ -IJNGS 1992-3 WITH WEIGHT
SEED -0.009 -0.025 -0.016 0.001 -0.049
-0.038 0.018 0054 0.007 -0.066 0.036 -0.031 0.031 -0.117 0.024
SD-WK 0.043“ 0.046“ 0.004 -0.003 0.091“
0.011 0.077 0.012 0.079 -0.038 0.051 0026 0.020 0.025 0.151
SD-IS -0.007 -0.013 0.016 -0.002 -0.006
-0.021 0.005 -0.048 0.026 -0.048 0.083 -0036 0.029 -0.089 0.077
SD-FS -0.022 -0.031“ -0.003 0.001 -0.055
-0.046 0.001 -0.061 -0.001 «0.067 0.055 -0.030 0.030 -0.140 0.034
WEEK -0.050“ ~0.034“ -0.046“ -0.017 -0.147“
-0.083 -0.022 -0.064 -0.006 -0.082 -0.009 -0043 0.010 «0.202 -0.090
WK-IS 0.021“ 0.045“ -0.024 0.021 0.063
0.006 0.037 0.007 0.079 «0.069 0.024 -0.009 0.048 -0.001 0.122
WK-FS 0.042“ 0.047“ 0.030 0.01 8 0.137“
0.018 0.070 0.017 0.078 -0.018 0.081 -0015 0.049 0.064 0.202
INITIAL -0.004 -0.013 0.018 0.004 0.006
-0.012 0.001 -0.042 0.016 -0.044 0.071 -0.029 0.038 -0068 0.081
IS-FS -0.012“ -0.026 0.044 -0.010 -0.004
-0.026 -0.001 -0.058 0.003 -0.049 0.114 -0.050 0.027 -0114 0.091
FALL -0.028“ -0.039“ 0.003 -0.013 -0.077
-0.053 -0.008 -0.071 -0.011 -0.082 0.073 -0.054 0.025 -0.185 0022
WDUNGS 1993-4
WEEK -0.012“ -0.012“ 0.017 0.007 -0.000
-0.021 0002 -0.022 «0.003 -0.034 0.073 -0.012 0.029 -0.062 0.069
WK-IS -0.001 -0.017 0.007 0.005 «0.005
-0.002 0.000 -0.036 0.005 -0.056 0.069 -0.017 0.025 -0.073 0.067
WK-FS 0.005“ -0.004 -0.014 -0.010 -0.024
0.001 0.008 -0.015 0.007 -0.095 0.060 -0.033 0.011 -0.111 0.049
INITIAL -0.000 -0.013 0.036 -0.018 0.005
-0.000 0.000 -0.031 0.004 -0.037 0.119 -0.052 0.015 -0.074 0.090
IS-FS 0.000 -0.000 0.011 «0.028 -0.017
-0.000 0.001 -0.014 0.013 -0.079 0.109 -0.062 0.007 -0.113 0.083
FALL -0.002 0.002 0.014 -0.013 -0.027
-0.003 -0.000 -0.010 0.013 -0.111 0.082 «0.040 0.014 -0.l20 0.075
F. PLANTED S_EEDLINGS 1993-4 WITH SEED WEIGHT
SEED -0.001 -0.025“ 0.084“ 0008 0.050
-0.010 0.007 -0.046 -0.004 0.016 0.147 -0.041 0.024 -0.023 0.125
SD-WK 0.019 -0.028“ 0.007 «0.003 -0.044
-0.038 0.002 -0.055 -0.002 -0.053 0.072 -0.028 0.019 -0.116 0.034
SD-IS 0.000 -0.012 0.075 -0.014 0.049
0004 0.004 .0040 0.018 -0.001 0.153 0.051 0.022 -0.035 0.136
SD-FS 0.008 0.000 0.053 -0.008 0.054
-0.001 0.017 -0.021 0.020 -0.033 0.137 -0.041 0.025 0.046 0.154
WEEK -0.015“ -0.015“ 0.036 0.010 0.017
-0.027 -0.005 -0.025 -0.004 -0.025 0.095 -0.010 0.033 -0.055 0.094
WK-IS -0.009 -0.020 0.000 0.004 -0.025
-0.018 0.001 -0.040 0.002 -0.063 0.067 «0.020 0.026 -0.104 0.045
WK-FS -0.003 -0.01 I -0.033 -0.013 «0.060
-0.013 0.006 -0.024 0.001 -0.118 0.056 -0.036 0.009 -0.150 0.037
INITIAL 0.000 -0.005 0.034 -0.014 0.016
-0.002 0.002 -0.029 0.016 -0.044 0.115 -0.045 0.023 -0.073 0.106
IS—FS 0.004 0.004 0.018 -0.022 0.003
-0.000 0.008 -0.018 0.024 -0.079 0.119 -0.057 0.013 -0.107 0.121
FALL 0.006 0.007 0.015 -0.005 0.023
-0003 0015 -0.011 0.024 -0.092 0.120 -0.037 0.026 -0.095 0.147

 

Table 11 (cont’d).

108

 

B. NON-m GRADIENT

 

 

 

TRAIT EPISODE
0 1 2 3 TOTAL
.NA LINGS 1992-3
WEEK 0.009“ 0.049“ 0.638“ 0.035 0.732“
0.002 0.016 0.023 0.080 0.182 1.432 -0.198 0.241 0.199 1.603
WK-IS 0.000 0.006 «0.695“ -0.023 -0.712“
0.000 0.000 -0025 0.034 -l.395 -0.233 -0.213 0.148 4.457 -0.230
WK-FS 0.000 0.007“ 1.206“ 0.074 1.286“
0.000 0.000 0.002 0.014 0.482 2.347 «0.204 0.330 0.490 2.439
INITIAL 0.000 -0.045“ 0.435“ 0.022 0.413“
0.000 0.000 -0.066 -0 024 0.087 1.002 -0.073 0.156 0.075 0.997
IS-FS 0.000 0.001 4”!“ -0.068 -0.868“
0.000 0.000 «0.004 0005 -1.662 -0.213 -0.275 0.083 -1.756 -0.292
FALL 0.000 0.001 1.259“ 0.135 1.395“
0.000 0.000 0.000 0.003 0.480 2.589 -0.114 0.412 0.541 2.788
2. NA! 93$ SEEDLINGS 19934
WEEK -0.038“ 0.01 1 0.022 0.000 -0.006
-0.052 -0.024 -0.013 0.034 -0.021 0.068 «0.014 0.014 -0.064 0.047
WK-IS 0.000 0.005 -0.000 0.029“ 0.033
0.000 0.000 -0.021 0.027 -0.050 0.058 0.004 0.051 -0.026 0.097
WK-FS 0.000 -0.000 0.046 -0.026 0.021
0.000 0.000 -0.001 0.000 -0.015 0.104 -0.051 «0.001 -0.050 0.080
INITIAL 0.000 -0.043“ 0.005 0.003 -0.035
0.000 0.000 «0.062 -0.025 -0.034 0.044 -0.019 0.028 -0.083 0.017
IS-FS 0.000 -0.000 0.045 0.012 0.058
0.000 0.000 -0.001 0.001 -0.012 0.111 -0.022 0.046 «0.008 0.138
FALL 0.000 0.000 -0.050“ -0.009 41.059“
0.000 0.000 -0.000 0.000 -0.093 -0.010 -0.032 0.018 -0.110 -0.010
W
WEEK -0.041“ -0.024“ -0.023 -0.008 -0.095“
-0.070 0013 -0.047 -0.000 -0.065 0.017 -0.026 0.011 -0.146 -0.045
WK-IS 0.000 0.042“ -0.026 0.015 0.030
0.000 0.000 0.009 0.078 -0.098 0.040 -0.019 0.047 -0.053 0.110
WK-FS 0.000 0.000 0.027 0.002 0.029
0.000 0.000 -0.000 0.001 -0.036 0.083 -0.033 0.033 -0.046 0.106
INITIAL 0.000 -0.004 -0.052 0.019 -0.037
0.000 0.000 -0.029 0.019 -0.117 0.022 -0.027 0.069 -0.118 0.060
IS-F S 0.000 -0.001 0.070 -0.011 0.058
0.000 0.000 «0.002 0.000 -0.033 0.157 -0.060 0.042 -0.063 0.159
FALL 0.000 -0.000 -0.039 0.005 -0.034
0.000 0.000 -0.000 0.000 -0.093 0.017 -0.031 0.041 -0.100 0.034
4. PLANTED SEEDLINGS 1992-3 WITH SEED WEIGHT
SEED -0.000 -0.018 -0.020 -0.001 -0.039
-0.027 0.025 -0.048 0.013 -0.069 0.029 -0.024 0.025 -0.106 0.028
SD-WK 0.038“ 0.030 -0.01 3 -0.016 0.039
0.006 0.068 -0.005 0.062 -0.064 0.038 -0.044 0.009 «0.034 0.105
SD-IS 0.000 0.005 0.009 -0.004 0.010
0.000 0.000 -0.037 0.044 -0.067 0.101 -0.043 0.036 -0.084 0.107
SD-FS 0.000 -0.000 -0.018 0.005 -0.014
0.000 0.000 -0.001 0.001 -0.093 0.053 -0.033 0.041 -0.091 0.066
WEEK -0.042“ -0.020 -0.01 7 -0.007 -0.086“
-0.070 -0.015 «0.046 0.005 -0.066 0.029 -0.026 0.012 «0.145 -0.027
WK-IS 0.000 0.029 0027 0.018 0.020
0.000 0.000 -0.007 0.065 -0.099 0.044 -0.022 0.058 -0.064 0.110
WK-FS 0.000 0.000 0.044 0.010 0.054
0.000 0.000 -0.001 0.001 -0.026 0.117 -0.033 0.044 -0.041 0.138
INITIAL 0.000 -0.003 -0.058 0.021 -0.040
0.000 0.000 -0.033 0.029 -0.129 0.023 -0.035 0.080 -0.132 0.068
IS-FS 0.000 -0.000 0.072 -0.009 0.062
0.000 0.000 -0.001 0.001 -0.054 0.171 -0.061 0.038 0.077 0.181
FALL 0.000 -0.000 -0.019 0.001 «0.018
0.000 0.000 -0.000 0000 -0.079 0.045 -0.028 0.035 -0.087 0.058

 

Table 11 (cont’d).

109

 

B. NON-m GRADIENT

 

 

 

 

TRAIT EPISODE
0 1 2 3 TOTAL
5. PLANTED SEEDLINGS 1993-4
WEEK -0.012“ ~0.011“ -0.020 ‘0.006 41.049“
-0.021 -0.002 -0.018 ‘0.002 -0.065 0.020 -0.015 0.003 -0.094 -0.004
WK-IS 0.000 -0.016 0.037 0.000 0.022
0.000 0.000 -0.035 0.006 -0.048 0.127 -0.028 0.029 -0.069 0.123
WK-FS 0.000 0.001 -0.066 -0.012 -0.077
0.000 0.000 0.000 0.001 -0.182 0.042 -0.044 0.015 -0.203 0.039
INITIAL 0.000 -0.013 0.015 0.022 0.024
0.000 0.000 -0.030 0.005 -0.060 0.084 -0.014 0.058 -0.059 0.114
IS-FS 0.000 0.001 0.048 -0.039 0.010
0.000 0.000 -0.000 0.002 -0.066 0.178 -0.084 0.011 -0.127 0.149
FALL 0.000 -0.000 -0.l05“ 0.016 -0.089
0.000 0.000 -0.000 -0.000 -0.l88 -0.030 -0.019 0.054 -0.175 0.000
6. PLANTED SEEDLINGS 1993-4 WITH SEED WEIGHT
SEED 0.005 -0.020 0.003 -0.003 -0.015
-0.004 0.015 -0.040 -0.000 -0.056 0.064 «0.021 0.016 -0.082 0.048
SD-WK -0.018 -0.019 0.040 0.004 0.007
-0.037 0.003 «0.046 0.010 -0.035 0.117 -0.024 0.031 .0073 0.084
SD-IS 0.000 0.001 0.052 -0.009 0.043
0.000 0.000 -0.034 0.036 -0.044 0.147 -0.053 0.035 -0.060 0.150
SD-FS 0.000 0.002 0.070 0.012 0.083
0.000 0.000 -0.001 0.004 -0.042 0.167 -0.030 0.051 -0.036 0.186
WEEK -0.009 -0.007 -0.030 «0.005 -0.051“
-0.020 0.001 -0.016 0.002 -0.072 0.010 -0.017 0.006 -0.098 -0.008
WK-IS 0.000 -0.009 0.006 0.004 0.001
0.000 0.000 -0.031 0.015 -0.085 0.095 -0.022 0.032 -0.106 0.100
WK-FS 0.000 0.001 -O.103 -0.017 -0.119
0.000 0.000 -0.000 0.001 -0.224 0.015 «0.050 0.015 -0.243 0.001
INITIAL 0.000 -0.001 -0.002 0.026 0.022
0.000 0.000 -0.027 0.023 0076 0.082 -0.011 0.063 -0.070 0.120
IS-FS 0.000 0.001 -0.005 «0.035 -0.040
0.000 0.000 -0.001 0.003 -0.126 0.106 «0.080 0.010 0182 0.089
FALL 0.000 -0.000 -0.129“ 0.012 -0.117“
0.000 0.000 -0.000 0.000 -0.210 -0.043 -0.021 0.045 -0.201 -0.021

 

l 10
in 1992 (Table l 1). There were no significant total non-linear selection difl‘erentials in

1993.

The pattern of direct selection in the non-linear gradient difl‘ered from the non-
linear difl‘erential. In 1992, the non-linear gradient showed an increased variance in week,
initial size, and fill size and increased covariance between week and fill size and decreased
covariance between week and initial size and initial size and fill size in the natural
seedlings (Figure 12D) and a decreased the variance in week in the planted seedlings
(Figure 14D). In 1993 selection acted directly to decrease the variance in fill size in the
natural seedlings (Figure 13D) and decrease the variance in week in the planted seedlings
(Figure 15D).

The difference between the non-linear selection differential and gradient illustrates
that correlated traits also produce changes in variances and covariances. In particular in
the natural seedlings in 1992, indirect selection nurst decrease the variance in all three
traits and increase the covariances between traits (compare Figure 12 C to D).

Four Trai Model with Seed Weigm

The inclusion of seed weight in the model showed that selection also altered the
variance in seed weight and its covariance with other traits in 1992 (Figure 17 CD), but
not in 1993 (Figure 18 CD). In 1992 the non-linear difi‘erential indicated a decreased
variance in emergence week and an increased covariance between seed weight and
emergence week and between emergence week and fill size, while the non-linear gradient
displayed a decreased variance in emergence week (Figure 17 CD). The number of

significant episodes of non-linear selection differentials indicates that there are numerous

l l 1
significant indirect efl‘ects. Direct non-linear selection on emergence week (Figure 17 D)

generates significant changes in the covariance between seed weight and emergence week
and between emergence week and fall size (Figure 17 C) by the negative phenotypic
covariances between these traits (Table 9).

iab' ' and f ' lecti n: ' des f lection

The magnitude and direction of selection on all four traits varied among episodes.
The selection differential showed significant positive selection for initial size and fill size“
among all episodes in all years with the one exception being the natural seedlings in 1992
that displayed negative selection for fill size in at emergence and through the fill (Figures
12-15A, Table 10). Thus, in general larger seedlings had higher survival in all viability
episodes and greater fecundity. However, selection on emergence week varied
substantially from episode to episode (Figures 12-15A). In both years selection fivored
late emergence for fill survival and early emergence for survival to spring (not significant
in all cases), while selection in the initial episode survival to establishment varied among
years, from positive in 1992 (Figure 12A) to negative in 1993 (Figures 13, 15A). Selection
on seed weight also varied among episodes: plants from heavier seeds performed best in
all episodes beyond establishment (Figures 17-18A).

Direct selection fivored larger fill size in all episodes (Figures 12-15B) with the
spring survival episode contributing most to the total selection gradient. In contrast, large
initial size was fivored for fill survival and small initial size was fivored for spring
survival, such that the total selection gradient was non-significant (Figures 12- 15B). Plants

that emerged later were fivored in all episodes beyond establishment (but significance

l 12
varies fiom year to year). However, plants that emerged earlier had the highest survival at

establishment in 1993-4. The contribution of selection in viability and fecundity episodes
to the total selection gradient on emergence week varied among years and among planted
and natural seedlings within years. Direct effects on seed weight were also variable
(Figures 17-18B). The selection gradient indicated that plants from lighter seeds survived
better at establishment and from winter to spring in 1992, while heavier seeds performed
better in other episodes (although not significantly) (Figure 17B). In 1993 plants from I
heavier seeds survived and were more fecund in all episodes except from winter to spring
(Figure 18B). Because these episodes differed in sign and magnitude, the total gradient on
seed weight was not significantly difl‘erent from zero in either year.

Indirect efl‘ects varied in magnitude among episodes, but were fiirly consistent in
sign among episodes (Figure 16). In general the total indirect selection differential was
influenced most by the effect of all three traits on spring survivorship with two exceptions
being fill size in 1992. When seed weight was included in the analysis, indirect efl‘ects
were also influenced most strongly by survivorship from the onset of winter to spring.

Non-linear selection difl‘erentials and gradients also varied among episodes
(Figures 12-15CD, Table 11). In 1992 non-linear selection in the spring survival episode
contributed most to the total non-linear selection difl‘erential and to the total non-linear
gradient in the natural seedlings Figure 12CD). While selection difl‘erartials and gradients
also varied among episodes for the planted seedlings in 1992, the episodes that
contributed most to the total values were survival to establishment and to the onset of

winter for both the three trait and four trait models (Figures l4, 17 CD). In 1993 the sign

1 13
and magnitude of selection difi'erentials and gradients varied among episodes and several

traits displayed significant episodes of non-linear selection despite the fict that the total
differentials and gradients were mostly not significant. In the two cases where the total
selection gradient was significant, survival to spring influenced the covariance between
initial size and fill size in the natural seedlings (Figure 13D). All episodes influenced the
variance in week, although only survival to establishment and to the onset of winter did so
significantly. .
Conditional vs. reconstructed selection anmsis

In order to quantify the total magnitude of selection on these traits across all
episodes, selection parameters in a given episode were weighted by the fiaction of
individuals surviving to that episode (Lynch and Arnold 1988). This weighting means that
later episodes contribute less to the total than earlier episodes, therefore, it changes the
relative magnitude of the episodes. A comparison of conditional (Figures 19-24, Tables 12
and 13) and additive parameters (Figures 12-15, 17, 18, Tables 10 and 11) shows the
decrease in the relative contribution of the spring survival and fecundity episodes to either
the total selection differential or gradient when parameters are additive.

