Ecotoxicology I, 3-16 (1992)

The threshold problem in ecotoxicology J O H N C A I R N S , JR Department of Biology and University Center for Environmental and Hazardous Materials Studies, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061-0415, USA

Received 30 January 1992; accepted 31 March 1992

The most commonly used threshold in environmental toxicology is the LCs0 (or modifications thereof) where 50% of the organisms die or are otherwise affected at a certain concentration of a chemical for a particular time of exposure under specified environmental conditions. Most commonly, this particular threshold is derived from single species laboratory tests low in environmental realism. If the field of ecotoxicology truly examines the effects of chemicals on ecosystems (i.e., complex multivariate systems), serious consideration must be given to thresholds other than those now commonly used in the field of environmental toxicology. Attributes at the community and ecosystem level of organization are not demonstrated at lower levels of biological organization, for example, energy flow and nutrient spiralling. Key issues are whether extrapolation is possible from one threshold to another within a level of biological organization and from one level of biological organization to another for thresholds that do not exist at many levels. Thresholds may be artefacts of testing procedures and may not exist in natural systems. Nevertheless, society must make management decisions about risk with availabe methods, including those designed to identify some point or threshold below which no deleterious effects are observed. However, these methods and their assumptions deserve more explicit and systematic examination than they have received thus far. Keywords: ecotoxicotogical thresholds; ecotoxicological break points; levels of biological organization; environmental toxicology; thresholds in pollution assessments.

The significantproblems we face cannot be solved at the same level of thinking we were at when we created them. Albert Einstein

Do ecological thresholds exist? A threshold is defined in Webster's Third International Dictionary as 'the point at which a physiological or psychological effect begins to be produced.' Finding this point might initially seem like a simple scientific problem, solved by finding a well-defined rate change in a d o s e - r e s p o n s e curve. Several types of curves are repeatedly found in studies of responses to stress (Fig. 1), Figure l a shows a distinct change in the rate of response that constitutes a threshold. An example of this pattern of response can be seen in the decline in bacterial numbers of microbial communities exposed to increasing levels of ammonia (Niederlehner and Cairns, 1990). In Fig. lb, stimulation at low doses may serve as a practical threshold (Odum et al., 1979). An example of this pattern of response can be found in the rate of population growth of aquatic oligochaetes exposed to cadmium (Niederlehner et al., 1984). Figure lc shows an asymptotic rate of change in response (Marcovich and Devoret, 1960) in which there is no absolute threshold, but the 0960-9292 © 1992 Chapman & Hall

4

Cairns A

B

C

D

Z 0 oo a~

DOSE

Fig. 1. Types of dose-response curves. Curve a shows an abrupt change in response with dose, i.e. a threshold. Curve b shows a subsidy at low doses that may serve as a practical threshold. Curve c is asymptotic with a practical threshold. Curve d shows no threshold. magnitude of increases in impairment become so small as to be biologically unimportant. In this case, there may be a practical threshold in the absence of an absolute threshold. Finally, Fig. ld shows a response that changes linearly from background to severe stress levels; there is no threshold. This is the typical pattern in acute lethality tests modelled as probit percent survival versus log concentration (Finney, 1971). It is also the pattern described by Woodwell (1974) for resistance to radiation effects on mutants in Tradescantia. However, the apparent simplicity in framing the problem of thresholds is misleading. As stewardship of environmental resources has improved, focus has shifted from gross and obvious environmental damage to the possible consequences of more subtle and rare impairments. With this shift to prediction of rare or subtle events at low stress levels, the ability to model accurately is exhausted. It is exceedingly difficult to extrapolate from nonoccurrence of zLrare event in a very restricted sample to nonoccurrence in a much larger, more complex, more variable and longer lived system. Woodwell (1974) asked the question "is it reasonable to assume that thresholds for effects of disturbance exist in natural ecosystems or are all disturbances effective, cumulative, and detrimental to the normal functioning of natural ecosystems?". Marcovich and Devoret (1960) take the position that the 'threshold concept' is theoretically absurd, but deviations exist in the low ranges of doses that may be considered in many cases as practical thresholds (Fig. lc). Campbell (1981), in a critique of articles I wrote on assimilative capacity (Cairns, 1977a), makes the point that any introduction or alteration of a natural system changes it. Poets and naturalists have made claims, such as "Thou canst not pluck a daisy without the troubling of a star" or "One can never walk through the same stream twice". But, as I have noted previously (Cairns, 1981), ecosystems are dynamic and change is normal. Changes can be neutral, subsidies or stresses (Odum et al., 1979). Furthermore, Odum et al. (1979) noted that variability is normal and described the normal range of variability as the nominative state. Although Odum et at. (1979) did not dwell on thresholds, they clearly acknowledged their existence by indicating concentrations within which normal variability occurred and

