Divergence in Proteins, Mitochondrial DNA, and Reproductive 
Compatibility Across the Isthmus of Panama BUSifl ? STOR 
Nancy Knowlton; Lee A. Weight; Luis Anibal Solorzano; DeEtta K. Mills; Eldredge 
Bermingham 
&%g?cg, New Series, Vol. 260, No. 5114 (Jun. 11,1993), 1629-1632. 
Stable URL: 
http://links.jstor.org/sici?sici=0036-8075%2819930611%293%3A260%3A5114%3C1629%3ADIPMDA%3E2.0.CO%3B2-%23 
Science is currently published by American Association for the Advancement of Science. 
Your use of the JSTOR archive indicates your acceptance of JSTOR' s Terms and Conditions of Use, available at 
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you 
have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and 
you may use content in the JSTOR archive only for your personal, non-commercial use. 
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at 
http://www.jstor.org/journals/aaas.html. 
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or 
printed page of such transmission. 
For more information on JSTOR contact j stor-info@umich.edu. 
?2003 JSTOR 
http://www.j stor.org/ 
TueNov 18 14:35:27 2003 
mm? 
ried out on the basis of only those species known to 
reach sizes >5 mm in any linear dimension, 
12. G. Rosenberg, Am. Malacol. Bull., in press. More 
than 1000 publications on western Atlantic mol- 
lusks were consulted in the compilation of this 
data base. 
13. W. D. Allmon, PalaiosB, 183 (1993). 
14. We estimate that the total shelled molluscan fauna 
of the Pinecrest is about 1050 species. We ob- 
tained this number by taking the estimated 801 
Pinecrest gastropods and bivalves and adding 
20% for species yet to be discovered and 5% for 
two less diverse classes, the chitons and scapho- 
pods (29). This number is the same as a widely 
cited estimate for the Pinecrest (9), which listed 
300 species, including 170 gastropods, and gave 
a total of about 1000 species. 
15. A. A. Olsson, in Miami Geological Society Fieldtrip 
Guidebook, R. D. Perkins, compiler (Miami, FL 
1968), p. 75. 
16. E. J. Petuch, Proc. Acad. Nat. Sci. Philadelphia 
134, 12 (1982). 
17. In 1977, the Florida fauna was tallied at 1085 
gastropods and bivalves (29); here we have 
tallied 1147 (827 gastropods and 320 bivalves), 
an increase of almost 6%. 
18. E. J. Petuch, Neogene History of Tropical Ameri- 
can Mollusks (Coastal Education and Research 
Foundation, Charlottesville, VA, 1988). 
19. J. K. Reed and P. M. Mikkelsen, Bull. Mar. Sci. 40, 
99 (1987); W. G. Lyons, Fla. Mar. Res. Inst. Publ. 
47(1989). 
20. S. M. Stanley, in Causes of Evolution, a Paleontolog- 
icalPerspective, R. M. Ross and W. D. Allmon, Eds. 
(Univ. of Chicago Press, Chicago, 1990), pp. 103- 
127. 
21. K. W. Flessa ef a/., in Patterns and Processes in 
the History of Life, D. M. Raup and D. Jablonski, 
Eds. (Springer-Verlag, Berlin, 1986) pp. 232-257. 
22. J B. C. Jackson, P. Jung, A. G. Coates, L. S 
Collins, Science 260, 1624 (1993). 
23. Eastern Pacific data from (4) as revised by C. 
Skoglund, Festivus (suppl.) 24, 1 (1992). 
24. T. M. Cronin, Ouat. Sci. Rev. 10, 175 (1991). 
25. L. W. Ward, R. H. Bailey, J. G. Carter, in The 
Geology of the Carolinas, J. W. Morton and V. A. 
Zullo, Eds. (Univ. of Tennessee Press, Knoxville, 
1991), pp. 274-289. 
26. S. M. Stanley, Paleobiology 12, 89 (1986); W. D. 
Allmon, Oxford Surv. Evol. Biol. 8, 219 (1992). 
27 B, W. Blackwelder, Palaeogeogr. Palaeoclimatol. 
Palaeoecol. 34, 87 (1981); T. M. Cronin, Geol. 
Soc. Am. Bull. 92, 812 (1981); H. J. Dowsett and 
T. M. Cronin, Geology 18, 435 (1990). 
28. W, D. Allmon, Palaeogeogr. Palaeoclimatol. 
Palaeoecol. 92, 41 (1992), and references therein. 
29. D. Nicol, Nautilus 91, 4 (1977). 
30. R. W. Portell, K. S. Schindler, G. M. Morgan, in 
"The Plio-Pleistocene stratigraphy and paleontol- 
ogy of southern Florida," T. M. Scott and W. D. 
