Attn: Colleen / ESSAY

Synthese (2008) 160:47–61 DOI 10.1007/s11229-006-9078-0

O R I G I N A L PA P E R

What justifies that?

Patrick Hawley

Received: 25 August 2005 / Accepted: 4 July 2006 / Published online: 7 November 2006 © Springer Science+Business Media B.V. 2006

Abstract I clarify and defuse an argument for skepticism about justification with the aid of some results from recent linguistic theory. These considerations illuminate debates about the structure of justification.

Keywords Justification · Foundationalism · Resultative · Skepticism

1 Introduction

If the house is surrounded, then the house is surrounded by something. If your glass is filled, then your glass is filled by something. If my belief is justified, is my belief justified by something? It is difficult to resist saying yes; in general, it is difficult to resist the following claim:

For any B, if B is justified, then B is justified by something.

Yet, resistance matters. The claim plays a role in a central argument for skepticism about justification. The claim also figures importantly in arguments against founda- tionalist accounts of justification.

In this essay, I examine whether the claim can be successfully resisted. As a test case, I consider a skeptical argument that relies on the claim. With the help of a little linguistics, I clarify the skeptical argument, arguing that we can both reveal flaws in the argument, and explain its apparent compellingness. I then draw some conclusions about discussion of the structure of justification.

Let us now see the claim at work in a skeptical argument.

P. Hawley (B) Department of Philosophy, Hong Kong University, Hong Kong, e-mail: [email protected]

48 Synthese (2008) 160:47–61

2 The Agrippan argument

One venerable argument for skepticism about justification goes like this. Suppose that one of my beliefs is justified. Since my belief is justified, it is justified by something. Now, if what justifies my belief is not itself justified, then my belief is not justified. So, what justifies my belief is justified. But if what justifies my belief is justified, then it is justified by something. And so on. This process either moves in a circle, or goes on infinitely. Since neither a circle nor an infinite chain suffices to justify my belief, my belief is not justified. Thus we have a reductio of the claim that my belief is justified. This argument can, at any time, be given by anyone, about any belief. So, no belief can ever be justified.1

The Agrippan argument, as just presented, is a reductio of the hypothesis that some belief is justified, relying on the following three premises:2

Premise I: For any A, if A is justified, then A is justified by something. Premise II: For any A and B, if A is justified, and A is justified by B, then B is justified. Premise III: Neither circular nor infinite chains suffice to confer justification.

Few philosophers are driven to skepticism about justification by the Agrippan argument on its own, for there are many ways to try to avoid the conclusion. A foun- dationalist thinks that each chain eventually ends. Others believe that circular—or even infinite—chains of justification can be acceptable. These views about the struc- ture of justification appear in various forms and combinations.

Although few philosophers are swayed by the Agrippan argument, the premises are each intuitively compelling. Indeed, many philosophers rely on parts of the Agrip- pan argument. The best argument for foundationalism relies on the Agrippan premise that neither circular nor infinite chains suffice for justification. Arguments for other views about the structure of justification often rely on Premise I and/or II.

Thus, plenty of non-skeptics should be interested in the Agrippan argument. An examination of the Agrippan argument will shed light on discussions about the struc- ture of justification. In what follows, I try to show that attention to linguistic issues reveals that the argument is unsound, and explains why it appears sound.

Undoubtedly, some readers will question one or more of the premises of the Agrippan argument. Some readers may accept (and have arguments for) views about justification which are incompatible with one or more of the premises. I emphasize that it is not my present purpose to directly engage with particular views about justifica- tion. I think one can agree that, before one starts developing theories of justification, each of the premises seems plausible. My project here is to see how far one can get in resisting the Agrippan argument without appealing to a particular theory of justification.

1 Of ancient lineage, such skeptical reasoning appears in Sextus Empiricus, Outlines of Pyrrhonism and Against the Mathematicians. Diogenes Laertius (IX 88) attributes it to one Agrippa, a philosopher otherwise unmentioned in ancient manuscripts. Barnes (1990, 1985) discusses the Agrippan reasoning helpfully in The Toils of Scepticism. 2 In presenting the premises, I make no mention of belief. In Premises I and II, the quantifiers could range over beliefs, collections of beliefs, or something else. For present purposes, assume that the quantifiers range over anything epistemically relevant. (Although I am discussing the argument as if the conclusion is that no belief can be justified, it should be noted that similar reasoning may be used to argue that no action, nor person, nor anything else can be justified.)

Synthese (2008) 160:47–61 49

3 Some ambiguity

Sentences of the form “X is justified” have at least three distinct readings. Call this the ambiguity thesis.

If the ambiguity thesis is correct, then Premise I and Premise II are ambiguous too. If so, then the soundness of the Agrippan argument is suspect. For the argument is sound only if those disambiguations of Premise I and II that are true, are sufficient (with Premise III) for the conclusion.

I will defend the claim that the Agrippan argument has an unfilled gap: the plau- sibly true readings of Premise I and II are not sufficient (with Premise III) for the conclusion.