The threshold problem in ecotoxicology

5

thresholds at which the variability of response exceeded this normal variability either upwards or downwards. As Woodwell (1974) noted, "The threshold concept takes on a special significance in that it is permissive; it allows some degree of pollution". The assumption that some level of anthropogenic activities may not cause disequilibrium or unfavorable trends in natural systems is at the core of the assimilative capacity hypothesis (Cairns, 1977a,b). This is also implicit in the Odum et al. (1979) manuscript on stresses and subsidies, because the assumption is that some substances may produce subsidies at certain concentrations and stresses at others. This is also the basis for much of the regulation of toxic substances in the United States. Since the purpose of most environmental toxicology or ecotoxicological tests is to predict the response of natural systems using tests in laboratories or surrogates of natural systems, ecotoxicologists must determine the best measurement end points in their quests for thresholds that mark important changes in ecosystem character but which transcend normal variation resulting from seasons, succession and other natural dynamics. In short, if we accept the premise that all ecosystems are dynamic and variable, the quest is for thresholds or attributes that signal deviations from Odum's nominative state. Unfortunately, although ecologists use the word ecosystem frequently and produce beautiful, theoretical models in scholarly journals, they have considerable difficulty establishing system boundaries in the natural environment. This may be because, while ecologists discuss ecosystems, most of them study populations. For example, Harte et al. (1992) surveyed relatively recent journal literature in ecology indicating a dominance in ecological research for ignoring the influence of climate, soil, water and air on species interactions. All 285 articles in four consecutive issues during 1987-88 of the journals Ecology, American Naturalist, Oecologia, and Conservation Biology were surveyed. Even in the 22% of the articles that mentioned at least one of these factors (climate, chemical/physical properties of soil, water and atmosphere), the reference was made as part of a general site description and played no role in the interpretation of results. In none of the papers surveyed were the interactions among these factors even mentioned. In a now classic review article, May (1977) noted a large body of empirical observations showing that many natural communities have a multiplicity of stable states. May believed that real ecosystems possess multiple stable states, as do plausible mathematical models. May (1977) also noted that it is unfortunate that "the complications inherent in multi-species systems almost invariably preclude any quantitative confrontation between theory and data. For multi-species communities, the empirical observations remain largely anecdotal and the theory remains largely metaphorical." If many alternative stable states exist, pollution and other accidents can be very important in shaping the ecosystem. For example, microbial communities developed under cadmium stress all attain a stable number of species, but this number is different for different cadmium levels (Fig. 2). Supporting theory for ecotoxicology is much weaker than it should be, and there is small comfort in the observations of Peters (1991) that indicate comparable difficulties in the field of ecology. Stop the world while we finish our research

Society and its representatives must make decisions now on various risks with whatever methodologies are available. This is similar to what George Bernard Shaw is reported to have responded when asked what he thought of life, "It is preferable to the alternative".

6

Cairns 40-

Control

$

*

*

20-

,0°:

. . . . . . . . . . . . . . . . . .

0,4

I

. . . . . . . .

'V'l

ug/L

. . . . . . . . .

i

*

. . . . .

'"v' •

'

,

•

I

-

. . . . . . . . .

-

*

20-

040-

. . . . . . . . .

i

,' ~""

. . . . . .

i

. . . . . . . .

~1

1.4 uglL

. . . . . . . . .

i

~,

. . . .

,,,,,,,

,

,

,

i

. . . . . . . . .

,

20,...,

E Z

0-

~40-

'5

2.7 ug/L

O.

0') 202

O.

40-

. . . . . . . . .

i

. . . . . . . . .

i

. . . . . . . . .

i

. . . . . . . . .

i

. . . . . . .

'""

i

. . . . . . . . .

5.6 ug/L 20,

J

4.

£1.

40-

. . . . . . . . .

i

. . . . . . . . .

i

. . . . . . .

" ' !

. . . . . . . . .

i

. . . .

, , - r - ~ ,

[

. . . . . . . . .

9.5 ug/L 20.

.!...!...,,..I I

3

..... 7

, .......

1.,,, 14

.......

,,I..,,,.,. 21

.....

