Proc. Natl. Acad. Sci. USAVol. 82, pp. 3707-3711, June 1985Ecology

Larval settlement rate: A leading determinant of structure in anecological community of the marine intertidal zone

(barnacles/Balanus glandula/population dynamics/demography)

S. GAINES AND J. ROUGHGARDENHopkins Marine Station, Stanford University, Pacific Grove, CA 93950

Communicated by L. R. Blinks, January 22, 1985

ABSTRACT Field studies demonstrate that the populationstructure of the barnacle Balanus glandula differs betweenlocations of high and low larval settlement rate. These observa-tions, together with results from a model for the demographyof an open, space-limited population, suggest that the settle-ment rate may be a more important determinant of rockyintertidal community structure than is presently realized.Locations with a low larval settlement rate exhibit a generallylow abundance of barnacles that varies slightly within yearsand greatly between years, reflecting yearly differences insettlement. Locations with a high-settlement rate exhibit agenerally high abudance of barnacles. However, the abundancevaries greatly within years with a significant oscillatory com-ponent (period, 30 weeks) and only slightly between yearsregardless of yearly differences in settlement. At the low-settle-ment location mortality of barnacles is independent ofthe areaoccupied by barnacles. At the high-settlement location mortal-ity is cover-dependent due to increased predation by starfish onareas of high barnacle cover. In both locations the cover-independent component of mortality does not vary with ageduring the first 60 weeks. As assumed in the demographicmodel, the kinetics of larval settlement can be described as aprocess in which the rate of settlement to a quadrat isproportional to the fraction of vacant space within the quadrat.Generalizations that the highest species diversity in a rockyintertidal community is found at locations of intermediatedisturbance, and that competition causes zonation betweenspecies of the barnacle genera Balanus and Chthamalus, seemto apply only to locations with high-settlement rates.

Many members of ecological communities in the marineintertidal zone have a life history consisting of pelagicallydispersed larvae and sessile, space-limited adults. Familiarexamples include barnacles and mussels (1-5). Larval phasesas short as 2-3 weeks may lead to transport oflarvae in excessof 100 km (6-10). As a result, recruitment to a local sectionof the shore is from larvae that likely originated at other sites.Hence, a local section of shore is an open population that isnot satisfactorily treated by the models of primarily closedpopulations applied to terrestrial ecological communitiesover the last decade (11-13).A model for the demography and population dynamics of

an open population with space-limited recruitment has re-cently been proposed for marine populations like barnacles(14-17). Here we report that qualitative aspects of thepopulation dynamics of barnacles accord with theoreticalexpectations of this new model. Specifically, with a lowlarval settlement rate, the population tends to approach asteady-state abundance. That abundance, however, is sensi-tive to stochastic variation in the settlement rate over time orspace; we term the condition a "stochastically sensitive"

steady state. With high settlement there is an oscillatorycomponent to the temporal pattern of abundance that, ac-cording to the model, may be caused in part by the time laginherent in the space-dependent recruitment of rapidly grow-ing organisms.

This study also confirms a key assumption of the open-population demographic model, that settlement to vacantspace can be treated as a process in which the rate ofsettlement in a quadrat is proportional to the fraction ofvacant space in it, with a constant of proportionality specificto location and time (including season). Further, this studyreveals that disturbance (mortality that removes space-oc-cupying organisms) is a cover-dependent process for bar-nacles subject to predation by the starfish Pisaster ochraceusand that the cover-independent component of survivorship isindependent of age for at least the first 60 weeks of life.

Finally, we note that the importance of the larval settle-ment rate in population dynamics implies that generalizationsabout the role of competition in causing patterns of zonationthrough mechanisms relying on physical contact (1, 2, 18) andthe "intermediate disturbance principle," a postulate that thehighest species diversity is found in communities subject toan intermediate degree of density-independent mortality(19-21), should be qualified for marine intertidal communi-ties. These generalizations seem to pertain only to situationsin which the settlement rate is high.