I will also defend the claim that the ambiguity thesis helps explain the plausibility of the Agrippan argument: Premise I and II are plausibly true on different disambig- uations of “A is justified”, and we fail to notice the shift between readings.

I will now present evidence to support the ambiguity thesis. After that, I will argue for the two further claims.

3.1 An old worry?

Some philosophers have worried that failure to attend to certain distinctions concern- ing justification causes confusion in epistemological discussion. These are concerns that, I think, are on the right track. Here are two examples.

1. Some philosophers observe that when we say that a belief was justified, sometimes we mean that the belief had a certain status, that of being justified. But sometimes when we say that a belief was justified we mean something else: we are describing the activity of showing that the belief had that status.3

2. Some philosophers have noticed that justification sometimes seems to come in degrees, and sometimes not. On the one hand, one belief may be more justified than another. On the other hand, it seems that having a status does not come in degrees; a belief either has the status of being justified or not.4

In other work, I plan to address the second distinction, and its relevance to skep- ticism about justification. My claims here are connected to the first distinction. One of my goals is to sharpen and clarify this first distinction by drawing on some recent linguistic results.

3 W. Alston remarks on the “pervasive confusion between the activity of justifying a belief- showing the belief to be reasonable, credible, or justified- and a belief’s being justified, where this is some kind of epistemic state or condition of the believer.” (“What’s Wrong with Immediate Knowledge?”, p. 70). See also (“Level Confusions in Epistemology”, p. 166). R. Audi makes a similar point in The Structure of Justification (pp. 25–26, also Chaps. 4 and 10), claiming that there is a difference between the process of justification and the property of justification. After drawing a similar distinction, Pryor (2000) (“The Skeptic and the Dogmatist”, p. 535), claims that the status of being justified is “episte- mologically primary”. Notably, Chisholm (1988) avoids forms of the word “justify” in his epistemic principles and definitions in Theory of Knowledge (pp. 135–140). 4 Goldman (1979, 1986), for example, notices this in “What is Justified Belief?” and in Chap. 4 of Epistemology and Cognition.

50 Synthese (2008) 160:47–61

3.2 Eventive and stative readings of passive sentences5

An analysis of sentences like “My belief is justified” is rather complicated, so let me begin with some simpler examples:

1. That door was closed 2. My house was surrounded

Passive sentences like 1 and 2 have two easily distinguishable readings. On one reading, 1 describes an event or a process, namely someone (or something) closing a door.6 On another reading, 1 describes a state, namely the state of the door after having been closed. Similarly, 2 can describe the event of someone (or something) surrounding my house; or it can describe the state of my house after having been surrounded. Many—perhaps most—English verbs form passive sentences with these two readings.

This difference, between a stative reading of 1 and an eventive reading of 1 can be even more clearly seen if we add a time modifier:

3. That door was closed yesterday at 3 pm

Was the door in a closed state just before 3 pm yesterday? On an eventive reading, the door was in an unclosed state just before 3 pm, and then someone or something closed it at 3 pm. On a stative reading, the door may or may not have been in a closed state just before 3 pm. So, the two readings of “That door was closed yesterday at 3 pm” have different entailments. Using a superscript E to indicate an eventive read- ing, and a superscript S for a stative reading, we can say that “That door was closed yesterday at 3 pm”E entails

4. That door was not in a closed state yesterday just before 3 pm

while “That door was closed yesterday at 3 pm.”S lacks this entailment. Similarly,

5. My house was surrounded by government agents yesterday at 3 pm

can describe the state of my house yesterday at 3 pm, or, instead, an untoward event at that time.

3.3 Another stative reading

Consider the following sentences:

6. That door was built closed 7. My house was built surrounded 8. * That door was built opened

5 The discussion in this section and the next is indebted to Embick (2004), Embick (2000), Marantz (2000), Marvin (2000), and Kratzer (1994). Recent work in this area includes Travis (2005a, b) and Jackson (2005). Many of the example sentences are from Embick (2000). Thanks to Tatjana Marvin and Karlos Arregi. 6 From now on I will just say “event” because I am not here examining the differences (if any) among events, processes and other non-states. Also, for convenience, I will sometimes speak loosely, saying that a sentence S entails the existence of some event. What I mean is that S entails S′, where S′ is a sentence that is often taken by philosophers to describe an event.

Synthese (2008) 160:47–61 51

9. * That door was built smashed 10. * That kettle was forged cleaned

8, 9 and 10 are very odd.7 However, 6 and 7 make perfect sense. The problem with 8 seems to be that “opened” indicates not only that the door was in an open state, but also that there was an earlier door opening. But, if the door was built that way, there was no earlier door opening. Similarly, the problem with 9 seems to be that “smashed” indicates not only that the door was in a smashed state, but also that there was an earlier door smashing. But, if the door was built that way, there was no earlier door smashing.

The data suggest the following. “That door was opened” (on a stative reading) entails the existence of an earlier event of opening that door. “That door was smashed” (on a stative reading) entails the existence of an earlier event of smashing that door. “That kettle was cleaned” (on a stative reading) entails the existence of an earlier event of cleaning that kettle.

