Are single species toxicity tests alone adequate for estimating environmental hazard?

John Cairns, Jr.Biology Department and University Center for Environmental Studies, Virginia Polytechnic Institute andState University, Blacksburg, VA 24061, U.S.A.

Keywords: single species toxicity tests, system level tests, application factors, hazard evaluation

Abstract

Most biologists agree that at each succeeding level of biological organization new properties appear thatwould not have been evident even by the most intense and careful examination of lower levels of organization.These levels might be crudely characterized as subcellular, cellular, organ, organism, population, multispe-cies, community, and ecosystem. The field of ecology developed because even the most meticulous study ofsingle species could not accurately predict how several such species might interact competitively or inpredator-prey interactions and the like. Moreover, interactions of biotic and abiotic materials at the level oforganization called ecosystem are so complex that they could not be predicted from a detailed examination ofisolated component parts. This preamble may seem platitudinous to most biologists who have heard thismany times before. This makes it all the more remarkable that in the field of toxicity testing an assumption ismade that responses at levels of biological organization above single species can be reliably predicted withsingle species toxicity tests. Unfortunately, this assumption is rarely explicitly stated and, therefore, oftenpasses unchallenged. When the assumption is challenged, a response is that single species tests have been usedfor years and no adverse ecosystem or multispecies effects were noted. This could be because single speciestests are overly protective when coupled with an enormous application factor or that such effects were simplynot detected because there were no systematic, scientifically sound studies carried out to detect them.Probably both of these possibilities occur. However, the important factor is that no scientifically justifiableevidence exists to indicate that degree of reliability with which one may use single species tests to predictresponses at higher levels of biological organization. One might speculate that the absence of such informa-tion is due to the paucity of reliable tests at higher levels of organization. This situation certainly exists butdoes not explain the lack of pressure to develop such tests. The most pressing need in the field of toxicitytesting is not further perfection of single species tests, but rather the development of parallel tests at higherlevels of organization. These need not be inordinately expensive, time consuming, or require any more skilledprofessionals than single species tests. Higher level tests merely require a different type of biologicalbackground. Theoretical ecologists have been notoriously reluctant to contribute to this effort, and, as aconsequence, such tests must be developed by associations of professional biologists and other organizationswith similar interests.

Introduction useful and are presently the major and only reliablemeans of estimating probable damage from an-

Although this discussion may appear hostile to thropogenic stress. However, many in the scientificsingle species toxicity testing efforts, it is not in- community are certainly aware of the need fortended to be. Single species tests are exceedingly community and system level toxicity testing. How

Hydrobiologia 100, 47-57 (1983).© Dr W. Junk Publishers, The Hague. Printed in The Netherlands.

48

then does one account for the difference betweenawareness and performance? As an illustration ofwhy such a difference exists, consider this scenariofrom a hypothetical workshop entitled 'The Con-tributions of Theoretical Ecology to Pollution As-sessment'. On the first day of this workshop, reas-suring exchange occurs among participants onideas of energy flow, ecosystem dynamics, multipleaggregate variables, niche packing, and the like. Onthe second day, the group which must make use ofthis information confronts the theoretical ecologistwith one or more site-specific problems and asksspecifically how theoretical ecology can be used in aparticular situation. It usually becomes abundantlyclear that no system level measurements exist onwhich a concensus occurs among theoretical ecolo-gists on use, interpretation, validity, and predictivevalue! Following this exercise, a retreat ensues tomeasurements associated with single species or atleast those that are clearly not ecosystem level pa-rameters. This is usually accompanied by a call formore research. A majority of scientists have at-tended such workshops; many have probably at-tended a number of such meetings with roughlysimilar scenarios.

The call for more research before specific re-commendations can be made usually falls on deafears. This is a pity because truly more research isneeded. I greatly fear that ecologists will lose credi-bility among practioners of pollution assessmentbecause they have correctly called attention to rath-er vast and significant problems without followingthrough with a professional concensus on whichsystem level tests to carry out, measurements tomake, methods to formally approve, and so on.There is even danger in calling attention to deficien-cies in single species tests in predicting system leveleffects because it may cause some practioners andregulators to doubt the efficacy of any biologicalmeasurements. In fact, single species tests haveproven remarkably effective to estimate responsesat high levels of biological organization despiteconsiderable theoretical deficiencies in using them.Nevertheless, if the field of environmental toxicol-ogy and chemistry is to continue to evolve, thesedeficiencies must be identified and corrective mea-sures taken.

