The structure of evidence law

Oxford Journal of Legal Studies

Volumn 26, Issue 4, 2006, Pages 805-822

  Redmayne, Mike ^a,b  

^a LONDON SCHOOL OF ECONOMICS   (United Kingdom)

^b AUSTRALIAN NATIONAL UNIVERSITY   (Australia)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 33845704020     PISSN: 01436503     EISSN: 14643820     Source Type: Journal    
DOI: 10.1093/ojls/gql024     Document Type: Review

References (103)

1. It is common to draw a distinction between 'risk' and 'uncertainty', the latter involving situations where probabilities (or precise probabilities) are not available (see, e.g. J. Runde, 'Clarifying Frank Knight's Discussion of the Meaning of Risk and Uncertainty' (1998) 22 Cambridge J Economics 539). The use of the latter term is not intended to invoke this distinction, as to do so would beg various questions.
2. An accessible introduction is I. Hacking, An Introduction to Probability and Inductive Logic (2001).
3. R. Friedman, 'Answering the Bayesioskeptical Challenge' (1997) 1 E & P 276
4. and R.O. Lempert, 'Modeling Relevance' (1977) 75 Michigan L Rev 643 are good accounts of the model sketched in this paragraph.
5. See D.A. Nance, 'The Best Evidence Principle' (1988) 73 Iowa LR 227.
6. See especially L.J. Cohen, The Probable and the Provable (1977). A more accessible introduction is his An Introduction to the Philosophy of Induction and Probability (1989)
7. various extracts from Cohen's work can be found in P. Murphy (ed.), Evidence, Proof and Facts: A Book of Sources (2003).
8. Unfortunately, it is quite common to distinguish Baconian from conventional probability in terms of a remark made by Cohen (at (1979) 7 Cognition 389): That conventional probability grades probabilification on the assumption that all relevant facts are specified in the evidence, while Baconian probability does so by the extent to which all relevant facts are so specified. As will become clear, this is a misinterpretation of conventional probability.
9. See P. Tillers and E.D. Green (eds), Probability and Inference in the Law of Evidence (1988)
10. the symposium on 'Decision and Inference in Litigation' (1991) 13 (2-3) Cardozo L Rev
11. the special issue on 'Bayesianism and Juridical Proof' (1997) 1 E & P 253-360.
12. A. Stein, Foundations of Evidence Law (2005) (hereafter referred to as Foundations).
13. Foundations, x, 133.
14. Ibid at 107.
15. The main exception is the drift away from exclusionary rules, almost complete in the UK in civil litigation and significant, especially after the Criminal Justice Act 2003, in criminal litigation.
16. The concept of weight is due originally to Keynes: See J.M. Keynes, A Treatise on Probability (1921), esp ch. 6.
17. A useful discussion is J. Runde, 'Keynesian Uncertainty and the Weight of Arguments' (1990) 6 Economics & Philosophy 275.
18. On the role played by generalizations, see T. Anderson, D. Schum, and W. Twining, Analysis of Evidence (2nd edn, 2005) at ch. 10.
19. This is odd. Stein makes the distinction at 98, where he suggests that use of a probability can solve the second problem (though in much of the preceding discussion he assumes that the second problem is significant, e.g. at 76). The problems are analogous, however, because both involve the refinement of a broad class (witnesses, disinterested witnesses) into a narrower one. If probability can solve the second problem, it can solve the first. Stein may draw the distinction in order to avoid an infinite regress in the refinement of generalizations, but that is not a good motivation for the distinction.
20. Foundations at 70.
21. Ibid at 83.
22. Ibid at 100.
23. It is difficult to say whether Stein takes these to be general epistemic principles, or ones confined to adjudicative fact-finding. The latter view of PMI is stressed at 72, but the phrase 'epistemologically unwarranted' to describe a non-weighty verdict at 121 (and similar at 101), as well as the coin tossing examples and the use of the lottery paradox (see below) suggest otherwise. A footnote backing up the claim at 72 fails to enlighten as it reflects confusion about what is involved in a 'partial belief' resolution of the lottery paradox.
24. Foundations at 12.
25. Ibid at 72.
26. Ibid at 69.
27. An excellent discussion of the paradoxes is D. Christensen, Putting Logic in its Place: Formal Constraints on Rational Belief (2004)
