R v Crupi (1): Forensic gait analysis comes to Australia (and is excluded)

The Supreme Court of Victoria (VSC) recently considered the admissibility of an expert opinion based on forensic gait analysis (using the study of the way people walk to identify them, among other things). Surprisingly, the VSC excluded the opinion. In this blog, I will provide a little more background and then discuss the decision (R v Crupi [2020] VSC 654).

Forensic gait analysis

Forensic gait analysis encompasses many tasks. One of these is looking at at the way an unidentified person walks and comparing that to the way a known person walks, and using that comparison to identify the person or say that they walk in a similar way.

There’s no scientific foundation for this practice. There have not been sufficiently large controlled studies to demonstrate that practitioners can do this task (i.e., to determine if they are correct when they say the two walkers are the same person, or that when they say they could be the same person, that they are).

Emma Cunliffe and Gary Edmond have written about the problems with forensic gait analysis (here and here).

I have written about it as well (here and here).

This type of evidence has been admitted in the UK and in Canada (see R v Aitken, which cites the UK decisions).

Australia has notoriously weak rules when it comes to scientific evidence (Uniform Evidence Law jurisdictions, which include the two largest states, do not consider the reliability of evidence when deciding whether to admit it, except perhaps in very extreme circumstances).

So, this decision was surprising.

R v Crupi

The background

Crupi was accused of murder. There was CCTV footage of someone that could been him hanging around the crime scene of the night of the murder and also a few days before.

The opinion

A University of Melbourne professor, Marcus Pandy, who studies human gait biomechanics, provided the opinion. He examined the CCTV footage mentioned above and known footage of Crupi and said:

While it cannot be concluded with any certainty that a subject appearing in any of the video clips 1–36 is the same as an individual appearing in any of the clips 38(sic)–95, the data clearly indicate that the subjects appearing in the two groups of video clips have similar spatiotemporal gait patterns…

And:

As regards the age of the subjects in the clips, Professor Pandy indicated that the mean step lengths and walking speeds of the subjects were ‘closely similar’ to results from a study of 26 older persons (mean age 66) studied by Judge et al in 1996 and dissimilar to results from a study of 16 healthy young adults studied by Besier et al in 2009. He concluded that …the subjects observed on the video clips walked with a gait pattern more consistent with that exhibited by older adults, who take shorter steps and walk more slowly than healthy young adults.

And:

Other opinions expressed by him were that: The spatio-temporal gait parameters of the subjects in the two groups of clips were closely similar; The mean step lengths of the subjects in the two groups of clips were remarkably similar; The mean step frequencies of the subjects in the two groups of clips were quite similar; The mean walking speeds of the subjects in the two groups of clips were quite similar; The gait of the subjects in the two groups of clip was more consistent with the gait of older persons, as was the degree of toe-out gait of the subjects.

Some of the reasons for this opinion were the speed at which the person was walking around (based on calculations made through the distances walked over time, using Google Earth distances and police measurements) and the way in which the person walked with their toes abducted out (which Pandy said was more extreme than is normal for healthy adults).

Pandy was fairly modest (especially as compared to experts in other cases), saying he had never done this before:

Professor Pandy said he had never before been asked to compare video recordings of subjects to establish whether they were similar or dissimilar. He said there are people with such expertise but he does not claim to be an expert in that field. He said he has never done any research regarding the potential pitfalls in making such a comparison. He said he was not familiar with the concept of confirmation bias

Should this have been admitted?

As noted above, I think it’s great that Pandy was modest about the knowledge he had and that he didn’t provide a very strong identification.

That said, I agree with the decision to exclude it. This is because we just don’t have any grounds beyond Pandy’s sayso to know how likely it is that his assessment is correct. When he says two people in videos share the same ‘spatiotemporal gait pattern’, how often is that right and how often is that wrong? How often does walking speed as judged from short video clips accurately predict the age of the walker? How often is walking speed correctly matched to the walking speed in another video? And, assuming some people can do this with some degree of reliability, can Pandy? Without knowing this, there is no way to assess his opinion. It’s useless.

Why was it excluded?

The VSC went a lot farther than I expected in excluding this decision.

First, it found the decision wasn’t relevant, and apparently did so because of a complete lack of reliability. This pretty astounding because courts typically assume reliability when determining relevance (for more context see the decision excerpted below):

What factors may have affected the reliability of Professor Pandy’s calculations? He acknowledged that he sometimes rounded-off the number of steps taken between point A and point B in a given clip but he did not keep a record of when he did so. When asked about the number of times he did this, he first said ‘infrequently’, then he said ‘no more than 20% of the time’ which strikes one as a significant number of occasions. If those occasions were spread across both groups they may have evened themselves out, so to speak, but Professor Pandy did not say positively that was the case. On the evidence, one cannot exclude the possibility that the rounding-offs were concentrated in one group. Professor Pandy said he sometimes had to make a subjective judgment as to the number of steps taken by the subject between points A and B. He acknowledged that the CCTV frame rates, about which he said he knew nothing, may have affected the accuracy of those subjective judgments. Professor Pandy used the time stamps on the CCTV clip to make his calculations but he did not check the accuracy of the various time stamps for the various CCTV cameras. Professor Pandy also acknowledged that the subjects did not always walk in a straight line between points A and B: at times, Professor Pandy noted excessive lateral movement by the subjects.

