ECOTOXICOLOGY: A Comprehensive Treatment - Chapter 36 (end) ppt

Clements: “3357_c036” — 2007/11/9 — 18:40 — page 811 — #1
Part VI
Ecotoxicology: A Comprehensive
Treatment—Conclusion
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 813 — #3
36
Conclusion
36.1 OVERARCHING ISSUES
science students accept theories on the authority of teacher and text, not because of evidence.
(Kuhn 1977)
Except in their occasional introductions, science textbooks do not describe the sorts of problems that
the professional may be asked to solve and the variety of techniques available for their solution. Rather,
these books exhibit concrete problem solutions that the profession has come to accept as paradigms
[however, students] must, we say, learn to recognize and evaluate problems to which no unequivocal
solution has been given; they must be supplied with an arsenal of techniques for approaching these future
problems
(Kuhn 1977)
So what is left to say? The twin goals of differentiation and integration (see Section 1.1) were
attained in Chapters 1 through 35. Facts and paradigms
1
relevant at each level of the biological
hierarchy were presented and then interconnected as much as presently possible. Relative to the
conduct of normal science,
2
this will foster the “determination of significant fact, matching of
facts with theory, and articulation of theory” (Kuhn 1970) at all relevant scales. Hopefully, an
appropriate balance was achieved by including more ecology than usually found in ecotoxicology
primers. The imbalance between autecotoxicological and synecotoxicological themes is generally
recognized as an important shortcoming of ecotoxicology as currently practiced (e.g., Cairns 1984,
1989; Chapman 2002). Amore congruent treatment was also attempted by including relevant human

effects information rather than taking the contrived approach of “asking humans to step out of the
picture” when discussing human influences on the biosphere. As justifying examples, discussion of
multidrug resistance transporter proteins (Chapter 3), inflammation (Chapter 4), endocrine modifiers
(Chapter 5), Seyle’s General Adaptation Syndrome (Chapter 9), and human epidemiology metrics
(Chapter 13) surely provided enriching detail to considerations of contaminant effects on nonhuman
individuals and populations. So, what more can be said? The last important issues to be explored
are ways of recognizing when innovation is needed and the best way of fostering change in this
extremely important science.
Statements about possible changes to the science of ecotoxicology were made throughout this
book. Yet, other than some general discussion in Chapter 1, no specific advice was given about how
exactly one recognizes the need for and then contributes to healthy change. As expressed in Kuhn’s
quote above, the absence of this kind of guidance is a fundamental shortcoming of most textbooks.
To avoid committing such a sin of omission, this chapter provides context and specific advice about
recognizing and then contributing to necessary change. Effective change is particularly crucial in
our applied science in which much harm can be done to the biosphere and our well-being if a failing
1
Kuhn’s concept of scientific paradigm is applied here because it is as useful as it is hackneyed. According to Kuhn
(1970), paradigms are “universally recognized scientific achievements that for a time provide model problems and solutions
to a community of practitioners.” They are the best explanations or approaches currently available to the scientific community.
2
Activities of scientists are divided into normal and innovative science (Kuhn 1970). When participating in innovative
science, a scientist intends to directly challenge an existing paradigm, or to propose a novel one. In contrast, the practice
of normal science involves filling in or refining important details surrounding an existing paradigm, or building ancillary
concepts or connections that reinforce or enrich a current paradigm. Normal and innovative science are complementary
activities essential to the health of any science.
813
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 814 — #4
814 Ecotoxicology: A Comprehensive Treatment
paradigm is maintained too long. The emerging global warming paradigm is one example where

irreparable harm ata global scale could occur if effective paradigmscrutiny and possible replacement
were put off any longer. The approach here will be to focus on foibles impeding innovation, hoping
that understanding impediments helps minimize their influence.
For if we learn more about resistance to scientific discovery, we shall know more also about the sources
of acceptance, just as we know more about health when we successfully study disease. By knowing more
about both resistance and acceptance in scientific discovery, we may be able to reduce the former by
a little bit and thereby increase the latter in the same measure.
(Barber 1961)
General cognitive and social psychology concepts will be blended with those more directly
concerned with sciences. General psychological concepts were included because understanding
them is pivotal to understanding how ecotoxicology might be moved out of its scientific nonage.
As put simply by Bourdieu (2004), “ the obstacles to the progress of science are fundamentally
social.” Footnotes will be applied liberally to improve continuity despite numerous digressions about
unfamiliar materials.
36.1.1 GENERATING AND INTEGRATING KNOWLEDGE IN
THE
HIERARCHICAL SCIENCE OF ECOTOXICOLOGY
The ecotoxicologist’s vocation is eminently important and enormous. In our opinion, this new
science could emerge as one of the most important for addressing society’s major challenges of
the millennium. The working knowledge needed to make useful ecotoxicological predictions spans
broad temporal, spatial, and informational content scales (Figure 36.1). Outstanding challenges
are the full articulation of major issues, development of predictive tools for handling problems
arising at every scale, and most critically, the integration of concepts and tools for every scale into
Temporal
Spatial
Informational
Explanation Obser
vation Significance
Cell/tissue
Molecule

Organ(s)
Organism
Community
Significance Observation Explanation
Ecosystem
Biosphere
Population
FIGURE 36.1 The temporal, spatial, and informational complexity scales relevant to ecotoxicology. In order
for ecotoxicology to emerge as a self-consistent science, more integration of facts and paradigms is essential
along all three scales. The explanation–observation-significance concatenation described in Chapter 1 is equally
useful at any level from molecule to biosphere. Although applied most often from the lower (explanation)
to higher (significance) levels, this concatenation is often useful in the opposite (“top-down”) direction.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 815 — #5
Conclusion 815
a congruent whole. A haphazard “feeling our way” (Platt 1964) approach is insufficient for this task
so thoughtful discussion is needed to meet these challenges.
Heterophilous
3
communications between dissimilar individuals may cause cognitive dissonance because
an individual is exposed to messages that are inconsistent with existing beliefs, an uncomfortable
psychological state.
(Rogers 1995)
Homophily can act as an invisible barrier to flow of innovations within a system. New ideas usually enter
a system through higher status and more innovative members. A high degree of homophily means that
these elite individuals interact mainly with each other,
(Rogers 1995)
Fundamental social processes determine how readily a novel idea, vantage, or technique dif-
fuses into any group, including a scientific community.
4