Another problem is the inescapable conclusionthat research (to provide system level responses ofthe type just described) is not sufficiently theoreti-

cal for some funding agencies and too theoreticalfor others. There is some evidence that this problemhas been recognized and has been addressed in aminor way. Other blocks to development of ade-quate ecosystem level tests include difficulties ingetting specialists in necessary disciplines to collab-orate when their salary increases and/ or tenure andpromotion may be judged by specialists with anuncharitable view toward group research.

A fundamental problem with present toxicitytesting protocols is that they often estimate effectson an ecosystem as if the ecosystem were merely acollection of species exposed to a single pure com-pound under constant conditions. The need to gobeyond single species testing to evaluate hazard tothe environment posed by toxic chemicals is gain-ing momentum. A parallel thrust involving thestudy of increasingly complex systems for evaluat-ing environmental fate of chemicals is also in pro-gress. Although the outcome of these developmentsis not evident, it is abundantly clear that the need toexamine both toxicity and environmental fate ofchemicals in a more environmentally realistic way isnow a sine qua non. As is the case with most devel-oping fields, it would be unfortunate if these newapproaches and new methods were used for regula-tion before a substantial and sound data base vali-dating their efficacy has been produced. It would beequally foolish to retard development of such meth-ods because they are not of immediate practicalbenefit. Both society and industry have much togain from the production of more accurate meansof estimating hazard of chemicals in the environ-ment. While single species tests are far from perfect,their development has far outstripped developmentof determination of toxicological responses athigher levels of biological organization. Althoughthis article addresses higher levels of biological or-ganization than single species, it is not intended todenigrate the research at lower levels (e.g., enzyme)which might enable predictive capabilities formechanisms of toxicity to be disclosed.

Robert MacArthur (1975) said 'Scientists'areperennially aware that it is best not to trust theoryuntil it is confirmed by evidence. It is equally true,as Eddington pointed out, that it is best not to puttoo much faith in facts until they have been con-firmed by theory'. Since Hart et al. (1945) produceda method for toxicity testing that was soon en-dorsed by a committee of the Water Pollution Con-

49

trol Federation (Doudoroff et al. 1951), quite alarge number of facts have been generated in thefield of aquatic toxicology: water quality often maymarkedly mediate the expression of toxicity, somechemicals will interact synergistically or antagonis-tically, life history stages of a single species may notbe comparably sensitive to different toxicants, anddifferent species may alter their relationships toeach other in terms of sensitivity to toxicants so thatknowing response relationships to chemical A doesnot ensure prediction of relative sensitivity to chem-ical B. As a consequence, predictive value of toxici-ty tests has remained low, and transferability ofinformation from one species to another and fromone level of biological organization to another hasnot been satisfactory.'Use of application factors tocompensate for areas of ignorance or absence ofdata has not reached a stage of development wheresubstantive scientific justification is available forthe efficacy of the factors presently used. A certaindegree of safety can be achieved by making thefactors as large as the worse possible case demands,but sanitary engineers attempting to achieve theselevels have found them technologically or economi-cally impossible. In short, the period since 1945might be characterized as an era where much evi-dence accumulated but little integrating theory sur-faced. No integrating hypothesis was produced topull these facts together or explain their relation-ship to each other.

Over the 30-year period from 1945 to 1975, anenormous toxicological data base foraquatic organ-isms had been generated. This data base was solarge that it was beyond the capability of any indi-vidual to fully comprehend all details, even whenthe information was subdivided into a single groupof organisms such as fish. However, an event oc-curred that showed that this very substantial database was inadequate in several notable aspects: (1)the amount of information on a particular chemicalwas probably inadequate, (2) the kinds of informa-tion generated on a particular chemical were gener-ally inadequate for making a scientifically justifia-ble estimate of hazard, and (3) the transfer ofinformation on one chemical to estimate with preci-sion the hazard of another did not appear feasible.