Allmon, Eds., Fla. Geol. Surv. Spec. Publ. 36 
(1992), pp. 181-194. 
31. R. W. Portell, K. S. Schindler, D. S. Jones, Tulane 
Stud. Geol. Paleontol., in press. 
32. We thank A. Collazos, K. Ketcher, and S. Schellen- 
berg for assistance in data analysis and T. Cronin, J. 
Jackson, D. Jones, and L. Ward for comments on 
earlier drafts. This paper is University of Florida 
Contribution to Paleontology no. 415. 
19 January 1993; accepted 23 April 1993 
Divergence in Proteins, Mitochondria! DNA, and 
Reproductive Compatibility Across the 
Isthmus of Panama 
Nancy Knowlton,* Lee A. Weigt, Luis Anibal Solorzano, 
DeEtta K. Mills,t Eldredge Bermingham 
It is widely believed that gene flow connected many shallow water populations of the 
Caribbean and eastern Pacific until the Panama seaway closed 3.0 to 3.5 million years ago. 
Measurements of biochemical and reproductive divergence for seven closely related, 
transisthmian pairs of snapping shrimps (Alpheus) indicate, however, that isolation was 
staggered rather than simultaneous. The four least divergent pairs provide the best es- 
timate for rates of molecular divergence and speciation. Ecological, genetic, and geological 
data suggest that gene flow was disrupted for the remaining three pairs by environmental 
change several million years before the land barrier was complete. 
Ideographic isolation is thought to permit 
divergence and speciation by disruption of 
gene flow (J). Pairs of marine sister taxa 
separated by the Isthmus of Panama are 
ideal for studying these processes (2-5) 
because isolation of the Caribbean and the 
eastern Pacific is well dated and relatively 
recent (6, 7). This geological framework 
Smithsonian Tropical Research Institute, Apartado 
2072, Balboa, Republic of Panama. 
*To whom correspondence should be addressed. 
Alternate mailing address: Smithsonian Tropical Re- 
search Institute, Unit 0948, APO AA 34002-0948, 
United States. 
tPresent address: Route 2, Box 538, Joshua, TX 
76058. 
has prompted study of transisthmian sister 
taxa to test the accuracy of molecular clocks 
and to estimate the timing of other evolu- 
tionary events (3, 4, 8). It has been difficult 
to interpret inconsistencies and estimate 
possible phylogenetic differences in diver- 
gence rates (9), however, because of the 
limited number of taxa and characters stud- 
ied. To address these problems, we investi- 
gated divergence in allozymes, mitochon- 
drial DNA (mtDNA), and reproductive 
compatibility for seven shallow water trans- 
isthmian pairs of sister taxa in the snapping 
shrimp genus Alpheus. 
We used the  taxonomic literature  to 
identify transisthmian pairs that were spe- 
cifically and unambiguously described as 
each other's closest relatives on the basis of 
morphological criteria (10). Collections 
along the two coasts and adjacent islands of 
central Panama at depths less than 5 m 
revealed unrecognized sibling species in ad- 
dition to these pairs (11). In total, we 
examined 17 taxa (Table 1): two unambig- 
uous pairs (P4-C4, P5-C5), three triplets 
(P3-C3, P3-C3'; P6-C6, P6'-C6; P7-C7, 
P7'-C7), and one quartet (Pl-Cl, P1-C2, 
P2-C1, P2-C2). We used shared anatomi- 
cal and color pattern character states (12) 
to posit relations within the triplets and 
quartet. The result was seven morphologi- 
cally defined transisthmian sister species 
pairs (bold in Tables 1 and 2). 
For each taxon, we characterized allo- 
zymes by using conventional starch gel elec- 
trophoresis (13) and sequenced a segment of 
the mtDNA cytochrome oxidase I (COI) 
gene (14). Aggressive behavior was used as 
an estimate of behavioral components of 
reproductive compatibility . (15) because 
snapping shrimp attack heterospecific indi- 
viduals and all conspecifics except potential 
mates (16). We calculated genetic diver- 
gence between transisthmian pairs using 
Nei's D for allozymes and Kimura's corrected 
percent sequence divergence for mtDNA 
(17). We estimated divergence in behavioral 
compatibility by standardizing measures of 
tolerance and intolerance for transisthmian 
pairs against values observed in intraoce- 
anic, conspecific control matings (15). 
