Synthese (2008) 160:215–228

DOI 10.1007/s11229-006-9108-y

O R I G I NA L PA P E R

Haack’s foundherentism is a foundationalism

Peter Tramel

Received: 26 September 2005 / Accepted: 30 August 2006 / Published online: 2 May 2007
© Springer Science+Business Media B.V. 2007

Abstract Susan Haack has always maintained that her unquestionably important
foundherentist theory of epistemic justification is not a foundationalism. In a 1997
Synthese exchange, Laurence BonJour questioned her right to this claim, and she dug
in and defended it. What was at stake is of timeless importance to epistemology: it
goes directly to the question, “What is foundationalism?” I inquire with greater care
than either Haack or BonJour took in 1997, and I find decisively in favor of the view
that foundherentism is a foundationalism. In the process, I explore the outer limits
of foundationalism: I examine just how far a foundationalism can go in allowing the
relevance of coherence to epistemic justification.

Keywords Haack, Susan · BonJour, Laurence · Foundationalism · Foundherentism ·

Coherentism · Epistemic justification

Suppose I believe that the animal I see rooting in my backyard is a groundhog. How
justified am I in believing this? According to a theory proposed by Susan Haack,
that depends (in part) on (a) how favorable my evidence is, i.e., how well the content
of my belief fits both what I see and my other relevant beliefs, (b) how secure my
reasons are, i.e., how justified my other relevant beliefs are, independently of their
connection with the one in question, and (c) how comprehensive my evidence is, i.e.,
(roughly) how well my evidence handles the groundhog-sighting contingencies that I
would have to take into account in order to be maximally justified.1
She derives these criteria by analogy with how we do crossword puzzles. When we
do crossword puzzles we consider (a*) how well potential answers fit both their clues
and already filled-in, intersecting answers, (b*) how well already filled-in, intersecting

1 Haack (1993, p. 82).

P. Tramel (B)
Department of English & Philosophy, United States Military Academy,
607 Cullum Rd., West Point,
10996 NY, USA
e-mail: peter.tramel@usma.edu
216 Synthese (2008) 160:215–228

answers fit both their clues and their already filled-in, intersecting answers, and (c*)
how many of the intersecting answers are already filled in.2
She opposes this crossword puzzle picture of the structure of epistemic justification
to the traditional foundationalist pyramid and coherentist raft, and she finds that her
view is unique, intermediate between and distinct from foundationalism and coher-
entism. She calls it “foundherentism,” a name she admits is ugly. I will not dispute
that it is unique (or that its name is ugly) but I will argue that it is a species of founda-
tionalism after all. In doing so I follow up on a suspicion voiced by Laurence BonJour
in his 1997 Synthese review of Haack’s Evidence and Inquiry, and I respond to her
published reply to BonJour.3

1 Haack’s original argument

In this section I introduce Haack’s conception of foundationalism and I summa-

rize her original (1993) argument for the claim that her foundherentism is not a
foundationalism.
She defines foundationalism as the view that:
(FD1) Some justified beliefs are basic; a basic belief is justified independently
of the support of any other belief;
and:
(FD2) All other justified beliefs are derived; a derived belief is justified via the
support, direct or indirect, of a basic belief or beliefs.4
Foundationalism thus requires a distinction between basic and derived beliefs. The
role of this distinction in the definition entails that justification is one-directional: it
moves from basic to derived beliefs but never vice versa.
Her original argument for the claim that her foundherentism is not a foundational-
ism, which I will call the Dependence Types argument, can be summarized as follows.
Consider the following four types of epistemic dependence relations among beliefs:
(i) a derived belief owes justification to a basic belief or beliefs, (ii) a derived belief
owes justification to another derived belief or beliefs, (iii) a basic belief owes justifi-
cation to another basic belief or beliefs, and (iv) a basic belief owes justification to
a derived belief or beliefs. Foundationalism is compatible with, at most, epistemic
dependence types (i) and (ii), since (FD1) rules out dependence types (iii) and (iv).
Foundherentism, on the other hand, is consistent with pervasive relations of mutual
support. That is, as her crossword puzzle model suggests, a belief anywhere in a found-
herentist structure of justified beliefs can owe justification to beliefs anywhere else in
the structure. Hence, foundherentism is not a foundationalism.5
2 Ibid.
3 Her reply to BonJour, Haack, 1997, was published in the same number of Synthese as his review.
4 Ibid: 14.
5 This is a summary of her more detailed version in Chapter 1 of Evidence and Inquiry; but all that
the more detailed version adds is a classification scheme for foundationalisms, including character-
izations of a number of foundationalism types. These details have no bearing on the soundness of
the argument: the point is always that foundationalism is limited to what I am calling dependence
relation types (i) and (ii)—indeed, this assumption is necessary to the whole classification scheme
(p. 14)—foundherentism, on the other hand, is not limited to types (i) and (ii): it is consistent with
types (iii) and (iv). So foundherentism is not a foundationalism (pp. 19–20).
Synthese (2008) 160:215–228 217

