1. J Goudkamp and D Nolan, ‘Contributory negligence in the twenty-first century: an empirical study of first instance decisions’ (2016) 79 MLR 575.

2. There is only a very sketchy and now dated discussion of the question in the classic English study of contributory negligence, GL Williams, Joint Torts and Contributory Negligence: A Study of Concurrent Fault in Great Britain, Ireland and the Common‐Law Dominions (London: Stevens & Sons, 1951) pp 480–484. For brief surveys of appellate review in comparative fault cases in the United States, see H Woods and B Deere, Comparative Fault (Deerfield, IL: Clark Boardman Callaghan, 3rd edn, 1996) ch 21; and V Schwartz, Comparative Negligence (New Providence, NJ: LexisNexis, 5th edn, 2010) pp 415–422.

3. The exception is Nixon v Thames Water Utilities Ltd [2006] CLY 2913, a decision of Romford County Court in the exercise of its appellate jurisdiction. The only decisions of the House of Lords or Supreme Court on contributory negligence in the relevant time period were two Scottish appeals, Robb v Salamis (M&I) Ltd [2006] UKHL 56, [2007] 2 All ER 97 and Jackson v Murray [2015] UKSC 5, [2015] 2 All ER 805, which were not included because our study was limited to decisions from England and Wales.

4. See G Drewry, L Blom‐Cooper and C Blake, The Court of Appeal (Oxford: Hart Publishing, 2007) p 2. This book is itself the most notable exception to the more general academic neglect that its authors highlight, though see also BM Atkins, ‘Interventions and power in judicial hierarchies: appellate courts in England and the United States’ (1990) 24 Law & Soc'y Rev 71; BM Atkins, ‘Party capability theory as an explanation for intervention behavior in the English Court of Appeal’ (1991) 35 Am J Polit Sci 881.

5. We are unconcerned in the present article with the justifiability of the law of contributory negligence, with whether the scope of the doctrine is appropriate, or with the desirability of appellate intervention in the contributory negligence context generally or in specific cases. Instead, the present study simply addresses the manner in which appellate courts apply the doctrine of contributory negligence.

6. Law Reform (Contributory Negligence) Act 1945, s 1(1).

7. Badger v Ministry of Defence [2005] EWHC 2941 (QB), [2006] 3 All ER 173 [16] (Stanley Burnton J).

8. Stapley v Gypsum Mines Ltd [1953] AC 663, 682 (Lord Reid).

9. CPR 52.3(1).

10. CPR 52.3(2).

11. CPR 52.3(6).

12. Tanfern Ltd v Cameron‐MacDonald [2000] 1 WLR 1311 [21] (Brooke LJ).

13. CPR 52.11(1).

14. CPR 52.11(3).

15. ‘[T]here is no single standard which is appropriate to every case’: South Cone v Bessant [2002] EWCA Civ 763 [26] (Robert Walker LJ).

16. Adrian Zuckerman, Zuckerman on Civil Procedure: Principles of Practice (London: Sweet & Maxwell, 3rd edn, 2013) para 24-197.

17. EI Du Pont de Nemours & Co v ST Dupont (Note) [2003] EWCA Civ 1368, [2006] 1 WLR 2793 [94] (May LJ).

18. Grant v Sun Shipping Co Ltd [1948] 2 All ER 238, 242 (Lord Porter), 246 (Lord du Parcq).

19. Ibid, 246 (Lord du Parcq).

20. Kerley v Downes [1973] RTR 188, 193 (Edmund Davies LJ).

21. See n 3 above.

22. Jackson [28].

23. Ibid, [35].

24. Ibid.

25. In 1999, the copyright entitlement over the republication of reserved judgments of the High Court and Court of Appeal previously held by court shorthand writers came to an end, and thereafter it would appear that most such judgments were uploaded onto the relevant online systems: see M Zander, The Law‐Making Process (Oxford: Hart Publishing, 7th edn, 2015) p 302. While reserved judgments of the Court of Appeal are supposed to be uploaded onto these systems (ibid, p 273), the same is not true of ex tempore judgments. During the study period, it appears that between one‐half and two‐thirds of judgments of the Civil Division of the Court of Appeal were reserved (ibid).

