Conference PaperPDF Available

Brain scaling in ants: body to brain size ratio

Authors:

Abstract and Figures

Ants have great size diversity and are one of the most dominant insect groups in the world. Within the leaf-cutting ants of the neotropics, the genus Atta exhibits a large degree of physical polymorphism, with workers ranging in size from only a few millimeters in length to over two centimeters. We used this size diversity in Atta colombica to study the brain to body size ratio within a single species, thus eliminating phylogenetic considerations when making comparisons across species. We also examined how different neuropil in the brain scale to overall brain and body size. Although the overall brain size increases with body size the ratio is not linear. We discuss how physical constraints on brain size may influence brain structure and behavior, how overall body size may influence this ratio and how it influences division of labor. We compare our data to theoretical predictions of how brain size scales to body size.
Scatterplots showing how brain size in ants scales to body size. a Allometry for individuals of a leaf-cutter ant, Atta colombica , based on volumetric brain reconstructions (n = 48), with an overall regression of y = 0.3503x + 17.583 (R 2 = 0.9301) (solid gray line). A piecewise regression analysis with a breakpoint at 1.4 mg body mass yields 2 significantly different regressions: allometry for larger ants (black symbols) is described by y = 0.2919x + 17.753 (R 2 = 0.9122) (blue line), and that for the smaller ones (gray symbols) by y = 0.6003x + 17.469 (R 2 = 0.7354) (red line). b Allometric relationships for 70 ant species (n = 261 ants), with an overall regression model of y = 0.5972x-3.0419 (R 2 = 0.9389) (gray line). A piecewise regression with a breakpoint at 0.9 mg body mass yields 2 significantly different regressions: allometry for larger ants (black symbols, and open blue circles) is described by y = 0.5506x-2.9446 (R 2 = 0.8477) (dashed blue line), and that for the smaller ones (gray symbols, and open red circles) by y = 0.802x-2.8089 (R 2 = 0.9359) (red line). Open circles indicate polymorphic species (see table 1). c Interspecific allometry for ants using a mean value for each species for those taxa represented by more than 1 individual, with an overall regression model of y = 0.671x-3.0582 (R 2 = 0.9731) (gray line). A piecewise regression with a breakpoint at 0.9 mg body mass yields 2 significantly different regressions: allometry for larger ants (black symbols) is described by y = 0.6692x-3.0681 (R 2 = 0.9258) (dashed blue line), and that for the smaller ones (gray symbols) by y = 0.7961x2.8451 (R 2 = 0.9557) (red line). d Scatterplot of the ratio of brain:body mass against body mass for 70 ant species.
… 
Content may be subject to copyright.
Fax +41 61 306 12 34
E-Mail karger@karger.ch
www.karger.com
Original Paper
Brain Behav Evol 2011;77:5–13
DOI: 10.1159/000322530
The Allometry of Brain Miniaturization
in Ants
MarcA.Seid a ArmandoCastillo a WilliamT.Wcislo a
Smithsonian Tropical Research Institute, Panama , Republic of Panama
Introduction
Haller’s Rule holds that the brains of smaller animals
are proportionally larger than those of large-bodied
fo rms [s ee Rensch, 19 48] . Thi s a llome tr ic rel at ion ship be-
tween body and brain size has been documented exten-
sively for vertebrates [Cuvier, 1845; Harvey and Krebs,
1990; Hanken and Wa ke, 1993; Striedter, 2005; Gonzalez-
Voyer et al., 2009]. In contrast, relatively little is known
about brain allometry for the numerous invertebrate taxa
with extremely small body sizes, such as tardigrades
[Zantke et al., 2008] or Arthropoda (insects, mites and
spiders) [see also Rensch, 1948; Cole, 1985]. This dearth
of information is surprising because such small animals
dominate eart h’s biodiversity [Grimaldi and Engel, 20 04],
and long ago Darwin [1871, p. 145] called attention to the
ant brain as being ‘mar vellous’, because such an ‘extreme-
ly small’ mass of nervous tissue could generate ‘extraor-
dinary mental activity. Kern [1985] presented data on
brain and body mass for 36 species in 8 orders of insects,
but did not statistically analyze the allometric relation-
ships. The most extensive allometric study on inverte-
brate brains presented data from 10 ant species, albeit
with limited taxon sampling: 5 species were in the genus
Cataglyphis and 5 were from 2 other genera (all from 1
subfamily, Formicinae) [Wehner et al., 2007]. That study
demonstrated that the allometric scaling component for
ants (b = 0.57) was similar to that for birds (b = 0.58) and
reptiles (b = 0.54) but significantly different from that of
Key Words
Allometry Ants Brain evolution Miniaturization
Polymorphism
Abstract
Extensive studies of vertebrates have shown that brain size
scales to body size following power law functions. Most
animals are substantially smaller than vertebrates, and ex-
tremely small animals face significant challenges relating
to nervous system design and function, yet little is known
about their brain allometry. Within a well-define d monophy-
letic taxon, Formicidae (ants), we analyzed how brain size
scales to body size. An analysis of brain allometry for indi-
viduals of a highly polymorphic leaf-cutter ant, Atta colom-
bica, shows that allometric coefficients differ significantly
for small ( ! 1.4 mg body mass) versus large individuals (b =
0.6003 and 0.2919, respectively). Interspecifically, allometric
patterns differ for small ( ! 0.9 mg body mass) versus large
species (n = 70 species). Using mean values for species, the
allometric coefficient for smaller species (b = 0.7961) is sig-
nificantly greater than that for larger ones (b = 0.669). The
smallest ants had brains that constitute 15% of their body
mass, yet their brains were relatively smaller than predicted
by an overall allometric coefficient of brain to body size. Our
comparative and intraspecific studies show the extent to
which nervous systems can be miniaturized in taxa exhibit-
ing behavior that is apparently comparable to that of larger
species or individuals. Copyr ight © 2011 S. K arger AG, Basel
W.T. Wci slo
Smithsonian Tropical Res earch Institute
Apartado 0843-03092
Balboa, Panama (Republic of Pana ma)
Tel. +507 212 8128, Fax +507 212 8148, E-Mai l WcisloW
@ si.edu
© 2011 S. Karger AG, Basel
0006–8977/11/0771–0005$38.00/0
Accessible online at:
www.karger.com/bbe
Seid /Castillo /Wcislo
Brain Behav Evol 2011;77:5–13
6
mammals (b = 0.77). Yet the smallest species reported in
that study was 2.5 mg, which is relatively large in com-
parison to many arthropods. In contrast, there are de-
ta iled volumetric bra in studie s of par ticular beetle (Cole-
optera) or Strepsiptera species with extremely small body
sizes [e.g. Beutel et al., 2005; Grebennikov, 2008; Polilov,
2008; Polilov and Beutel, 2010], but there are no data for
closely related large-bodied forms.