In addition, the additive selection parameters are standardized by the original
phenotypic variance-covariance matrix to reflect the total change in the phenotypic
distributions. Because the initial variances and covariances are not observed at the start of
selection, this original phenotypic variance-covariance matrix was reconstructed based on
observed selection gradients and the assumption that selection does not act directly on
traits before they are expressed. When one weights selection parameters by this

reconstructed phenotypic variance-covariance matrix, one can account for the indirect

114

Figure 19. Conditional selection parameters on three traits through four episodes of
selection for natural seedlings in 1992. Viability episodes include survival to establishment
(solid), survival to the onset of winter (vertical), survival to spring (diagonal), and the final
episode of fecundity selection (clear). Selection parameters include A) linear selection
difl‘erential, B) linear selection gradient, C) non-linear selection gradient, and D) non-linear
selection gradient. Trait abbreviations follow Table 9. All selection parameters are in Imits
of standard deviation. Sigrificant episodes of selection are denoted by *.

115

m—Nmm Em mam ASH—Z—

SI‘XM

)IHEIM

 

Mum—>2

 

o N V 00
O C O O
I I I 1

oo o v N
d d d d
.INEIICIVHD WEINI'I‘NON

”Q‘QYNONV’.
LNHICIVHDHVHM'I

a m m m

 

mN—m 47m.— mNHm 45.57: E

 

.2 2:5

ad-
dd-
vd-
Nd-

Nd
vd
dd
wd

d
'IVLLNHHEIJJICI WEINI'I'NON

~°9~Q<t<~1°r~1<t
mmauaaamuvam1

116

.2 2&5 8:8 2382
=< .82 E 32:83 ESE 8m cage—om .«o mafia—mo Bo.“ 5:85 8%.: 085 no B20883 8828 1.33.80 .8 0.32m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m u. ...... a m m m m a w

No-
2.9%
as-“
8.9%
W
8.00
..o m
neon

No

go

no

mN—m 4.22 mam 125% 2mm? mam .35 mam 25% 2mm?

”WWW. WMMHM

30.. A All. m
V .._=_________ H ”e W V ...:E:: m_=_=::__ no W
11 \ wwom I 25m
In Noam l...- meow.
“new... ”3

mm. <n~

 

 

 

 

 

 

 

 

 

 

 

117

.2 2:2". 32E 2358
=< .82 E 3:589. wean—a 8m E2823 me 8828 Sea swag: was: ooh: .8 Began cocoa—om 325650 ._N oBmE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m ...... m m .m m m ...... m
1. «a. was m H 1 M X
#6..“ «ad-W
29m. 27m
3.“ Sam
SW SW
0 o
o m o m
3 u S w.
do w 3 m
no 0 no
mam SE mam 25:7: 2mg, mam .35 mam 259:6 2mm?
No- S.
2 1
V 2.2.2... .. WEE o m R . \_\:: $23 m
\ .. x 3 W ... \\\. H \k “No m
* -
k . we 0 \ “Yo m
i r
0.0 m .. Mod n
* 4 a H .
.3 w. .3
m. _ <. _

118

.2 25mm 26:8 2852
5.. .89 E $2603 3283 8m 3:028 we 89.2% Sam @385 ER: 8:: co abs—58m c2628 WEE—:80 .NN Semi

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m m. m .m m m m .0 m m m m
1. fi M mm fl 1 mm W van .fi m
m.o-N m.o-N
. m . w
3.1 No-1
3.“ 27%
w w.
m m
S a S a
n w.
No W do w
o no 0 3
gm 2.2.2 mam .2522 2mm? mam SE mam .255 2mm?
. Mao- ... -wo-m
§ .~.o-1 wmd-
* t H . 9 .w \ *_I-.l|_ . . fl
Ill \\ “v0 m rl “v0 ..a
. * a
ll .85 TI .90 N.
... a .
I We m .22

 

 

 

 

 

 

 

 

 

 

119

.2 2:22 32S 2352
=< .82 a $3608 vows—a Hem 958—8 No 8.328 Bed @585 3B: Sou co mbgsa cocoa—om $53650 .mN oSwE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

S M S M
Mammmmmmflm msmmmmmmflm
1 H .03. m X H. m m m 1. «a. w m H M. «VI. m m
S- vS-N
N o
S-m S-m
Sm SE
SW SW
o D o m
S m S n
S M S m
o S o S
2.2... 2527: 2mm? :55? .3: 25:7: 0555 EBB:
-N o- .S- 1
E * :_:_ “fi-fic "I. V i::_ *E *\ filo m
l\ .. \N 22m V ...\\\ .. \ flS W
*\ ... W 1m- \\\ 8 ~ ... H G
11 \\\ «S 9 I... ....o a
l- 3. m 90 M
. . a 2 .
.l w o m w o
m2 _ <2 _ m

 

 

 

 

 

 

 

 

 

 

 

 

120

 

 

.2 2:5 25:8 £3;
=< .32 E “Ea—~68 385a 8m 8528 mo caveman 50m @585 3a: 50m co Egan a2828 1.53.650 .vm oSwE

 

 

S M S M
mmmmmmmmflm mamm.mmmnm
1 w % .9. X w m m m 1 M M m X H S m m
S- S-N
. N . m
SAM 8-..:
Sm S-“
S- S-w
w a
o u
Sm Sa
Sm Sm
. m _
S S
35— 453,: Mmma ESE? 3: SEE mmma $553
On Olnl.
.o-1 .o-m
o m o w
.o W .9 a
. m . a
o m o n
.o w. ‘o m
_ _

121

Table 12. Conditional linear selection difl‘erentials (A) and gradients (B) for natural and
planted seedlings in 1992 and 1993 for three trait or four trait models. Values are given
for each of four selection episodes. Below each value are the 95% confidence intervals
based on bootstrap resampling (n=500). Linear parameters reflect the change in the mean
in units of standard deviation. Values significant at p<0.05 are bold.

 

 

 

 

 

A LINEAR DIFFERENTIAL
TRAIT EPISODE
0 1 2 3
1. NATURAL SEEDLINGS 1992-3
WEEK 0.022“ 0.199“ 4.339“ -0.036
0.011 0.032 0.166 0.230 -0.422 -0.259 -0.142 0.068
INITIAL 0.000 0.158“ 0.199“ 0.114
0.000 0.000 0.121 0.199 0.118 0.294 -0.026 0.258
FALL 0.000 0.000 0.474“ 0.249“
0.000 0.000 0.000 0.000 0.403 0.548 0.127 0.361
2. NATURAL S_EEDLINGS 1993-4
WEEK -0.034“ 0.116“ -0.007 -0.082"
-0.049 -0.019 0.089 0.147 -0.071 0.052 -0. 160 -0.014
DII'I'IAL 0.000 0.107“ 0.162“ 0.206“
0.000 0.000 0.079 0.139 0.102 0.221 0.116 0.302
FALL 0.000 0.000 0.312“ 0.309“
0.000 0.000 0.000 0.000 0.243 0.372 0.236 0.393
3. PLANTED SEEDLINGS 1992-3
WEEK -0.005 0.058“ -0.068 0.004
-0.039 0.030 0.022 0.095 0147 0.010 -0.089 0.089
INITIAL 0.000 0.043“ 0.187“ 0.117“
0.000 0.000 0.003 0.081 0.102 0.268 0.004 0.229
FALL 0.000 0.000 0.363“ 0.257“
0.000 0.000 0.000 0.000 0.293 0.434 0.165 0.354
4. PLANTED SEEDLINGS 1992-3 WITH SEED WEIGHT
SEED -0.045“ 0.041“ 0.115“ 0.204“
-0.077 -0.012 0.003 0.078 0.031 0.194 0.113 0.300
WEEK -0.005 0.058“ -0.068 0.004
«0.039 0.031 0.022 0.098 -0.145 0.015 -0.081 0.093
INITIAL 0.000 0.043“ 0.187“ 0.117“
0.000 0.000 0.003 0.085 0.100 0.272 0.006 0.232
FALL 0.000 0.000 0.363“ 0.257“
0.000 0.000 0.000 0.000 0.296 0.436 0.167 0.358
WEEK -0.060“ 0.094“ -0.090“ -0.114“
«0.079 -0.044 0.069 0.119 -0.l61 -0.010 -0.213 -0.027
INITIAL 0.000 0.051“ 0.153“ 0.141“
0.000 0.000 0.024 0.078 0.064 0.238 0.045 0.240
FALL 0.000 0.000 0.382“ 0.257“
0.000 0.000 0.000 0.000 0.309 0.455 0.169 0.345
6. ELM 1 Q SEEDLINGS 1993-4 WITH SEED WEIGHT
SEED -0.000 0.092“ 0.107“ 0.186“
-0.013 0.014 0.065 0.120 0.024 0.192 0.087 0.287
WEEK -0.060“ 0.094“ -0.090“ -0.114“
-0.081 -0.042 0.068 0.119 -0.167 -0.012 -0.205 -0.020
INITIAL 0.000 0.051“ 0.153“ 0.141“
0.000 0.000 0.023 0.076 0.061 0.237 0.029 0.230
FALL 0.000 0.000 0.382“ 0.257“
0.000 0.000 0.000 0.000 0.308 0.462 0.177 0.339

 

Table 12 (cont’d).

122

 

B. LINEAR GRADIENT

 

 

 

 

 

 

TRAlT EPISODE
0 1 2 3
1. NA SEEDLINGS 1992-3
WEEK 0.022- 0.1771 0.026 0.158
0.011 0.032 0.147 0.207 0.070 0.131 0.006 0.331
mm 0.000 0.142- 0.042 0.079
0.000 0.000 0.109 0.181 0.179 0.095 0.247 0.099
FALL 0.000 0.000 0.513— 0.462“
0.000 0.000 0.000 0.000 0.361 0.669 0.229 0.695
2. NATURAL §_EEDLINGS 1993-4
WEEK 0.030 0.127- 0.168“ 0.061
0.049 0.019 0.100 0.159 0.103 0.240 0.023 0.142
INITIAL 0.000 0.109' 0.079 0.004
0.000 0.000 0.082 0.140 0.158 0.007 0.111 0.117
FALL 0.000 0.000 0.424- 0.386“
0.000 0.000 0.000 0.000 0.336 0.514 0.261 0.512
3. PIANI'EQSfEEDlmGS 1992-3
WEEK 0.005 0.067“ 0.086 0.196-
0.039 0.030 0.028 0.107 0.000 0.176 0.068 0.315
mm 0.000 0.050- 0.039 0.069
0.000 0.000 0.009 0.088 0.156 0.062 0.211 0.074
FALL 0.000 0.000 0.419 0.468-
0.000 0.000 0.000 0.000 0.306 0.523 0.312 0.630
4. PLANTED SEEDLINGS 1992-3 W1TH SEED WEIGHT
SEED 0.046~ 0.031 0.111- 0.030
0.079 0.013 0.011 0.074 0.212 0.012 0.099 0.172
WEEK 0.010 0.068“ 0.100- 0.188-
0.045 0.027 0.030 0.108 0.010 0.201 0.057 0.321
INITIAL 0.000 0.038 0.027 0.071
0.000 0.000 0.005 0.083 0.143 0.083 0.221 0.067
FALL 0.000 0.000 0.47m 0.447.
0.000 0.000 0.000 0.000 0.350 0.593 0.265 0.644
5. PLANTED S_EEDLINGS 1993-4
WEEK -0.060“ 0.1004 0.124- 0.020
0.079 0.044 0.073 0.127 0.030 0.219 0.132 0.100
INITIAL 0.000 0.047- 0.141- 0.01 1
0.000 0.000 0.019 0.072 0.252 0.032 0.150 0.125
FALL 0.000 0.000 0.511- 0.356“
0.000 0.000 0.000 0.000 0.409 0.624 0.199 0.539
6. PLANTED SEEDLINGS 1993-4 wrrH SEED WEIGHT
SEED 0.010 0.073- 0.131- 0.053
0.004 0.023 0.044 0.103 0.232 0.026 0.069 0.190
WEEK -0.062“ 0.090— 0.167“ 0.042
0.083 0.043 0.064 0.118 0.068 0.268 0.149 0.082
INITIAL 0.000 0.012 0.121 0.017
0.000 0.000 0.015 0.040 0.244 0.002 0.138 0.109
FALL 0.000 0.000 0.578“ 0.317-
0000 0.000 0.000 0.000 0.454 0.700 0.150 0.480

 

123

Table 13. Conditional non-linear selection differentials (A) and gradients (B) for natural
and planted seedlings in 1992 and 1993 for three trait or four trait models. Values are
given for each of four selection episodes. Below each value are the 95% confidence
intervals based on bootstrap resampling (n=500). Non-linear parameters reflect the change
in the variance in units standard deviation squared. Values significant at p<0.05 are bold.

 

 

 

 

 

A. NON-LINEAR DIFFERENTIAL
TRAIT EPISODE
0 1 2 3
1. NATURAL SEEDLINGS 1992-3
WEEK 0.052“ 0.340“ 41.639“ 0023
0.037 0.065 0.288 0.389 -O.811 -0.458 -0.197 0.179
WK-IS 0.000 0.004 0.176“ 0.053
0.000 0.000 -0.038 0.050 0.046 0.319 -0.209 0.393
WK-FS 0.000 0.000 0.390“ -0.059
0.000 0.000 0.000 0.000 0.293 0.482 -0.220 0.112
DIITIAL 0.000 -0.016 0.209 0.157
0.000 0.000 -0.084 0.051 -0.014 0.470 -0.351 0.742
IS-F S 0.000 0.000 0.026 0.122
0.000 0.000 0.000 0.000 0.087 0.154 -0.113 0.402
FALL 0.000 0.000 -0.084 0.354“
0.000 0.000 0.000 0.000 «0.194 0.044 0.140 0.589
2. NATURAL SEEDLINGS 1993-4
WEEK -0.070“ 0.022 -0.071 -0.035
-0.101 -0.041 -0.011 0.051 -0.139 0.004 -0.107 0.037
WK-IS 0.000 -0.006 0.068“ 0.088
0.000 0.000 -0 037 0.022 0.002 0.138 -0.017 0.207
WK-FS 0.000 0.000 0.069“ -0.008
0.000 0.000 0.000 0.000 0.006 0.133 -0.095 0.078
INITIAL 0.000 -0.069“ 0.103“ 0.174
0.000 0.000 -0.113 -0.025 0.009 0.199 -0.044 0.418
IS-FS 0.000 0.000 0.036 0.1 18
0.000 0.000 0.000 0.000 -0.045 0.112 -0.061 0.304
FALL 0.000 0.000 -0.051 0.185“
0.000 0.000 0.000 0.000 0137 0.037 0.005 0.394

Table 13(cont’d).

124

 

 

 

 

A. NON-LINEAR DIFFERENTIAL
TRAIT EPISODE
0 1 2 3
3. P GS 1992-3
WEEK -0.060“ -0.044 -0.196“ -0.071
-0.107 -0.018 -0.087 0.002 -0.292 -0.105 -0. 164 0.022
WK-IS 0.000 0.036 -0.012 0.043
0.000 0.000 -0.008 0.078 -0.088 0.054 ‘0 042 0.131
WK-FS 0.000 0.000 0.054 -0.006
0.000 0.000 0.000 0.000 -0.020 0.126 -0.104 0.074
INITIAL 0.000 -0.014 -0.041 0.145
0.000 0.000 -0.075 0.047 -0. 160 0.075 -0.077 0.365
IS-FS 0.000 0.000 -0.058 0.129
0.000 0.000 0.000 0.000 -0.141 0.042 -0.057 0.337
FALL 0.000 0.000 -0.132“ 0.265“
0.000 0.000 0.000 0.000 -0.232 -0.019 0.072 0.459
4. PLANTED SEEDLINGS 1992-3 WITH SEED WEIGHT
SEED -0.012 -0.048 0.086 0.076
-0.056 0.030 -0.102 0.001 -0. 188 0.015 -0.044 0.202
SD-WK 0.043“ 0.043 0.008 -0.032
0.009 0.080 -0.001 0.085 -0.077 0.085 -0.108 0.046
SD-IS 0.000 -0.024 «0.022 0.079
0.000 0.000 -0.062 0.013 -0.105 0.064 -0.048 0.202
SD-FS 0.000 0.000 0.005 0.165“
0.000 0.000 0.000 0.000 -0.082 0.093 0.038 0.301
WEEK -0.060“ -0.044 —0.196“ -0.071
-0.106 -0.019 -0.096 0.005 -0.296 -0.107 -0. 160 0.019
WK-IS 0.000 0.036 -0.012 0.043
0.000 0.000 -0.008 0.082 «0.087 0.053 -0.041 0.136
WK-F S 0.000 0.000 0.054 -0.006
0.000 0.000 0.000 0.000 -0.021 0.128 -0.100 0.084
INITIAL 0.000 -0.014 -0.041 0.145
0.000 0.000 -0.070 0.043 -0.154 0.076 -0.045 0.361
IS-FS 0.000 0.000 -0.058 0.129
0.000 0.000 0.000 0.000 -0.154 0.043 -0.063 0.335
FALL 0.000 0.000 —0.132“ 0.265“
0.000 0.000 0.000 0.000 -0.234 -0.024 0.074 0.469

 

Table 13 (cont’d).