I... 28

DQy

Fig. 2. Multiple equilibrium species numbers with different cadmium stress levels (from Niedertehner et al., 1985). To take an extreme view, suppose that every human activity causes some deleterious effects on natural systems. Would the consensus be that all activities stop? This is unlikely without robust supporting evidence for this view. The most likely social choice would be that some activities are more environmentally deleterious than others and, since the world's present population, size and quality of life depends on technology, some balance should be chosen that would permit a sustained use of the natural systems

The threshold problem in ecotoxicology

7

for an indefinite period of time. Attributes that appear to have thresholds, whether they were artefacts or not, would still have utility in the decision-making process. This practical approach need not suppress the theoretical arguments and controversies that should continue unabated. In addition, whenever a new theoretical advance has reached the point where it can be transferred to the applied field, it should have its impact. Ignoring relevant theory should not be condoned by regulatory or other organizations, but no action because of a lack of underlying theory, when the information available is better than none at all, should not be condoned either. False positives, false negatives and the burden of proof Tullock and Brady (1992) call attention to two primary economic costs of false environmental alarms: (1) the community mobilizes to protect itself from a danger that does not exist and, thus, is diverted from other more economically beneficial activities and (2) crying wolf a great many times when there is no wolf leads to the cry being ignored when a wolf is finally present, resulting in environmental damage. In the field of ecotoxicology, predictive models will never be robust enough to eliminate either false negatives (indications of safety when in fact a dangerous situation exists) or false positives (indications of danger when none in fact exists). But, as the field of ecotoxicology matures, the number of false negatives and false positives should decline appreciably, especially if more attention is given to confirmation or validation of the predictive models (e.g. Cairns et al., 1988). Tolerance of uncertainty: an illustrative example Uncertainty is at the core of the scientific process - it is initially high when a hypothesis is first proposed, but it is never eliminated entirely even when the hypothesis is generally thought to be quite sound. So, although the term scientific truth may be used in the news media, it is rarely used by practicing scientists. Instead, science works by developing hypotheses and then falsifying or supporting them. If, after a certain number of years, a body of evidence accumulates that seems to support a hypothesis, it is tentatively accepted and sometimes even called a fact or a law. However, if practicing scientists are pressed, they will almost invariably affirm that all hypotheses are tentative and even those that have the support of a large body of evidence may ultimately be shown to be unsound. There are differing interpretations of proper management response to uncertainty. Molina and Rowland (1974) found that safe, nonflammable refrigerants, chlorofluorocarbons (CFCs), were almost inert in the troposphere (that mixed portion of the atmosphere roughly below 50,000 feet). However, they slowly migrate upward until they reach the mid stratosphere (about 100,000 feet) where they are broken down by shortwave, ultraviolet radiation. When this occurs, the CFCs release atomic chlorine, which then reacts with ozone, converting it back to the common form of oxygen (02). During this process, the chlorine itself is not incorporated into inactive chlorine compounds but appears again as a reaction product and, thus, is available to react with other ozone molecules. Thus, it functions as a catalyst. Fairly robust evidence is now available for both the events just described and for potential damage to humans and other organisms from increased radiation that results from the thinning of the ozone layer (readers

8

Cairns

wishing a summary of this might find Sharon Rowan's Ozone Cris&: The Fifteen-Year Evolution of a Sudden Global Emergency, published by Wiley, New York, in 1989, useful). An-even more recent work is Anderson et al. (199t), although undoubtedly much, much more will come! However, effects are still disputed (e.g. Roberts, 1989). Such disputes are normal to the scientific process, but, in this instance, much more is at stake than the reputations of the scientists. Uncertainty about the magnitude of changes and the magnitude of their probable effects on ecosystem integrity and human health is interpreted by some as warranting delays in management action. Others feel that the potential for widespread, gross and not easily corrected harm makes management response mandatory at current levels of uncertainty. Uncertainty will never be eliminated. However, the degree of uncertainty that will be tolerated depends on such factors as the seriousness of the consequences, the reversibility of impact, the uniqueness of the resources at risk and who benefits and how much versus who loses and how much.