Here we ask whether the scaling relationships ob-
served in other taxa hold for animals with very small
body sizes compared with large-bodied relatives? We
studied intraspecific brain scaling in the leaf-cutter ant,
Atta colombica, in which body mass spans 3 orders of
magnitude, and polymorphic workers differ in behavior
and physiology [Weber, 1972; Hölldobler and Wilson,
1990]. Furthermore, Atta ants have a diphasic cephalic
allometry [Wilson, 1953], so we also tested for diphasic
allometry in brain volume. Interspecifically, we extend-
ed an earlier study of brain scaling in ants [Wehner et
al., 2007] by substantially increasing the number of spe-
cies (n = 70 vs. 10); the taxonomic coverage (31 genera
from 5 subfamilies and 2 informal groupings vs. 3 gen-
era from 1 subfamily); and we included species that are
over 60 times smaller (0.039 vs. 2.5 mg body mass),
spanning approximately 4 orders of magnitude among
species.
Materials and Methods
Ant Collections
For the intraspecific study, specimens were collected from a
single nest of Atta colombica in Gamboa, Colon Province, Repub-
lic of Panama . There is extensive continuous size variation among
Atta workers [Weber, 1972], and to sample the full range of size
variation we dug into various fungus chambers and collected
workers in or near t he gardens, along w ith soldiers; newly emerged
callow workers were excluded, but otherwise the ages of individu-
als were unknown. Individuals of 70 ant species were collected
either from queen-right laboratory colonies or as individual for-
agers, either in the vicinity of Gamboa, or near Gainesville, Fla.,
USA ( table1 ). Voucher specimens are deposited in the Dry Refer-
ence Collection of the Smithsonian Tropical Research Institute,
and the Museo de Invertebrados ‘Graham Fairchild’ de la Univer-
sidad de Panamá.
Interspecific Body and Brain Measurements
In most a nt species there is conti nuous var iation in worker size
[Hölldobler and Wilson, 1990], but some species are monomor-
phic (i.e. a single mode in worker body size distribution), while
others are polymorphic (i.e. multiple modes in the distribution of
worker body size). It may be problematic to compare workers hav-
ing different social roles among different species (e.g. large sol-
diers vs. small workers), or monomorphic and polymorphic spe-
cies. Furthermore, the sampling among taxa was uneven because
we included multiple individuals for the polymorphic species in
order to capture the full size range in polymorphic species ( ta-
ble1 ). To address these 2 problems, we conducted 1 set of analyses
using the full data set (n = 261 ants from 70 species), and a second
set of analyses using mean values for each species (n = 70); we re-
fer to these as full and reduced data sets, respectively.
Histological Brain Sectioning and Volumetric Reconstructions
for Atta colombica
Ants were weighed using a n AND GR-202 mic roba lance (ac-
curacy to 0.01 mg). The brain of each ant was quickly removed
from the head capsu le and immediately placed i n fixative (6% glu-
taraldehyde, 2% paraformaldehyde in 0.1
M cacodylate buffer) in
preparation for standard histological sectioning. After fixation
for 12–24 h, the brains were rinsed in cacodylate buffers and post-
fixed in 1–1.5% osmium tetroxide for 2–3 h. The brains were then
rinsed in buffer followed by H
2 O and dehydrated in DMP and
acetone in preparation for embedding in Epon . Brains were in-
filt rated in Epon, by first placing t hem in a 50/50 mixture of Epon/
acetone and then transferring them to 100% Epon. They were
then embedded in Epon in Beem capsules and cured at 60
° C
overnight. Embedded brains were sectioned in a microtome (Mi-
crom HM 355s) at 5- m sections using stainless steel disposable
knives. Serial sections were kept in order and placed individually
on glass slides and then stained with toluidine blue. Coverslips
were then placed over the sections using Permamont and the
sections were photographed using a Nikon 8700 camera at-
tached to a Nikon Eclipse E600 compound microscope. Serial
digital sections were then imported into a computer, and were
traced, a ligned and stacked usi ng the program Reconstr uct [Fiala,
2005] to calculate the 3-D volume of each brain.
Measurements of Brain Mass for Interspecific Comparisons
For the interspecific study we used brain mass as a measure of
size. Collected ants were weighed using a Sartorius CPA2P mi-
crobalance after their removal from laboratory colonies or usu-
ally w ithin 24 h after collection from the f ield. Collected ants were
dissected under cold Ringer’s solution (150 m
M NaCl, 24 m M KCl,
7.0 m
M CaCl2, 4.0 m M MgCl2, 5.0 m M HEPES buffer, and 131 m M
sucrose, pH = 7.0). The brain, including both the supra- and sub-
esophageal ganglion and all sensory lobes, was quickly removed
from the head c apsule, usual ly in less than 1 m in, and then cleaned
of all tracheae and fat bodies. Each brain was then placed on a
small piece of tared Parafilm within a small droplet of Ringer’s
solution. The Ringer’s solution was wicked away using finely
twisted pieces of Kimwipes and the brain was weighed within
4 s. To assess weight loss due to water evaporation from exposed
brains, we measured weight loss through time for 5 ants of differ-
ent body sizes. The steepest rate of water loss occurred within the
first 20 s following removal from Ringer’s solution (data not
shown). We used data points for the first 20 s starting at the time
we could detect weight loss, given the 1- g resolution of the bal-
ance, to calculate the slope of the rate of weight loss from a linear
regression, a nd we took thi s to be the maximu m rate loss. We used
this worst-case slope to calculate the expected loss of weight over
the interval needed to prepare and weigh the specimen, and then
expressed this weight loss as a percentage of total brain mass
( fig.1 ).
