Inquiry

ISSN: 0020-174X (Print) 1502-3923 (Online) Journal homepage: http://www.tandfonline.com/loi/sinq20

The epistemic significance of consensus

Aviezer Tucker

To cite this article: Aviezer Tucker (2003) The epistemic significance of consensus , Inquiry,
46:4, 501-521, DOI: 10.1080/00201740310003388

To link to this article: http://dx.doi.org/10.1080/00201740310003388

Published online: 25 May 2010.

Submit your article to this journal

Article views: 141

View related articles

Citing articles: 6 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=sinq20

Download by: [Erasmus University] Date: 28 September 2016, At: 09:11

 Inquiry, 46, 501–521

The Epistemic Significance of Consensus

Aviezer Tucker
Australian National University

Philosophers have often noted that science displays an uncommon degree of

consensus on beliefs among its practitioners. Yet consensus in the sciences is not a
goal in itself. I consider cases of consensus on beliefs as concrete events. Consensus
on beliefs is neither a sufficient nor a necessary condition for presuming that these
beliefs constitute knowledge. A concrete consensus on a set of beliefs by a group of
people at a given historical period may be explained by different factors according to
various hypotheses. A particularly interesting hypothesis from an epistemic
perspective is the knowledge hypothesis: shared knowledge explains a consensus on
beliefs. If all the alternative hypotheses to the knowledge hypotheses are false or are
not as good in explaining a concrete consensus on beliefs, the knowledge hypothesis
is the best explanation of the consensus. If the knowledge hypothesis is best, a
consensus becomes a plausible, though fallible, indicator of knowledge. I argue that if
a consensus on beliefs is uncoerced, uniquely heterogeneous and large, the gap
between the likelihood of the consensus given the knowledge hypothesis and its
likelihoods given competing hypotheses tends to increase significantly. Consensus is a
better indicator of knowledge than “success” or “human flourishing”.

Philosophers have often noted that science displays an uncommon degree of

consensus on beliefs among its practitioners. Yet, consensus in the sciences is
not a goal in itself (Roth 1987, pp. 127–8; Goldman 1999, pp. 70–1). I
consider cases of consensus on beliefs as concrete events. Such events do not
imply the intrinsic normative desirability of consensus. Consensus on beliefs
is neither a sufficient nor a necessary condition for presuming that these
beliefs constitute knowledge (Goldman 1999, p. 357). History is replete with
examples of consensuses on a variety of silly beliefs, and individuals may
possess knowledge against majority opinions.
A consensus on beliefs in a concrete group over a certain period is a
concrete event that can be explained by competing hypotheses. From an
epistemic perspective, the most interesting such hypothesis is that knowledge
is the best explanation of a consensus. In what follows I refer to this as the
knowledge hypothesis. The epistemic significance of the knowledge
hypothesis does not privilege it otherwise; it certainly does not make it true.
Alternative hypotheses may explain better a concrete consensus on beliefs.
For example, a consensus may emerge as a result of shared biases or common
political interests. Epistemology is interested in comparing the hypothesis
that shared knowledge generated a given consensus on beliefs with alternative
hypotheses that explain this consensus by shared class background, shared

DOI 10.1080/00201740310003388 ! 2003 Taylor & Francis

502 Aviezer Tucker

professional interests, a Zeitgeist, or for that matter the constellation of the

zodiac, etc. Obviously, not all competing hypotheses are of equal merit. This
paper is devoted to explicating the conditions under which the knowledge
hypothesis is a better explanation of consensus than other competing
hypotheses.
The evaluation of the knowledge hypothesis does not require taking sides
in the contentious philosophical debate over the definition of knowledge.
Nonetheless, whatever definition of knowledge is adopted, it must be
independent of the consensus it should explain, since if we define knowledge
in terms of consensus on beliefs, the knowledge hypothesis is circular and
vacuous. A version of the classical definition of knowledge as justified true
belief, or a revised version of it as a true belief that was fashioned by a reliable
process, or Dretske’s (1981, p. 86) proposal that knowledge is beliefs caused
or sustained by information, would all do just as well. If it can be shown that
shared knowledge is the best explanation of a concrete consensus on beliefs,
this consensus would mark a region in a network of beliefs that is likely to
reflect knowledge. This is a useful procedure prior to the examination of the
justification of beliefs, or the reliability of the process, or the flow of
information that lead to the beliefs.

The Case Against the Epistemic Significance of Consensus

Nicholas Rescher (1993, p. 13) criticized what he took to be Habermas’s
concept of consensus on the grounds that it begs the question of rationality.
Philosophers who consider consensus philosophically significant as an
indicator of truth are aware of the fact that consensus may emerge for a host of
different reasons. If by ‘consensus’ we mean an actual agreement between
real people, it may be based on shared mistakes or coercion. In Rescher’s
interpretation, Habermas limited his philosophic discussion of consensus to
what he considers its ideal rational type, according to an ideal process,
achieved exclusively by force of argumentation. But any account of epistemic
consensus that reduces consensus to a process that should generate the proper,
significant, type of consensus begs the question. If a consensus is
epistemically significant only if it follows a particular process, then the
process is important, not the ensuing consensus. It would be necessary to
study the process that leads to consensus, to examine its rationality, to decide
whether it resulted in knowledge or mere opinion. The use of consensus to
elucidate rationality must presuppose an account of rationality in order to
discover which consensus matters in the first place. Similarly, arguments for a
particular kind of consensus-generating process (Lehrer and Wagner 1981) or
against it (Levi 1985, Caws 1991) are not relevant for assessing the epistemic
significance of consensus, insofar as the concern in such debates ultimately
 The Epistemic Significance of Consensus 503