125

 

 

 

 

 

ANON-LINEAR DIFFERENTIAL
TRAIT EPISODE
0 1 2 3
W
WEEK 0.077- 0.017 0.156- 0.016
0.120 0.038 0.055 0.018 0.275 0.016 0.120 0.169
WK-IS 0.000 0.024 0.113. 0.003
0.000 0.000 0.050 0.002 0.032 0.199 0.117 0.110
WK-Fs 0.000 0.000 0.0954- 0.1 12
0.000 0.000 0.000 0.000 0.008 0.167 0.277 0.020
INITIAL 0.000 0.003 0.067 0.112
0.000 0.000 0.041 0.034 0.060 0.199 0.042 0.289
IS-FS 0.000 0.000 0.055 0.047
0.000 0.000 0.000 0.000 0.150 0.040 0.100 0.193
FALL 0.000 0.000 0.133- 0.206“
0.000 0.000 0.000 0.000 0.230 0.025 0.057 0.376
6. PLANTED S_EEDLINGS 1993-4 WITLSEED WEIGHT
SEED 0.007 0.025 0.076 0.047
0.013 0.026 0.062 0.013 0.074 0.226 0.136 0.238
SD-WK 0.014 0.039- 0.022 0.058
0.033 0.003 0.065 0.015 0.049 0.094 -O.148 0.036
SD-IS 0.000 0.019 0.088 0.039
0.000 0.000 0.051 0.011 0.009 0.183 0.095 0.182
SD-Fs 0.000 0.000 0.069 0.107
0.000 0.000 0.000 0.000 0.026 0.153 0.022 0.241
WEEK 0.077- 0.017 0156“ 0.016
0.130 0.037 -0.058 0.018 0.279 0.035 0.129 0.179
WKIS 0.000 0.024 0.1134- 0.003
0.000 0.000 0.049 0.002 0.031 0.198 0.119 0.111
WK-Fs 0.000 0.000 0.0951- 0.112
0.000 0.000 0.000 0.000 0.020 0.177 0.266 0.022
INITIAL 0.000 0.003 0.067 0.112
0.000 0.000 0.040 0.030 0.057 0.202 0.057 0.296
IS-FS 0.000 0.000 0.055 0.047
0.000 0.000 0.000 0.000 0.144 0.034 0.084 0.175
FALL 0.000 0.000 0.133- 0.206-
0.000 0.000 0.000 0.000 0.238 0.034 0.059 0.365

 

Table 13 (cont’d).

126

 

B. NON-LINEAR GRADIENT

 

 

 

TRAIT EPISODE
0 1 2 3

1. NA! QR_A_L SEEDLINGS 1992-3
WEEK 0.009“ 0.039“ 0.195“ «0.091

0.002 0.016 0.019 0.063 0.076 0.331 -0.312 0.113
WK-IS 0.000 0.009 -0.367“ 0.1 12

0.000 0.000 -0.018 0.036 -0.604 -0. 136 -0.234 0.463
WK-FS 0.000 0.000 0.639“ 0.061

0.000 0.000 0.000 0.000 0.326 0.982 -O.398 0.518
INITIAL 0.000 -0.048“ 0.067 -0.062

0.000 0.000 -0.069 -0.029 -0.072 0.188 -0.211 0.115
IS-F S 0.000 0.000 -0.233 -0. 125

0.000 0.000 0.000 0.000 -0.462 0.042 -0.502 0.1 13
FALL 0.000 0.000 0. 193 0.260

0.000 0.000 0.000 0.000 0.034 0.367 0.042 0.480
2. NATURAL SEEDLINGS 1993-4
WEEK -0.038“ 0.014 0.039 -0.012

-0.052 —0.024 «0.016 0.044 -0.038 0.129 «O. 100 0.080
WK-IS 0.000 0.005 0.002 0.160“

0.000 0.000 -0.025 0.033 -0.095 0.114 0.026 0.286
WK-F S 0.000 0.000 0.086 -0.159

0.000 0.000 0.000 0.000 -0.023 0.189 -0.325 0.001
INITIAL 0.000 -0.048“ 0.017 -0.01 1

0.000 0.000 -0.069 -0.028 -0.056 0.086 -0.112 0.101
IS-FS 0.000 0.000 0.083 0.089

0.000 0.000 0.000 0.000 -0.026 0.206 -0.11 1 0.288
FALL 0.000 0.000 -0.089“ -0.062

0.000 0.000 0.000 0.000 -0.165 -0.019 -0.223 0.117

 

Table 13 (cont’d).

127

 

B. NON-LINEAR GRADIENT

 

 

 

 

TRAIT EPISODE
0 1 2 3

3. LINGS 1992-3
WEEK -0.041“ 41.035“ -0.050 -0.047

-0.070 -0.013 -0.067 -0.001 -0.131 0.025 -0.187 0.075
WK-IS 0.000 0.056“ -0.051 0.081

0.000 0.000 0.012 0.103 -0.175 0.069 «0. 101 0.253
WK-FS 0.000 0.000 0.049 0.033

0.000 0.000 0.000 0.000 0.066 0.151 -0.213 0.245
INITIAL 0.000 -0.006 -0.078 0.069

0.000 0.000 -0.037 0.024 -0. 183 0.038 -0.101 0.234
IS-FS 0.000 0.000 0.1 1 1 -0.037

0.000 0.000 0.000 0.000 -0.054 0.256 -0.268 0.180
FALL 0.000 0.000 -0.061 0.012

0.000 0.000 0.000 0.000 -0.146 0.024 -0.172 0.197
4. PLANTED S_EEDLINGS 1992-3 WITH SEED WEIGHT
SEED -0.000 -0.026 -0.036 0.012

-0.027 0.025 -0.066 0.013 -0.121 0.048 -0.127 0.163
SD-WK 0.038“ 0.043 -0.013 -0.122

0.006 0.068 -0.005 0.086 -0.105 0.077 -0.316 0.052
SD-IS 0.000 0.004 0.023 -0.028

0.000 0.000 -0 048 0.055 -0.107 0.179 -0.208 0.161
SD-FS 0.000 0.000 -0.036 0.017

0.000 0.000 0.000 0.000 -0. 166 0.088 -0.208 0.250
WEEK -0.042“ -0.034 -0.042 -0.007

-0.070 -0.015 -0 070 0 003 -0.124 0.039 -0. 155 0.138
WK-IS 0.000 0.038 0057 0.101

0.000 0.000 -0.010 0.086 -0.177 0.068 -0.096 0.311
WK-FS 0.000 0.000 0.080 0.1 16

0.000 0.000 0.000 0.000 -0.047 0.205 -0.231 0.394
INITIAL 0.000 -0.004 -0.090 0.076

0.000 0.000 -0 043 0.036 -0.210 0.032 -0.1 14 0.278
IS«FS 0.000 0.000 0.117 -0.022

0.000 0.000 0.000 0.000 -0.091 0.282 -0.277 0.236
FALL 0.000 0.000 -0.030 -0.017

0.000 0.000 0.000 0.000 -0.119 0.066 -0.239 0.212

 

Table 13 (cont’d).

128

 

B. NON-LINEAR GRADIENT

 

 

 

TRAIT EPISODE
0 1 2 3
5. M122 SEEDLINGS 1993-4
WEEK ~0.012“ ‘0.014“ -0.031 -0.056
-0.021 -0.002 -0.024 -0.002 «0.106 0.035 -0.162 0.026
WK-IS 0.000 -0.019 0.059 0.011
0.000 0.000 0.041 0.007 -0.074 0.191 -0.150 0.194
WK-FS 0.000 0.000 -0.101 -0.110
0.000 0.000 0.000 0.000 -0.271 0.063 -0.381 0.123
INITIAL 0.000 -0.013 0.021 0.087
0.000 0.000 0.032 0.007 -0.078 0.117 -0.048 0.215
IS.FS 0.000 0.000 0.069 -0.181
0.000 0.000 0.000 0.000 -0.091 0.249 -0.444 0.094
FALL 0.000 0.000 -0.152“ 0.005
0.000 0.000 0.000 0.000 -0.269 0.040 -0.205 0.219
6. PLANTED SEEDLINGS 1993-4 WITH SEED WEIGHT
SEED 0.005 -0.020 0.009 -0.040
«0.004 0.015 -0.044 0.001 -0.073 0.094 -0.159 0.073
SD-WK -0.018 -0.024 0.069 0.058
-0.037 0.003 -0.056 0.013 -0.044 0.179 -0. 150 0.270
SD-IS 0.000 0.001 0.071 -0.015
0.000 0.000 -0.036 0.039 -0.059 0.203 -0.200 0.159
SDFS 0.000 0.000 0.100 0.099
0.000 0.000 0.000 0.000 -0.061 0.240 -0.192 0.387
WEEK -0.009 -0.010 -0.050 -0.074
-0.020 0.001 -0.022 0.002 -0.119 0.017 -0.213 0.048
WK-IS 0.000 -0.011 0.016 0.012
0.000 0.000 -0.038 0.018 -0.127 0.153 -0.154 0.179
WK-FS 0.000 0.000 -0.163 -0.172
0.000 0.000 0.000 0.000 -0.344 0.023 -0.456 0.119
INITIAL 0.000 -0.001 -0.001 0.094
0.000 0.000 -0.029 0.025 -0.103 0.1 13 -0.043 0.224
IS-FS 0.000 0.000 -0.003 -0.175
0.000 0.000 0.000 0.000 -0.175 0.159 -0.430 0.103
FALL 0.000 0.000 ~0.l98“ -0.047
0.000 0.000 0.000 0.000 -0.322 -0.070 -0.284 0.206

 

129
effects of selection on traits in episodes prior to their expression. Accormting for these

indirect efl‘ects means that the total additive linear and non-linear selection differentials
include selection on the traits before they are expressed. In this study, linear selection
differentials for fall size are significant for Stu-viva] to establishment and to the onset of
winter episodes (1 e. Figures 12, 13, 15A) as a result of its correlation with emergence
week and initial size. Non-linear difl‘erentials on fall size are also significant in episodes
prior to the onset of winter (i.e. Figures 12, 13, 14C). Thus, both the mean and variance in
fall size can change due to selection prior to the actual manifestation of the trait.
Furthermore, the covariance between fall size and other traits can also change due to
selection prior to the onset of winter on phenotypically correlated traits, i.e. covariance
between week and fall size in the first two episodes (Figures 12- 13C).

This reconstruction makes the assumption that selection does not act directly on
traits before they are manifested. Thus, reconstruction should not alter the linear and non-
linear selection gradient (compare Figures 12-15, 17, 18 with Figures 19-24). However,
when gradients are weighted by the original phenotypic variance-covariance matrix, small
(<0.03 standard deviation units), but statistically significant selection gradients occur for
fall size in the first episode in three of the four cases (Figures 12, 13, 15, Table 10),
probably a result of compounding errors in reconstruction.

Thus, making the selection parameters additive changes the relative magnitude of
selection in later episodes and sign and magnitude of the linear and non-linear selection
difl‘erentials in several episodes by accounting for unobserved changes in the phenotypic

variance-covariance matrix. The consequence is that both differentials and gradients were

130
of smaller magnitude because later episodes contributed less to the total Second, the

difl‘erentials now reflected indirect selection in early episodes that was not accounted for
by conditional difl‘erentials.

DISCUSSION
Changes in Trait Means

In this two year study of phenotypic selection on four juvenile traits, I found that
the direction of linear selection was consistent across years. The total selection difl‘erential,
the sum of direct selection and selection on phenotypically correlated traits, indicated that
plants with heavier seeds, larger initial size, and larger fall size had higher fitness, while
emergence week was unrelated to fitness. When the indirect efiects of correlated traits are
removed, the direct efl‘ects are quantified by the selection gradient which indicated that
plants that emerged earlier and were larger at the onset of winter had higher fitness, while
seed weight and initial size did not directly affect fitness. The magnitude of these changes
varied among traits and years. The magnitude of the predicted change in the mean due to
direct selection ranged fiom 0.1 to 0.47 standard deviations. The largest of these predicted
changes was for fall size where a 0.47 standard deviation shifi is equivalent to a 1.05 mm
change in mean fall cotyledon diameter.

If one assumed that fall size was inherited in a Mendelian fashion, then one could
predict the selection response according to the equation, R= hzs. With a selection
difl‘erential of 0.24 on fall size in the natural seedlings in 1992, the predicted change in the
mean would be fi'om 8.27 to 9.32 mm in diameter. Selection of this magnitude could

generate a substantial change in the trait fi'om one generation to the next. In a multivariate

13 1
framework, this change will be affected by phenotypic and genetic correlations with other

traits such that the response will be a function of the genetic variance-covariance matrix
and the vector of selection gradients (A E=GB). Iffall size were negatively genetically
correlated with another trait, its response might be constrained by selection on that trait.
In this study fall size was positively genetically correlated with seed weight and initial size
(Chapter 1) neither of which were directly selected in the natural population. Thus, its
response will not be influenced by these genetic correlations. However, because of these
genetic correlations seed weight and initial size should respond to direct selection on fall
size if one only considers Mendelian inheritance.
Maternal Inheritance

Response to selection will be more complicated than either the univariate or
multivariate equations would predict because three of these four traits display maternal
inheritance (Chapter 1). Seed weight, and initial and fall cotyledon diameter display
significant maternal additive variance and significant negative direct-maternal covariance,
while emergence week displays only standard Mendelian inheritance. The magnitude of the
negative direct-maternal covariance suggests that the evolution of these traits will be
strongly constrained by maternal inheritance. Direct selection may favor large fall size , but
this direct selection produces a correlated genetic response such that changes in maternal
performance will lead to smaller fall seedlings. Thus, the response to selection may be in a
direction opposite to selection depending on the integration of these maternal inheritance
parameters. Animal breeders have generally examined responses of single traits and thus,

base their expectations on a calculated realized heritability that incorporates maternal

132
genetic variances and covariances in the response. In Chapter 1 I showed that this realized

heritability for fall size is near zero in the greenhouse. Multivariate predictions about
evolutionary response require the incorporation of maternal inheritance in a specific
evolutionary model. Currently, there are two possible approaches to integrating
phenotypic selection and maternal inheritance: l) a covariance approach suggested by
Kirkpatrick and Lande (1989) that simplifies quantitative genetic parameters in one term
and 2) index selection incorporating maternal inheritance (Van Vleck 1970).

When traits display maternal inheritance the response to direct selection involves
both traits expressed at the juvenile stage in the life cycle and traits expressed in the
parental stage in the life cycle. While I have not documented phenotypic selection on
maternal traits in this population, the quantitative genetic analysis in Chapter 1
demonstrates that direct selection on maternally inherited juvenile traits will produce
evolutionary responses in the maternal phenotypic traits (unobserved) that determine the
nature of the observed maternal efl‘ects. Thus maternal inheritance results in a response to
selection for traits expressed much later in ontogeny. Direct phenotypic selection on those
maternal traits would also influence the evolutionary response of maternally influenced
juvenile traits. Thus, the genetic and phenotypic correlations between maternal and
ofl‘spring traits influence the joint evolution of these traits. One consequence of this fict is
that early-acting phenotypic selection can influence the distribution of maternal traits and
late-acting phenotypic selection on maternal traits influences the evolution of juvenile

traits. As Kirkpatrick and Lande (1989) point out, one interesting aspect of maternal

133
inheritance is that traits that are not under direct selection influence the evolutionary

response. When traits are inherited in a Mendelian fi'amework, this does not occur.
Direct Ed Indirect Efl‘ects

Equations for multivariate evolution incorporate indirect efl‘ects of selection by
incorporating the phenotypic variance-covariance matrix indicated by the correlations
among traits in Figure 8. Differences between the selection difl‘erential and the selection
gradient suggest how important these indirect effects will be for trait evolution. In this
study, I found that although larger seed size and larger initial size were favored (selection
differential), there was no direct selection on these traits (selection gradient). Therefore,
changes in these traits are due solely to the operation of direct selection on other traits. It
appears that strong positive directional selection on fill size indirectly selects for large
seed size and large initial size via the positive phenotypic correlations between these traits.
In addition, direct selection for later emergence week is countered by indirect effects
mediated via the negative phenotypic correlation between fill size and emergence week.
As a result, no change in emergence week is expected because although later emergence
directly enhances fitness, direct selection for larger fall size decrements the direct change
in emergence week via the negative phenotypic correlation.

The linear selection difl‘erentials observed in this study are consistent with
univariate studies of these juvenile traits: heavier seeds (Wulfl‘ 1986; Winn 1988), earlier
emerging seeds (Kalisz 1986; Miller 1987; van der Toom and Pons 1988; Biere 1991b),
and larger seedlings (Ross and Harper 1972) all experience enhanced fitness, although

these relationships can vary spatially and temporally (i.e. Kalisz 1986). Multivariate studies

134
that include emergence time and early seedling size also emphasize spatial and temporal

variation in selection (Kelly 1992; Stratton 1992b; but see Mitchell-Olds and Bergelson
1990b). The general pattern observed in these multivariate studies is that variation in
seedling size generates most of the variation in fitness, with larger seedlings displaying
higher survivorship and/or fecrmdity (Bennington and McGraw 1995b; Stratton 1992b;
Kelly 1992; Mitchell-Olds and Bergelson 1990b). Although emergence time directly
influences early sm'vival (Kelly 1992; Stratton 1992), most selection on emergence time
and seed weight is indirect via correlated traits expressed later in ontogeny (Bennington
and McGraw 1995b; Stratton 1992b; Mitchell-Olds and Bergelson 1990b). My study
documents the same pattern: fill size is the critical trait and once direct selection on it is
included in the model, then predictions about the changes in other traits are a function of
their correlation with fill size. Early traits like seed weight, emergence date, and initial size
influence survival in the early episodes, however, once fill size is expressed selection acts
only indirectly on seed weight and initial sin. Emergence week continues to directly
influence survivorship and in some cases fecundity even afler fill size is expressed.
Episodic Analysis

Path analysis represents an alternative approach to the episodic analysis of
selection that allows one to test alternative causal models (Crespi and Bookstein 1989;
Crespi 1990; Kingsolver and Schmeske 1991). In their multivariate study of juvenile traits
and early growth, Mitchell-Olds and Bergelson (1990b) constructed a specific path
analytic model to address the relative magnitude of direct and indirect efl‘ects of early

traits on fitness by allowing early traits to directly influence the expression of later traits as

135
fitness. In my analysis indirect efl‘ects are accounted for by the correlation structure alone.

One limitation of path analysis is the inability to account for early mortality eliminating
individuals from the analysis before they express all the relevant traits. In these cases, one
would be unable to quantify the total magnitude of selection in a path analytic framework
because individuals not expressing all traits would be excluded fi'om the analysis. Thus, for
my purposes an episodic analysis offers the most complete view of phenotypic selection in
this species with substantial mortality early in the life cycle. It also permits an examination
of selection through ontogeny and allows me to examine the nature of direct and indirect
efl‘ects in early vs. late episodes of selection, to identify if traits act directly in early
episodes, and only indirectly in later episodes.