Ant Brain Allometry Brain Behav Evol 2011;77:5–13
7
Genus Morph
category
Body size
rangea, mg
Brain size
rangea, mg
Mean brain
sizea, mg
n
Myrmicinae
Acromyrmex echinatior P 0.658–22.244 0.006–0.255 0.127 20
Apterostigma sp. 1 M 0.510–0.737 0.041–0.042 0.042 3
Apterostigma sp. 2 M 0.724–0.750 0.048–0.054 0.051 3
Apterostigma sp. 3 M 2.280–2.482 0.080–0.084 0.082 3
Atta colombica P 0.410–49.546 0.033–0.238 0.115 17
Cyphomyrmex cornutus M 0.347–0.433 0.025–0.032 0.029 3
Cyphomyrmex longiscapus M 0.347–0.429 0.023–0.035 0.028 3
Cyphomyrmex muelleri M 0.457–0.607 0.032–0.035 0.033 3
Cyphomyrmex sp. 1 M 0.385–0.435 0.027–0.035 0.030 3
Mycetophylax sp. M 1.399–1.574 0.044–0.050 0.047 3
Mycrocepurus smithii M 0.302–0.338 0.021–0.024 0.022 3
Myrmicocrypta cf. ednaella M 0.310–0.329 0.020–0.029 0.023 3
Sericomyrmex sp. M 1.127–1.514 0.048–0.058 0.053 3
Trachymyrmex coniktzi (1) M 0.863–0.939 0.045–0.050 0.047 3
Trachymyrmex cornetzi (2) M 1.115–1.356 0.051–0.074 0.060 3
Trachymyrmex sp. 1 M 1.117–1.248 0.048–0.052 0.050 3
Trachymyrmex sp. 2 M 2.149–2.373 0.062–0.074 0.070 3
Trachymyrmex zeteki M 1.971–2.282 0.061–0.066 0.064 3
Cephalotes atratus P 21.025–42.506 0.341–0.428 0.417 5
Cephalotes sp. 1 P 3.160–11.560 0.100–0.190 0.140 6
Cephalotes umbraculatus M 7.98 0.160 0.160 1
Crematogaster sp. M 0.795 0.044 0.044 1
Megalomyrmex sp. 1 M 0.137 0.010 0.010 1
Megalomyrmex sp. 2 M 1.147–1.231 0.044–0.051 0.057 2
Monomorium floricola M 0.065 0.006 0.006 1
Monomorium trageri M 0.098–0.104 0.009–0.011 0.010 3
Pheidole obscurithorax D 0.539–0.569 0.029–0.031 0.030 3
Pheidole sp. 1 D 0.093 0.008 0.008 1
Pheidole sp. 2 D 0.234 0.016 0.016 1
Pheidole sp. 3 D 0.069 0.008 0.008 1
Pheidole sp. 4 D 0.102 0.011 0.011 1
Pheidole sp. 5 D 0.254 0.020 0.020 1
Pheidole sp. 6 D 0.122 0.013 0.013 1
Pheidole sp. 7 D 0.873–1.181 0.047–0.055 0.049 3
Pheidole sp. 8 D 0.964–1.365 0.042–0.051 0.047 3
Pheidole sp. 9 D 0.471–0.587 0.028–0.029 0.028 3
Pogonomyrmex badius P 3.503–40.379 0.112–0.240 0.139 20
Solenopsis sp. 1 M 0.086–0.111 0.010–0.010 0.010 2
Solenopsis sp. 2 M 0.473 0.031 0.031 1
Wasmannia auropunctata M 0.109 0.007 0.007 1
Ectatomminae
Ectatomma ruidum M 8.184–12.878 0.208–0.232 0.220 2
Ectatomma tuberculatum M 15.950–21.090 0.380–0.400 0.387 3
Gnamptogenys sp. 1 M 0.576 0.028 0.028 1
Gnamptogenys sp. 2 M 9.873 0.244 0.244 1
Dolichoderinae
Azteca sp. 1 M 0.717 0.051 0.051 1
Azteca sp. 2 M 1.743 0.072 0.072 1
Dolichoderus sp. M 3.963–5.202 0.135–0.140 0.138 3
Tampinoma melanocephalum M 0.132 0.011 0.011 1
Brachymyrmex sp. 1 M 0.039–0.049 0.005–0.007 0.006 3
Brachymyrmex sp. 2 M 0.064–0.085 0.006–0.007 0.006 3
Tab le 1. List of taxa included in allometric analyses, with ranges for body and brain sizes, and mean brain size
Seid /Castillo /Wcislo
Brain Behav Evol 2011;77:5–13
8
Statistical Methods
We used the statistical package R for the piecewise regression
analyses, as well as for the comparisons of slopes by utilizing the
‘smatr’ library [Crawley, 2007]. Piecewise regression is a statistical
method to split a si ngle linear reg ression to assess whet her a 2-slope
model, or one with more slopes, would better fit the data than a
1-slope model [McGee and Carleton, 1970], and is commonly used
in identifying different growth trajectories for dimorphic or poly-
morphic phenotypes in insects [e.g. Eberhard and Gutierrez, 1991;
Eberhard et al., 2000]. The breakpoint is the point at which the tra-
jectory of 1 mor photype changes to another. Breakpoints are iden-
tified by analyzing multiple models and selecting the one with the
lowest residual standard errors (RSE) as providing the best fit. To
access the location of the breakpoint, we fitted models for different
breaks in the data using R [Crawley, 2007], and then visually in-
spected the RSE. To assess the robustness of the breakpoint, we
repeated the regression analyses manually at each of 4 additional
breakpoints, at 0.1- and 0.2-mg increments above our statistically
identified breakpoint, and at 0.1 and 0.2 mg below this initial
breakpoint. For each of the 4 new breakpoints we recalculated R
2
values and RSE. In each comparison the original breakpoint had
the highest R
2 and lowest RSE values. We also tested whether a
Genus Morph
category
Body size
rangea, mg
Brain size
rangea, mg
Mean brain
sizea, mg
n
Formicinae
Camponotus sericeiventri P 37.39 0.440 0.440 1
Camponotus sp. 1 M 5.0525 0.184 0.184 1
Camponotus sp. 2 P 7.720–36.540 0.190–410 0.295 13
Camponotus sp. 3 P 13.940–37.390 0.310–0.360 0.304 9
Paratrechina longicornis M 0.342 0.028 0.028 1
Dorylomorphs
Eciton burchellii P 1.503–33.395 0.079–0.305 0.202 31
Nomamyrmex sp. P 6.555–17.234 0.169–0.238 0.204 4
Poneromorphs
Odontomachus bauri M 15.626–15.939 0.292–0.332 0.312 2
Odontomachus brunneus M 5.828–6.557 0.190–0.193 0.192 2
Odontomachus hastatus M 22.970–27.680 0.410–0.460 0.430 3
