Diversity and Endemism of the Benthic Seamount Fauna in the Southwest Pacific

letters to nature

an exclusively to predominantly hei livorous~~ In overall A. morphology the dentition most closely resembles that of some We thank B. Rakotosamimanana. RasOakaramanana, P. Wright, B. Andriamihaja, the the omithischian dinosaurs, particularly stegosaurs and ankylosaurs, staff of the Institute for Conservationof Tropical Environments, the villagers of Berintra, and all expedition members for assistance with Beld work V. Heisq. and which are also generally regarded as herbivore~~~,~~.there are no J. Still, Holstein for assistance with specimen preparation; B. Demes, C Forster, N. Rybcrynski modem crocodylian analogues with which to compare the potential and S. Zehr for discussion; and the NSF for funding. function of Simosuchus' unique skull and dentition, and the suggestion of an herbivorous diet is speculatively based on dental Correspondence and requests for materials should be addressed to G.AB. comparisons with non-crocodylian reptiles. Given the cranial and dental specializations as well as the small size of the adult Sitnosuchirs, it can be assumed that Simosuchus did not bring down large prey. It is still plausible, however, that the diet may have consisted of arthropods, other invertebrates and potentially ...... small vertebrates such äs frogs. Similarities with ankylosaurid dinosaurs are not limited to dentition. Other features of Simusuchus appear to be convergent Diversity and endemism of the with those of ankylosaurs, such as the broad compact body, extensive dorsal and ventral shielding, bony protettion of the benthic seamount fauna skull above the supratemporal fenestrae .and orbits, and a deep the southwest Pacific cranium with a broad, short snout. A crocodyliform convergent in upon an omithischian dinosaur is intriguing in lightof the apparent Bertrand Richer de Forges+t,J. Anthony KoslowSg & G. C. B. Poo& absence of the latter from the Late Cretaceous of Madagascar. 0 .. *Centre IRD de Noumea, BPA5, 98848, Noumea Cedex, New Caledonia .I Received 8 February;accepted 13 April 2000. $ CSIRO Marine Research, GPO Box 1538, Hobart, Tnsmania 7001, Atrsfralia Victoria, 1. , Langston,W. inBiologyo~tlreRrptiliaVol.4(edsGans.C. &Panons,'T. S.) 263-284 (Academic, New §Museum of 71 Victoria Crescent, Abbotgord, Victoria 3067, Australia York. 1973). t These authors contributed equally to this work 2. Ruxoni,C Sobrereptilescreticcos del Uruyay (Urugriaysuclt~~na',~,n. g. n. sp.) ysus relaciones ...... con los notossúquidor de Patagonia. Bol Inst. Geolog. PeTf: Umg~uy19,3-64 (1932).

3. ~ Knuse.D. W.~t~l.TheLatcCretaceousvertebrarehunaofMadagas~:implicationsforGondwan~ Seamounts comprise a unique deep-sea environment, character- paleobiogeography. GSA Today 9,l-7 (1999). ized by substantially enhanced currents and a,fauna that is 4. Brochu,C. A. Closureofneurocentnlsuturesduringcmwdilianontogeny:implications formpity dominated by suspension feeders, such as corals'-? The potential assessment fossil archossurs. J. Vértebr.Paleonto% 16.49-62 (1996). J. 5. Pol, D. Basal mesoeucrocodylianrelationships: new dues to old conflicts. Víebr. Paleonîol. 19 importance of these steep-sided undersea mountains, which are (suppl. 3). 69A (1999). generally of volcanic origin, to ocean biogeography and diversity J. 6. Bonaparte. F. Losvertebrados fosiles de la formacion Rio Colorado, de la ciudad de Neuquen y was recognized over 40 years ago', but this environment has Cercanias, Cretacico Superior, Argentina. Rev. Mus. Argcnr. Cienc. Nat. 4.16-123 (1991). J. remained very poorly explored. A revie$ of seamount biota 7. Clark, M.. Jacobs,L. L.. &Dorms, W. R Mammal-likedentition in a Mesozoic uocndylian. Science