Under the provisions of the June 7, 1976, consentdecree, the U.S. Environmental Protection Agency(USEPA) was directed to issue Federal Water Pol-lution Control Act effluent limitations and guide-

lines, new source standards of performance, andthree treatment standards for 65 identified toxicpollutants. Implementation of the directives of thisdecree began when the USEPA drafted WaterQuality Criterion Documents for each of the indi-vidual 65 pollutants. These criterion documentsreviewed all pertinent literature available on theparticular chemical and attempted to determine ac-ceptable limits not to be exceeded for the protectionof aquatic life and human drinking water supplies.The Water Quality Committee of the USEPAScience Advisory Board was charged with evaluat-ing the 65 criterion documents. The report of thiscommittee to the Science Advisory. Board statedthat no documents had conclusions that were scien-tifically justifiable. This report was accepted by theExecutive Committee of the Science AdvisoryBoard and transmitted to the Administrator ofUSEPA (then Douglas Costel). It is worth emphas-izing that the court issuing the consent decree didnot allow sufficient time for the USEPA to generateits own data base, and, therefore, USEPA wasforced to prepare the criterion documents with dataalready available in the literature or documents inthe open literature that generally were prepared forsome other purpose. This event provided unmis-takable proof that data to be used for the hazardevaluation must be systematically generated forthat purpose.

Another event that had a major influence ontoxicity testing and hazard evaluation was the pas-sage of the Toxic Substances Control Act (TSCA)that became law with President Ford's signature onOctober 11, 1976. This act represents an attempt toestablish a mechanism whereby the hazard of achemical substance to human health and the envir-onment can be assessed before the substance isintroduced into the environment. The enactment ofTSCA served as a powerful new stimulant to devel-opment of testing procedures to evaluate hazardassociated with potentially toxic substances to hu-man health and the environment.

These events and others clearly indicated theneed for the development of a strategy for hazardevaluation that led to the production of a series ofbooks now commonly referred to in the professionas Pellston I, II, III, and IV. Pellston I (Cairns et al.1978) advocates linkage of the environmental con-centration of a chemical (the term is meant to in-clude such things as partitioning, transformation

50

processes, persistance, etc.) with the concentrationproducing no adverse biological effects. The degreeof uncertainty or lack of confidence in estimatingthese two concentrations was a function of theirproximity to each other. This view is summarizedgraphically in Fig. 1. Pellston II (Dickson et al.1979) examines protocols used in various indus-trialized countries for systematically generating thedata base necessary for a scientifically justifiablehazard evaluation. Pellston III (Maki et al. 1980)

Ca

4O-

.0

U

.4-

0

C0

40-a4-

COC0U

examines biotransformation processes and theirrole in estimating environmental concentration of achemical. Pellston IV (Dickson et al. 1982) dealswith modeling the fate of chemicals in the environ-ment.

These and other publications had as a primarygoal the development of an underlying strategy forestimating hazard. This strategy must be based onsound science and professional judgment andshould be as cost-effective as possible. The most

Highest test concentrationproducing no biologicaleffects.

Highest expected environmentalconcentration.

m

I 2 3 4 5 6Sequential Tests of Hazard Assessment

Procedure

Fig. 1. Diagrammatic representation of a sequential hazard-assessment procedure demonstrating increasingly narrow confidence limitsof estimates of no-biological effect concentration and actual-expected-environmental concentration. (Reprint with permission fromASTM STP 657, Estimating the Hazard of Chemical Substances to Aquatic Life, Copyright ASTM, 1916 Race St., Philadelphia, Pa.,19103.)

51

important consequence of these events just de-scribed has been to direct attention to the informa-tion content of data being generated (i.e., the facts)and ways in which data will be used! This will, inturn, add the additional requirement that data notonly be precise, reliable, reproducible, and so on,but also be suitable for the use or estimates pro-posed! If the problem is viewed in this fashion, itbecomes abundantly clear that the types of toxicitydata now being generated are qualitatively deficientfor their intended purpose! Toxicity tests shouldprovide information that will facilitate predictionsof the concentrations that will not harm livingthings in the environment at all levels of biologicalorganization! The purpose of this manuscript is topresent the view that single species tests alone areinadequate for this urpose.

Discussion

Some very important questions are related totesting at different levels of biological organizationthat deserve serious attention:

1. Can single species tests be used to predict re-sponses reliably at other levels of biological or-ganization?

2. In estimating the effects of chemicals on popula-tions, multispecies assemblages, communities,and ecosystems, what are the limitations of la-boratory science? In other words, are differentdegrees of environmental realism possible in thelaboratory or under laboratory conditions at dif-ferent levels of biological organization?

3. What should be the balance of toxicity tests atdifferent levels of biological organization inorder to make a valid estimate of hazard?

