- The document discusses integrating fossils and phylogenies in macroevolutionary analyses. It summarizes some key insights gained from doing so. - Incorporating fossils improves ancestral state estimation and the ability to detect evolutionary trends and non-Brownian models of trait evolution like early or late bursts. A few well-placed fossils can have a large impact. - On a per-taxon basis, fossils contain more macroevolutionary information than extant taxa alone. The impact of fossils depends on the underlying evolutionary process. - Paleontological data and perspectives can motivate new macroevolutionary hypotheses about processes like variation in rates, modes, and tempos of evolution over time that may not be detectable from

1. Phylogenetic Paleobiology: What do we
stand to gain from integrating fossils and
phylogenies in macroevolutionary
analyses?
Graham Slater
Department of Paleobiology, National Museum of Natural History
@grahamjslater
www.fourdimensionalbiology.com
Smithsonian
Smithsonian
National Museum of Natural History
National Museum of Natu

6. get a time-calibrated phylogeny
40
30
20
10
0

7. gather some trait data
40
30
20
10
0

8. fit some models

9. “insert_model” explains the
evolution of “insert_trait” in
“insert_clade” !

10. LETTER
doi:10.1038/nature10516
Multiple routes to mammalian diversity
Chris Venditti1, Andrew Meade2 & Mark Pagel2,3
The radiation of the mammals provides a 165-million-year test case
and that this was followed by a gradual slowdown towards the pre-

11. phylogenies don’t really look like this
40
30
20
10
0

12. they look like this
40
30
20
10
0

14. adding fossils improves ancestral state estimates
Finarelli and Flynn 2006 Sys. Biol.

15. do those extinct things
matter for testing
macroevolutionary
hypotheses?

16. do those extinct things matter for testing
macroevolutionary hypotheses?

17. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?

18. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?
• does a paleontological perspective change
the way we formulate our hypotheses?

19. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?
• does a paleontological perspective change
the way we formulate our hypotheses?
• can we use fossil information when we have
no phylogeny including extinct species?

20. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?
• does a paleontological perspective change
the way we formulate our hypotheses?
• can we use fossil information when we have
no phylogeny including extinct species?

23. simulate trait
evolution

24. simulate trait
evolution
prune to extant
taxa only

25. simulate trait
evolution
prune to extant
taxa only
fit models

26. simulate trait
evolution
prune to extant
taxa only
replace a proportion
of extant taxa for
fossils

27. simulate trait
evolution
prune to extant
taxa only
replace a proportion
of extant taxa for
fossils
fit models

28. Phenotype
Brownian motion
Time

29. swapping fossils for extant taxa has no effect
if BM is the true model of evolution
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
proportion of taxa that are extinct
1.0

30. swapping fossils for extant taxa has no effect
if BM is the true model of evolution
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
proportion of taxa that are extinct
1.0

31. BM is a special case of most current
models

32. BM is a special case of most current
models
AIC = 2k - 2ln(L)
# parameters
Likelihood

33. Phenotype
trend
Time

34. Phenotype
trend
Time

35. Cope’s rule -- an evolutionary trend
Benson et al. 2013 PLoS ONE

36. no trends can be detected from extant taxa
only
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
0/100 fossils
0
1
2
3
4
5
root - tip increase in mean
6

37. but a few fossils have a substantial effect
Akaike Weights
1.0
5/100
0.8
0.6
0.4
0.2
0.0
0/100 fossils
0
1
2
3
4
5
root - tip increase in mean
6

38. and more fossils improves ability to detect
weaker trends
95 /100
Akaike Weights
1.0
5/100
0.8
50 /100
0.6
0.4
0.2
0.0
0/100 fossils
0
1
2
3
4
5
root - tip increase in mean
6

