Abstract
Corroborative evidence may be understood as having two epistemic effects: a primary effect by which it offers direct evidence for some claim, and a secondary effect by which it bolsters the appraised probative, or evidential, value of some other piece of evidence for that claim. This paper argues that the bolstering effect of corroborative evidence is epistemically legitimate because corroboration provides a reason to count the belief based on the initial evidence as sensitive to, and safe from, defeat in a way that it was not previously recognized to be. Discovering that our initial evidence tracks the truth in a way we previously did not recognize provides a reason to positively reappraise the probative value of that evidence. The final section of the paper relates the proposed sensitivity- and safety-based account of corroboration to an explanation-based account.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Merriam-Webster’s Dictionary of Law (1996: 172).
Initial treatments of corroboration (e.g., Cohen 1977) ignored this bolstering function, and instead sought to understand and explain the primary effect of corroboration, understood as confirmation, by which an overall case is strengthened—i.e., how support is increased—through the addition of evidence. While the possibility of some bolstering effect of corroboration has largely gone unnoticed in the epistemological literature, it has been postulated in legal accounts of it. In this context, Redmayne (2000) has argued that the bolstering effect of corroborative evidence is epistemically illegitimate because it involves the fallacy of double counting: overvaluing the probative worth of some piece of evidence by counting it twice. Elsewhere (Godden 2010), I have offered a rebuttal of Redmayne’s argument, and (Godden 2014) argued that no vicious circularity results from the bolstering effect of corroboration properly explained.
Rather than choosing between sensitivity and safety as criteria of epistemic merit, the paper contends that, as they apply in paradigmatic cases of corroboration, they function equivalently. The paper seeks to show that the elements of a situation tracked by each criterion are similar in their epistemic salience and significance, and the functional responsiveness of each criterion to those conditions coincides to yield equivalent results in paradigmatic cases of corroboration.
A test, for instance, might be highly, but not perfectly, predictive of an extremely rare condition, such that a positive result only weakly supports a positive finding owing to the low incidence of the condition itself.
Doing so might, nevertheless, increase the dialectical or rhetorical support offered, by showing that certainty in the truth of the claim in question is readily and variously demonstrated.
The terms “trustworthy” and “credible” will be used to connote the two primary features of reliability and (where appropriate) sincerity when used to describe sources and their reports.
Notice, this decrease in the claim’s probative value does not come about by a decrease in the credence we place in the claim itself. Our confidence that the instrument gives some particular reading, n, remains unchanged. Rather, what changes is the evidentiary worth we attach to the reading when considering it as evidence, or as a reason, for other claims that we might infer on the basis of the reading.
I assume that the primary function of the evidence being considered (corroborating and corroborated) is both legitimate and that its operation can be satisfactorily explained.
Weak notions of corroboration that amount to either (a) the absence of inconsistency: such that any negatively relevant evidence, in counting against a claim, thereby fails to corroborate all positively relevant evidence, or (b) positive relevance: such that any positively relevant evidence, in counting in support of a claim, thereby corroborates all other positively relevant evidence, are not the subject of investigation. If corroboration amounts only to consistency or positive relevance then it is not particularly interesting in that it generally fails to produce the bolstering effect motivating this study and it presents no special problems for the theory of evidence.
Furthermore, I assume that the secondary, bolstering function does not operate in any of the following ways: (i) it does not act as premise support, by increasing the credence we place in the corroborated evidence itself; (ii) it does not work convergently, by supplying an additional, independent reason for the main claim; and (iii) it does not function by strengthening the warrant of some initial inference (i.e., from the initial, corroborated evidence to the main claim). I have argued for this latter assumption elsewhere (Godden 2010; 2014). And, if it turns out that I am wrong on one of these points, the bolstering effect of corroboration would not present a puzzling or theoretically interesting epistemological problem, since the functioning of (i)–(iii) is quite well understood in the theory of evidence and argument.
There is good reason to think that any account of knowledge on which knowledge claims are fallible, i.e., on which justification (or whatever takes its place) is defeasible, will require some nontrivial element of luck (Zagzebski 1994).
Nozick (1981: 76) actually proposed two counterfactual conditions, adherence and variation, maintaining that each is necessary since adherence stipulates that S’s doxastic states be sensitive to P’s truth, while variation stipulates that they be sensitive to P’s falsity. Together, they require that S’s doxastic states co-vary with the truth of P. Standardly, the adherence condition has been dropped, and the variation condition alone has come to be called sensitivity. See Luper (2012) for a comprehensive discussion of the problems adherence has been found to have.