4. How should tests at different levels of biologicalorganization be sequenced?

5. What criteria should be used to validate labora-tory predictions in 'real world' field situations?

The complexity and uniqueness of each ecosystemhas mitigated against ready transfer in general in-formation from one site to another. Thus, manyfield studies are situation bound and highly sitespecific. Can the transferability of informationfrom one site to another be enhanced by the devel-opment of mathematical models?

1. Can single species tests be used to predict re-sponses reliably at other levels of biological organi-zation?

One primary justification for using single speciestests as a basis for estimating concentrations thatwill not prove harmful to communities and ecosys-tems is that if the most sensitive species is selectedand concentration standards are set on that basis,then all other species will be protected. Since only asmall percentage (probably less than 1%) of allfreshwater species can be maintained in the labora-tory sufficiently well to satisfy the requirement thatno more than 10% of the control organisms expireduring the course of tests, it seems quite unlikely thatthe most sensitive species will be selected for testing.In virtually every instance, the most sensitive spe-cies is being selected from a limited array of testspecies and extrapolating is being done from thoseresults. Lest the discussion that follows be misun-derstood, it is not intended to be an attack on singlespecies toxicity testing. Such tests are essential forobtaining information on concentrations and dura-tions of exposures to chemicals that result inchanges in survival, reproduction, physiology, bio-chemistry, and behavior of individuals within par-ticular species. One can question the scientificjusti-fication of using single species tests to predictchanges in competition, predation, communityfunction, ecosystem energy flow, and nutrient cy-cling. These are only a few of the many characteris-tics of ecosystems that either cannot be predictedfrom single species tests or for which there is insuf-ficient evidence that the prediction is scientificallyjustifiable. Although practioners of single speciestoxicity testing may not state that the results ofthese tests can be used to protect biological systemsof greater complexity than single species, the impli-cation is definitely present. The public believes thatwhen a concentration is purported to produce noadverse biological effects then the effects so de-scribed go beyond the kinds of effects that are char-acteristic of single species responses. If the field ofenvironmental toxicology and chemistry is toprosper, 'truth in packaging' is mandatory in termsof the limitations, as well as the strengths, of singlespecies toxicity tests now so widely used.

One common argument advanced by people whofavor continuation of primary reliance on singlespecies testing is that no significant ecological disas-

52

ters have occurred when carefully carried out singlespecies tests were used. Of course, this could bemerely due to the fact that single species tests arerarely validated by extensive, carefully carried outecological investigations. It is not surprising that noadverse effects were noted because no extensiveinvestigations were carried out to support thisstatement. In short, it is a statement based more onabsence of information than on supporting infor-mation. It is quite likely that no dramatic events,such as a major fish kill, would be associated withwaste discharge practices based on carefully carriedout single species tests because, if the species werecarefully selected and the test conducted by exper-ienced professionals and a large application factorused, this should certainly not occur. On the otherhand, changes in the ecosystem that might reducefish population by impaired spawning rather thanlethality would be less likely to be noticed by casualobservers because no dramatic, highly visible evi-dence would be present to suggest major changeswere occurring. It would be extremely helpful ifpredictions made with single species tests were vali-dated by extensive field studies that would showwhether or not both ecosystem structure and func-tion were impaired at concentrations considered tohave no adverse biological effects based on singlespecies evidence alone.

There is also another intriguing possibility - sin-gle species tests are vastly overprotective. Ecolo-gists have made statements for years that ecosys-tems are fragile because of their extraordinarycomplexity. The intuitively reasonable argumentthat such highly complex systems may be put intodisequilibrium by disturbing any component of thesystem has been quite prevalent. The reasoning isthat such an interdependent, interlocking system isfragile because of these abundant linkages. Thiscomplexity and multitudinous first, second, third,etc. order interactions are so well accepted by ecol-ogists that any statement along these lines would beregarded as platitudinous. However, very little sub-stantive evidence exists that supports the statementthat complexity is necessarily associated with fragili-ty. Some of the most complex ecosystems known toman are periodically subjected to major naturaldisturbances that they are either able to resist or, ifdisplacement occurs, recover. In fact, Vogl (1980)and others have pointed out that some ecosystemsactually deteriorate if striking disturbances do not

>-

a-c-,

'U,

PERTURBATION-INDEPENDENT SYSTEMS

DISTURBANCE

FIRE-FLOODHURRICANE, ETC.