28. see also D. Nelkin, 'The Lottery Paradox, Knowledge, and Rationality' (2000) 109 Philosophical Review 373
29. B. Weatherson, 'Can We Do Without Pragmatic Encroachment?' (2005) 19 Philosophical Perspectives 417.
30. The best accounts of this paradox are J.J. Thompson, 'Liability and Individualized Evidence' in Rights, Restitution and Risk: Essays in Moral Theory (1986)
31. and D.T. Wasserman, 'The Morality of Statistical Proof and the Risk of Mistaken Liability' (1991) 13 Cardozo L Rev 935.
32. Stein, confusingly, wavers between the two views of belief: Fact-finders are said to assess probabilities (e.g. at 119), but cf. 74-5, 97-8, 101.
33. cf. Cohen's analysis in Introduction, above n 5, 203-11.
34. Foundations at 83. This is one of the most startling statements in the book. It is difficult to read it without wondering whether Stein believes, or thinks it probable, that the sun will rise tomorrow.
35. Foundations at 82-3. Puzzlingly, the fact-finder can unproblematically 'attach a 50 per cent probability of both heads and tails to each coin throw that belongs to the series' (82).
36. For a similar insight, see A. Hájek, 'Interpretations of Probability' in E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Summer 2003 Edition) http://plato.stanford.edu/archives/sum2003/entries/ probability-interpret/, 3.3.
37. See M. Strevens, 'Inferring Probabilities from Symmetries' (1998) 32 Noûs 231.
38. Foundations, 76 (I will not comment on the problematic notion of 'the long run'). Stein goes on to say that this assumption reflects the indifference principle, despite the fact that the source he cites for the principle (Cohen, Introduction, above n 5 at 47), stresses that the principle does not apply to empirical examples such as the one he discussing. Rather than discussing the indifference principle, in what follows we will attempt to reconstruct Stein's argument in more cogent terms.
39. It is possible that 'the hypothesis' should be read as 'the hypothesis that the ticket will lose', in which case the hypothesis is expected to be falsified on nearly every occasion.
40. Might it make sense if Stein is taken to be talking about chances, i.e. objective features of the world such as, in this case, the propensity of the coin to land heads given the experimental set-up? We can certainly be wrong about chances, as in the case where we do not know that a coin is biased and say that the chance of heads is 0.5. But it does not follow that we are wrong to say that, given what we do know, the probability is 0.5; indeed, it seems that any account of what it means for our probabilities to be objectively correct will relativize correctness to our knowledge of the set up. See H. Beebee and D. Papineau, 'Probability as a Guide to Life' (1997) 94 Journal of Philosophy 217.
41. Cohen, Introduction, above n 5 at 99-100.
42. See A. Hájek, 'Conditional Probability is the Very Guide of Life' in H.E. Kyburg, Jr and M. Thalos (eds), Probability is the Very Guide of Life: The Philosophical Uses of Chance (2003)
43. and 'The Reference Class Problem is Your Problem Too' (2006) Synthese (forthcoming).
44. In the real world, of course, we do have information about selection processes. Such information will be available in some cases from the very content of the witness's testimony: For example, many eyewitnesses will have witnessed the event by chance, and are unlikely to be biased. The same cannot be said of alibi witnesses.
45. T. McGrew, 'Direct Inference and the Problem of Induction' in Kyburg and Thalos, above n 34, 49.
46. See also S. Campbell and J. Franklin, 'Randomness and the Justification of Induction' (2004) 138 Synthese 79.
47. Stein's account of weight allows this: See 34-5, and the discussion of DNA evidence at 86-8.
48. The unlikely does of course happen: See R v Adams [1996] 2 Cr App R 467.
49. It is not certain proof of innocence, because the possibility of genetic mutation cannot be discounted.
50. This skirts over a complication: The example concentrates on a single trial. One of the insights provided by the best evidence principle is that it may be worth running an increased risk of error in this trial by excluding evidence in order to reduce the risk of error in other trials by encouraging parties to produce the best evidence.