As an anonymous reviewer pointed out, this example is vulnerable to the following issue. S’s belief about the time, which might plausibly be expressed by something like “It’s 3:00 p.m. now,” contains an indexical, making the content different on each occasion of conception or expression. Thus, were it not 3:00 p.m. now, S would not acquire the same belief as they would have, had they looked at the clock when it was actually 3:00 p.m. Recognizing this point, one should also recognize that nothing essential to the case made here, or concerning sensitivity and safety generally, depends on it. The example (which I borrow and develop from Pritchard 2008: 439 ff.) may readily be changed to avoid such an issue.
As DeRose (2010: 163) claims when describing his insensitivity account, “We have some at least fairly general—though perhaps not exceptionless—tendency to judge that insensitive beliefs are not knowledge.”
For a survey, see (Vogel 2012). Primary among its defects is that sensitivity is not closed under deduction (Hawthorne 2005): “One can perfectly well track a fact P and yet fail to track a fact Q that one knows to be entailed by P” (Sosa 1999: 143), thereby giving rise to Nozick’s “abominable conjunctions” (DeRose 1995: 28). Nozick (1981: 206-211) was aware of, and embraced, this daunting consequence. (But see, e.g., Roush 2005 and Baumann 2012 for attempts to reconcile Nozick’s truth-tracking with principles of epistemic closure.)
Further, as Cross (2010: 39) memorably put it, “Sensitivity has died the death of all other 1980s epistemic theories: it is buried under multitudes upon multitudes of clever, clever counterexamples.” First, some insensitive beliefs intuitively count as knowledge, as may be seen by Sosa’s (1999: 145-46) rubbish chute example, Vogel’s (1987: 206-208) ice on a hot day example, and Hawthorne’s (2004: 11) ticketless lottery winner example. Indeed, Sosa (2002: 265) observes that, in any case where S knows that P, the related belief that they are not mistaken to believe that P is insensitive, and hence unknown (cf. Murphy and Black 2012: 32). Second, some sensitive beliefs do not intuitively count as knowledge, e.g., Forbes’s (1984: 45) Gettier-like example of the vase hologram machine (cf. Nozick (1981: 190)).
Also, Kvanvig (2004, 2012) has sought to show that truth tracking, despite its avoidance of the swamping problem, cannot fully or adequately explain the value of knowledge.
Importantly, this is not the only way that one may strengthen the probative value of their evidence. Another way, for example, would be to confirm the reliability of the source directly—for example, one might calibrate their instruments to ensure their accuracy, or gather evidence that speaks to the trustworthiness of their reporting witness.
Two points should be noticed here. First, such methods do not change the credence attached to the evidence itself—our degree of belief in the report itself (e.g., that the instrument reads n, or that the witness said that P) remains unchanged pre- and post-confirmation of reliability. What changes here is the probative value we are prepared to attach to that evidence. On the account offered here, something quite similar happens, albeit indirectly, when corroborative evidence is obtained.
The second point to notice is that, in reliability confirmation tests, what I have called the “bolstering” function of corroboration is the primary and sole function of the new evidence, rather than its secondary or auxiliary function. By contrast, in cases of corroboration the corroborating evidence also performs a primary function of providing direct evidence for the claim at issue.
An important consequence of this assumption is that our ordinary belief forming practices are normally but defeasibly reliable. Notice that this presumption of normal-but-defeasible reliability underwrites our knowledge-acquisition practices and their underlying rationale. If, for example, we did not presume clocks to be normally reliable timepieces, we would not engage in the practice of consulting them in order to discover the time and nor would we take such a practice to be rational.
For the argument I am proposing to go through, the assumption required is actually weaker than the presumption of normalcy just given. Instead, all that is required is the assumption that for any two information sources, to the extent that they are actually independent, if they are unreliable, they are not normally so in the same way, such that they do not normally give the same incorrect or misleading information (Elgin 2014: 246).
Although this paper does not take its lead from DeRose, he might be read as having presaged a move of this kind. He (1995: 18, cf. 2004: 23–24, 2010: 161–163) considers the following example:
Yesterday’s Results:
I can know that [O] the Bulls won their game last night (by reading the score as reported under “Yesterday’s Results” in the Sports section of my daily paper) even though I don’t know that [not-H] the paper isn’t mistaken about whether the Bulls won last night. And, this is the case because my belief that O is ordinarily sensitive (satisfying the subjunctive conditional that had it not been the case I wouldn’t believe it) while my belief that not-H ordinarily is not (because I would still believe it, even if it hadn’t been the case).