These three measures of divergence con- 
sistently support assignments of transisth- 
mian sister species pairs on the basis of 
morphology and color pattern. Within the 
P6' 
P6 
C6 
P3 
C3 
ca- 
ps 
- C5 
<rP 
C4 
P4 
?c 
P1 
C1 
r-P2 
L- C2 
-C7 P7' P7 
C7 
Fig. 1. Single most parsimonious phylogenetic 
tree constructed on the basis of mtDNA se- 
quences with PAUP (18). Transitions were giv- 
en one-quarter the weight of transversions 
(based on the fourfold greater,abundance of 
transitions than transversions in our data), and 
trees were rooted by the P7-P7'-C7 clade. 
Taxon codes are as in Table 1. 
SCIENCE   ?   VOL. 260   ?    11 JUNE 1993 1629 
triplets and quartet, transisthmian sister 
species pairs generally show greater repro- 
ductive compatibility, lower allozyme di- 
vergence, and lower mtDNA divergence 
than nonsister transisthmian pairs (com- 
pare bold to normal type values in Tables 1 
and 2). Parsimony analysis (18) of mtDNA 
sequences also upholds these assignments 
(Fig. 1). 
The seven transisthmian sister species 
pairs varied nearly threefold or more in 
molecular and reproductive divergence 
(Tables 1 and 2). Furthermore, each mea- 
sure was strongly and significantly (al- 
though not perfectly) associated with the 
other two (Fig. 2). The most conspicuous 
discrepancy is the low Nei's D relative to 
mtDNA divergence for pair Pl-Cl, a pat- 
tern also exhibited by one of three studied 
transisthmian pairs of sea urchins (4). The 
overall agreement among the measures of 
divergence is best explained by staggered 
isolation. The null hypothesis, that isola- 
tion was simultaneous but rates of diver- 
gence are highly variable, is incompatible 
with the observed pattern because metabol- 
ic enzymes, mtDNA, and mate recognition 
share no mechanistic basis that would cause 
their divergence rates to be automatically 
associated. 
There are few other tenable explanations 
for this pattern of concordant variation, and 
none seems applicable here. Differences in 
intensity of natural or sexual selection are 
Table 1. Molecular comparisons (13, 14) of 
transisthmian taxa. The seven pairs of sister 
taxa are in bold. Currently recognized species 
names (10), with undescribed sympatric sib- 
ling species notations (11), are as follows: P1, 
Alpheus rostratus; C1, A. paracrinitus sp. b; P2, 
A. paracrinitus; C2, A. paracrinitus sp. a; P3, A, 
panamensis; C3, A. formosus sp. a; C3', A. 
formosus sp. b; P4, A. cylindricus; C4, A. cylin- 
dricus; P5, A. saxidomus; C5, A. simus; P6, A. 
canalis sp. b; C6, A. nuttingi; P6', A. canalis sp. 
a; P7 and P7', A. cristulifrons; C7, A. cristuli- 
frons (P, Pacific; C, Caribbean). Genetic diver- 
gence between pairs was calculated with Nei's 
D for allozymes and Kimura's corrected per- 
cent sequence divergence for mtDNA (17). 
unlikely to affect all three systems in a 
parallel fashion and cannot explain the com- 
parable pattern observed for silent mtDNA 
substitutions (19) in any case. Likewise, 
there is no evidence for differences among 
the pairs in historical effective population 
sizes or generation times that can be related 
to divergence (20). 
The conclusion that isolation was not 
simultaneous justifies elimination of the 
most dissimilar pairs in estimating rates of 
molecular divergence. Dividing the values 
of allozyme and mtDNA sequence diver- 
gence for pairs Pl-Cl, P2-C2, P3-C3, and 
P4-C4 by the estimate for time since final 
closure of the Panama seaway of 3.0 to 3.5 
million years ago (Ma) (6, 7) yields an 
approximate rate of divergence of 0.03 to 
Fig. 2. Relation between allozyme divergence 
(summarized as Nei's D), mtDNA sequence di- 
vergence, and male-female behavioral compat- 
ibility for seven transisthmian sister species pairs 
(bold values in Tables 1 and 2). For each pair, 
median behavioral compatibility is summarized 
in pie chart format within the circle indicating the 
relation between the two measures of biochem- 
ical divergence. Data for each pair are indepen- 
dent of data for other pairs. Each measure of 
divergence is significantly correlated with the 
other two measures. Spearman rank correlation 
coefficients are as follows: rs = 0.82, P < 0.02 
(Nei's 0 - mtDNA); rs = 0.93, P < 0.001 (Nei's 
0 - compatibility); and rs = 0.79, P < 0.03 
(mtDNA - compatibility). These coefficients are 
significant as determined by the sequential Bon- 
ferroni method (28). Habitats for Pacific mem- 
0.04 for Nei's D and 2.2  to 2.6% for 
mtDNA sequence per 106 years (21). Es- 
timated times since divergence for the 
remaining three pairs, using these calibra- 
tions, are 4.4 to 6.1 (P5-C5), 4.0 to 6.3 
(P6-C6), and 6.8 to 9.1 (P7-C7) Ma. 