2 BonJour’s objections

Doubting the soundness of what I am calling the Dependence Types argument,

BonJour wonders whether the best version of her theory will not turn out to be
foundationalist. He suggests that her argument both underestimates foundationalism
and overestimates foundherentism.
First, he points out that some “weak” forms of foundationalism are particularly
designed to allow basic beliefs to owe some part or kind of their justifications to other
basic, and even derived, beliefs. He writes
The more or less standard conception of weak foundationalism, after all, is one
in which basic beliefs have some relatively weak initial degree of justification,
which is then enhanced by something like coherence to a level sufficient for
knowledge.6
This is why he suspects that her argument underestimates foundationalism.
Second, even foundherentism explicitly rules out that beliefs can get support solely
from their relations to other beliefs. Indeed, this is the basis of her claim that foundher-
entism is not a coherentism. But “. . . this conclusion makes it doubtful that non-basic
beliefs have any support to give each other that is not ultimately derived from basic
beliefs.”7 This is why he suspects that her argument overestimates foundherentism.
Below I follow through on both of BonJour’s worries—first, that Haack under-
estimates foundationalism, and, second, that she overestimates foundherentism—to
achieve the conclusion that her foundherentism is a foundationalism. After a few
essential preliminaries in the next section, I argue in Sects. 4 and 5 that BonJour’s
weak foundationalism—I will follow Haack and call it “feeble”—is indeed a founda-
tionalism; and I argue in Sects. 6 and 7 that Haack’s foundherentism is just such a
foundationalism.

3 Fun with definitions

Haack calls BonJour’s “weak” foundationalisms “feeble,” since she used “weak foun-
dationalism” for another, unrelated purpose in the work at issue.8 Let’s go with feeble
and agree that by feeble foundationalism we mean foundationalism that is consistent
with basic beliefs owing some of their justification to other beliefs. In this section, I
argue that her original, stated definition of foundationalism is consistent with feeble
foundationalism, although she did not intend this.
Her original definition (quoted in Sect. 1) requires that “. . . a basic belief is justified
independently of the support of any other belief.” This requirement could be taken
in either of two ways. It could mean that basic beliefs can owe no justification to the
support of other beliefs: call this the isolationist reading. Or it could mean only that
basic beliefs must have some justification that they do not owe to the support of other
beliefs: call this the cosmopolitan reading.

6 BonJour (1997, pp. 16–17).

7 Ibid.
8 Haack (1997, p. 26). (It is hard to resist splitting the difference and calling them “weeble” foun-
dationalisms, since we could make some sense of the idea that they “. . . wobble but they don’t fall
down.”)
218 Synthese (2008) 160:215–228

As it happens, we can restate both readings of her original definition, preserving its
locutions and form, and wind up with a serviceable characterization of the distinction
between feeble and non-feeble foundationalism.
Non-Feeble Foundationalism is the view that:
(FD1I ) Some justified beliefs are basic; a basic belief is justified and owes no
justification to any other belief.
and
(FD2) All other justified beliefs are derived; a derived belief is justified via the
support, direct or indirect, of a basic belief or beliefs.
And:
Feeble Foundationalism is the view that:
(FD1C ) Some justified beliefs are basic; a basic belief possesses justification that
it does not owe to any other belief.
and
(FD2) All other justified beliefs are derived; a derived belief is justified via the
support, direct or indirect, of a basic belief or beliefs.
and
(FD3) All justification in a structure of justified beliefs derives, either directly
or indirectly, from the justificatory support of basic beliefs.
FD1I is inconsistent with the possibility that basic beliefs owe justification to other
beliefs, basic or derived; and FD1C is not. This is why we may call the former non-
feeble and the latter feeble. The extra condition, FD3, is implicit in the definition
of non-feeble foundationalism, but it needs to be made explicit in the definition of
feeble foundationalism, in order to preserve the same kind of recursive closure in
both definitions.
Although her original, stated definition of foundationalism is compatible with
either of these readings, we can see that she intended the non-feeble version, since the
feeble version would not support her confidence in the Dependence Types argument,
which turned on foundationalism being incompatible with dependence types (iii) a
basic belief owes justification to another basic belief or beliefs, and (iv) a basic belief
owes justification to a derived belief or beliefs.
Despite her original intention, she may later have become willing to concede that
feeble foundationalism is a foundationalism: at least she concedes it for the sake of
argument in her reply to BonJour.9 However that may be, in the next two sections I
argue that she has no right not to concede it.

4 Feeble foundationalism is a foundationalism, a prima facie case

Since Haack concedes to BonJour that feeble foundationalism is a foundationalism,

it might seem strange that I am now going to argue for this conclusion in detail. Why
look a gift-concession in the mouth? My reasons are threefold. First, she makes the