39. Phenotype
Early Burst - Declining rates
Time

40. Phenotype
Early Burst - Declining rates
Time

41. O R I G I NA L A RT I C L E
doi:10.1111/j.1558-5646.2010.01025.x
EARLY BURSTS OF BODY SIZE AND SHAPE
EVOLUTION ARE RARE IN COMPARATIVE
DATA
Luke J. Harmon,1,2,3 Jonathan B. Losos,4 T. Jonathan Davies,5 Rosemary G. Gillespie,6 John L. Gittleman,7
W. Bryan Jennings,8 Kenneth H. Kozak,9 Mark A. McPeek,10 Franck Moreno-Roark,11 Thomas J. Near,12
Andy Purvis,13 Robert E. Ricklefs,14 Dolph Schluter,2 James A. Schulte II,11 Ole Seehausen,15,16
Brian L. Sidlauskas,17,18 Omar Torres-Carvajal,19 Jason T. Weir,2 and Arne Ø. Mooers20
1
Department of Biological Sciences, University of Idaho, Moscow, Idaho 83844
2
Biodiversity Centre, University of British Columbia, Vancouver, BC V6T1Z4, Canada
3
E-mail: lukeh@uidaho.edu
4
Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, Massachusetts 02138
5
National Center for Ecological Analysis and Synthesis, University of California, Santa Barbara, 735 State Street, Suite 300,
Santa Barbara, California 93101
6
Department of Environmental Science, Policy and Management, University of California, Berkeley, California 94720
7
Odum School of Ecology, University of Georgia, Athens, Georgia 30602

42. early bursts need lots of taxa and big
changes in rate
Slater and Pennell (in press) Syst. Biol

43. early bursts need lots of taxa and big
weight
changes in rate
10
1.0
1.0
10
88
# of half lives
# elapsed
rate half
lives
0.8
0.8
66
0.6
0.6
44
0.4
0.4
22
00
Akaike
Weight
0.2
0.2
50
50
100
100
150
150
200
200
# taxa
number of taxa
Slater and Pennell (in press) Syst. Biol

44. early bursts need lots of taxa and big
weight
changes in rate
10
1.0
1.0
10
88
# of half lives
# elapsed
rate half
lives
0.8
0.8
66
0.6
0.6
44
0.4
0.4
22
00
Akaike
Weight
0.2
0.2
50
50
100
100
150
150
200
200
# taxa
number of taxa
Slater and Pennell (in press) Syst. Biol

45. early bursts need lots of taxa and big
weight
changes in rate
10
1.0
1.0
10
88
# of half lives
# elapsed
rate half
lives
0.8
0.8
66
0.6
0.6
44
0.4
0.4
22
00
Akaike
Weight
0.2
0.2
50
50
100
100
150
150
200
200
# taxa
number of taxa
Slater and Pennell (in press) Syst. Biol

46. we need a lot of fossils to detect weaker
early bursts
0.6
0
00
/1
50
5/
10
/10
0
0.8
95
Akaike Weights
1.0
00
/1
0
0.4
0.2
0.0
0
2
4
6
# elapsed rate half-lives
8

47. Phenotype
Late Burst - Accelerating rates
Time

48. Phenotype
Late Burst - Accelerating rates
Time

49. no ability to detect accelerating rates from
ultrametric trees
Akaike Weights
1.0
0.8
0.6
0.4
0/100 fossils
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

50. no ability to detect accelerating rates from
ultrametric trees
Akaike Weights
1.0
0.8
0.6
0.4
0/100 fossils
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

51. and may be mistaken for other “low-signal”
processes like Ornstein-Uhlenbeck
Akaike Weights
1.0
0.8
0.6
0.4
0/100 fossils
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

52. swapping extant tips for fossils increases
support for accelerating rates over OU
Akaike Weights
1.0
0.8
0.6
0.4
5/100 fossils
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

53. swapping extant tips for fossils increases
support for accelerating rates over OU
Akaike Weights
1.0
50/100 fossils
0.8
0.6
0.4
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

54. swapping extant tips for fossils increases
support for accelerating rates over OU
95/100 fossils
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
0
2
4
6
# elapsed rate doubling times
8

55. how much macroevolutionary information
do fossils hold relative to extant taxa?