26. We did not treat actions brought against an employee and his or her employer as involving multiple defendants since, in reality, the employer is the only defendant in such cases. Nor, in the case of Sykes v Harry [2001] EWCA Civ 169, [2001] QB 1014, did we treat an action against a defendant and his trustee in bankruptcy as involving multiple defendants, as again there is in effect only one defendant in such a case.

27. There were two cases involving multiple claimants, namely Goddard v Greenwood [2002] EWCA Civ 1590, [2003] RTR 10 (two claimants) and Minh Lac v Clayton [2009] EWCA Civ 106 (three claimants). There were four cases involving two defendants, namely Slack v Glenie 2000 WL 544172; Cook v Thorne [2001] EWCA Civ 81; Buyukardicli v Hammerson UK Properties Plc [2002] EWCA Civ 683; and Wells v Mutchmeats Ltd [2006] EWCA Civ 963.

28. There were eight such claims, namely Slack v Glenie 2000 WL 544172 (two claims); Watson v Skuse [2001] EWCA Civ 1158; Anderson v Newham College for Further Education [2002] EWCA Civ 505, [2003] ICR 212; Lunt v Khelifa [2002] EWCA Civ 801; Home Office v Lowles [2004] EWCA Civ 985; West Sussex CC v Russell [2010] EWCA Civ 71, [2010] RTR 19; and Phethean‐Hubble v Coles [2012] EWCA Civ 349, [2012] RTR 31. The existence of cross‐appeals was also made apparent by records for variables S–V in our dataset (see text following n 41 below for descriptions of these variables).

29. See, eg, Smith v Bailey [2014] EWHC 2569 (QB), [2015] RTR 6 (where the High Court was exercising its appellate jurisdiction to review interim orders made by a Master of the Queen's Bench Division).

30. See, eg, Goose v Wilson Sandford (No 2) [2001] Lloyd's Rep PN 189; Horner v Norman [2015] EWCA 1055; Rollinson v Dudley MBC [2015] EWHC 3330 (QB).

31. See, eg, Co‐operative Group (CWS) Ltd v Pritchard [2011] EWCA Civ 329, [2012] QB 320.

32. See, eg, Standard Chartered Bank v Pakistan National Shipping Corpn (Nos 2 and 4) [2002] UKHL 43, [2003] 1 AC 959.

33. See, eg, Vann v Ocidental‐Companhia De Seguros SA [2015] EWCA Civ 572 (Portuguese law).

34. See, eg, Sykes v Harry [2001] EWCA Civ 167, [2001] QB 1014.

35. See, eg, Tasci v Pekalp of London Ltd [2001] ICR 633.

36. Defined as claims in which the defendant was sued in their capacity as the claimant's employer. This was extended to relationships akin to employment: see, eg, Mullaney v Chief Constable of the West Midlands [2001] EWCA Civ 700, [2001] Pol LR 150 (police officer suing Chief Constable).

37. Defined as claims in which the defendant was sued in their capacity as the occupier of premises.

38. Defined as claims in which the defendant was sued in their capacity as a public authority.

39. If an employer, occupier or public authority was sued in that capacity, the claim was always classified as a type (2), (3) or (4) claim respectively, even if (for example) the claim also involved a road accident. This decision was made because our sense from reading the decisions in the (relatively uncommon) claims concerned was that the capacity in which the defendant was sued was the best and most straightforward guide as to how the claim in question should be categorised.

40. There were no claims in the dataset where the claimant was suing for more than one type of damage.

41. Robinson v Midland Bank plc (CA, 27 October 2000) was the only decision in the dataset in which this occurred.

42. CTO Gesellschaft Fur Containertransport MBH and Co v Dziennik [2006] EWCA Civ 1456.

43. See text to nn 26–28 above.

44. As can be seen, for example, by comparing the numbers of claims in the various contextual categories (Figure 1, below) with the overall number of appeals in those categories (Figure 3, below).