STUDY SITES AND METHODS

The study site consists of rocky intertidal habitat adjacent toHopkins Marine Station on Monterey Bay in central Califor-nia. The abundance of the high intertidal barnacle Balanusglandula varies from nearly complete cover on the seawardmargin ofthe habitat to sparse cover on shoreward rocks. Wereport data primarily from two sites: KLM, a group of rocksnear shore, and Pete's Rock, a similar group of rocks, butlocated between KLM and the seaward margin of the habitat.The sites are separated by ""30 m. Also, additional data froma site, Bird Rock, on the seaward margin of the habitat andfrom a site between KLM and Pete's Rock are mentioned inthe Discussion. Quadrats (34.6 cm2, four per site) were placedin the center of the Balanus-Endocladia zone (described inref. 22) and lie at approximately + 1 m above mean low lowwater. In this zone the barnacle B. glandula is usually thespecies occupying the greatest amount of space, and the algaEndocladia muricata also occurs conspicuously. In addition,the barnacles Chthamalus fissus and Chthamalus dalli andthe alga Gigartina-papillata are also sessile space occupiers.The mobile organisms that interact with B. glandula consistprimarily of the starfish P. ochraceus, the Thaidid snailsNucella emarginata, Acanthina punctulata, and Acanthinaspirata, and the limpets Collisella scabra and Collisellaparadigitalis. At both Pete's Rock and KLM the four smallquadrats appear to be representative of "10 m2 of nearbysubstrate with about the same species composition andabundance. They are thus descriptive of two points along a

3707

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 13

, 202

0

3708 Ecology: Gaines and Roughgarden

gradient of barnacle abundance and not of two large, discreteregions.The quadrats were monitored with photographs taken

using a camera mounted on a rigid frame. The frame wasplaced over stainless steel bolts seated in the rock to holdregister; the system provides a resolution of 100 am. Theoutline of the basal area of all organisms and the center of theaperture for all barnacles were digitized from the photo-graphs. The software for digitizing with a Hewlett Packard7221A plotter-digitizer, and for managing the large data setinvolved (>2000 individuals in this study), is written in Pascal(Oregon Software Pascal-2) and is available upon request.

Quadrats were monitored daily during August and Sep-tember 1982, weekly during April through July 1983 and 1984(the principal settlement season), and biweekly or monthly atother times. Also, during April to July in 1983 and 1984photographs on consecutive low tides (consecutive low-lowsin the mixed semidiurnal schedule that occurs at HopkinsMarine Station) were taken to detect any unusual mortalitythat might affect barnacles within hours after settlement. Theabundance and area occupied by barnacles and other orga-nisms and the growth rate and mortality rate of barnacleswere determined from these data. Mortality caused byPisaster predation could be scored by the characteristicdiscrete patches of basal plates. Thaidid and other predatorsthat leave single, empty tests are not distinguished from eachother by these photographic methods.

B. glandula settle as cyprid larvae (a terminal nonfeedingstage). The cyprids may move about (23) before taking thelocation at which they metamorphose into young adults(spat). They generally metamorphose only on unoccupiedspace; cyprids metamorphose on the tests of adult barnaclesonly if vacant space is <5% (unpublished data). Hence, thesettlement rate is expressed as the number of cyprids per unitvacant space per week.The relation between settlement rate and free space was

determined by subdividing each quadrat into equal sectorsand calculating the settlement rate into each sector as afunction of the percent free space in it. To compare settle-ment among quadrats that differ markedly in their overallsettlement rates, the rate in each sector is expressed relativeto the settlement in the whole quadrat. If settlement isproportional to the amount of free space, then the number ofcyprids settling in a sector (n,) should be ni = sFi, in whichs is the whole quadrat settlement rate, and Fi is the free spacein sector i. Rearranging and including the sector area (Ai),ni/(sAi) = Fi/Ai. Therefore, plotting the relative measure ofsettlement on the left as a function of the percent free spacein the sector should produce a 450 line if settlement rates area simple- linear function of free space. High-settlementquadrats were subdivided into 12 sectors; low-settlementquadrats were subdivided into only 6 sectors.

RESULTS

Uncrowded B. glandula individuals grow in basal areaaccording to a power law (basal area = 5.3 10-5*xl 99, R2 =0.9873, P << 0.01, units of basal area in cm2 and age inweeks). The growth rates at KLM (where all individuals wereuncrowded) are identical to those individuals at Pete's Rockhaving only zero or one contacting neighbor (F = 1.17, P >0.25 for comparison of logarithm transformed data). Theindividuals at Pete's Rock that contacted two or more otherindividuals usually showed lower growth rates thanuncrowded individuals.Weekly survivorship (Ps) was independent of age (x) at

both sites for the first year of life (Fig. 1). There was no highermortality ofjuveniles as suggested for this and other barnaclespecies (1, 2, 4, 24) despite the presence of limpets. Recruitstracked from the first low tide following settlement had the

1.00r

w 0.75

-0-0 500.