lies with the epistemic standing of the process in question rather than with the
consensus that stands behind it. It seems that a discussion of consensus is
epistemically superfluous.
It is possible to avoid the reduction of consensus to a process by discussing
an ideal consensus. The study of an ideal consensus in an ideal situation
permits ignoring actual cases of consensus that were irrational or founded on
false beliefs as deviations from the ideal. Charles Peirce (1877/1958, pp. 132–
3) believed that the scientific process will ultimately generate consensus on
objective truth. His faith in an unspecified scientific process of investigation is
combined with a quasi-religious teleological rhetoric about consensus as the
‘destiny’, ‘destined center’, ‘foreshadowed Goal’, and ‘predestined opinion’
of the history of science. Peirce offered no reason to suppose that the destiny
of science is consensus, or that science has any destiny. Consensus for Peirce
was not concrete, but an ideal to be realized at the end of time, an epistemic
kingdom of God. Before the promised consensus comes about at the end of
the scientific process, it is impossible to know which beliefs are objective
truth and which will be forsaken during the future history of science, just as it
is impossible to distinguish the righteous from the sinners before judgment
day. Peirce’s eschatological idealized consensus may be a kind of explication
of the concepts of ‘truth’ and the ‘real’, but it is not useful for epistemology.
Given a concrete consensus, Peirce would have nothing to tell us about the
probability that it is founded on knowledge.
The later Habermas followed Peirce in attempting to understand truth by
analyzing an ideal speech situation, and the pragmatic conditions for the
achievement of consensus. Rescher noted that if ‘consensus’ is used in the
sense of an ideal consensus under ideal circumstances that allow for a process
of unconstrained rational consideration, consensus is reduced to the outcome
of unbridled rationality. Unbridled rationality does not require a consensus. If
a single individual or a machine possesses all the relevant evidence,
background knowledge and pure rationality, a rational result will be inferred
inevitably and a consensus would be again superfluous. Rescher (1993, pp.
55–7) rejected the project of deciding on what is true on the basis of
consensus. He concluded that consensus has no intrinsic significance.
Rescher’s rejection of what he took to be Habermas’s project of
understanding rationality and truth through consensus led him (1993, pp.
17–20) to argue that consensus is philosophically insignificant. Still, Rescher
(1993, pp. 37–8) acknowledged that scientists find consensus useful for
correcting individual mistakes in measurement, and for assuring replication
and verifiability (in the sense of precise description). Rescher acknowledged
further that consensus among experts and practitioners on a belief is ‘a
decisive probative consideration’ in its favor: ‘Agreement is indicative of an
invariance of sorts: it reflects the fact that, not withstanding the variation of
the particular epistemic stance of different individuals, people’s efforts at
504 Aviezer Tucker

epistemic problem-solving favor one particular outcome. And this sort of

invariance is clearly a positive evidential factor’ (Rescher 1993, p. 61).
I endorse Rescher’s criticism of attempts to reduce epistemically
significant consensus to the proper outcome of real or ideal processes. In
what follows I attempt to clarify and develop Rescher’s acknowledgement
that consensus can be a positive evidential factor and therefore has an
instrumental value in discovering groups that share knowledge.

The Best Explanation of Consensus

A concrete consensus on a set of beliefs by a group of people at a given
historical period may be explained by different factors according to various
hypotheses. The knowledge hypothesis holds that, given background
conditions, shared knowledge explains a consensus on beliefs. If all the
alternative hypotheses to the knowledge hypotheses are false or are not as
good in explaining a concrete consensus on beliefs, the knowledge hypothesis
is the best explanation of the consensus. If the knowledge hypothesis is best, a
consensus becomes a plausible, though fallible, indicator of knowledge.
Though the knowledge hypothesis may be a better explanation of a consensus
than all the competing hypotheses, this explanation is still fallible because a
better hypothesis than the knowledge hypotheses is always possible.
One may object that the number of alternative hypotheses to the knowledge
hypothesis is infinite. ‘All sorts of methods can yield consensus: brainwash-
ing, the threat of the rack or burning at the stake, totalitarian control of the
sources of information….’ (Goldman 1987, p. 120). If Goldman is right, the
labor of eliminating alternative hypotheses or proving them to be of inferior
quality to the knowledge hypothesis is Sisyphean. But Goldman did not
consider that it may be unnecessary to examine all the logically possible
alternatives to decide whether a concrete consensus on beliefs is a likely
indicator of knowledge. In concrete cases of consensus on beliefs, the actual
number of moot alternative hypotheses is usually quite manageable because
alternative explanations are usually associated with general theories in the
sociology of knowledge. The knowledge hypothesis does not have to prove its
absolute truth, only its comparative advantage over existing competitors.
I argue that if a consensus on beliefs satisfies the following three
conditions, the gap between the likelihood of the consensus given the
knowledge hypothesis and its likelihoods given competing hypotheses tends
to increase significantly. The group of individuals that share a consensus on
beliefs should be: (1) uncoerced; (2) uniquely heterogeneous; (3) sufficiently
large. Let’s consider these three conditions in turn.
 The Epistemic Significance of Consensus 505

Uncoerced Consensus
An epistemically significant consensus must be uncoerced. As Caws (1991, p.
379) answered Goldman, coercion is no foundation for real consensus, but for
‘unwilling acquiescence’. The coercion of individuals may take many forms.
Some may be intimidated by threats. Other may be manipulated by their
economic dependence. Still others may be browbeaten. The literature about
deliberative democracy considers various types of ‘coercion’ that interfere
with the democratic process (Dryzek 2000, pp. 57–80). Iris Marion Young
(quoted in Dryzek 2000, p. 68) suggested that some may be punished by the
‘normalizing gaze of the group’. Arguably, rhetoric and argument by
authority may be coercive. Some people may argue better than others, or may
deny others information about their premises (ibid., pp. 68–72). All these
types of influences over the opinions of individuals may lead to the
establishment of a discursive hegemony or to the institutional domination of a
particular segment of the population. But they are insufficient for coercing a
consensus on beliefs. Some people are not easily intimidated, others have a
strong character and do not react to threats, and many people take their own
opinions sufficiently seriously to express them regardless of the effect it may
have on their personal fate. Historical experience demonstrates that the
establishment of even a local coerced consensus on public expressions of
belief in a state-sanctioned dogma requires the extensive and extreme use of
violence. Both the Inquisition and the Soviet NKVD secret police required
more than intimidation and institutional hegemony, they required killing and
imprisoning those who would not be coerced by other means. The overt
public consensus in the Soviet Union on Lysenko’s theory of biological
adaptation did not imply the presence of knowledge. Soviet scientists did not
consider the merits of the theory, but the high probability of a long
incarceration in the Gulag if they denied it. Indeed, since some scientists
defied this official dogma despite the threat, consensus was achieved
ultimately only by the physical elimination of dissent. Even that was
insufficient for achieving consensus because scientists in other countries
continued their work in genetics; those Soviet geneticists who survived the
Gulags or managed to study genetics privately re-established the field during
the 1960s. An average contemporary academic environment in a democratic
country contains many positive and negative sanctions for publicly espousing
this or that belief. These sanctions probably skew the aggregate public
reporting of views of individual academics, in comparison with what honest
reporting of their beliefs would have looked like. Such sanctions can coerce
public expressions of belief by weak, opportunistic, or economically
dependent individuals but cannot coerce a consensus. Even an all powerful
totalitarian state and its secret police can only coerce a local involuntary
acquiescence by the extensive use of extreme violence.
506 Aviezer Tucker