244. 1064-1066 (1989). @ and biogeography reported a total of 597 invertebrate species 8. Gomani.E. M.AcrocodyIiformfrom theEarly Cretaceous dinosaurbeds, northernMalaw¡.l Vertebr. recorded from seamounts worldwide since the CharZengm expedi- PaleontoL 17,280-294 (1997). tion of Most reports, based on a single taxonomic group, 9. 1872. Medem, F. Osteología aaneal, distribución geográfica y ecologia de MeInriosuchlisniser (Spix) were extremely limited: 5 seamounts of the estimated more than (Crocodylia, Alligatoridae). Rw. Amd Colomb. Cieirc Emd W y Nat lt.5-19 (1963). J. 10. Balouet, C. & Buffetaut, E. Mekosucltis inexpenanrs, n.g., usp., aocodilien nouveaude I'Holocene 30,000 seamountsin the world's oceans4"accounted for 72% of the de Nouvelle Calédonie. C R. Acad. Sci. 304,853-856 (1987). species recorded. Only 15% of the species occurring on seamounts 11. W~llis,P. M. A New crocodilians fmm the Late Oligocene White Hunter Site, Riversleigh; north- were considered potential seamount endemics. Here we report the western Queensland. Me~nQld Mus. 41,423-438 (1997). discovery of more than macro- and megafaunal species from 12. Cawalho, I. S. &Campos, D. A Um mamífero triconodonte do Cretáceo Inferior do Madào, 850 BmiL An. Arad Bru. Cienc 60,437-446 (1988). seamounts in the Tasman Sea and southeast Coral Sea, of which 13. Canalho. L S. Candidodon:um crocodilo com heterodontia(Notosuchia, CretáceoInferior - Brasil). 29-3496 are new to science and potential seamount endemics. An. Amd Bras. Cienc. 6633'1-345 (1994). . Low species overlap between seamounts in different portions of 14. Wu, X -C., Sues, H. -D.&Sun, A. A plant-eating crocodyliformreptile from the Cretaceousof Cllink Nanrre 376.678480 (1995). the region indicates that the seamounts in clusters or along ridge J. 15. Benf0n.M. I. &Clark, M. in ThePhylo*~yandCInsrificntion ofthe TetrapodsVol. 1 (ed. Bcuton,M. systems function as 'island groups' or 'chains,' leading to highly J.) 295-338 (Clarendon, Oxford. 1988). localized species distributions and apparent speciation between 16. Krause, D. W., Prad G. V. R., Koenigwald,W. von, SAN, A & Grine. E E. Cosmopolitanism groups or ridge systems that is exceptional for the deep sea. These among Gondiv&n Late Creraceous m3mmals. Nnture 390,504-507 (1997). S. this 17. Sampron. D. et aL Predatory dinosaur remains fromMada&car: implications for the Cretaceous results have substantial implications for the conservation of biogeobphy of Gondwana. Science 280,1048-1051 (1998). fauna, which is threatened by fishing activity'. C. 18. Buckley, 6. A., Brochu. A. &Krause, D. WIHy&rdiversityand the paleobiogeographicorigins of Whereas previous studies of the seamount fauna have often the iate Cretaceous crocodyliforms 0fMadagascar.J. Vertebr.PakontoL 17 (suppl. (1997). 3). 35A focused on particular taxa, the present study sought to describe 19. Iordansky, N. N. in Biology ofthe Reptilia Vol. 4 (eds Gans, C &Parsons, T. S.) 201-262 (Academic, New York, 1973). the benthic community as a whole and enlisted broad taxonomic (J. 20. Gans, C Biomechanics:An Approach to Vkiebmte Biology B. Lippinmtt, Philadelphia. 1974). support (see Acknowledgements). In all, 516 species of fish and The J. 21. Wake, M.H. in SLuIl VOL 3 (edsHanken, & Hau, B. IC) 197-240 (University of Chicago, macro-invertebrates were obtained from 6 seamounts along the Chicago, 1993). 22. Schumacher,G.-H.inBiolog)'oftheReprilinVol.4(edsGans, C. &Parsons,T.S.)101-199 (Academic, Norfolk Ridge, 108 from 4 seamounts on the Lord Howe Rise, and NewYÖrk. 1973). . 297 from 14 seamounts south of Tasmania (Table 1, Fig. 1).Thirty- 23. Hott0n.N. III Asurveyofadaptive relationships ofdentition to diet in theNorth AmeriunIguanidae. six per cent of species from the Norfolk Ridge seamounts were new Am. Mid. Nat 53,138-114 (1955). to science and not known from sampling of the open seafloor and 24. Suffi,H.-D. &Reisz, R. R Origins and earlyevolutionof herbivory in tetrapods. TrendsEcoLEVO% 13, 141-145 (1998). are therefore potential endemic species, along with 31% of species 25. Galton, P. M. in The Dinosauria (eds Weisbampel, D. B., Dodson, P. & Osmóska, H.) 456-483 (Unk from the Lord Howe seamounts and between 16 and 33% from the California. Berkeley, 1990). Tasmanian seamounts. 26. Coombs, W. P. &Maryanska, T. in The Dinosauria (cds Weisbampel, D. B., Dodson, P. & Osmólska. H.) 456-483 (Univ. California, Berkeley, 1990). - Over the range of sampling carried out on Norfolk and Lord 27. Swofford, D. L. PAW': PhyfogeneticAnalysirUn'ngParsimony (*and Other Methods], Version 4.0bZa. Howe Ridges, we found a linear relationship between number of (SinauerAssociates. Sunderland, Massachusetts, 1999). species recorded from each seamount and the number of samples íñ6~käiion-is-äSdäEdë obtained there (Fig. 2). This relationship implies that the greater SuippE,mGtäG on'N&ë's WÕrlp I.M;ide Web site (http://www.nature.com) or as paper copy from the London editorial office of Nature. number of species obtained from seamounts on Norfolk Ridge is -- I22 NATURE IVOL 405 JUNE 2000 I www.nature.com - .-. g44 l~l~lll~~~~~l~~~~ll' 0 10022029 letters to nature