Pachycondyla apicalis M 39.685–41.749 0.606–0.711 0.659 2
Pachycondyla obscuricornis M 15.129–15.226 0.369–0.424 0.397 2
Pachycondyla sp. 1 M 9.7 0.190 0.190 1
Pachycondyla sp. 2 M 4.358–4.599 0.148–0.162 0.155 2
Pachycondyla sp. 3 M 2.723 0.075 0.075 1
Pachycondyla villosa M 41.480–51.390 0.470–0.521 0.503 3
Paraponera clavata M 153.89–183.68 1.620–1.750 1.730 3
Pseudomyrmecinae
Pseudomyrmex sp. 1 M 2.889 0.135 0.135 1
Pseudomyrmex sp. 2 M 0.460–0.537 0.038–0.050 0.045 3
Pseudomyrmex sp. 3 M 4.850–5.606 0.259–0.281 0.277 3
P = Polymorphic; M = monomorphic; D = dimorphic, but only the minor subcaste was used.
a Range and mean not given for these taxa represented by singletons.
Table 1 (conti nued)
0
Estimated maximum percent weight loss
0.319 0.263 0.133
Brain weight (mg)
0.058 0.021
0.5
1.0
1.5
2.0
2.5
Fig. 1. Estimated maximum weight loss due to water evaporation
as a function of brain weight during the time needed to prepare a
specimen.
Ant Brain Allometry Brain Behav Evol 2011;77:5–13
9
model of 3 slopes would significantly improve the fit, because of
the well-known observation that more complex models better fit
the data than less complex ones [see discussion in Striedter, 2005].
We used major axis regression to account for error in both the
x-axis and y-axis, and comparisons of slopes were performed as
outlined by Warton et al. [2006], using the sample correlation be-
tween residuals and fitted values. We calculated phylogenetically
independent contrasts to account for an expected lack of indepen-
dence associated with phylogenetic structure [Felsenstein, 1985;
Harvey and Pagel, 1991; Ricklefs and Starck, 1996], using log-
tr ansforme d da ta a nalyz ed w it h th e PDAP mo dule [Mi dfo rd et al .,
2005] in MESQUITE v. 2.72 [Maddison and Maddison, 2007]. The
phylogeny from Brady et al. [2006] was used to create a tree at the
generic level for the ta xa in our analysis. Ants with in a single genus
were designated in the tree as an unresolved polytomy for this
analysis. We set branch lengths to 1, and subtracted 1 degree of
freedom for each polytomy in our analysis to yield our phyloge-
netic independent contrast slope for our data [Harvey and Pagel,
1991]. In additional analyses, we removed all polymorphic species
[ sensu Hölldobler and Wilson, 1990], and in cases where there
were multiple individuals per species, we calculated a mean value
for each species’ brain and body mass, and repeated the analyses.
R e s u l t s
The allometric relationship between brain volume and
body mass was significant for individuals of the highly
polymorphic species Atta colombica (n = 48; F
1,46 = 612.2,
p ! 0.0001) ( fig.2 a). A piecewise regression model with 2
regression equations fitted the data significantly better
than a single-slope model, with a breakpoint at 1.4 mg
body mass (ANOVA, F
1,2 = 20.024, p ! 0.0001; 2-slope
model – R
2 = 0.9634 and RSE = 0.09517; 1-slope model –
R
2 = 0.9301 and RSE = 0.1286); a model with 3 slopes was
not significantly different from the 2-slope model (F
1,2 =
1.3937, p = 0.2594). The RSE for the 8 0.1 and 8 0.2 incre-
mental steps from this breakpoint (i.e. body masses of 1.2,
1.3, 1.5, and 1.6 mg) were 0.1149, 0.1012, 0.1009, 0.1009,
respectively, and all were greater than the RSE at the 1.4
mg breakpoint. The allometric coefficient (the slope of
the model) from the 2-slope model was significantly
greater for the set of small A. colombica ( ! 1.4 mg) than
the coefficient for larger individuals, or from the single-
slope model (BLR = 7.8269; p = 0.0052; BLR = 5.659, p =
0.0017, respectively; fig.2 a). The coefficient for the larger
A. colombica was significantly different from that of the
single-slope model (BLR = 6.6981, p = 0.0096).
Similar patterns of brain scaling were observed among
70 species of ants (n = 261 individuals) using mass as a
measure of brain size ( fig.2 b). The allometric relationship
was significantly different for ant species above and be-
low a breakpoint of 0.9 mg body mass ( fig.2 b). A piece-
wise regression model with 2 equations fit the data sig-
nificantly better than a single-slope model (2-slope mod-
el: R
2 = 0.9475 and RSE = 0.2568; single-slope model –
R
2 = 0.9389 and RSE = 0.276; F
1,2 = 21.07, p ! 0.0001,
fig.2 b), but a model with 3 slopes did not improve the fit
(F
1,2 = 1.5871, p = 0.2065). The RSE for the 8 0.1 and
8 0.2 incremental steps from the 0.9 mg breakpoint (i.e.
body masses of 0.7, 0.8, 1.0, and 1.1 mg) were 0.2572,
0.2570, 0.2574, and 0.2580, respectively, and all were
greater than the RSE at the 0.9-mg breakpoint. As with
the intraspecific analysis, the allometric coefficient for
smaller ants ( ! 0.9 mg) from the 2-slope model was great-
er than that from the piecewise slope for the larger ants,
or from the single-slope model [Bartlett-corrected likeli-
hood ratio (BLR) = 54.67, p ! 0.0001; BLR = 51.01, p !