Uniquely Heterogeneous Consensus on Beliefs

The unique heterogeneity of a consensus group generates the strongest
evidence against alternative hypotheses to the knowledge hypothesis.
Consensus in a heterogeneous group can be used to eliminate extraneous
variables, insofar as uniformity of results in a heterogeneous group can
substantiate a causal link in controlled scientific experiments. When it is
impossible either to control extraneous intervening variables in a laboratory
setting, or to select a random sample and subject it to statistical analysis,
scientists attempt to control extraneous variables by selecting an experimental
group composed of members that do not share any of the properties they
attempt to control. For example, when a pharmaceutical manufacturer wishes
to examine the effectiveness of a new drug, it is impossible to select a random
sample of humanity, infect it with a disease, inject it with the cure and see
what happens. Nor is it possible to find the effects of age or gender on the
progress of the disease and cure by manipulating at will the age and gender of
human guinea pigs. Instead, scientists select a heterogeneous group of people
who are already infected with the disease but otherwise do not share
properties (such as age, genetic makeup, socio-economic background,
medical history etc.) that could be responsible for their recovery. If after
taking the new drug all the members of the heterogeneous experimental group
are cured, it is safe to assume that the medicine was the common cause, rather
than the various factors that members of the group do not share. If all the
members of the experimental group are cured except a sub-group that is
homogeneous in say, sharing the stage of development of the disease or in
having a particular genetic makeup, it is still warranted to assign the cure to
the new drug because the best explanation of the exceptional, dissenting,
group is its homogeneous makeup. As long as the cured group is uniquely
heterogeneous, i.e., all the sub-groups composed of patients who are not
cured are homogeneous in certain relevant respects, the best explanation of
the cure is the medicine.
In the absence of coercion, the more heterogeneous the consensus group,
the easier it is to refute alternative hypotheses that link properties that only
some members of the consensus group share with the consensus. Alternative
hypotheses may attempt to explain a concrete consensus on beliefs with
particular power relations, or political interests, or ideological convictions, or
gender bias, or cultural contexts. But if the consensus group is composed of
some persons who are not connected to each other in any power relationship,
others who do not share political interests and ideologies, of both genders and
many cultures, such hypotheses are less probable than the knowledge
hypothesis. If an alternative to the knowledge hypothesis maintains that a
concrete consensus, e.g., on the law of conservation of energy or Darwinian
evolution, is caused by properties like the respective national cultures of
 The Epistemic Significance of Consensus 507

Germany and England, and if some of the members of the consensus group do
not share this property, then the knowledge hypothesis is more plausible than
the alternative hypothesis.
The evaluation of a consensus group as sufficiently heterogeneous depends
on existing alternative hypotheses to the knowledge hypothesis. To be
considered sufficiently heterogeneous, members of the consensus group
should not share any of the properties that alternative hypotheses single out as
causes of that concrete consensus. A group that has been considered
heterogeneous may lose this status if all its members share a biasing property
that a new hypothesis posits as the cause of the consensus. For example,
feminist philosophers of science developed hypotheses that connect male
biases in scientific communities in medicine and biology with their scientific
beliefs (cf. Longino 1990, pp. 103–214; Okruhlik 1994). Before the
introduction of feminist philosophy of science, historians and philosophers
of science had not considered the gender composition of scientific groups to
be significant for the explanation of their consensus on certain beliefs.
Conversely, an alternative hypothesis to the knowledge hypothesis may be
dropped because a theory with which it is associated is proven false, and a
group that had not been considered heterogeneous is suddenly heterogeneous
enough. It is possible to guard only against conceivable biases that are
connected with alternative hypotheses to the knowledge hypothesis. It is
always possible that a relevant bias has been overlooked and the knowledge
hypothesis is false despite the apparent heterogeneity of the consensus group.
Confidence in the knowledge hypothesis is directly related to the variety and
diversity of alternative hypotheses. The more varied alternative hypotheses
there are, the easier it is to guard against bias.
People may reach a consensus on beliefs because of their various biases,
rather than despite them. A complex alternative theory to the knowledge
hypothesis may combine conjunctions and disjunctions of different factors to
explain the consensus. An analysis of the consensus group may discover
separate sub-groups that share beliefs because of different biases. Miriam
Solomon (2001) advocated such an explanation of scientific consensus.
Solomon suggested that individual scientists choose among competing
theories according to their different biases. Scientific consensus on beliefs
results from different biases (‘decision vectors’ in her terminology) that
somehow have the same effect. Better scientific theories are affected by a
greater variety and distribution of biases than theories that follow a more
narrow range of biases. Solomon did not regard biases, ‘decision vectors’
such as ideology or pride, deference to authority or agreement with scriptures
as impediments to the achievement of scientific knowledge, but as inevitable
necessary prerequisites for scientific progress. If Solomon is right, scientists
do not agree despite their different genders, cultures, interests etc., but exactly
because they are men and women of different cultures and of various
508 Aviezer Tucker

interests. Solomon collapsed the distinction among evidence (‘empirical

vectors’ in her terminology), cognitive values (‘theoretical vectors’), and
biases, which she calls ‘social vectors.’ Altogether, Solomon (2001, pp. 62–3)
estimated there are 50–100 decision vectors of all kinds that affect theory
choice by themselves or in interaction with each other.
Whereas I have proposed to use heterogeneity to prove that biases are
unlikely, Solomon suggested that diversity of biases is the best explanation of
concrete historical consensuses on beliefs among scientists. Both explana-
tions are possible, but ceteris paribus they are not equally plausible. Scientific
theories are constantly in the process of being accepted or rejected by new
scientists who enter science or become acquainted with a theory they were not
familiar with earlier. If Solomon is right, each new scientist brings into the
mental act of theory-acceptance a new set of individual biases, yet all result in
the same set of beliefs. Ceteris paribus, the likelihood of consensus on beliefs
in a heterogeneous group is higher given common knowledge than given
diverse biases because of the low probability that all biases will imply the
same set of beliefs. Take for example conflicting interests: the knowledge
hypothesis suggests that a consensus is likely given that scientists with
conflicting interests set these interests aside to form beliefs according to what
they know. Solomon would claim that consensus on beliefs is likely given
conflicting interests. This is possible though unlikely, since it is more
probable that causally efficacious conflicting interests would lead to different
beliefs. For example, partisan interpretations of the histories of conflicts in
places such as Northern Ireland or the Middle-East are inconsistent because
their authors have conflicting political, national and religious interests and
identities. If historians are capable of reaching consensus on many beliefs
concerning the history of these conflicts, it is more likely a result of their
shared knowledge than their conflicting interests. Formally, if C stands for a
consensus; K stands for knowledge; and B1, B2,… Bn stand for different
biases, then for Solomon:
Pr!C"B1 # $ Pr!C"B2 # $ ! ! ! !Pr!C"Bn # " Pr!C"K#
That is to say, the likelihood of the consensus given a bias multiplied by its
likelihoods given all the other biases must still be higher than its likelihood
given shared knowledge. For this to be the case, the likelihood of the
consensus given each of the biases must be higher than its likelihood given
shared knowledge. This can be the case only if the likelihood of the consensus
given all the biases is extremely high, or if the prior probability of knowledge
is extremely low. Ceteris paribus, the knowledge hypothesis is also simpler
and of a wider scope than complex alternative theories.
The position I present here is fallibilistic. It is conceivable, though unlikely,
that an examination of the history of science will discover that factors other
 The Epistemic Significance of Consensus 509