150"E 165'E 180" 300 15"s 15"s A 250

200

150 30"s 30"s

O0 1 BR LordHowe X Total Lord Howe 50

O 45"s 45% I 50 1 O0 Number of samples 150"E I65"E 180" of Figure 2 Relationship between the number species recorded at each seamount site of Figure 1Location'of seamount sampling sites (triangles)and topographic and and the number samples obtained from that site. oceanographic features mentioned in the text. Depths are given in fathoms (1 fathom = 1.83 m).

owing to greater sampling effort there rather than greater species seamounts in that region (C = O). Thus there are significant richness. The lack of an asymptote in the relationship implies that differences in species composition of the seamount communities the number of species present on these seamounts is far greater than between ridge systems at a similar latitude and separated by -1,000 the number currently recorded. Species richness on the Tasmanian km across the northern Tasman Sea'and southeast Coral Sea, and a seamounts appears to be significantly greater, but it is unclear complete changeover of species between the northern and southern whether this is due to greater diversity or is an artifact of sampling sectors of the Tasman Sea. C did not decline significantly with over a greater number of seamounts, albeit within a limited area. distance on smaller scales (tens to several hundred kilometres) Although sampling on the Tasmanian seamounts extended deeper (Fig. 3). than on Lord Howe Ridge and the Norfolk Rise, sampling in the Although the region around southern Australia and New northern Tasman Sea extended into shallower depths (Table 1). Caledonia is noted for its relatively isolated €aunas''J', the complete Most species on the Tasmanian seamounts exhibited a broad depth lack of affinity betyeen the seamount communities across this area distribution and few were found only at the greatest depths appears remarkable. For comparison, about 60% of the 93 species sampled7. High species diversity has been previously noted off of non-mesopelagic decapod crustaceans (Anomura, Caridea, southeastern Australiass. Palinura, Brachyura and Astacidea) recorded from depths greater The coefficient of community (C) between seamounts decreased than 200m on the continental slope off southeastern Australia significantly from a median value of 0.21 (range 0.08-0.49) for (south of -35" S latitude) are recorded from the Indo-West Pacific adjacent seamounts on the same ridge to a median of 0.04 (range 0- region as well (G.C.B.P., unpublished data). This mirrors the general 0.17) for seamounts on different ridge systems within the northern pattern for decapods on the continental shelf in this region, sector of the Tasman Sea and southeast Coral Sea (that is, the although isopod affinities are mostly with Antarctic and deep-sea NorfolkRidge and LordHowe Rise) (analysis ofvariance (ANOVA), faunas'. Thus, the isolation of the seamount faunas between the FIAS(C) = 46.56, P < 0.001) (Fig. 3). Just four species from the southem and northern Tasman Sea contrasts markedly with the Tasmanian seamounts are known from the seafloor around New generally strong affinity displayed by the soft-sediment slope fauna Caledonia (the bryozoan Hmwelliporina multiaviculata, the bivalve between these two regions. Acesta sp., the barnacle Hexelasmn sp., and the hydroid It was also striking to find such little overlap in community Symplectoscyphus commensalis), but none of these was found on composition in the deep sea between sites sharing the same habitat