Thresholds for single species For many years, there was an intuitively reasonable but practically unapplicable belief on the part of the regulatory agencies in the United States that if one could find the most sensitive species and designate concentrations that would protect it, all other species would be protected as well. Cairns (1986) challenged the most sensitive species approach, and, subsequently, a rather large compendia of information confirmed that there is no single, most sensitive species (e.g. Mayer and Ellersieck, 1986). The most important fundamental flaw in the most sensitive species approach is that a species that is most sensitive to a particular chemical compound is not likely to be the most sensitive species to all chemical compounds. Although the belief in the most sensitive species still persists, the widespread belief was probably ended by a splendid editorial by Mount (1987), suggesting the abandonment of the search for the most sensitive species and concentration on the predictive value of toxicity tests. Even if the most sensitive species were a widely accepted hypothesis, there would still be a threshold problem. Obviously, lethality is not an acceptable final threshold, since scientists want organisms to do more than merely survive. Organisms should not suffer reproductive impairment, arrested growth, behavioral aberrations, tumors, impaired resistance to disease and the like. Additionally, all life-history stages should be protected, and abundant evidence shows that sensitivity to a particular compound may change from one life-history stage to another. Regrettably, a particular life-history stage is not invariably the most sensitive stage to all chemical compounds. Also, one compound may affect behavior while another affects visual acuity. Therefore, absence of an effect in one attribute does not mean protection of all attributes. An array of information, ideally including tests on a variety of relevant responses at different levels of biological organization, is needed to predict likely biological effects. In addition, predictions must be validated in natural systems. Although the term validation or confirmation has appeared in the literature with great regularity in the last few years, there have been various approaches. There are four illustrative associations of explicit predictions and validation criteria (Cairns, 1988). 1. Only the test species is expected to be protected fully in natural systems, and validation means confirming this assumption in the field. 2. Other species than the one(s) actually tested are thought to be protected at the no-

The thresholdproblem in ecotoxicology

9

observable-effects concentration, and a list of these species inhabiting the natural systems in question is included. Validation is carried out with field observation on all of these species. 3. No adverse effects at the community level of biological organization will occur, and the characteristics used to validate this assumption are listed. Validation may be carried out in microcosms or mesocosms (e.g. Odum, 1984) under certain circumstances but should be based on field observations whenever possible. 4. No adverse effects will occur at the ecosystem level of organization, and the characteristics at the ecosystem level used to validate this assumption are listed. Validation in field enclosures or in natural systems over relevant time and spatial scales should probably be mandatory. By coupling the explicit predictions being made with the explicit end points being used to validate these predictions, a more systematic and orderly process of validation will result. Until more hypothesis testing is carried out in the field of hazard evaluation than is in place presently, it is unlikely that the field will get the confidence it desires.

Signal-to-noise ratio Another way of looking at both the threshold and the validation problem is the signal-tonoise ratio. Figure 3a illustrates a situation where a signal is measured (for example, respiratory frequency of a fish or species richness of a microbial community), showing that the mean response appears to have changed by a small amount directly after the stress stimulus. This change in response is the signal of environmental effects, but the amplitude or variation or noise is so great that the change in response would be questionable. Figure 3b shows an idealized state where the change clearly exceeds the normal variability. In Fig. 3a, the signal appears to be well within the noise range, whereas in Fig. 3b the signal, in response to a stress, is clearly outside the normal range of variability. The change could be up as well as down, the point being that the signal is well outside the noise. Whether the change in Fig. 3b is called a threshold might be considered strictly a theoretical, or even a semantic, problem since unquestionably the system has responded. If the attribute is species richness, a dramatic decline would be regarded by most ecologists as unfavorable even if the degree to which species richness can decline without seriously affecting the biological integrity of the system cannot be determined.

Multispecies and community level toxicity testing The most important question in ecotoxicology is how well the test systems predict responses in complex, multivariate, natural systems. Single species tests are the best means of measuring such attributes as growth, reproductive success and the like, but there is little evidence that they are good predictors of events in natural systems at higher levels of biological organization (e.g. Cairns, 1983a). Many of the attributes found in multispecies or community level tests, such as species richness or nutrient spiralling, are not attributes displayed by single species tests. Furthermore, validation of multispecies predictions in natural systems is easier than predictions from surrogate single species end points since similar attributes can be used in both the laboratory and the field. This is usually not possible when single species toxicity tests are used, particularly those

10

Cairns

A STRESS

B

STRESS TIME

Fig. 3. Differences in signal-to-noise ratios of end points in biomonitoring. Part a illustrates a low signal-to-noise ratio. The signal, i.e. the change in response after stress, is small but the response varies widely over time, so there is a large amount of noise. Part b illustrates a high signal-to-noise ratio. The change in response after stress is large and the response is less variable over time. In this case, it is easy to see the impact of the stress, organisms commonly used in standard US Environmental Protection Agency tests since these species are not indigenous to many of the natural systems into which toxic substances are discharged. Thus, multispecies toxicity tests can provide predictions of toxicant effects that can be extrapolated to natural ecosystems with less uncertainty. A strong ecological argument can be made for testing at higher levels of biological organization (e.g. Cairns, 1983b). It is generally accepted that higher levels of biological organization (e.g. communities and ecosystems) possess properties that are not present at lower levels (e.g. populations) (McMahon et al., 1978; Webster, 1979). Despite a convincing theoretical argument for the use of multispecies tests, several scientific and regulatory concerns remain. Criticisms of multispecies tests (e.g. Giesy, 1985; Loewengart and Maki, 1985; Mount, 1985) generally focus on the following six points. 1. Are multispecies tests more sensitive to toxic compounds than single species tests? The central question is not sensitivity, but predictive accuracy, as Mount (1987) points out. 2. Can data obtained from laboratory multispecies test systems be extrapolated to natural ecosystems with more certainty than data from single species?