The only, albeit important, distinction
between scientific and nonscientific communities is the rules by which beliefs acquire favored status
(Chapter 1). Despite this distinction, scientists remain subject to rules governing acceptance of ideas
and innovation in any social group. As stated by Barber (1961), the rational, open-minded tradi-
tion has a powerful influence in scientific communities yet it “works in conjunction with a number
of other cultural and social elements, which sometimes reinforce it, sometimes give it limits.” As
reflected in the above quotes, an obstacle to accepting new ideas is often the barrier presented by
homophily: scientists trained in a particular discipline encounter cognitive dissonance, resist, and
then react against ideas from outside their immediate training or practice; for example, a systems
ecologist’s negative response to ideas of a mammalian toxicologist or vice versa. The discomfort
invoked by dissonance prompts a person to seek association with those sharing similar ideas and to
actively thwart, or minimally isolate, those with different ideas. Most likely, this is the true root of the
distracting reductionism–holism debate criticized inChapters 1 (Section 1.2) and20 (Section 20.2.1),
not any definitive superiority of one or the other as an investigative vantage. These dynamics are
altered only when the discomfort of maintaining a failing paradigm or stance becomes harder to bear
than that associated with confronting cognitive dissonance. Kuhn (1977) describes such a crisis in
scientific communities as one of “pronounced professional insecurity” that forces resolution. Con-
trary to popular belief,
5
most scientists only abandon “the idol of certainty” (Popper 1959) under
duress.
Described as an essential tension by Kuhn (1970, 1977), scientists resist the discomfort of
change until that of bolstering a failing paradigm becomes harder to bear. It should come as no
surprise that strong conformists are most often the opinion leaders of groups, including those of
scientific communities, and innovators are the most actively censored members (Bourdieu 2004,
Rogers 1995) except during special occasions requiring change.
It is our opinion that ecotoxicology is in a period of tension relative to the integration of core
concepts from all pertinent scales of organization. Out of immediate necessity, the individual-based
paradigms and approaches of mammalian toxicology were wisely adopted in our young science to
address immediate problems. However, enough data and collective experience has accumulated

to expose the discomforting inconsistencies among level-specific paradigms and approaches that,
together, constitute a self-contradictory system. For example, most risk assessments reluctantly
3
Homophily is the degree to which two interacting or communicating individuals are similar in relevant attributes. The
opposite of homophily is heterophily (McPherson et al. 2001, Rogers 1995).
4
A scientific community is a “group whose members are united by a common objective and culture” (Hagstrom 1965).
5
Typifying the presumption of scientists being open-minded is Berkeley’s quote, “He must surely be either very weak,
or very little acquainted with the sciences, who shall reject a truth, that is capable of demonstration, for no other reason but
because it is newly known and contrary to the prejudices of mankind” (Berkeley 1710). In reality, strong-willed and informed
scientists often display these behaviors. Barber (1961) provides an insightful counterpoint to this conventional image.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 816 — #6
816 Ecotoxicology: A Comprehensive Treatment
ignore central ecological principles, relying on LC50 and NOEL data for effects to individuals.
Recently, a pragmatic species sensitivity distribution approach combining individual LC50 or NOEC
values has been proposed to estimate a concentration protective of ecosystems (see Posthuma et al.
2002 for details). Such tests tacitly ignore critical species interactions such as those described in
Chapter 27 among killer whales, pinnipeds, otters, urchins, and kelp.
6
As explained by Popper
(1959), such self-contradictory systems eventually fail because they are uninformative systems from
which “ no statement can be singled out, either as incompatible or as derivable, since all are
derivable.” As evidenced in the primary literature and reflected in this textbook, a paradigm shift
seems to have begun in which ecological paradigms and techniques are applied in an increasingly
congruent and concerted manner with those emerging from mammaliantoxicology. In times likethis,
the ideal opinion leader is more innovative than typical and innovators are given more credibility
than otherwise warranted.
7

Ecotoxicology will shatter into a loose collection of interface discip-
lines (Odum 1996) in the absence of such leadership and the recognition of contributions made by
innovators.
Sociology provides clues about how best to act as an individual practitioner in a science under-
going necessary change. The most useful is related to the above discussion of homophilic and
heterophilic diffusion of innovations. Homophilic communication, the form of exchange that is
most frequent and least likely to produce cognitive dissonance, is also the form least likely to
contain novel information with which to solve an emerging problem. The more challenging and
frustrating, heterophilic communication is more likely to produce rapid diffusion of crucial know-
ledge into one’s evolving social group. This is the basis of the Strength of Weak Ties Theory,
i.e., weak, heterophilic communication forms bridging links containing the most novel information
with which to address challenges faced by a social group (Rogers 1995). What is the specific mes-
sage to be taken from this by a nonleader practitioner of ecotoxicology today? Instead of seeing
ecotoxicologists working at a different scale than your immediate group as erring competitors to
be coped with, consider the possibility that your scientific activities would be most enhanced by
exploring the vantage or ideas of these heterophilic ecotoxicologists. In doing this, strong emphasis
should be on discovering consistency among levels and enriching interpretation at any particular
level.
Heterophilic communication among ecotoxicologists working at different hierarchical levels is
not enough to meet the challenge: the communication must also be thoughtful in order to reach
sound judgments. For example, satisficing is a common danger in decision making by well-intended
individuals. Satisficing is a flawed form of decision making that often emerges in groups of indi-
viduals with very different agendas and vantages (i.e., heterophilic groups). Instead of seeking
the best possible decision, the group takes “a course of action that is ‘good enough,’ that meets
a minimal set of requirements” of all parties (Janis and Mann 1977). The operating premise is
that a “barely ‘acceptable’ course [is] better than the way things are now.” Janis and Mann
(1977) and similar books ondecision making theory and practiceprovide concrete meansof minimiz-
ing suboptimal heterophilic interactions like satisficing. Certainly, the influence of satisficing during
heterophilic scientific deliberations can be reduced by adopting Chamberlin’s multiple working
hypothesis scheme (Chamberlin 1897) in combination with Bayesian abductive inference methods