0.0001, respectively; fig.2 b].
Using a reduced data set based on mean values for spe-
cies, again a 2-slope model with a breakpoint at 0.9 mg
bo dy w eig ht f itted the dat a signifi cantly better than a sin-
gle-slope model (2-slope model – R
2 = 0.9762 and RSE =
0.2071; single-slope model – R
2 = 0.972 and RSE = 0.2173;
F
1,2 = 4.443, p = 0.016; fig.2 c), but a 3-slope model did not
(F
1,2 = 0.289, p = 0.7433). The RSE for the 8 0.1 and 8 0.2
incremental steps (i.e. body masses of 0.7, 0.8, 1.0, and 1.1
mg) were 0.2113, 0.2072, 0.2074, 0.2088, respectively, and
all were greater than the RSE at the 0.9-mg breakpoint.
The slope for the smaller ants ( ! 0.9 mg) is again signifi-
cantly steeper t han for larger ants or from the single-slope
model (BLR = 5.7094, p = 0.016; BLR = 11.9467, p = 0.21,
p = 0.0005, respectively; fig. 2 c). Taking phylogenetic
structure into account, the slope was not significantly
different (sample correlation between residuals and fit-
ted values = –0.016, p = 0.8979). Plotting the ratio of
brain:body mass against body mass yields a steep expo-
nential decay function (y = 0.0473e
–0.3876x , R 2 = 0.8347)
su ch th at th e pr opor tion o f body m ass c ompri sed of bra in
is substantially greater for the smallest ants; the smallest
ants ( Brachymyrmex sp.) had brains that accounted for
1 15% of their body mass ( fig.2 d). Small ants ( ! 0.9 mg)
occur in all the major taxa that we sampled, except for
those in the informal groups dorylomorphs and ponero-
morphs ( fig.3 ), so there is no basis to suspect that the dif-
ferences are taxon dependent rather than size dependent.
Discussion
Even though the smallest ants in our study had rela-
tively massive brains, constituting 15% of their body
mass, we observed no morphological modifications of
Seid /Castillo /Wcislo
Brain Behav Evol 2011;77:5–13
10
17.0
–1 0 1 2
In (body weight [mg])
In (brain volume [μm3])
345
17.5
18.0
18.5
19.0
a
–5.5
–3.5 –2.5 –1.5 –0.5 0.5 1.5 2.5
In (body weight [mg])
In (brain weight [μg])
3.5 4.5 5.5
–4.5
–2.5
–3.5
–1.5
–0.5
0.5
b
–5.5
–3.5 –2.5 –1.5 –0.5 0.5 1.5 2.5
In (body weight [mg])
In (brain weight [μg])
3.5 4.5 5.5
–4.5
–2.5
–3.5
–1.5
–0.5
0.5
c
0
–4 –2 0 2
In (body weight [mg])
Brain:body mass
46
0.04
0.08
0.12
0.02
0.06
0.10
0.16
0.14
d
Fig. 2. Scatterplots showing how brain size in ants scales to body
size.
a Allometry for individuals of a leaf-cutter ant, Atta colom-
bica , based on volumetric brain reconstructions (n = 48), with an
overall regression of y = 0.3503x + 17.583 (R
2 = 0.9301) (solid gray
line). A piecewise regression analysis with a breakpoint at 1.4 mg
body mass yields 2 significantly different regressions: allometry
for larger ants (black symbols) is described by y = 0.2919x + 17.753
(R
2 = 0.9122) (blue line), and that for the smaller ones (gray sym-
bols) by y = 0.6003x + 17.469 (R
2 = 0.7354) (red line). b Allometric
relationships for 70 ant species (n = 261 ants), with an overall re-
gression model of y = 0.5972x – 3.0419 (R
2 = 0.9389) (gray line).
A piecewise regression with a breakpoint at 0.9 mg body mass
yields 2 significantly different regressions: allometry for larger
ants (black symbols, and open blue circles) is described by y =
0.5506x – 2.9446 (R
2 = 0.8477) (dashed blue line), and that for
the smaller ones (gray symbols, and open red circles) by y =
0.802x – 2.8089 (R
2 = 0.9359) (red line). Open circles indicate
polymorphic species (see table 1).
c Interspecific allometry for
ants using a mean value for each species for those taxa represent-
ed by more than 1 individual, with an overall regression model of
y = 0.671x – 3.0582 (R
2 = 0.9731) (gray line). A piecewise regres-
sion with a breakpoint at 0.9 mg body mass yields 2 significantly
different regressions: allometry for larger ants (black symbols) is
described by y = 0.6692x – 3.0681 (R
2 = 0.9258) (dashed blue line),
and that for the smaller ones (gray symbols) by y = 0.7961x –
2.8451 (R
2 = 0.9557) (red line). d Scatterplot of the ratio of
brain:body mass against body mass for 70 ant species.
Ant Brain Allometry Brain Behav Evol 2011;77:5–13
11
head shape or size to accommodate these proportionally
large brains. Other small animals facing similar chal-
lenges from brain miniaturization have morphological
adaptations in which the brain invades other body parts,
such as the prothorax in larval insects [Beutel et al., 2005;
Grebennikov, 2008], and the coxae in legs of spiders
[Quesada et al., in prep.]. Although their brains are dis-
proportionately large as expected by Haller’s Rule, ex-
tremely small ants, both intra- and interspecifically, have
br ai ns t hat are smal ler than wou ld b e exp ected if t hey fol-
lowed the allometric slope of larger ants. We do not know
whether this shift represents a compensatory mechanism
to cap energetic costs, or is related to constraints on head
morphology. The latter alternative seems unlikely, how-
ever, given that macrocephaly has evolved repeatedly in
ants [Hölldobler and Wilson, 1990], suggesting that brain
size is not necessarily limited by head size.
