DISCERNING THE FORM AT THE SECOND STAGE OF VICARIOUS LIABILITY

Abstract This article uses Atiyah and Summer's categorisation of the attributes of formal legal reasoning in Form and Substance in Anglo-American Law to examine the type of legal reasoning process used by the courts in England and Wales when determining the second stage of vicarious liability. The analysis shows that, although remaining formal in nature, the shift away from the Salmond test has resulted in a shift in the type of form used by the courts. It is suggested that future guidance issued by the Supreme Court to lower courts when determining the second stage of vicarious liability needs to take account of this change for the guidance to be effective.


I. INTRODUCTION
The law of vicarious liability has undergone considerable development in the past 30 years, 1 particularly the second stage (the requirement that the tort is committed "in the course of employment", or, since 2001, 2 that there is a sufficiently close connection between the tortfeasor's employment and her conduct to render the employer vicariously liable). In 2020, in Various Claimants v Wm Morrison Supermarkets plc 3 ("Morrison II"), Lord Reed sought to rein in that development. In particular, Lord Reed sought to clarify the earlier reasoning of Lord Toulson in Mohamud v Wm Morrison Supermarkets plc 4 ("Morrison I"). In discussing whether there was a sufficiently close connection to impose vicarious liability in that case, Lord Toulson had invoked the "principle of social justice" 5 and referred to the importance of "an unbroken sequence of events" that led from the employee in question carrying out his duties to the assault. 6 According to Lord Reed in Morrison II, lower courts in subsequent cases had taken these phrases out of context and interpreted them as if they represented a new approach to determining the second stage of vicarious liability. 7 This was wrong, said Lord Reed, because Lord Toulson had not intended to change the law nor move away from the precedents cited. 8 Lord Reed then placed use of existing precedents at the centre of the process for determining the second stage of vicarious liability. Referring to the criterion offered by Lord Nicholls in Dubai Aluminium Co. Ltd. v Salaam 9 that there is a sufficiently close connection if an act of an employee may fairly and properly be regarded as done by an employee while acting in the ordinary course of his employment, Lord Reed observed: The words "fairly and properly" are not, therefore, intended as an invitation to judges to decide cases according to their personal sense of justice, but require them to consider how the guidance derived from decided cases furnishes a solution to the case before the court. Judges should therefore identify from the decided cases the factors or principles which point towards or away from vicarious liability in the case before the court, and which explain why it should or should not be imposed. Following that approach, cases can be decided on a basis which is principled and consistent. 10 This article considers whether a renewed focus on precedent is likely to be sufficient to achieve Lord Reed's aim of reining in vicarious liability. It does this by comparing the approach of the English courts to determining the second stage of vicarious liability under both the traditional test (known as the "Salmond" test) and the more recent "close-connection" test.
Detecting a change in the nature of the legal reasoning process adopted when applying the different tests to determine the second stage of vicarious liability is undoubtedly more difficult than detecting the change in the test itself. To assist in this task, this article utilises a framework developed by Atiyah and Summers when comparing the nature of the legal reasoning processes adopted by English and American courts. 11 Use of this framework is appropriate for the following reasons.
First, as observed by Atiyah and Summers, "form" is systemic. Although not attempting to draw specific causal connections, they argued that the more "formal" nature of the legal reasoning process in England conformed 5 Ibid., at 693, [45] 6 Ibid., at 693, [47]. 7

496
[2022] The Cambridge Law Journal to and was consistent with other features of the English legal system. 12 Those features included: development of and adherence to a positivist legal tradition by legal theorists; 13 the general trust of legal decision makers in the "rest of the legal-political establishment", including "governments, legislatures, officials, police"; 14 and the small and somewhat elitist nature of the bar and bench. 15 This explains why English courts have been reluctant to incorporate substantive justifications into the process for determining the second stage of vicarious liability. 16 The framework rightly focuses on better understanding the nature of the legal reasoning process (including the type of form) utilised by the courts when determining the second stage of vicarious liability rather than dissecting a range of possible substantive justifications for that liability.
Second, Atiyah and Summers' examination of "form" was wider-ranging than just so-called legal rules. It included a broad range of phenomena considered by judges and other officials vested with legal decision-making power within the legal reasoning process, including the use of precedents, the outcome of a case (in terms of the judgment or verdict reached), matters of status and "rules crystallising rights or duties". 17 The framework expressly contemplates the existence of different types of form that might be utilised by judges when determining the second stage of vicarious liability.
Finally, Atiyah and Summers identified four attributes of formal legal reasoning. 18 Those attributes break down the process of formal legal reasoning into its component parts, enabling different processes to be analysed and compared.
The analysis in this article demonstrates that the nature of the legal reasoning process adopted by the English courts under the Salmond and closeconnection tests has indeed changed. To be sure, the legal reasoning process adopted under the close-connection test remains formal in nature. What has changed is the type of form utilised by the English courts. Any guidance issued by the Supreme Court to lower courts on determining the second stage of vicarious liability must take account of this change to be effective.
The analysis starts in Section II by describing and distinguishing the four different attributes of formal legal reasoning identified by Atiyah and Summers. Those attributes are used in Section III to analyse the approach of the English courts to legal reasoning under the Salmond test. The analysis shows that the legal reasoning process utilised by the English courts 12 Ibid., at 35. 13 Ibid. 14 Ibid., at 39. 15 Ibid. 16 See the discussion on Lister v Hesley Hall [2002] 1 A.C. 215, note 95 below. 17 Atiyah and Summers, Form and Substance, 7-11. 18 Ibid., at 11-28. C.L.J. 497 Discerning the Form at the Second Stage of Vicarious under the Salmond test is underpinned by an analogy drawn with agency (used by Salmond in a broad, non-contractual sense). This analogy grounds the legal reasoning process and provides a mechanism (outside cases of deliberate misconduct) to limit the potential for under or over-inclusive results. In Section IV of this article, the four attributes of formal legal reasoning identified by Atiyah and Summers are used to examine the approach of the English courts to legal reasoning under the close-connection test. Unlike the Salmond test, legal reasoning under the close-connection test is not informed by an analogy, and is more rule-based in nature. When transplanting the test from Canada, the English courts expressly declined to adopt the substantive justifications identified by the Supreme Court of Canada as informing the application of the test. 19 Without the benefit of those substantive justifications or the analogy underpinning the Salmond test, the close-connection test is shown to have a much higher degree of content formality than the Salmond test. The higher degree of content formality that results from use of the close-connection test at the second stage of vicarious liability increases the risk of arbitrary results. This risk exists due to the failure of the Supreme Court to recognise the shift in the type of form utilised under the close-connection test and to provide appropriate guidance to lower courts on how to minimise the potential for under or over-inclusive results.

II. ATTRIBUTES OF "FORMALITY"
The first attribute of formal legal reasoning identified by Atiyah and Summers is authoritative formality. 20 This term describes the extent to which any "form" examined as part of the legal reasoning process can be viewed as legally authoritative. There are two aspects of authoritative formality: validity formality and rank formality. 21 Validity formality refers to the extent to which the relevant "form" is considered valid in and of itself. This attribute of formality directs attention to the source of the "form" to be analysed. For instance, a rule derived from statute will have high validity formality if the statute is properly passed in accordance with the rules governing the enactment of statutes. In contrast, rank formality directs attention to the priority to be afforded to different "forms" once they have been ascribed validity formality. For instance, a rule derived from statute will have higher rank formality than a common law rule.
The term content formality is used by Atiyah and Summers to describe the extent to which any "form" examined as part of the legal reasoning process reflects its content. 22 The lower the degree of correlation between the 19 See note 101 below. 20 Ibid., at 12-13. 21 Ibid. 22 Ibid., at 13-14.