Genetic divergence before final closure 
may have been facilitated by changing 
oceanographic conditions. Fossil foramin- 
iferal assemblages suggest a cessation of 
circulatory connections across the Panama 
seaway between 12.9 and 7.0 Ma as a 
result of altered current patterns, followed 
by return of a restricted shallow water 
connection that shoaled to a depth of less 
than 50 m by 6.3 Ma (7). By 5.0 Ma, 
strombinid gastropods showed substantial 
divergence at the subgeneric level (22), 
20 
.15 
'10 
H   5 
Island - low Qy 
(5 Island - mid 
i Mainland - low 
Mainland - mid 
0.0 0.1 0.2 0.3 
Nei's D 
bers of pairs are indicated (25). The dotted line connecting the origin and the most divergent pair 
shows a linear relation between the two biochemical measures that is consistent with the 
assumption of an initial absence of genetic differentiation. 
Table 2. Behavioral tolerance and intolerance of male-female transisthmian pairs (T) relative to 
intraoceanic control pairs (I) (15). Shown are the proportion paired, number of passive contacts, 
number of snaps, number of aggressive contacts, and overall compatibility (median of the four 
measures). Species are as described in Table 1. No data are available for the P7'-C7 combination. 
Low values indicate that transisthmian pairs showed little tolerant behavior or much intolerant 
behavior relative to intraoceanic pairs of the same taxa. Note that behavioral pairing does not 
necessarily lead to production of fertile clutches. During a 30-day period after these behavioral 
observations, only one replicate of transisthmian pair P3-C3 produced fertile clutches (representing 
1 % of all transisthmian pairs tested, in contrast to 60% production of fertile clutches by intraoceanic 
control pairs). 
Taxa Allo- 
mtDNA COI Tolerance (T/l) Intolerance (l/T) 
zymes Mean (range) Taxa Paired Passive 
contacts Snaps 
Aggressive 
contacts 
Compatibility 
P1,C1 
P1, C2 
0.028 
?0.183 
7.7 (7.3-8.2) 
17.3(17.0-17.6) P1,C1 0.86 1.97 2.18 0.49 1.42 
P2, C2 0.114 6.6 (6.4-6.7) P1, C2 0.00 0.00 0.11 0.10 0.05 
P2, C1 0.119 17.4(16.7-18.2) P2, C2 0.67 0.33 1.40 0.48 0.58 
P3, C3 0.109 7.7 (7.2-8.1) P2, C1 0.00 0.02 0.15 0.10 0.06 
P3, C3' 0.124 13.4(13.1-13.6) P3, C3 0.45 0.66 0.49 0.31 0.47 
P4, C4 0.121 8.5 (8.4-8.7) P3, C3' 0.00 0.03 0.35 0.26 0.15 
P5, C5 0.177 13.4(13.4) P4, C4 0.00 0.07 0.33 0.53 0.20 
P6, C6 0.188 10.5(10.4-10.7) P5, C5 0.33 0.51 0.14 0.13 0.24 
P6', C6 0.224 9.0 (8.7-9.4) P6, C6 0.00 0.01 0.33 0.19 0.10 
P7, C7 0.272 19.2(18.6-19.5) P6\ C6 0.00 0.00 0.09 004 0.02 
P7', C7 0.231 19.7(19.3-20.4) P7, C7 0.00 0.01 0.14 0.00 0.01 
1630 SCIENCE '   VOL.260   ' 11 JUNE 1993 
REPORTS 
and carbonate-associated benthic foramin- 
iferal communities of the southern Carib- 
bean were established (23). Hence, pairs 
P5-C5 and P6-C6 probably separated dur- 
ing the period of marked shoaling and 
environmental divergence preceding final 
closure. The isolation of P7/P7' from C7 
perhaps occurred when the hypothesized 
earlier circulatory barrier was in place, 
followed by failure to interbreed when 
partial connection between the oceans 
was reestablished. Environmental transi- 
tions also appear to have prompted intrao- 
ceanic divergences (24). 
All the shrimps we studied are shallow 
water, fully marine forms with planktonic 
larvae. However, they do show some distri- 
butional differences that could affect sensi- 
tivity to changing conditions associated 
with gradual rise of the isthmus. Pacific 
members of the most divergent pairs are 
found deeper in the intertidal or are rare in 
habitats with heavy sedimentation (25) 
(Fig. 2). Thus, larval avoidance (26) of 
shoaling waters over the rising isthmus (6, 
7) may have accelerated genetic isolation of 
these pairs. 