56. how much macroevolutionary information
do fossils hold relative to extant taxa?
on a “per-taxon” basis, fossils
contribute more macroevolutionary
information than extant taxa

57. how much macroevolutionary information
do fossils hold relative to extant taxa?
on a “per-taxon” basis, fossils
contribute more macroevolutionary
information than extant taxa
impact of fossils depends on the
underlying evolutionary process

58. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?
• does a paleontological perspective change
the way we formulate our hypotheses?
• can we use fossil information when we have
no phylogeny including extinct species?

60. do we test the right
models?

61. How fast...do animals
evolve...? That is one of
the fundamental
questions regarding
evolution
Simpson (1944, 1953)
Photo: Florida Museum of Natural History

63. Illustration by Mark Hallet

64. Eocene
Oligocene
Miocene

65. fossils suggest an increase in mean and
variance of body size after the K-Pg
mean mass
K
Pg
standard deviation mass
Ng
K
Pg
Ng
Alroy (1999) Systematic Biology

66. Phylogenetic approaches find no rate
increase in the Cenozoic
relative
rate
J
K
Pg
Ng
Venditti et al. (2011) Nature

67. do we really think mammals
changed their rate of body
size evolution?

68. Simpson’s adaptive zones
From Simpson (1953)

69. the mammalian adaptive zone
body size
T
J
Mesozoic
K
Pg
Ng
Cenozoic

70. the mammalian adaptive zone
body size
T
J
Mesozoic
K
Pg
Ng
Cenozoic

71. variation in tempo
body size
T
J
Mesozoic
K
Pg
Ng
Cenozoic

72. variation in tempo
body size
evolution slow
T
J
Mesozoic
K
Pg
Ng
Cenozoic

73. variation in tempo
evolution slow
T
J
Mesozoic
K
Pg
Ng
Cenozoic
body size
evolution fast

74. body size
T
J
Mesozoic
K
Pg
Ng
Cenozoic
images from http://dinosaurs.about.com

75. variation in mode
body size
T
J
Mesozoic
K
Pg
Ng
Cenozoic
images from http://dinosaurs.about.com

76. variation in mode
body size
evolution constrained
T
J
Mesozoic
K
Pg
Ng
Cenozoic
images from http://dinosaurs.about.com

77. variation in mode
body size
evolution
unconstrained
evolution constrained
T
J
Mesozoic
K
Pg
Ng
Cenozoic
images from http://dinosaurs.about.com

78. 3 paleo-motivated models for mammalian
body size evolution
Slater (2013) Methods Ecol. Evol.

79. 3 paleo-motivated models for mammalian
body size evolution
K-Pg Shift
Mesozoic
Cenozoic
BM rate 1
BM rate 2
Slater (2013) Methods Ecol. Evol.

80. 3 paleo-motivated models for mammalian
body size evolution
K-Pg Shift
Cenozoic
BM rate 1
ecological release
Mesozoic
BM rate 2
Mesozoic
Cenozoic
Ornstein-Uhlenbeck
BM
Slater (2013) Methods Ecol. Evol.

81. 3 paleo-motivated models for mammalian
body size evolution
K-Pg Shift
Cenozoic
BM rate 1
ecological release
Mesozoic
BM rate 2
Mesozoic
Cenozoic
Ornstein-Uhlenbeck
release and radiate
Mesozoic
BM
Cenozoic
Ornstein-Uhlenbeck BM*
Slater (2013) Methods Ecol. Evol.

82. time calibrated phylogeny of living and fossil
mammals
0
2.59
Q
Ng
Pg
Cenozoic
23
66
K
J
Mesozoic
145
201.3
T
252.2
Pz
P
264.94
Slater (2013) Methods Ecol. Evol.

