THAD M.

BOTHAM

PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS

ABSTRACT. In response to a collection of essays in Jonathan Kvanvigs (1996) Warrant

in Contemporary Epistemology: Essays in Honor of Plantingas Theory of Knowledge,
Alvin Plantinga notices that certain Gettier-style examples undermine his (1993b) canon-
ical account of epistemic warrant as delineated in Warrant and Proper Function. In hopes
to clarify how his account survives Gettiers purchase, he (1996; 2000) argues that a belief
has warrant sufficient for knowledge only when produced in a favorable cognitive mini-
environment. In Warranted Christian Belief Plantinga (2000) specifies a condition required
for a cognitive mini-environments favorability. I argue that this condition falls prey to
counterexample. Then I investigate a possible solution, which I reason fails as well.

Plantinga (1993a, 4) takes warrant to be that normative, possibly complex

quantity that comes in degrees, enough of which is what distinguishes
knowledge from mere true belief. Roughly, a belief has warrant for
someone just when it is produced and maintained by cognitive faculties
functioning properly in a suitable environment according to a design plan
successfully aimed at truth.1 At the very least, an adequate epistemolo-
gical theory should yield the right results for paradigm cases of knowledge
(or lack thereof), and Plantinga thinks his theory can handle them. Yet
some philosophers, citing certain Gettier-style examples, have argued that
Plantingas canonical theory in WPF implies that some warranted true
beliefs count as knowledge when knowledge is clearly absent.2
Consider three Gettier-style counterexamples used to putatively under-
mine Plantingas account. First, an example predating Gettier: Bertrand
Russells stopped clock case. You glance at an analogue clock that reads
2:00. You believe it is 2:00, and your belief is true. However, the clock
just happened to stop exactly twenty-four hours ago. So your belief is
accidentally true from a cognitive point of view. Thus you have a justified
true belief that does not count as knowledge. You do not know it is 2:00,
since the clock could just as well have stopped at some other time. You
have a true belief only by sheer epistemic fortuitousness. Second, Carl
Ginets barn case: youre driving through Wisconsin. For every real barn in
a particular region of Wisconsin, the locals have erected three fake barns
to appear more prosperous. You drive through this area, see one of the
real barns, and form the belief Thats a fine barn. Your belief is true
and justified but does not constitute knowledge since your belief is only

Synthese 135: 431441, 2003.

2003 Kluwer Academic Publishers. Printed in the Netherlands.
432 THAD M. BOTHAM

coincidentally true given that there are so many barn facades in the vicinity.
You could just as well have formed the belief by looking at a fake barn.
Third, the identical twin case: you glance across the street and see your
good friend Paul leaving someones house. You believe the proposition
Theres Paul. However, Pauls identical twin brother, Peter, happens to
be visiting Paul for the weekend. Peter is still in the house, putting on his
coat. Furthermore, you are not even aware Paul has a sibling Peter lives
in Paris. So, you have a true justified belief that you see Paul. However,
you do not know that Paul is the person you see, since it is only a cognitive
coincidence that you see Paul instead of Peter. If Peter had emerged first,
then you still would have believed the proposition Theres Paul.
Critics would argue that in each of these cases Plantingas theory mis-
takenly implies that you do know, since you have a true belief produced by
faculties functioning properly in an environment presumably suitable for
those faculties according to a design plan successfully aimed at truth. They
reason along the following lines. Consider the nearest possible worlds
where the clock works just fine, where there are no fake barns bordering
the roadway, or where Pauls identical twin is nowhere in the vicinity. You
possess knowledge in all of these close worlds, yet Plantingas theory does
not capture the relevant difference between your actual situation and the
possible situations. For example, in the nearest worlds where the clock
does not stop, your belief that it is 2:00 is presumably produced by the
same exercise of cognitive powers as it is in the actual world. Furthermore,
the environment and design plan appears the same. However, in the actual
case you do not know that it is 2:00, while in the nearest worlds you do.
Similarly, for either the fake barn or identical twin case, your exercises
of cognitive power, your environments, and your design plans appear the
same in the nearest worlds where there are no fake barns or where Peter
is nowhere nearby, respectively.3 Plantinga (1996, 312) recapitulates the
objection, saying, The same bit of the cognitive system, governed by the
same bit of the design plan produces the same belief in the two situations
[e.g., in the actual Gettier situation and in any one of the situations in the
nearby worlds]; if the process in question is governed by a bit of the design
plan aimed directly at the truth in the one case, the same goes in the other.
In response, Plantinga claims that the environment in each of these
cases where the subject does not know is misleading or unfavorable
with respect to the subjects exercise of cognitive powers. For true beliefs
to possess enough warrant for knowledge, the environment must be favor-
able for their production.4 In WPF Plantinga thought of a suitable cognitive
environment (mostly) as one very much like the one we humans enjoy here
on earth: devoid of evil demons, of experimenting Alpha Centaurians, of
 PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS 433