Our data can also be used to estimate 
rates of divergence in reproductive com- 
patibility. Even the least divergent pairs 
show substantial reproductive isolation, 
but considerable behavioral compatibility 
and sporadic production of fertile clutches 
occur (Table 2). This observation suggests 
that, as in other groups (5, 27), 3.0 to 3.5 
million years may be the minimum time 
required for development of strong repro- 
ductive isolation under the classic allopat- 
ric model of division into two large popu- 
lations without secondary contact. 
REFERENCES AND NOTES 
1. E. Mayr, Animal Species and Evolution (Harvard 
Univ. Press, Cambridge, 1963). 
2. D. S. Jordan, Am. Nat. 42, 73 (1908); E. Mayr, 
Evolutions, 1 (1954); I. Rubinoff, thesis, Harvard 
University (1963); G. J. Vermeij, Biogeography 
and Adaptation (Harvard Univ. Press, Cam- 
bridge, MA, 1978); 0. S. Jones and P. F. Hasson, 
in The Great American Biotic Interchange, F. G. 
Stehli and S. D. Webb, Eds. (Plenum, New York, 
1985), pp. 325-355; J. R. Weinberg and V. R. 
Starczak, Mar. Biol. 103, 143 (1989); H. A. 
Lessios, Am. Nat. 135, 1 (1990). 
3. H. A. Lessios, Nature 280, 599 (1979); A. T. 
Vawter, R. Rosenblatt, G. C. Gorman, Evolution 
34, 705 (1980); H. A. Lessios, ibid. 35, 618 
(1981); T. Collins, Paleontol. Soc. Spec. Pub. 6, 
68(1992). 
4. E. Bermingham and H. A. Lessios, Proc. War/. 
Acad. Sci. U.S.A. 90, 2734 (1993). 
5. R. W. Rubinoff and I. Rubinoff, Evolution 25, 88 
(1971);   H.  A.   Lessios,   ibid.  38,   1144  (1984); 
  and C. W. Cunningham, ibid. 44, 933 
(1990). 
6. T. Saito, Geology 4, 305 (1976); L D. Keigwin, Jr., 
Science 217, 350 (1982), A. G. Coates et a/., 
Geol. Soc. Am. Bull. 104, 814 (1992); but see G. 
Keller, C. E. Zenker, S. M. Stone, J. S. Am. Earth 
Sci. 2, 73 (1989). 
7. H. Duque-Caro, Palaeogeogr. Palaeoclimatol. 
Palaeoecol. 77, 203 (1990); but see P. F. Hasson 
and A.  G.  Fischer,  Micropaleontology 32,  32 
(1986). 
8. S. A. McCommas, Mar. Biol. 68, 169 (1982); A. P. 
Martin, G. J. P. Naylor, S. R. Palumbi, Nature357, 
153 (1992). 
9. J. H. Gillespie, The Causes of Molecular Evolution 
(Oxford Univ. Press, New York, 1991). 
10. Pacific taxa: W. Kim and L. G. Abele, Smithson. 
Contrib. Zool. 454, 1 (1988); L. B. Holthuis, Zool. 
Meded. 55, 47 (1980). Caribbean taxa: F. A. 
Chace, Jr., Smithson. Contrib. Zool. 98,1 (1972). 
11. N. Knowlton and D. K. Mills, Proc. San Diego Soc. 
Nat. Hist. 18, 1 (1992). 
12. Character states linking transisthmian pairs within 
the triplets and quartet are as follows: presence or 
absence of dorsal spots and divided or single 
terminal abdominal band (pairs P1-C1 and P2-C2, 
respectively), balaeniceps minor chela in both 
sexes (pair P3-C3), and blue antennae and ab- 
sence of movable spine on ischium of 3rd pere- 
opod (pair P6-C6). Currently P7 and P7' can only 
be distinguished biochemically, and they appar- 
ently diverged after the Caribbean-Pacific split 
(see Fig. 1). 
13. We used standard horizontal starch gel electro- 
phoresis techniques and terminology [R. W. 
Murphy, J. W. Sites, Jr., D. G. Buth, C. H. 