9 Haack (1997, p. 27).

Synthese (2008) 160:215–228 219

concession only for the sake of argument; she may not think that she must make
it.10 Second, as I explain below in Sect. 5, one can doubt that she fully grasps what
she concedes. Finally, the details of this argument will matter as I go on to argue in
Sects. 6 and 7 that her foundherentism is a feeble foundherentism.
My argument of this section, i.e., my prima facie case, has the following general
form.
1. If a theory of justification stops all dependence regresses of justification—e.g.,
belief A owes justification to belief B, which B in turn owes to belief C, and so
on—with basic beliefs, then it is a foundationalism.
2. Non-feeble foundationalisms stop all dependence regresses of justification with
basic beliefs.
3. If non-feeble foundationalisms stop all dependence regresses of justification with
basic beliefs, then so do feeble foundationalisms.
4. So feeble foundationalism is a foundationalism.
In this section I argue for 2 and 3; I take up 1 in the next section.
Non-feeble foundationalism straightforwardly and abruptly stops all dependence
regresses with basic beliefs. Basic beliefs can generate no dependence regresses be-
cause they owe no justification to other beliefs; and, since derived beliefs must ulti-
mately owe all of their justification to basic beliefs, the dependence regresses they
generate must always stop with basic beliefs.
Feeble foundationalism, on the other hand, is consistent with basic beliefs that
generate dependence regresses. Although it requires basic beliefs to have some jus-
tification that they do not owe to other beliefs, it also allows them to have additional
justification that they do owe to other beliefs. A basic belief, p, might owe part of
its justification directly to experience, call this part
, and another part, call it , to
other beliefs. With respect to
p does not generate a regress but with respect to  it
generates one. So feeble foundationalism’s cosmopolitan conception of basic beliefs
(FD1C ), unlike its non-feeble, isolationist counterpart (FD1I ) fails to guarantee, all
by itself, that basic beliefs stop all regresses; it only guarantees that basic beliefs stop
some of them.
Since feeble foundationalism allows basic beliefs to generate dependence regresses,
we need FD3 to ensure that they all stop with basic beliefs; i.e., to get the same recur-
sive closure as the definition of non-feeble foundationalism. Perhaps p owes part of its
justification  to another basic belief, q, and the buck stops there. Or perhaps p owes
 to a derived belief, r, and r owes it, directly or indirectly, to another basic belief, s.
Feeble foundationalism implies no limit to how much complexity of this sort of belief
interdependence structures can have. Non-feeble foundationalism is consistent with
all of the same sorts of belief interdependence, but only in the superstructure; i.e.,
only among derived beliefs.
FD3 stops the regress; but does feeble foundationalism have a right to add FD3?
I assume that if non-feeble foundationalism has a right to the general way that it
stops regresses, then feeble foundationalism has a right to stop them in the same
general way. FD3 is no more than a general description of the recursive function
by which non-feeble foundationalism stops dependence regresses of justification; so
feeble foundationalism is entitled to FD3 if non-feeble foundationalism is. Since both
views stop the regress in the same way, if non-feeble foundationalism is a foundation-
alism, then so is feeble foundationalism.
10 Her reservations upon granting it suggest that she doubts it very much. See her 1997: 26–28.
220 Synthese (2008) 160:215–228

5 Feeble foundationalism is a foundationalism: Confirmed

In this section I complete my argument for the prima facie case by arguing for its first
premise, and, in the process, I respond to the only two potential objections to it I find
in Haack’s work.
Haack’s characterization of foundationalism is idiosyncratic in its inattention to
the regress problem (of justification): the problem of how dependence regresses stop.
John Greco first drew attention to this idiosyncrasy and accordingly, as BonJour
would later do, wondered whether foundherentism is not really a foundationalism.11
The common thread, perhaps the only common thread, in foundationalism from Aris-
totle to the latest definitions in encyclopedias and dictionaries of philosophy, is this:
foundationalism solves the regress problem with what we are calling basic beliefs; i.e.,
by stopping every (doxastic) dependence regress with justification that beliefs have
and do not owe to other beliefs.
She knows that foundationalism is committed to basic beliefs, and she conceives
basic beliefs such that they solve the regress problem; however, she does not conceive
them, as is usual, in terms of how they solve the regress problem. Feeble and non-fee-
ble foundationalism, despite their important differences, use basic beliefs in the same
way, qua solutions to the regress problem: they both stop all dependence regresses
with basic beliefs. Why, then, not call them both foundationalisms? How can one jus-
tify cutting off the feeble? I find only two answers in her work, and neither one holds
water. First, she thinks that foundationalism is committed to a “one-directionality”
requirement that rules out feeble foundationalism; second, she thinks that hardly any
foundationalists find feeble foundationalism tempting. I argue that on both counts
she is wrong.
By way of introducing her original characterization of the relevant sense of foun-
dationalism (quoted in Sect. 1 above), she writes:
. . . ‘foundationalism’ will refer to theories of justification which require a dis-
tinction, among justified beliefs, between those which are basic and those which
are derived, and a conception of justification as one-directional, i.e., as requiring
basic to support derived beliefs, never vise versa.12
Despite her “. . . will refer to . . .” she is not just stipulating; for in that case the
truth conditions of her claim to have found an alternative to foundationalism and
coherentism—an alternative she conceives as a great new hope for traditional episte-
mology—would be no more than rules for a private language.13 We may assume that
is not what she intends. So we may reject any of these requirements for which there is
no good reason, especially if there are good reasons to reject them.
We may, consistently with my prima facie case, concede all of her conditions except
the “never vise versa” (NVV) part of her one-directionality condition. However, if
she has a right to NVV, then we are in trouble. For then there can be no feeble
foundationalism, since feeble foundationalism is consistent with ‘vise versa’, i.e., it is
consistent with basic beliefs owing justification to derived beliefs. What, then, does she
have to say for NVV? Originally, nothing (although she appeals to it early and often).
In her response to BonJour she tries to motivate NVV in a footnote by citing two