Diphasic allometries have been demonstrated for oth-
er traits in insects [Niven and Scharlemann, 2005; Eber-
hard et al., 2000], including studies of cephalic allometry
in ants [Wilson, 1953], but were unknown for brain scal-
ing. Wilson [1953] hypothesized that diphasic allometry
reflects a mechanism that helps to stabilize the head size
of very small workers, while enabling the production of
very different workers with only small differences in
body size. We speculate that for polymorphic fungus-
growing ants, such as Atta, it may be advantageous to
have a smaller-than-expected head size at the small end
of the size spectrum because of the need for tiny workers
to move about within the interstices of the fungus garden
[Weber, 1972], while maintaining the neural capabilities
to process information relating to the health of the fungal
cultivar, detecting the presence of pathogens in the fun-
gal gardens, and implementing disease-control measures
[e.g. Fernández-Marín et al., 2009]. Small versus large
workers of 2 species of Atta differ in the relative size
of brain components [Kleineidam et al., 2005], and these
differences may be related to differential behavioral re-
sponses in the 2 size classes [Kleineidam et al., 2007]. A
detailed volumetric analysis of brain region size relative
to head size will elucidate how brain regions influence
total brain size [Seid, Elizondo and Wcislo, in prep.].
In light of known behavioral differences among sub-
castes of Atta workers [Weber, 1972; Hölldobler and Wil-
son, 1990], and a report of diphasic cephalic allometry in
A. texana [Wilson, 1953], we expected an allometric shift
in brain allometry for A. colombica. A similar allometric
shift in the interspecific study was surprising and may
point to a general rule governing how ant brains are con-
structed beyond a critical size threshold. Our finding that
Myrmicinae
Formicinae
Ectatomminae
Dolichoderinae
Pseudomyrmecinae
Dorylomorphs
Poneromorphs
Range of absolute brain mass (mg)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fig. 3. Range of absolute brain mass for major subfamilies and
taxonomic groupings used in this study. The phylogeny is taken
f r o m B r a d y e t a l . [ 2 0 0 6 ] .
Seid /Castillo /Wcislo
Brain Behav Evol 2011;77:5–13
12
References Beutel RG, Pohl H, Hunefeld F (2005): Strep-
sipteran brains and effects of miniaturiza-
tion (Insecta). Arth Struct Develop 34:
301
313.
Bonner JT (2006): Why Size Matters: From Bac-
teria to Blue Whales. Princeton, Princeton
University Press.
Brady SG, Schultz TR, Fisher BL, Ward PS
(2006): Eva luating alter native hypot heses for
the early evolution and diversif ication of
ants. Proc Natl Acad Sci USA 103:
18172–
18177.
Chittka L, Niven JE (2009): Are bigger brains
better? Curr Biol 19:R995–R1008 .
Cole BJ (1985): Size and behavior in ants: con-
straints on complexity. Proc Natl Acad Sci
USA 82:
8548–8551.
Crawley MJ (2007): The R Book. Chichester,
Wiley.
Cuvier G (1845): Leçons danatomie comparée.
Vol 3: contenant le système ner veux et les or-
ganes des sens, ed 2. Paris , Fortin, Masson et
Cie.
the allometric rules governing ant brain size change at
extremely small body sizes, both within and among spe-
cies, was derived from 2 methods of measuring brain size,
and therefore it is unlikely to be an artifact of the tech-
niques u sed . An allometric co eff ici ent a s hi gh a s 0.8 for
small ants differs significantly from that reported previ-
ously for a survey of 10 ant species [Wehner et al., 2007],
which gave a scaling coefficient of 0.567. Our study and
that of Wehner et al. [2007] included the same substruc-
tures in calculating total brain mass, so it is unlikely that
any methodological differences account for the discor-
dant findings. Indeed, for the larger ants in our study
( 1 0.9 mg), the scaling coefficient from the full data set
(b = 0.5506; fig.2 b) was nearly identical to that of Weh-
ner et al. [2007]. Notably, the smallest ants used by Weh-
ner et al. [2007] were substantially larger than the break-
point (0.9 mg) in our study, which could account for the
discrepancy between these studies. To our knowledge,
extremely small ants ( ! 0.9 mg body mass) are the only
animals known to have a brain allometric coefficient
comparable to that of mammals [White et al., 2009].
It is difficult to interpret the biological significance of
diphasic allometry in the interspecific comparison be-
cause so little is known about how brain size relates to
behavior in ants, but we speculate that it may be associ-
ated with energetics. The disproportionate investment in
brain mass by small animals implies disproportionately
high energetic costs, given that neuronal tissue is expen-
sive to maintain [Niven and Laughlin, 2008]. Diphasic
brain allometry should limit these energetic costs at very
small body sizes because it produces smaller brains than
expected from a monophasic allometry. An additional
mechanism to reduce costs or the size of the nervous sys-
tem involves relative investment in glia and neurons; lim-
ited data show that ants from 1 large species have more
glia processes than ants from a smaller species [pers. obs.].
An extremely small animal may pay a severe energetic
cost to maintain a disproportionately large CNS in order
to have information processing capabilities equivalent to
large-bodied species [Niven and Laughlin, 2008; Niven et
al., 2007]. Alternatively they may have evolved life history
or behavioral traits, or more elaborate peripheral sensory
systems, to reduce the need for relatively large and sophis-
ticated information processing systems, and hence mini-
mize energetic costs. Unfortunately, the functional conse-
quences of brain miniaturization are not well understood
and there are few data to distinguish between these 2 al-
ternatives. Both Snell-Rood et al. [2009] and Mares et al.
[2005] hypothesize that brain size may limit learning abil-
ities in insects [also Rensch, 1956, 1959]. Limited empiri-
cal studies on the behavior of extremely small animals
indicate that they do not suffer from inferior behavioral
capabilities [Eberhard, 2007; Hesselberg, 2010], and we are
not aware of any behavioral deficiencies in small ants.
Many animals with tiny brains express behavior compa-
rable to large-brained animals [Miklos, 1998; Chittka and
Niven, 2009] and ants are no exception, but more studies
are needed to understand compensatory mechanisms for
adapting to small body sizes and the energetic costs for
maintaining a relatively large brain in small-bodied ani-
mals.