It may well be, as P submitted, that none of these matters significantly affected the accuracy of Professor Pandy’s calculations but, on the present state of the evidence, there is no way of knowing whether that is the case. If the jury is to act rationally, there must be a proper basis for the jury to conclude that the accuracy of his calculations were not significantly affected by these matters. The evidence fails to provide a proper basis for such a conclusion.

Another significant issue with Professor Pandy’s evidence is the difficulty in concluding rationally that he was “comparing apples with apples”. He conceded in cross-examination that there was no way of knowing whether the subjects in clips 1–95 were walking at their preferred speed. And yet he based his opinion that the subjects had gait patterns more consistent with older persons primarily on a comparison of the data with results obtained in gait studies of young and old adults whom it was known were walking at their preferred speed.

This is remarkable. The court isn’t just saying, we don’t have evidence that he’s any good at this or that this procedure works. Rather, it is looking at the method itself and saying there are all kinds of inaccuracies and flaws in it. It isn’t just saying it can’t tell if the method works or not, it’s saying that based on all these flaws, it’s unlikely it does.

Section 79

The VSC then went on to say not only is this not relevant, but it fails section 79 (the expert evidence rule/exception) because it wasn’t convinced Pandy was qualified to provide the opinion; recall, he hadn’t done it before. The court didn’t seem to doubt whether forensic gait analysis is a field of specialized knowledge or that this practice constituted specialized knowledge. I would say it’s not.

The Tendency Rule and Section 137

Finally, the Court said the evidence should also be excluded for being tendency evidence (without substantial probative value) and that 137 also applied due to its low probative value and high risk of prejudice. As to the risk of prejudice, it said a jury may misuse the opinion by conflating ‘remarkably similar’ for an identification.

Conclusion

I’m not entirely sure this decision was legally correct. Courts have been pretty firm in saying that trial judges should assume the jury will find evidence is reliable when determining its relevance and probative value.

That said, I think it’s wonderful that this court took a hard look at the practices being employed here and found they were simply messy and untrustworthy. This is especially remarkable given how often precedent is influential in guiding expert evidence decisions (although it shouldn’t be). It takes real determination to exclude a form of evidence that other courts have given their stamp of approval to. In other words, it’s hard to not follow those courts’ footsteps (I apologize for this sentence).

Note: much of the unformatted text of the Crupi judgment is excerpted below because that context is essential for understanding my analysis (as well as advancing the principles of open justice).

R v Crupi [2020] VSC 654

BACKGROUND

1 This is a ruling about the admissibility of evidence of forensic gait analysis (FGA). Such evidence has been admitted in the UK1 and Canada2 but, to my knowledge, this is the first time its admissibility has been the subject of a ruling in Australia.

2 All references to parts and sections below are to the Evidence Act 2008, unless otherwise indicated.

3 Vincenzo Crupi (D) is charged with the murder of Joseph Acquaro (V). V was fatally shot at close range near his Brunswick East restaurant, the Gelo Bar, shortly before 1:00am on 15 March 2016.

4 The prosecution (P) allege that D was the shooter. This is disputed and is the central issue in the case.

5 P’s case that D was the shooter is a circumstantial one. In brief, it includes the following evidence in addition to the impugned evidence:

• Evidence of D’s animus towards V, giving him a motive to murder V; CCTV footage of a person similar in appearance and gait to D conducting surveillance of the Gelo Bar on the night of 11–12 March 2016;

• Evidence that D was identified by witness Reardon as the person conducting surveillance on the Gelo Bar that night;

• CCTV footage of a vehicle very similar to D’s vehicle being in the vicinity of the Gelo Bar on the night of 11–12 March 2016;

• CCTV footage of a person similar in appearance and gait to D near the crime scene on the night of 14–15 March 2016, the night of the murder;

• CCTV footage of a vehicle very similar to D’s vehicle being in the vicinity of the Gelo Bar on the night of 14–15 March 2016 and, after the shooting, travelling to a place near D’s home;

• Alleged incriminating conduct by D post offence (disposal of a jacket);

• Evidence of alleged gunshot residue on a cap found in D’s car;3

• Alleged lies told by D to investigators.

6 P’s summary of prosecution opening begins with what it calls an Executive Summary of the evidence briefly outlined above. It is convenient to include that Executive Summary as an annexure to this ruling.4

7 Before turning to the impugned FGA evidence, it is appropriate to say something more about the identification evidence of Mr Reardon.

8 On the night of 11–12 March 2016, Mr Reardon, who worked at a fish and chip shop near the Gelo Bar, noticed a man in Lygon St who appeared to be keeping watch on the Gelo Bar. Mr Reardon was concerned that the man might have been stalking someone. Just before midnight, Mr Reardon approached the man, who walked off. Around 12:45am, the man returned, wearing a different top, and appeared to resume surveillance of the Gelo Bar. Mr Reardon approached him again and, this time, spoke to him briefly. The man told Mr Reardon he was waiting on a friend and walked off. Both occasions where Mr Reardon approached the man were recorded on CCTV.5

9 On the 21 March 2016, police showed Mr Reardon a photoboard. He identified a photograph of D as the person he had seen watching the Gelo Bar on the night of 11–12 March 2016.