The threshold problem in ecotoxicology

11

Simply stated, if the dynamics of laboratory multispecies systems do not mimic those of natural systems, then their predictive capability is suspect (e.g., Heath, 1980). The theoretical argument here, of course, includes the familiar scale effect. It seems intuitively reasonable, however, that the further apart two systems are (e.g. single species toxicity tests and complex, multivariate natural systems) the greater the scale effect, and the closer they are together (e.g. communities versus natural ecosystems) the less distortion will result from differences in scale. On an empirical basis, there is evidence that the correspondence between microcosm tests involving indigenous communities to a chemical compound and the response of similar communities in natural systems is fairly close (e.g. Tolle et al., 1985; Niederlehner et al., 1990). 3. How replicable and reproducible are results gathered from multispecies toxicity testing systems? It is a sine qua non that increased variability in response reduces confidence associated with statistical estimations (e.g. a lowest-observable-effect concentration or LOEC). As a consequence, an environmentally realistic test system with great complexity and low replicability may provide predictions that are not as powerful as those of single species tests that exhibit increased replicability but are less realistic. Restated, whereas single species toxicity tests may be less accurate than multispecies toxicity tests in predicting responses at higher levels of biological organization, they may be more precise. Although these two problems are quite different in some regards, from an ecological and statistical standpoint, they pose similar problems in attempts to estimate hazard. Again, although the underlying theory is not yet robust, circumstantial evidence indicates that when a large array of species gathered on an artificial substrate in an undisturbed ecosystem is tested in the laboratory against response to a toxicant, the variability is remarkably small. This is true even though the species composition of the community accrued on the artificial substrate may never twice be identical. The hypothesis explaining this is that a large array of species will be distributed with regard to sensitivity along a normal curve, and, if sufficient species richness is present, the curves will appear virtually identical even though the species composition is not (e.g. Cairns et al., 1991). 4. Will the magnitude of natural variation evident in multispecies systems (including natural ecosystems) make it impossible to detect stress effects? This is the basis of Loewengart and Maki's (1985) doubts of the value of multispecies laboratory tests on the basis of well-documented variability in natural communities and ecosystems. As Odum et al. (1979) noted, variability is one of the attributes of undisturbed, natural ecosystems. Therefore, their variability is not itself a valid criticism of multispecies tests unless the variability of these tests substantially exceeds that of natural systems. If one assumes that the goal of any toxicity testing exercise is to determine a no-observable-adversebiological-effects concentration for natural ecosystems, a consideration of natural variability should be a critical portion of the hazard evaluation process. In short, while intraand inter-ecosystem variability may not fit comfortably into the current regulatory framework, these attributes may be essential if the goal is to increase the predictive power of hazard assessment procedures. 5. Is it possible to develop suitable end points for multispecies and community level tests? Some single species toxicity tests provide what appear to be crisp thresholds. But, there is some evidence that these thresholds may be artefacts of sample size rather than real thresholds. Resistance in a population to a particular chemical substance frequently follows a normal curve; as such, median effect levels are much easier to determine than