498
[2022] The Cambridge Law Journal "form" and any substantive reason 23 for that "form", the higher the degree of content formality. They suggest that the degree of content formality of a particular legal "form" can be determined by reference to two discrete factors: the extent to which the "form" itself is shaped by fiat (e.g. a rule that requires people to drive on the left as opposed to the right 24 ); and the extent to which the "form", when applied, can produce results that are either under or over-inclusive in terms of any substantive reason identified as the basis for introducing that "form". A legal reasoning process based on "form" that is either shaped by fiat and/or produces under or over-inclusive results is viewed by Atiyah and Summers as potentially more arbitrary and thus having higher content formality.
Atiyah and Summers used the term interpretive formality to describe the nature of the interpretation process engaged in by legal decision makers with respect to a particular legal "form". 25 In their view, the more literal the interpretation process adopted in respect of that "form", the higher the degree of interpretive formality. Correspondingly, the degree of interpretive formality is reduced to the extent legal decision makers consider the purpose of or substantive reasons underpinning the particular "form" as part of the interpretation process.
The final attribute of formality identified by Atiyah and Summers is mandatory formality. 26 This term describes the extent to which a legal decision maker is required to apply a particular legal "form" without regard to any substantive reasons. The less freedom that a legal decision maker has to override the particular legal "form" or refer to substantive reasons, the higher the degree of mandatory formality. Atiyah and Summers identified a variety of different ways in which the mandatory formality of a particular legal "form" might be reduced, including: express or implied qualifications; limits in the scope of application; capacity for discretionary non-enforcement; and defeasibility. 27 Atiyah and Summers acknowledged that each of four attributes of formality might be present to varying degrees in a particular legal reasoning process. 28 The degree of overall "formality" within a legal reasoning process is not to be determined formulaically, and different types of legal reasoning processes might have different combinations of the four attributes of formality and still reasonably be viewed as "formal".
Atiyah and Summers also noted that "form" was not necessarily determinative of the legal reasoning process. 29 Prioritising "form" might 23 As in a second level substantive reason, discussed ibid., at 23-28. 24 Ibid., at 13. 25 Ibid., at 14-16. 26 Ibid., at 16-17. 27  C.L.J. 499 Discerning the Form at the Second Stage of Vicarious "usually exclude from consideration" or "diminish the weight of" other reasons, but it did not always do so. 30 In a "formal" system it might sometimes be necessary to go beyond "form" where there are "gaps in the law", the existing law is "nascent and fragmentary" or the "form" produces "excessive arbitrariness". 31 In Atiyah and Summers' view, a legal system in which legal decision makers routinely ignored such difficulties was more likely to be categorised as "formalistic" rather than "formal". 32 Vigilance is therefore required to ensure that such legal reasoning processes do not become formalistic.

III. THE SALMOND TEST
This section analyses the approach of the English courts to legal reasoning under the Salmond test in terms of each of the four attributes of formality identified by Atiyah and Summers. Before assessing the extent to which that legal reasoning process can be described as "formal", it is first necessary to outline a brief history of how the Salmond test developed. The analysis focuses on Salmond simply because it was the test adopted by the English courts. Other formulations of the second stage of vicarious liability that existed at the time ultimately proved less successful. 33

A. Historical Development
Sir John Salmond was born in North Shields, Northumberland, and emigrated with his family to New Zealand in his teens. 34 He studied law at University College of London and later became a judge in New Zealand. Prior to that he was an academic at the University of Adelaide and it was there that he wrote most of The Law of Torts. The textbook remained in print until 1996 with subsequent editions baring the title Salmond on the Law of Torts, and finally, Salmond and Heuston on the Law of Torts to reflect the contribution of Robert Heuston who took over authorship after Salmond's death in 1924.
The formulation of the Salmond test used by the English courts was present in the very first edition of Salmond's textbook, The Law of Torts, published in 1907: 30 Atiyah and Summers, Form and Substance, 2. 31 Ibid., at 28. 32 Ibid. 33 The history of modern formulations of vicarious liability is usefully traced in W. Swain, "A Historical Examination of Vicarious Liability: A 'Veritable Upas Tree'?" [2019] C.L.J. 640. It should be noted that in outlining this history, Swain treats all uses of the label "vicarious liability" or the phrase "qui facit per alium, facit per se" as instances of the same type of liability without making further inquiry as to whether or not the liability imposed by the courts in what is a wide array of factually distinct cases is sufficiently similar to be usefully compared. This approach is problematic. See generally C. Beuermann, Reconceptualising Strict Liability for the Tort of Another (Oxford 2019). 34 R.F.V. Heuston, "Sir John Salmond" (1964) Adel. L. Rev. 220, 220.

500
[2022] The Cambridge Law Journal §29 -The Course of Employment 1. A master is not responsible for a wrongful act done by his servant unless it is done in the course of his employment. It is deemed to be so done if it is either (a) a wrongful act authorised by the master, or (b) a wrongful and unauthorised mode of doing some act authorised by his master. 35 What was it about this test that made it so influential? One possible answer lies in how Salmond presented the test. Critically, when Salmond wrote the Law of Torts, he was an academic. Furthermore, he was an academic who had been trained in a university. He was not a barrister and had not learned the legal craft from other barristers at the Inns of Court. As an academic, the intended audience for Salmond's textbook included students. 36 He therefore had to adopt a style of presentation that was easier for students to digest. 37 That style included conveniently labelled, discrete sections within each article of the text and the inclusion of a convenient summary (or, less charitably, reduction) of the content at the beginning of the section.
The style adopted by Salmond was in stark contrast to many of the texts that were produced by and for barristers trained at the Inns of Courts at the time, of which Clerk and Lindsell's Law of Torts was an example. 38 There were other textbooks written by barristers in publication, but (as will be seen below) there is evidence that Salmond was directly influenced by Clerk and Lindsell's text when formulating the test that came to bear his name.
First published in 1889, the style adopted by Clerk and Lindsell's text is perhaps best described as minimal. It contained long sections of prose divided into articles, with headings kept to a minimum. Each article provided extensive coverage of the range of cases that a barrister might need to consider in respect of a particular legal issue. The only real navigation aids were margin notes and the extensive case list provided at the beginning of the text. The different priorities of the textbooks in terms of style can be gleaned by comparing the respective length of the contents' pages (Salmondsix pages; Clerk and Lindsellthree pages) and the case lists (Salmond -16 pages; Clerk and Lindsell -46 pages). 39 35 J.W. Salmond, The Law of Torts: A Treatise on the English Law of Liability for Civil Injuries (London 1907), 78-83. 36 The preface to the first edition was as follows: "I have endeavoured in this book to set forth the principles of the law of torts with as much precision, coherence and system as the subject admits of, and with as much detailed consideration as is necessary to make the work one of practical utility. No book is justified by the good intent of its authority, but I hope that the present work will be found of use to lawyers and to students of law, as a general exposition in moderate compass, of an extensive and in some respects difficult and imperfectly development department of our legal system"; Salmond, Law of Torts. The thoroughness of Clerk and Lindsell's text had its uses. Salmond cited Clerk and Lindsell in his textbook when explaining the "Course of Employment" element of his test for vicarious liability. 40 The section from Clerk and Lindsell to which Salmond referred is as follows: An act is said to be within the scope of the servant's employment when, although itself unauthorised, it is so directly incidental to some act or class of acts which the servant was authorised to do that it may be said to be a mode, though no doubt an improper mode, of performing them. For an impropriety or excess on the part of the servant in the course of doing something which was authorised the master will be responsible, but not for an act wholly unconnected with the class of acts which the servant was authorised to do. In short, the master's liability for the unauthorised acts of his servant is limited to unauthorised modes of doing authorised acts. 41 The derivative nature of Salmond's test is immediately apparent. So too is the denseness of Clerk and Lindsell's text. The succinctness of Salmond's test may therefore have been a factor in its use and longevity.
It was not just the structure of Salmond's textbook that was shaped by his time in universities, but also its content. As evident throughout the textbook, Salmond was familiar with legal history. This knowledge was most likely acquired during his time in universities. With this knowledge, he could suggest a taxonomy of the different forms of liability in tort and, more relevantly for present purposes, outline the reasons why vicarious liability might be imposed. 42 It should be noted that Salmond's ambitions in outlining a taxonomy of tort law were relatively conservative. Unlike other academic authors, Pollock for instance, 43 Salmond had no desire to develop a unifying theory of tort law. Indeed he was criticised by other academics for this lack of ambition. 44 The concept of vicarious liability can be traced to Roman law and the well-documented liability of a master for the wrongdoing of a slave. 45 In Roman law, this was because the slave was the property of the master and could therefore be viewed as an extension of the master himself.