Haufler, in Molecular Systematics, D. M. Hillis 
and C. Moritz, Eds. (Sinauer, Sunderland, MA, 
1990), pp. 45-126]. Sixteen loci are shared 
across all species [except P1, C1, P2, C2 (no 
activity for OPDH) and P5, C5 (no activity for 
OPDH, FUMH, ALD, LDH-1)]. Five loci were 
monomorphic (PER, GP, FUMH, LDH-1, CK), 
and 11 were polymorphic (ALD, MDHP, LDH-2, 
GDH, ICD, MDH-1, MDH-2, OPDH, GPI, TPI-2, 
PGM). Mean sample size per locus averaged 
across all species was 18.7, and the percentage 
of polymorphic loci ranged from 12.5 to 37.5%, 
with an average of 20.9%. Data were analyzed 
with BIOSYS-1 [D. L Swofford and R. B. So- 
under, J. Hered. 72, 281 (1981)]. 
14. Genomic DNA was extracted from individual 
shrimp, and for most species a 681-nucleotide 
pair (np) region was amplified by the polymerase 
chain reaction and mtDNA COI a and f primers [S. 
Palumbi ef a/., The Simple Fool's Guide to PCR 
(Department of Zoology, University of Hawaii, 
Honolulu, 1991)]. A 640-np region of species P1, 
C1, P2, C3, and P7 was amplified with an Al- 
pheus-specific primer (H7188 5'-CATTTAG- 
GCCTAAGAAGTGTTG-3') with COI f, and a 511- 
np region of species C2 was amplified with two 
/Wp/ieus-specific primers (H7083 5'-AATARGGG- 
GAATCAGTGGGCAAT-3' and L6595 5'-TATAT- 
CAACACTTATTTTGATT-3'). Both the light and 
heavy mtDNA strands were sequenced, in sepa- 
rate experiments, after K exonuclease digestion of 
the double-stranded amplification product or with 
double-stranded sequencing methods. Two indi- 
viduals from each of the 17 taxa were sequenced, 
yielding four divergence values for each pair of 
taxa. On average, 76% of the sequence analyzed 
was verified by the overlapping of light and heavy 
strands. 
15. We observed the behavior of a male and female 
placed in a glass dish with a small shelter for 30 
min and recorded two measures of tolerance 
(number of passive contacts, paired versus sep- 
arate at end of observation period) and two 
measures of intolerance (number of snaps, num- 
ber of aggressive contacts) [see (16) for meth- 
ods]. Males and females were above minimum 
reproductive size, matched in size, and were 
used only once. For each related pair of trans- 
isthmian taxa, we performed four categories of 
experiments: two intraoceanic (Caribbean male 
+ Caribbean female, Pacific male + Pacific 
female) and two transisthmian (Caribbean male 
+ Pacific female, Caribbean female + Pacific 
male). Replicates of each (generally three to 
four) were used to calculate the percentage of 
trials resulting in pairing and the mean numbers 
of snaps, aggressive contacts, and passive con- 
tacts per trial. From these data, mean intraoce- 
anic and transisthmian values were calculated 
for each of the four behavioral measures. We 
calibrated transisthmian encounters against in- 
traoceanic encounters involving the same taxa 
by dividing one by the other (transisthmian di- 
vided by intraoceanic values for tolerant interac- 
tions, intraoceanic divided by transisthmian val- 
ues for intolerant interactions), so that the larger 
the value, the greater is transisthmian behavioral 
compatibility relative to intraoceanic controls. 
16. B. A. Nolan and M. Salmon, Forma Functio2, 289 
(1970); N. Knowlton and B. D. Keller, Behav. Ecol. 
Sociobiol. 10, 289 (1982); Bull. Mar. Sci. 37, 893 
(1985). 
17. M. Nei, Genetics 89, 583 (1978); M. Kimura, J. 
Mol. Evol. 16, 111 (1980). 
18. D. L Swofford, PAUP: Phylogenetic Analysis Us- 
ing Parsimony, version 3.1; (Illinois Natural History 
Survey, Champaign, IL, 1993). 
19. The mean percentages of fourfold degenerate 
sites with transversions were 7.2 (P1-C1), 4.2 
(P2-C2), 3.6 (P3-C3), 3.0 (P4-C4), 10.1 (P5-C5), 
10.6 (P6-C6), and 26.9 (P7-C7). 