*4W-*-4I*.-

U. i-I-

0.00 L0 15 30 45 60

Age, weeks

FIG. 1. Weekly survivorship versus age for B. glandula. Aster-isks represent data from Pete's Rock, and dots represent data fromKLM. There are no effects of age on survivorship.

same survivorship as older age classes. Barnacle age ex-plained <5% of the variance in survivorship at each site. ForKLM, Px = 0.985 + (6.71.10-5)x, R2 = 0.0032, P > 0.5; forPete's Rock, Px = 0.939 + (5.67X10-4)x, R2 = 0.038, P > 0.5.Weekly survivorship depends on the amount of free space

(Fig. 2). There is a precipitous decline in weekly survivorshipwhen barnacle cover is high caused primarily by predationfrom the starfish P. ochraceus. In fact, all 21 points repre-senting survivorship values <0.8 in Fig. 2 were associatedwith discrete patches of basal plates characteristic of Pisasterpredation.

Settlement rate at the two sites is directly proportional tothe amount of unoccupied space (Fig. 3). More specifically,the relative number of individuals settling in a subsector of aquadrat is directly proportional to the amount of free spacein the subsector. Settlement is proportional to free spaceprovided the free space is distributed around existing adults.In contrast, large patches of bare space (>50 cm2) tend to becolonized first around the perimeter of the patch (25).There were large differences in the rate of settlement

between sites and between years (Table 1). For each of the3 years settlement was =20-fold higher at Pete's Rock thanat KLM. Also, 1983 had two to four times the recruitmentrates of 1982 and 1984 for all quadrats but one (quadrat 4 atKLM).At KLM there was little within-year variation in free space,

yet large between-year changes that mirrored yearly differ-ences in settlement rates (Table 1). A representative trajec-tory is shown in Fig. 4. Notice that the large decline in free

1.00 r

0.75 I-00.0eCL 0.501-.

0.25 -

0 20 40 60 80 100% free space

FIG. 2. Probability of survival through 1 week for B. glandula asa function of the percentage of free space in the quadrat. Data arefrom the Pete's Rock site only. Quadrats at KLM had free space>50% and no density-dependent mortality. The high mortality undercrowded conditions primarily represents predation by starfish.

Proc. Natl. Acad. Sci. USA 82 (1985)

.:. .. "I -1.1 . 4 Z;, .,w

. . .. 7

Dow

nloa

ded

by g

uest

on

Dec

embe

r 13

, 202

0

Proc. Natl. Acad. Sci. USA 82 (1985) 3709

1.00-r

0.75p-v:u

CO

100

750

0.(A

w 5020

25

0.50 -

0.25 -

0 20 40 60 80 100% free space

FIG. 3. Settlement (S) ofB. glandula larvae relative to the percentfree space. Data were obtained by subdividing quadrats and plottingthe relative settlement rate against the percent free space in thatsubquadrat. The approximately 450 line through the origin indicates"mass action" settlement kinetics.

space corresponds to the large 1983 settlement burst and thatthe average free space for 1984 is increasing corresponding tothe lower settlement of 1984. For all KLM quadrats in allyears, the year's average free space reflects the intensity ofsettlement for that year (Table 1). An increase in settlementrelative to the previous year leads to a decrease in averagefree space and vice versa. Indeed, quadrat 4 at KLM, thequadrat where settlement rates were aberrant and increased

Table 1. Yearly patterns of free space availability andrecruitment

Within-year Recruits,% free space variance as S no. per

Year Mean Variance of total variance cm2/week

KLM1982 88.42 2.46 0.89 0.071

89.01 5.30 1.98 0.06387.70 15.04 4.41 0.07294.00 6.33 1.67 0.044

1983 52.89 19.79 7.13 0.2862.36 23.19 8.68 0.2561.93 31.01 9.08 0.1881.77 17.12 4.51 0.059

1984 64.40 19.39 6.98 0.07565.92 26.54 9.94 0.06664.04 16.56 4.85 0.05360.59 22.67 5.97 0.074

Pete's Rock1982 50.33 207.62 72.94 1.08

43.32 181.61 65.47 1.2447.82 203.91 84.11 1.3836.61 188.99 72.69 1.06

1983 50.54 181.33 63.75 4.0242.89 274.78 99.06 4.3649.33 220.76 91.07 3.3442.72 229.23 87.86 3.22