On the other hand, “That door was closed” (on a stative reading) need not entail that there was an earlier event of closing that door. And “My house was surrounded” (on a stative reading) need not entail that there was an earlier surrounding event.

In short, it appears that there are two stative readings available for passive sen- tences: a resultative stative reading, which describes a state as resulting from an earlier event, and a pure stative reading which merely describes a state. The resultative read- ing entails the existence of a certain earlier event, while the pure stative reading lacks this entailment. I’ll indicate the resultative reading with a superscript R, reserving the superscript S for the pure stative reading. Thus, if “X was V’ed” has both a resultative and a pure stative reading, then “X was V’ed”R entails “Someone (or something) V’ed X”,8 while “X was V’ed”S lacks this entailment.

One test for the presence or absence of these two stative readings has just been presented. If V can occur with a verb of creation (as in “built closed” and “built sur- rounded”) then “X was V’ed” has a pure stative reading. If V cannot sensibly occur with a verb of creation (as in “built opened” and “forged cleaned”) then “X was V’ed” has a resultative reading, but not a pure stative reading.9

Given that two stative readings are available for passive sentences, one might ask whether any single passive sentence exhibits both stative readings. I believe that the answer is yes, although I do not have an example to show this conclusively.

7 The ‘*’ indicates that most native English speaking informants judge that sentence to be ungram- matical. Some informants do find some of the starred examples acceptable, but I have found no one who finds all such examples acceptable. 8 Where this is a (disambiguated) sentence describing an event. Note that in this generalization ‘ed’ represents the appropriate passive or past tense ending. For example, the passive “The window was broken” entails the past tense “Someone (or something) broke the window”. 9 There are other linguistic tests to help determine whether a participle can yield a resultative read- ing. (See Kratzer (1994) and Embick (2004).) One test is to see if modification with an adverbial is possible:

The package remained carefully opened. *The package remained carefully open.

Another test is whether prefixing with “un-” is possible: This door was unopened.

*This door was unopen. (The second test is not conclusive. As Kratzer points out, some participles which can be prefixed by “un-” like “unshaven” does not appear to form passive sentences with a resultative reading.)

52 Synthese (2008) 160:47–61

Still there is the thought from the last section: “On another reading, 1 describes a state, namely the state of a door after having been closed.” Did you find that peculiar when I said it? If not, you have a reason to think that “That door was closed”R is available. And, since “That door was built closed” makes sense, there is reason to think that “That door was closed"S is available too.10

For present purposes, I will maintain that some passive sentences (like “That door was closed”) have at least three readings: the eventive, the resultative and the pure stative. I adopt this hypothesis to help simplify the exposition of my argument; the argument survives (in a modified form) without it. Sufficient for a version of my argu- ment is a weaker claim: some passive sentences have both an eventive and resultative reading, and some passive sentences have both an eventive and a pure stative reading. I believe that the weaker claim is well supported by the examples I have given.

3.4 Summary

The following summarizes the three readings of some passive sentences:

Eventive SE describes the occurrence of an event Resultative SR a state resulting from an earlier event Pure stative SS a “pure” state: no earlier event implied

3.5 A syntactic ambiguity

I have just suggested that some passive sentences have at least three readings. The argument was based on intuitive judgments about meaning and entailment.

That there are three readings of sentences like “That door was closed” is compat- ible with recent linguistic theory. Some generative linguists theorize that sentences like “That door was closed” and “My house was surrounded” are syntactically ambig- uous. According to them, such English sentences correspond to at least three distinct syntactic structures.11

10 According to the tests from footnote 9, “That door was closed” has a resultative reading: both “That door remained carefully closed” and “That door was unclosed” are grammatical. According to the verb of creation test, “That door was closed” has a pure stative reading: “That door was built closed” makes sense. So the linguistic tests indicate that both stative readings are present. 11 Embick (2004) develops a proposal in the framework of Distributed Morphology. Embick’s pro- posal is that the structure for the (pure) stative lacks v, the verbalizing head, while the difference between the eventive and the resultative lies in a feature on v: v[AG] yields the eventive interpreta- tion; v[FIENT] yields the interpretation of a state resulting from an event.

(a) (b) (c)

Synthese (2008) 160:47–61 53

One might try to explain the differences without positing a syntactic ambiguity. What is important here is that, however the differences between SE, SR, and SS are explained, they have different entailments. I assume that this claim has been made rea- sonable. My main claims in this essay neither depend on the claim that the ambiguity is syntactic, nor depend on a particular syntactic theory. My claim so far is that some passive sentences have three readings, and these readings have different entailments.12

However, that there is a reasonably well developed syntactic theory, attempting to explain the three readings as a case of syntactic ambiguity, supports my claims in at least two ways. First, the notion of syntactic ambiguity seems to me to be both less controversial and easier to understand than, for example, various mysterious claims that philosophers are prone to make about context dependence. Second, the syntactic theory is intended to explain other kinds of data. It has a degree of empirical support, both by English and by other languages. So if a syntactic ambiguity is posited by such a theory, we have reason, beyond the rough intuitive evidence above, to believe that it is present. In particular, we have a reason to believe that a syntactic ambiguity is present that is independent of philosophical exigencies.