502
[2022] The Cambridge Law Journal for the misconduct of his wife and a father for the misconduct of a child. Blackstone was aware of, and utilised, the Roman conception of liability for the misconduct of another when explaining the vicarious liability of a master for the misconduct of a servant in the Commentaries. 46 Notwithstanding the largely commercial nature of the relationship between master and servant, Blackstone followed the Roman law and equated the relationship between master and servant with the other "great relations in private life": husband and wife and parent and child. 47 Blackstone also adopted the Roman explanation of the liability: qui facit per alium facit per se ("he who acts through another does the act himself"). 48 After outlining a number of English cases in which a principal had been held liable for the wrongdoing of an agent, Blackstone explained the basis of the liability in the following terms: We may observe that, in all the cases here put, the master may be frequently a loser by the trust reposed in his servant, but never can be a gainer: he may frequently be answerable for his servant's misbehaviour, but never can shelter himself from punishment by laying blame on his agent. The reason of this is still uniform and the same; that the wrong done by the servant is looked upon in law as the wrong of the master himself; and it is a standing maxim that no man shall be allowed to make any advantage of his own wrong. 49 It should be noted that Blackstone appeared to use the terms "agent" and "servant" interchangeably.
In writing the Law of Torts, Salmond appeared to take his lead from Blackstone, labelling the vicarious liability of a master for the tort of a servant, 50 along with the vicarious liability imposed on a husband for the torts of his wife, as forms of "absolute liability" (although Salmond noted that the latter was less significant at the time the textbook was written). 51 Unlike Blackstone, however, Salmond attempted to more clearly distinguish the relationships between principal and agent and master and servant. This distinction had been highlighted in the English cases decided after Blackstone's Commentaries and had, once again, been examined in detail by Clerk and Lindsell, with the authors' digesting the two categories of case separately.
In Salmond's view, a principal could only be held vicariously liable for an agent where the commission of the actual tort engaged in by the agent had been authorised by the principal. 52 Note, Salmond does not use the 46 W. Blackstone, Commentaries on the Laws of England, vol. 3 (Oxford 1768), 429. 47 Ibid., at 429. 48 Ibid. Presumably the source of the "master's tort" theory of vicarious liability. See R. Stevens, "Vicarious Liability or Vicarious Action" (2007) 123 L.Q.R. 30. 49 Blackstone, Commentaries, 432. 50 The other forms of liability identified by Salmond for taxonomy purposes was liability occasioned by "wrongful intent" or "culpable malice". Salmond, Law of Torts, 7. 51 Ibid., at 14. The categorisation of the liability as "absolute" would be challenged today; Lunney, "Sir Professor John Salmond", 115.
C.L.J. 503 Discerning the Form at the Second Stage of Vicarious term "agent" in a contractual sense. 53 Instead, he uses the term more generally to refer to one person doing something on behalf of another person; specifically, being "authorised" to commit a tort on behalf of another person. There need not be any ongoing relationship between the defendant and the so-called "agent" for the commission of such a tort to be authorised. Nor need there be any ongoing relationship between the defendant and the claimant, as would exist if the "agent" was to act in a contractual capacity. 54 A defendant engaging a thug to maim the claimant is a good example of what Salmond had in mind when he discussed this form of vicarious liability.
In contrast, an employer was not only vicariously liable for the tort of an employee the employer had authorised the employee to commit, but for the tort of an employee that was "so connected with [non-tortious] acts which [the employer] has so authorised" 55 that it could be treated in the same way. The way Salmond draws the link between the two forms of liability is significant. The basis of the two forms of vicarious liability is the sameauthorisation of conduct. It is broader in the employment relationship, but only by analogy with the vicarious liability that could be imposed when the defendant authorises an "agent" to commit a tort on the defendant's behalf. The extent of the analogy drawn by Salmond is clear from the text of the Salmond test itself. Both elements of the test rest upon the idea of authorisation of conduct by the person held vicariously liable for the commission of a tort by another person. 56 In formulating the test in this way, Salmond effectively combined the strengths of both Blackstone and Clerk and Lindsell. Blackstone had already explained the basis of vicarious liability arising in category (a), the tort being seen, by extension, as that of the employer herself. 57 Salmond adopted that explanation and developed it by analogy in category (b) to explain the extensive range of cases digested by Clerk and Lindsell in which vicarious liability was imposed by reason of an employment relationship in circumstances in which the commission of the tort had not been directly authorised by the employer.
What then is it about the employment relationship that attracts this extended form of vicarious liability? The most likely answer seems to be the authority vested in an employer to direct the conduct of an employee 52 Salmond, Law of Torts, 75. 53 Cf. C. Beuermann, "Disassociating the Two Forms of So-called Vicarious Liability" in S.G.A. Pitel, J.W. Neyers and E. Chamberlain (eds.), Tort Law: Challenging Orthodoxy (Oxford 2013). 54 This distinction is significant. See note 127 below. 55 Salmond, Law of Torts, 84. 56 It is worth noting, as observed by Swain ("Historical Examination", 650-51), that Salmond did not commit in The Law of Torts to expressly adopting either the master's or servant's theory of vicarious liability. He appears to have found neither theory particularly satisfactory. For a critique of the utility of the distinction drawn by the two theories see Beuermann, Reconceptualising Strict Liability, 125-27. 57 Qui facit per alium facit per se ("he who acts through another does the act himself").