11 Greco (1996).
12 Haack (1993, p. 14).
13 For these claims as to the significance of her view, see her 1993: 1–9.
Synthese (2008) 160:215–228 221

authorities: William Alston and Ernest Sosa. I doubt that they intended her NVV; but
let’s suppose that they did.14 So what? She cites no arguments for NVV from them,
and when we visit the articles by them that she cites we find none.
So she gives us no relevant reason, almost no reason at all, for NVV. Yet the reasons
to reject it are strong. First, what is NVV besides a denial of the possibility of feeble
foundationalism? Nothing. As a consideration against feeble foundationalism it is
therefore question-begging, as a necessary condition for foundationalism it is ad hoc.
Second, conceiving foundationalism in the traditional way, i.e., as a certain kind of
solution to the regress problem, makes the foundationalist/non-foundationalist divide
a logical chasm: the distinction pulls real conceptual freight. Why shave off only a bit
of the former and consider that bit no man’s land? It is, after all, a mere bit of putative
foundationalism that NVV shaves off: NVV is consistent with a quasi-feeble foun-
dationalism, one that allows basic beliefs to owe justification to other basic beliefs
(and thus it does not require non-feeble foundationalism); it only rules out that basic
beliefs can owe justification to derived beliefs. But if you allow basic beliefs to owe
justification to other basic beliefs, why prohibit their owing justification to derived
beliefs that owe it, in turn, to other basic beliefs?
The only potentially relevant thing she says about this is that almost no one defends
a feeble foundationalism. According to her, no one but Roderick Firth and maybe
Bertrand Russell ever entertained the possibility of feeble foundationalism, and nei-
ther of them paid it more than passing attention.15 Of course the issue is not the
popularity of feeble foundationalism. But maybe she means to exclude feeble foun-
dationalism as uncommon usage. That does not seem like a very strong thing to oppose
to the usual distinction in terms of logical space; but we do not have to explore that
because she is wrong. No less a Roderick than Chisholm himself, who thought (at
least by the third edition of his Theory of Knowledge) that the justification of basic
beliefs can be improved upon through coherence with other justified beliefs, held what
we are calling “feeble foundationalism”.16 Many, if not most, foundationalists since
Chisholm have held feeble foundationalisms.17 She seems not to get foundationalism

14 She quotes Alston and Sosa (her 1997: 34, n.3), to support NVV, as follows:

Alston: “foundationalism is a view concerning the structure of justified beliefs possessed by

a given individual. Such a system is divided into ‘foundation’ and ‘superstructure’, so related
that beliefs in the latter depend on the former for their justification but not vice versa” (Alston,
1992, p. 144, second italics mine [Haack’s]). Sosa: according to foundationalism “every piece
of knowledge stands at the apex of a pyramid that rests on stable and secure foundations
whose stability and security does not derive from the upper stories” (Sosa, 1980, pp. 23–24, my
[Haack’s] italics).
Why does she think that Alston’s dependence claim is not merely ultimate, and thus consistent with
feeble foundationalism, which requires that all justification enters the system at the basic belief level?
One finds nothing in Alston 1992 inconsistent with that interpretation. Like Sosa she later makes uses
of “stability” and “security,” as we will observe in Sect. 8, but even if Sosa’s uses of these are hers
there is no reason to think that they are inconsistent with feeble foundationalism (especially since, as
I argue in Sects. 7 and 8, her uses of them commit her to feeble foundationalism).
15 Haack (1997, p. 27–28).
16 Chisholm (1989, pp. 61–74).
17 She knows enough about Robert Audi’s foundationalism (Audi, 1993) to make fun of it in a
parenthetical (Haack, 1997: 29–30), yet she fails to notice that the whole book she mocks is an artic-
ulation and defense of a version of what we are calling feeble foundationalism. Of course Audi, like
Chisholm, does not call it “feeble”: he calls it modest, and later, moderate, foundationalism. Even
some coherentist critics of foundationalism during the last 20 years tend to think that the feeble are
the most plausible foundationalisms: (e.g., Geoffrey Sayre-McCord, 1996, pp. 137–189).
222 Synthese (2008) 160:215–228

since Chisholm. No wonder she adopts a foundationalism advanced by C.I. Lewis in

1947 as her paradigm of foundationalism.18

6 Haack’s foundherentism is a feeble foundationalism, a prima facie case

Haack’s foundherentism is what she calls a “dual-aspect” theory of epistemic justifi-