TIME-

Fig. 2. Disturbances in general ecosystems create vegetationalsetbacks and complete recovery is slow, whereas disturbances inperturbation-dependent ecosystems usually stimulate pulses ofgrowth which rapidly decline unless disturbed again.

occur at periodic intervals. The difference betweendisturbance-dependent and disturbance-independ-ent ecosystems is given in Fig. 2. An alternativehypothesis equally tenable is that ecosystems aretough because they are complex and that damage toone of several similar pathways may result in shunt-ing to alternative pathways of nearly comparablefunction. In this case, complexity would increaserather than decrease resistance to disturbance.Functional redundancy in ecosystems has been rec-ognized for years. There may be several predatorson a single prey species. The river continum hy-pothesis (Vannote et al. 1980) indicates that certainprocesses, such as leaf degradation, may be carriedout by different taxonomic groups in the upper andlower reaches of a stream. The end functional re-sult, namely increased availability of nutrients andenergy in the leaf, is unchanged.

In addition, the low environmental realism of thevery simple, common toxicity test (where organ-isms are tested in a container with water but with nomud, rocks, vegetation, etc.) means that transfor-mation of hazardous chemicals might occur lessrapidly than in the 'real world'. Rapid transforma-

53

tion in the latter might produce secondary productsless harmful than the original and result in de-creased ecosystem vulnerability. Similarly, varioustypes of environmental sinks for chemicals are notincorporated into most commonly used single spe-cies tests.

A final argument given by those who accept theneed for going beyond single species testing will bethat these are sufficient in instances when the esti-mated environmental concentration of the chemi-cal is so far below the estimated no adverse effectsconcentration that it would be ridiculous to gobeyond simple and inexpensive single speciesscreening tests. Kimerle (1979) has noted, however,that the actual environmental concentration might

be far higher than was estimated from simple labor-atory screening tests and that the no adverse biolog-ical effects concentration might be far lower in the'real world' than was estimated from simple screen-ing laboratory tests (Fig. 3). In short, the screen-ing tests did not accurately predict 'real world'events! In one case, the effects were vastly de-creased and in the other (the environmental con-centration) vastly increased (Fig. 3). The end resultis two concentrations that appeared comfortablyseparated in the laboratory were in the 'real world'quite close together. Of course, the errors could bein the opposite direction in both cases and end upwith two concentrations that appeared quite closetogether from laboratory evidence being quite dis-

CLEAN WATERLABORATORY

TOXICITY TESTS

Apparent BiologicalEffects Concentration

EstimatedEnvironmental Concentration

1 2 3Tiers

Screening Predictive

NATURAL WATERSOF

REAL WORLD

Tox city Incre esedSynergistic Environmental

Effects

Real Cc

U....' I..--. U----.

biologicaloncentrati

Eflectson

C

'I

00

Synergistic Chemical

Measured EnvironmentalConcentration

4

Confirmalive

CLEAN WATERLABORATORY

TOXICITY TESTS

Apparent BiologicalEffects Concentration

EstimatedEnvironmental Concentration

1 2 3Tiers

Screening Predictive

NATURAL WATERSOF

REAL WORLD

Real BiologicalEffects Concentration

ToxMitigate

city Red cedng Environmental

Effects

Measured TotalEnvironmental Concentration

I ISuspended Solids

Colloidal AndDissolved Matter

Biologically AvailableConcentration

4

Confirmative

Fig. 3. Hypothetical situation of an apparent small margin ofsafety from clean water laboratory toxicity data actually beingmuch smaller because of synergistic effects of natural waters.(From Kimerle, 1979, in Workshop on Hazard Assessment.)

Fig. 4. Hypothetical situation of an apparent large margin ofsafety from clean water laboratory toxicity data actually beingmuch greater because of mitigating effects of natural waters.(From Kimerle, 1979, in Workshop on Hazard Assessment.)

M

c0

.i;

CSrI

00

r I

I I

l1li- mmm- pillill

MEN _ ..6-

. _ _..__.... _ .._ .,| _| _

54

tant from each other in the real world (Fig. 4).These possibilities provide support for doing toxici-ty testing at more than one level of biological organ-ization even for screening purposes.

From an economic standpoint, the soundest pos-sible evidence on which to base management andregulatory decisions must be demanded. At thepresent time, sufficient evidence is not available todetermine how accurately prediction can be done oftoxicological response from one level of biologicalorganization to another, but both theoretical biol-ogy and the rapidly accumulating data base on thissubject seem to indicate that such predictions arerelatively weak.