In the context of this example, DeRose (1995: 25) writes:
by checking appropriately independent sources, I could get myself into a position in which I seemingly would know that [not-H] the newspaper isn’t mistaken about whether the Bulls won last night. But the checks that would seemingly allow this knowledge would also make it seem that if the paper were mistaken, I would not believe that it wasn’t.
Later DeRose (1995: 33) writes “one could gather further evidence, strengthen one’s epistemic position with respect both to not-H and O, and make even one’s belief that not-H sensitive.” Here, DeRose anticipates much of what I say about corroboration, claiming that checking an independent source (i.e., finding corroborating evidence) can make a belief sensitive and that doing so can sometimes transform non-knowledge into knowledge. By contrast, I claim that making a belief sensitive increases its acceptability, and grant that this could sometimes have the effect that DeRose claims.
In this way, the account offered here further embraces some of the contextualist features of DeRose’s theory (1995, 2010) by advising that the strengthening function of corroboration is strongest when it serves to eliminate the most relevant alternatives, i.e., defeating conditions in the nearest-by possible worlds, and that the relevance of an alternative can, at least partly, be a function of its discursive features.
The two are importantly different since subjunctive conditionals do not contrapose.
This is not to say that safety-based accounts are without their problems. Like sensitivity, safety has problems with closure (Murphy 2005; Alspector-Kelly 2011). There also seem to be cases of unsafe knowledge (Comesaña 2005). Further, Neta and Rohrbaugh (2004: 404) argue that the basic idea of safety—that in order to know, one could not have easily have been wrong—is an unacceptably stringent condition on knowledge.
Sosa (1999: 142 notation adapted) immediately proceeds to give the following alternative formulations: “Alternatively, a belief by S that P is ‘safe’ iff: S would not believe P without it being the case that P; or, better, iff: as a matter of fact, though perhaps not as a matter of strict necessity, not easily would S believe that P without it being the case that P,” and later (1999: 146 italics removed) as “A belief is sensitive iff had it been false, S would not have held it, whereas a belief is safe iff S would not have held it without its being true” (cf. DeRose 2004: 26, 31).
See fn. 12, on the issue of indexicality as it pertains to the content of the resultant beliefs.
It is well known that sensitivity and safety each solve ordinary Gettier-like cases adequately. And, for my purposes, they function similarly with the kinds of cases involving corroboration. A central difference between the operation of sensitivity and safety concerns the possible worlds to which one is directed in their testing. With sensitivity, one looks for possible worlds where P fails to obtain, and checks to see whether S still believes that P. With safety, by contrast, one looks for possible worlds where S still believes that P, and checks to see whether P obtains (cf. Pritchard 2012: 174). The difference is often in the relative proximity of these worlds to the base world. For example, in the case of skeptical hypotheses, the nearest-by possible world in which S still believes that P is much closer than the nearest-by possible world where not-P. The result is that our common-sense beliefs about skeptical hypotheses tend to be quite safe even though they are not sensitive. As Pritchard (2012: 179) writes:
In cases where the possibility of error is very close, then the belief will be subject to a high degree of epistemic risk and hence very unsafe; while in cases where the possibility of error is further out the belief will be subject to a much lower degree of epistemic risk, and hence will be more likely to qualify as safe.
DeRose (2004: 33-35) likens safety to his own contextualist position on epistemic strength, writing “Sosa’s rough account of knowledge is that of safe enough true belief, while mine is that of strong enough true belief.” Both views agree that the modal proximity of the possible world falsifying the conditional matters.
This view is consistent with my own position since, I would maintain, it is relatively easy to find corroborating evidence for ordinary, yet relatively unsafe, beliefs where the possibility of error is quite close by. While doing so could eliminate the most antecedently likely defeating conditions, and thereby increase the epistemic worth of the belief, other defeaters might, nevertheless, remain quite proximate. By contrast, it can be relatively difficult (if not impossible in the case of skeptical hypotheses) to find properly corroborating evidence for relatively safe beliefs, whose only possibilities for error are rather far-fetched.