Table 1 Location and depth range of seamounts sampled Seamount ...... Latitude Longitude...... Seamount depth range (m)

Aztec 24" 07' 168" 45' 55-1.000 57-800 16 Sponge 24" 55' 168" 20' 450-2,000 494-920 84 Jumeau East .23" 42' 168" 17' 250-1,500 380-950 30 Jumeau West 23" 40' 168" 00' 200-1,000 230-730 51 Kaimon Maru 23" 26' 168" 05' 400-2,000 223-600 58 Stylaster 23" 40' 167" 40' 400-1,000 418-970 29 Lord Howe mo 23" 13' 159" 35' 150-2,000 178-300 6 Kelso 23" 51 159" 24' 150-2,000 160-840 12 Nova 22" 28' 159" 15' 250-2,000 277-930 9 Capel 159" 24" 50'44" 35' 200-2,000 223-540 8 Tasmanian*...... 44" 12' - 24' ...... 660-2,000 660-1.700 34 'The Tasmanian seamountsare aggregated owng to II .. .. 405 2000 NATuRElVOL 122 7UNE Iwww.nature.com 945

?." 8. 8. 'i letters to nature

time. Seventeen new genera were obtained from the Norfolk Ridge samples, four from the Lord Howe Ridge and seven or eight from the Tasmanian samples. Some appear to be relicts of groups earlier believed to have disappeared in the Me~ozoic'~-~. These seamounts thus appear to be isolated marine systems and provide an exceptional opportunity to examine evolution and speciation in the deep sea. Y El The highly localized distribution of many seamount species has Et Eo O profound implications for their conservation. Seamount commu- 1,000 2,000 3,000 4,000 5,000 6,000 7,000 nities are extremely derable to the impacts of fishing: their Distance (km) limited fixed habitat, the extreme longevity of many species (of the order of 100 years and more2'-"), and the apparently limited The (4 Figure 3 coefficient of community in relation to distance. Cis shown between recruitment between seamounts, all compound the uncertainty of seamount sites (inverted solid diamonds), hydrothermal vent sites from the East Pacific recovery from trawling, which sweeps away the benthic epifaunal Rise in the North and South Pacific and Galapagos Riff (open circles), and vent sites from community as a bycatch7. The global status of seamount benthic disjunct ridges in the northeast Pacific at 41-49' N (crossed circles). communities is unknown; however, the localized distribution of many benthic seamount species greatly increases the threat of extinction and may require that conservation and protection of seamount communities be undertaken on a local scale. U type, at similar latitude and depth, and only -1,000 kilometres apart. This distribution pattern is consistent with Johannesson's Methods. hypothesis that many taxa adapted to seamount conditions limit Our study is based on 23 cruises that the Muséum National d'Histoire naturelle and the their dispersal to maintain their populations, because of the gen- Institut François de Recherche Scientifiquepour le Développement en Coopération (MUSORSTOM) carried out since 1984 to study the fauna of the seamounts, ridges and erally small size-of seamounts, the considerable distance between adjacentseafloor within the FEZ of New Caledonia at the northem edge of the Tasman Sea, them and their unique oceanographic envir~nment""~.Circulation 11 of which focused on seamounts,primarily along the Norfolk and Lord Ho& Ridgesio; a cells in the vicinity of seamounts and along seamount chains have single cruise Lamed out by the Commonwealth Scientific and Industrial Research 1998 14 been shown to facilitate recruitment to seamounts within relatively Organization (CSIRO) in to study the benthic fauna of small seamounts off close proximity of each other and along ridge southem Tasmania at the southern extreme of the Tasman Sea'; and sumeys by the Museum of Victoria of the soft-bottom fauna of the continental slope off southeastern i0 Endemicity of the benthic fauna on southwest Pacific seamount Australia baied cruises 1984,1986 and 1994 (ref. 9) (Table 1, Fig. 1). ORSTOM carried groupslchains appears to be greater than that of deep-sea vent out 462 trawls and 748 dredge hauls using a cod end with 10 mm meshz; many results are communities, another specialized deep-sea habitat. For compari- presented in the 21 volumes of Restrlfafsdes campagnes MUSORST0M6. On the CSIRO son, we calculated Con the basis of available data for vent sites from mise, so far as was possible, each seamount was sampled on its summit, slope and base usingadredgewithacodendofZSmmofstretchedmesh;34sampleswereobtainedinall. the Explorer, Juan de Fuca and Gorda Ridges in the northeast Pacific To compare the similarity in fauna between two seamounts,we used the coefficient of at 41-49' N latitude, several site's along the northern East Pacific community (variously referred to as Sorensen's, Dice's or the Bray-Curtis similarity 27" Rise (9-13', 21" and N latitude), the Galapagos Rift on the statistic for binary data): C= 2a/(btc), where a is the number of species in commo; to equator and 17-18' Slatitude on the southern East