12

Cairns

rare effects. An apparent threshold or leveling off in response rate may actually be the exhaustion of statistical power for a given sample size. With more complex systems, all of the responses of component parts are represented and integrated into the emergent end point. More of the total response gradient at all hierarchial levels is exposed immediately. A single threshold of the type generated by single species toxicity tests appears more difficult to obtain. Therefore, people who favor single species toxicity tests tend to state that the end points for single species tests are more decisive than those of multispecies tests when, in fact, they just represent a more limited spectrmn of a broad response range. A more extended discussion of this complex issue may be found in Cairns and McCormick (1992). Despite the fact that obtaining thresholds is more difficult or perhaps impossible in complex systems, a number of end points are eminently suitable for hazard evaluation and ecological risk analysis. While ecologists might disagree over whether species richness or trophic complexity is a better end point for predicting effects, most would agree that changes on either one would be a more justifiable indicator of community impairment than the rate of mortality of individuals of a few species, some or all of which may not be found in the particular system under consideration. Although more research is definitely needed to assess the reliability of community and ecosystem end points, evidence is persuasive that several end points yield reliable results for predicting noadverse-biological-effects concentrations (see review in Cairns and Niederlehner, 1987). It has been assumed erroneously that, because a widely touted advantage of multispecies tests is the incorporation of species interactions, measurements of these interactions must be used as end points for these tests (e.g. Mount, 1985). While the direct measurement of stress effects on interactions among species is sometimes useful, what is more important is that these emergent properties are functioning in the test system, so that the effects on more easily quantifiable end points, such as species richness and evenness, are more comparable with those routinely used to assess ecological health in natural systems! Indeed, the use of multispecies tests to measure effects on populations, although criticized (Loewengart and Maki, 1985; Mount, 1985), does provide an assessment under more realistic conditions than do single species tests, assuming that multispecies test systems mimic the natural environment of the populations of interest. 6. Are tests at higher levels of biological organization (multispecies, community, ecosystem) cost effective? There is more than one way to measure cost effectiveness[ The most commonly used way is to consider only the cost of the test itself without considering the cost of inadequate or inappropriate information. This is made easier for industries wishing to carry out only a few tests because they can do what the law requires and then use the fact that they complied with the law as a defense if the information is inadequate or inappropriate. There are notable examples in the professional literature where standard single species protocols resulted in gross underprotection of the environment (e.g. Kimball and Levin, 1985). However, cost effectiveness can be seriously diminished by overprotection as well as underprotecfion. Carlson et al. (1986) have estimated that single species protocols may overestimate environmental risk by an order of magnitude. Because of a general lack of attention to error control (monitoring) in hazard assessment protocols, the actual predictive success of single species tests is hard to assess (Cairns, 1983a). Regrettably, there is even less information with which to judge the success of multispecies tests in instances where single species tests failed. This would be an important criterion for including multispecies tests in the regulatory process (Mount,

The threshold problem in ecotoxicology

13

1985). If the cost of inappropriate decisions is considered in the broader view, the most important criteria for choosing a test system would be (1) the degree to which predictions can be extrapolated directly to natural systems with confidence and (2) the replicability and reproducibility of the test system, which will affect the degree of statistical uncertainty surrounding predictions. Of course, other concerns are related to cost effectiveness, but these two seem to be the most important at the moment. All decisions have some economic component. However, if ecotoxicology is to progress and utilize new thresholds, their cost effectiveness must be evaluated in a broader context than the simple cost of the test itself! Even in the limited context, multispecies tests may well prove cost effective as their development improves, because they are capable of furnishing information not available in single species tests, some of which (e.g. environmental partitioning) is essential to the hazard-assessment process.

Ecotoxicology at the landscape level The field of ecology has gradually realized that both the temporal and spatial scales of most studies on thresholds in ecotoxicology are inadequate. While population ecology still remains the dominant area of interest, ecosystem level and now landscape level ecology are receiving increased attention. The field of landscape ecology focuses on: (1) landscape structure, that is, the spatial arrangement of ecosystems within landscapes; (2) landscape function, that is, the interaction among the component ecosystems through flow of energy, materials and organisms; and (3) alterations of this structure and function through natural or anthropogenic stress or natural successional processes, etc. (Forman and Godron, 1986; Risser, 1987). The development of the field of landscape ecology has incorporated the conceptual base of hierarchy theory, i.e. ecosystem processes occur within a hierarchy of different spatial and temporal scales (Allen and Starr, 1982; O'Neill et al., 1986). O'NeiU et al. (1986) make the point that ecological systems are constrained both by the range of potential behavior of lower scales and the environmental limits of higher scales. Forman and Godron (1986) have hypothesized that measures of ecological structure and function must be taken at a scale appropriate to the process being observed. If ecotoxicology is to combine the fields of ecology and toxicology, then ultimately some recognition must be given to landscape components.

Concluding statement Peter Drucker said: "Management is doing things right; leadership is doing the right things". When toxicologists added the prefix eco to the field of toxicology, so that the word became ecotoxicology, they continued primarily to make the same measurements they had made before the name was changed. Universities have followed the same pattern by adding environmental to a variety of traditional disciplines without any substantive change in the courses taught, the faculty employed and the like. Behind both facades, there is little substance. 'Both devices are so transparent that only the most gullible will be fooled. Academic institutions must learn that simply adding a few words to an administrative structure does not constitute reform, and toxicologists must learn that creating the word ecotoxicology implies a paradigm shift of considerable magnitude. We have perfected single species tests to a high degree over the last 40 years while neglecting validations of predictions based on single species tests in natural systems,

14

Cairns

incorporation of much more environmental realism in laboratory tests, discussions of extrapolation from a single species theshold to an ecosystem response and a whole variety of other factors discussed here. A diligent search through the current literature revealed practically no attention to ecological thresholds in purportedly ecotoxicotogical manuscripts. Perhaps this manuscript will stimulate some discussion. Ecotoxicologists, having assumed the label, must now demonstrate leadership as well as tactics. This development is long overdue.