504
[2022] The Cambridge Law Journal for the employer's benefit. 58 The term "authority" is used to refer to the general capacity of an employer to direct the conduct of an employee for the employer's benefit. 59 The vesting of such "authority" in an employer by an employee under an employment contract is a natural precursor to the employer giving the employee a specific direction to commit a tort (which can then be said to have been "authorised" by the employer). The existence of such "authority" is broader than that found within an "agency" relationship (in the broader sense used by Salmond). It extends beyond "authorising" the commission of a particular tort to a capacity to direct more generally the conduct of an employee whilst at work for the employer's benefit. 60 The importance of such authority can be seen in Salmond's attempt to justify the extended form of vicarious liability that arose by reason of an employment relationship: Its rational justification is to be found in the presumption that the negligence and other torts of a servant in the execution of his master's business are either actually authorised by the master, or are at least the result of some want of care on the master's part in the choice of competent servants or in the superintendence and control of their work. Very often this presumption does not correspond with the facts; but the difficulty of actually proving some default on the part of the master would be so great, that it is better on the whole to create a legal presumption against the master, and to even make that presumption irrebuttable. 61 What is being suggested is that vicarious liability is imposed by reason of an employment relationship because the nature of the employment relationship (specifically the authority an employer has to direct the conduct of an employee) means that an employer may very well have had a role in causing the tort to be committed, either directly by actual authorisation (category (a) of the Salmond test) or indirectly by engaging in some other conduct when directing the employee without which the tort would not have been committed (category (b) of the Salmond test). 62 As a stranger to the employment relationship, an injured claimant would be at a distinct disadvantage in having to prove the extent of an employer's involvement in the commission of the tort by the employee. In Salmond's view, vicarious liability is imposed on an employer to overcome such evidential difficulties. This explains the significance of an employment relationship to the 58 See more generally Beuermann, Reconceptualising Strict Liability, ch. 5. 59 See C. Beuermann, "Vicarious Liability: A Case Study in the Failure of General Principles" (2017) 33 Journal of Professional Negligence 179. 60 It is acknowledged that it is rare for an employer to direct "all" of an employee's conduct. Employees are given discretion to perform their tasks as required. What is important is that an employer has the capacity to so direct, if desired, and can take disciplinary measures if such a direction is not complied with. 61 Salmond, Law of Torts, 78, emphasis added. 62 Salmond expanded on this justification in his book Jurisprudence, but never incorporated such thoughts in The Law of Torts; Lunney, Scholars of Tort Law, 117-18. C.L.J. 505 Discerning the Form at the Second Stage of Vicarious imposition of vicarious liability. It also explains why an employer has a right of indemnity against an employee at common law. 63 As Salmond notes, "very often this presumption does not accord with the facts". 64 Despite the popularity of the Salmond test, there does not appear to be any judicial reference by the English courts to the unique justification proffered by Salmond for the vicarious liability imposed by reason of his test. Although unique, Salmond's justification is not anomalous. Other examples of evidential "gap-filling" devices being utilised within the law of torts include res ipsa loquitur and the controversial Fairchild exception to causation. 65

B. Attributes of "Formality"
The considerable period in which the Salmond test was used by the English courts means that there is extensive evidence of the legal reasoning process engaged in by judges when applying the test. When analysed in terms of Atiyah and Summers' four attributes of formality, the legal reasoning process appears relatively "formal". Each of the different attributes of formality will be examined in turn, starting with authoritative formality.
It will be recalled that there are two aspects of authoritative formality: validity formality and rank formality. Validity formality refers to the source of the relevant legal "form", in this case, the Salmond test itself. For obvious reasons, very little weight is usually attributed to academic textbooks as a source of law, academics having no recognised status as law makers. The Salmond test was, however, subsequently adopted by the House of Lords. 66 For this reason the Salmond test can be viewed as having relatively high validity formality notwithstanding the test originating in an academic textbook.
It should be noted that there was a significant delay between the publication of Salmond's textbook and the Salmond test being formally adopted by the House of Lords. The first reference to the Salmond test by judges who sat on courts in England is found in a 1942 judgment by the Privy Council. The Salmond test was relied upon by the Supreme Court of Canada in Canadian Pacific Railway Co. v Lockhart 67 to impose vicarious liability on an employer and the Privy Council confirmed on appeal that use of the test had been appropriate. 68

506
[2022] The Cambridge Law Journal decision of Williams v A. W. Hemphill Ltd. 69 (this time on an appeal from Scotland, Scottish law in this respect being viewed as the same as English law). One obvious explanation for the delay was the convention of the English courts at the time to not refer to living authors. 70 Another possible explanation is the doctrine of common employment which prevented an action being brought for vicarious liability by an employee in respect of the tort of a fellow employee. This doctrine reduced the range of circumstances in which an action for vicarious liability might be brought, thus reducing the opportunities for the English courts to clarify and develop the law. The doctrine of common employment was not fully abolished until 1948. 71 In terms of rank validity, the legal reasoning process encapsulated by the Salmond test can be described as having relatively high rank validity during the period of its operation. Once adopted by the House of Lords, use of the test could only be limited by legislation or the House of Lords itself (as finally occurred with the test's replacement in Lister v Hesley Hall Ltd. 72 ). The abolition of the doctrine of common employment saw also the removal of any legislative intervention with respect to vicarious liability. It follows that the adoption of the Salmond test by the House of Lords afforded the test its highest possible ranking short of being incorporated into legislation.
It is more difficult to assess the content formality of the legal reasoning process encapsulated by the Salmond test due to the historic reluctance of the English courts to expressly incorporate substantive reasons as part of the legal reasoning process. 73 As guided by Atiyah and Summers, the content formality of the Salmond test is to be determined by reference to the extent to which the test was shaped by fiat and/or produced results that were either under or over-inclusive in terms of any substantive reason identified as the basis for introducing that test.
It cannot be said that the Salmond test was shaped by fiat. As the history of the test demonstrates, Salmond exercised considerable care when formulating the test to encapsulate both the historical development of vicarious liability as a concept (as described by Blackstone) and the more recent cases decided by the English courts distinguishing vicarious liability within the agency and employment relationships (as digested by Clerk and Lindsell). Although the English courts do not appear to have expressly adopted the substantive reason outlined by Salmond for his test, they did  73 It will be recalled that, having applied the four attributes of formality to the English legal system, Atiyah and Summers classified the English legal system as a system of "formal" legal reasoning.
C.L.J. 507 Discerning the Form at the Second Stage of Vicarious adopt the precise formulation of the test prescribed by Salmond, quoting it verbatim on numerous occasions. To the extent that Salmond himself did not shape the test by fiat, neither then did the English courts.
But how can it be determined whether the results of applying the Salmond test are under or over-inclusive in terms of any substantive reason identified as the basis for introducing that test when the English courts failed to adopt any substantive reason as part of the legal reasoning process when applying the test? The nature of the challenge is self-evident. Some assistance in addressing the challenge can be found in the particular way in which Salmond formulated the test. It will be recalled that at the core of the Salmond test lies an analogy. The analogy is drawn between the vicarious liability that is imposed on a principal for the tort of an "agent" because the principal has authorised the commission of the tort (agent being defined in a broad, non-contractual sense) and the more extensive vicarious liability that is imposed on an employer for the tort of an employee because the commission of the tort is either authorised or "so connected with acts which [the employer] has so authorised" 74 that it can be treated the same way. To the extent this analogy has been made explicit within the Salmond test itself, it can be argued that it provided the English courts with a tool to reduce the risk that application of the test might produce results that were either under or over-inclusive in terms of any substantive reason identified as the basis for introducing the test (whether the substantive reason outlined by Salmond or some other derived by the courts).
To be sure, it is not a substantive reason that is incorporated into the Salmond test, but an analogy from which a substantive reason might be extrapolated. The process of reasoning by analogy has been criticised. 75 It is argued that its effectiveness as a legal reasoning process is limited to the extent that a legal decision maker can have confidence that she is drawing the analogy by reference to the specific attributes in the source material that are relevant for the purposes of addressing the particular legal issue at hand. 76 The absence of a substantive reason explaining why the analogy is being drawn can sometimes make determining the relevance of different attributes within the source material more difficult. It follows that the results produced at the limits of an analogy are at greater risk of challenge than those results produced within the core areas of its operation. The criticism attracted by the approach of the English courts to legal reasoning under the Salmond test corresponds directly to the concerns that are raised with the process of reasoning by analogy more generally. In most cases of vicarious