A second advantage of the episodic approach is that it allowed me to examine the
association between selective agents observed in particular episodes relative to the total
change in the trait mean, i.e. it can suggest the relative impact of selective agents
associated with particular episodes. For instance, in this study slug herbivory was a major
source of mortality at establishment and likely to be one from establishment to the onset of
winter. An examination of selection gradient (Figures 12-15, 17, 18B) in the first two
episodes shows that slugs generally select for earlier emergence in the first episode
(significant only in 1993 ), but later emergence in the second episode. This corresponds to
observations of slug activity. Slug activity prior to leaf drop is low, but once a thick layer
of leaf litter covers the seedlings, many more shrgs were observed. Thus, seedlings
emerging early avoided slug herbivory at establishment, while seedlings that emerged afier

leaf drop were consumed by slugs as they emerged. In late fill leaf litter begins to freeze

136
and shrg activity decreases. In the second episode from establishment to the onset of

winter, earlier emerging seedlings were available for slug consumption when slugs were
most active ie. afier leaf drop and prior to cold temperatures, while later emerging
seedlings were more likely to avoid slug consumption afier establishment.

In the spring viability episode, direct selection acted primarily on fill size favoring
large individuals, but strong direct non-linear selection in this episode (Figure 12D)
suggests that while large seedlings survived best, intermediately sized seedlings were lost
from the population. Deer herbivory is one possible mechanism for this disruptive
selection if deer consume medium-sized plants more fiequently than the rare small or large
plants. This efl"ect of deer herbivory on spring survivorship could not be quantified
because herbivory was scored only on those plants that survived. However, 20.7% and
15.9% of plants surviving to spring were browsed by deer in 1992 and 1993, respectively
and browed plants produced fewer seeds than unbrowsed plants (Figure 10). Fall size was
significantly smaller in unbrowsed plants, but no more variable than the larger browsed
plants. Thus, deer did have an impact on fecundity selection that contributed significantly
to the total linear selection difl‘erentials and gradients, but not significantly to the non-
linear parameters.

Reconstruction

The reconstruction of the phenotypic variance-covariance matrix in this analysis is
a powerful technique because it allowed me to estimate the total forces of selection on
particular phenotypic traits as well as examine the indirect and direct effects of selection

through ontogeny. The larger context of the present study is the integration of maternal

137
inheritance of juvenile traits and phenotypic selection on them in order to illustrate how

maternal inheritance may affect the evolution of juvenile traits in this natural population.
Conditional estimates of selection derived directly from the multiple regression analysis
would not have allowed me to predict how direct and indirect selection on these traits
through the life cycle would generate total change in traits means and variances.

My comparison of conditional and additive estimates demonstrates that additive
estimates difl‘er in the relative contribution of later episodes to the total selection gradient.
Second, additive difl‘erentials incorporate indirect effects that were essentially unobserved
in the conditional analysis, thereby providing a more complete picture of phenotypic
selection.

Conclusions

Phenotypic selection acts on phenotypic values and produces changes in means,
variances, and covariances (Figure 8). The trans-generational response to these phenotypic
changes depends on the underlying inheritance and genetic correlations. In a separate
study, 1 demonstrated that each of these traits displays significant additive genetic
variance. Three traits, seed weight, and initial and fill cotyledon diameter display
significant maternal additive variance and significant negative direct-maternal covariance,
while the fourth trait, emergence week, displays only standard Mendelian inheritance. In
addition, seed weight is positively genetically correlated with the other three traits and
initial size and fill size are positively genetically correlated. Thus, while the genetic
correlations among traits will enhance the response to selection via indirect effects (i.e.
direct selection on fill size will generate concomitant increases in seed weight and initial

size), the negative direct-maternal genetic covariances will constrain the responses. This

138
constraint is due to direct selection on offspring traits altering the distribution of maternal

phenotypes and their subsequent impact on offspring in the next generation. The
integration of phenotypic selection and maternal inheritance requires the nnrltivariate
analysis of maternal inheritance suggested by Kirkpatrick and Lande (1989). Once that
analysis is completed, I will be able to integrate this complex interaction between maternal
inheritance and phenotypic selection and make quantitative predictions about response to
selection. Clearly, in this natural plant population, maternal inheritance will have an impact

on the evolution of four juvenile traits related to individual survival and fecundity.

Chapter 3
THE OPPORTUNITY FOR MATERNAL SELECTION

IN A NATURAL POPULATION 0F COLLINSIA VERNA
(SCROPHULARIACEAE).

INTRODUCTION

The biological complexity of hierarchically structured populations generates the
opportunity for selection at multiple levels of organization (Wilson 1975,1980; Wade
1978, 1982, 1985). While evolutionary biologists have argued that individual selection is
most parsimonious (Williams 1966; reviewed by Sober 1984), much theoretical work in
population and quantitative genetics has focused on how the diflerential extinction or
proliferation of groups, group selection, may also contribute to changes in allele
frequencies (Wilson 1975, 1980; Wade 1978, 1980, 1985; Breden 1990) or phenotypic
distributions (Yokoyama and Felsenstein 1978; see references in Cheverud 1984).
Empirical demonstrations of the components of group selection utilize artificially
constructed populations in which hierarchical levels of interaction can be easily
manipulated (Wade 1978; Breden and Wade 1989; Goodnight 1990ab; except see Stevens
et al. 1995; Kelly 1996). In contrast, in natural populations hierarchical interactions can be
more complex (e. g. Brandon 1990).

One approach to incorporating multiple levels of biological complexity is

partitioning the covariance between phenotype and fitness, a measure of phenotypic

139

140
selection, into within and among group components (Wade 1978, 1980). In this approach

the term phenotypic selection refers only to the relationship between phenotype and fitness
which is distinct from Endler’s (1986) definition of natural selection that incorporates both
phenotypic selection, heritability, and between generation response in the term natural
selection (Lande and Arnold 1983; Brodie et al. 1995 distinguish these two difl‘erent
processes as natural selection and evolution, respectively). Utilizing Price’s (1970, 1972)
covariance partitioning approach, Wade (1985) demonstrated the relationships between
hard selection, soft selection, group selection, and kin selection by illustrating how these
difierent selection models afl‘ect the within and among group components of selection as
well as the genetic variance between groups. Thus, the covariance approach provides a
generalized framework for describing multiple levels of selection given different
assumptions about the nature of selection, i.e. hard vs. sofi selection. Furthermore, Wade
(1985) derived the relationship between the within and among group covariances and
corresponding partial regression coeficients. This link between covariances, selection
differentials, and partial regression coeflicients, selection gradients, paved the way for a
new approach to the analysis of group selection, contextual analysis (Heisler and Damuth
1987 ; Goodnight et a1 1992). Contextual analysis is an extension of Lande and Arnold’s
(1983) multiple regression approach to selection analysis that allows one to include
contextual traits ie. both aggregate and emergent characters of groups as well as
individual traits. In hierarchically structured populations this statistical technique allows
one to identify the magnitude of selection at various scales i.e. to compare group and

individual selection. One advantage of the regression approach relative to the covariance

l4 1
partitioning approach is the ability to disentangle direct from indirect effects of selection at

difl‘erent levels of organization.

In nature, fimily groups often interact. For example, the efl‘ect of a mother on her
ofl‘spring is one of the most ubiquitous interactions with the potential to generate group
Structure in natural populations. These family interactions can be considered as kin
selection, a type of group selection in which interactions among related individuals have
fitness consequences (Hamilton 1964; Wade 1980; Kelly 1994 and many others).
Cheverud (1984) explored the effect of pleiotropy on kin selection between mothers and
their offspring to illustrate how genetic constraints may prohibit the evolution of altruism
Kirkpatrick and Lande (1989, 1992) termed this type of kin selection when mothers
directly afl‘ect the survival or fecundity of their offspring maternal selection and
demonstrated that maternal selection can difler fiom individual selection in two ways
(Kirkpatrick and Lande 1989, 1992). First, like maternal inheritance, maternal selection
can result in maladaptive evolution. Using the Karn and Penrose (1951) example of
stabilizing selection on human birth weight as a model of a maternal selection, Kirkpatrick
and Lande (1989, 1992) demonstrated that traits influencing maternal selection could
evolve maladaptively when stabilizing selection acted more strongly on the maternal trait
than on the ofl‘spring trait, when there was strong correlational selection between a mother
and her ofl‘spring, and when heritability was low. Second, they showed that maternal
selection was unlike other forms of selection because the magnitude depended on the
resemblance between parents and ofl‘spring.

Maternal selection is analogous to fimily selection utilized by animal and plant

breeders (Falconer 1981; England 1977). When heritability is low, family size is large, and

142
common environmental effects do not affect resemblance among fimily members, fimily

selection yields better response to selection than individual selection by providing a more
accurate assessment of individual breeding values (Falconer 1981). It is interesting to note,
however, that the combination of individual, fimily, and within fimily selection (i.e. index
selection) can produce greater responses to selection than individual selection alone in
certain cases. Thus, these artificial forms of selection, analogous to individual and group
selection in nature, demonstrate the possibility that different forms of selection are likely
influence evolutionary change in natural populations.

Maternal selection can include both pre-natal provisioning traits or post-natal
parental care traits by which a mother directly impacts her ofl‘spring’s survival in the
juvenile period and beyond. When mothers differ in their influence on the subsequent
generation, they create the opportunity for selection among fimilies because group
membership can influence both individual phenotype and/or individual fitness components.
Several fictors may contribute to the among maternal fimily group variance: 1) maternal
inheritance, the contribution of maternal phenotypic traits to phenotypic attributes in her
ofispring that may cause maternal fimily groups to difler phenotypically, 2) maternal
selection, the contribution of maternal phenotypic traits to the fitness of her ofl‘spring, and
3) spatial variation in the environment that can influence phenotypic expression and/or the
nature of phenotypic selection. In this chapter, I present a preliminary investigation of
maternal selection in which no maternal phenotypic attributes are included, and I explore
the Opportunity for maternal selection as the variance among maternal fimilies in relative
fitness. This among fimily variance is the first prerequisite for maternal selection to occur.

Furthermore, I evaluate the extent to which fimily membership may influence individual

143
relative fitness. This fimily efi‘ect indicates the maximum amount of variance that any

particular maternal trait may enquain (Heisler and Damuth 1987).

In natural populations, variance among maternal fimilies in relative fitness can
exist when mothers difl‘er in their effects on individual ofl‘spring fitness or when maternal
fimilies experience difierent selective environments. In plants limited seed dispersal means
that ofl‘spring from a maternal individual may experience selection on a local scale.
Evidence for fine scale variation in phenotypic selection in natural plant populations is
extensive. Phenotypic selection can vary spatially on a scale fiom meters to centimeters
(Kalisz 1986; Stewart and Schoen 1987; Scheiner 1989; Kelly 1992; Stratton 1992a).
Stratton (1994, 1995) has demonstrated that relative fitness of difl‘erent Erigeron
genotypes varies at a scale of 20 cm suggesting fine-scale spatial variation in selection in
experimental populations. In their extensive study of variation in selection on gall size in
Solidago, Weis et al. (1992) note that variable selection can result fiom: l) variation in the
relationship between phenotype and fitness, termed fluctuating fitness function, 2)
variation in the underlying phenotypic distribution, or 3) fiom a combination of these two
components. Thus, at the population level, variance among maternal fimilies can be
influenced both by maternal effects on fitness and by spatial variation in selection. In
contrast, paternal effects on fitness, paternal selection, would likely be manifested over
larger spatial scales and integrate fine scale variation in phenotypic selection because
pollen flow tends to be more widespread than seed dispersal (Levin and Kerster 1974).

In this chapter I evaluate the potential for maternal selection in a natural plant
population in which a number of juvenile traits display significant maternal inheritance

(Chapter 1) and are subject to individual phenotypic selection at various stages in the life

144
cycle (Chapter 2). I explore empirically maternal selection at two spatial scales that reflect

how difierent processes might contribute to among fimily variance. At the scale of the
population, families may difl‘er in fitness either because ofisprmg differ in their phenotypic
attributes or because maternal effects on ofl‘spring fitness difl‘er among fimilies. At this
scale, all families are planted into all environments and thus experience the average
selective environment. I refer to this average selective environment as the global scale.
However, in nature fimilies typically experience only a single environment due to limited
seed dispersal distance. In this case, in addition to the two factors above, among firnily
variance can also be influenced by spatial variation in selection. I refer to this local
selective environment as the local scale. In this chapter I address the following questions:
1) is there opportunity for selection by maternal fimily when fimilies experience the
average selective environment,
2) for these fimilies, can the variation in individual fitness be attributed to fimily
membership,
3) is there opportunity for selection among maternal fimilies experiencing single
environments,
a) does the direction and magnitude of linear selection vary spatially,
b) if so, at what spatial scale does it vary,
0) what fictors might account for spatial variation in selection,
4) on a local scale can the variation in individual fitness be attributed to maternal fimily
membership?
This study is the first to examine the potential for maternal selection on two spatial scales

in a natural plant population.

145
MATERIALS AND METHODS

WE

Collimia verna Nutt. (Scrophulariaceae), a winter annual inhabiting mesic forests
of the eastern United States (Femald 1970), displays substantial phenotypic variation in
juvenile traits including seed weight, emergence time, initial cotyledon diameter (hereafier
initial size), and fall cotyledon diameter (hereafler fill size) in a natural population in
Kalamazoo County, Michigan. In previous chapters I have demonstrated that seed weight,
initial size and fill size are maternally inherited (Chapter 1) and that phenotypic selection
fivors earlier emergence week and larger fill size in the natural population (Chapter 2).

To explore the Opportunity for maternal selection, I document among fimily
variance in relative fimily fitness. To explore the fictors influencing among fimily
variance in mean fimily fitness, I planted offspring at two spatial scales: global and local.
For the global scale, offspring from a single mother were planted at random across
multiple blocks (n=12 blocks) in the population. Families planted at this global scale allow
me to examine the first two factors independent of spatial variation in phenotypic
selection. For the local scale, offspring from a single mother were planted back into a
single block in the population. These single environment (hereafier local) ofl’spring
consisted of two types: 1) home ofl‘spring were planted back into the block fiom which
their mother was collected and 2) away offspring were planted into a randomly chosen
block. In preliminary analyses, home and away fimilies paired by block showed no
significant difierence in any fitness component in 1992 and a significant difference in only
two of four components of fitness in 1993 (Wilcoxin sign rank test). Therefore, home and

away offspring were combined in all subsequent analyses.

146
The study site and life cycle of C. vema were previously described (Chapter 2).

Naturally emerging and planted seedlings were individually tagged and scored for three
phenotypic traits, emergence week, initial size, and fill size. In addition, the planted
individuals were scored for seed weight. The measurement of these traits allowed me to
assess survivorship at three stages in the life cycle: 1) survival to establishment, 2) survival
to the onset of winter, and 3) survival to spring. The number of seed produced served as
an estimate of fecundity on all individuals that survived to the spring.

Spatial and temporal variation in biotic and abiotic fictors influencing survival and
fecundity previously described in Chapter 2 suggest that spatial variation in phenotypic
selection may affect the opportunity for maternal selection. The most notable spatial
variation in this population resulted fiom a sharp boundary between the woods and an
adjacent agricultural field that resulted in higher light levels along the edge and presumably
greater drought stress as well. I did observe wilting in the first few emergence censuses
along this edge. Seedling densities were also highest along this forest edge. In order to
account for this spatial variation in biotic and abiotic fictors influencing survival and
fecundity, I censused naturally emerging and planted individuals along two transects, one
along the edge and the other 25 m away in the interior. Ten blocks were arrayed along
each 100 m transect at 10 m intervals. To ensure adequate sample sizes in the analyses
described below, spatial variation was assessed at two spatial scales transect and block
nested within transect.

To document the opportunity for selection among maternal fimilies, I monitored
the survival and reproduction of locally planted ofl‘spring fiom 77 and 191 naturally

produced maternal fimilies in 1992 (n=844 seeds) and 1993 (n=1882 seeds), respectively,

147
and of globally planted offspring from 56 naturally produced maternal families in 1993

(n=5 52 seeds). Relatedness of offspring fiom a maternal fimily must fill between 0.25 and
0.5 for half-sibs and full-sibs, respectively. At TU Ave. the outcrossing rate was
consistently greater than 0.85 with a high probability of correlated mating (Holtsford et al.
in prep) suggesting that these maternal fimilies have an average relatedness close to 0. 5.
Relatedness between mothers and ofi‘spring was also likely to be 0.50 given the
infrequency of selfing in this population. Twelve seeds were planted per firnily, while all
seeds were planted for smaller fimilies. In 1993 some fimilies had more than 12 seeds
planted to fill in the planting array. On average 10.96 :1: 0.20 (:t 1 SE) and 9.86 i 0.18
seeds per local fimily were planted in 1992 and 1993, respectively, and 9.86 :1: 0.0.29
seeds per global fimily in 1993. This approach eliminated initial differences in the number
of ofl‘spring per fimily, hence among fimily difierences are conservative with respect to
initial fimily size.
A_na_1yai§
Maternal selection at a global scale

For global fimilies relative fimily fitness was calculated as the mean of individual
ofl‘spring fitness in a maternal fimily standardized by the grand mean of fimily mean
fitness for survival in each of three viability episodes and fimily mean fecundity in the final
episode of reproduction. The among fimily variance in relative family fitness is the
opportrmity for maternal selection.

Difl‘erences in survivorship among global families were examined by failure time
analysis using the life table method in the LIFETEST procedure in SAS (Fox, 1993; SAS

Institute, Inc., 1994). Subsequently, variation in survivorship in each selection episode was

148
analyzed by one-way analysis of variance for all fimilies with at least two ofl‘spring in a

given episode. Significance tests for these analyses are likely to be compromised by non-
normality of residuals. Variation among fimilies in phenotypic traits was determined by a
multivariate analysis of variance.