45. See n 3 above, regarding the exclusion from the study of appeals to the ultimate appellate level from Scotland.

46. Owners of the Selat Arjuna v Owners of the Contship Success [2000] 1 All ER (Comm) 905.

47. Collins‐Williamson v Silverlink Train Services Ltd [2009] EWCA Civ 850.

48. St George v Home Office [2008] EWCA Civ 1068, [2009] 1 WLR 1670.

49. Chubb Fire Ltd v Vicar of Spalding [2010] EWCA Civ 981, [2010] 2 CLC 277.

50. See text at nn 95–103 below.

51. Regarding this distinction between claims and appeals, see the text to nn 43–44 above.

52. See the text to nn 21–24 above.

53. We should be clear about what we mean by these statements. When we say, for example, that the appellate court overturned the determination of the first instance court regarding guilt in x% of cases, we mean that the appellate court acceded to a request to overturn the determination of the first instance court regarding guilt in x% of the cases in which it was asked to do so. See further the descriptions of variables O and P of our dataset (text following n 41 above), from which results on intervention rates were generated.

54. The intervention rate on guilt in occupiers' liability cases was zero, which explains the presence of a line rather than a bar in the relevant section of Figure 4.

55. While intervention regarding guilt was more likely than intervention in connection with apportionment in road accident claims (50% and 44% respectively), this was not true in employers' liability claims (rounding up to 38% in both cases).

56. Logistic regression analysis showed that when employers' liability claims were compared with road accident claims, the odds of appellate intervention were lower in the former, but not significantly so (p = 0.449). More generally, a Pearson's chi‐square test on the null hypothesis that the appellate intervention rate is the same across all types of claim returned a relatively high p‐value of 0.299 (suggesting no strong association between intervention rate and claim type at a general level). A p‐value is the probability of obtaining an effect at least as extreme as the one in the sample data, assuming the truth of the null hypothesis. Hence, the lower the p‐value, the more likely it is that the null hypothesis is false. The p‐value in this instance is quite high, which suggests that the disparity in the interference rate in issue may well be attributable to chance (in the form of the sample of claims we happened to survey).

57. Logistic regression analysis showed that when occupiers' liability claims were compared with road accident claims, the odds of appellate intervention were substantially lower in the former (p = 0.055).

58. Another investigator obtained a similar result in the context of a study of all decisions of the Court of Appeal handed down between 1983 and 1985: Atkins (1991), above n 4, p 894.

59. A Pearson's chi‐square test on the null hypothesis that the appellate intervention rate with regard to guilt is the same for both types of first instance court returned a p‐value of 0.493. Running the same test with regard to appellate intervention with respect to apportionment also yielded a high p‐value (p = 0.512).

60. See the text following n 52 above.

61. A Welch two‐sample t‐test comparing the average increase in discount with the average reduction in discount returned a p‐value of 0.032. This result was broadly consistent with the result of a Wilcoxon rank sum test with continuity correction (p = 0.04).

62. In one employers' liability case, Toole v Bolton MBC [2002] EWCA Civ 588, the Court of Appeal found that there had been no contributory negligence at all, when at first instance a 75% discount had been imposed for contributory fault.

63. Eyres v Atkinsons Kitchens & Bedrooms Ltd [2007] EWCA Civ 365, 151 SJLB 576 (where the Court of Appeal increased the discount from one‐quarter to one‐third).

64. It is common for investigators to organise data into four equally sized bands according to their distribution. The ‘quartiles’ are the three values that divide the data into these four bands.

65. A Pearson's chi‐square test on the null hypothesis that the post‐appeal success rate of the plea of contributory negligence is the same across all types of claim returned a p‐value of 0.083.

66. Logistic regression analysis showed that when claims in the ‘other’ category were compared with road accident claims (the largest category of claim), the odds of the plea of contributory negligence succeeding post‐appeal were significantly lower in ‘other’ claims (p = 0.01).

67. Logistic regression analysis revealed that when professional negligence claims were compared with road accident claims, the odds of the plea of contributory negligence succeeding post‐appeal were lower in professional negligence claims, but not significantly so (p = 0.304).

68. See Goudkamp and Nolan, above n 1, p 593.

69. Logistic regression analysis revealed that when employers' liability claims were compared with road accident claims, the odds of the plea of contributory negligence succeeding post‐appeal were lower in employers' liability claims (p = 0.114).