3.6 “Justified”, finally

Like “That door was closed” and “My house was surrounded”,

11. My belief was justified

has at least three readings. As before, it is easy to distinguish an eventive reading from a stative reading by adding an indication of time:

12. My belief was justified yesterday at 3 pm

For an eventive reading, one can imagine that the evidence arrived yesterday at 3 pm, thereby justifying my belief. On a stative reading, the sentence describes the state of my belief yesterday at 3 pm.13

As before, we can test for a pure stative reading. The sentence

13. My belief was formed justified

is as innocuous as “That door was built closed”, and “That house was built sur- rounded”. Moreover, “My belief was always justified” is acceptable too. Thus, it is reasonable to conclude, as before, that there is a stative reading of “My belief was

12 Note that Kratzer (2000) develops a slightly different distinction (due to Parsons (1990)) between target state passives and resultant state passives: “If I throw a ball onto the roof, the target state of this event is the ball’s being on the roof, a state that may or may not last for a long time. What I am calling the resultant-state is different; it is the state of my having thrown the ball onto the roof, and it is a state that cannot cease holding at some later time.” (Parsons, 1990, p. 234 quoted by Kratzer, 2000, p. 2.). I don’t believe that this way of drawing the distinction is correct because it mistakenly predicts that “That door is opened” can be true if someone opened the door, closed it again, and now the door remains firmly shut. In any event, Kratzer agrees that some passive sentences are ambiguous between a reading which is event entailing and a reading which is not. 13 Both “Rico’s belief was justified by him” and “Rico’s belief was justified by the evidence” have eventive readings. So the eventive/stative distinction I am drawing is not the one mentioned earlier between a person’s performing an activity, and a belief’s having a certain status (see footnote 3).

54 Synthese (2008) 160:47–61

justified” that entails that there was a justifying event, and a stative reading that lacks this entailment.14,15

3.7 A further complication

Before I can connect these observations to the Agrippan argument, I need to clarify the entailments of different passive sentences.

So far, I have concluded that, where S is a passive sentence with all three readings, both SE and SR entail the existence of a certain event, while SS does not. “X was V’ed”E and “X was V’ed”R both entail “Someone (or something) V’ed X”. “X was V’ed”S does not. For example, “That door was closed”E entails “Someone (or some- thing) closed the door”, as does “That door was closed.”R “That door was closed”S

does not entail “Someone (or something) closed the door”. “That door was closed”S

14 The data is not completely clear. A minority of native English speakers I have sampled find “My belief was formed justified” odd. Thus, according to them, “My belief was justified” resembles “The door was opened”, in having only the eventive and the resultative readings. (Alec Marantz (p.c.) also suggests that there are some morphological reasons to expect that “justified” is like “opened”, and thus that “My belief was justified” lacks the pure stative reading.) The tests in footnote 9 suggest that a resultative reading is available for “My belief was justified”: “My belief remained carefully justified” and “My belief was unjustified” are both grammatical. So the issue is whether or not a pure stative reading is available.

If there are indeed two rather than three readings of “My belief was justified”, the argument I will make survives in a modified form. As long as some passive sentences have a resultative reading, and some passive sentences have a pure stative reading, one may claim that people can be unsure or confused or even disagree about whether a non-eventive reading of “My belief is justified” is a resultative or a pure stative. This is enough for the kind of argument I am making. 15 While I have discussed past tense passive sentences like “My belief was justified” the Agrippan argument used the present tense: “My belief is justified”. I focused on past tense sentences because the different readings are easier to hear in the past tense. However, pure stative, resultative, and non-stative readings are also available in the present tense.

Non-stative readings are easily recognized. “That door is closed” has the habitual reading that someone usually closes the door. There is also the so-called “sportscaster” reading. Imagine an announcer narrating the action in a play. “She walks toward the open door. Suddenly, the door is closed. ‘Must’ve been the wind’ she says.” “My belief is justified” has these non-stative readings too.

Now that I have identified them, in what follows, I will ignore non-stative readings of “A is justi- fied”. I do so because non-stative readings are easy to distinguish from stative readings. For example, neither the habitual reading nor the sportscaster reading is likely to cause confusion in our under- standing of the Agrippan argument. Perhaps confusion between stative and non-stative readings plays a role in other philosophical discussions. I just doubt that that is the case here.

More importantly, the pure stative/resultative distinction is also seen in the present tense. Note the contrast between “That door is closed” and “That door is opened.” While “That door is closed” may be true because that door was built closed, “That door is opened” is, on a stative reading, only true if there was an earlier door opening event.