Pacific Riselfi.In both sites and b and care the total number of species found at each site. C is commonly assumed tovarybetween Oand 1(ref.25), butin factthemaximumvalueissensitivetothe contrast to the seamounts, there was substantial similarity in relative number of species present at the two sites" and is conservativelybiased when the community composition, even at distances of 3,000-4,000 km samplingand number of spedes at two sites differs significantly. The proportion of species (Fig. 3). There was some indication that similarity might be reduced at the less well-sampled site in common to the two sites (that is, C' = ab,where b < c) is biased in the opposite direction and was used as an upper limit to the coefficient of between disjunct ridge systems (for example, between the Juan de two community between sites, where there were large differences of sampling intensity; Fuca and East Pacific Rise) but these differenceswere not significant however, use of C' did not materially affect any of our conclusions. (analysis of covariance (ANCOVA), Fl,12 = 3.02, P = 0-11), and the similarity of vent communities between disjunct ridge systems was Received 22 November 1999; accepted 28 March 2000 still clearly greater than between seamounts in the northern and 1. Rodcn,G.I.inErivironmetrtandRe~ou~esofS~mounfsintl~NorfhPatific(edsUchida,RN.,Hayasi. (US southern Tasman Sea (Fig. 3). Vents are relatively ephemeral S. & BoebIert,G. W.)NOAA Teclmical ReporfNMFS 43,9-12 Dept. of Commerce, Seattle, 1986). environments, whereas seamounts are typically millions to tens of 2. Genin, A., Dayton, P. K, Lonsdale, P. F. &Spies, E N. Co& onieamountsprovide evidence of millions ofyearsin age, so dispersal between vent sites is essential to cuncnt acceleration over dee sea topography. Nature 3% 59-61 (1986). 3. Wion, R R &Kaufmann, R S. Seamount biota and biogeography. Geophp Monqraplzs 43,355- the continuance of these communities. Cold seeps, on the other 377 (1987). hand, appear to have higher levels of endemicity: most species are 4. Rogers, R D. The biology of seamounts. Adv. Mnr. BioL 30,305-350 (1994). apparently restricted to single regions", which has been related to 5. Hubbs, C. L. Initial discoveries of fish faunas on seamounts and offshore banks in the eastan Pacific. Pa$ Sci 13,311-316 (1959). J. the stability and isolation of these habitats. However, direct com- 6. Smith, D. K. & lordon, T. H. Scanount statistics in the Pac5c Occan. Geophys.Res. 93,2899-2919 parison with cold seep communities would be premature, as most (1988). A. cold seeps are poorly explored with only a few species recorded per 7. Koslow, I. et al. The seamount benthic macrofauna off southem Tasmania: community strumtre and impacts of trawling. Mar. EcoL Prog. (submitted). , site. J. Ser. 8. Koslow, A, Bulman, C M. &Lyle, J. M. The mid-slope demersal fish community off southeastan Wilson and Kauhann3 estimated the endemism of fishes and Australia. Deep-sea Res 41,113-141 (1994). 15%, J., the invertebrates on seamounts as 12 and respectively, but half of 9. Poore, G. C. B., Jnst, &Cohen, B. E Compositionand divenityof Crustacea Iropada of the reports that they cited from reasonably well-studied regions southeastern Australian continental slope. Deep-sea Res. 41,677-693 (1994). 10. Richer de ForgesB.Explontions forbathyd faunaintheNewC;lledonianeconomiczone.Mem.Mur (that is, the Hawaiian Ridge, Mid-Pacific Mountains, Kyushu-Palau Nat2 Hirt. Nor. 145,9-54 (1990). Ridge, Vema seamount) indicated levels of endemism of 22-36%, 11. Vaughan. T. W. &Wells, I.W. Revision ofthe suborders, Wiies, and genera ofthe Scleractiiina. GOL consistent with our results. Citing Hubb's 'stepping-stone' hypo- Soc Am SpeCidPapm 44,363 (1943). thesis5, Wilson and Kaufmann3 emphasized the faunistic links 12. Johannesron. K The paradox of Rockall: wby is a brooding gastropod (Littorim 5amhlis) more widespread than one having a planktonichnal dispersal stage (LJitturm)? Mar. Bid. 99,507-513 among seamount chains extending around the rim of the North (1988). oceanic Pacific, along the East Pacific Rise into the southeast Pacific and 13. Parker, T. & TunnicWe, V. Dispersal strategies of the biota on an scaaouut: implications for through the Emperor Chain and Hawaiian Ridge into the central ecology and biogeography.BioL Bull 187,336-345 (1994). North Pacific. The apparently distinct seamount chains of the 14. Lutjbs.1. R E. & Heydorn, A. E E The rocklobstcr~asus&ani on Vema seamount drifting buoys suggest a possible recruiting mechanism. Deep-Sa Res. 28,631-636 (1981). A southwest Pacific, Indian and Southern Oceans and much of the 15. Mullineaur, L. S. & Mills, S. W. test of the larval retention hypothesis in seamount-generatedflows. Indo-West Pacific" were, however, .virtually unexplored at that Deep-sea Rec. 44,745-770 (1997).