70. Goudkamp and Nolan, above n 1, pp 593–594. Leniency towards employees was, however, manifested at the apportionment stage: ibid, p 597.

71. Ibid, p 594.

72. A Welch two‐sample t‐test of the difference between the average discount in the first instance study and the average post‐appeal discount in the current study gave a very low value (p < 0.001). This result was consistent with the result of a Wilcoxon rank sum test with continuity correction (p < 0.001).

73. In the claims in the appellate study dataset, the average discount at first instance was 52%, which is (again) considerably higher than the average discount in our first instance study of 40.5%.

74. A Welch two‐sample t‐test comparing the average post‐appeal discounts in road accident and employers' liability claims returned a p‐value of 0.051, although the p‐value rose to 0.135 when we ran a Wilcoxon rank sum test with continuity correction. The p‐values for the same tests comparing the average post‐appeal discounts in employers' liability and professional negligence claims were 0.099 and 0.033 respectively. However, when the average post‐appeal discounts in road accident and professional negligence claims were compared using the same tests, the p‐values were considerably higher (0.225 and 0.278 respectively).

75. See Goudkamp and Nolan, above n 1, pp 595–598.

76. The average discount in such cases at first instance (51%) was some 10.5% higher than the overall average discount of 40.5% (ibid, p 597). By contrast, in the current study the disparity was a mere 3% (53% as against 50%).

77. See Goudkamp and Nolan, above n 1, p 599.

78. C (a child) v Imperial Design Ltd [2001] Env LR 33; Paramasivan v Wicks [2013] EWCA Civ 262.

79. Palfrey v WM Morrisons Supermarkets Plc [2012] EWCA Civ 1917.

80. Judges have often emphasised that the fact that the claimant is a child is a particularly relevant consideration in the contributory negligence context: see, eg, Gough v Thorne [1966] 1 WLR 1387.

81. Pearson's chi‐square test with Yates's continuity correction, p = 0.017.

82. In those three claims, the appellate court intervened with respect to guilt in one claim (Osei‐Antwi v South East London & Kent Bus Co [2010] EWCA Civ 132), and refused to do so in the other two (Stanton v Collinson [2010] EWCA Civ 81, [2010] RTR 26; Phethean‐Hubble v Coles [2012] EWCA Civ 349, [2012] RTR 31). The equivalent figures in claims brought by adults were 18 and 23 respectively.

83. Pearson's chi‐square test with Yates's continuity correction, p = 0.774. Logistic regression (controlling for claim type and gender) returned a p‐value of 0.767.

84. The results were as follows: 10–19 years, 79% post‐appeal success rate; 20–29 years, 83%; 30–39 years, 75%; 40–49 years, 78%; 50+ years, 75%.

85. Logistic regression analysis comparing the odds of a finding of contributory negligence post‐appeal where the claimant was in the 10–19 age range with the odds of such a finding in the two other age ranges in which there were substantial numbers of claimants returned very high p‐values (20–29 years: p = 0.759; 40–49 years: p = 0.964).

86. A Welch two‐sample t‐test of the difference between the average post‐appeal discounts in child claimant and adult claimant claims returned a p‐value of 0.298. This result was broadly consistent with the result of a Wilcoxon rank sum test with continuity correction (p = 0.342). Linear regression analysis (controlling for case type and gender) returned a p‐value of 0.468.

87. See Goudkamp and Nolan, above n 1, pp 603–604.

88. Linear regression analysis of the change in the discount with age as a continuous variable returned a p‐value of 0.05.

89. Linear regression analysis of the change in the discount with age as a continuous variable (controlling for case type) returned a p‐value of 0.114.

90. Goudkamp and Nolan, above n 1, p 604.

91. See Goudkamp and Nolan, above n 1, p 604. The results of a study of over 150 published articles regarding gender differences in the propensity to run risks gave ‘clear support’ to the conclusion that male participants were more likely to take risks than female participants: JP Byrnes, DC Miller and WD Schafer, ‘Gender differences in risk taking: A meta-analysis’ (1999) 125 Psychol Bull 367, 377.