2. In estimating the effects of chemicals on popula-tions, multispecies assemblages, communities, andecosystems, what are the limitations of laboratoryscience?

Are there different degrees of environmental real-ism possible under laboratory conditions at differ-ent levels of biological organization? In a visitingscholar address given at the Mountain Lake Biolog-ical Station, Virginia, Eugene P. Odum (1981 ) gavean address entitled 'The Limitations of LaboratoryScience' that beautifully illustrates some areas inwhich laboratory investigations, however skillfullycarried out, will not suffice. It seems intuitivelyreasonable that environmental realism is more easi-ly achieved in the laboratory at lower levels ofbiological organization (e.g., species) than at higherlevels of organization (e.g., community or ecosys-tem). The report of the National Research CouncilCommittee on Ecotoxicology (Cairns et al. 1981)takes note of the fact that situations occur wherelaboratory evidence will not be adequate and fieldtesting will be mandatory. The report even makes adistinction between field situations in which theexposure is contained and those in which it is not.This merely recognizes that both effects of scale andtime as well as degree of complexity must be consi-dered that are not amenable to laboratory study. Ifthis hypothesis is correct, then predictions of toxi-cological effects from one level of biological organ-ization to another are not scientifically justifiable.In this event, an array of toxicity tests at differentlevels of biological organization are necessary forscientifically justifiable estimates of hazard. If thehypothesis is accepted that the limitations of labor-

atory science become greater as the complexity orlevel of biological organization increases, then bothmicrocosms and field tests will become more com-mon than they are now. Of course, both these hy-potheses need careful and detailed testing so thatthe judgement of their soundness can be basedon solid evidence. In addition, evidence must beobtained on the limitations of laboratory sciencein making predictions at different levels of bio-logical organization. In other words, to what degreecan more complex biological systems be simulatedin the laboratory?

3. What should be the balance of toxicity tests atdifferent levels of biological organization in orderto make a valid estimate of hazard?

A rigid or specific balance of tests at each level ofbiological organization would not serve equallywell for all situations and all categories of chemicalsubstances. Therefore, protocols need to incorpo-rate procedures for adjusting the balance of tests atdifferent levels of biological organization as infor-mation on the level of complexity most likely to beaffected is determined. Presumably, the most evendistribution would be at the outset and become lessand less even as critical sensitive components areidentified.

The determination of balance, particularly as oneproceeds through a sequential protocol at differentlevels of organization (i.e., a sequence for each ofthe major levels), will pose some interesting prob-lems. For example, suppose that a particular singlespecies test shows that deleterious effects wouldoccur with that species but that major ecosystemfunctions would continue undisturbed. An addi-tional condition, not explicitly stated but implied inthe previous statement, could be made that alterna-tive species were available to carry out similar eco-logical functions if the species suffering the delete-rious effects represented a significant portion of thebiomass. This would still pose a problem requiringconsiderable professional judgment and analysisbecause the loss of functional redundancy (i.e., re-ducing the number of species carrying out a particu-lar function) is a deleterious ecosystem effect. Aneven more interesting decision would be forced byan effect on a transitory species that would soon belost anyway because of normal successional pro-cesses with no other effects being discernible. A

55

number of other interesting arrays of mixed testresults could be furnished that would illustrate thepoint that by adding more levels of biological or-ganization to the test system that the increasedcomplexity of the situation requires much moreprofessional judgment than has ever been neces-sary. However, this merely emphasizes the pointthat judgments were probably being made on evi-dence that was far too simplistic.

4. How should tests at different levels of biologicalorganization be sequenced?

Sequencing in a toxicity testing protocol canserve various purposes: (1) adding entirely differentinformation from that already gathered, (2) ex-panding on information shown to be critical byearlier tests, (3) validating evidence gathered in pre-vious tests. In many of the existing toxicity testingprotocols, multispecies and system level tests arecarried out only when the probability appears greatthat some deleterious effects might occur. In caseswhere the no adverse biological effects concentra-tion (based on single species tests) was very marked-ly higher than the estimated environment concen-tration of the chemical, carrying out additionaltests at higher levels of biological organization wasusually considered unnecessary.