Similarly, while “My belief is justified” may be true because there was an earlier justifying event, that need not be the case. “My belief is justified” may be true because my belief was formed justified. To see more clearly that sentences containing “is justified” can have both stative readings, consider the sentence “Killing in self-defense is justified”. This sentence could be true as a result of a court decision: before the court decision, the law said that killing in self-defense was not justified, but now, as a result of the decision, killing in self-defense is justified. This is the resultative reading. On the other hand “Killing in self-defense is justified” may be true because of the intrinsic nature of such actions. No event makes such actions justified; such an action simply is justified. This is the pure stative.

In addition, the linguistic tests mentioned earlier indicate that “My belief is justified” has both stative readings. “My belief is justified” may be true because my belief was formed justified. That indicates that a pure stative reading is available. “My belief is unjustified” and “My belief remains carefully justified” are available, which suggests that a resultative reading is present.

Synthese (2008) 160:47–61 55

describes the door as being in a certain state. “That door was closed”R describes the door as being in a certain state as the result of a closing event.

However these conclusions need to be qualified. Sometimes “X was V’ed”S does entail “Someone (or something) V’ed X”. To see this, note that “surround” differs from “closed”. If the door was built closed, then it is not the case that someone closed the door. However, if my house was built surrounded, then something surrounded my house. “My house was surrounded”S entails “Something surrounded my house”.

Noting a further ambiguity shows how to repair the conclusions. The active sen- tence “Something surrounded my house” has both an eventive and a stative reading. One can see the contrast by comparing “Someone closed the door at 3 pm yester- day” with “Something surrounded my house at 3 pm yesterday”. The former has one, eventive, reading. The latter has both stative and eventive readings. That suggests the following generalizations. “X was V’ed”S does not entail an eventive “Someone (or something) V’ed X”. However, for some, but not all, choices of V, “X was V’ed”S does entail a stative “Someone (or thing) V’ed X”. Accepting these generalizations, “My belief was justified”S does not entail an eventive “Someone (or something) justified my belief”. But we should now explore whether “My belief was justified”S entails a stative “Someone (or something) justified my belief”.

One notices immediately that in one respect “justify” patterns with “surround” rather than “close”: “Someone (or something) justified my belief” has a stative read- ing. So there is an open question: does “My belief was justified”S entail a stative “Someone (or something) justified my belief”? I have some intuition that the answer is no. Imagine that my belief was formed justified. That is, when it came into existence it was in the state of being justified. In that case, it seems to me, my belief need not have had a justifier (something that justified it); it simply was justified. If so, then we have a counterexample to the claim that “My belief was justified”S entails a stative “Someone (or something) justified my belief”.

Of course an intuition—let alone a weak intuition—is not an argument. But my point here is not to demonstrate that the entailment fails to hold. Instead, my point is that neither linguistics nor intuition compel us to think that the entailment does hold. Absent further argument, we lack reason to believe that “My belief was justified”S

entails a stative “Someone (or something) justified my belief”. What that means, as we shall see in a moment, is that we have no reason to accept such a premise in a skeptical argument.16

3.8 A is justified by B

Before returning to the Agrippan argument, I must discuss different readings of pas- sive sentences including a by-phrase: “That door was opened by John”, or “My belief is justified by the evidence”.

In “The rabid dog was killed by the laws of Texas”, the word “by” means “according to” or “in virtue of”. But in “The ball was kicked by John”, “by” does not mean “in virtue of”. Moreover, in “The rabid dog was killed by the police by the laws of Texas”,

16 Although I will then switch to the present tense “My belief is justified”. But the point holds for the present tense case also. Suppose my belief is justified because it was formed justified. That is, when it came into existence it was in the state of being justified, and it still is. Is it the case that something justifies my belief? Again, my intuition is that the answer is no, so I conclude that, absent argument, we lack reason to accept that “My belief is justified”S entails “Someone (or something) justifies my belief”.

56 Synthese (2008) 160:47–61

only the second “by” has the “in virtue of”/“according to” meaning. Now suppose that A is justified by B by induction. If so, then A is justified by B, and A is justified by induction. But this does not mean that induction bears the same relation to A as B bears to A.

For simplicity, I will initially assume that the “by” in “A is justified by B” does not mean “in virtue of” or “according to”. I will later dispense with this assumption.

As far as whether passive sentences with a by-phrase show an eventive/stative ambiguity, some facts are clear. In some cases, the by-phrase forces an eventive read- ing. “That door was smashed” has both eventive and stative readings; however, “That door was smashed by John” only describes an event in which John smashed the door. Many—perhaps most—verbs follow this pattern. Other verbs form unambiguously stative active sentences (“John owned the house”) and unambiguously stative passive sentences (“The house was owned by John”). There is also a third class of verbs. As just observed, verbs like “surround”, and “justify” form active sentences that have both stative and eventive readings. In the passive, with a by-phrase, such verbs yield both stative and eventive readings. “My house was surrounded by Federal agents at 3 pm yesterday” can either describe the state of my house at 3 pm or an untoward event at 3 pm.