than knowledge usually suffice for the explanation of consensus. Solomon

attempted to support her thesis with case studies from the history of science.
However, I cannot agree that her examples support her thesis either because
cases that Solomon classified as examples of consensus are actually cases of
scientific dissent, or because I do not consider cases where cognitive values
determined a consensus on a theory to be cases of externalist biases. If
cognitive values are often conducive to attaining knowledge, as I argue later
in this article, the effectiveness of cognitive values in generating a consensus
on beliefs is consistent with the knowledge hypothesis.
Solomon (2001, p. 109) suggested that the consensus on plate tectonics in
geology emerged gradually and was affected by prestige and informal
networks. Non-empirical vectors (biases) operated on both sides of the plate
tectonics debate (ibid., p. 121). But if there were two sides, there was no
consensus. Eventually, when consensus on plate tectonics was established, it
was based on overcoming biases, on shared knowledge. The molecular
approach to genetics was supported by a ‘decision vector’ that favored
theories that unite disparate areas of science, in this case by reducing biology
to chemistry. What Solomon (2001, pp. 111–4) considers a metaphysical bias
are the cognitive values of simplicity and scope. Solomon claimed that
consensus on what she termed ‘the central dogma’ in genetics had negative
results because it led to ignoring important work that did not fit this dogma.
But if there was such research, there was no consensus on the ‘dogma’
because a heterogeneous group of scientists whose work was ignored by the
majority dissented. If the work of Jewish and female scientists was ignored by
scientists working within the ‘nuclear monopoly’ (ibid., p. 122), there was no
heterogeneous consensus on the central dogma. Solomon described a
historical consensus on the hypothesis that males display greater genetic
variety than females and explained it by a sexist bias:
Even though well-respected educational psychologists (mostly women such as Helen
Bradford Thompson (Wooley) and Leta Stetter Holdingworth, but also a man: G. W.
Fraiser in a 1919 study that found no differences in variability) criticized the
hypothesis, giving both contrary data and alternative environmental explanations of
the existing data, broad consensus on the variability hypothesis continued. Much had
to happen – scientifically, politically, and socially – before this consensus was
dismantled (ibid., p. 123).
But if women scientists and at least one male expressed their disagreement
with the sex differences thesis, there was no consensus of beliefs about it!
Solomon uses ‘consensus’ here to describe a majority opinion or the position
of the academic establishment, but these do not amount to a true unity of
opinions. Solomon’s concept of ‘consensus’ is accordingly left vague:
‘Consensus is typically not all encompassing… since usually some dissent
remains. The decision whether or not to call a state of affairs “consensus” or
“dissent” is to some extent arbitrary.’ (ibid., p. 118) With a vague concept of
510 Aviezer Tucker

consensus, it is possible to adapt the judgment of what is considered a case of

consensus or dissent to the argument one wishes to make.
Paul Seabright (1988) asked a similar question to the one I have been
discussing: how can we explain various degrees of consensus in different
disciplines? Seabright answered that social and cultural reasons are sufficient
for explaining various levels of consensus. In his opinion, appeals to what he
calls ‘metaphysical’ factors such as truth, objectivity, and knowledge are
superfluous. The main flaw in Seabright’s argument is that his only example
for the redundancy of knowledge in explaining consensus is from South
Indian dietary traditions, which he described as: ‘what, with some latitude
will be described as a scientific discipline’ (Seabright 1988, p. 31). ‘Some
latitude’ is quite an understatement. The multiplication of dietary habits and
folk beliefs associated with them, has nothing to do with science, it represents
a normal pattern of cladistic diversification of traditions. But even more
serious, sophisticated and convincing externalist interpretations of the history
of science (cf. Shapin and Schaffer, 1985) would find it difficult if not
impossible to explain large and heterogeneous consensuses on beliefs as the
result of external political and other interests and biases, not just at the initial
stage of scientific innovation and immediate reception, which must emerge
somewhere sometime among some people, but also in the many subsequent
cases of belief-acceptance by an increasingly diversified population of
varying interests and biases.
Helen Longino (1994) also argued that consensus on beliefs results from a
wide distribution of personal differences. She added however a Habermasian
stipulation that the consensus must result from critical dialogues among
individuals and groups from different points of view, an ‘interactive dialogic
community’ (ibid., pp. 142–3). Longino stipulated further the properties of
the community that can transmute subjective points of view into scientific
objectivity: there must be public fora for criticism; the community must be
responsive to criticism; there must be revisable public standards, cognitive
values, to which participants in the discourse can appeal to reach agreement;
and the community should be intellectually egalitarian, where decisions about
beliefs are not taken by appeal to authority. Critical discourse should explain
how different biases and points of view are transformed into consensus on
beliefs. However, Rescher’s criticism is just as relevant here: Longino cannot
be interested in consensus per se, but in an ideal process for generating
knowledge. The conditions she stipulated have never been satisfied in any
historical case of consensus, and are of doubtful utility in explaining actual
historical cases of consensus on beliefs.
A consensus on beliefs may emerge as a result of different theoretical
assumptions, of a heterogeneous theoretical background. For example a
consensus on the date of an archeological artifact may emerge as a result of
carbon dating, analysis of the architecture of the remains of the house where it
 The Epistemic Significance of Consensus 511