6
This is an example of the behavior called collusive lying, that is, “two parties, knowing full well that what they are
saying or doing is false, collude in ignoring the falsity” (Bailey 1991) (see also Section 16.4). It is a technique of groups
attempting to establish a useful paradigm in which they push away an inconsistent fact until the fledgling paradigm has been
given sufficient clarity and detail to be assessed properly. Unfortunately, this understandable behavior also carries the risk
of uncritical acceptance by the ecotoxicology community based on the stature of advocates with consequent long delay in
eliminating such a compromised paradigm when its shortcomings are eventually revealed.
7
It is equally important to understand that innovators are very poor opinion leaders when no change is needed (Rogers
1995). Their distracting exploration of unhelpful innovations can slow the accumulation of facts and ancillary concepts by
normal science around useful, established paradigms.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 817 — #7
Conclusion 817
(Howson and Urbach 1989, Josephson and Josephson 1996, Pearl 2000) that will be discussed at the
end of this chapter.
36.1.2 OPTIMAL BALANCE OF IMITATION,INNOVATION, AND
INFERENCE
Transmission [of ideas and innovations] withers on the vine when the present is taken as the only model.
And innovation itself withers with it, scorn for the past being the greatest enemy of progress.
(Debray 2000)
Very often the successful scientist must simultaneously display the characteristics of the traditionalist and
of the iconoclast.
8
(Kuhn 1977)
In the rest of this chapter, we try to sketch out the most efficient way of incorporating crucial
innovations while preserving existing, valuable ecotoxicological theory and practice. The next two
sections explore the relative virtues of resisting and embracing change to scientific knowledge.
36.1.2.1 The Virtues of Imitation
Among the forces that support social rules there is the imperative of regularization of “falling into the

line with the rule”
(Bourdieu 2004)
The dominant players [in a science] impose by their very existence, as a universal norm, the principles
that they engage in their own practice. This is what is called into question by revolutionary innovation
A major scientific innovation may destroy whole swathes of research and researchers as a side-effect,
without being inspired by the slightest intention of doing damage It is not surprising that innovations
are not well received, that they arouse formidable resistance
(Bourdieu 2004)
Invention of alternates is just what scientists seldom undertake except during the pre-paradigm stage
of their science’s development and at very special occasions during its subsequent evolution. So long
as the tools a paradigm supplies continue to prove capable of solving the problems it defines, science
moves fastest and penetrates most deeply through confident employment of those tools. The reason is
clear. As in manufacture so in science retooling is an extravagance to be reserved for the occasion that
demands it.
(Kuhn 1977)
Change isresisted for obvious reasons, some high minded, and othersnot. Often ascientist’s behavior
and activities have committed them to a paradigm that is now questioned, resulting in their beliefs
and hard work also being questioned—“A threat to theory is therefore a threat to the scientific
life” (Kuhn 1977). To expect open and immediate acceptance from such a scientist is to expect the
superhuman. Such a person will resist dismissing the value of their past work or giving up their
hard-earned professional status. More importantly, if a certain level of resistance to change were
not present in a scientific community, forward progress would be stymied by frequent detours or
side trips to explore novel paradigms that eventually turned out to be dead ends. That is the point
being made in the quote above from Kuhn (1977). It is inefficient to keep “retooling” a science
8
Kuhn (1977) adds an important footnote to this statement, “Strictly speaking, it is the professional group rather than the
individual scientist that must display both these characteristic simultaneously.”
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 818 — #8
818 Ecotoxicology: A Comprehensive Treatment

when there is no need. Furthermore, Loehle (1987) notes that insistence on testing and potentially
rejecting a concept before sufficient related details have accumulated by normal science can lead
to premature (“dogmatic”) falsification of a perfectly sound paradigm. A current case in point is
the species sensitivity distribution approach mentioned briefly above. There is real virtue to healthy
resistance to changewhenfaced withanovel explanation orsolution. The keyis avoiding pathological
resistance.
36.1.2.2 The Wisdom of Insecurity
Our innate social psychology is probably that bequeathed to us by our Pleistocene ancestors.
(Richerson and Boyd 2005)
We have analyzed this problem using several mathematical models of the evolution of imitation, and all
of them tell the same story. Selection favors a heavy reliance on imitation whenever individual learning
is error prone or costly, and environments are neither too variable nor too stable. When these conditions
are satisfied, our models suggest that natural selection can favor individuals who pay almost no attention
to their own experience, and are almost totally bound to what Francis Bacon called the “dead hand of
custom.”
(Richerson and Boyd 2005)
social influences exist that tend to form habits of thought leading to inadequate and erroneous
beliefs.
(Dewey 1910)
As evidenced by the above quotes, the strategy of imitation—uncritical acceptance of actions or
beliefs of those around us—is prominent in human interactions. Evolution favored imitation in our
Pleistocene ancestors as they attempted to survive in bands of hunter-gatherers. It remains intact
in modern social groups (Richerson and Boyd 2005), including scientific communities. However,
modern groups exist in a very different environment and have goals other then hunting and gath-
ering: sometimes our evolved behaviors remain useful, but in others, they are maladaptive. One
healthy modern manifestationis informational mimicry, a behavior in financial corporations in which
a group mimics another particularly knowledgeable group, rather than attempting to formulate its
own strategy (Vernimmen et al. 2005). A maladaptive example is faculty mobbing, an odd tendency
of university faculty to mob
9

overachieving members “who stand out from the crowd,” endangering
the common good (Westhues 1998, 2005). These evolved strategies, which were optimized for Pleis-
tocene hunter-gatherers social groups, arguably do not always result in optimal fitness of a modern
group of scientists and, consequently, require some understanding and coping behavior. As already
discussed, optimal fitness relative to a scientific community’s goal involves a thoughtful balance of
adherence to traditional explanation (mimicking) and openness to plausible alternatives. Imitation
has clear advantages as long as a minimum number of competent innovators exist when change is
required.
A group’s ability to respond to change is placed at risk when too few innovators exist, either
because of too active exclusion or because of passive neglect in teaching innovation skills. Clearly
articulating to students the valued role of innovation and then actively teaching problem solving
skills are two pedagogical necessities in a healthy modern science, especially a nascent one like
ecotoxicology.
Fitting Gigerenzer’s (2000) metaphor to the present topic, trying to change a scientific discipline
can be like trying to move a cemetery. Given the social psychological backdrop described above, the
9
“Workplace mobbing is ‘a common and bloodless form of workplace mayhem’ (Maguire 1999), usually carried out
politely and without violence” (Westhues 2005).
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 819 — #9
Conclusion 819
challenge for leaders in our field is facilitating necessary change in the presence of natural resistance.
An inappropriately conservative (or innovative) opinion leader will be swept aside after a period of
ineffective confusion. The challenge for opinion leaders is discerning when and where innovation is
truly needed. Amajor component of any solution is adopting a way of reliably discerning the relative
plausibility of candidate explanations, and then effectively updating these plausibility estimates
as new information emerges from normal science activities. An ideal system would allow the most
effective identification and then communication of the need for change to the group’s members. Such
a system should be resistant to processes such as satisficing or groupthink (a suboptimal process
prevalent in homophilic group decision making in which group concurrence is the objective, not