Importantly, this condition is not always satisfied. For example, someone’s guilt is not the best explanation of their having motive or opportunity. Similarly, the wrongness of the death penalty is not the best explanation its cost ineffectiveness, non-deterrence, or moral intolerability. Thus, that C could explain R1 seems to provide a condition for the possibility of corroborative, rather than merely convergent, evidence.
Coady (1992) addresses testimony, whereas here I employ a broader notion of evidential sources and their reports.
The point is not that corroboration brings evidence from the realm of the incredible to the realm of the credible, although this can happen (see Godden 2014 for examples). Rather, corroboration bolsters the probative value of existing but defeasible evidence.
See (Coady 1992: ch. 10) on when such skepticism is rationally warranted.
Here, “Wn” denotes a set of possible worlds fixed by the truth of the claims listed P1, P2, …, Pn. So, Wn is the set of possible worlds at each of which P1, P2, …, and Pn are true. Often, I will express this as follows: “Wn (P1, P2, …, Pn).”
Recent considerations of sensitivity (Black and Murphy 2007, Cross 2010, DeRose 2010, Murphy and Black 2012) have sought to link the notions of sensitivity and explanation in an effort to preserve a sensitivity-based account of knowledge. As Cross (2010) points out, compelling skeptical scenarios have an explanatory function—their plausibility, or cognitive hold over us, is not judged merely according to their antecedent probability, since we judge more probable ones (e.g., I do not have hands) to be less plausible than improbable ones (e.g., I am the handless victim of an evil genius) (cf. DeRose 1995: 23). Generally, the point unifying these revised accounts of sensitivity is to add an explanatory condition such that sensitivity-defeaters (i.e., conditions of insensitivity) must explain how it is that the subject believes falsely. Skeptical or defeating scenarios that merely insist that a subject believes wrongly in the nearest-by possible world where the claim at issue fails to obtain are to be ignored, despite their probability, precisely because they lack any explanatory value.
The explanation-based account of corroboration presently on offer finds similar epistemic merit in the explanatory power of claims and their defeaters. By contrast, though, it seeks to incorporate the positive value of explanation into the account, rather than merely precluding scenarios that lack explanatory value. For example, it relies on the idea that evidence that is best explained by the-truth-of-what-it-is-evidence-for has a greater probative value than evidence best explained by something inconsistent with the-truth-of-what-it-is-evidence-for. Most strongly, it claims that that the bolstering function of corroboration only occurs in cases where the claim, C, is an explanation of the corroborated evidence for it.
References
Alspector-Kelly, M. (2011). Why safety doesn’t save closure. Synthese, 183, 127–142.
Baumann, P. (2012). Nozick’s defense of closure. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 81–97). Cambridge: Cambridge UP.
Becker, K. (2012). Methods and how to individuate them. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 1–8). Cambridge: Cambridge UP.
Black, T., & Murphy, P. (2007). In defense of sensitivity. Synthese, 154, 53–71.
Carnap, R. (1962). Logical foundations of probability (2nd ed.). Chicago: University of Chicago Press.
Coady, C. A. (1992). Testimony: a philosophical study. Oxford: Clarendon Press.
Cohen, L. J. (1977). The probable and the provable. Oxford: Oxford UP.
Comesaña, J. (2005). Unsafe knowledge. Synthese, 146, 393–402.
Cross, T. (2010). Skeptical success. In T. Gendler & J. Hawthorne (Eds.), Oxford studies in epistemology (Vol. 3, pp. 35–62). Oxford: Oxford UP.
DeRose, K. (1995). Solving the skeptical problem. Philosophical Review, 104, 1–52.
DeRose, K. (2004). Sosa, safety, sensitivity, and skeptical problems. In J. Greco (Ed.), Ernest Sosa and his critics (pp. 22–41). Oxford: Blackwell.
DeRose, K. (2010). Insensitivity is back, baby! Philosophical Perspectives, 24, 161–187.
Elgin, C. (2014). Non-foundationalist epistemology: holism, coherence and tenability. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (2nd ed., pp. 244–255). Oxford: Wiley Blackwell.
Forbes, G. (1984). Nozick on scepticism. The Philosophical Quarterly, 34, 43–52.
Gettier, E. (1963). Is justified true belief knowledge? Analysis, 23, 121–123.
Godden, D. (2010). Corroborative evidence. In C. Reed & C.W. Tindale (Eds.), Dialectics, dialogue and argumentation: An examination of Douglas Walton's theories of reasoning and argument (pp. 201-212). London: College Publications.