was found, its materials and techniques, the discovery of coins nearby, and
analysis of the language in which texts found there were written. Accordingly
this consensus is founded on various theories from physics, archeology,
numismatics, and comparative linguistics (Kosso 2001, Wylie 1999). When
scientists use different theories to reach an identical conclusion, the theories
mutually confirm each other as well as the conclusion. These theories
‘bootstrap’ each other, to use Glymour’s (1980) expression. The best
explanation of these cases of consensus is still the knowledge hypothesis.
Their convergence on a single interpretation of the evidence is highly likely,
given the knowledge hypotheses.
There is no concrete universal consensus on beliefs because some people
always dissent. There are still people who deny that the earth is round or that
it revolves around the sun; and others who deny that there was a Holocaust or
that Fundamentalist Moslems destroyed the World Trade Center. This led
philosophers who consider consensus to be philosophically valuable to
attempt to prescribe whose opinions matter for determining whether there is a
consensus or dissent on beliefs. Reliance on professional organizations or
academic institutions, affiliations and certifications is a tempting approach. If
all professional astronomers agree that the earth is round and revolves around
the sun, it must be because they share knowledge of the heavens and those
who honestly dissent must be either ignorant or irrational. At their best,
professional and academic institutions should indeed reflect the qualities that
make consensuses among their members epistemically significant. However,
this appeal to authority too often failed in history. Academic and professional
institutions have proved themselves to be too susceptible to political coercion,
economic interests and graft, or the common biases of a class of people who
share professional interest and often social background, preferring their
institutional, class and professional interests to intellectual integrity.
The alternative approach I present here does not have to rely on authority or
esoteric knowledge, nor does it require universal consensus. It merely
requires a unique non-coerced heterogeneity. It does not matter if some
dissent remains, so long as the people who do dissent are sufficiently different
from each other to eliminate alternative hypotheses to the knowledge
hypothesis, and those who dissent are sufficiently homogenous to support
hypotheses that explain their dissent by particular biases. For example, the
group that reached consensus on Darwinian evolution is uniquely hetero-
geneous, it includes people who are secular and religious, and of many
different faiths. The community that upholds the alternative ‘creationism’ is
homogenous, composed exclusively of biblical fundamentalists, primarily
American protestants, though many American protestants believe in
Darwinian evolution. Their bias in favor of an anachronistic, historically
insensitive interpretation of Genesis is the best explanation of their beliefs.1
512 Aviezer Tucker

Sufficiently Large Consensus

The uncoerced heterogeneous group that reaches consensus must be
sufficiently large to avoid accidental results. Small groups can never be
sufficiently heterogeneous to exclude hidden biases. Therefore, statistical
samples have minimal sizes. If only four scientists work on a particular
problem and they agree on a set of related beliefs, it does not imply that their
agreement was caused by common knowledge. The four may be a professor,
her assistant, a former student, and an untenured member of faculty who
needs the professor’s vote to secure tenure. The minimal size of a significant
consensus depends on local circumstances such as whether the people who
develop a consensus are related socially, the nature of their relations, and
whether they attempted to replicate the process that generated the beliefs or
merely accepted the conclusions of others on faith or authority (cf. Sarkar
1997, p. 510). Usually, when the consensus involves hundreds of people who
are geographically, institutionally and professionally dispersed, it is safe to
assume that it is large enough. If the consensus is on an esoteric topic, and
only a handful of experts are competent or interested enough to reach the
consensus, they may possess knowledge, but their consensus cannot function
as an indicator of knowledge. It is necessary to follow their reasoning to
evaluate the status of their beliefs.

Comparison with Realism

Standard realist philosophy of science considers knowledge to be the best
explanation of the success of science (Putnam 1975). Why attempt to defend
the thesis that knowledge is the best explanation of consensus rather than
success? The short answer is that it is far easier to examine and measure
consensus than success. Success is value laden, persons that hold diverse
values may value the same results as ‘successful’ or ‘unsuccessful’. Solomon
divided scientific success into ‘empirical’ and ‘theoretical’. Though she
regards empirical success as necessary:
[n]o one has a good, comprehensive definition of empirical success…. Empirical
successes can be observational, predictive, retrodictive, experimental, explanatory or
technological – to mention some of the main categories. During different historical
periods, and in different fields, there has been focus on one or another kind of
empirical success…. There is no ‘paradigmatic’ or ‘typical’ kind of empirical
success…. Attempts to define ‘empirical success’ run the risk of being too specific,
and not including all empirical successes, or too vague, and not characterizing
anything. (Solomon 2001, pp. 21–2) There is a variety in what counts as a theoretical
success. Different historical periods, different scientific fields and different scientists
all can have different – even contrary – views about theoretical success (ibid., p. 17).
It is easier to agree on what constitutes uncoerced, uniquely heterogeneous
 The Epistemic Significance of Consensus 513

and sufficiently large consensus than on what constitutes success. ‘Success’

may be in prediction and in the generation of technological innovation
(Putnam 1975). But some kinds of scientific knowledge can neither generate
predictions (Hobbs 1993), nor have pragmatic applications that make human
life more tolerable. It is debatable whether disciplines such as evolutionary
biology, linguistics or some branches of psychology can generate any
predictions but extremely vague ones (Solomon 2001, p. 22). Sciences that
aim to know the past such as cosmology, comparative linguistics or
historiography have no obvious technological applications.

Cognitive Values and Consensus on Beliefs

It may be argued that the uncoerced, sufficiently large, and seemingly
heterogeneous groups whose consensus on beliefs should indicate knowledge
are actually homogenous because they share cognitive values. Values
describe what should be considered worthy of preference, as distinct from just
being preferred. Cognitive values discriminate which statements are worthy
of being considered knowledge, whether absolutely or in comparison with
competing statements. An alternative reformulation of the knowledge
hypothesis could suggest then that consensus on beliefs results from shared
cognitive values that determine what a community considers knowledge in
the first place. For example, the birth of science was accompanied by a shift
from cognitive values that considered knowledge to be the result of faith,
revelation, or the wisdom of the ancients, to ones that consider knowledge to
be the result of empirical investigations. Kuhn (1996, pp. 184–6) suggested
that the scientific community is constituted by values (accuracy, consistency,
scope, simplicity and fruitfulness) that are shared more widely than
theoretical models. Such values have changed more slowly than theories
during the history of science. Scientists may share cognitive values, but still
disagree because they dispute the evidence or have different theoretical
backgrounds. If shared cognitive values are necessary but insufficient
conditions for the emergence of consensus on beliefs, the knowledge
hypothesis may have to be qualified as relative to particular sets of cognitive
values.
It is possible to avoid this conclusion by denying that consensus on facts,
methods and theories necessitates consensus on cognitive values (Laudan
1984, pp. 43–6; Laudan and Laudan 1989). The logical relations between
cognitive values and methods and knowledge are neither sufficient nor
necessary, though shared axiomatic cognitive values ‘sure help’, as Laudan
(1984, p. 46) put it. Assessing the relations between cognitive values and
concrete consensuses on beliefs would require a more extensive survey of the
history of science than is possible here. The historical correlation between
514 Aviezer Tucker