wayward bursts of radiation, of complete darkness, of amnesia-inducing

drinking water, and so forth. This global type of environment is what
Plantinga now calls a maxi-environment. He still holds that knowledge
requires a suitable maxi-environment. But that is not all. If someones
cognitive faculties are to yield a true belief with enough warrant for know-
ledge, they need to operate within a favorable mini-environment as well.
Plantingas notion of a mini-environment is designed to account for the
lack of resolution exercises of cognitive powers display in certain scen-
arios, such as in the three Gettier cases mentioned above. For example, in
the stopped clock case your exercise of cognitive powers is not fine-grained
enough to detect at a glance the second hands failing to move. Intuitively,
a cognitive mini-environment with respect to someones exercise of cog-
nitive power is favorable only when the environment is not epistemically
misleading, i.e., only when the agents exercise of cognitive powers can be
counted on to yield some or other true belief in that mini-environment.
In short, Plantinga holds that each of the Gettier cases above illus-
trates that even though the maxi-environment is epistemically favorable,
the agents exercise of cognitive powers cannot be counted on to yield
a true belief with respect to its local cognitive mini-environment. Thus
Plantinga adjusts his environment condition. Here are three steps out-
lining his strategy to circumvent Gettier cases: (i) define a cognitive
mini-environment MBE for some subjects exercise E of cognitive powers;
(ii) specify the necessary conditions under which a mini-environment is
favorable with respect to E; and (iii) add the following resolution condition
to his theory of warrant:
(RC) A belief B produced by an exercise E of cognitive powers in
a cognitive mini-environment MBE has warrant (sufficient for
knowledge) only if MBE is favorable for E.5
In what follows I argue that there are counterexamples to Plantingas
most recent proposal for satisfying (ii). In addition, I argue that one way
Plantinga could respond to my criticism fails as well.6 I shall assume
Plantinga succeeds in meeting (i), i.e., he satisfactorily defines a cognit-
ive mini-environment MBE for some exercise E of an agents cognitive
powers. Roughly, Plantinga (1996, 314315) lets MBE be as large a frag-
ment of the actual world as you like. However, there is an exception: MBE
neither includes nor excludes Es producing a true belief.7 Motivating this
exception is the fact that if a subject has a true belief in a Gettier scenario,
then his E can surely be counted on to produce a true belief with respect
to an environment that includes Es producing a true belief.
To review, Plantinga hopes to deflect certain Gettier-style counter-
examples by refining his environment condition. A true belief has enough
434 THAD M. BOTHAM