1984 43.54 298.95 105.10 1.5147.92 153.52 58.84 1.33

Years are assumed to start on May 1. Recruitment is measured asthe number of cyprids metamorphosing per cm2 of available freespace per week averaged over the settlement season. Settlementrates for 1982 were partly estimated by extrapolating backwards frombarnacle size distributions present when the studies were initiated.Data from only two quadrats are shown in Pete's Rock for 1984because the remaining quadrats were invaded and dominated by thealga E. muricata. Within-year variances can exceed variancescalculated from the entire 3-year trajectory (as in 1984 at Pete's Rock)if variability for a quadrat within 1 year is substantially larger thanwithin the other 2 years.

0.60

0.40

0.20

8/82 2/83 8/83 2/84 8/84

FIG. 4. Percent free space (solid line, scale on left vertical axis)and rate of larval settlement (dashed line, scale on right vertical axis)for 2 years in a quadrat from the KLM area at Hopkins MarineStation. 1983 had about four times the settlement of 1984. Thetrajectory of free space reflects the settlement history.

progressively from 1982 through 1984, illustrates the sensitiv-ity of average free space levels to fluctuations in the magni-tude of settlement. This quadrat showed a progressivedecline in average free space each year.

Within-year variances in free space at KLM are consis-tently 1 to 2 orders of magnitude less than total samplevariances for all quadrats (Table 1); most of the totalvariability comes from year-to-year differences in averagefree space. All within-year variances are significantly lessthan total variances (variance ratio test on arcsin transformeddata, P < 0.001 in each test). Also, at KLM barnacles of allsizes are intermingled with free space.At Pete's Rock the average free space is consistently lower

than at KLM (t tests; P < 0.01 for each year) and shows nosignificant difference between years (Table 1) despite similar3- to 4-fold fluctuations in recruitment as seen at KLM.Within years, however, there are large swings in the amountof free space. A representative trajectory appears in Fig. 5.The large swings are not driven by bursts of settlement, andthere is no correspondence between increasing settlementand decreasing average free space (or vice versa) for any ofthe Pete's Rock quadrats (Table 1). Within-year samplevariances for free space are not significantly different (vari-ance ratio test, P > 0.10 for each) from the total samplevariance.Time series analysis of the free space trajectories was used

to detect cyclical changes. Autocorrelation functions forquadrats from KLM and Pete's Rock appear in Fig. 6. KLMquadrats show the classical pattern for nonperiodic randomfluctuations; the autocorrelation rapidly declines to zero asthe lag increases. The autocorrelation for the Pete's Rockquadrats indicate a statistically significant oscillatory com-ponent with a period of ==30 weeks. The cycling processstarts with heavy settlement onto vacant space (typicallygenerated by Pisaster predation), followed by rapid growth,and then by recurrence of starfish predation when free spaceis nearly exhausted. Moreover, free space tends to bedistributed in patches that are later filled by members of asingle cohort, with the result that barnacles of the same sizetend to occur in patches.

DISCUSSION

Spatial variation in mortality after settlement is not theprimary cause of spatial variation in barnacle density atHopkins Marine Station. Density-independent mortalityrates are nearly identical for the two sites. And, whencover-dependent mortality is also considered, barnacles havehigher overall survivorship at the low-density shorewardsites (Fig. 1). Furthermore, dissections of barnacles during

Ecology: Gaines and Roughgarden

II

I

6.

111.

II 1. f 11I I 'Ik.- I ,.10 a 0 , 11

Dow

nloa

ded

by g

uest

on

Dec

embe

r 13

, 202

0

3710 Ecology: Gaines and Roughgarden

C)~~~~~~~~~~~~1co

5

8/82 2/83 8/83 2/84 8/84

FIG. 5. Percent free space (solid line, scale on left vertical axis)and rate of larval settlement (dashed line, scale on right vertical axis)for 2 years in a quadrat from the Pete's Rock area of Hopkins MarineStation. The overall settlement rates are -20 times as high as in theKLM region. (Note the change in scale on the right vertical axis fromFig. 4.) The trajectory of free space is effectively independent of thesettlement history and displays an oscillatory component.