The impugned FGA evidence

10 The impugned evidence the subject of this ruling is the FGA evidence of Professor Marcus Pandy, whose academic qualifications and output are impressive. Professor Pandy did his Ph.D in human gait biomechanics and has been the Chair of Mechanical and Biomedical Engineering at the University of Melbourne since 2005.

11 In 2016, the police supplied him with 95 CCTV clips for analysis. Professor Pandy was, in his own words, asked to ‘identify and document any physical characteristics observed in the video clips.’6

12 Professor Pandy was not informed that: clips 1–36 were of the person allegedly conducting surveillance of the Gelo Bar on the night of 11–12 March 2016; clips 37–94 were allegedly of the shooter on the night of 14–15 March 2016, both before and after the shooting; and clip 95 was of D (his face redacted), walking along a corridor on 18 March 2016.7

13 For his first report dated 23 November 2016, Professor Pandy measured certain spatio-temporal features of the gait of the subjects in the clips by using a Google Earth measurement tool and only those clips that he considered suitable for the purpose. The gait features that Professor Pandy measured were step length, step frequency and walking speed. According to Professor Pandy:

Step length was calculated by dividing the distance walked by the number of steps taken by the subject in each clip. Walking speed was found by dividing the distance walked by the elapsed time noted on the video recording. Step frequency then was calculated by dividing walking speed by step length, as v = s.f, where v is walking speed in metres/sec; s, step length in metres; and f, step frequency in s-1 or Hz.8

14 Professor Pandy found that the step lengths of the subjects were consistently around the same value. Mean step length was calculated to be 0.64m with a standard deviation (SD) of 0.04m or 4cm. Mean step frequency was calculated to be 1.68hz (SD 0.14hz). Mean walking speed was calculated to be 1.07m/s (SD 0.13m/s).

15 Other features which Professor Pandy noted about the subjects but did not measure, included the toe-out gait (abduction) of the subjects. Professor Pandy opined that the degree of abduction that he observed in a number of clips was larger in both feet than is normal for healthy young adults.

16 In his first report, Professor Pandy did not express any opinion as to whether or not the subjects of the clips were the same person.

17 Professor Pandy compared the spatio-temporal data from the clips with data from spatio-temporal studies done on the gait of young and old (65+) he althy adults, walking at their ‘preferred speed’.9 Professor Pandy opined that the mean step length and mean walking speed of the subjects were more consistent with the subjects being elderly, as was the degree of toe-out gait. I note that D, whose date of birth is 11 March 1949, was aged 67 at the time V was murdered.

18 The police were more prescriptive in commissioning Professor Pandy’s second report dated 21 January 2018. The police asked Professor Pandy to say whether: …a person recorded in any of the CCTV footage contained in clips 1 to 36 is the same as a person recorded in any of the CCTV footage contained in clips 37 to 95.10

19 Police also provided precise measurements between various geographical features shown in the clips so that Professor Pandy could more accurately calculate the relevant spatio-temporal gait features of the subjects, rather than relying on the Google Earth measuring tool.1120 In re-doing his calculations, Professor Pandy utilized 13 clips from clips 1–36;12 26 clips from clips 37–94;13 and divided clip 95 into parts A and B.

21 Step length for the clips selected from clips 1–36 ranged from 0.55m to 0.73m. The mean step length was 0.60m (SD 0.05m). Step length for clips selected from clips 37 to 95 ranged from 0.52m to 0.71m. The mean step length was 0.61m (SD 0.05m). Professor Pandy commented that the mean step lengths for the subjects of the two groups of clips were ‘remarkably similar’14 with the difference being only 1cm, less than 2%.15

22 Step frequency for clips selected from clips 1–36 ranged from 1.42hz to 1.85hz with a mean step frequency of 1.65hz (SD 0.15hz). Step frequency for clips selected from clips 37–95 ranged from 1.40hz to 2.14hz with a mean of 1.80hz (SD 0.19hz). Professor Pandy remarked that the mean step frequencies for the two groups of clipswere ‘quite similar’ with a difference of 9%. He said the absolute difference may be considered ‘relatively small’.16

23 Walking speed for clips selected from clips 1–36 ranged from 0.78m/s to 1.09m/s with a mean walking speed of 0.98m/s (SD 0.11m/s). Walking speed for clips selected from clips 37–95 ranged from 0.74m/s to 1.36m/s with a mean walking speed of 1.11m/s (SD 0.17m/s). Professor Pandy remarked that the mean walking speeds for the two groups of clips were ‘quite similar’17 with a difference of 13%.18 He said the absolute difference in mean walking speed (0.13m/s) may be considered ‘relatively small’.19

24 As he did in his first report, Professor Pandy noted that in some clips the subjects walked with toe-out gait larger than normal for healthy young adults but, again, he did not measure the degree of abduction.