To evahrate the effect of fimily membership on individual fitness, fimily was added
to univariate models of phenotypic selection for global seedlings in each of the four
selection episodes. Heisler and Damuth (1987) suggested including group membership as
a class variable in an analysis of covariance as an indicator of the potential for group
selection. In this study a statistically significant fimily efl‘ect indicates the maximum
amount of variance that any maternal or fimily group attribute may explain (see Firebaugh
1979 in Heisler and Damuth 1987). First, 1 evaluated whether the relationship between the
trait and fitness component might vary among fimilies testing for heterogeneity among
slopes by including an interaction term last in a sequential sums of squares analysis. If
there was no indication of heterogeneous slopes, a reduced model including only the trait
and fimily were subsequently analyzed by partial sum of squares. Significant heterogeneity
of slopes in the full model or fimily effects in the reduced model both indicate the
potential for maternal selection. These analyses of covariance for each selection episode
are termed conditional because analyses were based only on individuals surviving to the
beginning of a particular episode.

Maternal selection at a local scale
The opportunity for maternal selection was calculated in the same way for local

fimilies as for global fimilies and therefore includes spatial variation in relative fitness.

149
Spatial variation in fitness components

 

Spatial variation in survivorship and fecundity in each of four episodes among the
edge and interior transects and among blocks within transects was analyzed in a separate
nested analysis of variance for each episode (GLM, SAS Institute Inc. 1994). Transects
and blocks nested within transects were treated as random efl"ects with appropriate F-tests
calculated using the Random statement in SAS using a type [[1 sum of squares. Residuals
from these analyses were not normally distributed because of the bivariate nature of the
survivorship data and the non-normal distribution of fecundity. While AN OVA is robust to
departures from normality (N eter, Waserman, and Kutner 1985), significance tests may be
compromised. Stratton (1995) found that significance for randomization tests did not
deviate substantially from AN OVA F—tests for similar sruvivorship data. These ANOVA
results are preferable in this case to non-parametric tests such as Kruskal-Wallis because
transect and block can not be treated as random fictors nor can those spatial efl‘ects be
appropriately nested in non-parametric approaches.

Spatial variation in phenotypic traps

The extent of spatial variation in the phenotypic traits, seed weight, emergence
week, initial size, and fill size was determined by a multivariate nested analysis of variance
using GLM (SAS Institute Inc. 1994). MANOVA is preferable to separate univariate
analyses because it does not inflate type I error. Secondly, it evaluates not only differences
in multivariate means, but also in correlation structure (Scheiner 1993 ).

Spatial variation in phenogpic selection

An episodic analysis of phenotypic selection for three viability and one fecundity

selection episode according to the reconstruction techniques of Lynch and Arnold (1988)

150
described in Chapter 1 was used to estimate spatial variation in the total linear selection

difl‘erentials and total linear selection gradients for natural and planted seedlings in 1992
and 1993. Only linear components of selection were examined because small sample sizes
at the block scale made estimates on non-linear components unreliable. Furthermore, only
the total magnitude of selection is described here for simplicity. This spatial analysis was
performed at two spatial scales in separate analyses ie. at the transect and block levels
with the number of bootstrap resampling rounds at 250 per transect and 50 per block,
respectively. The number of observations in the resampled data sets matched the original
number of observations fiom that transect or block.

Local fimily effects

AN COVA models for each selection episode were utilized to evaluate the potential
contribution of families to local variation in individual relative fitness. These models
included all phenotypic traits and a class variable, fimily. Separate analyses were
performed by block. Slopes were not tested for heterogeneity. In addition a single nested
AN COVA that included block and fimily nested within block as well as all phenotypic
traits in a given episode were utilized to explore the efl‘ect of family on Mess when spatial
variation could be accounted for by the inclusion of block in the analysis. This hierarchical
statistical model is a preliminary approach to analyzing selection at rmrltiple levels of
biological organization indicating the potential for phenotypic selection at individual,

fimily, and block level in this population.

15 1
RESULTS

Maternal selection at a global scale

For globally planted seedlings in 1993 the mean number (i standard deviation) of
offspring in a family (n=56) decreased from 9.85 i 2.19 at planting, 6.73 :1: 2.58 fiom
germination to establishment, 6.09 i 2.35 fiom establishment to the onset of winter, 4.66
i 2.18 fiom winter to spring, to 1.58 :t 1.28 in the final fecundity episode when only 43
fimilies were represented. A comparison of survivorship among fimilies did not difl‘er for
the three survivorship episodes (log-rank chi-square, 18:60.07, df=5 5, p>0.2973), but
difl‘ered when all planted seeds were included such that seeds not germinating were
censored in the analysis (log-rank chi-square, x2=108.85, df=55, p>0.0001) suggesting
that differences in dormancy among families contributed to among fimily difi‘erences in
survivorship curves. Mean survivorship only difiered among fimilies in the second
episode, survival to the onset of winter (Table 14). The opportunity for selection among
globally planted fimilies was greatest in survivorship to spring and fecundity episodes
(Figure 25).

Global fimilies differed significantly in phenotype (MANOVA, Wilk’s A=0.24,
numerator df=220, denominator df=809.3, p<0.0001). Univariate analyses for each of the
four phenotypic traits showed that only seed weight differed significantly (AN OVA,
F=2.53, df=55, 205, p<0.0001), while emergence week (ANOVA, F=1.37, df=55, 205,
p<0.0586) and fall size (ANOVA, F=1.36, df=55, 205, p<0.0644) were marginally

significant.

l 5 2
Did fimily membership explain variation in individual survivorship and fecundity?

In Chapter 2 the conditional selection differentials for all planted seedlings in 1993-4, ie.
both local and global, showed significant selection on all traits except seed weight in the
first episode (see Chapter 2, Table 12). A separate analysis for global seedlings only
showed significant linear selection on emergence week in the first two survivorship
episodes, on fill size in survivorship to spring, on seed weight, initial size, and fill size in
the fecundity episode (Table 15). The magnitudes of the selection coefficients were similar
between analyses based on all planted seedlings and only on globally planted seedlings.
The AN COVA including fimily and its interaction showed that the relationship between
each trait and fitness components varied among fimilies for two traits, emergence week
during survival to establishment (Table 16A) and seed weight during survival to the onset
of winter (marginal interaction term) (Figure 26). These early episodes provided the most
power for testing for heterogeneous slopes because they had more observations per family
than later episodes (see survivorship above). When slopes were not heterogeneous among
families, one trait, initial size, approached significance for fimily efi‘ect on survival to the
onset of winter (Figure 28, Table 16B), while other traits showed no significant fimily
efieas (Figure 27, Table 16B).

hiatemal selection 111g local scale

 

Spatial variation in biotic and abiotic fictors can affect among fimily differences in
phenotypic traits, fitness components, and in the relationship between phenotype and

Mess at two scales: transects (25 m apart) and blocks (adjacent pairs separated by 10 m).

153

Table 14. One-way analysis of variance of fimily variation in survivorship for globally
planted seedlings in 1993 at TU Ave for four episodes: 1) survival through establishment
(A), 2) survival to the onset of winter (B), 3) survival to spring (C), and 4) final fecundity
(D). Only fimilies with two or more ofl‘spring in a given episode were included in this
analysis. Degrees of freedom are for the numerator and denominator, respectively.

 

 

 

Episode

Source 1?.2 df MS F P

A. Survival to establishment

Family 0.12 51,321 0.08 0.91 0.6490
B. Survival to the onset of winter

Family 0.20 51, 336 0.24 1.38 0.0538
C. Survivalto spring

Family 0.17 48, 205 0.20 0.88 0.6892
D. Fecundity

Family 0.32 27, 45 92.65 0.81 0.7213

154

Figure 25. The opportunity for maternal selection portrayed as the among fimily variance
in relative Mess in each episode for locally and globally planted seedlings. Relative

fimily fitness is calculated as mean family fitness standardized by the grand mean of fimily
fitnesses for the sample population.

155

 

Eon—52338 9 332% I

has}, “—0 wows—O 05 OH Egan—m

 

madam 9 3:33 A

 

 

 

 

€886; _ _

 

 

 

 

 

moa— N03

.35qu 4.24001— 17..qu

~°9~Q<t<~1°
GOOD

v-4v—4

v-dv—C

SSHNJJJ ElAIlV'IE-DI XIIWVJ NJ HONVRIV A

N99919:

156

Table 15. Univariate regression models for globally planted seedlings in 1993 in four
selection episodes: 1) sruvival through establishment (A), 2) survival to the onset of
winter (B), 3) survival to spring (C), and 4) final fecundity (D). Significant linear selection
coeflicients (B) are bold.

 

 

 

Source
R2 df [3 F P
A. Survival to establishment
Seed weight 0.00 1, 375 0.01 0.13 0.7208
Emergence week 0.05 l, 375 - 20.16 0.0001
0.07
B. Survival to onset of winter
Seed weight 0.01 l, 339 0.05 2.40 0.1223
Emergence week 0.02 l, 339 0.08 6.70 0.0100
Initial size 0.01 1, 339 0.05 2.92 0.0886
C. Survival to spring
Seed weight 0.01 1, 259 0.14 2.36 0.1257
Emergence week 0.01 l, 259 - 1.83 0.1770
0.12
Initial size 0.00 1, 259 0.07 0.60 0.4400
Fall size 0.05 1, 259 0.33 15.25 0.0001

D. Fecundity

Seed weight 0.08 1, 86 0.38 8.00 0.0058
Emergence week 0.00 1, 86 0.01 0.00 0.9530
Initial size 0.09 1, 86 0.38 8.91 0.0037
Fallsize 0.13 l, 86 0.47 12.45 0.0007

157

 

 

 

 

 

28S a: .8 88¢ E .3. 235...
SSS M: .2 5S 2L .2 $52
38... E 4 Ed 88.: >2 ._ 3.6 09mm :3
336 2 .nm 236 3— .3 Quad mmm .3 SIRE—
Emad 2 .3 3E8 62 .3 Seed mmn .33 ESE
385 2 ._ cod named c2 .~ 36 3.86 m8 4 Rd one—5::
336 mm .3 good 22 .9 396 mmm .3 38.: com .3 2.3283."—
meaod mm .9 «~36 SH .3 Sofie mmN .3 836 com .33 ESE
336 mu .2 $6 $86 3: ._ 36 $8.: mmm ._ and 38.: own 4 end 0.83 383225
wcvmd 2 .5 082. 3— .x... $86 mmm .3 $56 33 .3 210.226
SSS S .9 MES 82 .2 88d SN .2 ES SN .2 £832
88.: S ._ 36 326 3— ._ and 32d mmm ._ and wwwnd ooN ._ vmd Emma)» 30m
mono—m 9583880: .<

m :0 Aw— m .5. mm m a. “a m .20 A...

baa—Sou main». 3 3325 88:23.26 828 8 3225 “BASE—€80 3 :2sz
383m 836m

 

8:33 we 88 3:8 .3 £2.33 no 88 E 2.3. was Soho amaze me 850%2 Be 323 083

A5 moan—a macaw mono—m .«e bmouoweouo: wan—am? 8: was: =a Hem £358 303.0% $33 re:— 8353 3.5% 8.893 .«e 88
Egg—com .3 £8.33 we 83 _ 23,—. .8 v3.3 83 oogufiwmm A3 mono—m we £08388.— 8255 3 «EB ecuofioafi 05 v2.22:
.868 S 2: .22 E 2:563 .22» BL 8:628 .2635: EL 3838 65 36.286 é 286:. «582.4 65 .2 65.:

158

mgad 3. .9. 396 SN .33 hams..—

 

 

 

S85 3. ._ :3 885 3a ._ mg as... E
223 319. 8...; EN .2 88d 38 .2 SEE
$85 3. ._ $5 323 EN ._ 2 .o 23.: gm ._ 12 8a 3:5
585 3. .S 83° «.2 .2 32 .o as .2 $8....—
325 3. ._ two $86 3N ._ 23 :22 3a ._ o3 “.83 semsam
38d 3. .2. 8:3 SN .2 Soho 8m .2 Ea:
as... 3. ._ 33 EN; EN ._ a; 285 can ._ 2d ammo: Bow
magma £5 .m
m .6 «a m a. .... m a. N... n. a. R
gun—com mam—mm 8 3225 52:3 ,3 $28 3 3225 2585:2350 3 33.55
885m ooh—om

 

3.253 2 03.:

159

Figure 26. The phenotypic distribution of seed weight for all globally planted seedlings
(A), the distribution of family means (I) in standard deviation units with 10, 25, median,
75, and 95 percentiles depicted (B), and the overall linear relationship between seed
weight and survival to the onset of winter (C).

 

 

 

 

 

 

 

‘
n,
A ,

160

 

 

 

 

 

/

 

 

 

 

 

_* a _ =—

_ .3:
_ a an

 

 

 

_
_
_

 

 

 

 

 

 

 

 

 

 

 

 

 

3.2. 19.8. 6

7
0.
A<>~>M3m m>F<4md

l
l

gays-116‘

>Ag<m

 

SEED WEIGHT (STD)

80

70
> 60
‘73 50
g 40

30
20
10

muA
2am

mN.N

mmm
5N".
oF—iv—I

mmd
mNd-

tn
5.
9

mm.—.

V3

l‘.

F—I
.

mN.N-

mud-

SEED WEIGHT MJDPOINT

Figure 26.

161

Figure 27. The phenotypic distribution of emergence week for all globally planted
seedlings (A), the distribution of family means (I) in standard deviation units with 10, 25,
median, 75, and 95 percentiles depicted (B), and the overall linear relationship between
emergence week and survival to the onset of winter (C).

162

 

t—i
vb

 

DJ

 

 

p—ny—up—a
n—IN

 

 

 

l LilL llll lllJ 111] [Ill 1111

RELATIVE SURVIVAL

 

 

99.0.0
aquacu—

 

 

 

 

 

 

. [111 111
.4

FAMILY

 

 

   

I

 

.IIIIII
lIII'I

.I .
In“
I II

I

I]
I:

I
I
I
I.
II

I

 

 

 

    
 
    
  
   

 

 

 

Figure 27.

<-3

-2.5

I
I

III 'I“
I ;I III 1

I
I
I
I

|
1

 

I
I I
. I
|
I I

I

 

II

 

I...

 

 

 

I

 

    

' I
In.

I
I

II
III" II
II

'IIII

I
I

 

 

I I

III
I

   

I
III

I

 

 

-1 0 1 2

EMERGENCE WEEK STD

-1.5 -o.5 0.5 1.5 2.5
EMERGENCE WEEK MIDPOINT

3.5

>=4

 

 

163

Figure 28. The phenotypic distribution of initial size for all globally planted seedlings (A),
the distribution of family means (I) in standard deviation units with 10, 25, median, 75,
and 95 percentiles depicted (B), and the overall linear relationship between initial size and
survival to the onset of winter (C). The two lines in (C) are for overall regression and
weighted average of within family regressions, respectively.

164

 

 

 

 

 

 

FREQUENCY

 

<-2.5 -2.25 -1.75 -1.25 -0.75 -0.25 0.25 0.75 1.25 1.75 2.25 >=2.5
INITIAL SIZE MIDPOINT

Figure 28.

165
Spatial variation in fitness conmonents

 

In general, there was little significant variation among transects in survivorship or
fecundity (Table 17). Transects difl‘ered significantly in survivorship only for two episodes
in 1992 for both natural and planted seedlings: survival to the onset of winter and survival
to spring (natural seedlings only). In contrast, blocks difi'ered significantly in survivorship
and fecundity across all episodes for both natural and planted seedlings (Table 17). The
one exception to significant spatial variation in survivorship at the scale of blocks were the
1992 planted seedlings, probably due to the limited sample size in that category. The
AN OVA models accounted for 3 to 20% of the variation in fitness components when only
transect and block were included in the model. For planted seedlings, the inclusion of
nested family efl‘ects increased the R2 ; these models accounted for 20 to 59% of the
variance in fitness.

The spatial pattern of variation in survivorship and fecundity among blocks
difl‘ered across episodes (Figures 29-30BCEF). When contiguous blocks were sampled for
all episodes in 1993, survival to establishment was more uniform across adjacent blocks
than survivorship in subsequent episodes. Also the variance among blocks increased
through subsequent episodes. Coeflicients of variation for the grand mean across blocks
demonstrated that spring survivorship (CV ranged from 31.4 to 46.0 across years), and
fecundity (range=30.5 to 46.2) were much more variable than survivorship through
establishment (range=5.5 to 8.0) and survival to the onset of winter (range=12.6 to 29.4).
In addition, some blocks showed consistent patterns among years, while others varied
across years. For example, spring survival was relatively low in blocks 11 and 12 in both

years, however, survival to spring was high in block 1 in 1992, but relatively poor in 1993

166

 

0000.0 8.. .0... S. :8... e... 8.... 0: 38.80.08".

 

 

 

0.8... .....N .00... 0. 88... 00.0 000.0 0. 888......8...
80.... 00.~ 08.0 . 83.0 8.0 80... . .808...
.08... 00.. 80.0 30 8.0 88.0 3.. 8.... 3.. 8.0 .082
000. 00.8... d

.80... ....0 R... 0. .80... 3.0 2.0... 0. .38....8...
.32. 00.0 02.... . .00.... 8.0 0.00... . 88.8..
88.0 8.0 o... 00.~ 3... .80... .0... 000.0 ~00. 8.0 .08.).
000. 8302.0

0.8... 8.. 00.... .8 800.0 8.0 00.... S 0.8.0.008...
SR... 8.. $0... 0. 080.0 0:. 000.0 0 .35....8...
008... 00.0 08.0 . 00.0... 8.0 000.0 . .808...
03.0.0 00.. 80.0 000 .0... 8.0.... 00.0 .02. 000 .0... .252
000. .025... ...