1983 indicated the size-specific reproductive output was thesame at both sites (unpublished data). The density-independent component ofgrowth is also the same at the twosites. Thus, barnacles perform as well or better at shorewardsites, where their abundance is lowest, than at seaward sites,where their abundance is higher.The primary difference among the sites that could account

for the observed variation in average barnacle abundance isthe settlement rate. At shoreward sites settlement is limiting;the abundance each year is low and mirrors the magnitude ofsettlement that year (Table 1). Similar sensitivity of apopulation's abundance to the settlement rate has been notedin coral reef fish (26, 27), commercial fish stocks (28), andepiphytic bryozoans (29). At seaward locations, settlementrates, though different each year, were sufficient to producenearly the same high average abundance of barnacles eachyear. Further studies are necessary to determine why thesettlement rate varies throughout the habitat.At Pete's Rock, a high-settlement location, a significant

periodic component in the temporal pattern of barnacle coverwas detected (Fig.. 6). According to a model for thedemography of an open population with space-limited re-cruitment (14, 15), a possible explanation for the oscillatorycomponent is the destabilizing effect of growth in a sessileorganism on population dynamics. Growth in basal areaeffectively introduces a time lag into the population dynamicsbecause recruitment is proportional to the amount of un-occupied space currently available and not to the amount ofspace that will be available when the animals have grown totheir expected area. More larvae thus may settle than thesystem can ultimately support, and cover-dependent mortal-ity can periodically reopen space for subsequent recruitment.The effect becomes increasingly pronounced as the settle-ment rate increases.

Oscillations in barnacle cover may also be seen in the agedistribution. Waves of cohorts are observed (14) as patchesof free space undergo the cycle of high recruitment, rapidgrowth, and finally enhanced mortality when free space isnearly exhausted. Glynn (22), in a study ofB. glandula in theearly 1960s, presents 3 years of size class distributions (figure64 in ref. 22) that clearly show such waves of dominant sizeclasses for a site within a few meters of our present Pete'sRock site. This suggests that the periodicity of barnacleabundance may have a long history at this site.The large differences in barnacle population dynamics

between high- and low-settlement sites pose the question ofwhat happens at sites with intermediate settlement rates. Wehave been following six quadrats on a somewhat less regular

A

-1.00

0°O 1.00KB

0.50

-0.50 -

0 10 20 30 40 50Time lag, weeks

FIG. 6. Autocorrelation functions for percent free space overtime. -, Pete's Rock; * * , K, KLM. Quadrats 1 and 2 from each siteare presented in A and quadrats 3 and 4 from each site are presentedin B. In A the quadrats were monitored for 2 years, and in B thequadrats were monitored for 75 weeks (see Table 1 legend). Dashedhorizontal lines are confidence limits (95%) based on the nullhypothesis of white noise fluctuation. Confidence limits in B areslightly larger than inA due to the shorter time course ofthe quadrats.The Pete's Rock autocorrelations indicate statistically significantoscillations with a period of about 30 weeks. Trends due to year-to-year variation in mean free space levels were removed for allquadrats (i.e., the autocorrelation functions are calculated as theproducts of deviations from yearly means rather than the grandmean). Autocorrelations were computed by using values interpolatedto achieve weekly intervals during the times when settlement waslight and sampling effort reduced. However, confidence intervals forthe null hypothesis were computed by using actual sample sizes.

basis that have had settlement rates between those at KLMand Pete's Rock. Barnacle dynamics have switched withyearly changes in recruitment rates. In 1982, settlement rateswere generally low and free space was abundant, as at KLM.However, in 1983, three of the quadrats had sufficiently highrecruitment rates that free space levels declined to <30%.Each of these three quadrats attracted predation by Pisaster,generating the initial phases of the cycling seen at Pete'sRock. The periodicity was not sustained, however, becauserecruitment declined again in 1984, and barnacle abundanceshave persisted at low levels.The kinetics of larval settlement by barnacle cyprids was

found to be a mass action process (Fig. 3)-that is, represent-able as an expression of the form: (rate of settlement) =(const) x (free space), in which (const) is a constant ofproportionality that reflects many factors, including theabundance of larvae in the water, the duration of exposure towater, average bulk flow over the surface, and so forth. Thisresult is surprising because a large repertoire of behavioraltraits has been shown for larvae that lead to substrateselectivity, gregariousness, spacing, and sensitivity tomicroclimate (23, 30-32). These capabilities either are notrealized in the field for B. glandula or combining a relativelylarge number of larvae with individually complex behaviorsmay make the aggregate phenomenon describable by a simple