25 In relation to whether any of the subjects in clips 1–36 were the same as the subjects in clips 37–95, Professor Pandy said this:20 While it cannot be concluded with any certainty that a subject appearing in any of the video clips 1–36 is the same as an individual appearing in any of the clips 38(sic)–95, the data clearly indicate that the subjects appearing in the two groups of video clips have similar21 spatiotemporal gait patterns…

26 As regards the age of the subjects in the clips, Professor Pand y indicated that the mean step lengths and walking speeds of the subjects were ‘closely similar’ to results from a study of 26 older persons (mean age 66) studied by Judge et al in 1996 anddissimilar to results from a study of 16 healthy young adults studied by Besier et al in 2009. He concluded that …the subjects observed on the video clips walked with a gait pattern more consistent with that exhibited by older adults, who take shorter steps and walk more slowly than healthy young adults.22

27 Professor Pandy gave oral evidence on 18 and 19 May 2020 at a hearing conducted pursuant to s 198B of the Criminal Procedure Act 2009.

28 He said this was the first time he had ever given FGA evidence.23

29 Professor Pandy said he had never before been asked to compare video recordings of subjects to establish whether they were similar or dissimilar.24 He said there are people with such expertise but he does not claim to be an expert in that field. 25 He said he has never done any research regarding the potential pitfal ls in making such a comparison.26 He said he was not familiar with the concept of confirmation bias.27

30 Professor Pandy said he was not a statistician but that he had sufficient knowledge of statistics to calculate the standard deviations referred to in his report.28

31 He agreed there was no way of assessing whether the subjects in any of the clips were walking at their preferred speed.29 Professor Pandy said that many studies indicate that preferred walking speed for healthy young adults is between 1.3 to 1.5 m/s.30

32 In working out his spatio-temporal calculations for his second report, Professor Pandy said he sometimes had to round up or round down the number of steps taken between points A and B.31 He did not keep a record of when he did this but he said he did it ‘infrequently … no more than 20% of the time’.32 Professor Pandy said he sometimes had to make a judgment as to when a subject’s heel struck the ground.33

33 Professor Pandy agreed that distances covered by the subjects were measured from point A to point B but that the subjects did not always walk in a straight line between those two points: Professor Pandy sometimes observed excessive lateral movement in the gait of the subjects.34

34 Professor Pandy said he had never previously been asked to compare CCTV clips of subjects.35 He said he did not know anything about the make or model of the CCTV cameras capturing the images or whether or not they distorted the imagery. 36 He said he had no information about the frame rates of the various CCTV clips. 37 He agreed that the frame rates could affect the accuracy of his judgments as to when the heel of the subjects struck the ground.38

35 Professor Pandy said he could not shed any light on the statistical significance of his findings compared with the population at large.39 When he said in his second report that the mean step lengths of the subjects in the two groups of clips were ‘remarkably similar’, he was not comparing mean step length with any population data.40

36 Professor Pandy said that in his reports he did not say the subjects were old.41 He said ‘you can’t determine anything about age’.42

37 In relation to the toe-out gait of the subjects, Professor Pandy said both feet of the subjects exhibited toe-out gait. He said he did not make measurements of the toe-out gait43 but it appeared greater than normal for healthy young adults, normal being 5 degrees to 8 degrees.44 Professor Pandy said he did not know the prevalence of a greater than normal toe-out gait in the community.45

SUBMISSIONS

P’s submissions

38 P made written46 and oral submissions47 in support of the admissibility of Professor Pandy’s evidence.

39 Cases relied upon by P included R v Aitken,48 Hashi v R,49 and Otway v R.50 In each of those cases, expert evidence of similarities between the gait of the alleged offender and the accused was admitted.

Pt 3.1

40 P submitted that the relevant “fact in issue” is the identity of the shooter and that Professor Pandy’s evidence bears upon that issue.

41 P relied on the following aspects of Professor Pandy’s evidence in support of its case that D was the shooter:

• his measurements and comparison of the spatio-temporal gait features of the subjects of the CCTV footage, namely, their step length, step frequency and walking speed;

• his opinion that the spatio-temporal features of the subjects’ gait are more consistent with the gait of an older person;51 and

• his opinion that the toe-out gait of the subjects is more consistent with the gait of an older person.52

42 P submitted that Professor Pandy’s evidence is relevant because it rationally increases the probabilities that D was the shooter. P highlighted Professor Pandy’s finding that the mean step lengths of the subjects in the two groups of clips were ‘remarkably similar’. He also said the mean step lengths were ‘almost exactly the same’.53

43 P submitted that the jury could only make general comparisons of the gait of the subjects whereas Professor Pandy’s measurements in his second report provided a level of detail which will be of assistance to the jury.

44 P submitted that the jury will also be assisted by Professor Pandy’s expertise regarding studies on the gait of younger and older adults, which informed his opinion that the spatio-temporal gait features of the subjects were more consistent with the gait of older persons.

45 As regards the toe-out gait of the subjects, P submitted that whilst the jury could observe it themselves, they would not be able to say whether the subjects display a larger than normal degree of abduction. Without Professor Pandy’s assistance, they would not be able to properly appreciate the significance of this feature.

46 P submitted that it was unlikely that any of the limitations of the CCTV footage mentioned in D’s written submissions54 — eg, number of frames per second, accuracy of the time stamps — would significantly impact on the accuracy of Professor Pandy’s measurements.