Acknowledgements
We thank J. Bonner, L. Chittka, J. Douglass, W. Eberhard, H.
Fe rn án de z- Ma rí n, A. Fa rj i, J. N iv en , S . Ti er ne y, a nd R. Weh ne r f or
helpful comments on different drafts of the manuscript or useful
discussions; H. Fernández-Marín and G. Bruner for providing
some ants; L. Jiménez and L. Elizondo for laboratory assistance;
B. Turner for the use of a microbalance; and STRI staff, especial-
ly P. Galgani, for logistical support.
Funding was provided by generous support from the F.H.
Levinson Fund to the STRI Laboratory of Behavior and Evolu-
tionary Neurobiology (W.T.W., principal investigator), and by the
Smithsonian Institution’s Scholarly Studies Program (W.T.W.,
principal investigator). We are grateful to t he Autoridad Nacional
del Medio Ambiente (ANAM) of the Republic of Panama for re-
search and collecting permits.
Ant Brain Allometry Brain Behav Evol 2011;77:5–13
13
Da rwin C (1871): The Des cent of Man, and Selec-
ti on i n Rel at ion t o Se x. P rin cet on, P ri nce ton
University Press. (1981 reprint of 1871 ed).
Eb erhard WG (2007): Min iaturiz ed orb-weaving
spiders: behavioural precision is not limited
by small size. Pro Roy Soc Ser B 274:
2203–
2209.
Eberhard WG, Garcia JM, Lobo J (2000): Size-
specif ic defensive structures in a horned
weevil confirm a classic battle plan: avoid
fights with larger opponents. Proc Biol Sci
267:
1129 –1134 .
Eberha rd WG, Gutier rez EE (1991): Male dimor-
phisms in beetles and earwigs and the ques-
tion of developmenta l constrai nts. Evolution
45:
18–28.
Felsenstein J (1985): Phylogenies and the com-
parative method. Amer Nat 125:
1–15.
Fernández-Marín H, Zimmerman JK, Nash DR,
Boomsma J J, Wc islo WT (2009): Reduced bi-
ological control and enhance d chemical pest
management in the evolution of fung us-
farm ing in ants. Pro Roy S oc Ser B 276:
2263–
2269.
Fiala JC (2005): Reconstruct: a free editor for se-
rial s ection microsc opy. J Microsc 218:
52–61.
Gonzalez-Voyer A, Winberg S, Kolm N (2009):
Distinct evolutionary patterns of brain and
body size during adaptive radiation. Evolu-
tion 63:
2266–2274.
Grebennikov VV (2008): How small you can go:
factors limiting body miniaturization in
winged insects with a review of the pantrop-
ical genus Discheramocephalus and descrip-
tion of six new species of the smal lest beetles
(Pterygota: Coleoptera: Ptiliidae). Eur J En-
tomol 105:
313–328.
Grimaldi D, Engel MS (2004): Evolution of the
Insects. New York, Cambridge University
Press.
Hanken J, Wake DB (1993): Miniaturization of
body size: organismal consequences and
evolutionar y significance. Annu Rev Ecol
Syst 24:
501–519.
Harvey PH, Krebs JR (1990): Comparing brains.
Science 249:
140 –146 .
Harvey PH, Pagel MD (1991): The Comparative
Method in Evolutionary Biology. New York,
Oxford University Press.
Hes selberg T (2010): Ontogenetic changes i n web
design in two orb-web spiders. Ethology 116:
535–545.
Höl ldobler B, Wilson EO (1990): The Ants . Cam-
bridge, Harvard University Press.
Kern MJ (1985): Metabolic rate of the insect
brain in relation to body size and phylogeny.
Comp Biochem Physiol 81:
501–506.
Kleineidam CJ, Obermayer M, Halbich W,
Rössler W (2005): A macrog lomerulus in the
antennal lobe of leaf-cutting ant workers and
its possible functional significa nce. Chem
Senses 30:
383–392.
Kle ineidam CJ, Rössler W, Hölldobler B, Roces F
(2007): Percept ual dif ferences in tra il follow-
ing leaf-cutting ants relate to body size. J In-
sect Physiol 53:
1233–1241.
Maddison WP, Maddison DR (2007): Mesquite:
A Modular System for Evolutionary Analy-
sis. Version 2.0. http://mesquiteproject.org
Mares S, Ash L, Gronenberg W (2005): Brain al-
lometry in bumblebee and honeybee work-
ers. Brain Behav Evol 66:
50–61.
McGee VE, Carleton WT (1970): Piecewise re-
gression. J Am Stat Assoc 65:
1109 –1124 .
Midford PE, Garland T Jr, Maddison W (2005):
PDAP Package for Mesquite. Version 1.07.
http://mesquiteproject.org/pdap_mesquite/
Miklos GLG (1998): The evolution and modifi-
cation of brains and sensory systems. Daeda-
lus 127:197–216.
Niven JE, Scharlemann JPW (2005): Do insect
metabol ic rates at rest and duri ng flig ht scale
with body mass? Biol Lett 1:
346–349.
Niven JE, Anderson JC, Laughlin SB (2007): Fly
photoreceptors demonstrate energy-infor-
mation trade-offs in neural coding. PLoS
Biol 5:
828–840.
Niven JE, Laughlin SB (2008): Energy limitation
as a selective pressure on the evolution of
sensory systems. J Exp Biol 211:
1792 –1804 .
Polilov AA, Beutel RG (2010): Developmental
stages of the hooded beetle Sericoderus late-
ralis (Coleoptera: Corylophidae) with com-
ments on the phylogenetic position and ef-
fects of miniaturization. Arth Struct Devel-
op 39:
52–69.
Polilov AA (2008): Anatomy of the smallest Co-
leoptera, feat herwing beetles of the tribe
Nanosellini (Coleoptera, Ptiliidae), and lim-
its of insect miniaturization. Ent Rev 88:
26–
33.
Rensch B (1948): Histological change with evo-
lutionary changes of body size. Evolution 2:
218–230.
Rensch B (1956): Increase of learning capabilit y
with increase of brain size. Amer Nat 90:
81–
95.
Rensch B (1959): Evolution above the Species
Level. Ne w York , Columbia University P ress.