If several levels of biological organization aretested at the outset of a toxicity testing protocol,several alternative courses of action are possible: (1)if some or all the tests at various levels in the firstpart of the sequence show a probability of adversebiological effects or there is uncertainty about thisprobability, additional tests would be carried out atall levels of biological organization in the early partof the sequence unless compelling evidence existsfor omission of one or more levels, (2) the mostcritical level of biological organization could beselected for these tests and additional tests could beconducted only in that sequence designed for thatparticular level (i.e., a sequence would be designedfor each level of biological organization), (3) thelevels of biological organization likely to give theleast useful information could be eliminated butseveral levels each with its own sequence could beretained as a means of validating the presumedrelationship identified in earlier parts of the se-quence.

Since so few tests are now routinely utilized for

hazard evaluation at levels of organization higherthan the species, providing a detailed scenario onhow they might be used is difficult. Some factorsthat would influence sequencing in the alternativesystem proposed would be the amount of informa-tion redundancy, the predictive capability within aparticular level of organization from one functionto another, and so on. Since so little informationexists now on these factors, speculation on detailsof sequencing is difficult.

5. What criteria should be used to validate labora-tory predictions in 'real world'field situations?

If predictions from one level of biological organi-zation to another are unsound, validations of pred-ications made using one level with a higher or lowerlevel of biological organization would not workwell. On the other hand, if a prediction is beingvalidated, then it does not matter what level ofbiological organization was used in the test but onlywhat level of biological organization is being pro-tected by the prediction. Thus, in the first case theaccuracy of a test in terms of the degree of environ-mental realism incorporated into the laboratorytest is being validated. In the second case, the accu-racy of the prediction based on the laboratory test isbeing validated. The prediction may be that ecosys-tem integrity will not be impaired at or below acertain concentration of a chemical. At the veryleast, putting this procedure into place would in-troduce a note of caution into the predictions beingmade, particularly where the protection of ecosys-tems is concernced. Furthermore, the basic as-sumptions underlying most present practices wouldbe more rigorously examined. Finally, the ecolo-gists would be forced to play a more active role asproblem solvers and to endorse professionallythose methods now available for making realisticecosystem measurements and predictions. If noneare immediately suitable for formal endorsementby professional ecologists, strenuous efforts wouldbe made to determine why this situation exists.

Concluding remarks

When I entered the field of pollution assessmentin 1948, the burning issue was whether biologicaltesting had any role to play in pollution assessment.

56

Students in my classes find it difficult to believe thatanyone would doubt the value of biological evi-dence, but an examination of the literature of theperiod from 1945 to 1950 will show this to be true(Patrick 1949). This was the period when the newlypublished simple batch toxicity testing method(Hart et al. 1945) was being acknowledged by acommittee of the organization now called the Wa-ter Pollution Control Federation (Doudoroff et al.1951) and was eventually incorporated as an Amer-ican Society for Testing and Materials standardmethod. Had Hart et al. (1945) or the Doudoroffcommittee (1951) called for the various toxicitytests now commonly used with continuous flowrequirements, embryo larvae tests, generationaltests, tests at different trophic or functional levels,etc., they would have been regarded as hopelesslyvisionary. This is merely a consequence of the ex-plosive and rapid development of a field that hadonly a few practioners in the late 1940s. However,the last 10 years have shown remarkable changes inboth attitudes and methodology. Almost all theadvances have occurred in single species toxicitytesting. Testing at higher levels of biological organ-ization has not kept pace with advances in singlespecies testing, and an uncharitable person mightsay that a practioner using ecological methods ofthe late 1940s and early 1950s could still get bytoday.

There is abundant evidence, however, that a pe-riod of explosive development is already beginningin the use of laboratory microcosms, as well as inthe use of artificial streams and larger simulationunits which E. P. Odum has called mesocosms.Papers are beginning to appear in the professionalliterature validating laboratory tests in natural sys-tems with a frequency that is in notable contrast tothe virtual absence of such publications only a fewyears ago.

I recognize the considerable temerity of callingattention to the need for going beyond single spe-cies testing when such tests are just now beingcommonly used and no system level tests have beenformally endorsed as standard methods. However,precise predictions of ecosystem effects will not bepossible until methodologies and capabilities notnow available have been developed. It has been saidthat looking into the future is equivalent to peeringthrough a brick wall. Nevertheless, it seems intui-tively reasonable that the following will be land-

marks in the development of toxicity tests at higherlevels of biological organization than single species:

I. The first professionally endorsed ecosystem levelmethod appearing as a standard method in oneof the presently recognized systems for doing so.