cation. That is, it implies conditions for the justification of beliefs both qua states and
qua contents.19 I will argue that both of its aspects entail feeble foundationalism.
Recall that a key component of BonJour’s original objection was that Haack over-
estimates the resources of foundherentism in the following way. She thinks that found-
herentism allows too large an epistemic role to coherence for any foundationalism
to handle; however, since she also denies that mere mutual support among beliefs,
e.g., mere coherence, can have any epistemic force on its own (this, after all, is why
she rejects coherentism) she cannot really allow coherence any role that feeble foun-
dationalism cannot allow. So foundherentism is a foundationalism. This is the prima
facie case.
To put this another way, her reason for rejecting coherentism commits her (in
context20 ) to the foundationalist solution to the regress problem; but for a view to
be committed to the foundationalist solution to the regress problem implies that it
is a foundationalism. Hence, her foundherentism is a foundationalism. Her Depen-
dence Types argument proves that foundherentism is not a non-feeble foundational-
ism (or a merely quasi-feeble foundationalism); therefore foundherentism is a feeble
foundationalism.
Against BonJour’s version of this, Haack argues that foundherentism must allow
too much epistemic significance to coherence for foundationalism to handle; for found-
herentism admits of no basic/derived belief distinction. She writes:

What he [BonJour] seems to have lost hold of is that no form of foundational-

ism can accommodate the mutual interpenetration, the quasi-holism, of beliefs,
which is the real insight of coherentism; for if one takes the coherentist insight
seriously, one realizes that no distinction of basic and derived beliefs, such as all
forms of foundationalism require, is viable.21

Apparently, she is saying that this “mutual interpenetration, . . . quasi-holism, of

beliefs” is something greater than the Dependence Types argument suggests; it breaks
the bounds of any distinction between basic/derived beliefs altogether.22

18 Her 1993, Chapter 2 (34–51), entitled, “Foundationalism Undermined”, is no more nor less than
an examination and rejection of Lewis’s 1947 view.
19 Haack (1993, pp. 73–74).
20 i.e., since she also rejects every other non-foundationalist solution to the regress problem. See her
1993: 21–31.
21 Haack (1997, p. 26).
22 Her idea that foundationalism is inconsistent with “the mutual interpenetration, etc.” of doxastic
justification, in conjunction with her apparently rationale-less “one-directionality” requirement for
foundationalism (see Sect. 5 above), tempts me to speculate. Foundationalism is of course named for
the architectural metaphor with which it is often introduced. Owing justification, in the metaphor, is
like a brick depending for support against gravity upon another brick. One has to be careful, though,
not to carry the analogy too far: structures of justified beliefs, for instance, do not have south and
north ends. Perhaps she carries the analogy too far, as follows. Take a pyramid. Bricks on the bottom
Synthese (2008) 160:215–228 223

The problem with this argument is that the basis of her anti-coherentism, i.e., the
basis of the prima facie case, ensures that this “mutual interpenetration” etc., can-
not break the bounds of a distinction between basic/derived beliefs. For this basis
entails that if there are any justified beliefs then there are some beliefs that possess
justification that they do not owe to any other beliefs; and thus it entails the feeble
foundationalist basic/derived belief distinction.
Indeed, we can find the feeble foundationalist basic / derived belief distinction doing
real work in her own account of foundherentism: we can pinpoint its explicit existence
in both her account of the justification of beliefs qua states, i.e., S-justification, and
in her account of the justification of beliefs qua contents, i.e., C-justification.23 This is
relatively easy to show in the case of S-justification; trickier, but more interesting, in
the case of C-justification.

6.1 The case of S-justification

Let’s take the case of S-justification first. To block coherentism in S-foundherentism,

she requires that
Experiential S-evidence evidentially sustains/inhibits S-beliefs, but not vice-
versa.
A’s experiential S-evidence is, one might say, his ultimate evidence.24
In other words, all S-justified beliefs must owe all of their positive S-justification to
justifying non-belief sources, i.e., must be caused by experience or caused by beliefs
that were caused by experience, etc.
Notice that this implies the S-equivalent of (FD3) in the definition of feeble foun-
dationalism. Moreover, it is easy to conceptualize a difference between S-justified
beliefs directly and indirectly sustained by what she calls “experiential S-evidence:”
the former—call them S-basic beliefs—derive some of their S-justification directly
from experiential states rather than other beliefs, and the latter—call them S-derived
beliefs—derive none of their justification directly from experiential states but instead
derive all of their justification from other beliefs, all of which are, or ultimately derive
all of their justification from, S-basic beliefs. Hence, her anti-S-coherentism commits
her to an S-aspect distinction between basic and derived beliefs.

6.2 The case of C-justification

Her picture of the structure of C-justification is the crossword puzzle picture I

mentioned in the beginning. She writes:

Footnote 22 continued
support bricks above, but never vice versa: this would metaphorically translate into her
one-directionality requirement. Perhaps the bottom bricks could, in addition to being supported
by the ground, lean on one another a bit (quasi-feeble foundationalism), but they could not lean on
the ones above them (like in feeble foundationalism, or foundherentism). But if we back up and ask
what is being compared in the analogy, we must see that, on the foundationalist side, it is a certain kind
of solution to the regress problem, a kind which has no features that bear out the comparison as far as
she takes it. If this is what she does (I am only speculating), then she goes wrong by putting an analogy
ahead of what is being compared, like someone who would infer from the foundationalist metaphor
that there must be south and north beliefs. Perhaps we could cure this by pointing out old-fashioned
arches, in which the keystones are held in place by the bricks beside and above them, and vice versa.
23 Haack (1993, pp. 73–74).
24 Haack (1993, p. 77).
224 Synthese (2008) 160:215–228