the best decision Box 36.1) (Janis and Mann 1977). Unfortunately, the unstructured expert opinion
systems on which we currently depend for making regulatory and many scientific judgments are
prone to these kinds of decision making errors (Cooke 1991). The approach of melding Chamberlin’s
multiple working hypothesis schemes and Bayesian abductive inference methods described at the
end of this chapter provides one possible means of doing this.
Box 36.1 Minimizing Groupthink
Groupthink, a pervasive problem in homophilic group decision making, has significantly
impacted our recent history. A classic example is the flawed decision making that occurred
in the Kennedy administration that led to the failed Bay of Pigs invasion. More recently, NASA
groupthink contributed substantially to the 1986 Challenger space shuttle disaster. Groupthink
will be examined closer here because of our pervasive dependence on the error-prone expert
opinion approach for establishing much ecotoxicological consensus and assessing ecological
risk. Although underexploited by natural scientists in their decision making, there are simple
ways of reducing groupthink’s influence during group activities.
The qualitiesof groupthinkare described by Janis and Mann (1977). Importantly, groupthink
occurs with “in-groups” (homophilic groups) in which concurrence is a highly valued charac-
teristic of the group dynamics. Rationalizations are invoked to preserve and foster concurrence.
In-groups experiencing groupthink often manifest eight characteristics:
1. Members tend to feel minimally vulnerable to mistakes and become overconfident
in their abilities. Consequently, they take on more risk in decisions than warranted
by facts.
2. There is a consorted effort to dismiss contrary facts or opinions, and to rationalize.
3. The group’s “inherent morality” becomes a given during deliberations.
4. Foibles are exaggerated and strengths trivialized for rivals or those with contrasting
opinions.
5. Direct pressure is applied to any member who questions the group’s actions or
stances.
6. Members who have doubts practice self-censoring in which they remain silent
despite their misgivings.
7. That silence indicates concurrence is a shared assumption of the group.

8. Self-appointed “mindguards”emerge whoaggressively act to protect thegroup from
erring members or information inconsistent with emerging consensus.
Groupthink can erode the quality of a group’s deliberations. Fortunately, there are means
of reducing its influence (Table 36.1). Panels, committees, and less-formal scientific teams
can benefit greatly from being aware of them. It is easy to imagine groupthink emerging during
a panel’s deliberations when applying Fox’s or Miller’s qualitative rules (see Section 13.2.1
and Box 13.2 in Chapter 13, and Box 22.3 in Chapter 22) or codified Environmental Protection
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 820 — #10
820 Ecotoxicology: A Comprehensive Treatment
TABLE 36.1
Tools for Reducing Groupthink
1. The leader or empowering agency should objectively emphasize the need for impartiality at the onset of the decision-
making process.
2. The leader or empowering agency should state the importance of expressing objections and concerns, making this
an obligation of each group member.
3. One member of the group should be assigned the role of skeptic or challenger during each decision-making session.
4. Perhaps by dividing the group occasionally to conduct separate assessments, the group should periodically assess
the feasibility of the group’s current stance.
5. If a rival group with contrasting views can be identified, enough members of that group should be as engaged as
possible in establishing possible alternate decisions.
6. After the consensus-building meetings have occurred, a “second chance upon reflection” meeting should be planned
to air any concerns emerging after the group breaks up.
7. Significant experts not associated with the core group should be invited to engage with the group with the request to
act as a “fresh pair of eyes.”
8. Each group member should be asked to discuss the group’s thoughts and progress with trusted peers and report back
the results of these independent exchanges.
9. If possible, separate groups can beestablished that address the same problemor question. The decisions from different
groups are then used to come to a final decision.
Source: Summarized from pages 399–400 of Janis and Mann (1977).

Agency (EPA) guidance (2000) to determine the cause or risk associated with a particular
scenario.
36.1.2.3 Strongest Possible Inference: Bounding Opinion and
Knowledge
It is therefore worth while to search out the bounds between opinion and knowledge.
(Locke 1690, reprinted 1959)
What current approach best balances conceptual stability and change? How can we define the bounds
between mere opinion and sound knowledge? Clues can be found in many places. Some may be
familiar while others may induce a level of discomfort, which was identified earlier as a characteristic
of heterophilic exploration of concepts.
Previously, we explored the concept of strong inference as articulated in the classic Science
paper by Platt (1964). The conventional Baconian scientific method is advocated by Platt with
emphasis being placedon consistent applicationof such aninductiveinference approachina scientific
discipline. He holds up as an exemplary example of a conditional logic tree the process that a chemist
might employ to qualitatively determine the nature of a substance. A series of positive/negative tests
are conducted in an exclusionary manner until only one possibility remains. Unfortunately, there are
substantial shortcomings associated with advocating such an approach as the kingpin of scientific
inquiry. Even assuming that the qualitative(positive/negative) natureof sucha processis adequatefor
all tasks, the presence of Type I and II error ratesrestricts the value of such a simple approach, making
it prone to error in the hands of the naïve practitioner (e.g., Box 10.2). Type I and II errors must be
considered in orderto make sensibledecisions.Also, manyecotoxicological judgments ofplausibility
are based on quantitative information for which such a dichotomous approach is suboptimal or prone
to logical error (see again Box 10.2). Some require the adaptive inference strategy described in
Section 29.5.3.
An even more serious problem with using such a strictly Popperian falsification scheme arises
from trying to incorporate the second major component of Platt’s strong inference approach, the
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 821 — #11
Conclusion 821
method of multiple hypotheses, which Platt (1964) claims to be “ the second great intellectual