warrant for knowledge only if an exercise E of cognitive powers produces

the belief in a mini-environment that is favorable with respect to E. Hence,
the remainder of this essay addresses this crucial question: in virtue of what
is a mini-environment MBE favorable with respect to a subjects exercise
E of cognitive powers?
Plantinga (2000, 160) observes that B is probable with respect to MBE
when the standard conditions for warrant have been met. Yet in certain
Gettier cases the cognizers E lacks sufficient resolution. Sometimes there
are epistemically relevant features of ones environment that are not detect-
able by E. Plantinga distinguishes between (a) states of affairs included in
MBE that are detectable, i.e., cognitively accessible, by E and (b) states of
affairs included in MBE that E cannot competently detect. Call the former
conjunctive state of affairs DMBE. For example, there being a clock, the
clocks reading 2:00, and the clocks being indoors are detectable states
of affairs in the stopped clock case. Similarly, in the barn case your E
can detect there being some cows feeding beside the structure you think is
a barn. In contrast, the clocks having stopped exactly twenty-four hours
ago and there being fake barns in the neighborhood are not cognitively
accessible states of affairs.
A worry surfaces: does the notion of detectability smuggle in know-
ledge, since (perhaps) a proposition p is detectable by S just if S can
readily know that p?8 If so, Plantingas account appears in trouble, since
this is prima facie circular. However, Im not convinced that this is a prob-
lem. A merchants scale detects its being the case that the goods weigh
400 pounds, but the scale cannot detect its being the case that a handful of
chocolates weighs three ounces. A thermostat detects its being the case that
the rooms ambient temperature is eighty degrees Fahrenheit but cannot
detect the temperature to four significant digits. A speedometer detects the
automobiles moving 50 miles per hour but cannot detect the precise rate
of 50.2 miles per hour. Even though the scale, thermostat, and speedometer
detect various states of affairs only when they function properly, they do
not have knowledge. Neither do they have anything close to justified or
warranted beliefs. Thus its not clear that detection entails anything as
strong as knowledge or justified belief. At any rate, I assume Plantinga
can sidestep this difficulty.
 PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS 435

Given the distinction between detectable and undetectable states of af-

fairs (with respect to an exercise of cognitive powers), Plantinga proffers
the following condition for a favorable mini-environment:

(FC) MBE is favorable [with respect to E] just if there is no state of

affairs S included in MBE but not in DMBE such that the object-
ive probability of B with respect to the conjunction of DMBE
and S falls below r, where r is some real number representing
a reasonably high probability.9

Notice that in each of the three Gettier cases outlined above the MBE turns
out unfavorable according to (FC); thus, your true belief does not have
warrant sufficient for knowledge. Hence, Plantingas canonical warrant
account together with (FC) and (RC) entails that you do not know the
propositions Its 2:00, Thats a fine barn, and Theres Paul. Why?
Consider the following candidates for S that when coupled with the relev-
ant DMBE renders the objective probability of your belief not reasonably
high:

The stopped clock case:

B1 = Its 2:00,

S1 = The clocks being broken.

The barn case:

B2 = Thats a fine barn,

S2 = There being many barn facades in the region.

The identical twin case:

B3 = Theres Paul,

S3 = There being Pauls identical twin Peter in the vicinity.

Nevertheless, (FC) is problematic, since it renders many intuitively favor-

able mini-environments unfavorable. Consider three new examples. First,
consider a version of the identical twin case where we stipulate that Paul
is an only child, thereby removing Peter from the scene. In addition, the
moment before you believe the proposition Theres Paul, his uncle
436 THAD M. BOTHAM