.80... 0: 08.. 0 .08... 00... 000... . 8800.50.88
008... 00.0. 3...... . 28.0 00... .00.. . 88.8...
.80... 0.... 000.0 000. 8.0 .80... 00.0 09.... ...: 8.0 .25:
~00. 83024

.. 0 02 .... N... .. ... 02 0.. a...
0.83 ..o .008. o. .3385 808883800 0. .3385
0.508% 880m

 

8.0088. 0... 8.. 0.0 82.00... .0 00080.. .08... 8.0. 880 0... 8.. 0.0 .0008 0... 8.. 88.00... .0 000805
.030. .0... 6.88.0... 3%. $30 8 8080.30 80.80.. 0... w80= 3.08.085 0.03 0.00.... 0.880880

.80 080b0 80.80. 00 3.00.. 0.03 3.8. 0w8...000 00.80.... 0.00... 8....3 .0300: E80,. .80 .8008...
8....3 60.00.. 0.00... .8080... 5.8.8.00. .08. ... .80 088.2.0 e. .3380 Am £0.83... .008. 0... e. .3380
.0 088.0080» .08.... .328 2 02.8...» :5. 8. 3.. E a 08. 0.0 ~00. 0. 00.0.88 038...
.80 8.80080 3:08.08 8.. 3.8800,. .80 8808823 8 88.0.03 .0..0..0 0... ..o <>OZ< 60.007. .... 030...

 

167

Es... 00.. .00.? 0o. 03.... o... 0.0... 3. 288.88.

 

 

 

3. o... 3.. N. 0.0m. 0. Sec... NV... mam... w. 300088.... 800:.
RH»... we... find . ... mm... a... max... . 80080....
ammo... on. $3.3 cm. mm .o 88... ...... mam... 02. cu... .08.).
«do. ..080... .Q

.25... Run an. .wwu w. Soc... 3... 08.. 0. 300088.... 0.00...
None... .3... nmmdo. . mam... no. 00.~. . 80080....
88... vc.~ .mcdmm mmm 0o... .08... no... no. .. E... cod .08.).
mg. .8807. .0

22.... mm. 00m .3 mm Sued ...... an... 8 3.00.... 8.80....
5.2.... we... owwdm 5 Non. .c 8.. v9... 5 300088.... 0.08m
032. 0.... $8.0m . 00...... co... Nae... . 80080....
.wcmd mm. mafimm m m on .o mm . od 3 .. mom... o... 3.... .08.).
moo. ..080... .m

28... rm... 0 . ...... n 38... em :0. mnm .o .. $00088... 0.08m
mammd 3... «00.00 . 035... a... moo... . 80080....
o. co... mm... Sm .3. v. m we... coco... 9mm mm»... ..3 mod .08.).
~00. .8802 .<

a. n. m.). .... my. a. n. m2 .... ...
3.8800“. 88.80 o. .0823
0.508”. 3ch

 

.2388. b. 050...

168
(Figures 29-30BCEF). Although the transects ran parallel to each other, there did not

appear to be any association between paired distances along the transects.

Spatial variation in phenotypic traits

Traits did not differ significantly among transects with one exception (Table 18).
Only planted seedlings in 1993 displayed significant variation in phenotypic traits among
transects, largely due to differences in emergence week. In contrast, phenotypic traits
difl'ered significantly among blocks within transect (Table 18). Univariate ANOVA’s
suggested that fall size varied significantly among blocks for both natural and planted
seedlings in both years (Figures 29-30AD), while the significance of other traits varied
among years (Table 18).

Spatial variation in phenoflpic selection

Both linear selection difierentials and gradients, measures of phenotypic selection,
showed significant variation among transects (Figures 31-32). Ninety-five percent
confidence intervals did not overlap for both the linear differential and gradient for fall size
in the natural seedlings in 1993 and for the linear difl‘erential for emergence week in the
planted seedlings in 1992. Other traits showed little overlap in confidence intervals: 1992
linear differentials for emergence week and fall size and 1993 linear differential on initial
size. In this analysis 95% confidence intervals were based on limited resampling (n=250)
hindering my ability to detect spatial variation. Despite this limitation, I detected spatial

variation in the linear components of phenotypic selection across transects.

169

Figure 29. The spatial scale of phenotypic and demographic variation for natural (A, B,
C) and planted seedlings (D,E,F) in 1992. Block means (i 1 standard error ) for
phenotypic variation in seed weight (k), emergence week (I), initial size (0), and fall
size (A) (A, D), survivorship through three episodes, survival to establishment (I),
survival to the onset of winter (0), survival to spring (A) (B, E), and fecundity in the final
episode (36) (C, F). The grand mean across blocks is depicted by a line for each episode of
survival or reproduction. Blocks are located on forest edge and interior.

170

 

 

 

 

 

 

 

 

 

A NATURAL D PLANTED
10 10
9 9
7 7
cZD 6 6 O Q
4 4
E 3 3
E 2 2
1 1
2’ . 1
: I 3
0.9-‘+—l—-l L+——,J+ 09—:
0.8— 08-;
«0.73? 9 9 Q i 0.7—g
06‘E—-r9--———----—-F§‘ 06‘;
05— i 0.5—;
$0.41} 4 I i i 04‘?
E 03-: ........ I. . . ................. TM} ....... in 03—;
02—; 02‘?
E014: i 0.1—2
OdC 0 F
104 14

 

 

 

 

MEAN FECUNDITY
.1:- u.
1.1
H—I
H1
5....
01 co
1 1
Hm—I
1-—-l—1
Hut
*
l-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o o“
1248911121517 1248911121517
BLOCK BLOCK
EDGE INTERIOR EDGE INTERIOR

Figure 29.

171

Figure 30. The spatial scale of phenotypic and demographic variation for natural (A B,
C) and planted seedlings (D,E,F) in 1993. Block means (i 1 standard error ) for
phenotypic variation in seed weight (*), emergence week (I), initial size (I), and fall
size (A) (A, D), survivorship through three episodes, survival to establishment (I),
survival to the onset of winter (0), survival to spring (A) (B, E), and fecundity in the final
episode (36) (C, F). The grand mean across blocks is depicted by a line for each episode of
survival or reproduction. Blocks are located on forest edge and interior.

172

A NATURAL

MEAN PHENOTYPE
-- N w a. u- o \1 co

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

D PLANTED

r—‘NW-fi-MO‘INOO

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 B
I ll!

:32 3 9
20.7 if
Eon-3 ‘i u “ ‘
s”? 9 i l}
w0.4f
303—355} }
E3333

0-

c F

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32 1:33 3
if: 1113513 W N
E]: 15513323 PE :3 h Hm“

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F‘NMVWNDI‘QChofiNMVVDwaO‘o
—————————— N

EDGE | INTERIOR

Figure 30.

 

v—NMV‘WOFwO‘O—‘NMVWOFQQO

o—nu—u—u—u—Qv—u—n—v—u—N

EDGE INTERIOR

173

 

 

 

88¢ 88... 885 885 885 women one a; Sad 98:: ESE

885 82:. £85 :83 885 28 NE SN 32 9855 E85

982 386 28.: ~33 285 2 v 9:. 23 88:5

82 8.55 d

885 m8? 385 885 «.89 3 Rd 83 €825 E85

$86 23.... 3:... 385 2 m and :3 6855

82 1352.0

:86 :25 :85 386 «83. c2 SN 5: N35 $8.6 Eaam

$85 $35 83.... 8E5 885 E: ”a m: 83 833.: :88

$85 33.3 33° :26 2.25 e v 02 83 685:.

82 BEE .m

886 885 885 885 SE a Sam on? Egg: :85

5:3 285 NEE 326 n m a? $3 8855

82 3532 .<

85 E83 EEQB m E E. m E 888
ii 03m GEE 88385 woom hog—«£8289 838852 .853

Egg B<E<ESE

 

$18 momcmha $58388 3% tab :23 8m 3882; 98 magnanoa 82.33:: $83.28 =« Sm comma:

29$ 33:3 mo 88 E 09E. Ema—38 v8.83 8m 853:? “825%“. mm v8.3: 203 Emma? 38 wfiwéofi 3mg 3cm .qumvoom
$3880 $332. Sm 852:3 82:893. ma BSA—3a 89$ 3% Em 28 “3mm FEE £83 oofiwhoao «Eu: 8:: Avg— .0:—
.3:qu m<mv 215 E 808036 <>OZ<E 2: a 333% 385m 33.—mega 4E3 Somme 822:: mu @885 803 “we—nag 338
.2 303832 .82 Ba 3: E :2 Es En: is. $503. 8823 9.85 E5? moEEE 23 .202 E 2": 2a 82 E
SHE $0855 £53 383 ANNE 33mg: Sm 3382 was gang? £58 33% 2F .o>< E 3 32 23 mom: 3 mwfiuoom
v85:— Ea wagon—o £23.:— uom 3:5 own—bogs; E scat? 33% 05.? 0235?? abs; come: gum—«>332 .m— 032.

174
Phenotypic selection also varied across blocks (Figure 33-34). Non-overlapping

confidence intervals in the linear differential for initial size in the planted seedlings and for
emergence week and initial size in natural seedlings and for the linear gradient for
emergence week in the natural seedlings all indicated spatial variability (Figure 34). Other
traits showed minimal overlap in confidence intervals ie. linear differential for fall size
(Figure 34E). Despite the very limited resampling efi‘ort (n=50), selection varied at a scale
of 10 meters for some traits i.e. emergence week and initial size. In contrast, selection on
fall size appeared more consistent across larger spatial scales. The spatial scale of variation
in phenotypic selection also difi‘ered among episodes.

W32

For maternal families planted locally in both years, the variance in relative fitness at
the family level indicates the opportunity for maternal selection is greatest in two episodes:
spring survival and fecundity (Figure 25). The total variance across episodes is greater for
locally planted families than for globally planted families. This difference between families
planted at these two spatial scales may indicative of the extent to which spatial variation in
biotic and abiotic factors may afi‘ect phenotypic differences among families either in traits,
in components of Mess, or in phenotypic selection. This comparison is compromised by
the different numbers of families considered at these two scales, but the pattern suggests
that when families experience local conditions only, the among family variance in relative

fitness is greater.

175

Figure 3 1. Spatial variation in phenotypic selection between transects along forest edge
(1) and interior (2) for natural and experimentally planted seedlings in 1992. The linear
selection difierentials (A, C) and linear selection gradients (B, D) for three traits,
emergence week, initial size, and fall size, in each viability and fecundity episode. Codes
for each episode as in Figure 25. Total magnitude of selection across all episodes is
depicted by (C) and is based on reconstructed phenotypic variance-covariance matrix (see
Chapter 2). 95% confidence intervals for total values are also depicted by (O) and are
based on 250 resampled data sets for each transect. Natural seedlings along transect 2 in
1992 have confidence intervals that exceed the upper and lower axis values.

SELECTION DIFFERENTIAL

SELECTION GRADIENT

0100-3
-0200 j
-0300 ‘

-0100 j
-0200 j
-0300

176

NATURAL

INITIAL FALL

WEEK SIZE

SIZE

 

0.700 ,

 

0.600 1

 

0.500

 

0.400 2

 

0.300 j

 

 

0.200 2
0100-3
0000-:

 

 

 

 

 

 

 

-0.100-3

 

 

 

 

 

 

 

 

-0.200-:

 

 

-0.300 ‘

 

0.700 ,

 

0.600 Z
0500-:

in

 

 

0.400 I
0300-?
0200—3
0100—5
0000-3

$2 4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E

-0. 100

 

 

 

 

 

 

 

-0.200 .
-0.300 ‘

 

 

Figure31.

2

l 2
TRANSECT

PLANTED

INITIAL FALL
WEEK SIZE

SIZE

 

0.700 .

 

0.600 ‘

 

0.500

 

0.400 I

 

0.300

 

0.200 I

 

0.100 j
0.000—j

 

 

 

ma

 

 

HZ

 

 

 

 

 

 

 

 

 

 

0.700 .

 

0.600 j

 

0.500 ‘

LLL

 

 

0.400 .

 

0.300

 

0.200 ,

 

0.100 j
0.000%

 

 

I 1

 

 

 

 

 

 

 

 

 

 

 

2

l 2
TRANSECT

l

177

Figure 32. Spatial variation in phenotypic selection between transects along forest edge
(1) and interior (2) for natural and experimentally planted seedlings in 1993. The linear
selection differentials (A, C) and linear selection gradients (B, D) for three traits,
emergence week, initial size, and fall size, in each viability and fecundity episode. Codes
for each episode as in Figure 25. Total magnitude of selection across all episodes is
depicted by (O) and is based on reconstructed phenotypic variance-covariance matrix (see
Chapter 2). 95% confidence intervals for total values are also depicted by (O) and are
based on 250 resampled data sets for each transect.

SELECTION DIFFERENTIAL

SELECTION GRADIENT

-0100 Z
-0200 j
-0300

-0100
-0200
-0300 q

178

NATURAL

WEEK INITIAL FALL

SIZE

SIZE

 

0.700 1

A

 

0.600 1

 

i
0.500 ,

 

0.400 2

 

0.300

 

0.200 1
0100-1
0000—;

 

i

 

 

-0.100-§

 

 

 

 

 

 

 

0200 j
A

 

-0300 ‘

 

 

0.700 1

 

0.600

?

 

0.500

 

0.400

 

0.300
0.200—j
0.100-i
0.000—j

 

 

IW

 

 

 

 

-0100 j

 

 

 

 

 

 

 

-0.200

 

 

Figure 32.

2

1 2

TRANSECT

-0300 ‘

PLANTED

WEEK INITIAL FALL
SIZE

 

0.700 ,

 

0.600 j

 

0.500

 

0.400 ‘

 

0.300

 

0.200

 

0.100
,
0.000—j

Eire

 

Wit

 

 

 

 

 

 

 

 

 

 

 

 

0.700

 

0.600 1

 

0.500 f

 

0.400 I

 

0.300 ‘
0200-3
0100-3
0000-;

 

_@‘

 

 

 

 

 

 

 

 

 

 

 

i

 

 

 

 

2

l

TRANSECT

2

179

Figure 33. Spatial variability in phenotypic selection among blocks along the edge (1-10)
and interior (1 1-20) transects for natural seedlings in 1993. Linear selection differentials
(A,B,C) and linear selection gradients (D,E,F) in each viability and fecundity episode for
three traits, emergence week, initial size, and fall size are shown. The total magnitude of
selection across all episodes is depicted by (O) and is based on reconstructed phenotypic
variance-covariance matrix (see Chapter 2). 95% confidence intervals for total values are

based on 50 resampled data sets for each block. Total values are connected by a line for
visual clarity.

180

.8 28..

 

060.... .60....
MMUNHHUHN68L9SVEZI WMUWHHUHM68L9SV£ZI

8.? 8.?

o 8.? 8.? s
8.? 8...-m
8.? 8.? 3
8... 8... w
8... 8... N
8... 8... 9
8... 8... m
8... 8... m
8.. 8.. w.
8.. 8..

a

8.? 8.?
8.? 8.?m
8.? 8.? M
8.? 8.?w
8... 8... N
8... 8... a
8... 8... m
8... 8... m
8... 8...
...... 8.. m
8.. 8..

 

ma 1535.5 U Emma <

181

 

 

 

€86.83 5:2... 3.38

.853 .«o .88 on. 3 R>E=m
macaw 8 E225

368.com

DIEI

 

 

z
0

.....68. 8 2a....

M0015

IIIIIIII
8L9$V€IO68L9SV€ZI

 

mNfi 417$ m

 

and-
cod-
ovd-
omd-
cod N
end
ovd
cod
ow .o
co. _
ON. _

OLLDEI'IEIS

TVILNEIH'EIJlICI

182

Figure 34. Spatial variability in phenotypic selection among blocks along the edge (1- 10)
and interior (1 1-20) transects for experimental seedlings in 1993. Linear selection
differentials (A,B,C) and linear selection gradients (D,E,F) in each viability and fecundity
episode for three traits, emergence week, initial size, and fall size are shown. The total
magnitude of selection across all episodes is depicted by (O) and is based on reconstructed
phenotypic variance-covariance matrix (see Chapter 2). 95% confidence intervals for total
values are based on 50 resampled data sets for each block. Total values are connected by a
line for visual clarity. Confidence intervals exceed the upper or lower axis values in five
cases in blocks 3 and 12.

183

 

MUOAm
83322232222 a m h o n v m N—

 

 

NNB A<FEZ~ U

.8 28..

v.00...
..No....s.o.n....m.~...... a 8 .. e n .. m N .

..8..- ..8..-
. S
8.. ..- ..8 ..-n
a
8...? 8...? m.
o
..8... 8.... N
9
8.... 8.... m
m
..8... ..8... w.

..8. ..8.

..8... ..8..-
m
..8? ..8? m
m
8...? 8...? o
N
8.... 8.... m
.d
8.... 8.... m
..8... ..8... m

..8. ..8.

 

mmm—B <

184

 

 

.saemafifi ......e... $.28
8.83 we «once 05 8 123.6
mac 8 3 33.5%

369.com

DIEI

 

 

.....88. ..m 2......

MUOAm
Oman—22232222 O m A. O n v m N _

 

WNE ‘3’; m

OON. T
OOm .O..

8...?

O'
.INEIICIVHD NOIIDEI'IEIS

OOVO

OOQO

OONA

OON.T

oow .O-

OOVO-

OOOO

OOVO

oow .O

OON.

'IVLLNEIHEIJJICI NOLLDH'IHS

185
Locg am effects

In most cases spatial variation among blocks within transects accounted for the
observed variation in survival and fecrmdity (Table 17BD). In three cases, however,
differences among local families accounted for variation in survival to establishment and in
final fecundity (marginally significant) in 1993 and in survival to spring in 1992. In
contrast, a multivariate AN OVA demonstrated that in addition to significant variation
among blocks in 1993 and no significant spatial variation in 1992, local families differed in
their phenotypic attributes in both years (Table 18BD). The univariate analyses suggested
that all traits except initial size differed among families.

Does family membership account for any of the observed variation in individual
fitness on a local scale? The inclusion of family in an ANCOVA including all phenotypic
traits in a given episode demonstrated significant family effects in 10 out of 17 models
significantly explaining variation in individual relative fitness in both years ( out of a
possible number of 106 separate regression models). Limited sample sizes per block were
not sufiicient for these statistical descriptions of selection. In more complicated contextual
models where a family efi‘ect was nested within blocks, family contributed significantly to
the variance in survival to establishment and final fecundity in 1993 and to spring

survivorship in 1992 (Table 19).