invention which is what is needed to round out the Baconian scheme.” Chamberlin’s (1897)
multiple hypotheses approach attempts to reduce the bias toward any particular hypothesis(ses)
during testing by requiring that all plausible hypotheses be given equal amounts of effort during
testing. Unfortunately, adhering tothis approachis extremelydifficult and often becomes contrived in
ecotoxicology. The complexity (high dimensionality) and high uncertainty of many ecotoxicological
issues limits the value of any testing method that requires the classic dichotomous “accept/reject”
context advocated by Popper and Platt, and institutionalized in Fisher’s significance testing. While
current null hypothesis-based testing remains invaluable, dogmatic rejection of any other logical
approach of gauging plausibility of an explanation creates an impasse for the ecotoxicologist.
As long as there is an institutionalized [null hypothesis testing] methodology that does not encourage
researchers to specify their hypotheses, there is little incentive to think hard and develop theories from
which such hypotheses could be derived.
(Gigerenzer 2000)
Fortunately, an approach for reducing these difficulties exists for which the approach advocated
by Platt(1964) is a special case. Itis calledthe Strongest PossibleInference approachin this bookonly
for the purposes of identifying it as a simple extension of Platt’s Strong Inference and emphasizing
the conditional nature of any ensuing judgments of scientific hypothesis/explanation plausibility.
It is not novel, being a straightforward application of quantitative abductive inference. The strongest
possible inferences available to ecotoxicologists can be made at this time with abductive inference
as formalized in Bayesian inference formulations.
10
Associated Bayesian methods are pervasive
and widely available, and many textbooks (e.g., Gelman et al. 1995, Neapolitan 2004, Pearl 2000,
Woodworth 2004) and software [e.g., Netica
®
(Norsys Software)] facilitate their implementation.
Explanation of Strongest Possible Inference will begin by repeating Locke’s premise that “our
assent ought to be regulated by the grounds of probabilities” (Locke 1690). This seventeenth-century
quote contains the essence of abductive inference and Bayesian statistical inference. Abductive
inference is simply inference that favors the most probable explanation or hypothesis (Newman and

Evans 2002). Josephson and Josephson (1996) use the following syllogism for abductive inference:
D is a data collection about a phenomenon.
H explains the data collection, D.
No alternate hypothesis (H
A
) explains D as effectively as H does.
∴ H is probably true.
The key to applying abductive inference is quantifying “as effectively as” and “probably.” Bayesian
techniques permit quantification of abductive inference. The logic can be shown for using data to
judge a single hypothesis:
D provides support for H if P(H | D)>P(H).
D draws support away from H if P(H | D)<P(H).
D provides neither undermining nor supportive information if P(H | D) = P(H).
where P(H) = the probability of the hypothesis being true before any consideration of the data, and
P(H | D) = the probability of the hypothesis being true given the data. The task becomes estimating
the probabilities. Evidenceis combinedwith aprior probabilityof H being trueto producea statement
of probability given the evidence—a new probability of an explanation being true is established.
If more evidence (D
NEW
) was then collected during an inquiry, the newly established probability
can be used as the new “prior probability”
11
and combined with D
NEW
to calculate a new post
10
See Boxes 10.2 and 13.3 as instances in which we have already applied Bayesian methods for this purpose.
11
The probability is a “prior probability” relative to the collection of the new data.
© 2008 by Taylor & Francis Group, LLC

Clements: “3357_c036” — 2007/11/9 — 18:40 — page 822 — #12
822 Ecotoxicology: A Comprehensive Treatment
probability reflecting the plausibility of the H, given D and D
NEW
. Bayes’s theorem (Equation 36.1)
can be used to estimate P(H | D) in this case
P(H | D) =
P(H)P(D | H)
P(D)
, (36.1)
where P(D | H) = the probability of getting the data given the hypothesis was true, and P(D) = the
probability of getting the data regardless of whether or not the hypothesis was true. The resulting
P(H | D) can become the new prior probability (P(H
NEW
)) with the collection of additional data
P(H | D
NEW
) =
P(H
NEW
)P(D
NEW
| H)
P(D
NEW
)
. (36.2)
This process can be repeated with the addition of data until the associated probability is “good
enough” to make an evidence-based judgment of hypothesis or explanation plausibility. Obviously,
if more information becomes available, it can be modified again. This same process can be applied

to judging any hypothesis against its negation (∼H), a single alternate (H
A
), several alternate hypo-
theses (e.g., H
A1
, H
A2
, H
A3
). Equation 36.3 illustrates how the posterior odds for H versus
H
A
being true (P(H | D)/P(H
A
| D)) can be calculated from the prior odds (P(H)/P(H
A
)) and
likelihood ratio (P(D | H)/P(D | H
A
)):
P(H | D)
P(H
A
| D)
=
P(H)
P(H
A
)
P(D | H)

P(D | H
A
)
. (36.3)
Equation 36.4estimates theprobability of an hypothesis from its prior (P(H)), the prior of its negation
(P(∼H)), P(D | H), and P(D |∼H):
P(H | D) =
P(H)P(D | H)
P(H)P(D | H) + P(∼H)P(D | P(D |∼H))
. (36.4)
If one thinks carefully for a moment about Equation 36.4, it will become clear that P(H | D)
expresses the Positive Predictive Value (PPV) of a test, the usefulness of which was illustrated in
Box 10.2 for conventional hypothesis testing. (See page 81 in Gigerenzer 2000 for more details.)
Equation 36.5 is a generalization in which the probability of the ith hypothesis or explanation (H
i
)
of n hypotheses/explanations being true given information (D)
P(H
i
| D) =
P(D | H
i
)P(H
i
)