946 NATUREIVOL 405 122 JUNE 2000 Iwww.nature.com

II II .. ers to nature

the 16. Tunnicliffe, V., McArthur, A G. & McHugh, D. A blogeographicd perspective of deep-sea model of cortical processing as an electronic circuit that emulater hydrothermalventfaunaAdv.Mar. Bio¿ 34,353-442 (1998). this hybrid operation, and so is able to perform computationsthai 17. Sibuet, M. & Olu, K. Biogeography, biodiversity and fluid dependence of deep-sa-cold-seep gain communitiesat active and passivemargins. Deep-sea Ra. II 45,517-567 (1998). . are similar to stimulus selection, modulation and spatio 18. Heezen, B. C. & Tharp, M. World Oman Floor (Map). United States Nay. Office of Naval Research temporal pattern generation in the neocortex (Washington, D.C., 1977). The multistability of digital circuits and the linear amplification 19. Bouchet, P. & Metivier, B. Linng Pleurotomariidae (MoUusck. Gzs~opoda)from the South Pacific. New ZealnndJ. Zoo., 9,309-318 (1982J. achieved in analogue circuits are generally seen as incompatible J. two 20. Ameziane-Cominarh, N., Bourreau, P. & Rau, M. Les crinoides pedoncules de Nouvelle- functions and are separated into classes of electronic technol- Caledonie (S.W.Paci6que) :une fauneancestralebue delaMesogeemesozoique. C.R.4cad Sc Park, ogy. However, the neuronal circuits of the neocortex do not respect 304.15-18 (1987). this distinction. There, multistability coexists with analogue 21. lauin, B. Decouverte d’w squelette de du lophophore de type “crura” chez un brachiopode soutien response. For example, when a visual stimulus is attended at the inarticule: description de Neoancistrouaniay norfolld gen. sp. nov. (Cranidae). C.R Acnd. Sci. Parir, 314,343-350 (1992). expense of other visual stimuli-the subject is concentratingon one 22. Richer de Forges, B. La Divenite du Benthos Marin de Nouvelle-Caledonie:de l’&pece a la Notion de object in a field of vision-then many cortical neurons tend to Patrimonine. PhD Thesis, Museum National d’Histoire Naturelle. Paris (1998). respond in a graded way to the sensory attributes of the attended 23.VaceletJ.etnLAcolonialsphinaozoanspongerelatedtoTnassicreefbuilderssurvivingindeep water off New Caledonia. CRAcnd. Sci 314,379-385 (1992). stimulus, as if it were presented alone’. . 24. Grigg, R W.Resource managementofprecious corals. a review and application to shallow water reef- We have designed a simple electronic circuit that emulates this building corals. MurinefioL 5,57-74 (1984). hybrid behaviour. The circuit comprises a ring of 16 excitatory 25. Pielou, E. C in Population and Community Ecology 311-312 (Gordon and Breach Science, NewYork, 1974). neurons, each of which makes synaptic connections of variable 26. Richer de Forges. B. ín Remifats der Campaper MUSORSTOM Volume 10 (ed. Crosnier, A) Mem. strength onto itself and onto its nearest and next nearest neigh- Mus NntlHist. Nat. 156,475-491 (1993). J. bours. These localized excitatory interactions reflect the preference 27. Koslow, A Community structure in North Atlantic deep-sea &shes. Prog. Oceanogr. 31,321-338 for local connections observed in neocortex. At the centre of the ring (1993). is a single inhibitory neuron that receives synaptic input from all the