47 P submitted that the absence of a population database of gait spatio -temporal features did not render the evidence irrelevant. It simply went to the weight a jury could give the evidence.55

Pt 3.3

48 P submitted that Professor Pandy’s evidence was admissible expert opinion evidence under s 79.56

49 In oral submissions, P submitted that the relevant field of specialised knowledge was gait analysis and Professor Pandy was an expert in that field, even though he had never given FGA evidence before and had never previously analysed gait using CCTV clips. Every expert has to give evidence for the first time and the CCTV clips were just the medium for the application of his expertise on this occasion.

50 P conceded that Professor Pandy’s calculations of spatio-temporal gait features were not based on Professor Pandy’s expertise in gait analysis but rather on simple mathematical calculations which anyone could do (eg speed = distance divided by time).57 But P noted that s 8058 abolishes the common knowledge rule.59 Further, P submitted that the mathematical calculations made by Professor Pandy were one small part of the exercise he performed.60 His opinions were ‘substantially’ based on specialised knowledge of gait analysis, ‘substantially’ in s 79 meaning ‘largely’ but not ‘primarily’.61

Pt 3.6

51 P submitted Professor Pandy’s evidence was not tendency evidence.

52 In discussion, P relied on Elomar v R62 and Higgins v R,63 both terrorism cases where the court found that the impugned evidence was not evidence of a tendency to have a state of mind but evidence of a continuous state of mind.

53 P submitted that the impugned evidence was not evidence of D’s tendency to walk in a particular way but simply evidence that he did walk in a particular way. P said this:

[T]his isn’t to do with a propensity to walk in a particular way; it is the way in which he walks... It’s not his propensity to act in a particular way on particular occasion, he did act [in that way] on that occasion.64

54 P likened the evidence of the gait of the subjects to evidence that an offender and an accused had similar identifying features (eg, both had blue eyes or tipped hair or a certain build or a cleft palate or an eye patch or a hunched back65). P submitted that the way a person walked was an ‘inherent physical attribute’66 or physical trait,67 not evidence of conduct.68 No syllogistic tendency reasoning was involved.

55 P submitted that the rationale of tendency evidence in criminal proceedings is to exclude evidence of misconduct other than that the subject of the charge. One should have regard to that rationale in determining whether FGA evidence is tendency evidence: gait is not misconduct.

56 If Professor Pandy’s evidence was tendency evidence, P conceded that it only had modest probative value, not significant probative value.

Pt 3.11

57 P at first submitted that the probative value of Professor Pandy’s evidence was ‘fairly high’69 but then appeared to resile from that position.

58 P submitted that the only danger of unfair prejudice that fell for consideration was whether the jury might overvalue the evidence but P said it would not be submitting to the jury that Professor Pandy’s evidence had any more than modest probative value. Further, cross-examination would highlight the limitations of the evidence, as would appropriate jury directions. P submitted that these considerations meant there was no danger of the jury overvaluing the evidence.

D’s submissions

59 D made written70 and oral submissions.

60 D did not concede that he is the subject of clips 1–94. Nor did he concede that clips 37–94 show the shooter.71 D concedes only that he is the subject of clip 95.

61 D submitted that the evidence is inadmissible under Parts 3.1, 3.3, 3.6 and 3.11.

Pt 3.1

62 D submitted that Professor Pandy’s evidence failed the test of relevance for several reasons.

63 First, the jury can observe and compare the gait of the subjects themselves. Professor Pandy’s evidence really adds nothing.72

64 Second, Professor Pandy’s evidence is merely consistent with D being the subject of the clips and something more than mere consistency is required for relevance.73

65 Third, Professor Pandy’s methodology was so deficient that his conclusions could not rationally affect the probabilities that D was the shooter. He did not take into account difficulties or limitations associated with making accurate measurements based on CCTV footage. There is no evidence that those difficulties or limitations are inconsequential. P asserts they are inconsequential but that is nothing more than mere assertion.

66 Fourth, the significance of Professor Pandy’s calculations could not be meaningfully evaluated by a jury in the absence of information about the frequency of these gait parameters in the general population. D said this:

So relevance fundamentally is about … rationally affecting the probability of the fact in issue. That requires in this case some comment to be made about how probable it is that someone other than the accused would display the same features. Now when it’s in terms of this looks like an old person, this person walks slowly, that’s something that the jury can certainly do for themselves and that is the point ultimately that the prosecution will seek to make. The numerical values devoid of any statistical comparison with anything else don’t assist in that task.74

Pt 3.3

67 D submitted that on the evidence adduced, one could not be satisfied that FGA is a field of specialised knowledge.75

68 In the alternative, D submitted that the relevant area of FGA here is, as P stated in its written submissions, forensic gait comparison analysis, not merely gait analysis. On his own admission, Professor Pandy is not an expert in that area because this is the first time he has ever undertaken the task.

69 Further, Professor Pandy’s task was to compare the gait of subjects recorded not on video recordings made in ideal laboratory conditions but on CCTV, another task he has never previously undertaken.