n
i=1
P(D | H
i

)P(H
i
)
. (36.5)
So, the plausibility of an explanation or relative plausibilities of a set of alternate explanations
can be judged using evidence-based probabilities. These estimates of belief warranted by evidence
can be recalculated periodically, and associated explanations reinforced or discarded as evidence
accumulates. Calculations can includethe simple“accept/reject” contextdescribed by Popper(1959),
or more complex contexts with higher uncertainty.
An excellent illustration of this approach can be found in the publications of Lane, Hutchinson,
and coworkers (Cowell et al. 1991, Hutchinson et al. 1989, Lane 1989, Lane et al. 1987) although
they focus on applications during medical diagnosis. Newman and Evans (2002) and Newman
et al. (2007) discuss their direct application to ecological risk assessment. Gigerenzer (2000, 2002)
provides many examples of applying these methods in risk assessment and communication.
For the readerwho isputoff by theabove equations, it maybe helpful torealize thatfairlycomplex
situations can be rendered to this framework by using natural frequencies instead of probability
equations (Gigerenzer 2000, 2002). The reader is urged to review Box 13.3 in which Bayesian
methods were applied with more intuitive diagrams.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 823 — #13
Conclusion 823
Box 36.2 Fish Kills due to Toxic Dinoflagellate Blooms or Hypoxia?
P. piscicida was implicated as the causative agent of 52±7% of the major fish kills on an annual
basis in North Carolina estuaries and coastal waters.
(Burkholder et al. 1995)
A series of large fish kills began in mid-Atlantic USA estuaries and coastal waters in the early
1990s. Regional resource managers and politicians asked scientists to determine the cause so
a remedy could be found. Early in the process, the notionally toxic dinoflagellate Pfiesteria
piscicida was identified as the cause by North Carolina researchers (e.g., Burkholder et al.
1992). This conclusion leaned heavily on statements like the one quoted above. The basis

for this statement was 3 years of monitoring data in which high levels of P. piscicida were
found at 8 of 15, 5 of 8, and 4 of 10 large fish kills. When counterarguments were presented
that episodic low oxygen events could be the cause, a confrontation took place that “became
mired in accusations of ethical misconduct, risk exaggeration, and legislative stonewalling”
(Newman et al. 2007). The maladaptive features of heterophilic exchange manifested, leading
to the resource managers being poorly served.
Newman et al. (2007) recently used this situation to illustrate how Bayesian methods could
quantify the relative plausibility of two competing explanations. They began by reiterating the
point of Stow (Stow 1999, Stow and Borsuk 2003) that the above quote demonstrates a common
error. The above quote describes data with which P(Pfiesteria | Fish Kill) can be estimated, yet
the direct conclusion was made about P(Fish Kill | Pfiesteria). These probabilities are different.
Bayes’s theorem (Equation 36.1) can be used to show how they are related,
P(Fish Kill | Pfiesteria) =
P(Fish Kill)P(Pfiesteria | Fish Kill)
P(Pfiesteria)
,
where P(Fish Kill) = the probability of a fish kill, P(Pfiesteria) = the probability of finding
high Pfiesteria levels at an estuarine or coastal location regardless of whether or not a fish kill
occurred. Newman and Evans (2002) used published reports to derive estimates of P(Pfiesteria)
(either .205 or .345 depending on the data applied) and P(Fish Kill) (approximately .081), and
used these probabilities along with Burholder et al.’s P(Pfiesteria | Fish Kill) to estimate the
P(Fish Kill | Pfiesteria). Depending on which data were used for P(Pfiesteria), the probability
of getting a large fish kill given the presence of high Pfiesteria levels (P(Fish Kill | Pfiesteria))
was estimated to be between .122 and .205. The likelihood of getting a large fish kill if high
levels of Pfiesteria were present (12–21%) was lower than the originally inferred 52%.
What about the relative likelihoods of large fish kills being associated with Pfiesteria
versus hypoxia? Rearrangement of Equation 36.3 produces a likelihood ratio that answers this
question:
P(H | D)
P(H

A
| D)
=
P(H)
P(H
A
)
P(D | H)
P(D | H
A
)
P(Pfiesteria | Fish Kill)
P(Low DO | Fish Kill)
=
P(Pfiesteria)
P(Low DO)
P(Fish Kill | Pfiesteria)
P(Fish Kill | Low DO)
P(Fish Kill | Pfiesteria)
P(Fish Kill | Low DO)
=
P(Pfiesteria | Fish Kill)
P(Low DO | Fish Kill)
P(Low DO)
P(Pfiesteria)
where P(Low DO | Fish Kill) = probability of low dissolved oxygen concentration if a large
fish kill occurred, P(Low DO) = probability of low dissolved oxygen concentration at
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 824 — #14
824 Ecotoxicology: A Comprehensive Treatment

a location, and P(Fish Kill | Low DO) =the probability of a large fish kill given low dissolved
oxygen concentration at a location. Newman et al. (2007) used published or North Carolina
State agency records for the period and region studied in Burkholder et al. (1995) to obtain
estimates of the probabilities, P(Low DO | Fish Kill) = .220 and P(Low DO) = .095. These
probabilities were inserted into the equation for the likelihood ratio:
P(Fish Kill | Pfiesteria)
P(Fish Kill | Low DO)
=
P(Pfiesteria | Fish Kill)P(Low DO)
P(Low DO | Fish Kill)P(Pfiesteria)
=
(0.520)(0.095)
(0.220)(0.345 or 0.205)
= 0.651 or 1.095.
A conditional, evidence-based inference from these likelihoods is that hypoxia is as, or
slightly more, likely to have caused a large fish kill than Pfiesteria.
This kind of reasoning is more easily understood and communicated using natural
frequencies (e.g., “42 out of 100”) as argued persuasively by Gigerenzer (2000, 2002) and
illustrated in Box 13.3. Figure 36.2 illustrates how the natural frequency approach can lead to
these same inferences using the P(Pfiesteria) estimate of 0.205. From that figure, the likelihood
ratio can be calculated:
421 Cases of large Fish Kills with high Pfiesteria levels
1884 Cases of no large Fish Kills with high Pfiesteria levels
= 0.22346
178 Cases of large Fish Kills with low dissolved oxygen concentrations
873 Cases of no large Fish Kills with low dissolved oxygen concentrations
= 0.20389
Likelihood ratio =
0.22346
0.20389

= 1.096.
The advantage of Bayesian inference is the ability to quantitatively express the degree of
belief warranted by evidence in a particular explanation. Any quantitative expression can be
FIGURE 36.2 An example of applying
Bayesian inference using natural frequen-
cies. Probabilitiesareplacedalongsidearrows
associated with different states. The num-
bers associated with each branch are based
on 10,000 cases. As examples, 810 of 10,000
cases will involve a large fish kill or 421 of
10,000 cases will involve a large fish kill and
the presence of high levels of Pfiesteria.
Large
fish kill
High
Pfiesteria
High
Pfiesteria
Large
fish kill
Low
oxygen
Low
oxygen
Yes
0.081
(810)
No
0.919
(9190)