unbeknownst to you lies to a friend, telling her that Paul has an identical
twin brother whos visiting Paul at that very moment and that he just spoke
with each of them on the telephone. Pauls uncle asserts this falsehood
while in London, thousands of miles away. In this case, you know the
proposition Theres Paul.10 However, the objective probability of your
belief is not reasonably high given DMBE4 together with the state of affairs
S4 viz., its being the case that Pauls uncle just told a friend that Paul has
an identical twin who is presently in the vicinity of Paul. Thus, by (FC)
your MBE4 is unfavorable with respect to E4 , which together with (RC)
entails that you do not have knowledge. But you do know that it is Paul.
His uncles false assertion, spoken thousands of miles away, is irrelevant
to your particular epistemic circumstance. So (FC) is inadequate.
Second, consider again the barn case with the following alteration.
Rather than constructing life-sized barn facades, the locals manufacture
model barns so tiny that standard passersby cannot view them.11 Perhaps
the tiny barns are only two inches in height, built to suit ant communities.
As you drive through the heart of this anomalous territory, you see one
of the only real barns and believe the proposition Thats a fine barn. In
this miniature barn case, you clearly possess knowledge; however, there is
a state of affairs S5 there being many barn simulations in the region
that together with DMBE5 makes the objective probability of your belief
less than reasonably high. Thus, by (FC) your MBE5 is unfavorable with
respect to E5 , which when coupled with (RC) entails that you do not know.
But you do know that its a fine barn. So (FC) is inadequate.
Third, suppose you observe a Dalmatian by looking through a win-
dow into the backyard. You believe the proposition Theres a Dalmatian.
However, there is a state of affairs S6 being such that there are ten
mechanical Dalmatians in the backyard, each of which appears like a real
Dalmatian that together with DMBE6 renders the objective probability
of your belief less than reasonably high. Nonetheless, the owner locked all
of the robotic Dalmatians in a shed in his backyard to prevent them from
rusting. Indeed, theyve resided in the shed for about ten years, the lock is
rusted shut, and no one has known the keys whereabouts for at least two
years. So presumably you know that it is a Dalmatian, yet by (FC) your
MBE6 is unfavorable, which with (RC) entails that you do not know. Thus,
(FC) is false.
There are two ways to respond to the last three counterexamples against
(FC). One could deny that they are paradigm cases of knowledge, i.e.,
one might hold that they are merely borderline cases of knowledge. If the
cases are vague, they do not count much against Plantingas theory (unless
one also thinks that an adequate epistemological theory should reflect the
 PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS 437

vagueness inherent in such cases). Notice that this response assumes that
there are no clear cases of knowledge when there are any merely mislead-
ing states of affairs with respect to an exercise of cognitive powers. I find
this assumption somewhat implausible but will not argue against it here.
The alternative response accepts that the case of either the dishonest
uncle, the miniature barns, or the mechanical Dalmatians is a clear enough
instance of knowledge. If this is right, then salvaging Plantingas theory
requires modifying (FC) to cope with what might be considered misleading
states of affairs that are candidates for S. Such a person might be tempted
to put forward the following favorability condition:12

(FC ) MBE is favorable with respect to E only if for every state of af-
fairs S included in MBE but not in DMBE such that the objective
probability of B with respect to the conjunction of DMBE and
S falls below r, there is another state of affairs S such that: (1)
MBE includes S ; (2) the conjunction of S and DMBE does
not entail B; (3) S includes S; and (4) it is not the case that
the objective probability of B with respect to the conjunction of
DMBE and S falls below r.13
(FC ) copes well with the last three cases introduced against (FC) above,
as illustrated by the following relevant assignments where you intuitively
know the belief in question:

The dishonest uncle case:

B4 = Theres Paul,

S4 = Its being the case that Pauls uncle just told a friend that
Paul has an identical twin who is presently in the vicinity of
Paul,
S4 = S4 and its being the case that Pauls uncle is lying.

The miniature barn case:

B5 = Thats a fine barn,

S5 = There being many barn simulations in the region,

S5 = S5 and its being the case that all of the barn simulations
are about two inches in height.
The Dalmatian case:

B6 = Theres a Dalmatian,
438 THAD M. BOTHAM

S6 = There being ten mechanical Dalmatians in the backyard,

each of which appears like a real Dalmatian,

S6 = S6 and its being the case that the owner keeps all of the
robotic Dalmatians locked in a shed in his backyard to prevent
them from rusting, etc.

The crucial point is that even though the objective probability of B is not
reasonably high with respect to the conjunction of S and DMBE, it is on
the conjunction of S and DMBE. Furthermore, in typical (non-Gettier)
cases of knowledge, (FC ) will vacuously hold since there will not be any
S satisfying the conditionals antecedent. So far so good.
Nevertheless, (FC ) now fails to handle the Gettier cases for which
(FC) was originally designed. In brief, (FC ) does not imply that intuit-
ively unfavorable mini-environments are unfavorable. Recall the first three
cases where you do not know the true belief in question. Now adjust the
background information just enough to fit the following assignments:

The stopped clock case:

B1 = Its 2:00,

S1 = The clocks being broken,

S1 = S1 and its being the case that if you were to request the
time from three nearby, random strangers who just looked at
their wristwatches, they all would report that it is 2:00.