I am not sure whether sentences like “My house is surrounded by Federal agents” which have at least one stative reading, have more than one stative reading. For one thing, it is not clear what an eventive stative reading would be. Perhaps we may find different stative readings by searching for different entailments. It may be that, on one reading, “A is justified by B” entails “A is justified”S, while, on another reading, this entailment is lacking. Or, perhaps, “A is justified by B” can either mean that A is fully justified by B, or that A is partly justified by B. That is, “A is (fully) justified by B” and “B is justified”S together entail “A is justified”S, while “A is (partly) justified by B” and “B is justified”S do not.

The situation is confusing. In order to chart a clear path, I will ignore this possible plethora of readings, and make the following simplifying assumption: “A is justified by B” has only one reading, a stative reading that lacks these entailments.

4 Back to the Agrippan argument

We are now in a position to examine the Agrippan argument. One goal of this section is to defend the thesis that the Agrippan argument has an unfilled gap: the plausi- bly true readings of Premise I and II are not sufficient (with Premise III) for the conclusion. A second goal is to explain how that thesis helps a non-skeptic.

4.1 Resultative version

Let me now clarify the Agrippan argument by disentangling resultatives from pure statives. Recall the premises:

Premise I: For any A, if A is justified, then A is justified by something. Premise II: For any A and B, if A is justified, and A is justified by B, then B is justified. Premise III: Neither circular nor infinite chains suffice to confer justification.

Synthese (2008) 160:47–61 57

Since Premise III does not contain “is justified”, I will not question Premise III at this time. Premises I and II contain as parts sentences of the form “X is justified”. Since such sentences are ambiguous, plausibly, Premises I and II are ambiguous too. So we can try to disambiguate Premises I and II. First, take the ambiguous parts as resultatives:17

Premise IR: if A is justifiedR, then A is justified by something. Premise IIRR: if A is justifiedR, and A is justified by B, then B is justifiedR.

These premises (coupled with Premise III) are sufficient for the skeptical conclu- sion. Should we accept them?

There is reason to accept Premise IR. As seen earlier, “A was justified”R entails “Something justified A”. So, arguably, “A is justified”R entails “Something justifies A”. Thus, “A is justified”R arguably entails “A is justified by something”.18

However, Premise IIRR is in doubt. The likely thought that makes Premise II plausible is that only a B in a justified state can serve as a justifier; if B is not in a justified state then it can’t make A justified. This thought, however, is not a reason to accept that only a B in a justified state as a result of an earlier justifying event can serve as a justifier. This thought does not adequately support Premise IIRR.

Thus, I conclude that it is not clear whether or not to accept Premise IIRR. Further argument is needed. Perhaps someone could provide further argument. I merely point out that Premise IIRR lacks the intuitive plausibility of Premise II. This version of the Agrippan argument is incomplete, as it stands.

4.2 Pure stative version

Perhaps disambiguating in favor of the pure stative holds promise:

Premise IS: if A is justifiedS, then A is justified by something. Premise IISS: if A is justifiedS, and A is justified by B, then B is justifiedS.

These two premises (coupled with Premise III) are also sufficient for the skeptical conclusion.

We have reason to accept Premise IISS. Unlike Premise IIRR, Premise IISS con- tains “B is justified”S rather than “B is justified”R. The earlier thought that there are no unjustified justifiers does support Premise IISS.

However, we lack reason to accept Premise IS. As discussed earlier, absent further argument, we lack reason to believe “A is justified”S entails a stative “Someone or something justifies A”. Premise IS lacks the intuitive plausibility of Premise I. Thus, absent further argument in favor of Premise IS, we should not accept Premise IS.19

This version of the argument is also incomplete, as it stands.

17 I suppress the initial quantifiers, and also abuse notation slightly by attaching the superscript to “justified”, but the meaning should be clear. 18 There is an extra step here. “A is justified”R entails “Something justified A” (on an eventive reading). So, plausibly, it entails “Something justifies A” (on a stative reading). 19 An anonymous reviewer worries that I am presupposing that Premise IS is false, and that to do so would be to (illegitimately) assume foundationalism. I do not think I am assuming foundationalism. To claim that we lack reason, without further argument, to believe Premise IS, is not to assume that Premise IS is false. For one may suspend judgment about Premise IS.

58 Synthese (2008) 160:47–61

4.3 Mixed version

So far, we have seen that disambiguating in favor of the pure stative makes Premise II look good, at the expense of casting doubt on Premise I. And disambiguating in favor of the resultative makes Premise I looks good, at the expense of casting doubt on Premise II. Since each of the two premises has a plausible disambiguation, we can try combining them together:

Premise IR: if A is justifiedR, then A is justified by something. Premise IISS: if A is justifiedS, and A is justified by B, then B is justifiedS.

Despite their plausibility, Premise IR and Premise IISS (coupled with Premise III) are not sufficient for the skeptical conclusion. Drawing the skeptical conclusion would be to equivocate between the pure stative and the resultative. The argument needs something more to link these premises together. These two supplementary premises will do the job:

Premise R–S: if A is justifiedR, then A is justifiedS. Premise S–R: if A is justifiedS, then A is justifiedR.