Yes
0.520
No
0.480
Ye s
0.205
No
0.795
421 Cases
of kills with
Pfiesteria
389 Cases
of kills without
Pfiesteria
1884 Cases
of no kills
with Pfiesteria
7306 Cases
of no kills without
Pfiesteria
Yes
0.081
(810)
No
0.919
(9190)
Yes
0.220
No
0.780

Ye s
0.095
No
0.905
178 Cases
of kills with
low DO
632 Cases
of kills without
low DO
873 Cases
of no kills
with low DO
8317 Cases
of no kills without
low DO
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 825 — #15
Conclusion 825
updated as new evidence is obtained. The equations above also suggest what information is
most needed to permit belief-based action (i.e., funding and legislative decision making) to be
effectively aligned with or withdrawn from a particular explanation. In this case, more accurate
and precise information for the associated probabilities would be extremely helpful for both
the hypoxia and Pfiesteria explanations for fish kills.
The one issue unresolved relative to the Strongest Inference Possible approach is establish-
ing the probability at which confidence is sufficiently high to accept a choice or explanation. Just
as we concluded in Chapter 10 relative to hypothesis and equivalence testing, there is no simple
answer because the issue covers a broad array of situations. However, this difficulty does not
force us to fall back on qualitative approaches such as Hill’s rules: conditional answers can be
provided.

Acceptable probabilities can be chosen depending on the stage of scientific inquiry or the seri-
ousness of decision error consequences. This logic is similar to that employed by the EPA and other
regulatory agencies in establishing thresholds for acceptable “excess mortality” (i.e., the 1 in 10
4
to
1in10
6
rule). To be of most utility, probabilities should be agreed on a priori. As an overly sim-
plified illustration, a civil action in U.S. courts requires a preponderance of evidence (i.e., P > .50),
but a criminal trial requires a level of belief “beyond a reasonable doubt” (P > .70–.90?), which
experience has shown to be more difficult to pinpoint (Cohen 1977). Attempts have been made
during expert elicitation exercises to define probabilities associated with expressions of warranted
belief. The Kent chart provided by Cooke (Table 2.4) is one example:
Highly likely to near certainty 90–99%
Probable 60%
Even or about even chances 40–50%
Improbable 10%
Nearly impossible 1%
Cooke (1991) describes techniques to achieve best consensus about probabilities during expert
opinion elicitations. Obviously, the same type of process is needed to choose between alternate
explanations or decisions. How different do the associated probabilities have to be in order to decide
to conditionally accept one and reject another?
36.2 SUMMARY: SAPERE AUDE
12
Ecotoxicology’s ambitious goals, immediate obligations to society, and unquestionable success
in generating a rich information base have created the need for integration of information and
explanations into a congruent whole. Drawing on the concepts sketched out in this chapter, informa-
tion created via normal science has succeeded in producing enough cognitive dissonance that change
in existing paradigms must occur. Suggestions about how an ecotoxicologist might recognize and
facilitate effective change are provided and a Strongest Inference Possible approach advocated as

the best means of judging the relative merits of hypotheses or explanations in situations ranging
from high to very low certainty. The Strongest Inference Possible approach is simply an exten-
sion of Platt’s Strong Inference approach. Platt himself extended the Baconian approach by simply
insisting on consistent application of the exclusionary “scientific method” and incorporation of
Chamberlin’s multiple hypotheses approach. The Strongest Inference Possible approach only adds
12
“Dare to Know,” an intellectual challenge made famous by Immanual Kant.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 826 — #16
826 Ecotoxicology: A Comprehensive Treatment
to Platt’s Strong Inference the integration of Bayesian inference methods that allow most effect-
ive inferences in situations varying in levels of uncertainty. The ecotoxicologist’s task of judging,
explaining, and integrating information for all relevant levels requires the most effective approach.
We suggest that the Strongest Inference Possible approach is the most effective approach currently
available.
36.2.1 SUMMARY OF FOUNDATION CONCEPTS AND PARADIGMS
• Ways of fostering appropriate innovation in ecotoxicology were highlighted in this
chapter. Effective change is crucial in this applied science in which delay can result in
substantial harm to the biosphere and our well-being.
• Unnecessary stasis or change can impede the forward movement of any science.
• The major impediments to scientific progress are social; therefore, social solutions are
required.
• Diffusion of innovation can be accelerated by understanding the value and shortcomings
of heterophilic and homophilic interactions.
• On the basis of the Strength of Weakest Links theory, the most productive (and challen-
ging) interactions for an ecotoxicologist working at a particular scale will be those with
ecotoxicologists working at a different scale.
• During periods requiring change, opinion leaders should become more innovative and
heterophilic interactions should increase. During periods in which unnecessary change
would impede normal science, the opposite is most useful in a scientific community.

• Ecotoxicology is currently in a period of tension that requires more innovation, especially
in connecting factsand concepts emerging at different temporal, spatial, and informational
content scales (Figure 36.1).
• Platt’s classic Strong Inference approach can be extended via Bayesian inference
techniques to create a powerful tool for judging plausibility of explanations.
REFERENCES
Bailey, F.G., The Prevalence of Deceit, Cornell University Press, Ithaca, NY, 1991.
Barber, B., Resistance by scientists to scientific discovery, Science, 134, 596–602, 1961.
Berkeley, G., Principles of Human Knowledge/Three Dialogues, Penguin Books Ltd., London, UK, 1710.
Bourdieu, P., Science of Science and Reflexivity (translated by R. Nice), University of Chicago Press, Chicago,
IL, 2004.
Burkholder, J.M., Glasgow, H.B., Jr., and Hobbs, C.W., Fish kills linked to a toxic ambush-predator
dinoflagellate: Distribution and environmental conditions, Mar. Biol. Prog. Series, 124, 43–61,
1995.
Burkholder, J.M., Noga, E.J., Hobbs, C.H., Glasgow, H.B., Jr., and Smith, S.A., New “phantom” dinoflagellate
is the causative agent of major fish kills, Nature, 358, 407–410, 1992.
Cairns, J., Jr., Are single species toxicity tests alone adequate for estimating environmental hazard? Environ.
Monit. Assess., 4, 259–273, 1984.
Cairns, J., Jr., Will the real ecotoxicologist please stand up, Environ. Toxicol. Chem., 8, 843–844, 1997.
Chamberlin, T.C., The method of multiple working hypotheses, J. Geol., 5, 837–848, 1897.
Chapman, P.M., Integrating toxicology and ecology: Putting the “eco” into ecotoxicology, Mar. Pollut. Bull.,
44, 7–15, 2002.
Cohen, L.J., The Probable and the Provable, Oxford University Press, Oxford, UK, 1977.
Cooke, R.M., Experts in Uncertainty. Opinion and Subjective Probability in Science, Oxford University Press,
New York, 1991.
Cowell, R.G., Dawid, A.P., Hutchinson, T., and Spiegelhalter, D.J., A Bayesian expert system for the analysis
of an adverse drug reaction, Artif. Intell. Med., 3, 257–270, 1991.
Debray, R., Transmitting Culture (translated by E. Rauth), Columbia University Press, New York, 2000.
© 2008 by Taylor & Francis Group, LLC
Clements: “3357_c036” — 2007/11/9 — 18:40 — page 827 — #17