consensuses on beliefs and consensuses on cognitive values may suffice to

warrant a discussion of cognitive values as partial causes of concrete
consensuses on beliefs. If Laudan’s analysis corresponds with much of what
happened in history, the following discussion is redundant.
Cognitive values emerged, became dominant and acceptable in the course
of history. If cognitive values come and go for no discernable reason, it is
difficult to explain why large uncoerced and heterogeneous groups of
scientists and laymen come to agree on them. Some philosophers retreated
into metaphysics and ‘explained’ the evolution of cognitive values by
reducing it to developments in the Zeitgeist (Hegel) or the history of being
(Heidegger). Yet, apart from making the often interesting observation that
some aspects of culture change simultaneously (Spengler), and perhaps
influence each other or are influenced by a common cause, these metaphysical
appeals do not say much beyond recognizing that various aspects of culture
relate to each other and change together, while claiming that human agency
can do little or nothing to affect historical changes in cognitive values.

The Best Explanation of Consensus on Cognitive Values

Various hypotheses may explain historical shifts in cognitive values. A
particularly interesting hypothesis from an epistemic perspective would
suggest that a concrete consensus on cognitive values emerges because new
cognitive values are more conducive to the attainment of knowledge than
those that preceded them; henceforth: the conduciveness hypothesis. Proving
the fallible plausibility of this thesis requires proving only that it is better than
competing explanations of concrete uncoerced, heterogeneous and suffi-
ciently large consensuses on cognitive values. The conduciveness hypothesis
can be supported by using the strategies that support the knowledge
hypothesis: when an uncoerced, sufficiently large and uniquely hetero-
geneous group agrees on a set of cognitive values, hypotheses that attempt to
explain the consensus by factors that only a part of the group share are at a
disadvantage in comparison with the conduciveness hypothesis.
Hypotheses that explain a uniquely heterogeneous consensus on cognitive
values by any particular social, cultural or historical variables would find it
quite difficult to explain the unique appeal of these values to very different
people. Sarkar (1997) criticized Kuhn for on the one hand claiming that all
scientists choose theories according to five cognitive values (accuracy,
consistency, scope, simplicity and fruitfulness) while on the other hand
maintaining that scientists form their opinions according to their various
social, cultural, national, historical, etc., circumstances. In Sarkar’s
interpretation, Kuhn suggested that though scientists have different starting
points, backgrounds, nationalities or upbringings, they travel different roads
 The Epistemic Significance of Consensus 515

to reach the same destination, the five cognitive values. Yet, according to
Kuhn, different backgrounds may affect the hierarchy of the five values, some
scientists may prefer scope to simplicity or vice versa. Sarkar is quite right to
describe Kuhn’s hypothesis as assuming an incredible accident, various
biases transmute into exactly the same five cognitive values. This incredible
accident parallels Solomon’s incredible accident, the transmutation of the
different biases of scientists into consensuses on beliefs. It is possible that
various biases would generate a uniquely heterogeneous consensus on
cognitive values. But ceteris paribus, it is less plausible than the
conduciveness hypothesis because it implies that the likelihood of a
heterogeneous consensus on cognitive values given each of the biases must
be higher than its likelihood given the conduciveness hypothesis.
If the uniquely heterogeneous, uncoerced, and large group that achieves a
consensus on certain beliefs assumes cognitive values that are supported by a
uniquely heterogeneous, uncoerced, and large consensus, it is likely to share
knowledge, because the consensus on cognitive values is likely given their
conduciveness to knowledge. However, if the uncoerced, uniquely hetero-
geneous and large group that reaches consensus on beliefs assumes cognitive
values that are upheld only by a homogenous group, the posterior probability
of the knowledge hypothesis is low. The value of the knowledge hypothesis is
the likelihood of the consensus on beliefs given shared knowledge multiplied
by the prior probability that the assumed cognitive values are conducive to
knowledge.
For example, scientific cognitive values replaced traditionalist cognitive
values. But the rules that recognize which tradition matters and who are the
legitimate bearers of tradition differ from one homogenous community to
another. A consensus on beliefs in a community that upholds a traditionalist
system of cognitive values can be achieved over time only through coercion.
Otherwise, natural variations and mutations fragment the tradition into
diverse traditions. The propositions that traditionalist cognitive values
recognize as knowledge evolve even in the presence of homogenizing
coercion. Coercion can only ensure that the evolution of tradition is uniform.
A crisis in traditionalist cognitive values emerges when members of a
community that share a tradition encounter other groups that have different
kinds of traditionalist cognitive values, other rules for recognition of
legitimate traditions. The realization that other communities have different
traditionalist cognitive values can shake the confidence that members of
homogenous communities have in their own traditionalist cognitive values.
Some members may react to the discovery of other traditions with fanaticism,
others may endorse value skepticism or search for the kind of cognitive values
that could be accepted by a heterogeneous community without coercion. The
first two solutions are unstable; fanaticism is self-destructive and skepticism
unfruitful. Better cognitive values are the only stable solution for the crisis of
516 Aviezer Tucker

traditionalism. Scientific cognitive values may be introduced either while

social acceptance of traditionalism is already in decline, or they may ‘sneak’
into an entrenched set of accepted but inconsistent cognitive values and
generate beliefs inconsistent with those resulting from older cognitive values.
Laudan (1984, p. 60ff) showed how a contradiction between implicit and
explicit cognitive values can cause change. Once these contradictions appear,
a revolutionary or gradual process of cognitive values replacement follows.