70 Whilst it may be conceded that there has to be a first time for an expert to give evidence, D submitted that Professor Pandy is not an expert because he has not performed the relevant task in the past.76

Pt 3.6

71 D submitted that Professor Pandy’s evidence is evidence of conduct (gait) which P relies on for tendency reasoning and as such the evidence is caught by the exclusionary rule in s 97.77

72 D submitted that P inaccurately claims that Professor Pandy’s evidence of gait is simply evidence of how the subjects walk, not evidence of their tendency to walk in a certain way. But the way a person walks is not fixed, like the colour of their eyes.

73 D submitted that Professor Pandy’s evidence in relation to clips 1–36 and 95 is in reality evidence of D’s tendency to walk in a certain way. P also relies on Professor Pandy’s evidence as demonstrating that the shooter walked in a way that was closely similar to the gait tendency exhibited by D. The jury will infer from D’s tendency that on the occasion in question (clips 37–94), D walked in conformity with that tendency. In this way, tendency reasoning is engaged, even though Professor Pandy does not go so far as to say that the subjects of the clips are one and the same person.

74 D submitted it is tendency evidence which lacks significant probative value and so is inadmissible (s 97).

Pt 3.11

75 D submitted that the probative value of Professor Pandy’s evidence is low. It is merely evidence of similarities of gait. It is not even evidence that the subjects were older persons, just that their gait is more consistent with being older. Further, Professor Pandy does not provide any information about the prevalence or otherwise of the gait observed in the general population.

76 D submitted that there is a real risk that the jury will overvalue Professor Pandy’s evidence and use it as evidence of identification not of similarity, notwithstanding directions. Consequently, the evidence should be excluded under s 137.78

ANALYSIS

Pt 3.1

77 The ultimate fact in issue is the identity of the shooter.

78 Applying the test of relevance (s 55),79 could the jury rationally find that Professor Pandy’s evidence increased the probabilities that D was the shooter?

79 In answering that question it is important to keep in mind three things. First, the test of relevance is to be applied on the assumption that the jury will accept his evidence, unless no reasonable jury could accept it: credibility and reliability issues are normally jury issues, not admissibility issues.80 Second, the impugned evidence need only increase the probabilities slightly to pass the test of relevance.81 Third, the evidence is not to be considered in isolation but in the context of all the evidence relied on by P.82

80 It is conceded that D is the subject of clip 95. It is also open to a jury to find that the subject in clips 1–36 is D. This is so primarily because of Mr Reardon’s identification evidence that the male he saw watching the Gelo Bar on the night of 11 and 12 March 2016 was D. Consequently, a jury could find there were a significant number of clips showing D which informed Professor Pandy’s FGA.

81 It is also open to the jury to rationally find that the subject of clips 37 –94 is the shooter.83 Though this point was not conceded by D, it was only faintly contested.

82 Whilst Professor Pandy does not say that the two groups of clips show the same person, he does not merely say that the subjects have similar gaits. Based on his calculations of the mean step length, step frequency and walking speed, he says there is close similarity in the gait patterns of the subjects. In particular, he says the mean step lengths for the two groups, is ‘remarkably similar’ or ‘almost exactly the same’.84 Because those observations were based on measurements, his comparative analysis goes deeper than what a jury could discern from simply viewing the clips, which answers the objection that Professor Pandy’s evidence adds nothing to what a jury could determine for themselves.

83 Assuming the correctness of his calculations, it would be open to a jury to find that Professor Pandy’s evidence as to the close similarities of the spatio-temporal gait features of the subjects in the two groups of clips, when considered in combination with the other evidence relied on by P, increases the probabilities that D was the shooter.

84 As mentioned above, the general rule when assessing relevance is that one assumes that the impugned evidence will be accepted by the jury, unless no reasonable jury could accept it. Hence, one assumes the accuracy of Professor Pandy’s calculations, unless no reasonable jury could accept them.

85 What factors may have affected the reliability of Professor Pandy’s calculations? He acknowledged that he sometimes rounded-off the number of steps taken between point A and point B in a given clip but he did not keep a record of when he did so. When asked about the number of times he did this, he first said ‘infrequently’, then he said ‘no more than 20% of the time’ which strikes one as a significant number of occasions. If those occasions were spread across both groups they may have evened themselves out, so to speak, but Professor Pandy did not say positively that was the case. On the evidence, one cannot exclude the possibility that the rounding-offs were concentrated in one group. Professor Pandy said he sometimes had to make a subjective judgment as to the number of steps taken by the subject between points A and B. He acknowledged that the CCTV frame rates, about which he said he knew nothing, may have affected the accuracy of those subjective judgments. Professor Pandy used the time stamps on the CCTV clip to make his calculations but he did not check the accuracy of the various time stamps for the various CCTV cameras. Professor Pandy also acknowledged that the subjects did not always walk in a straight line between points A and B: at times, Professor Pandy noted excessive lateral movement by the subjects.

86 It may well be, as P submitted, that none of these matters significantly affected the accuracy of Professor Pandy’s calculations85 but, on the present state of the evidence, there is no way of knowing whether that is the case. If the jury is to act rationally, there must be a proper basis for the jury to conclude that the accuracy of his calculations were not significantly affected by these matters. The evidence fails to provide a proper basis for such a conclusion.