Proc. Natl. Acad Sci. USA 82 (1985)

Dow

nloa

ded

by g

uest

on

Dec

embe

r 13

, 202

0

Proc. Natl. Acad. Sci. USA 82 (1985) 3711

expression. Nonetheless, the finding supports the use ofmassaction formulae in efforts to model the dynamics of marinepopulations (14, 15).Mechanisms of mortality that lead to intraspecific density

or cover dependence have been known for years in marinesystems. "Hummocking," or the bulging of crowded bar-nacles away from the rock surface, can increase a barnacle'ssusceptibility to removal by waves (4, 33). In this study,crowded barnacles are more susceptible to predation bystarfish than barnacles sparsely intermingled with unoccu-

pied space (Fig. 2). In contrast to the density-independentmortality caused by waves or wave-borne logs (4, 34, 35) theeffects of cover-dependent mortality are "biologically tar-geted" to locations with high cover. Thus, in this community,disturbance should not be viewed as an abiotic factor thatcontinually resets a community to an earlier point on itstrajectory of development; instead, the disturbance is a

mechanism that depends on the internal state of the com-

munity.Settlement in the high intertidal community adjacent to

Hopkins Marine Station appears to play as important a roleas postsettlement processes such as predation and competi-tion. Indeed, the rate of recruitment is a (causally) priorconsideration in determining community structure since itslevel selects the set of factors that subsequently affect adults.At low settlement the community is recruitment-limited andsensitive to stochastic fluctuations in settlement (36, 37). Athigh settlement it is a "mosaic of patches" (34, 35) withintrinsic oscillatory components.

This importance of settlement in determining the com-

munity structure of rocky intertidal communities implies thattwo well-known generalizations should be qualified as per-taining only to high-settlement communities:

(i) A classical pattern of zonation in the intertidal zone

between barnacles of the genera Balanus (Semibalanus) andChthamalus (1, 2, 38) is caused by physical contact; individu-als ofBalanus can overgrow, crush, and pry loose individualsof Chthamalus, resulting in Chthamalus remaining only in azone with physical conditions that Balanus cannot physio-logically tolerate. For this pattern to form, settlement ratesmust be consistently high enough to generate extensivecontact among individuals; otherwise the species will havecompletely overlapping distributions (2). In fact, at HopkinsMarine Station, the combined abundance of the barnacles C.fissus and C. dalli is independent of Balanus abundanceexcept when free space is nearly exhausted. Specifically,when B. glandula occupies <75% of the surface, Balanus andChthamalus densities are uncorrelated (R2 = 0.0049, P >0.5). Balanus has a significant negative effect on Chthamalusabundance where the Balanus abundance is >75% of thesurface. At Hopkins, such high cover of Balanus is onlyreached at the extreme seaward edge of the intertidal habitat(Bird Rock) where settlement rates have generally beenhigher than those at Pete's Rock (unpublished data). At bothKLM and Pete's Rock, Balanus and Chthamalus havecompletely overlapping distributions.

(ii) The "intermediate disturbance principle," a hypothesisthat the highest species diversity in a community occurs atsome intermediate level of disturbance (19, 21), has a tacitassumption of high settlement (36). If settlement rates are lowsuch that extensive contact fails to develop among space-oc-cupiers, then any decline in diversity caused by a possiblehierarchy in competitive ability cannot be expressed. Withlow-settlement rates diversity and disturbance may be in-versely related, regardless of whether interspecific competi-tion is hierarchical.

We thank Charles Baxter, Lawrence Blinks, Sally Blower,Stephen Brown, Thomas Hahn, Steven Hamburg, Peggy

Lubchenco, Jane, Lubchenco, Bill Rice, and Rene Toulson forassistance during the course of this research and for comments on themanuscript. We also gratefully acknowledge support from theDepartment of Energy (Contract EV10108).