Acknowledgements I am much indebted to my colleagues Bruce Wallace (for calling to my attention the Marcovich and Devoret paper) and Barbara R. Niederlehner for reading an early draft of this manuscript. Teresa Moody transcribed the dictation of the first draft of this manuscript and made adjustments on the word processor for successive drafts. Darla Donald prepared the manuscript for publication according to the specific directions furnished by the publisher.

References Allen, T.F,H. and Starr, T,B. (1982) Hierarchy: Perspectives in Ecological Complexity. Chicago: University of Chicago Press. Anderson, J., Toohey, D. and Brune, W. (1991) Free radicals within the Antarctic vortex: The role of CFCs in Antarctic ozone loss. Science 251, 39-45. Cairns, J., Jr. (1977a) Aquatic ecosystem assimilative capacity. Fisheries 2(2), 5-7, 24. Cairns, J., Jr. (1977b) Quantification of biological integrity. In Integrity of Water (Ballentine, R.K. and Guarria, L.J., eds) pp. 171-87. Washington, DC: US Environmental Protection Agency, Office of Water and Hazardous Materials, 055-001-010680-1. Cairns, J., Jr. (1981) Of: A critique of assimilative capacity by I.C. Campbell. J. Water Pollut. Control Fed. 53(11), 1653-5. Cairns, J., Jr. (1983a) Are single species tests alone adequate for estimating hazard? Hydrobiologia 100, 47-57. Cairns, J., Jr. (1983b) The case for simultaneous toxicity testing at different levels of biological organization. In Aquatic Toxicology and Hazard Assessment: Sixth Symposium (Bishop, W.E., CardweU, R.D. and Heidolph, B.B., eds) pp. 111-27. Philadelphia: American Society for Testing and Materials, STP802. Cairns, J., Jr. (1986) The myth of the most sensitive species. BioScience 36(10), 670-2. Cairns, J., Jr. (1988) What constitutes field validation of predictions based on laboratory evidence? In Aquatic Toxicology and Hazard Assessment: Tenth Volume (Adams, W.J., Chapman, G.A. and Landis, W.G., eds) pp. 361-8. Philadelphia: American Society for Testing and Materials, 8TP971. Cairns, J., Jr. and Niederlehner, B.R. (1987) Problems associated with selecting the most sensitive species for toxicity testing. Hydrobiologia 153, 87-94. Cairns, J., Jr. and McCormick, P.V. (1992) Developing an ecosystem-based capability for ecological risk assessments. Environ. Professional, in press. Cairns, J., Jr., Smith, E.P. and Orvos, D. (1988) The problem of validating simulation of hazardous exposure in natural systems. In Proceedings of the 1988 Summer Computer Simulation Conference (Barnett, C.C. and Holmes, W.M., eds) pp. 448-54. San Diego: The Society for Computer Simulation International.