. . . how good A’s C-evidence with respect to p is would depend on;

1. how favorable A’s direct C-evidence with respect to p is;
2. how secure A’s direct C-reasons with respect to p are, independently of the
C-belief that p;
3. how comprehensive A’s C-evidence with respect to p is.
It should be noted that, although clause 2 mentions explicitly only A’s direct
C-reasons with respect to p, its application takes one progressively outward, to
the appraisal of A’s indirect1 , indirect2 . . .etc., C-evidence with respect to p.25
As I explain below, the crossword puzzle analogy interpretation of this creates
an appearance of a foundherentism that is not a foundationalism, perhaps not even
a feeble foundationalism. However, I argue that when we examine this account of
C-justification, in light of her view of the connection between S- and C-evidence, this
appearance turns out to be an illusion.
First, the appearance. In the analogy, the support that puzzle answers get from their
clues is like the support beliefs get directly from relevant experience; the support that
puzzle answers get from fitting intersecting, filled-in answers is like the support beliefs
get from coherence with other relevant beliefs. Now Haack denies coherentism, which
she defines as the view that relations of mutual support can alone evidentially sup-
port beliefs.26 She can avoid it in 1 and 2 by, first, requiring favorability for “good”
C-evidence, and requiring fit with experience for favorability. The requirement gets
reiterated at every step in the “progressively outward” moving appraisals that we do
when we assess security, since they will require assessments of favorability in every
case. Thus, fit with filled-in answers provides evidential support only if those filled-in
answers fit their clues, and their fit with further filled-in answers provides them further
evidential support only if the further filled-in answers fit their clues, and so on. So we
avoid coherentism: we never get a case of coherence providing evidential support on
its own. Nevertheless, the support coherence gives seems not to reduce to the requisite
fit with the clues/experience, since favorability is a matter of both fit with clues and
fit with filled-in intersecting answers, with no reduction of the role of the latter to
the role of the former. Since there is no such reduction, it does not seem like we can
capture the justificatory role of coherence by tracing all justificatory support back to
fit with the clues, as foundationalism, even the feeble foundationalist (FD3), requires.
So the analogy may tempt us to think that there is a middle way between coher-
entism and even the feeble foundationalist (FD3). However, at least in the case of
her foundherentism, the apparent middle way turns out to be an illusion. To see this,
we must examine what her distinction between direct and indirect C-evidence really
is—and not just what the analogy suggests that it is—in light of her way of connecting
S- and C-justification.
The analogy tempts us to understand the distinction between direct and indirect
support in strictly crossword puzzle terms. An answer to 1 down may owe direct
support to both its clue and to fit with the filled-in, intersecting answer to 2 across.
It may owe indirect1 support to the support that the filled-in answer to 2 across gets
from intersection with the filled-in answer to 3 down; indirect2 support to the sup-
port 3 down gets from intersection with the filled-in answer to 4 across; and so on.
This analogy-driven way of distinguishing directness/indirectness leaves no room for
application of the distinction inside of judgments of favorability; thus it supports the
25 Haack (1993, p. 82).
26 Haack (1993, pp. 18–19; 53–72).
Synthese (2008) 160:215–228 225

appearance that the role of coherence inside of favorability is not reducible, in any
foundationalist way, to the role of requisite support from experience.
However, the analogy-driven way of understanding directness is not, and cannot
be, the only relevant one to which she is committed. This becomes apparent when
we examine her account of the connection between S- and C-justification—which she
needs in order for foundherentism to be a “dual-aspect” theory, and not just two com-
peting theories of justification. She connects S- and C- justification by correlating both
experiential S- and C-justification, on one hand, and S- and C-directness/indirectness,
on the other. Let’s take the S-side of this first.

. . . A’s experiential S-evidence with respect to p consists of non-belief states of

A, not the kind of thing with respect to which A has, or needs, evidence. Experi-
ential S-evidence evidentially sustains/inhibits S-beliefs, but not vice versa. A’s
experiential S-evidence is, one might say, his ultimate S-evidence.27

(I quoted part of this above, in the discussion of S-evidence—that part turns out to be
decisive in both cases). S-evidence is thus direct or indirect according to how causally
close or remote it is to experiential S-evidence. For instance, relative to experience E,
p is a direct S-belief if it is caused by E; an indirect1 S-belief if it is caused by a belief
that is caused by E; an indirect2 S-belief if it is caused by a belief that is caused by a
belief that is caused by E; and so on.28
Now for the C-side. Haack claims that A’s experiential C-evidence for his belief
that p consists of

. . . sentences or propositions to the effect that A is in a certain state or states—the

state(s) which constitute(s) A’s experiential S-evidence for believing that p.29

She further claims that the directness/indirectness of C-evidence are distinguished