1. Connell, J. H. (1961) Ecology 42, 710-723.2. Connell, J. H. (1961) Ecol. Monogr. 31, 61-104.3. Paine, R. T. (1974) Oecologia 15, 93-120.4. Dayton, P. K. (1971) Ecol. Monogr. 41, 351-389.5. Menge, B. A. (1976) Ecol. Monogr. 46, 355-393.6. Crisp, D. J. (1958) J. Mar. Biol. Assoc. U.K. 37, 483-520.7. Okubo, A. (1971) Deep-Sea Res. Oceanogr. Abstr. 18,

789-802.8. Strathmann, R. R. (1974) Am. Nat. 108, 29-44.9. Scheltema, R. S. (1977) in Concepts and Methods of

Biostratigraphy, eds. Kauffman, E. G. & Hazel, J. E.(Dowden, Hutchison, and Ross, Stroudsburg, PA), pp. 73-108.

10. Waldichuk, M. (1958) J. Fish. Res. Bd. Can. 15, 1065-1102.11. May, R. M. (1973) Stability and Complexity in Model Ecosys-

tems (Princeton Univ. Press, Princeton, NJ) 235 pp.12. Roughgarden, J. (1979) Theory of Population Genetics and

Evolutionary Ecology: An Introduction (Macmillan, NewYork).

13. Diamond, J. & Case, T. (1985) Community Ecology (Harper &Row, New York).

14. Roughgarden, J., Iwasa, Y. & Baxter, C. (1985) Ecology 66,54-67.

15. Roughgarden, J., Gaines, S. & Iwasa, Y. (1984) in Exploitationof Marine Communities, ed. May, R. M. (Springer-Verlag,New York), pp. 111-128.

16. Roughgarden, J. & Iwasa, Y. (1985) Theor. Popul. Biol., inpress.

17. Iwasa, Y. & Roughgarden, J. (1985) Theor. Popul. Biol., inpress.

18. Wethey, D. (1983) Biol. Bull. 165, 330-341.19. Paine, R. T. & Vadas, R. L. (1969) Limnol. Oceanogr. 14,

710-719.20. Connell, J. H. (1975) in Ecology and Evolution of Communi-

ties, eds. Cody, M. L. & Diamond, J. M. (Harvard Univ.Press, Cambridge, MA), pp. 460-490.

21. Lubchenco, J. (1978) Am. Nat. 112, 23-39.22. Glynn, P. (1965) Beaufortia 12, 1-198.23. Crisp, D. J. (1974) in Chemoreception in Marine Organisms,

eds. Grant, P. T. & Mackie, A. M. (Academic, London), pp.178-265.

24. Hawkins, S. J. (1979) Dissertation (Univ. Liverpool,Liverpool, England).

25. Toulson, R. (1984) Dissertation (Stanford Univ., Stanford,CA).

26. Victor, B. (1983) Science 219, 419-420.27. Sale, P. (1984) in Ecological Communities Conceptual Issues

and the Evidence, eds. Strong, D. R., Simberloff, D., Abele,L. & Thistle, A. B. (Princeton Univ. Press, Princeton, NJ), pp.478-490.

28. Sissenwine, M. P. (1984) in Exploitation of Marine Communi-ties, ed. May, R. M. (Springer-Verlag, New York) pp. 59-94.

29. Yoshioka, P. M. (1982) Ecology 63, 457-468.30. Strathmann, R. R. & Branscomb, E. S. (1979) in Reproductive

Ecology of Marine Invertebrates, ed. Stancyk, S. E. (Univ.South Carolina, Columbia), pp. 77-89.

31. Lewis, C. A. (1978) in Settlement and Metamorphosis ofMarine Invertebrate Larvae, eds. Chia, F. S. & Rice, M. E.(Elsevier, New York), pp. 207-218.

32. Wethey, D. (1984) J. Mar. Biol. Assoc. U.K. 64, 687-698.33. Barnes, H. & Powell, H. (1950) J. Anim. Ecol. 19, 175-179.34. Levin, S. A. & Paine, R. T. (1974) Proc. Natl. Acad. Sci. USA

71, 2744-274735. Paine, R. T. & Levin, S. A. (1981) Ecol. Monogr. 51, 145-178.36. Underwood, A. & Denley, E. (1984) in Ecological Communi-

ties: Conceptual Issues and the Evidence, eds. Strong, D. R.,Simberloff, D., Abele, L. G. & Thistle, A. B. (Princeton Univ.Press, Princeton, NJ), pp. 151-180.

37. Underwood, A., Denley, E. & Moran, M. (1983) Oecologia 56,202-219.

38. Wethey, D. S. (1983) Biol. Bull. 165, 330-341.

Ecology: Gaines and Roughgarden

Dow

nloa

ded

by g

uest

on

Dec

embe

r 13

, 202

0