Comparison with Putnam’s Realist Justification of Cognitive

Values
Putnam (1990, pp. 135–41, pp. 163–78) accepted that theory-choice in
science progresses not just according to judgments of fact, but also according
to judgments of values such as simplicity, coherence, conservatism, and
instrumental efficacy. Cognitive values are necessary conditions for knowl-
edge of facts. Putnam avoided, though, the conclusion that knowledge is
relative to values. If a culture upholds certain values, it does not imply that
these values are not ‘objective’, in the sense of being valuable for what
Putnam (following Aristotle) called ‘human flourishing’ for more than one
culture. Cognitive values make an indirect contribution to human flourishing,
through the acquisition of knowledge. The acquisition of knowledge helps us
to orient ourselves in the world and consequently should improve our chances
for survival and flourishing. Putnam justified objective cognitive values by
using a different set of values, the ones that decide what should constitute
human flourishing. However, it is easier to agree on cognitive values than on
the values that constitute human flourishing. Putnam’s justification of
cognitive values resembles his justification of science in his realist stage. Both
are justified because they lead to a kind of success. In the context of such an
ideal set of universal values, cognitive values should indeed further contribute
to human flourishing. But whether or not there is a set of values that
universally maximize human flourishing, depending on whether a single
universal human flourishing can fit all, it is quite clear that lay persons as well
as philosophers have been unable to agree on the composition of such an ideal
set of universal values: hedonistic shoppers and Buddhist monks, Plato and
Aquinas have had very different ideas about what constitutes human
flourishing. There is more than a single version of the good life.2 Rawls,
Habermas, and Hampshire appealed to a hypothetical consensus on universal
values under ideal circumstances. But these circumstances abstract people
from their actual situation and are highly oversimplified. Universal values are
either too abstract and meaningless, or inapplicable in some circumstances
(Parekh 1999).
In concrete historical cases, the value of cognitive values for human
 The Epistemic Significance of Consensus 517

flourishing is context dependent. A simple example from the history and

philosophy of law may illustrate this point. Suppose that a legal system is
founded on two values: First, reverence of the past, leading to assignment of
authority to laws according to their alleged ancient origins. This reverence
may be accompanied by faith in divine revelation as the origin of the law, or
belief in a tradition that tells of ancient destructive social strife, halted by a
unique social compromise devised by an outstanding legislator of unique
wisdom and savvy, a Solon, who is the founder of the present constitutional
order. The achievement of the ancient legislator may be considered precious,
unique, and not fully comprehensible. It may be feared that any
reconsideration, any tinkering with the ancient constitution may reopen the
long buried conflicts that threatened society with self-destruction. Second,
while reverence for the past requires any verdict to be based on ancient law,
traditionalist cognitive values recognize propositions as knowledge according
to their connection to a chain of oral tradition that has allegedly been
preserving the original interpretation of the ancient texts or norms. This
second value of a traditional legal system is cognitive: it values a
jurisprudential tradition as the exclusive source of knowledge of the ancient
laws or norms. Litigants and society respect the authority of their judges
because they believe in an undistorted tradition that connects their judges
with the ancient law.
The authority of the judges relies on popular reverence of an ancient
constitution in conjunction with faith in the authority of a tradition that
authenticates the contemporary interpretation of the authoritative ancient
constitution. The outcome of the application of these two values to the legal
system can actually promote what many would consider human flourishing by
combining legal order with progress. Popular reverence of an ancient
constitution preserves social order. Popular faith in tradition as legal
knowledge, allows the judges to adapt the law to changing circumstances,
while preserving its authority. For example, Jewish law achieved in the last
centuries BCE, after the canonization of the bible, flexibility comparable to
that of Common Law. The rabbis revised and adapted Jewish law to changing
circumstances, while relying on the authority of its alleged divine origin.
They practically eliminated capital and corporal punishments by setting
impossible conditions for such sentences, set legal conditions for slavery that
made it highly uneconomical, and permitted charging interest to adapt Jewish
law to a vibrant Hellenic market economy that required assuming risk in
business transactions. The oral tradition allowed revisions, protected by the
traditional prohibition on writing itself down3. Respect for the authenticity of
tradition and the authority of the past allow legal authorities to reform, revise,
and modernize the tenets of the law, while preserving their traditional
authority. This result is quite similar to Dworkin’s (1986) theory of legal
518 Aviezer Tucker

interpretation in allowing judges to reform the law, while preserving the

authority of precedents.
In this context, traditionalist cognitive values promote what many would
consider human flourishing by obstructing the attainment of knowledge. Oral
tradition is an unreliable means for the preservation of information from the
past. But in the context of a society that worships its past, an inferior cognitive
value may have a higher value for human flourishing. If society maintains
reverence for the authority of the past, but at the same time drops its
traditionalist cognitive values, the question of authenticity of tradition will
lead either to skepticism about knowledge of the past or to the development of
something like the critical method in historiography. In the first case, if no
other authority or method replaces established tradition, myriad contested and
rivaling interpretations of ancient laws and folklore will emerge, leading to
civil strife over the correct interpretation of the ancient lore, like the religious
wars of early modern Europe. If traditionalist cognitive values are replaced by
the cognitive values of critical historiography, they would be used to discover
the authentic, or at least most ancient knowable form of the law. The
community would restore its constitution to its original or most ancient form
and interpretation. This would lead to the imposition of Draconian laws, laws
that were enacted at and for a particular historical context in a different
context, without revision or adaptation. In the context of reverence for the
past, forsaking traditionalist cognitive values and replacing them with
scientific cognitive values leads to decline in what many would consider
human flourishing either through civil strife, or through the imposition of
Draconian laws. The value of cognitive values for human flourishing is
context dependent. In some contexts, this value may be negative. The choice
of consensus instead of flourishing as an indicator of cognitive values most
conducive to knowledge, supported by the conduciveness hypothesis, is far
more convincing.

Conclusion: The Usefulness of Consensus on Beliefs

Consensus on beliefs may be epistemically useful as an indicator of
knowledge in three cases. First, a justified appeal to consensus is useful when
tracing a particular belief-forming process is impossible or very difficult
because evidence for the process is missing, or because following this process
or justifications is too time consuming, or if understanding the process or the
justifications requires specialized arcane background knowledge. For
example, lay patients ask for second and even third medical opinions because
they are unable to evaluate by themselves medical justifications for opinions
of specialized doctors, but presume that the best explanation of a medical
consensus is that it reflects common knowledge. Second, consensus on beliefs
 The Epistemic Significance of Consensus 519