Godden, D. (2014). Modeling corroborative evidence: Inference to the best explanation as counter-rebuttal. Argumentation, 28, 187-220.
Goldman, A. (1976). Discrimination and perceptual knowledge. Journal of Philosophy, 73, 771–791.
Greco, J. (2012). Better safe than sensitive. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 193–206). Cambridge: Cambridge UP.
Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford UP.
Hawthorne, J. (2005). The case for closure. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (1st ed., pp. 26–41). Oxford: Blackwell.
Kvanvig, J. (2004). Nozickian epistemology and the value of knowledge. Philosophical Issues, 14, 201–218.
Kvanvig, J. (2012). Truth-tracking and the value of knowledge. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 101–121). Cambridge: Cambridge UP.
Lipton, P. (1990). Contrastive explanation. Royal Institute of Philosophy Supplement, 27, 247–266.
Luper, S. (2012). False negatives. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 207–226). Cambridge: Cambridge UP.
Luper-Foy. S. [now “Luper”]. (1984). The epistemic predicament: knowledge, Nozickian tracking, and scepticism. Australasian Journal of Philosophy, 89, 47–58.
Murphy, P. (2005). Closure failures for safety. Philosophia, 33, 331–334.
Murphy, P., & Black, T. (2012). Sensitivity meets explanation: an improved counterfactual condition on knowledge. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 28–42). Cambridge: Cambridge UP.
Neta, R., & Rohrbaugh, G. (2004). Luminosity and the safety of knowledge. Pacific Philosophical Quarterly, 85, 396–406.
Nozick, R. (1981). Philosophical explanations. Cambridge: Harvard UP.
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford UP.
Pritchard, D. (2008). Sensitivity, safety, and antiluck epistemology. In J. Greco (Ed.), The Oxford handbook of scepticism (pp. 437–455). Oxford: Oxford UP.
Pritchard, D. (2012). In defence of modest anti-luck epistemology. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 173–192). Cambridge: Cambridge UP.
Redmayne, M. (2000). A corroboration approach to recovered memories of sexual abuse: a note of caution. Law Quarterly Review, 116, 147–155.
Roush, S. (2005). Tracking truth: Knowledge, evidence, and science. Oxford: Oxford UP.
Russell, B. (1948 1992). Human knowledge: Its scope and limits. New York: Routledge.
Sosa, E. (1999). How to defeat opposition to Moore. Philosophical Perspectives, 13, 141–154.
Sosa, E. (2002). Tracking, competence, and knowledge. In P. Moser (Ed.), The Oxford handbook of epistemology (pp. 264–286). Oxford: Oxford UP.
Vogel, J. (1987). Tracking, closure, and inductive knowledge. In S. Luper-Foy (Ed.), The possibility of knowledge: Nozick and his critics (pp. 197–215). Totowa: Rowman & Littlefield.
Vogel, J. (2012). The enduring trouble with tracking. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 122–151). Cambridge: Cambridge UP.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford UP.
Zagzebski, L. (1994). The inescapability of Gettier problems. The Philosophical Quarterly, 44, 65–73.
Acknowledgements
Earlier versions of this paper were presented as conference presentations at the Canadian Society for Epistemology, Université de Sherbrooke, Québec, on 23 November 2012, and the Southern Society for Philosophy and Psychology 105th Annual Meeting, Austin, Texas, on 28 February 2013, and as philosophy department talks at Old Dominion University, on 5 February 2013, the University of Waterloo, on 7 April 2016, McMaster University, on 8 April 2016, and the University of Nevada, Las Vegas, on 7 October 2016. Of those audiences, thanks are especially due to Dale Miller, David Matheson, Cecilea Mun, and Bill Roche for their insightful and instructive comments. Thanks to David Matheson for, early on in this project, calling my attention to the fact that, in addition to making a view defeasibly sensitive, corroboration similarly makes it defeasibly safe. To Scott Aikin, Keith DeRose, Catherine Elgin, John Grey, Matt McKeon, and Duncan Pritchard I offer my sincere thanks for their careful readings of earlier versions of this paper, for their illuminating and constructive criticisms, and for their encouragement of the project generally. Additionally, the anonymous reviewers of this paper have my thanks.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Godden, D. Corroboration: Sensitivity, Safety, and Explanation. Acta Anal 34, 15–38 (2019). https://doi.org/10.1007/s12136-018-0351-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12136-018-0351-x