“ . . . in parallel to the corresponding distinctions for A’s S-evidence”.30 This is crucial.
What can it mean for C-directness to run parallel to S-directness? Only this: just as we
distinguish S-directness/indirectness in terms of causal-dependence closeness/remote-
ness to experiential S-evidence, we must distinguish C-directness/indirectness in terms
of logical-dependence closeness/remoteness to experiential C-evidence. Notice that
this sense of directness is more than the crossword puzzle sense: it can be applied
inside judgments of favorability, in a way that requires the reducibility of the role of
coherence in those judgments to the role of fit with experience (however remote that
experience may turn out to be, in crossword puzzle terms).
The correlations between experiential S-evidence and experiential C-evidence, on
one hand, and S-directness and C-directness, on the other, thus combine to guaran-
tee the same feeble foundationalist structure of S- and C-justification. We identified
the class of S-beliefs that directly depend on experiential S-evidence as S-basic be-
liefs, in the feeble foundationalist sense, and we identified S-directness (and thus the
requirement that “Experiential S-evidence evidentially sustains/inhibits S-beliefs, but
not vice versa”31 ) as implying the S-equivalent of (FD3) in the definition of feeble

27 Haack (1993, p. 77).

28 Ibid.
29 Haack (1993, p. 80).
30 Ibid.
31 Haack (1993, p. 77).
226 Synthese (2008) 160:215–228

foundationalism. The C-correlations imply that we may, with equal justice, identify
the beliefs that comprise experiential C-evidence as, themselves, C-basic beliefs, and
C-directness (and thus a corresponding requirement that “experiential C-evidence
evidentially sustains/inhibits C-beliefs, but not vice versa”) as implying the
C-equivalent of (FD3).
Therefore, even if there is promise of a middle way in the crossword puzzle analogy,
her foundherentism cannot realize this promise. For her account of S-justification is
straightforwardly feeble foundationalist, and her account of the connection between
S-justification and C-justification ensures the same result for C-justification.32 Nor is
that the product of a mere technicality: it is necessary for the unity of foundherentism,
as she conceives it.

7 Foundherentism is a feeble foundationalism, confirmed

Above I considered and replied to all but one element of Haack’s reply to BonJour. In
this final section I turn to her one remaining point, which seems to be, in summary, this:
‘make your foundationalism feeble enough to accommodate my foundherentism and
then all beliefs will be actually basic, and so there will be no non-ad-hoc reason to dis-
tinguish basic from derived beliefs’.33 In order to show this, she considers and rejects
two possible reasons for a feeble foundationalist basic/derived belief distinction.
First, in order to try to hold onto the distinction, the feeble foundationalist can
try to say that basic beliefs are those that are at first (in time) only supported by
experience, but may later gain justification in virtue of their coherence with one’s
other relevant beliefs. But why deny the initial relevance of what one admits can be
relevant later? Such a move is decidedly arbitrary.34
Second, the feeble foundationalist can try to hold that basic beliefs are in part
directly supported by experience, whereas derived beliefs are entirely supported by
other beliefs. However, in this case there will probably be no derived beliefs, and
so no useful reason to make the distinction. For “dim memory traces of what one
earlier saw, read, etc.” will count as direct support by evidence.35 And what justified
belief does not owe some of its justification to such sources? Moreover, this strategy

32 She anticipates a worry to this effect, in Evidence and Inquiry, p. 86, but her response shows that
she does not fully appreciate the grounds of the worry. She writes that in assessing security
. . . one will eventually reach a point where the issue is not how well some belief is supported
by other C-beliefs, but how well it is supported by experiential C-evidence. But doesn’t this
mean that the account is lapsing into some kind of foundationalism? No. What it means is that
“justified” eventually drops out of the explicans as one reaches the question, how well some
belief(s) is (are) supported by experiential C-evidence; this does not require that any beliefs
be justified exclusively by experiential C-evidence, nor, a fortiori, that all other justified beliefs
be justified by the support of such beliefs. (86)
What this mainly misses, of course, is that feeble foundationalism does not “require that any beliefs
be justified exclusively by” basic beliefs. Also, dropping justification out of the explicans looks at best
ad hoc; and it is inconsistent with her account of the connection between S- and C-evidence. Finally, it
seems to commit foundherentism to contextualism—the view that beliefs can owe justificatory support
to beliefs that are not, in context, themselves in need of justification—a possibility that she elsewhere
denies (e.g., p. 20).
33 In my words, this is the argument in Haack (1997, p. 28–29).
34 Haack (1997, p. 28).
35 Ibid.
Synthese (2008) 160:215–228 227