186

 

 

 

 

88.. 88. .88..
888... ..8... 88... . 08.8.8.8...
.83
88... 8.8 888.8 . ...8... ......8 888.8 . 8880......
88... 8.... ....8... 8. 8.8... 88.. 8:... ..8. 385.888....
88... 8.8 88... .... .88... ......8 ....8... .... .85
...8... .... 2.8... 88 88... 88... 8.... 8...... 8.... 88... .082
88. 83...... ...
88 888 .35
88.... 88... 888.. . 08.8.8.8...
.83
88.... 8.. 888... . 5.8... ..88 88... . 888085
.88.... 8.... 88... 8... 88...... ...... ......o .... £855.88...
88... 88.8 8...... 8 88... 88.. 88... 8 .85
888... 8.... 888... 888 88... 2.8.... 8.... 8.... 2.8 88... .08.).
88. 8.5.... a...
.. .. 8.). .... ... .. h. 8.). .... ...
55.3.... .08.... c. .8>.>....m 80.5.8388 8 .8>.>....m
0.58am. 3.50m

 

.880. 00:85.58 8.. 008...... 0.03 80.8.68 .... .....8

... 25 8.88.08.88.18 ...... 8...... 3 .558 .8 5.8.3... .38 2.. 8 .258 .8 888......800 .88....
.8>.>....8 A. 8008...». 8.6,. 8.. 03.. D... .8 mg. ...... moo. ... $5.000... .558... 8.. 828.0238 ... 5.8.0.... 3.2.8
8.. wggcooa 0....3 8805.. 5303...... ..e 8.80.08 .8888... 0588.0 3 00:85.60 .... 8.8.8.8.... .0807. .o. 058...

187

 

 

 

 

8...... 88... .088 . 8...... 00.... 888.88 . .80...
8.8.... 8.... 08. .. . S80... 80... ..88.. . 08.8 ......
8.88.... 88.8 08.... . 8...... 88.. 888.8 . 08.8 1......
88...... 88.. 000... 8... 088... .... 0:8 80. .88.... 88......
88..... 08.8 808.. ... 8...... 88... 800.8 8. .8...
8...... 08.. 8.0... 88. 80... 8...... 88.. 880.8 088 88... .0802
88... 88...... ...
0.. ..8 .308
8... ... 88.8 000.. . 808...... 08.8 .8. .8 . 08.8 ......
808.... 88.8 88.... . 0088... 88... .2... . 08.8 ......8.
.003
8.08.... 8.8 088.. . 888... 88.. ...8.. . 00:08.08...
8088... ..8.. 808... 88 888.... 88.. 88.8 80 .88.... 88......
0.8.... ...... 88..... 8 888.... 88.. 088.8 8 .00...
.888... 88.. 088.... 88 80... :8... 88.. 888.8 08. ..8... .0802
888. 80.8.... .<
.. ... 8.). ... .8 .. .. 8s. 8.. .8
5.0.5008. $.38 ... .8>.>..=m
0008...... 00.2%

 

......80. .8. 0.888

188

DISCUSSION

In this population of C. verna the variance among maternal families in mean
relative Mess of individual family members relative to the family mean in the population
demonstrates that the prerequisite for maternal selection is met both at the local and global
scales. Furthermore, the decline in number of families through the episodes of selection .4
(from 56 to 43 for global families and from 77 to 43 in 1992 and 191 to 128 in 1993 for
local families) indicates that there is difl‘erential extinction and proliferation of maternal
family groups. The examination the relationship between family membership and fitness
components at two spatial scales indicates that 1) variation in absolute survivorship, 2)
variation in phenotypic traits, 3) variation in the relationship between phenotype and
fitness and 4) spatial variation in 1-3 all contribute to among family variance in this natural
population.
Maternal selection on a global scale

For globally planted families experiencing the average in environmental conditions,
the variance among families does not appear to be influenced by differences in absolute
survivorship or fecundity (Table 14). However, significant differences in the multivariate
phenotype among global families could contribute to the opportunity for maternal
selection. These familial phenotypic differences are consistent with the evidence for
maternal inheritance of these traits described in Chapter 1.

Two lines of evidence also suggest that the opportunity for maternal selection is
influenced by variation in phenotypic selection among families: 1) families difi‘ered in their

fitness functions, ie. slopes were heterogeneous in AN COVA, or 2) families perceived

189
selection similarly (slopes were homogeneous), but difl‘ered in relative Mess for other

reasons i. e. significant family effects in AN COVA excluding non-significant interaction
term. In the mivariate selection analyses for global families, slopes were significantly
heterogeneous for two traits, emergence week on the survival to establishment and seed
weight on survival to the onset of winter (Figure 26, P<0.0548) (Table 15). In later
episodes this AN COVA approach is compromised by few observations per family. In
general, the significance of this interaction would be best evaluated by many observations
within families. In the absence of heterogeneity of phenotypic selection among families,
one trait, initial size displayed marginally significant family effects on individual survival to
the onset of winter (Figure 27, Table 15).

The description of phenotypic selection for global families is limited by the small
number of individuals observed. This efi‘ect of this small sample size is evident in the
comparison of significance of selection coefficients between all planted seedlings (Chapter
2, Table 18) and global seedlings only (Table 15). When only global seedlings are
included, the selection coefficients are similar in magnitude but lack significance in a
number of episodes. As a result of this statistical limitation, the detection of family efl'ects
is also limited. However, in spite of these limitations both types of family effects are
evident. These two types of family efl‘ects indicate the potential for maternal selection. The
amount of variation accounted for by family efl‘ects represents the maximum amount of
variance that any given maternal family attribute may contribute to the model ( see Heisler
and Damuth 1987 ). This maternal family attribute is a property of the family group and
could include, for example, the family mean phenotype, a specific maternal trait, or an

emergent property of the family group. Contextual analysis separates individual fiom

190
group effects on Mess by including both individual traits and group attributes (Heisler

and Damuth, 1987 ; Goodnight et a]. 1992; Stevens et al. 1995). In contextual models,
therefore, one could compare selection at the individual and group level. For example, one
could evaluate whether group selection favored an attribute that was not favored by
individual selection, a common assrunption of theoretical models for the evolution of
altruism (Breden 1990). Furthermore, one could determine whether individual selection
indirectly generates selection at the group level or vice versa (Goodnight et a1 1992). In
this study maternal family attributes were not inchrded in univariate selection models, so
the nature of selection at different levels can not be evaluated. The significance of this
study is that it indicates the potential for maternal selection on seed weight, emergence
week and initial size in the early viability selection episodes in a natural population.
Measures of maternal phenotype and larger sample sizes would allow contextual analysis
of phenotypic selection.
M_aternal selection on a local scale

Spatial variation in biotic and abiotic factors can also affect the opportunity for
maternal selection. The opportunity of maternal selection is greater for locally planted
seedlings relative to globally planted seedlings indicating that spatial variation may
contribute to among family variance. This comparison is based on difi‘erent numbers of
families between groups which could bias the variance in either direction (Figure 25).
However, the evidence for spatial variation in absolute survivorship (Table 17, Figures 29-
3OBCEF), in the multivariate phenotype (Table 18, Figures 29-30AD), and in phenotypic
selection (Figures 31-34) especially at the block scale supports the conclusion that spatial

variation contributes to the opportunity for maternal selection. In addition, the spatial

191
pattern of absolute survivorship among the episodes suggests that biotic and abiotic

factors acting as agents of selection in different episodes operate at difierent spatial scales
(Figures 31-32).

To evaluate the contribution of family effects on individual relative Mess, I
accounted for this spatial variation in two ways. First, I analyzed rrmltivariate phenotypic
selection by block including a family efiect in an AN COVA model This analysis involved
a large number of regression models for each episode and each block (n=106 models). As
in the analysis of global seedlings, few individuals in each block limited these statistical
descriptions of selection. In the 17 significant models, 10 showed significant family efi‘ects.
This result suggests the potential for maternal selection in these blocks during certain
episodes. IfI corrected for the large number of regression models tested by adjusting for
table-wide significance (Rice 1989), however, this evidence for the potential for maternal
selection is no longer significant. Second, in a single AN COVA I examined differences in
individual Mess among blocks and families within blocks by accounting for the average
multivariate Mess Motion across the whole population in each episode (Table 19). In
these models the statistical description of selection was more robust because it was based
on larger sample sizes. Significant efl‘ects of block and family within block indicated the
potential for selection at two hierarchical scales, among blocks and among families within
blocks. In contrast to AN COVAs by block, these models do not allow for spatial variation
in phenotypic selection. Rather, they demonstrate that when phenotypic selection is
homogeneous across the population, blocks and families within them differed in relative

Mess.

192
In their contextual analysis of individual size in Impatiens, Stevens et al. (1995)

found evidence for group selection operating among patches distributed over similar
spatial scales as the block in this study. Selection coemcients on individual size and mean
size of the group difi‘ered in sign indicating opposition across levels of selection. Kelly
(1996) experimentally manipulated plant architecture to demonstrate that interactions
among near neighbors can have Mess consequences on target individuals in Impatiens.
His description of this interaction as kin selection depends on his assumption that
interacting individuals were relatives. In this study interaction among related offspring was
minimized because offspring were separated by a minimum distance of 8 cm when seeds
were planted. Therefore, evidence for kin selection is most likely to due to mother-
ofi‘spring interactions, not sibling interactions afier germination.

Studies of spatial variation in individual phenotypic selection have demonstrated
significant variation in selection over similar spatial scales (Kalisz 1986; Scheiner 1989;
Kelly 1992; Stratton 1992a). The spatial scale of variation in phenotypic selection relative
to gene flow and the strength of selection interact to determine the rate and scale of local
adaptive evolution. Local adaptation is a common feature in many natural plant
populations (e. g. Bennington and McGraw 1995b). Differences in phenotypic selection
can produce locally adapted phenotypes over very short spatial scales (Antonovics et
31.1971). In this study there is very little evidence for local adaptation because home and
away families did not differ significantly in any Mess component. Nevertheless, spatial
variation in selection can be a potent force for maintaining genetic variation in populations

(Haldane and Jaykar 1963; Barton and Turelli 1987).

193
Inheritance of ggup traits

It is possible that response to selection of group attributes may be a fimction of
indirect selection on correlated traits at the individual level that are heritable (Goodnight
1990 a,b). One interesting feature of maternal selection, a type of kin selection , however
is the possibility that maternal inheritance (Chapter 1) may provide a mechanistic model
for the inheritance of group attributes. While Cheverud (1984) has demonstrated how
genetic covariances can afi‘ect the evolution of altruistic interactions between mothers and
their offspring and produce unusual evolutionary responses, it is also possible that in the
absence of pleiotropy the heritability of a maternal attribute with direct efi‘ects on ofi‘spring
Mess could allow response to selection at the maternal family level For example,
genetically based variation among mothers in provisioning could cause difi'erential survival
among maternal families. The selection difi‘erential on this maternal attribute mediated by
offspring survival will determine the mean provisioning value among mothers in the next
generation. Genetic covariances among this provisioning trait and offspring traits could
constrain or enhance this selection response (see Chapter 1). Thus, maternal inheritance
can provide an alternative mechanism for the inheritance of group attributes in a maternal
selection model. Understanding the interplay between maternal inheritance and maternal
selection and their influence on multivariate evolution would provide a unique view of the
role of maternal effects on levels of selection.

Conclusions

In plant populations maternal family groups are spatially structured as a result of

limited seed dispersal. Differences in the relative survival or fecundity of individual

ofispring in these maternal family groups creates the opportunity for selection at two

194
hierarchical levels: among individuals and among maternal families. This study clearly

demonstrates the opportunity for maternal selection at two spatial scales. A number of
factors contribute to the among maternal family variance in relative Mess. Maternal
families vary phenotypically. This phenotypic variation is likely to be due both to maternal
inheritance of juvenile traits and to spatial variability in the environment. Furthermore,
families vary in survival and fecundity. Variation in relative individual survival and
fecundity can be attributed both to phenotypic attributes of the individuals, to family
membership, and to spatial location (i. e. block). The magritude of variation attributed to
family or block represents the maximum amount of variance that any specific maternal
attribute or group attribute may explain indicating the potential for selection at these
hierarchical levels. Furthermore, maternal inheritance of these traits (Chapter 1) provides a

mechanism for the inheritance of group level effects.

LIST OF REFERENCES

 

LIST OF REFERENCES

Antonovics, J., and J. Schmitt. 1986. Paternal and maternal efi‘ects on propagule size in
Anthoxanthum odoraium. Oecologia 69:277-282.

Antonovics, J, A D. Bradshaw, And R G. Turner. 1971. Heavy metal tolerance in plants.
Advances in Ecological Research 7: 1-85.

Arnold, S. J., and M. J. Wade. 1984a. On the measurement of natural and sexual
selection:theory. Evolution 38:709-734.

Arnold, S. J., and M. J. Wade. 1984b. On the measurement of natural and sexual
selection: applications. Evolution 38:720-734.

Atchley, W. R. 1984. Ontogeny, timing of development, and genetic variance-covariance
structure. American Naturalist 123:519-540.

Barton, N. H, and M. Turelli. 1989. Evolutionary quantitative genetics: How little do we
know? Annual Review of Genetics 23:337-370.

Baskin, J. M., and C. C. Baskin. 1983. Germination ecology of Collinsia verna, a winter
annual of rich deciduous woodlands. Bulletin of the Torrey Botanical Club
110:311-315.

Bennington, C. C., and J. B. McGraw. 1995a. Phenotypic selection in an artificial
population of Impatiens pallida: The importance of the invisible fraction.
Evolution 49:317-324.

Bennington, C. C., and J. B. McGraw. 1995b. Natural selection and ecotypic
differentiation in Impatien pallida. Ecological Monographs 65:303-323.

Biere, A 1991a. Parental efl‘ects in Lychnisflos-cuculi. 1: Seed size, germination and
seedling performance in a controlled environment. Journal of evolutionary Biology
32447-465.

Biere, A 1991b. Parental efl‘ects in Lychnisflos-cuculi. 1]. Selection on time of

emergence and seedling performance in the field. Journal of evolutionary Biology
3:467-486.

195

196

Bondari, K, R L. Wilham, and A. B. Freeman. 1978. Estimates of direct and maternal
genetic correlations for pupa weight and family size of T ribolium. Journal of
Animal Science 47:358-365.

Brandon, R N. 1990. Adaptation and environment. Princeton University Press, Princeton,
NJ.

Breden, F. 1990. Partitioning of covariance as a method for studying kin selection. Trends
in Ecology and Evolution 5 :224-228.

Breden, F. and M. J. Wade. 1989. Selection within and between kin groups of the
imported willow leaf beetle. American Naturalist 134: 35-50.

Brodie HI, E. D., A J. Moore, and F. J. Janzen. 1995. Visualizing and quantifying natural
selection. Trends in Ecology and Evolution 10:313-318.

Bull, J. J. 1987. Evolution of phenotypic variance. Evolution 41:303-315.

Cantet, R J. C., D. D. Kress, D. C. Anderson, D. E. Doombos, P. J. Burfening, and R L.
Blackwell. 1988. Direct and maternal variances and covariances and maternal
phenotypic effects on preweaning grth of beef cattle. Journal of Animal Science
66:648-660.

Cantet, R J. C. 1990. Estimation and prediction problems in mixed linear models for
maternal genetic effects. Doctoral Dissertation, Animal Sciences, University of
Illinois, Urbana-Champaign.

Cantet, R J. C., R L. Fernando, and D. Gianola. 1992a. Bayesian inference about
dispersion parameters of univariate mixed models with maternal effects:
Theoretical considerations. Genetics, Selection, and Evolution 24: 107-135.

Cantet, R J. C., R L. Fernando, D. Gianola, And I Misztal. 1992b. Genetic grouping for
direct and maternal efi‘ects with difi‘erential assignment of groups. Genetics,
Selection, and Evolution 24: 211-223.

Carriere, Y. 1994. Evolution of phenotypic variance: non-Mendelian parental influences
on phentypic and genotypic components of life history traits in a generalist
herbivore. Heredity 72:420-430.

Charlesworth, B. 1990. Optimization models, quantitative genetics, and nnrtation.
Evolution 44:520-538.

Cheverud, J. M., L. J. Leamy, W. R Atchley, and J. J. Rutledge. 1983. Quantitative
genetics and the evolution of ontogeny. Genetical Research 42:65-75.

197

Cheverud, J. M. 1984. Evolution by kin selection: A quantitative genetic model illustrated
by maternal performance in mice. Evolution 38:766-777.

Cheverud, J. M., and A J. Moore. 1994. Quantitative genetics and the role of the
environment provided by relatives in behavioral evohrtion. In C. R B. Boake (ed)
Quantitative Genetic Studies of Behavioral Evolution. The University of Chicago
Press. Chicago, IL.

Chiu, W.L., and B. B. Sears. 1993. Plastome-genome interactions affect plastid
transmission in Oenothera. Genetics 133:989-997.

Cockerham, C. C., and B. S. Weir. 1977. Quadratic analyses of reciprocal crosses.
Biometrics 33:187-203.

Cowley, D. E. 1991. Prenatal effects on mammalian growth: embryo transfer results. In E.
C. Dudley (ed) The unity of evolutionary biology: proceedings of the fourth

international congress of systematic and evolutionary biology. Dioscorides Press,
Portland, OR

Crespi, B. J ., and F. L. Bookstein. 1989. A path-analytic model for the measurement of
selection on morphology. Evolution 43: 18-28.

Crespi, B. J. 1990. Measuring the efi‘ect of natural selection on phenotypic interaction
systems. American Naturalist 135:32-47.

Dickerson, G. E. 1947. Composition of hog carcasses as influenced by heritable

differences in rate and economy of gain. Iowa Agricultural Experiment Station
Research Bulletin 354:489-524.

Dixon, P. M., J. Weiner, T. M'nchell-Olds, and R Woodley. 1987. Bootstrapping the Gini
coefficient of inequality. Ecology 68: 1548-1551.

Dixon, P. M. 1993. The bootstrap and the jacknife: Describing the precision of ecological
indices. In S. M. Scheiner and J. Gurevitch (eds) Design and analysis of ecological
experiments. Chapman and Hall, New York, NY.

Efl‘ron, B. 1982. The jacknife, the bootstrap, and other resampling plans. Soc. Ind. Appl.
Math. CBMS-Nationall Science Foundation. Monograph 38.

Eisen, E. J. 1967. Mating designs for estimating direct and maternal genetic variances and
direct-maternal genetic covariances. Canadian Journal of Genetic Cytology 9: 13-
22.