508
[2022] The Cambridge Law Journal liability, the Salmond test has been effective in limiting the risk that application of the test might produce results that are either under or overinclusive in terms of any substantive reason that might be identified as the basis for introducing that test due to the explicit analogy made within the test with agency. The results produced from applying the test were largely consistent and coherent when viewed across the period of its use, particularly in circumstances in which the tort engaged in by the employee was negligent. 77 Towards the limits of the analogy, however, there were some difficulties.
The general effectiveness of the analogy incorporated by the Salmond test can be seen in cases such as Williams v A W Hemphill Ltd, 78 the case in which the Salmond test was formally adopted by the House of Lords. The case involved a lorry driver instructed by his employer to drive a party from Benderloch to Glasgow. At the request of some of the members of the party, the driver took a more circuitous route. During the journey, the lorry was involved in an accident caused by the employee's negligence. The House of Lords held the employer vicariously liable for the injuries caused by the employee's negligence, notwithstanding the deviation from the employer's instructions. After stating the Salmond test, Lord Pearce (with whom the other Lords agreed) identified what it was the employee was authorised to do. He then compared what the employee was authorised to do with what the employee actually did do. The extent to which the analogy shaped Lord Pearce's reasoning can be clearly seen in the following extract from his judgment: Had the driver in the present case been driving a lorry which was empty or contained nothing of real importance, I think that so substantial a deviation might well have constituted a frolic of his own. The presence of passengers, however, whom the servant is charged qua servant to drive to their ultimate destination makes it impossible (at all events, provided that they are not all parties to the plans for deviation) to say that the deviation is entirely for the servant's purposes. Their presence and transport is a dominant purpose of the authorised journey, and, although they are transported deviously, continues to play an essential part. . .The more dominant are the current obligations of the master's business in connection with the lorry, the less weight is to be attached to the disobedient navigational extravagances of the servant. 79 A similar process was repeated time after time in cases involving negligence by an employee with considerable success. 80 Less successful were cases in which an employee had committed a tort deliberately.  79 Ibid., at 46, emphasis added. 80 In the sense that application of the process produced results which were generally accepted as consistent. 81 Harrison v Michelin Tyre Co. Ltd.  a practical joke on a second employee whilst at work by deliberately steering an empty two-wheeled hand truck out of the designated route marking so that a duck board would tip and cause the second employee to fall. A review of the cases led the judge to express considerable frustration. He noted that there had been previous cases in which employers had not been held vicariously liable for the consequences of practical jokes, 82 but found them difficult to reconcile with cases such as Williams v A W Hemphill Ltd. in which an otherwise deliberate deviation from what had been authorised by the employer was insufficient to deny the imposition of vicarious liability. 83 Nor did the judge find it particularly useful to consider whether the employee was on a "frolic of his own" such that the tort engaged in by the employee could not be considered an "unauthorised mode of doing some act authorised by his master", variously interchanging "frolic" with terms such as "joke", "wickedness" and "madness". 84 In an effort to remove himself from such difficulties, the judge in Harrison opted to devise his own test for determining vicarious liability: I will come to a few cases in a moment but, in searching for a principle, I have presumptuously invented one of my own and I have done it in the form of two questions, one which if answered "yes" makes the employer liable, the other if answered "yes" makes the employer not liable. Question 1: was the incident part and parcel of the employment in the sense of being incidental to it although and albeit unauthorised or prohibited? The test, of course, for that question as for the other being a reasonable approach. The alternative question is: or was it so divergent from the employment as to be plainly alien to and wholly distinguishable from the employment? If it was, then the employer is not liable.
Applying this test, the judge found the employer vicariously liable.
The judge in Harrison evidently had some sympathy for the injured plaintiff. Rather than be limited by the process of reasoning by analogy inherent in the Salmond test, the judge opted to devise his own test which was unconstrained by such a reasoning process. Notwithstanding the criticisms that had been made of the Salmond test at the time, the superior courts in England gave the proposed new test short shrift and moved to quickly reinstate the process of reasoning by analogy incorporated by the Salmond test. The Master of the Rolls in Aldred v Nacanco Ltd., another case concerning a practical joke in the workplace, could not have made the position clearer when he said: I mention this merely in the hope that in future cases the classic test, namely that to which my Lord Sir Frederick Lawton referred, which is set out in the 18 th ed of Salmond on Tort, will be applied. If one applies it to the facts of this

510
[2022] The Cambridge Law Journal case and one takes account, as the learned author rightly says one has to take account, of all the factors, such as time, place and the nature of the act, whether it was wilful or malicious, whether it was expressly or impliedly authorised by the employer and so on, I have no doubt that the unauthorised and wrongful act of Miss Perry was not so connected with the authorised act of going into the washroom for the usual purposes, and that the mode of doing it was an unauthorised act. Accordingly, the defendants are not liable. 85 In reinstating the Salmond test, the Master of the Rolls reinstated the process of reasoning by analogy incorporated by the Salmond test (which he then used to determine the outcome of the case).
Despite the difficulties associated with the Salmond test, it seems reasonable to suggest that the English courts were conscious of, and deliberately sought to retain, a process of reasoning by analogy when determining whether to impose vicarious liability under that test. It can therefore be argued that, notwithstanding the failure of the English courts to adopt a substantive reason for the Salmond test, the results produced by application of the test were generally not under or over-conclusive in terms of any substantive reason that might be derived from that analogy. It follows that the degree of content formality present within the approach of the English courts to legal reasoning under the Salmond test can be described as relatively low in the sense that the results produced from applying that test were generally not arbitrary.
The above analysis has implications when it comes to assessing the degree of interpretive formality present within the approach of the English courts to legal reasoning under the Salmond test. There is evidence to suggest, as indicated by the comments of the judge in Harrison, that the English courts could adopt a quite literal approach to interpreting terms such as "unauthorised mode" and "frolic of their own". Indeed, such interplay of terms led Atiyah himself to describe the reasoning process under the Salmond test as "largely a semantic one". 86 Such interpretive techniques were, however, generally only engaged when the process of reasoning by analogy incorporated by the Salmond test had reached its limits. As explained above, this was in a proportionately small number of cases, such as those involving torts committed deliberately by an employee. It follows that whilst the English courts sometimes adopted a literal approach to interpretation under the Salmond test, they did not always do so. For this reason, the degree of interpretive formality present within the approach of the English courts to legal reasoning under the Salmond test can be described as middling; not particularly low, as there were instances of terms being interpreted literally, but not particularly high, as those instances were not extensive. 85 Aldred v Nacanco Ltd. The final attribute of formality identified by Atiyah and Summers for assessing the degree of formality within a legal reasoning process was mandatory formality. There were no recognised exceptions or defences to the Salmond test. If the test was satisfied, it followed that the employer would be held vicariously liable. To this extent, the degree of mandatory formality within the approach of the English courts to legal reasoning under the Salmond test might be described as quite high. As has already been noted, however, the process of reasoning by analogy incorporated by the Salmond test had its limits. The degree of mandatory formality is consequently necessarily reduced to account for the degree of uncertainty that could arise as to whether the different elements of the Salmond test might be satisfied.
In overview, the approach of the English courts to legal reasoning under the Salmond test can be described as having relatively high authoritative and mandatory formality, middling interpretive formality and relatively low content formality. It will be recalled that Atiyah and Summers did not approach the assessment of overall "formality" formulaically. They also noted that it was not unusual for "form" in the English legal system to have a relatively low degree of content formality where the substantive reason for that "form" had been incorporated into the manner in which the "form" was expressed. It therefore does not seem out of line to suggest that the approach of the English courts to legal reasoning under the Salmond test could be described as relatively "formal". Furthermore, the analysis suggests that the risk of the Salmond test producing arbitrary results was relatively low, existing only at the margins of its application and pertaining specifically to the deliberate misconduct by an employee.