The barn case:

B2 = Thats a fine barn,

S2 = There being many barn facades in the region,

S2 = S2 and its being the case that of all the barn-like structures
only the fake barns face the road.14

The identical twin case:

B3 = Theres Paul,

S3 = There being Pauls identical twin Peter in the vicinity,

 PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS 439

S3 = S3 and Peters remaining in the foyer as he puts on his

coat.
These assignments clearly illustrate the inadequacy of (FC ). Since each
example is a standard Gettier case, the belief in each example is accident-
ally true. And since the belief is true, there are states of affairs that make
it true and there are states of affairs with respect to which the belief is
nontrivially probable. The important point is that even though the objective
probability of B on both DMBE and S is not reasonably high, there is (I
think) always a larger state of affairs S that both includes S and when
coupled with DMBE renders the objective probability of B sufficiently
high.15
For example, consider the identical twin case. You do not know it is
Paul, since Peter could just as well have emerged from the house before
Paul did. Notice that the objective probability of the proposition Theres
Paul given both DMBE3 and the state of affairs S3 , there being Pauls
identical twin Peter in the vicinity, is not reasonably high. However, there
is a larger state of affairs S3 viz., there being Pauls identical twin Peter
in the vicinity and Peters remaining in the foyer as he puts on his coat
that together with DMBE3 renders the objective probability of the pro-
position Theres Paul reasonably high. Thus, since (FC ) does not imply
that certain mini-environments are unfavorable as it should, (FC ) does not
sidestep the Gettier cases raised against Plantingas warrant account.
In sum, Ive argued that (FC) is too restrictive because it labels some
intuitively favorable mini-environments as unfavorable, and (FC ) is too
lenient, not branding intuitively unfavorable mini-environments unfavor-
able. Thus, sidestepping Gettiers stranglehold by specifying the necessary
conditions for a mini-environments favorability requires finding some
(FC ) that falls between (FC) and (FC ).

ACKNOWLEDGEMENTS

Special thanks to Thomas Crisp, Marian David, Alvin Plantinga, Warren

Shrader, and two anonymous referees for helpful advice.

NOTES

1 For a more precise rendition of the necessary and sufficient conditions for warrant, see
Plantinga (1993b, 194); hereafter WPF.
2 See Feldman (1996), Klein (1996), and Swain (1996).
440 THAD M. BOTHAM