identifies “no plausible kind of belief as basic.”36 One might try to counter that it
identifies beliefs at the lowest levels of generality as basic. However, beliefs at the
lowest levels of generality are as interdependent as beliefs at the highest levels of
generality.
. . . think of the belief that this is green, when the subject is playing a novel
version of Kim’s game in which, now blindfolded, he is trying to recall the color
of a soft, round object from the collection he earlier saw; of an American’s belief
that there’s a dog in the yard, and then of his belief that there’s a wombat; or the
scientist prompted by what he sees in the bubble chamber to draw the conclusion
that electrons are composed thus and so.37
She thus supposes that without some independently plausible reasons to think that
some actual beliefs are derived, the distinction falls apart so that there is no reason to
identify any as basic, either. The distinction becomes ad hoc; the only reason to make
it is to preserve the foundationalism of the theory.38
First, I straightaway concede her criticism of feeble foundationalisms that distin-
guish basic beliefs solely on the basis of their being first, in time, justified only by
experience. Second, for the sake of argument I can concede (although I have to say I
am unconvinced) that all justified beliefs owe some justification directly to experience.
What I deny, first, is that this shows that we can identify “no plausible kind [of beliefs]
as basic.” Instead, it shows that all actual justified beliefs are basic. What I deny,
second, is her implicit assumption that if all justified beliefs are actually basic, then
the basic/derived distinction itself is ad hoc, i.e., dictated by theory alone. Whether
or not we are psychologically capable of having derived beliefs, the distinction is
still an essential part of the foundationalist solution to the regress problem, and so
all feeble foundationalisms, including foundherentism, are committed to it by their
anti-coherentism.
Suppose Descartes had a sister, say, Lisa. Suppose that Lisa agreed with all of the
prior assumptions and conclusions of his Meditations up until, but not including, his
proof of the certainty of what he “clearly and distinctly perceives.” She thus becomes
a solipsist of the present moment. She believes that she exists, now, and that is all she
believes (or, more likely, all she thinks she justifiedly believes). She does not follow
her brother even far enough to believe that she is a thinking thing, as opposed to
any other sort of thing. After all, the Evil Genius might fool around with the trace
memories she needs (on Haack’s psychological view) to reason from her recognition
of the Cogito at any given time, t1, to her conclusion that she is a thinking thing at any
given subsequent time, t2 . She therefore holds that she has only one justified belief,
the Cogito, and it is basic. Does that prove that she is not a foundationalist? Hardly.
Instead, foundationalism is a necessary premise in her argument for the conclusion
that she has no (justified) derived beliefs.
The fact that a recursive clause in a definition fails to pick out any real thing does
not prove that it is not valid or important. Consider, for instance, a legal will that
recursively picks out secondary beneficiaries on the basis of their relations to primary
beneficiaries, although, sadly, all of the actual secondary beneficiaries have passed
away. The lawyers do not therefore strike the uninstantiated recursive clause; for it

36 Ibid.
37 Ibid.
38 Haack (1993, p. 29).
228 Synthese (2008) 160:215–228

still reflects the testator’s intentions, although they now apply only counterfactually,
and part of the purpose of a will, above and beyond its distributive implications, is to
reflect the testator’s intentions. Also, the lawyers keep the uninstantiated recursive
clause for reasons of epistemic modesty. Perhaps their information is wrong. Perhaps
an unexpected secondary beneficiary will turn up. Similarly, perhaps Haack’s psy-
chology of belief is wrong. Perhaps a foundherentist derived belief will turn up.39
Foundherentism is prepared to handle that eventuality if it sticks to its feeble founda-
tionalist response to the regress problem, but it could not handle such a discovery if it
arbitrarily turns its back on its non-coherentist birthright: the basic/derived belief dis-
tinction. Hence, foundherentism should hold onto its basic /derived belief distinction
in order to encounter, and survive, unexpected discoveries in psychology.
Finally, even if there are no foundherentist derived beliefs, simplicator, within
foundherentism we can distinguish basic and derived beliefs relative to particular
experiential grounds. With respect to experiential ground X, belief Y is basic and
belief Z is derived, even if, with respect to experiential ground R, Z is basic and Y is
derived. This is enough for feeble foundationalism; and so foundherentism is a feeble
foundationalism, and, as I argued above, feeble foundationalism is a foundationalism.
Therefore, Haack’s foundherentism is a foundationalism.

Acknowledgements I wish to thank Robert Audi, Melissa Bergeron, William Cornwell, John Greco,
Louis Pojman, Stephen Tramel, Dan Zupan, and two anonymous Synthese reviewers for helpful
comments on earlier versions of this paper.

References

Audi, R. (1993). The structure of justification. Cambridge: Cambridge University Press.

BonJour, L. (1997). Haack on justification and experience. Synthese, 112, 13–25.
Chisholm, R. (1989). Theory of knowledge (3rd ed.) Englewood Cliffs, NY: Prentice Hall.
Greco, J. (1996). Review of Susan Haack, Evidence and Inquiry, in International Philosophical
Quarterly, XXXVI(2), 231–234.
Haack, S. (1993). Evidence and inquiry: Towards reconstruction in epistemology. Oxford:
Blackwell Publishers.
Haack, S. (1997). Reply to BonJour. Synthese, 112, 25–35.
Sayre-McCord, G. (1996). Coherentist epistemology and moral theory. In W. S. Armstrong, &
M. Timmons (Eds.), Moral knowledge? New readings in moral epistemology (pp. 137–189).
Oxford: Oxford University Press.

39 This is not so unlikely. Perhaps, for instance, we will discover that the best theory of how we
learn the meanings of some words requires that, at some time, we are justified in believing of them
that they have particular meanings, not because of anything that happens when we read them (i.e.,
not because of their clues), but only because of justifications for believing that the words that form
their contexts have certain meanings. Hence, they will be like crossword puzzle answers concerning
which the puzzler never understood the clue but nevertheless earns deserved confidence in his answer
because he filled it in entirely on the basis of other, well-justified (i.e., secure) intersecting answers.