51. Bentham, Rationale of Judicial Evidence, quoted in Murphy, above n 5 at 61.
52. See also Beebee and Papineau, above n 32 at 238-42
53. for discussion of why it is rational to seek more evidence and, in this respect, note Ramsey's reference to 'weight' in F.P. Ramsey, 'Weight or the Value of Knowledge' (1990) 41 Brit J Phil Sci 1.
54. Foundations at 123.
55. Ibid at 124.
56. The closest Stein gets to an argument that the exclusion of non-weighty evidence allocates error fairly is at 229-30, where he suggests that the availability of evidence to each party is 'fortuitous: Some litigants have more evidence than others'. But this is to deny that evidence is correlated with truth, a correlation that Stein accepts at 188: 'criminals leave behind more incriminating traces than innocents do ... reasons discrediting the defendant's self-exonerating testimony would predominantly appear in cases featuring guilty defendants'.
57. Unless the evidence is amenable to 'adverse utilization': See below.
58. 'Largely', because it may sometimes be a litigant's fault that evidence is unavailable. Stein has some very interesting ideas about the significance of this: See his elaboration of the 'evidential damage doctrine' in A. Porat and A. Stein, Tort Liability Under Uncertainty (2001) and Foundations, 167-71, 223-5.
59. See also D.A. Nance, 'Evidential Completeness and the Burden of Proof' (1998) 49 Hastings LJ 621.
60. Foundations at 12.
61. It is not clear whether a simple assertion is enough to make a possibility 'evidenced'. Nor is anything said about how this framework deals with proof of subjective mental states, which will typically involve broad generalizations, e.g. 'someone who punches a person in the face intends to do so'. There is some evidence that Stein thinks it would be appropriate to abandon subjective mens rea standards because of proof difficulties: See the discussion of provocation at 3-4.
62. Foundations at 227.
63. Ibid at 71-2.
64. Ibid at 183-4.
65. Confrontation is the idea that criminal defendants should be able to confront their accusers face to face, even if such confrontation is unlikely to enhance fact-finding. It is recognized in the Sixth Amendment to the US Constitution (reinvigorated by Crawford v Washington 124 S Ct 1254 (2004)) and, to some extent, in the ECHR jurisprudence on article 6. One of confrontation's chief proponents suggests that, beyond the instrumental values served by cross-examination, it is difficult to say much other than that confrontation responds to 'something deep in human nature'. See R.D. Friedman, 'The Confrontation Clause Re-Rooted and Transformed' [2004] Cato Sup Ct Rev 439, 442.
66. Foundations at 179, 184.
67. United States v Veysey 334 F 3d 600 (2003).
68. The calculation was based on the implausible assumption that accidental fires are probabilistically independent. There is also the problem that, while a series of accidental fires is improbable, so too is serial arson, thus the meaning of this figure is not intuitively obvious. For discussion of such problems, see A.P. Dawid, 'Bayes's Theorem and the Weighing of Evidence by Juries' in R. Swinburne (ed.), Bayes's Theorem (2002).
69. Perhaps, Stein suggests, Veysey had an enemy who kept burning down his houses. But even under his own account of criminal proof, it is not clear that this implausible scenario is a problem, as it is presumably 'unevidenced', and thus creates only abstract doubts.
70. Worse still, given that PMI applies both to individual items of evidence and to the case as a whole, it looks as though circumstantial evidence - for example, the fact that D was seen close to the scene of crime at the relevant time - will not even be admissible. Having said that, in some examples Stein resorts to a synergy principle, which allows admissibility and proof where 'naked statistical evidence', such as DNA, combines with other evidence, such as a victim's testimony: See 87-8. Mysteriously, however, this principle does not apply to character evidence, which is simply excluded.
71. R v Straffen [1952] 2 QB 911
72. R v Smith (1915) 11 Cr App R 229
73. Makin v AG for New South Wales [1894] AC 57.
74. Ibid.
75. Foundations, 184-6.
76. See R.O. Lempert, S.R. Gross and J.S. Liebman, A Modern Approach to Evidence (3rd edn, 2000), 338-9.