Conclusion 827
Dewey, J., How We Think, Reprinted in 2005 by Barnes & Nobles Books, New York, 1910.
EPA., Stressor Identification Guidance Document, EPA 822-B-00-025, December 2000, Office of Research and
Development, Washington, D.C., 2000.
Gelman, A., Carlin, J.B., Stern, H.S., and Rubin, D.B., Bayesian Data Analysis, Chapman & Hall/CRC,
Boca Raton, FL, 1995.
Gigerenzer, G. Adaptive Thinking, Oxford University Press, Oxford, UK, 2000.
Gigerenzer, G., Calculated Risks, Simon & Schuster, New York, 2002.
Hagstrom, W.O., The Scientific Community, Basic Books, New York, 1965.
Howson, C. and Urbach, P., Scientific Reasoning. The Bayesian Approach, Open Court, La Salle, IL, 1989.
Hutchinson, T.A. and Lane, D.A., Assesing methods for causality assessment of suspected adverse drug
reactions, J. Clin. Epidemiol., 42, 5–16, 1989.
Janis, I.L. and Mann, L., Decision Making. A Psychological Analysis of Conflict, Choice, and Commitment,
The Free Press, New York, 1977.
Josephson, J.R. and Josephson, S.G., Abductive Inference. Computation, Philosophy, Technology, Cambridge
University Press, Cambridge, UK, 1996.
Kuhn, T.S., The Structure of Scientific Revolutions, 2nd ed., The University of Chicago Press, Chicago, IL,
1970.
Kuhn, T.S., The Essential Tension, The University of Chicago Press, Chicago, IL, 1977.
Lane, D.A., Subjective probability and causality assessment, Appl. Stochastic Models Data Analy., 5, 53–76,
1989.
Lane, D.A., Kramer, M.S., Hutchinson, T.A., Jones, J.K., and Naranjo, C., The causality assessment of adverse
drug reactions using a Bayesian approach, Pharmaceut. Med., 2, 265–283, 1987.
Locke, J., An Essay Concerning Human Understanding, Collated by A.C. Fraser and reprinted by Dover
Publications, Inc., New York, 1690, reprinted 1959.
Loehle, C., Hypothesis testing in ecology: Psychological aspects and the importance of theory maturation,
Q. Rev. Biol., 31, 145–164, 1987.
McPherson, M., Smith-Lovin, L., and Cook, J.M., Birds of a feather: Homophily in social networks,
Annu. Rev. Sociol., 27, 415–444, 2001.
Neapolitan, R.E., Learning Bayesian Networks, Pearson Prentice Press, Upper Saddle River, NJ, 2004.

Newman, M.C. and Evans, D.A., Enhancing belief during causality assessments: Cognitive idols or Bayes’s
theorem? InCoastal and Estuarine Risk Assessment, Newman, M.C., Roberts, M.H., Jr., andHale,R.C.,
(eds.), CRC Press, Boca Raton, FL, 2002, pp. 73–96.
Newman, M.C., Zhao, Y., and Carriger, J.F., Coastal and estuarine ecological risk assessment: The need for
a more formal approach to stressor identification, Hydrobiol., 577, 31–40, 2007.
Odum, E.P., Preface, In Ecotoxicology. A Hierarchical Treatment, Newman, M.C. and Jagoe, C.H. (eds.),
CRC Press/Lewis Publishers, Boca Raton, FL, 1996.
Pearl, J., Causality, Models, Reasoning and Inference, Cambridge University Press, Cambridge, 2000.
Platt, J.R., Strong inference, Science, 146, 347–352, 1964.
Popper, K.R., The Logic of Scientific Discovery, Routledge, London, UK, 1959.
Posthuma, L., Suter, G.W., II, and Traas, T.P. (eds.), Species Sensitivity Distributions in Ecotoxicology,
CRC Press/Lewis Publishers, Boca Raton, FL, 2002.
Richerson, P.J. and Boyd, R., Not By Genes Alone, University of Chicago Press, Chicago, IL, 2005.
Rogers, E.M., Diffusion of Innovations, 4th ed., The Free Press, New York, 1995.
Stow, C.A., Assessing the relationship between Pfiesteria and estuarine fish kills, Ecosystems, 2, 237–241,
1999.
Stow, C.A. and Borsuk, M.E., Enhancing casual assessment of estuarinc fish kills using graphical models,
Ecosystems, 6, 11–19, 2003.
Vernimmen, P., Quiry, P., Le Fur, Y., Dallocchio, M., and Salvi, A., Corporate Finance. Theory and Practice,
Wiley & Sons, New York, 2005.
Westhues, K., Eliminating Professors. A Guide to the Dismissal Process, Kempner Collegium Publications,
Queenston, ON, 1998.
Westhues, K., The Envy of Excellence. Administrative Mobbing of High-achieving Professors, The Edwin
Mellen Press, Lewiston, NY, 2005.
Woodworth, G.G., Biostatistics. A Bayesian Introduction, John Wiley & Sons, Hoboken, NJ, 2004.
© 2008 by Taylor & Francis Group, LLC