92. Pearson's chi‐square test with Yates's continuity correction, p = 0.029.

93. However, statistical analysis suggests that this disparity may well be attributable to chance: Pearson's chi‐square test with Yates's continuity correction, p = 0.268.

94. Success rate of the plea post‐appeal and claimant gender: Pearson's chi‐square test with Yates's continuity correction, p = 0.667. Average discount post‐appeal and claimant gender: Welch two‐sample t‐test, p = 0.718.

95. It would have been interesting to know whether the decline in the number of appeals was specific to contributory negligence appeals or was true of appeals generally. However, while data are available on the number of appeals determined by the Court of Appeal (Ministry of Justice, The Royal Courts of Justice Annual Tables: Supplementary Data for Additional Courts (London: Ministry of Justice, 2015) Table 3.9 (https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/527027/rcj‐tables.xlsx, accessed 31 July 2016), unfortunately these only date back to 2003.

96. We are grateful to Adrian Zuckerman and Andrew Higgins for sharing their thoughts with us on these points.

97. See the text to nn 9–12 above.

98. See Smith v Cosworth Casting Processes Ltd [1997] 1 WLR 1538, 1538 (Lord Woolf MR), explaining that, at the time, the Court of Appeal could grant leave to appeal even if it was not satisfied that the applicant had a realistic prospect of success. Cf the current test for permission to appeal: text to n 11 above.

99. In our dataset, the average number of days that passed between the date of the first instance decision and the date of the appellate decision where both dates were known was 338. The equivalent figure for claims in our first instance study that we knew had been the subject of an appeal was 328 days: Goudkamp and Nolan, above n 1, p 583 n 43.

100. Access to Justice Act 1999, s 27.

101. In 2000–2003 there was an average of 5.5 claimant appeals per year in our dataset, and an average of 8.25 defendant appeals. From 2004–2015 there was an average of 1.9 claimant appeals per year, and an average of 2.6 defendant appeals.

102. See the onerous test in CPR 52.13. This change was described as ‘major’ in Tanfern Ltd v Cameron‐Macdonald [2000] 1 WLR 1311 [42] (Brooke LJ).

103. M Burton (ed), Civil Appeals (London: Sweet & Maxwell, 2nd edn, 2013) para 1-300.

104. S Deakin, A Johnston and B Markesinis, Markesinis and Deakin’s Tort Law (Oxford: Oxford University Press, 7th edn, 2013) p 756.

105. T Weir, A Casebook on Tort (London: Sweet & Maxwell, 10th edn, 2004) p 247.

106. According to the Ministry of Justice, 48% of all appeals heard by the Court of Appeal between 2003 and 2015 were allowed (disregarding appeals that were dismissed by consent, struck out or disposed of by other means): Ministry of Justice, above n 95, Table 3.9. In an earlier study, Atkins found that the Court of Appeal intervened on average in 35% of appeals in all types of case decided between 1952 and 1983, with ‘a remarkable degree of consistency through’ this period: Atkins (1990) above n 4, p 83.

107. See the text to nn 21–24 above.

108. According to statistics published by the Ministry of Justice, in the period 2008–2015 the rate of appellate intervention with respect to decisions of the Queen's Bench Division of the High Court was 40%. The corresponding intervention rate for decisions from county courts (excluding appeals in family and admiralty matters) was 47%. See Ministry of Justice, above n 95, Table 3.9. Accordingly, while in our study the rate of appellate intervention with respect to High Court decisions was similar to the intervention rate in relation to such decisions generally, the intervention rate regarding county court decisions was somewhat lower.

109. See Figure 5 above.

110. See the text to nn 63–64 above.

111. See the text to nn 21–24 above.

112. Jackson [38].

113. For this suggestion, see ibid.

114. See Table 1 above.

115. Goudkamp and Nolan, above n 1, p 599.

116. Ibid, pp 611–612.

117. See Butcher v Cornwall CC [2002] EWCA Civ 1640 [13] (Sedley LJ); Cooper v Carillion plc [2003] EWCA Civ 1811 [11] (Keene LJ).

118. See Toole v Bolton MBC [2002] EWCA Civ 588 [13] (Buxton LJ).