2. The first protocol in which system structure andfunction are given equal attention.

3. The first protocol in which tests at higher levelsof biological organization than single speciesplay a major role in generating the initial infor-mation on which subsequent testing is based.

4. The first formally endorsed field method for val-idating laboratory tests of any kind.

The November 1981 issue of Science announces as amatter of general interest the development of asealed microcosm which has stable characteristicsand species composition for over a year. Grantedthat this is a rather simple system, it neverthelessdisplays characteristics long sought by those whowish to carry out chronic microcosm tests undercontrolled conditions. Presumably now that thisinitial breakthrough has lead the way, additionalmethods will quickly appear as is often the casewhen a major new field begins to open.

Acknowledgements

This paper is based on invited presentations tothe Society for Environmental Toxicology andChemistry in November 1981 and to the AlbertaSociety of Professional Biologists in April 1982. Iam indebted to Darla Donald for editorial assist-ance in the publication of the manuscript and toJudy Vest for typing it. Dr. M. A. Mehlman andMobil Foundation supplied funds that made prep-aration of this manuscript possible.

References

Cairns, J. Jr., Alexander, M., Cummins, K. W., Edmondson,W. T., Goldman, C. R., Harte, J., Hartung, R., Isensee,A. R., Levins, R., McCormick, J. F., Peterle, T. J. & Zar,J. H., 1981. Testing for Effects of Chemicals on Ecosystems.Nat. Acad. Press, Wash. D.C. 103 pp.

Cairns, J. Jr., Dickson, K. L. & Maki, A. (eds.) 1978. Estimatingthe Hazard of Chemical Substances to Aquatic Life. Am.Soc. Test. Mater. Philad. Pa. 278 pp.

57

Dickson, K. L., Cairns, J. Jr. & Maki, A. (eds.) 1979. Analyzingthe Hazard Evaluation Process. Am. Fish. Soc., Wash. D.C.159 pp.

Dickson, K. L., Maki, A. W. & Cairns, J. Jr. (eds.) 1982.Modeling for Estimating the Fate of Chemicals in AquaticEnvironments. Ann Arbor Sci. Publ., Inc., Ann Arbor,Mich., 413 pp.

Doudoroff, P., Anderson, B. G., Burdick, G. E., Galtsoff, P. S.,Hart, W. B., Patrick, R., Strong, E. R., Surber, E. W. & VanHorn, W. M., 1951. Bioassay for the evaluation of acutetoxicity of industrial wastes to fish. Sewage ind. Wastes 23:1380-1397.

Hart, W. B., Doudoroff, P. & Greenback, J., 1945. The Evolu-tion of the Toxicity of Industrial Wastes, Chemicals andOther Substances to Freshwater Fishes. Waste Cont. Lab.,Atlantic Refining Co., Philad. Pa., 349 pp.

Kimerle, R. A., 1979. Aquatic hazard assessment- concepts andapplication. In: Bratzel, M. P., Jr. (ed.) Workshop on Ha-zard Assessment. Great Lakes Water Quality Board of theInternational Joint Commission, Windsor, Ontario, Cana-da: 221-230.

MacArthur, R., as cited in Cody, M. L. & Diamond, J. M. (eds.)1975. Ecology and Evolution of Communities. BelknapPress of Harvard Univ. Press, Cambridge, Mass.

Maki, A. W., Dickson, K. L. & Cairns, J. Jr. (eds.) 1980. Bio-transformation and Fate of Chemicals in the Aquatic Envir-onment. Am. Soc. Microbiol., Wash. D.C. 150 pp.

Odum, E. P., 1981. Revival of non-laboratory science. As. S.-east. Biol. Bull. 28: 119-121.

Patrick, R., 1949. A proposed biological measure of streamconditions, based on a survey of the Conestoga Basin, Lan-caster County, Pennsylvania. Proc. Acad. nat. Sci. Phila.101: 277-341.

Vannote, R. L., Minshall, G. W., Cummins, K. W., Sedell, J. R.& Cushing, C. E., 1980. The river continuum concept. Can.J. Fish. aquat. Sci. 37: 130-137.

Vogl, R. J., 1980. The ecological factors that produce perturba-tion-dependent ecosystems. In: Cairns, Jr. Jr. (ed.). TheRecovery Process in Damaged Ecosystems. Ann Arbor Sci.Publ., Inc., Ann Arbor, Mich.: 63-94.