The threshold problem in ecotoxicology

15

Cairns, J., Jr., McCormick, P.V. and Niederlehner, B.R. (1992) Estimating ecotoxicological risk and impact using indigenous aquatic microbial communities. Hydrobiologia 229, 1-15. Campbell, I.C. (1981) A critique of assimilative capacity. J. Water Pollut. Control Fed. 53(5), 604-7. Carlson, A.R., Nelson, H. and Hammermeister, D. (1986) Development and validation of sitespecific water quality criteria for copper. Environ. Toxicol. Chem. 5, 997-1012. Finney, D.J. (1971) Probit Analysis. London: Cambridge University Press. Forman, R.T.T. and Godron, M. (1986) Landscape Ecology. New York: Wiley. Giesy, J.P. (1985) Multispecies tests: Research needs to assess the effects of chemicals on aquatic life. In Aquatic Toxicology and Hazard Assessment (Bahner, R.C. and Hansen, D.J., eds) pp. 67-77. Philadelphia: American Society for Testing and Materials, STP891. Harte, J., Torn, M. and Jensen, D. (1992) Consequences of indirect linkages between climate change and biological diversity. In Global Warming and Biological diversity (Peters, R. and Lovejoy, T.E., eds) pp. 325-43. New Haven: Yale University Press. Heath, R.T. (1980) Are microcosms useful for ecosystem analysis? In Microcosms in Ecological Research (Giesy, J.P., ed.), pp. 33-347. Springfield, VA: Technical Information Center. Kimball, K.D. and Levin, S.A. (1985) Limitations of laboratory bioassays: The need for ecosystem level testing. Bioscience 35, 165-71. Loewengart, G. and Maki, A.W. (1985) Multispecies toxicity tests in the safety assessment of chemicals: Necessity or curiosity. In Multispecies Toxicity Testing (Cairns, J., Jr., ed.), pp. 112. New York: Pergamon Press. Marcovich, H. and Devoret, R. (1960) The effect of small doses of ionizing radiations on Escherichia coli. In Immediate and Low Level Effects of Ionizing Radiations (BuzzatiTraverso, A.A., ed.), pp. 293-303. London: Taylor & Francis. May, R.M. (1977) Thresholds and breakpoints in ecosystems with a multiplicity of stable states. Nature 269, 471-7. Mayer, F.L., Jr. and Ellersieck, M.R. (1986) Manual of Acute Toxicity: Interpretation and Data Base for 410 Chemicals and 66 Species of Freshwater Animals. Washington, DC: Fish and Wildlife Service, US Department of the Interior, Resource Publ. 160. McMahon, J.A., Phillips, D.L., Robinson, J.V. and Schimpf, D.J. (1978) Levels of biological organization: An organisms-centered approach. BioScience 28, 700-4. Molina, M. and Rowland, F.S. (1974) Stratospheric sink for chlorofluromethanes: Chlorine atom catalyzed destruction of ozone. Nature 249, 810-4. Mount, D.L. (1985) Scientific problems in using multispecies toxicity tests for regulatory purposes. In Multispecies Toxicity Testing (Cairns, J., Jr., ed.), pp. 13-18. New York: Pergamon Press. Mount, D.I. (1987) Editorial: A changing profession. Environ. Toxicol. Chem. 6, 83. Niederlehner, B.R. and Cairns, J., Jr. (1990) Effects of ammonia on periphytic communities. Environ. Pollut. 66, 207-21. Niederlehner, B.R., Buikema, A.L., Jr., Pittinger, C.A. and Cairns, J., Jr. (1984) Effects of cadmium on the population growth of a benthic invertebrate Aeolosoma headleyi (Oligochaeta). Environ. Toxicol. Chem. 3: 255-62. Niederlehner, B.R., Pratt, J.R., Buikema, A.L., Jr. and Cairns, J., Jr. (1985) Laboratory tests evaluating the effect of cadmium on freshwater protozoan communities. Environ. Toxicol. Chem. 4, 155-5. Niederlehner, B.R., Pontasch, K.W., Pratt, J.R. and Cairns, J., Jr. (1990) Field evaluation of predictions of environmental effects from a multispecies-microcosm toxicity test. Arch. Environ. Contamination Toxicol. 19(1), 62-71. Odum, E.P. (1984) The mesocosm. BioScience 34(9), 558-62. Odum, E.P., Finn, J.T. and Franz, E.H. (t979) Perturbation theory and the subsidy-stress gradient. BioScience 29(6), 349-52.

16

Cairns

O'Neill, R.V., De Angelis, D.K., Waide, J.B. and Allen, T.F.H. (1986) A Hierarchical Concept of Ecosystems. Princeton: Princeton University Press. Peters, R.H. (1991) A Critique for Ecology. New York: Cambridge University Press. Risser, P.G. (1987) Landscape ecology: State of the art. In Landscape Heterogeneity and Disturbance (Turner, M.G., ed.), pp. 1-14. New York: Springer-Verlag. Roberts, L. (1989) Does the ozone hole threaten Antarctic life? Science 244, 288-9. Tolle, D.A., Arthur, M.F., Chesson, J. and Van Voris, P. (1985) Comparison of pots versus microcosms for predicting acroecosystem effects due to waste amendment. Environ. Toxicol. Chem. 4, 501-9. Tullock, G. and Brady, G. (1992) The risk of crying wolf. In Predicting Ecosystem Risk (Cairns, J., Jr., Niederlehner, B.R. and Orvos, D.R., eds)pp. 77-91. Princeton: Princeton Scientific Publishing. Webster, J.R. (1979) Hierarchical organization of ecosystems. In Theoretical Systems Ecology (Halfon, E., ed.), pp. 119-29. New York: Academic Press. Woodwell, G.M. (1974) The threshold problem in ecosystems. In Ecosystem Analysis and Prediction (Lcvin, S.A., ed.), pp. 9-21. Alta, UT: SIAM Institute for Mathematics and Society.