83. 0
2.59
Q
Ng
Pg
Cenozoic
23
66
K
J
Mesozoic
145
201.3
T
252.2
Pz
P
264.94
Slater (2013) Methods Ecol. Evol.

84. Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
Brownian
Motion
Directional
Trend
Ornstein
Uhlenbeck
AC
/DC
standard models
White
Noise
K-Pg
shift
Ecological
release
Release
& radiate
paleo-inspired models

85. release & radiate fits best
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
Brownian
Motion
Directional
Trend
Ornstein
Uhlenbeck
AC
/DC
standard models
White
Noise
K-Pg
shift
Ecological
release
Release
& radiate
paleo-inspired models

86. faster rates of body size evolution in the
Mesozoic?
Parameters Mesozoic
Cenozoic
rate (σ2)
0.97
0.1
OU param (α)
0.01
-

87. faster rates of body size evolution in the
Mesozoic?
Parameters Mesozoic
Cenozoic
rate (σ2)
0.97
0.1
OU param (α)
0.01
-

88. phenotype
Brownian motion is a diversifying process
time

89. phenotype
Brownian motion is a diversifying process
starting state
time

90. phenotype
Brownian motion is a diversifying process
starting state
rate σ2
time

91. phenotype
Brownian motion is a diversifying process
starting state
rate σ2
time

92. phenotype
Brownian motion is a diversifying process
starting state
starting state
rate σ2
time

93. phenotype
Brownian motion is a diversifying process
starting state
starting state
rate σ2
σ2 * time
time

94. phenotype
OU is an equilibrium process
time

95. phenotype
OU is an equilibrium process
starting state
time

96. phenotype
OU is an equilibrium process
starting state
rate σ2
time

97. phenotype
OU is an equilibrium process
starting state
rate σ2
rubber band parameter α
time

98. phenotype
OU is an equilibrium process
starting state
rate σ2
rubber band parameter α
time

99. phenotype
OU is an equilibrium process
starting state
starting state
rate σ2
rubber band parameter α
time

100. phenotype
OU is an equilibrium process
starting state
starting state
rate σ2
rubber band parameter α
time
σ2 / 2α

101. BM and OU simulated at the same rate give very
different disparities
phenotype
Brownian motion
Ornstein-Uhlenbeck
time

102. the OU process has an equilibrium disparity
200
50
variance
250
millions of years ago
150
100
Mesozoic
Cenozoic
0

103. the OU process has an equilibrium disparity
200
50
variance
250
millions of years ago
150
100
Mesozoic
Cenozoic
0

104. a low BM rate increases disparity
200
50
variance
250
millions of years ago
150
100
Mesozoic
Cenozoic
0

105. do we really think mammals
changed their rate of body
size evolution?

106. ✗
do we really think mammals
changed their rate of body
size evolution?

107. How fast...do animals
evolve...? That is one of
the fundamental
questions regarding
evolution
Simpson (1944, 1953)
Photo: Florida Museum of Natural History

108. How fast...do animals
evolve...? That is one of
the fundamental
questions regarding
evolution
Simpson (1944, 1953)
Photo: Florida Museum of Natural History

109. ...hang on a minute

110. A
B
C
D
8
6
4
2
0

111. A
B
C
D
8
6
4
2
0

112. A
B
C
D
A
A
8.24 0.00 0.00 0.00
B
B
0.00 8.24 0.61 0.61
C
0.00 0.61 4.65 4.10
D
0.00 0.61 4.10 8.24
C
D
8
6
4
2
0

113. A
B
C
D
A
A
8.24 0.00 0.00 0.00
B
B
0.00 8.24 0.61 0.61
C
0.00 0.61 4.65 4.10
D
0.00 0.61 4.10 8.24
C
D
8
6
4
2
0

114. A
B
C
D
A
A
8.24 0.00 0.00 0.00
B
B
0.00 8.24 0.61 0.61
C
0.00 0.61 4.65 4.10
D
0.00 0.61 4.10 8.24
C
D
8
6
4
2
0

115. OU VCV
transformation
A
B
C
D
A
A
4.04 0.00 0.00 0.00
B
B
0.00 4.04 0.18 0.13
C
0.00 0.18 3.03 1.75
D
0.00 0.13 1.75 4.04
C
D
8
6
4
2
0