20. If differences in historic population sizes were 
responsible for the concordant variation in diver- 
gence, then the most divergent pairs should show 
the lowest heterozygosities [M. Nei, T. Maruyama, 
R. Chakraborty, Evolution29, 1 (1975); R. Chakra- 
borty and M. Nei, ibid. 31, 347 (1977)]. Average 
heterozygosity values ranged from 0.016 to 0.080 
(observed) and 0.015 to 0.125 (Hardy-Weinberg 
expected). There was no relation between het- 
erozygosity (minimum, maximum, or mean of Car- 
ibbean and Pacific values) and any measure of 
divergence [for Spearman rank correlations, table- 
wide P values are insignificant for a sequential 
Bonferroni test (28); the only two individually sig- 
nificant values showed a correlation in the direction 
opposite that expected]. Similarly, if differences in 
generation time produced the concordant pattern, 
then the smallest taxa should show the greatest 
divergence because of a positive correlation be- 
tween body size and generation time [R. H. Peters, 
The Ecological Implications of Body Size (Cam- 
bridge Univ. Press, Cambridge, 1983), p. 132] and 
a negative correlation between generation time 
and rate of evolutionary -change (9). Published 
maximum carapace lengths ranged from 6.1 to 
15.1 mm for the Pacific species (10). There is no 
correlation between maximum size and any mea- 
sure of divergence (for Spearman rank correla- 
tions, all Pvalues >0.6). 
21. Medians of mean mtDNA divergences and Nei 
distances (Table 1) for the four pairs were used. 
These calibrations assume a linear relation be- 
tween time and divergence for the spans over 
which they are calculated or applied. They are 
broadly consistent with previous studies (3, 4, 8), 
although enzyme systems and mtDNA regions 
are not equivalent. 
22. J. B. C. Jackson, P. Jung, A. G. Coates, L. 5. 
Collins, Science 260, 1624 (1993). This study also 
reveals variation in the timing of divergence com- 
parable to that exhibited by Alpheus. 
23. L. 5. Collins, Paleontol. Soc. Spec. Pub. 6, 67 
(1992). 
24. Values of mean corrected percent sequence di- 
vergences for sympatric members of sibling spe- 
cies complexes are 8.1 (P1-P2), 16.4 (C1-C2), 
12.1 (C3-C3'), 7.6 (P6-P6'), and 6.3 (P7-P7'). 
Nei's 0 values are 0.109 (P1-P2), 0.194 (C1-C2), 
0.165 (C3-C3'), 0.216 (P6-P6'), and 0.019 (P7- 
P7'). 
25. Habitat differences are most conspicuous in the 
Pacific [P. W. Glynn, Bull. Biol. Soc. Wash. 2, 13 
(1972)]. Species P7 is normally collected in the 
shallow subtidal of offshore islands in dead coral," 
species P6 is only collected during extreme low 
tides (-60 to -85 cm), and species P5 is mid- 
intertidal but restricted to clearer waters of off- 
shore islands. In contrast, species P1, P2, P3, and 
P4 are abundant in the mid- to low-intertidal zone 
along the mainland. 
26. The swimming abilities of crustacean larvae per- 
mit considerable habitat choice [R. S. Burton and 
M. W. Feldman, in Estuarine Comparisons, V. 5. 
Kennedy, Ed. (Academic Press, New York, 1982), 
SCIENCE   ?   VOL. 260   ?    11 JUNE 1993 1631 
pp. 537-551], and distributions of larvae in the 
water column may reflect depth distributions of 
shallow-water adults [R. K. Grosberg, Ecology 63, 
894 (1982)]. 
27. J. A. Coyne and H. A. Orr, Evolution43, 362 (1989). 
28. W. R. Rice, ibid., p. 223. 
29. We thank J. Jara, F. Bouche, E. Gomez, and R. 
Hamilton for technical assistance; E. Duffy, A. 
Herre, J. Jackson, H, Lessios, A. Martin, P. Morris, 
R. Rowan, and D. Zeh for comments on the 
manuscript; the Smithsonian Institution for finan- 
cial support; and the Government of Panama 
(Recursos Marines) and the Kuna Nation for per- 
mission to collect specimens. 
19 January 1993; accepted 26 April 1993 
Large Odd-Numbered Carbon Clusters from 
Fullerene-Ozone Reactions 
Stephen W. McElvany, John H. Callahan, Mark M. Ross, 
Lowell D. Lamb, Donald R. Huffman 
The odd-numbered carbon clusters C119, C12g, and C139 have been observed in the mass 
spectra of toluene extracts of fullerene soots and of the products of ozone-fullerene 
reactions. Specifically, ozone-C60 reactions yield C119, ozone-C70 reactions yield C139, and 
ozone-(C60/C70) reactions produce C11g, C12g, and C139. These unexpected species 
correspond to dimers of C60, C60/C70, and C70, respectively, less one carbon atom, and 
are stable gas-phase ions with behavior similar to that of fullerenes. The results suggest 
a new route to functionalization and derivatization of fullerenes through controlled ozone- 
catalyzed cage-opening reactions. 
JNumerous studies have shown that there 
is a wide variety of fullerene chemical 
reactions (1). For example, one unusual 
aspect of fullerenes is their ability to un- 
dergo coalescence reactions that result in 
larger fullerenes (2-4). Although the de- 
tails of these reactions vary, in all cases 
observed to date coalescence apparently 
was caused by photon-induced radiation 
damage of fullerenes, and in all cases the 
reaction products had even numbers of 
carbon atoms. 