Endler, J. A 1986. Natural selection in the wild. Princeton University Press, Princeton,
NJ.

198

England, F. 1977. Response to fanrily selection based on replicated trials. Journal of
Agricultural Science 88: 127-134.

Falconer, D. S. 1955. Patterns of response in selection experiments with mice. Cold
Spring Harbour. Symposium on Quantitative Biology. 20: 178-196.

Falconer, D. S. 1965. Maternal effects and selection response. In S. J. Geerts (ed)
Genetics today, Proceedings of the XI International Congress on Genetics, Vol. 3,
p. 763-774. Pergamon, Oxford.

Falconer, D. S. 1981. Introduction to quantitative genetics 3rd ed. New York: Longman.

Fenner, M. 1983. Relationships between seed weight, ash content and seedling grth in
twenty-four species of Compositae. New Phytologist 95:697-706.

Femald, M. L. 1970. Gray's manual of botany. 8th ed. New York: D. Van Nostrand
Company.

Firebaugh, G. 1979. Assessing group effects: a comparison of two methods. Social
Methods Research 7: 384-395.

Foulley, J. L., and G. Lefort. 1978. Methodes d'estimation des efi’ets directs et matemels
en selection animale. Annales de Genetique et de Selection Animale. 10:475-496.

Fox, G. A. 1993. Failure-time analysis: emergence, flowering, survivorship, and other
waiting times. In S. M. Scheiner and J. Gurevitch (eds) Design and analysis of
ecological experiments. Chapman and Hall, New York, NY.

Gilham, N. W. 1994. Organelle genes and genomes. Oxford University Press, Oxford.

Goodnight, C. J. 1990a. Experimental studies of community evolution 1: The response to
selection at the community level Evolution 44: 1614-1624.

Goodnight, C. J. 1990b. Experimental studies of comnumity evohrtion II: The ecological
basis of the response to comnumity selection. Evolution 44: 1625-1636.

Goodnight, C. J., J. M. Schwartz, and L. Stevans. 1992. Contextual analysis of modes of
group selection, sofi selection, hard selection, and the evolution of altruism
American Naturalist 140:743-761.

Gross, K L. 1984. Efi‘ects of seed size and growth form on seedling establishment of six
monocarpic perennial plants. Journal of Ecology 72:369-3 87.

199

Gross, K L., and A D. Smith. 1991. Seed mass and emergence time efi‘ects on
performance of Panicum dichotomrflorum Michx across environments. Oecologia
87:270-278.

Haldane, J. B. S., and S. D. Jayakar. 1963. Polymorphism due to selection of varying
direction. Genetics 58: 237-242.

Hamilton, W. D. 1964. The genetical evolution of social behavior. 1. Journal of
Theoretical Biology 7: 1-16.

Heisler, I. L., and J. Damuth. 1987. A method for analyzing selection in hierarchically
structured populations. American Naturalist 130:582-602.

Houle, D. 1991. Genetic covariance of Mess correlates: what genetic correlations are
made of and why it matters. Evolution 45:630-648.

Janssen, G. M., G. d. Jong, E. N. G. Joosse, and W. Scharloo. 1988. A negative maternal
efiect in springtails. Evolution 42:828-834.

Karn, M. N. and L. S. Penrose. 1951. Birth weight and gestation time in relation to
maternal age, parity, and infant survival. Annals of Human Genetics 16: 147-164.

Kalisz, S. 1986. Variable selection on the timing of germination in Collinsia verna
(Scrophulariaceae). Evolution 40:479-491.

Kelly, C. A 1992a. Spatial and temporal variation in selection on correlated life-history
traits and plant size in Chamaecristafasiculata. Evolution 46: 1658-1673.

Kelly, J. K. 1994. The efiect of scale dependent processes on kin selection: Mating and
density regulation. Theoretical Population Biology 46 : 32-57.

Kelly, J. K 1996. Kin selection in the annual plant Impatiens capensis. American
Naturalist 147: 899-918.

Kingsolver, J. G., and D. W. Schemske. 1991. Path analyses of selection. Trends in
Ecology and Evolution 6:276-280.

Kirkpatrick, M., and R Lande. 1989. The evolution of maternal characters. Evolution
43:485-503.

Kirkpatrick, M., and R Lande. 1992. The evolution of maternal characters: errata.
Evolution 46:284.

Lacey, E. P. 1991. Parental effects on life-history traits on plants. In E. C. Dudley (ed)
The rmity of evolutionary biology: proceedings of the fourth international congress
of systematic and evolutionary biology. Dioscorides Press, Portland, OR

 

200

Lande, R 1979. Quantitative genetic analysis of multivariate evolution, applied to brain:
body size allometry. Evolution 33:402-416.

Lande, R 1982. A quantitative genetic theory of life history evolution. Ecology 63:607-
615.

Lande, R, and S. J. Arnold. 1983. The measurement of selection on correlated characters.
Evolution 37: 1210- 1226.

Lande, R and M. Kirkpatrick. 1990. Selection response in traits with maternal inheritance.
Genetical Research 55:189-197.

Lande, R, and T. Price. 1989. Genetic correlations and maternal efi‘ect coeficients
obtained fiom offspring-parent regression. Genetics 122:915-922.

Levin, D. A and H. W. Kerster. 1974. Gene flow in seed plants. In T. Dobzhansky, M. K
Hecht, and W. C. Steere (eds. ), Evolutionary biology, vol 7, 139-220. Plenum,
New York, NY.

Lynch, M. 1987. Evolution of intrafamilial interactions. Proceedings of the National
Academy of Science 84:8507-8511.

Lynch, M., and S. J. Arnold. 1988. The measurement of selection on size and growth. In
B. Ebenman and L. Persson (eds) Size-structured populations. Springer-Verlag,
Berlin.

Lynch, M., and B. Walsh. 1996. Quantitative genetics. Sinauer Associates (in press).

Mazer, S. J. 1987. The quantitative genetics of life history and Mess components in
Raphanus raphanistrum L. (Brassicaceae): Ecological and evolutionary
consequences of seed-weight variation. American Naturalist 130:891-914.

Mazer, S. J. 1987. Parental effects on seed development and seed yield in Raphanus
raphanistrum: implications for natural and sexual selection. Evolution 41:355-371.

Mazer, S. J. and C. T. Schick. 1991. Constancy of population parameters for life-history
and floral traits in Raphanus sativus L. 11. Effects of planting density on phenotype
and heritability estimates. Evolution 45: 1888-1907.

Meyer, K 1991. Estimating variances and covariances for multivariate animal models by
restricted maximum likelihood. Genetics, Selection, and Evolution 23:67-83.

Meyer, K 1992. Bias and sampling covariances of estimates of variance components due
to maternal effects. Genetics, Selection, and Evolution 24:487-509.

201

Miao, S. L., F. A Bazzaz, and R B. Prirnack. 1991. Persistence of maternal nutrient
effects in Plantago major: the third generation. Ecology 72: 1634-1642.

Miller, T. E. 1987 . Efi‘ects of emergence time on survival and growth in an early old-field
plant community. Oecologia 72:272-278.

Mitchell-Olds, T. 1986. Quantitative genetics of survival and growth in Impatiens
capensis. Evolution 40: 107-1 16.

Mitchell-Olds, T., and R G. Shaw. 1987. Regression analysis of natural selection:
Statistical inference and biological interpretation. Evolution 41: 1149-1161.

Mitchell-Olds, T., and J. Bergelson. 1990a. Statistical genetics of an annual plant,
Impatiens capensis. 1. Genetic basis of quantitative variation. Genetics 1242407-
4 l 5.

Mitchell-Olds, T., and J. Bergelson. 1990b. Statistical genetics of an annual plant,
Impatiens capensis. 11. Natural selection. Genetics 124:417-421.

Montalvo, A M., and R G. Shaw. 1994. Quantitative genetics of sequential life-history
and juvenile traits in the partially selfing perennial, Aquilegia caerulea. Evolution
48: 82 8- 84 1.

Neter, J ., W. Wasserman, and M. H Kutner. 1985. Applied linear statistical models. 2nd
ed. Irwin Homewood, IL.

Noreen, E. W. 1989. Computer-intensive methods for testing hypotheses: an introduction.
John Wiley & Sons, New York, NY.

Orzack, S. 1993. Stochastic modeling of extinction in plant populations. In P.L Fiedler
and S. K Jain (eds) Conservation biology: The theory and practice of nature
conservation, preservation, and management. Chapman and Hall, New York, NY.

Parrish, J. A D., and F. A Bazzaz 1985. Nutrient content of Abutilon theophrasti seeds
and the competititve abe of the resuhing plants. Oecologia 65:247-251.

Phillips, P. C., and S. J. Arnold. 1989. Visualizing multivariate selection. Evolution
43: 1209- 1222.

Platenkamp, G. A J., and R G. Shaw. 1993. Environmental and genetic maternal effects
on seed characters in Nemophila menzisii. Evolution 47:540-555.

Price, G. R 1970. Selection and covariance. Nature 227: 520-521..

202

Price, G. R 1972. Extension of covariance selection mathematics. Annals of Human
Genetics 35: 485-490.

Rice, W. R 1989. Analyzing tables of statistical tests. Evolution 43: 223-225.

Riska, B., J. J. Rutledge, and W. R Atchley. 1985. Covariance between direct and
maternal genetic effects in mice, with a model of persistent environmental
influences. Genetical Reseach 45:287-297.

Riska, B. 1991. Introduction to the symposium In E. C. Dudley (ed) The unity of
evolutionary biology: proceedings of the fourth international congress of
systematic and evolutionary biology. Dioscorides Press, Portland, OR

Roach, D. A, and R D. Wulfl‘. 1987. Maternal efiects in plants. Annual Review of
Ecology and Systematics 18:209-235.

Rocha, O. J., and A G. Stephenson. 1990. Effect of ovule position on seed production,
seed weight, and progeny performance in Phaseolus cocineus L. (Leguminosae).
American Journal of Botany 77: 1320-1329.

Ross, M. A, and J. L. Harper. 1972. Occupation of biological space during seedling
establishment. Journal of Ecology 60:77-80.

SAS Institute, Inc. 1994. SAS user’s guide: Statistics, version 6. SAS Institute, Inc. Cary,
NC.

Scheiner, S. M. 1989. Variable selection along a successful gradient. Evolution 43:548-
562.

Scheiner, S. 1993. MANOVA: Multiple response variables and multispecies interactions.
In S. M. Scheiner and J. Gurevitch (eds) Design and analysis of ecological
experiments. Chapman and Hall, New York, NY.

Schluter, D., and L. Gustafsson. 1993. Maternal inheritance of condition and clutch size in
the collared flycatcher. Evolution 47:658-667.

Schluter, D. 1994. Exploring Mess surfaces. American Naturalist 143: 597-616.

Schluter, D. 1988. Estimating the form of natural selection on a quantitative trait.
Evolution 42: 849-861.

Schmid, B., and C. Dolt. 1994. Efl‘ects of maternal and paternal environment and

genotype on offspring phenotype in Solidago altissima L. Evolution 4821525-
1549. .

203

Schmitt, J ., J. Niles, and R D. Wulfll 1992. Norms of reaction of seed traits to maternal
environments in Plantago lanceolata. American Naturalist 139:451-466.

Schwaegerle, K E., and D. A Levin. 1991. Quantitative genetics of Mess traits in a wild
population of Phlox Evolution 45:169-177.

Sewell, M. M., Y. Qiu, C. R Parks, and M. W. Chase. 1993. Genetic evidence for trace
paternal transmission of plastids in Liriodendron and Magnolia (Magnoliaceae).
American Journal of Botany 80:854-858.

Shaw, F. H, R G. Shaw, G. S. Wilkinson, and M. Turelli 1995. Changes in genetic
variances and covariances: G whiz? Evolution 49: 1260-1267.

Shaw, R G. 1987. Maxinurnr-likelihood approaches applied to quantitative genetics of
natural populations. Evolution 41:812-826.

Shaw, R G., and F. H Shaw. 1992. QuercuszPrograms for quantitative genetic analysis
using maximum likelihood. Published electronically on th e Internet; available via
anonymous tip from flp.bio.indiana.edu; directory path biology/quantgen/quercus.

Shaw, R G., and N. M. Waser. 1994. Quantitative genetic interpretations of
postpollination reproductive traits in plants. American Naturalist 143:617-635.

Shi, M. J., D. Laloe, F. Menissier, and G. Renaud. 1993. Estimation of genetic parameters

of preweaning performance in the French Limousin cattle breed. Genetics,
Selection, and Evolution 25:177-189.

Sinervo, B. 1991. Experimental and comparative analyses of egg size in lizards:
constraints on the adaptive evolution of maternal investment per ofl‘spring. In E. C.
Dudley (ed) The unity of evolutionary biology: proceedings of the fourth
international congress of systematic and evolutionary biology. Dioscorides Press,
Portland, OR

Sober, E. 1984. The nature of selection: evolutionary theory in philosophical focus. MIT
Press, Cambridge, MA

Southwood, O. I., and B. W. Kennedy. 1990. Estimation of direct and maternal genetic
variance for litter size in Canadian Yorkshire and Landrace swine using an animal
model. Journal of Animal Science 68: 1841-1847.

Stanton, M. L. 1985. Seed size and emergence time within a stand ofwild radish
(Raphanus raphanistrum L. ): the establishment of a Mess hierarchy. Oecologia
67: 524-53 1.

Stearns, S. C. 1992. The evolution of life histories. Oxford University Press, Oxford.

 

204

Stevens, L., C. J. Goodnight, and S. Kalisz 1995. Multi-level selection in natural
populations of Impatiens capensis. American Naturalist 145: 513-526.

Stewart, S. C. and D. J. Schoen. 1987 . Pattern of phenotypic viability and fecundity
selection in a natural plant population of Impatiens pallida. Evolution 41: 1290-
1301.

Stratton, D. A 1995. Spatial scale of variation in Mess of Erigeron annuus. American
Naturalist. 146: 608-624.

Stratton, D. A 1994. Genotype-environment interactions for Mess of Erigeron annuus
show fine-grained heterogeneity in selection. Evohrtion 48: 1607- 1618.

Stratton, D. A 1989. Competition prolongs expression of maternal effects in seedlings of
Erigeron annus(Asteraceae). American Journal of Botany 76: 1646-1653.

Stratton, D. A I992a. Life-cycle components of selection in Erigeron annuus: I.
Phenotypic selection. Evolution 46:92-106.

Stratton, D. A. 1992b. Life-cycle components of selection in Erigeron annuus: Genetic
variation. Evolution 46: 107- 120.

Thompson, R 1976. The estimation of maternal genetic variances. Biometrics 32:903-
917. .

Thompson, E. A, and R G. Shaw. 1992. Estimating polygenic models for multivariate
data on large pedigrees. Genetics 131:971-978.

Topham, P. B. 1966. Diallel analysis involving maternal and maternal interaction effects.
Heredity 21: 665-674.

Tuljapurkar, S. D. 1989. An uncertain life: Demography in random environments.
Theoretical Population Biology 35: 227-294.

van der Toom, J., and T. L. Pons. 1988. Establishment of Plantago lanceolata L. and
Plantago major L. among grass II. Shade tolerance of seedlings and selection on
time of germination. Oecologia 76:341-347.

van Sanford, D. A, and D. F. Matzinger. 1982. Direct and maternal genetic variances and
covariances fo seedling characters in Tobacco. Crop Science 22: 1213-1218.

Van Vleck, L. D. 1970. Index selection for direct and maternal genetic components of
economic traits. Biometrics 26:477-483. .

205

Van Vleck, L. D. 1976. Selection for direct, maternal, and grandmatemal genetic
components of econonric traits. Biometrics 32:173-181.

Wade, M. J., and S. Kalisz 1989. The additive partitioning of selection gradients.
Evolution 43: 1567-1569.

Wade, M. J. 1978. A critical review of the models of group selection. The Quarterly
Review of Biology 53: 101-114.

Wade, M. J. 1980. Kin selection: Its components. Science 210: 665-667.

Wade, M. J. 1982. Evolution of interference competition by individual, family, and group
selection. Proceedings of the National Academy of Science USA 79: 3575-3578.

Wade, M. J. 1985. Soil selection, hard selection, kin selection, and group selection.
American Naturalist 125: 61-73.

Waller, D. M. 1985. The genesis of size hierarchies in seedling populations of Impatiens
Capensis meerb. New Phytology 100:243-260.

Weiner, J. 1985. Size hierarchies in experimental populations of annual plants. Ecology
66:743-752.

Weiner, J. 1990. Asymetric competition in plant populations. Trends in Ecology and
Evolutionz360-364.

Weis, A. E., W. G. Abrahamson, and M. C. Anderson. 1992. Variable selection on
Eurosta's gall size, I: The extent and nature of variation in phenotypic selection.
Evolution 46: 1674- 1697.

Wilham, R L. 1963. The covariance between relatives for characters composed of
components contributed by related individuals. Biometrics 19:18-27.

Wilham, R L. 1972. The role of maternal effects in animal breeding: III. Biometrical
aspects of maternal effects in animals. Journal of Animal Science 35:1288-1293.

Wilham, R L. 1980. Problems in estimating maternal effects. Livestock Production
Science 72405-418.

Williams, G. C. 1966. Adaptation and natural selection. Princeton University Press,
Princeton, NJ.

Williams, G. C. 1957. Pleitropy, natural selection, and the evolution of senescence.
Evolution 11 (398-411). '

206

Wilson, D. S. 1975. A theory of group selection. Proceedings of the National Academy of
Science USA 72: 143-146.

Wilson, D. S. 1980. The natural selection of population and comnumities.
Benjamin/Cummings, Menlo Park, CA

Winn, A A 1988. Ecological and evolutionary consequences of seed size in Prunella
vulgaris. Ecology 69: 1537-1544.

Wulfl‘, R D. 1986. Seed size variation in Desmodium paniculatum. III. Effects on
reproductive yield and competitive ability. Journal of Ecology 74: 1 15-121.

Yokoyama, S. and J. Felsenstein. 1978. A model of kin selection for an altruistic trait
considered as a quantitative character. Proceedings of the National Academy of
Science USA 75: 420-422.

Young, C. W., and J. E. Legates. 1965. Genetic, phenotypic, and maternal
interrelationships of growth in mice. Genetics 52:563-576.