3 Of course the criticism need not rely on the apparatus of possible worlds; one could talk
of ways things might go that are sufficiently similar to the original examples.
4 Recall that warrant comes in degrees, hence the locution enough warrant for know-
ledge.
5 Plantinga (1996, 328; 2000, 159). The parenthetical in (RC) is designed leave open the
possibility that beliefs produced in an unfavorable mini-environment have some degree of
warrant, albeit a degree insufficient for knowledge. For a more complete context of the
present discussion especially concerning what motivates the distinction between maxi-
environments and mini-environments see Plantinga (1996, 313317, 326329; 2000,
156161) and Crisp (2000).
6 It might be worth noting that one can find Plantingas first attempt to meet (ii) in his
(1996, 328). However, in response to difficulties advanced by Thomas Crisp (2000, 46
47), Plantinga (2000, 160) offers a second approximation. Incidentally, Crisp (2000) offers
a solution to both (i) and (ii) of his own, and I want to register my conviction that I think
Crisp at least succeeds in satisfying (ii). Nonetheless, my essay focuses on the adequacy of
Plantingas own specification of mini-environmental favorability.
7 Say a state of affairs A includes a state of affairs B just in case that, necessarily, if A
obtains then B obtains. Say that a state of affairs A entails a proposition p if and only if it
is not possible (in the broadly logical sense) that A obtain and p fail to be true.
8 Marian David brought this worry to my attention.
9 Clearly the following condition should be added: the conjunction of S and DMBE does
not entail B. More perspicuously, MBE is favorable for E iff S{(MBE includes S) &
(DMBE includes S) & ((S&DMBE) entails B) & Pr(B/(DMBE&S)) < r}.
10 This example is a version of the Grabit case first advanced by Keith Lehrer and Thomas
Paxson (1969).
11 I owe this example to Thomas Crisp and the next example, the Dalmatian case, to Donald
Smith.
12 Plantinga (personal conversation, Spring, 2001) is not wholly convinced that these are
clear cases of knowledge. However, he suggests that if one were moved by the counter-
examples against (FC), then it might be fixed with something in the neighborhood of
(FC ).
13 More perspicuously, MBE is favorable for E only if S{((MBE includes S) & (DMBE
includes S) & Pr(B/(DMBE&S)) < r) S ((MBE includes S ) & ((DMBE&S )
entails B) & (S includes S) & Pr(B/(DMBE&S )) r)}.
14 Notice the original barn example did not include the fact that of all the barn-like struc-
tures only the fake barns face the road. Nonetheless, we can just stipulate that they do and
then adjust MBE2 and DMBE2 accordingly. Or perhaps of all the barn-like structures only
the fake barns happen to be obscured by thick fog. Then we have MBE2b , DMBE2b , and
the relevant assignment: S2b = S and its being the case that of all the barn-like structures
2b
only the fake barns are hidden by thick fog.
15 For the reader steeped in post-Gettier epistemology, the above discussion no doubt re-
minds her of how various contemporary epistemologists have used the notions of epistemic
defeaters and misleading defeaters in order to deal with Gettier-style counterexamples.
Plantingas (RC) together with (FC) is clearly analogous with standard no-defeater
conditions. Similarly, (FC ) basically amounts to a no-misleading-defeater condition.
 PLANTINGA AND FAVORABLE MINI-ENVIRONMENTS 441

REFERENCES

Crisp, T.: 2000, Gettier and Plantingas Revised Account of Warrant, Analysis 60, 4250.
Feldman, R.: 1996, Plantinga, Gettier, and Warrant, in J. Kvanvig (ed.), Warrant in
Contemporary Epistemology: Essays in Honor of Plantingas Theory of Knowledge,
Rowman and Littlefield, London, pp. 199220.
Klein, P.: 1996, Warrant, Proper Function, Reliabilism, and Defeasibility, in J. Kvanvig
(ed.), Warrant in Contemporary Epistemology: Essays in Honor of Plantingas Theory
of Knowledge, Rowman and Littlefield, London, pp. 97130.
Kvanvig, J. (ed.): 1996, Warrant in Contemporary Epistemology: Essays in Honor of
Plantingas Theory of Knowledge, Rowman and Littlefield, London.
Lehrer, K. and T. Paxson: 1969, Knowledge: Undefeated Justified True Belief, Journal
of Philosophy 66, 225237.
Plantinga, A.: 1993a, Warrant: The Current Debate, Oxford University Press, Oxford.
Plantinga, A.: 1993b, Warrant and Proper Function, Oxford University Press, Oxford.
Plantinga, A.: 1996, Respondeo, in J. Kvanvig (ed.), Warrant in Contemporary Epistem-
ology: Essays in Honor of Plantingas Theory of Knowledge, Rowman and Littlefield,
London, pp. 307378.
Plantinga, A.: 2000, Warranted Christian Belief, Oxford University Press, Oxford.
Swain, M.: 1996, Warrant versus Indefeasible Justification, in J. Kvanvig (ed.), Warrant
in Contemporary Epistemology: Essays in Honor of Plantingas Theory of Knowledge,
Rowman and Littlefield, London, pp. 131146.

T. M. Botham
University of Notre Dame
Department of Philosophy
Malloy Hall 100
Notre Dame, IN 46556-4619
U.S.A.
Phone: 219.287.9230
E-mail: tbotham@nd.edu