IV. THE CLOSE-CONNECTION TEST
The Salmond test remained in use until 2001 when it was replaced by the House of Lords in Lister 87 with the close-connection test. This section briefly outlines how the close-connection test came to fruition. It then repeats the analysis undertaken in respect of the Salmond test by examining the approach of the English courts to legal reasoning under the closeconnection test in terms of each of the four attributes of formality identified by Atiyah and Summers.

512
[2022] The Cambridge Law Journal churches and residential homes and was typically perpetrated by adults employed by or associated with those institutions. The likely impecunity of the perpetrators and standard insurance exclusions for criminal acts meant that most claims were brought against the institution responsible for the care of the child, rather than the individual perpetrator.
The Supreme Court of Canada was the first common law apex court 88 to hear a claim for vicarious liability brought by a child sexually abused by an employee of an institution charged with their care. The defendant institution in Bazley v Curry 89 operated "two residential care facilities for the treatment of emotionally troubled children". 90 The plaintiff had been placed into fulltime care at one of those facilities as a young boy. 91 Whilst at the facility, the plaintiff was sexually abused repeatedly by Leslie Curry. 92 Curry had been employed by the institution as a childcare counsellor. 93 His role was to act as a surrogate parent to the children in the facility. 94 The Supreme Court of Canada found the defendant institution vicariously liable for the sexual abuse.
The difficulties in applying the Salmond test in cases of torts committed deliberately have already been noted. The prospect of extending the concept of an "unauthorised mode" of an act otherwise authorised by an employer to such self-serving deliberate conduct as sexual abuse would have been to stretch the analogy incorporated by the Salmond test to breaking-point. 95 As the judge in Harrison had done before her, McLachlin J. (giving judgment for the majority) found it necessary to change the test in order to impose vicarious liability on the defendant institution. Significantly, she did not do this by abandoning the Salmond test altogether. Nor did she abandon the cases that had been decided under it. Instead, McLachlin J. recognised the limits of the analogy incorporated by the Salmond test and held that the only way to determine the boundary of the liability beyond the limits of the analogy was to openly resort to the substantive reasons for imposing vicarious liability: This review suggests that the second branch of the Salmond test may usefully be approached in two steps. First, a court should determine whether there are precedents which unambiguously determine on which side of the line between vicarious liability and no liability the case falls. If prior cases do not clearly suggest a solution, the next step is to determine whether vicarious liability should be imposed in light of the broader policy rationales behind strict liability. 96 Reviewing the jurisprudence, and considering the policy issues involved, I conclude that in determining whether an employer is vicariously liable for an employee's unauthorized, intentional wrong in cases where precedent is inconclusive, courts should be guided by the following principles: (1) They should openly confront the question of whether liability should lie against the employer, rather than obscuring the decision beneath semantic discussions of "scope of employment" and "mode of conduct". (2) The fundamental question is whether the wrongful act is sufficiently related to conduct authorized by the employer to justify the imposition of vicarious liability. Vicarious liability is generally appropriate where there is a significant connection between the creation or enhancement of a risk and the wrong that accrues therefrom, even if unrelated to the employer's desires. Where this is so, vicarious liability will serve the policy considerations of provision of an adequate and just remedy and deterrence. Incidental connections to the employment enterprise, like time and place (without more), will not suffice. Once engaged in a particular business, it is fair that an employer be made to pay the generally foreseeable costs of that business. In contrast, to impose liability for costs unrelated to the risk would effectively make the employer an involuntary insurer.
(3) In determining the sufficiency of the connection between the employer's creation or enhancement of the risk and the wrong complained of, subsidiary factors may be considered. These may vary with the nature of the case. When related to intentional torts, the relevant factors may include, but are not limited to, the following: (a) the opportunity that the enterprise afforded the employee to abuse his or her power; (b) the extent to which the wrongful act may have furthered the employer's aims (and hence be more likely to have been committed by the employee); (c) the extent to which the wrongful act was related to friction, confrontation or intimacy inherent in the employer's enterprise; (d) the extent of power conferred on the employee in relation to the victim; (e) the vulnerability of potential victims to wrongful exercise of the employee's power. 97 Note the caveats at the beginning of the test; it only applies in respect of torts committed deliberately by an employee when the analogy incorporated by the Salmond test, as represented by established cases, has been exhausted. The analysis in the previous section as to the general effectiveness of the Salmond test within the limits of the analogy with "agency" is

514
[2022] The Cambridge Law Journal therefore confirmed. Note further the starting point for applying the test; as with the Salmond test, judges are to consider what conduct has been authorised by the employer and whether the conduct engaged in by the employee is "sufficiently related" to "justify the imposition of vicarious liability". Where the test differs from the Salmond test is that it sets the outer limits of vicarious liability for torts deliberately committed by reference to various substantive concerns, namely enterprise liability, rather than resting on a process of reasoning by analogy.
Not long after Bazley v Curry was decided, the House of Lords was given the opportunity to consider the prospect of vicarious liability for historic sexual abuse in Lister. 98 The defendant institution in that case operated a school and boarding annex for boys. A number of boys were sexually abused at the boarding annex by Dennis Grain, who was employed as warden and housemaster of the boarding annex. The House of Lords followed the Supreme Court of Canada and imposed vicarious liability on the defendant institution. The House of Lords did not, however, adopt the reasoning of the Supreme Court of Canada.
Apart from Lord Hutton who agreed with the reasoning of Lord Steyn, each of the Lords in Lister delivered a separate judgment. Although there are differences across the judgments, the approach of each of the Lords is largely similar. For this reason, it is convenient to focus on the judgment of Lord Steyn, although differences in approach will be highlighted where relevant.
Lord Steyn started by noting that vicarious liability had been imposed by the English courts on at least some occasions for torts committed deliberately by an employee. He found, however, that the language used in the Salmond test was unhelpful in determining the outer limits of that liability. After discussing Lloyd v Grace Smith & Co., a case in which a bank had been held vicariously liable for the fraud of an employee, Lord Steyn commented: It is not necessary to embark on a detailed examination of the development of the modern principle of vicarious liability. But it is necessary to face up to the way in which the law of vicarious liability sometimes may embrace intentional wrongdoing by an employee. If one mechanically applies Salmond's test, the result might at first glance be thought to be that a bank is not liable to a customer where a bank employee defrauds a customer by giving him only half the foreign exchange which he paid for, the employee pocketing the difference. A preoccupation with conceptualistic reasoning may lead to the absurd conclusion that there can only be vicarious liability if the bank carries on the business of defrauding its customers. Ideas divorced from reality have never held much attraction for judges steeped in the tradition that their task is to deliver principled but practical justice. How the courts set the law on a sensible course is a matter to which I now turn. The reason for replacing the Salmond test was therefore the potential for absurd results at the outer limits of its application. It will be recalled that Atiyah and Summers had noted that one of the reasons judges might look to substantive reasons in a "formal" legal system, without affecting the overall "formality" of that legal system, is in response to use of a particular legal "form" producing absurd results. Unlike the Supreme Court of Canada, the House of Lords did not consider it appropriate to incorporate substantive reasons into the new test for vicarious liability. Various reasons were given.
Lord Steyn noted the inconsistencies between some of the substantive reasons given for vicarious liability; specifically, the need to compensate the injured claimant as against the risk of imposing an undue burden on business. 100 Unable to resolve such inconsistencies, Lord Steyn found that was it unnecessary to express any views on the policy considerations examined by the Supreme Court of Canada. 101 Lord Clyde noted the various substantive reasons outlined by the Supreme Court of Canada, but preferred to search for a principle underlying those substantive reasons. 102 Lord Hobhouse went the furthest in denouncing the appropriateness of the English courts adopting a legal reasoning process based upon the identification of any such substantive reasons: I do not believe it appropriate to follow the lead given by the Supreme Court of Canada in Bazley v Curry. The judgments contain a useful and impressive discussion of the social and economic reasons for having a principle of vicarious liability as part of the law of torts which extends to embrace acts of child abuse. But an exposition of the policy reasons for a rule (or even a description) is not the same as defining the criteria for its application. Legal rules have to have a greater degree of clarity and definition than is provided by simply explaining the reasons for the existence of the rule and the social need for it, instructive though that may be. In English law that clarity is provided by the application of the criterion to which I have referred derived from English authorities. 103 Lord Millett did spend some time in his judgment exploring the underlying substantive reasons for vicarious liability. His willingness to examine those substantive reasons did not extend to incorporating any of those substantive reasons within the formulation of the test for vicarious liability. 104 Having ruled out the incorporation of substantive reasons, the new test for vicarious liability was put in relatively straightforward terms by Lord Steyn: "The question is whether the warden's torts were so closely connected with his employment that it would be fair and just to hold the 99 Lister v Hesley Hall [2001] 1 A.C. 215, 224, emphasis added. 100 Ibid., at 223. Also acknowledged by McLachlin J., note 97 above. 101 Ibid., at 230. 102 Ibid., at 237. 103 Ibid., at 242. 104 Ibid., at 245.