Acknowledgements ~ Resnlts from the northem Tasman Sea were obtained as part of the MUSORSTOM program, a collaborationbetween ORSTOM and the Natural History Museum in Paris, and involved some 181 researchers from 92 Institutions and 24 nations. Work around Tasmania was supported by CSIRO and grants from Environment Australia and the Fisheries Research Development Corporation and involved 25 researchers. Their contributions are gratefully abowledged; in particular, wethank R. Grandperrin, A. Crosnier, P. Bouchet, K. Gowlett-Holmes,M. Lewis, S. Cummings, J.S. Philippe and the crews of the RVAlu and Southern Surveyor. JAKreceived support from a National ‘i Research Council Senior Research Fellowship. Correspondence and requests for materials should be addressed to 1.A.K. (e-mail: [email protected]).

...... *...... I Excitatory neuron Digital selection and analogue amplification coexist in a . cortex-inspiredsilicon circuit Richard H. R. Hahnloser*S, Rahul Sarpeshkart5, Misha A. MahoWald*, Rodney J. Douglas* & H. Sebastian SeungtS 4 .. inhibitory neuron *Institute of Neiiroinformatiu ETHZ/UNIZ, Wnterthiirerstrasse 190, 8057 Zürich, Swikedand 7Bell Laboratories, Murray Hill, New Jersey 07974, USA $Department of Brain and Cognitive Sciences and $Department of Electrical Engineering and Compiitez Science, Mn; Cambridge2Massachusetts 02139, USA

Digital circuits such as the flip-flop use feedback to achieve multistability and nonlinearity to restore signals to logical levels, for example 0 and 1. Analogue feedback circuits are generally designed to operate linearly, so that signals are over a range, and the response is unique. By contrast, the response of cortical circuits to sensory stimulation can be both multistable and graded’“. We propose that the neocortex combines digital selec- Figure 1 Silicon implementation of a recurrent network with a ring architecture. The tion of an active set.of neurons with analowe response by synaptic connections are consistently coloured in the a-c. a, Each excitatory neuron E and dynamically varying the positive feedback inherent in its recur- makes connections @lue) with itself four neighbours on the ring, of strength a,,Thf the p. rent connections. Strong positive feedback causes differential inhibitory neuron I makes global recurrent connections on ring of strength Green the instabilities that drive the selection of a set of active neurons presynaptic connections; red, postsynaptic connections. 6, The core of excitatory under the constraints embedded in the synaptic weights. Once neuron is the current mirror labelled rectification which contains a large capacitance. selected, the active neurons generateweaker, stable feedback that c, The core of the inhibitory neuron is the bipolar current mirror that sets the gain of provides analogue amplification of the input. Here we present our

NATUREIVOL 405 122 JUNE 2000 (m.nature.com 947 ..