87 Another significant issue with Professor Pandy’s evidence is the difficulty in concluding rationally that he was “comparing apples with apples”. He conceded in cross-examination that there was no way of knowing whether the subjects in clips 1–95 were walking at their preferred speed. And yet he based his opinion that the subjects had gait patterns more consistent with older persons primarily on a comparison of the data with results obtained in gait studies of young and old adults whom it was known were walking at their preferred speed.

88 In my view, these issues with Professor Pandy’s evidence mean that it fails the test of relevance (s 55) and renders his evidence inadmissible (s 56).86

Pt 3.3

89 Professor Pandy expressed a number of purportedly expert opinions in his reports and in his oral evidence.

90 Because Professor Pandy’s spatio-temporal calculations involved some subjective judgments, I regard those calculations as expressions of opinion, rather than statements of fact.

91 Other opinions expressed by him were that:

• The spatio-temporal gait parameters of the subjects in the two groups of clips were closely similar;

• The mean step lengths of the subjects in the two groups of clips were remarkably similar;

• The mean step frequencies of the subjects in the two groups of clips were quite similar;

• The mean walking speeds of the subjects in the two groups of clips were quite similar;

• The gait of the subjects in the two groups of clip was more consistent with the gait of older persons, as was the degree of toe-out gait of the subjects.

92 Dasreef Pty Ltd v Hawchar87 is a High Court decision which highlights the importance of identifying the relevant area of specialised knowledge.

93 In its written submissions, P submitted that the relevant field of speciali sed knowledge was forensic gait comparison analysis. In oral submissions, P submitted that the relevant field was simply “gait analysis”. P downplayed the significance of the use of CCTV footage to perform that analysis, submitting that the video clips were merely the medium by which Professor Pandy went about his analysis.

94 In my view, the relevant area of specialised knowledge should be determined by reference to the specific task that Professor Pandy was asked to undertake by the police in producing his second report. This was forensic gait comparison analysis of subjects recorded on CCTV footage. There were two aspects of the task in which Professor Pandy acknowledged he was a novice. First, the comparison aspect — whether any of the subjects were the same. Second, the CCTV aspect — Professor Pandy indicated he was familiar with gait analysis from recordings made under ideal laboratory conditions but had never undertaken gait analysis using CCTV recordings.

95 The comparison aspect might not appear troubling given that Professor Pandy’s evidence is evidence of similarities only, not evidence of identification. But Professor Pandy’s failure to discuss dissimilarities of gait strikes me as a significant omission and calls into question his expertise to undertake comparative gait analysis. 88

96 The second aspect is also concerning. If the relevant field of specialised knowledge is forensic gait comparison analysis of subjects recorded on CCTV footage, it is difficult to see how Professor Pandy can be considered an expert in that field when this is the first time he has performed the task. On his own admissions, Professor Pandy had limited knowledge of the shortcomings of CCTV footage and could not say the extent to which these shortcomings impacted on his perceptions and calculations.

97 For all his undoubted learning and experience, I am not satisfied that Professor Pandy is an expert in the relevant area of FGA. Thus, even if his evidence passes the test of relevance, it engages the exclusionary opinion rule (s 76)89 and is not “saved” by the exception for opinions based on specialised knowledge (s 79).

Pt 3.6

98 Tendency reasoning involves two steps. The first step moves from the particular to the general (induction90): typically, conduct on multiple occasions is used to infer a tendency to engage in such conduct. The second step moves from the general to the particular (deduction91); it is inferred that the person acted in conformity with their tendency on the occasion in question.

99 P disavows any reliance on tendency reasoning but, as was observed in Elomar v R,92 that is not the end of the matter. Without doubting the sincerity of P’s disavowal, it is necessary to consider whether tendency reasoning actually underlies the use that P would have the jury make of Professor Pandy’s evidence.

100 P submits that the impugned evidence is simply evidence of how D walks compared with how the shooter walks. P submits that it is not inviting the jury to infer that D has a tendency to walk in a certain way or to infer that he acted in conformity with that tendency on the night of the murder. P says it will simply be inviting the jury to consider the close similarities in how D and the shooter walk in support of the contention that they are one and the same person.

101 But we know that the spatio-temporal features of a person’s gait are not fixed physical characteristics like the colour of a person’s eyes, or even reasonably constant features like a person’s religious convictions (cf Elomar v R93 and Higgins v R94). Gait can change markedly even in the course of a single day depending on the exigencies of the moment. Consequently, it strikes me as inaccurate to say that the evidence here is just evidence of how D (and/or the shooter) walks.

102 It is more accurate, in my view, to say that this is evidence of how D walked on a limited number of occasions which is really to say that it is evidence of a tendency on his part to walk in that way. He may not walk that way all the time but he does sometimes. And because he has walked that way on some occasions, he can be expected to walk that way on other occasions. Why? Because he has a tendency to

walk that way.