516
[2022] The Cambridge Law Journal employers vicariously liable. On the facts of the case the answer is yes." 105 Similar references to the need for a close connection were made by the other Lords, hence the label being applied to the new test. The language of "connection" had been previously used both by the Supreme Court of Canada and by Salmond in his textbook when attempting to further clarify the operation of his initial test. 106 A number of points can be made in respect of how the close-connection test adopted by the House of Lords is formulated in contrast to the test adopted by the Supreme Court of Canada. First, there is no attempt under the close-connection test to retain a role for either the Salmond test itself, or the analogy with "agency" (in a broad, non-contractual sense) upon which the Salmond test was based. Second, the relevance of existing precedent created under the Salmond test is somewhat unclear given the stated intention to move away from the test entirely, not just alter the legal reasoning process at the limits of the test. Finally, the language used in the close-connection test is relatively open-textured. At least on Lord Steyn's formulation, it offers little in way of guidance to legal decision makers other than it must be "fair and just" to impose vicarious liability. In each of these respects, the close-connection test is not dissimilar to the test proposed by the judge in Harrison which was subsequently rejected by the Court of Appeal in Aldred v Nananco Ltd. 107

B. Attributes of "Formality"
The evidence of the legal reasoning process engaged in by the English courts when applying the close-connection test is more limited due to the shorter period in which it has been in operation. Examined in terms of Atiyah and Summers' four attributes of formality, the evidence to hand suggests that the nature of the legal reasoning process conducted by the English courts under the close-connection test has changed in certain respects from that utilised under the Salmond test.
The position with respect to authoritative formality is very similar under both the close-connection and the Salmond tests. Having been formulated and implemented by the House of Lords, the close-connection test has relatively high validity formality notwithstanding the origins of the test can be traced not only to Salmond's textbook, but developments in the Canadian courts. The legal reasoning process encapsulated by the close-connection test also has relatively high rank validity, there being no current legislation which affects its sphere of operation.
The first significant difference between the close-connection and Salmond tests is found when assessing the degree of content formality. It 517 Discerning the Form at the Second Stage of Vicarious cannot be said that the close-connection test is shaped any more by fiat than the Salmond test. The Lords in Lister did attempt to derive a principle from the existing cases and, whether they succeeded or not, that attempt is sufficient to discount any claim that the close-connection test was shaped by fiat. There is, however, arguably the potential for the close-connection test to produce results that are under or over-inclusive in terms of any substantive reason that might be identified as the basis for introducing that test.
The refusal of the English courts to identify or incorporate substantive reasons underpinning the close-connection test does not, of itself, produce under or over-inclusive results. The refusal is consistent with the English legal system being classified as a "formal" legal system more generally. It does though place the onus on the English courts to find some other way to address the potential for arbitrary results that might arise in the absence of such substantive reasons. As Atiyah and Summers noted, persistent failure to do so puts the English legal system at risk of becoming "formalistic". 108 A similar deficiency in the Salmond test was overcome by expressly incorporating within the operation of the test an analogy with "agency". Unlike the Salmond test, the close-connection test does not prescribe a process of legal reasoning (reasoning by analogy), but a rule. The rule stipulates that vicarious liability can be imposed where the connection between the employee's conduct and their employment is sufficiently close that it is "fair and just" to impose vicarious liability. It was this stipulation that enabled the Lords in Lister to overcome the limitations of the Salmond test. The advantages of such an open-ended resort to justice were subsequently noted by Lord Dyson in Morrison I: The importance of Lister is that it recognised the difficulty created by the second limb of the Salmond test. This was not effective for determining the circumstances in which it was just to hold an employer vicariously liable for committing an act not authorised by the employer. The close connection test was introduced in order to remedy this shortcoming. This improvement was achieved by the simple expedient of explicitly incorporating the concept of justice into the close connection test. 109 The potential for such an open-ended resort to justice to produce under or over-inclusive results in terms of any substantive reason identified as the basis for introducing the close-connection test has subsequently been recognised by the Supreme Court. It was this potential that led Lord Reed in Morrison II to refine the approach to be taken by lower courts when applying the test. 110 Specifically, Lord Reed sought to encourage lower courts to move from a rule-based legal reasoning process to a process of developing 108 See note 32 above. 109