Premise R–S is clearly true. “A is justified”R entails “A is justified”S: it is impossible for A to be in a state (and be in that state as a result of a justifying event), and fail to be in that state.

Premise S–R is a substantial claim in need of argument. It needs to be argued that any justified A is in that state as the result of a justifying event. Or at least, for a restricted version of the Agrippan argument, every justified belief is in that state as the result of a justifying event.

Perhaps a philosopher could defend Premise S–R, or find some other way to make this version of the argument valid. But, as it stands, this version of the argument is, like the others, incomplete.

4.4 Other versions

We have examined three versions of the argument so far. There are eight versions of the argument in all, because there are two disambiguations of Premise I, and four disambiguations for Premise II. For the sake of completeness, here are all of them:

Premise IR: if A is justifiedR, then A is justified by something. Premise IS: if A is justifiedS, then A is justified by something. Premise IIRR: if A is justifiedR, and A is justified by B, then B is justifiedR. Premise IIRS: if A is justifiedR, and A is justified by B, then B is justifiedS. Premise IISR: if A is justifiedS, and A is justified by B, then B is justifiedR. Premise IISS: if A is justifiedS, and A is justified by B, then B is justifiedS.

We need not tarry long on other versions of the argument. Premise IIRS is plausible because, given R–S, IIRS follows from the plausible IISS.

However, IIRS does not support an intuitively plausible disambiguated Agrippan argument. As we have seen, we lack reason to accept Premise IS without further argument. And Premise IR won’t help either: Premise IIRS together with Premise IR are not sufficient for the skeptical conclusion. Adding Premise S–R would make the argument valid. But, as we have seen, we lack reason to accept Premise S–R absent further argument.

Synthese (2008) 160:47–61 59

Premise IISR can be successfully combined with IR to make a valid argument for the skeptical conclusion. However, IISR is at least as controversial as the controversial IIRR. For, like IIRR, IISR has a resultative in the consequent.

Finally, although three further versions of the argument include IS they are not worth examining; if IS were intuitively plausible than the pure stative version would be intuitively plausible.

4.5 The gap

The intuitively plausible premises—IR, IISS and IIRS—cannot be combined with Premise III to produce an argument sufficient for the Agrippan conclusion. I con- clude that the argument has an unfilled gap.

Each of the following three premises will fill the gap: Premise IS, Premise IIRR, Premise S–R. The last premise yields the other two. For suppose that Premise S–R is true. Then IS follows from IR. And (given Premise R–S) IIRR follows from IISS.

4.6 Shifty by-phrases

I have been assuming that “A is justified by B” has a univocal, stative meaning. But, earlier I pointed out an ambiguity in passive sentences with by-phrases. In a sentence like “A is justified by B by induction”, the second “by” means something like “in virtue of” or “according to”, while the first does not. Perhaps shiftiness in the meaning of the by-phrase also plays a role in the Agrippan argument. Let’s have a look.

The controversial Premise IS,

Premise IS: If A is justifiedS, then A is justified by something.

Seems trivially true if “by” here means “in virtue of”. Surely “A is justified”S, if true, is true in virtue of something; something makes it the case that A is in that state. If, instead, “by” takes on its other meaning then Premise IS is a controversial claim.

Moreover, any version of Premise II looks implausible on the “in virtue of” read- ing. For if A is justified, and A is justified in virtue of B, why should we think that B is justified? For example, I may be justified in believing some things in virtue of my expe- rience. But why think that my experience is something that needs to be justified—or even can be justified?

Shiftiness in the meaning of the by-phrase is one more source of confusion in discus- sions of justification. And, perhaps, it is one more reason why the Agrippan argument appears compelling. On the “in virtue of” reading, Premise IS looks trivially true. On the other reading, Premise IISS is plausible.

4.7 Two small steps forward for the non-skeptic

The thesis that there is an unfilled gap helps a non-skeptic in two ways. First, the non-skeptic can explain away the force of the Agrippan argument. Premise

I appears true because Premise IR is true. Premise II appears true because Premise IISS is true. And the Agrippan argument appears valid because, on some disambigu- ations of the premises, the argument is valid. If it is one part of the non-skeptic’s task to explain why this skeptical argument appears plausible, then the non-skeptic has strengthened her position.

60 Synthese (2008) 160:47–61

Second, the Agrippan argument poses no threat to the non-skeptic. The Agrippan argument is incomplete without reasons to accept an additional premise like IS, IIRR or S–R, and an incomplete argument does not threaten the non-skeptic’s position.

These two points may not be strong enough for some tastes. One might wish, for example, for an argument that IS, IIRR and S–R are false. Still, even these two limited points are interesting. Consider the following dialectic between skeptic and foundationalist. The skeptic presents Premises I, II and III to argue that no belief can be justified. The foundationalist, assuming that some beliefs are justified, argues that Premise I is false. Is the burden of proof on the skeptic or the foundationalist? Is one side begging the question? If the skeptic’s argument has an unfilled gap the founda- tionalist is free to offer her argument without responding to the skeptic’s argument. Worries about question beggingness and burden of proof can be set aside, at least until some other skeptical argument rears its head.