103 It seems to me that the prosecution are really inviting the jury to reason as follows. First, clips 1–36 and 95 show D walking. Second, Professor Pandy’s analysis of D’s gait in those clips shows that D has a tendency to walk a certain way. Third, because D has such a tendency, he can be expected to walk that way on other occasions. Fourth, clips 37–94 show the shooter walking. Fifth, the way he walks is closely similar to the gait tendency exhibited by D. Taken together, the evidence of D’s gait tendency, and its close similarity to the shooter’s gait, support P’s case that D was the shooter.

104 P submitted that the rationale for the tendency rule militates against a conclusion that Professor Pandy’s evidence is tendency evidence. P submitted that the rationale for the exclusionary tendency rule is the danger of unfair prejudice to an accused from the jury learning of other misconduct by that accused.95 P submitted that the determination of whether evidence is tendency evidence should be informed by that rationale and, since FGA evidence is not evidence of misconduct other than that charged, it does not engage Part 3.6.

105 There are at least seven problems with this submission.

106 First, the wording of s 97 does not support P’s submission that tendency evidence is limited to evidence of other misconduct. Section 97 speaks of evidence of (inter alia) conduct, not misconduct. It speaks of evidence of a tendency, not of a good or bad tendency. The focus of s 97 is whether tendency reasoning is engaged, not the type of conduct that founds that kind of reasoning.

107 Second, it is s 101,96 not s 97 which is concerned with the danger of prejudicial effect.

108 Third, the fact that s 97 applies in civil proceedings makes it impossible to maintain that tendency evidence is limited to evidence of other misconduct. For example, tendency evidence might be adduced in a case where the issue is whether the defendant agreed to a particular term of an oral contract: evidence that in other similar contractual transactions, the defendant had agreed to such a term, would be evidence of a tendency but not of misconduct.

109 Fourth, various expositions of tendency evidence by the High Court and intermediate courts of appeal do not support P’s submission. For example, in IMM v R,97 Gageler J described tendency evidence this way:

The nature of tendency evidence adduced by the prosecution in a criminal trial is that it is evidence of another occasion or occasions on which the accused acted in a particular way. The evidence is adduced in order to provide a foundation for an inference that the accused has or had a tendency to act in that way or to have a particular state of mind, the existence of which tendency makes it more probable that the accused acted in a particular way or had a particular state of mind at the time or in the circumstances of the alleged offence. Tendency evidence is thus evidence the relevance of which lies in its capacity indirectly to affect the assessment of the probability of the existence of the fact in issue of the accused’s action or state of mind at the time or in the circumstances of the alleged offence.98

110 Gageler J was even more to the point in his description of tendency evidence in Hughes v R:

99 … Applied to evidence of past conduct, tendency reasoning is no more sophisticated than: he did it before; he has a propensity to do this sort of thing; the likelihood is that he did it again on the occasion in issue.

111 Fifth, the rationale of the tendency rule is not confined to the danger of unfair prejudice from the disclosure of other misconduct by an accused. As can be seen from the relevant ALRC reports, one of the rationales for the tendency rule is the potential unreliability of tendency evidence. Psychological studies relied on by the ALRC indicated that conduct by an accused on other occasions may be of little use as a predictive tool unless the circumstances are similar to the occasion in question.100 This rationale for the tendency rule is pertinent whether or not the other conduct is misconduct.

112 Sixth, if tendency evidence was limited to evidence of misconduct it would not have been necessary in s 110101 to provide that the tendency rule does not apply to evidence of the good character of an accused generally or particularly.

113 Seventh, even in criminal proceedings, the tendency evidence may not be about the accused. An accused may rely on tendency evidence about the complainant and such evidence may be of conduct that could not be described as misconduct.

114 Accordingly, I find that Professor Pandy’s evidence is tendency evidence which engages the exclusionary rule in s 97.

115 In my view, P rightly conceded that Professor Pandy’s evidence does not have significant probative value and so it is inadmissible tendency evidence.

Pt 3.11

116 I will assume the following for the purpose of applying s 137 — that Professor Pandy’s evidence is relevant, that it is admissible expert evidence and that it is not tendency evidence.

117 The burden is on D to show that the probative value of Professor Pandy’s evidence is outweighed by the danger of unfair prejudice to D.

118 When discussing Part 3.6, P conceded that Professor Pandy’s evidence does not have significant probative value. In my view, that concession was properly made. The probative value of the evidence is modest. It is modest because, taken at its highest, it is evidence of similarity, not identity. Further, there is no evidence as to how common or uncommon in the general population are the mean step length, step frequency and walking speed of the subjects in the clips. Further, Professor Pandy does not assert that the subjects of the relevant clips are elderly, just that the spatio- temporal gait features are more consistent with the gait of older persons.

119 There are, in my view, two ways in which Professor Pandy’s evidence may be misused. First, the jury might attach more weight to it than deserves. Professor Pandy’s use of the phrase “remarkably similar” creates the greatest risk: it tends to suggest identification, not just close similarity. But it seems to me that danger can be dealt with by directions about the limitations of the evidence.

120 The other way in which Professor Pandy’s evidence may be misused is for tendency reasoning, which P disavows. I have already set out the train of reasoning above and will not repeat it here. In my view, the risk of that reasoning is very real because it is such an obvious line of reasoning. I do not consider that risk is likely to be cured by directions.

121 Accordingly, I also exclude Professor Pandy’s evidence under s 137.