77. In connection with the predictive/trace distinction, Stein cites C.W. Sanchirico, 'Character Evidence and the Object of Trial' (2001) 101 Columbia L Rev 1227, which uses the distinction to give an account of the character evidence rule in terms of the incentive structure provided to prospective offenders. Sanchirico explicitly acknowledges that this framework cannot justify the admissibility of motive evidence, and concludes that motive should therefore be inadmissible.
78. Stein does suggest, at 185, that such evidence is susceptible to individualized testing, but this is not obvious: One can no more test whether a motive was a cause of crime than whether a propensity to commit crime was.
79. See C. Tapper, Cross and Tapper on Evidence (10th edn, 2005), 597-605
80. Lempert et al., above n 61 at 575-7.
81. Foundations at 192.
82. R v Andrews [1987] AC 281, 300-1.
83. (1879) 14 Cox CC 341.
84. Stein makes the additional point (232) that the witness can be cross-examined about all of the events surrounding the hearsay statement, and thus, in the civil paradigm, her evidence is subject to 'adverse utilization' by the opponent because she may contradict elements of the proponent's case. In Bedingfield, for example, she might testify that the husband was not the only person in the room at the time of the attack. This might seem to open up some distance between PMI and the dangers approach but, while interesting, this point is not terribly significant, because whether or not the hearsay is admissible, the witness will surely be called to shed light on the circumstances surrounding the attack.
85. Foundations at 231.
86. Ibid at 234.
87. See Tapper, above n 63 at 641-4.
88. Foundations at 233; see also the discussion of public records at 136-7.
89. Ibid; cf. 229-30. There is, of course, no such guarantee: The most we can say is that we do not expect either claimants or defendants to be significantly disadvantaged in the long run.
90. See, for example, the demand for 'specific' evidence in US v Shonubi 998 F2d 84 (1993).
91. Cf. P. Tillers, 'If Wishes were Horses: Discursive Comments on Attempts to Prevent Individuals from Being Unfairly Burdened by their Reference Classes' (2005) 4 Law, Probability & Risk 33.
92. 'Warner Commutes Death Sentence' Washington Post, 30 November 2005.
93. Though PMI perhaps explains too much: It might well endorse the quashing of Lovitt's conviction, rather than just the commutation of the death sentence.
94. Foundations, 199. It is unclear what 'unavailable' means: Would it apply where a witness has died, or evidence was destroyed because the culprit set fire to the crime scene?
95. There may also be a more complex reason for responding sympathetically to Lovitt. The destruction of the weapon deprives him of an opportunity to establish his innocence (it seems that the weapon should not have been destroyed); and as destruction was the State's fault, the State should not be able to gain an advantage by carrying out the death penalty. Lovitt thus benefits, not from doubts about his guilt, but because the State's actions have reduced his chances of overturning his conviction. There are obvious links here with Stein's ideas about 'evidentiary damage' (see n 46 above), but there does not seem to be any reason to connect this rationale with PMI. Note that Lovitt would not benefit had the weapon decayed through natural processes.
96. P. Slovic, 'Perception of Risk' (1987) 236 Science 280.
97. See also K.J. Heller, 'The Paradox of Circumstantial Evidence' (2006) 105 Michigan LR 241.
98. See B. Weatherson, 'Keynes, Uncertainty and Interest Rates' (2002) 26 Cambridge J Economics 47
99. cf. J.M. Joyce, 'How Probabilities Reflect Evidence' (2005) 19 Philosophical Perspectives 153.
100. See D.H. Mellor, Probability: A Philosophical Introduction (2005) at 90-126.
101. On one account, the probabilities we attach to everyday propositions will be 'extraordinarily vague':R. Swinburne, Epistemic Justification (2001) at 68.
102. See S. Haack, Evidence and Inquiry: Towards Reconstruction in Epistemology (1993). Haack's theory includes a weight-like criterion in her account of comprehensiveness of evidence as an element in justification: See esp. 87-8.
103. P. Achinstein, The Book of Evidence (2001) at ch. 5.