5 Concluding remarks

I began this essay with the following difficult to deny claim:

Premise I: For any B, if B is justified, then B is justified by something.

I displayed the central role Premise I plays in the Agrippan argument. Examining English passive sentences revealed that this premise, as well as Premise II of the Agrippan argument, are ambiguous. I concluded that the argument has an unfilled gap, and that this helps the non-skeptic.

What further conclusions can be drawn from this exercise? 1. As I noted earlier, few philosophers are driven to skepticism by the Agrippan

argument, but the argument is interesting nonetheless. The premises of the Agrippan argument form a common core around which turns the debate about the structure of justification. All non-skeptics I know of accept one or more of the premises of the Agrippan argument. They just disagree about which premises to accept.

What I have shown here helps clarify this debate. To successfully argue against foundationalism, one should argue for IS, IIRR or S–R. Foundationalists must resist all three of IS, IIRR and S–R. In resisting, foundationalists can add that Premise I and II appear true because they are true on some readings. For instance, is Premise I true? If a belief is justified is it justified by something? A foundationalist might say that this seems so because Premise IS is true.

2. The present linguistic investigation provides distinctions useful for philosophers. Here are some examples. As noted earlier, Alston (1989a, b, c) draw a distinction between a person’s performing the activity of justifying a belief and a belief’s having the status of being justified. Now we can gain more precision. First, there is the dis- tinction between a state and an event. A belief may be in a certain state, or an event may occur in which a belief enters that state. That event need not involve an activity of “justifying a belief” performed by a person. Second, there is a difference between being in a state as a result of a certain event, and merely being in that state. These distinctions are blurred by the English passive.

3. Finally, the investigation begun here about passive sentences may bear fruit beyond epistemology. If an event was caused, was it caused by something? If a prop- erty was exemplified, was it exemplified by something? Getting clear about passive sentences like “X was caused” and “Y was exemplified” may help us clarify these

Synthese (2008) 160:47–61 61

questions, and the discussions where they are asked. It may sound unfashionable, but I am encouraged, by the results here, to believe that patient and careful study of language will help move philosophical discussion forward, a few small steps at a time.

Acknowledgements The author acknowledges support for this research from grant 20062159005 from The University of Hong Kong Seed Funding Programme for Basic Research.

References

Alston, W. (1989a). Epistemic justification. Ithaca, New York: Cornell University Press. Alston, W. (1989b). Level confusions in epistemology. In Epistemic justification (pp. 153–171). Ithaca:

Cornell University Press. Alston, W. (1989c). What’s wrong with immediate knowledge? In Epistemic justification (pp. 57–78).

Ithaca: Cornell University Press. Audi, R. (1993). The structure of justification. Cambridge, UK: Cambridge University Press. Barnes, J. (1990). The toils of scepticism. Cambridge, MA: Cambridge University Press. BonJour, L. (1978). Can empirical knowledge have a foundation? American Philosophical Quarterly,

15, 1–13. BonJour, L. (1985). The structure of empirical knowledge. Cambridge, Massachusetts: Harvard

University Press. Chisholm, R. (1988). Theory of knowledge (3rd ed.). Englewood Cliffs, NJ: Prentice-Hall. Embick, D. (1997). Voice and the interfaces of syntax. PhD thesis, University of Pennsylvania. Embick, D. (2000). Participial structures and participial asymmetries. MIT Handout. Embick, D. (2004). On the structure of resultative participle in english. Linguistic Inquiry, 35, 355–392. Feldman, R. (2002). Epistemology. Upper Saddle River, NJ: Prentice Hall. Goldman, A. (1979). What is justified belief? In G. S. Pappas (Ed.), Justification and knowledge

(pp. 1–23). Dordrecht: D. Reidel. Goldman, A. (1986). Epistemology and cognition. Cambridge, MA: Harvard University Press. Jackson, E. (2005). Pima resultatives. UCLA Handout. Klein, P. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13,

297–325. Kratzer, A. (1994). The event argument and the semantics of voice. Ms. Kratzer, A. (2000) Building statives. In Conathan et al. (Eds.), Proceedings of the Twenty-sixth annual

meeting of the Berkeley Linguistics Society (pp. 385–399). Berkeley: University of California, Berkeley Linguistics Society.

Marantz, A. (2000). Reconstructing the lexical domain with a single generative engine. MIT Handout. Marvin, T. (2000). Past participles in reduced relatives. Ms. Parsons, T. (1990). Events in the semantics of english. Cambridge, MA: MIT Press. Pryor, J. (2000). The skeptic and the dogmatist. Noûs, 34, 517–549. Steup, M. (2003). Foundationalism, circularity, and infinitism. Ms. deMena Travis, L. (2005a) Passives, states, and roots in Malagasy. AFLA XII handout. deMena Travis, L. (2005). States, abilities and accidents. CLA handout.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.