In a recent mass-spectral investigation 
of large fullerenes, we detected the pres- 
ence of the odd-numbered, pure carbon 
clusters C119, C129, and CI39 in toluene 
extracts of several fullerene soot samples, 
which we speculated to be the products 
of coalescence of two C60's, a C60 and 
C70 and two C70's, respectively. These 
large, odd-numbered carbon clusters are 
unexpected, given the overwhelming evi- 
dence for the preferential stability of large, 
even-numbered carbon clusters (5). Re- 
sults from a subsequent series of ozonolysis 
experiments support this interpretation and 
suggest that oxidation plays a key role in 
the production of these unusual species. 
These results have implications for several 
important issues in fullerene chemistry, in- 
cluding chemical reaction mechanisms and 
the resulting fullerene-based products, coa- 
lescence of fullerenes, and the molecular 
structure consideration raised by the exis- 
tence of odd-numbered "fullerenes." 
S. W. McElvany, J. H. Callahan, M. M. Ross, Code 
6113/Chemistry Division, Naval Research Laboratory, 
Washington, DC 20375-5320. 
L 0. Lamb and 0. R. Huffman, Department of Physics, 
University of Arizona, Tucson, AZ 85721. 
A typical thermal desorption-negative 
ion mass spectrum (6, 7) of a toluene 
extract (8) of a commercial soot sample 
(9) is shown in Fig. 1. Although not 
shown, CL0~ and C70~ are approximately 
103 to lCr times more abundant than the 
base peak in this spectrum. As expected, 
the abundances of fullerene ions (Cn~, n 
> 74) generally decrease with increasing 
size, with anomalously abundant C84 and 
C90. However, in addition to the even- 
numbered carbon clusters, ions are ob- 
served corresponding to C119, CI29, and 
C139. These odd-numbered carbon clusters 
were detected in a commercially available, 
unchromatographed mixture of fullerenes 
(9), as well as in toluene extracts of 
various fullerene soots [Polygon, SES (9), 
and "homemade" soot (7) produced at the 
Naval Research Laboratory]. 
In order to test the interpretation of the 
100 
60 
?Q 
a 
9 ?40 
20 
Cn 
.jkllui 
mass-spectral results, two alternative ex- 
planations had to be ruled out. Namely, it 
was possible that these clusters were not 
stable species but only artifacts of the mass 
spectrometry, that is, these ions were gen- 
erated in the desorption-ionization pro- 
cess. In contrast to positive ion or laser 
desorption analysis, which may be compli- 
cated by fragmentation or coalescence of 
molecular species, previous studies of 
fullerenes indicate that thermal desorp- 
tion-negative ion analysis is much less 
prone to these artifacts (7). A second 
possibility was that these were not pure 
carbon molecules. The relative ion abun- 
dances in the distribution from mass-to- 
change (mix) ratios 1428 to 1433 were 
measured to be identical (within experi- 
mental error) to those calculated for C119 
based on the natural 13C abundance (see 
inset of Fig. 1). Similar results were ob- 
tained for C129 which indicated that these 
ions correspond to odd-numbered all-car- 
bon species. (The abundance of C139, 
however, was too low to allow this type of 
analysis.) Further investigations showed 
that the presence of the odd-numbered 
carbon clusters is not affected by the sol- 
vent, as they are also observed in hexane 
and benzene extracts of fullerene-rich 
soot. Careful inspection of mass spectra 
(not shown) of the raw (unextracted) soot 
revealed very low abundances of these ions 
(?500 times less abundant than neighbor- 
ing even-n Cn). This observation and the 
analysis of the soot toluene extract (Fig. 1) 
suggest that the odd-numbered clusters are 
more soluble than the comparably sized 
even-numbered fullerenes. In addition, 
the use of ammonia or argon instead of 
methane as the buffer gas in the analysis 
yielded similar results, indicating that the 
formation of odd-numbered clusters is not 
due to the buffer gas. 
As a complement to the negative ion 
analyses described above, electron ioniza- 
tion to generate positive ions of thermally 
1427 14291431 1433 
^lii t |),l,l, t , ,,i,yl|j,i,ul|,,,| i,,, 
1000 1200 1400 1600 
mlz 
Fig. 1. Negative ion mass 
spectrum of a toluene ex- 
tract of graphitic soot with 
the inset showing the ex- 
panded     region     around 
Cl19~ 
1800 2000 
1632 SCIENCE VOL. 260 11 JUNE 1993