116. OU VCV
transformation
A
B
C
D
A
A
4.04 0.00 0.00 0.00
B
B
0.00 4.04 0.18 0.13
C
0.00 0.18 3.03 1.75
D
0.00 0.13 1.75 4.04
C
D
8
6
4
2
0

117. OU VCV
transformation
A
B
C
D
A
A
4.04 0.00 0.00 0.00
B
B
0.00 4.04 0.18 0.13
C
0.00 0.18 3.03 1.75
D
0.00 0.13 1.75 4.04
C
D
8
6
4
2
0

118. OU VCV
transformation
A
B
C
D
A
A
4.04 0.00 0.00 0.00
B
B
0.00 4.04 0.18 0.13
C
0.00 0.18 3.03 1.75
D
0.00 0.13 1.75 4.04
C
D
8
6
4
2
0

119. OU branch length
transformation
A
B
C
D
A
A
4.04 0.00 0.00 0.00
B
B
0.00 4.04 0.13 0.13
C
0.00 0.13 1.48 1.22
D
0.00 0.13 1.22 4.04
C
D
8
6
4
2
0

120. release & radiate still fits best...
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
Brownian
Motion
Directional
Trend
Ornstein
Uhlenbeck
AC
/DC
standard models
White
Noise
K-Pg
shift
Ecological
release
Release
& radiate
paleo-inspired models

121. but ecological release is almost as good
Akaike Weights
1.0
0.8
0.6
0.4
0.2
0.0
Brownian
Motion
Directional
Trend
Ornstein
Uhlenbeck
AC
/DC
standard models
White
Noise
K-Pg
shift
Ecological
release
Release
& radiate
paleo-inspired models

122. a less pronounced rate decrease ...
Parameters Mesozoic
Cenozoic
rate (σ2)
0.2
0.1
OU param (α)
0.03
-

123. but
200
/ 2α makes more sense
millions of years ago
150
100
50
variance
250
2
σ
Mesozoic
Cenozoic
0

124. but
200
/ 2α makes more sense
millions of years ago
150
100
50
variance
250
2
σ
Mesozoic
Cenozoic
0

126. Tree transformations under OU
don’t work for non-ultrametric
trees!

127. do those extinct things matter for testing
macroevolutionary hypotheses?
• how much macroevolutionary information
do fossils hold relative to extant taxa?
• does a paleontological perspective change
the way we formulate our hypotheses?
• can we use fossil information when we have
no phylogeny including extinct species?

128. a touch of realism

129. most comparative biologists don’t have this
40
30
20
10
0

130. but have this
†
†
†
†
40
30
20
10
0

131. †
40
30
20
10
0

132. †
40
30
20
10
0

133. †
40
30
20
10
0

134. 0.00
Density
0.10
0.20
0.30
density.default(x = X)
40
-4
-2
0
2
ln(mass)
30
4
†
6
20
10
0

135. caniform carnivores span a huge range of body sizes

136. caniform carnivores span a huge range of body sizes
30-250 grams

137. caniform carnivores span a huge range of body sizes
30-250 grams
> 3, 000 Kg

138. do caniform carnivores exhibit a trend towards large
body size?
Finarelli and Flynn 2006 Sys. Biol.