may follow many different routes or different justifications, as when different

theories or methods yield identical conclusions. Instead of examining each of
these routes to the shared beliefs, it is easier to treat the consensus on beliefs
as evidence for shared knowledge, irrespective of how it was arrived at. For
example, if various professionals using different scientific methodologies and
background theories, agree in their dating of an archeological site, we can
presume that they are right without following their different methods.
Thirdly, we may be interested in examining in detail the belief-forming
process in a broad epistemic area where it is not obvious where to look for
knowledge-generating processes. For example, it is not always patently
obvious which parts of the social sciences are based on scientific knowledge,
and which present partisan opinions. Dividing the beliefs that uncoerced and
uniquely heterogeneous large groups of social scientists hold into those that
enjoy consensus and those that do not may indicate which parts are scientific
and which are merely the opinions of a section of the scientific community.
An uncoerced and uniquely heterogeneous consensus on beliefs, as a fallible
indicator of knowledge, can delineate an area of inquiry into justification of
beliefs when it is unclear where an epistemic inquiry should commence. Once
such an area in the net of beliefs is delineated, philosophers may inquire how
scientists or ordinary people reach this consensus. For example, if a
philosopher wishes to investigate the epistemology of our knowledge of
history, it is not obvious where to commence this inquiry. Beliefs about the
past are varied and have various epistemic values. It may be useful to
commence such an investigation with those beliefs that an uncoerced,
uniquely heterogeneous and large group of historians agrees on and examine
what are the theories and methods that participate in generating these beliefs.
An uncoerced, uniquely heterogeneous and large community of historians
emerged for the first time in the second quarter of the nineteenth century. A
philosophical examination of the epistemology of history may commence
there.

NOTES

1 Dissent, the simultaneous presence and persistence over time of uncoerced, heterogeneous
and large groups that have mutually inconsistent beliefs, e.g., on the nature of light or the time
and place where proto-Indo-European was spoken, does not prove that there is no objective
fact of the matter, or that consensus is impossible in principle, or that a group does not possess
knowledge, though unable or unwilling to convince other groups (Rescher, 1993, 45–50).
There may be natural constraints on obtaining certain types of knowledge as a result of the
structure of the world as it interacts with certain kinds of human theories and methods. Under
such constraints there may be more than one and only rational answer to a question under a set
of assumptions and information constraints (Rescher 1993, 9–10).
2 In Parekh’s (1999) opinion the only two values we can find across all cultures are patriarchy
and xenophobia.
3 When the tradition of unwritten law was broken in the third century CE and a version of
traditional law was written and codified as the Mishnah, it resulted in a more rigid system of
520 Aviezer Tucker

rabbinic law. Subsequent codifications of interpretations of this codified law (the Talmud,
Maimonidas’ Mishne Torah, Joseph Caro’s sixteenth century Shulhan Arukh etc.) gradually
increased the rigidity of the law further, until the emergence of reformed Judaism during the
Enlightenment.

REFERENCES

Caws, Peter 1991 ‘Committees and Consensus: How Many Heads Are Better than One.’
Journal of Medicine and Philosophy 16, 375–91.
Dretske, Fred I 1981. Knowledge and the Flow of Information. Cambridge MA: MIT Press.
Dryzek, John S 2000. Deliberative Democracy and Beyond: Liberals, Critics, Contestations.
Oxford: Oxford University Press.
Dworkin, Ronald 1986. Law’s Empire. Cambridge MA: Harvard University Press.
Glymour, Clark 1980. Theory and Evidence. Princeton: Princeton University Press.
Goldman, Alvin 1987 ‘Foundations of Social Epistemics.’ Synthese 73, 109–44.
Goldman, Alvin 1999. Knowledge in a Social World. Oxford: Oxford University Press.
Hobbs, Jesse 1993 ‘Ex Post Facto Explanations.’ Journal of Philosophy 90, 117–36.
Kosso, Peter 2001. Knowing the Past: Philosophical Issues of History and Archeology.
Amherst NY: Humanity Books.
Kuhn, Thomas S 1996. The Structure of Scientific Revolutions, 3rd edn. Chicago: Chicago
University Press.
Laudan, Larry 1984. Science and Values: The Aims of Science and Their Role in Scientific
Debate. Berkeley: University of California Press.
Laudan, Rachel and Laudan, Larry 1989 ‘Dominance and the Disunity of Method: Solving the
Problems of Innovation and Consensus.’ Philosophy of Science 56, 221–37.
Lehrer, Keith and Wagner, Carl 1981. Rational Consensus in Science and Society: A
Philosophical and Mathematical Study. Dordrecht Holland: Reidel.
Levi, Isaac 1985 ‘Consensus as Shared Agreement and Outcome of Inquiry.’ Synthese 62, 3–
11.
Longino, Helen E 1990. Science as Social Knowledge: Values and Objectivity in Scientific
Inquiry. Princeton: Princeton University Press.
Longino, Helen E 1994 ‘The Fate of Knowledge in Social Theories of Science.’ in Socializing
Epistemology: The Social Dimensions of Knowledge. Frederick F. Schmitt ed., Lanham MD:
Rowman & Littlefield 135–57.
Okruhlik, Kathleen 1994 ‘Biology and Society.’ Canadian Journal of Philosophy Suppl. 20,
21–42.
Parekh, Bhikhu 1999 ‘Non-ethnocentric universalism.’ in Human Rights in Global Politics.
Tim Dunne and Nicholas J Wheeler eds. Cambridge: Cambridge University Press 128–59.
Peirce, Charles S 1958 [1877] ‘How to Make Our Ideas Clear.’ in Selected Writings. Philip P.
Weiner ed. New York: Dover 113–36.
Putnam, Hilary 1975. Mind, Language and Reality, Philosophical Papers. Cambridge:
Cambridge University Press.
Putnam, Hilary 1990. Realism with a Human Face. James Conant, ed. Cambridge MA:
Harvard University Press.
Rescher, Nicholas 1993. Pluralism: Against the Demand for Consensus. Oxford: The
Clarendon Press.
Roth, Paul A 1987. Meaning and Method in the Social Sciences. Ithaca: Cornell University
Press.
Sarkar, Husain 1997 “The Task of Group Rationality: The Subjectivist’s View–Part II”,
Studies in History and Philosophy of Science. 28, 497–520.
Seabright, Paul 1988 ‘Objectivity, Disagreement, and Projectibility.’ Inquiry 31, 25–51.
Shapin, Steven and Schaffer, Simon 1985. Leviathan and the Air-Pump: Hobbes, Boyle, and
the Experimental Life. Princeton: Princeton University Press.
Solomon, Miriam 2001. Social Empiricism. Cambridge MA: MIT Press.
 The Epistemic Significance of Consensus 521

Wylie, Alison 1999 ‘Rethinking Unity as a “Working Hypothesis” for Philosophy of Science:
How Archaeologists Exploit the Disunities of Science.’Perspectives on Science 7, 293–317.

Received 4 November 2002

Aviezer Tucker, Social & Political Theory Program, Research School of Social Sciences,
Australian National University, Canberra ACT 0200, Australia. E-mail: avitucker@yahoo.com