Ricklefs RE, Starck M (1996): Applications of
phylogenetically independent contrasts: a
mixed progress report. Oikos 77:
167–172 .
Roth G, Blan ke J, Ohle M (1995): Brain size and
morphology in miniaturized plethodontid
salamanders. Brain Behav Evol 45: 84–95.
Snel l-Rood EC, Papaj DR, Gronenber g W (2009):
Brain size: a global or induced cost of learn-
ing? Brain Behav Evol 73:
111–128.
Striedter GF (2005): Principles of Brain Evolu-
tion. Sunderland, Sinauer.
Warton DI, Wright IJ, Falster DS, Westoby M
(2006): Bivariate line-fitting methods for al-
lometry. Biol Rev 81:
259–291.
Weber NA (1972): Gardening Ants, The Attines.
Philadelphia, American Philosophical Soci-
ety.
Wehner R, Fukushi T, Isler K (2007): On being
small: brain allometry in ants. Brain Behav
Evol 69:
220–228.
White CR, Blackburn TM, Seymour RS (2009):
Phylogenetically informed analysis of the al-
lometry of mammalian basal metabolic rate
supports neither geometric nor quarter-
power scaling. Evolution 63:
2658–2667.
Wilson EO (1953): The origin and evolution of
polymorphism in ants. Q Rev Biol 28:
136–
156.
Zantke J, Wolff C, Scholtz G (2008): Three-di-
mensional reconstruction of t he central ner-
vous system of Macrobiotus hufelandi (Eu-
tardigrada, Parachela): implications for the
phylogenetic position of Tardigrada. Zoo-
morphology 127:
21–36.
Reproducedwith permissionofthe copyrightowner.Furtherreproduction prohibitedwithoutpermission.
ResearchGate has not been able to resolve any citations for this publication.
Article
John Tyler Bonner, one of our most distinguished and creative biologists, here offers a completely new perspective on the role of size in biology. In his hallmark friendly style, he explores the universal impact of being the right size. By examining stories ranging from Alice in Wonderland to Gulliver's Travels, he shows that humans have always been fascinated by things big and small. Why then does size always reside on the fringes of science and never on the center stage? Why do biologists and others ponder size only when studying something else--running speed, life span, or metabolism. Why Size Matters, a pioneering book of big ideas in a compact size, gives size its due by presenting a profound yet lucid overview of what we know about its role in the living world. Bonner argues that size really does matter--that it is the supreme and universal determinant of what any organism can be and do. For example, because tiny creatures are subject primarily to forces of cohesion and larger beasts to gravity, a fly can easily walk up a wall, something we humans cannot even begin to imagine doing.
Article
The recently described and originally monotypic genus Discheramocephalus Johnson, 2007 from the Solomon Islands is revised. Six new species are described, illustrated and keyed: Discheramocephalus brucei sp. n. (Cameroon), D. elisabethae sp. n. (Cameroon), D. mikaeli sp. n. (Tanzania), D. stewarti sp. n. (Bolivia), D. jarmilae sp. n. (Bolivia), D. minutissimus sp. n. (Indonesia). Adults of D. minutissimus have a body length of about 400-426 μm, which is at the lower limit among non-egg-parasitoid insects. Evidence is provided that an egg size large enough to produce a viable larva is the main factor limiting miniaturisation of female insects. Females and males of egg-parasitoids are able to overcome the 400 μm threshold and reach limits of 180 μm and 130 μm, respectively. Brain size is likely the second most important factor limiting miniaturisation in insects.
Article
The form of the relationship between the basal metabolic rate (BMR) and body mass (M) of mammals has been at issue for almost seven decades, with debate focusing on the value of the scaling exponent (b, where BMR is proportional to M(b)) and the relative merits of b= 0.67 (geometric scaling) and b= 0.75 (quarter-power scaling). However, most analyses are not phylogenetically informed (PI) and therefore fail to account for the shared evolutionary history of the species they consider. Here, we reanalyze the most rigorously selected and comprehensive mammalian BMR dataset presently available, and investigate the effects of data selection and phylogenetic method (phylogenetic generalized least squares and independent contrasts) on estimation of the scaling exponent relating mammalian BMR to M. Contrary to the results of a non-PI analysis of these data, which found an exponent of 0.67-0.69, we find that most of the PI scaling exponents are significantly different from both 0.67 and 0.75. Similarly, the scaling exponents differ between lineages, and these exponents are also often different from 0.67 or 0.75. Thus, we conclude that no single value of b adequately characterizes the allometric relationship between body mass and BMR.
Article
Analysis of 17 species from 6 families indicates that male dimorphisms in weapon design may be common, at least in horned beetles. -from Authors
Article
A difficult regression parameter estimation problem is posed when the data sample is hypothesized to have been generated by more than a single regression model. To find the best-fitting number and location of underlying regression systems, the investigator must specify both the statistical criterion and the search-estimation procedure to be used. The approach outlined in this article is essentially a wedding of hierarchical clustering and standard regression theory. As the name suggests, piecewise regression may be described as a method of finding that piecewise continuous function which best describes the data sample. Computational procedures and a fully-worked example, together with possible extensions, are provided.
Article
In the current resurgence of interest in the biological basis of animal behavior and social organization, the ideas and questions pursued by Charles Darwin remain fresh and insightful. This is especially true of The Descent of Man and Selection in Relation to Sex, Darwin's second most important work. This edition is a facsimile reprint of the first printing of the first edition (1871), not previously available in paperback. The work is divided into two parts. Part One marshals behavioral and morphological evidence to argue that humans evolved from other animals. Darwin shoes that human mental and emotional capacities, far from making human beings unique, are evidence of an animal origin and evolutionary development. Part Two is an extended discussion of the differences between the sexes of many species and how they arose as a result of selection. Here Darwin lays the foundation for much contemporary research by arguing that many characteristics of animals have evolved not in response to the selective pressures exerted by their physical and biological environment, but rather to confer an advantage in sexual competition. These two themes are drawn together in two final chapters on the role of sexual selection in humans. In their Introduction, Professors Bonner and May discuss the place of The Descent in its own time and relation to current work in biology and other disciplines.