139. Procyonidae
Mustelidae
Ailuridae
Mephitidae
Phocidae
Otariidae
Odobenidae
Ursidae
Canidae
50
40
30
20
10
Millions of Years Ago
0
-2.5
7.5
ln(body mass)
Slater et al. 2012 Evolution

140. 12 node priors
- 11 internal
- root
Procyonidae
Mustelidae
Ailuridae
Mephitidae
Phocidae
Otariidae
Odobenidae
Ursidae
Canidae
50
40
30
20
10
Millions of Years Ago
0
-2.5
7.5
ln(body mass)
Slater et al. 2012 Evolution

141. ancestral size is too large based on extant taxa ...
1.0
BM
1.0
0.8
density
density
0.8
Trend
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0.1
1
10
100
1000
0.1
1
mass
1.0
AC/DC
0.8
density
0.8
mass
OU
0.6
1000
extant
nodes + root
0.6
0.4
100
nodes
density
1.0
10
0.4
0.2
0.2
0.0
0.0
0.1
1
10
mass
100
1000
0.1
1
10
mass
100
1000

142. ... but is more realistic with fossil priors
1.0
BM
1.0
0.8
density
density
0.8
Trend
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0.1
1
10
100
1000
0.1
1
mass
1.0
AC/DC
0.8
density
0.8
mass
0.6
OU
1000
extant
nodes + root
0.6
0.4
100
extant
fossils
nodes
fossils + root
density
1.0
10
0.4
0.2
0.2
0.0
0.0
0.1
1
10
mass
100
1000
0.1
1
10
mass
100
1000

143. 3
extant
extant.only
fossils
nodes
fossils.plus.root
-1
2*Ln(Bayes Factor)
0
1
2
nodes + root
Trend
Trend
AC/DC
ACDC
OU
OU

144. 3
extant
extant.only
fossils
nodes
fossils.plus.root
nodes + root
-1
2*Ln(Bayes Factor)
0
1
2
positive
Trend
Trend
AC/DC
ACDC
OU
OU

145. 3
extant
extant.only
fossils
nodes
fossils.plus.root
-1
2*Ln(Bayes Factor)
0
1
2
nodes + root
Trend
Trend
AC/DC
ACDC
OU
OU

146. 3
extant
extant.only
fossils
nodes
fossils.plus.root
-1
2*Ln(Bayes Factor)
0
1
2
nodes + root
Trend
Trend
AC/DC
ACDC
OU
OU

147. the estimated change in mean mass is subtle
25
mode = 1.55
20
density
15
10
5
0
-1
0
1
2
3
4
root-tip increase in Ln(mass)
5

148. which is difficult to detect using AIC
95 /100
Akaike Weights
1.0
5/100
0.8
50 /100
0.6
0.4
0.2
0.0
0/100 fossils
0
1
2
3
4
root - tip change in mean
5
6

149. root-tip increase in Ln(mass)
joint marginal distribution of root state and trend
parameter
5
4
3
2
1
0
-1
0.5
1
10
ancestral mass (Kg)

150. root-tip increase in Ln(mass)
joint marginal distribution of root state and trend
parameter
5
4
3
2
1
0
-1
0.5
1
10
ancestral mass (Kg)

151. root-tip increase in Ln(mass)
joint marginal distribution of root state and trend
parameter
PP(Mu> 0) = 0.97
5
4
3
2
1
0
-1
0.5
1
10
ancestral mass (Kg)

152. how do fossils change our picture of caniform
size evolution?
no fossils
ancestral size
mode of evolution
with fossils

153. how do fossils change our picture of caniform
size evolution?
no fossils
ancestral size
mode of evolution
with fossils
large (~25kg)
small (~2 kg)

154. how do fossils change our picture of caniform
size evolution?
no fossils
ancestral size
mode of evolution
with fossils
large (~25kg)
small (~2 kg)
Brownian motion
Brownian motion
+ trend to large
size

155. can we use fossil information when we have no
phylogeny including extinct species?

156. can we use fossil information when we have no
phylogeny including extinct species?
even using fossil traits as
informative node priors
improves model fitting

157. do those extinct things
matter for testing
macroevolutionary
hypotheses?

158. today’s model systems for macroevolutionary
studies
images: wikipedia

159. tomorrow’s model systems for macroevolutionary
studies
images: www.amnh.oig., wikipedia