518
[2022] The Cambridge Law Journal the law by analogy with existing cases to ensure that future determinations of the second stage of vicarious liability would be "decided on a basis which is principled and consistent". 111 Whether introducing a process of reasoning by analogy into the closeconnection test can, on its own, reduce the risk of the test producing under or over-inclusive results is doubtful. The process of analogical reasoning suggested by Lord Reed in Morrison II is significantly different to the process of analogical reasoning incorporated into the Salmond test. Under the Salmond test, the broader vicarious liability imposed by reason of an employment relationship is to be determined by analogy with the vicarious liability imposed on a principal who authorises an "agent" (in the broader, non-contractual sense) to commit a tort. No similar basis for drawing an analogy is recognised by Lord Reed in Morrison II. It is therefore more difficult for judges to ascertain which features of the existing precedents might be relevant for the purposes of drawing an analogy when using Lord Reed's revised approach.
It is most likely for this reason that Lord Reed sought to give an example of the types of features present in the precedents that might be relevant for determining the second stage of vicarious liability. Specifically, he noted that the close-connection test had been applied differently in cases of child sexual abuse, the very cases that led to the test being introduced. 112 In Lord Reed's view, there was a common feature of the child sexual abuse cases that resulted in the close-connection test being satisfied when applied to cases of child sexual abuse. That feature was a conferral of authority by the institution charged with the care of the child to the tortfeasor over the child given that there was the potential for that authority to have been exploited by the tortfeasor in order to engage in the sexual abuse. 113 Lord Reed's observation about the child sexual abuse cases is significant. It explains what it is about those cases that prompted the English courts to set aside the Salmond test and adopt the close-connection test, specifically in respect of the deliberate misconduct of an employee. The closeconnection test does not, however, apply only to cases of child sexual abuse. It applies to all cases of vicarious liability in England. In most of those cases, such as those that would have previously been viewed as within the core of the Salmond test, there is no such conferral of authority. It cannot even be said that such a conferral of authority is present in all cases of vicarious liability imposed for deliberate misconduct (consider 111  C.L.J. 519 Discerning the Form at the Second Stage of Vicarious the facts of Morrison I, which has not been overturned and remains good law). It is consequently not clear how lower courts are meant to use this feature of the child sexual abuse cases to reason by analogy in determining whether the second stage of vicarious liability is satisfied in future cases.
It follows that Lord Reed's direction to lower courts to apply the closeconnection test by analogy with existing cases does little to reduce the risk of the test producing under or over-inclusive results in terms of any substantive reason that might be identified as the basis for introducing that test. There is insufficient guidance to enable lower courts to ascertain which features of the existing precedents might be relevant for the purposes of drawing an appropriate analogy and in what circumstances. The degree of content formality in the approach of the English courts to legal reasoning under the close-connection test is therefore relatively high.
The high degree of content formality in the close-connection test necessarily impacts the degree of interpretive formality. There is little doubt that, at least initially, the test had a high degree of interpretive formality. It will be recalled that it was the literal interpretation of phrases such as "the principal of social justice" and an "unbroken sequence of events" that led Lord Reed to refine the close-connection test in Morrison II. In Lord Reed's view, such phrases had been treated as critical by the lower courts, 114 and elevated to a point where their application "would constitute a major change in the law". 115 Lord Reed was adamant that no such change had been intended. 116 Such phrases had instead been taken out of context and interpreted too literally. 117 At the very least, by highlighting the problem of the close-connection test being interpreted literally and directing the lower courts to focus more on previous cases than pithy phrases, Lord Reed has reduced the chances of the close-connection test being interpreted literally in the future. It is very unlikely, however, in the absence of sufficient guidance being provided to assist lower court judges to make appropriate analogies from existing cases, that the risk has been removed. Consider the phrase "conferral of authority". Does a hospital have authority over its patients which is then conferred on medical practitioners? 118 Does a football club confer authority on a scout when bringing junior players to the club's attention? 119 Without an explanation as to why such a conferral of authority is relevant for the purposes of determining the second stage of vicarious liability more generally, there is the potential for the phrase to be taken out of context. If this occurs, it may be applied to circumstances which cannot be supported by any substantive reasons that might have led to the feature initially being identified as key to the imposition of vicarious liability in the child sexual abuse cases. Without such difficulties being addressed, the degree of interpretive formality under the close-connection test might not be that dissimilar to that which existed at the limits of the Salmond test and ultimately resulted in the Salmond test being replaced.
The increased content formality associated with the close-connection test may also affect the degree of mandatory authority associated with the test. It will be recalled that Atiyah and Summers noted that a high degree of mandatory authority was required for not only an approach to legal reasoning to be considered "formal", but also for the approach to legal reasoning to be considered "law". 120 The reduction in mandatory authority comes not from the availability of exclusions or defences to the imposition of vicarious liability, of which there are none, but from the resulting uncertainty surrounding which analogies might be drawn from the existing precedents by lower courts when determining the second stage of vicarious liability. This uncertainty gives lower courts a certain amount of discretion to select an appropriate analogy, thus increasing the capacity of the lower courts to override the close-connection test and impose liability in a way that might not be supported by any substantive reason that might be identified as the basis for introducing that test.
Analysing the approach of the English courts to legal reasoning under the close-connection test using Atiyah and Summer's four attributes of formality shows the approach to be significantly different from that previously adopted under the Salmond test. The freedom with which judges may identify relevant analogies under the close-connection test means that the approach of the English courts when applying the test has a higher degree of content formality and potentially a lower degree of mandatory formality. There is consequently a risk that the close-connection test will produce even more arbitrary results when used to determine the second stage of vicarious liability than the Salmond test which it replaced.

V. CONCLUSION
Lord Reed hoped a renewed emphasis on precedent would ensure that lower courts applying the close-connection test to determine the second stage of vicarious liability would do so in a manner which was "principled and consistent". 121 This guidance failed to appreciate the change in the type of form utilised by the courts when the Salmond test was replaced with the close-connection test. This change in form, as demonstrated by this article, 120 Atiyah and Summers, Form and Substance, 19. 121 Morrison II [2020] UKSC 12, [2020] A.C. 989. C.L.J. 521 Discerning the Form at the Second Stage of Vicarious increased the degree of content formality in the application of the closeconnection test and hence the risk of arbitrary results. Recognising the change in the type of form now gives the Supreme Court the opportunity to reconsider the appropriateness of the close-connection test and/or the guidance given to lower courts when applying it. Valuable insights from the preceding analysis may assist the Supreme Court in this regard. They are outlined in conclusion.
First, the open-textured nature of the close-connection test is problematic in terms of the potential to produce under or over-inclusive, and therefore arbitrary, results. The English courts have adopted similarly open-textured tests based on "fairness and justice" in other troublesome areas of the law of torts, including to establish a duty of care in the tort of negligence. 122 Such tests have typically been introduced to overcome the perceived limitations of the previous legal reasoning process engaged by the English courts. The Supreme Court is to be applauded for its recent steps to limit the operation of such tests. 123 Second, the type of "form" incorporated into the Salmond test was more "process" based, whereas the type of "form" incorporated into the closeconnection test was more "rule" based. Differences in the type of "form" employed within a legal reasoning process can result in differences in the degree of overall "formality" within that legal reasoning process, in turn creating a risk of arbitrary results. Different types of "form" might therefore be more appropriate in particular circumstances. It follows that the English courts need to be more mindful of the type of "form" adopted within a legal reasoning process and consider its appropriateness for those particular circumstances. With respect to vicarious liability, a more "process" based type of "form" may arguably be more effective, given the absence of any clear substantive reasons that might be identified as underpinning the imposition of vicarious liability. Even if such reasons were to be identified, they might not be precise enough to provide courts with appropriate guidance when determining cases of vicarious liability. Consider the idea of vicarious liability as an evidential gap-filling device which was the original substantive reason proffered by Salmond when devising the test that bore his name. 124 Third, care needs to be taken when transplanting legal tests from other jurisdictions. As Atiyah and Summers note, "form" is systemic. The nature of the broader legal system in which a test operates will necessarily impact how that test is framed. All legal transplants need to be reviewed to 122

522
[2022] The Cambridge Law Journal ascertain what modifications might be necessary to adapt it to the "formal" nature of legal reasoning in the English legal system.
Finally, it is difficult for a process of reasoning by analogy to be useful when a legal decision maker cannot have confidence that she is drawing the analogy by reference to specific attributes in the source material that are relevant for the purposes of addressing the particular legal issue at hand. 125 Considerable care therefore needs to be taken in identifying relevant analogies and the circumstances in which they might be used. Looking back at the transition from the Salmond test to the close-connection test, the primary difficulty has been with cases of deliberate misconduct by an employee. Where an employee has been negligent, the original analogy of "agency" (used in a broad, non-contractual sense) under the Salmond test seems to have been relatively effective. Similarly, a conferral of authority by a defendant on the tortfeasor over the claimant appears to distinguish the liability imposed for deliberate misconduct in the child sexual abuse cases. Unfortunately, the two analogies cannot work together. But they might work separately.
On close examination, the two analogies point to different relationships. The "agency" analogy focuses attention on the relationship between the defendant employer and the tortfeasor employee. In contrast, a conferral of authority focuses attention on the relationship between the defendant and the claimant, over whom authority to the tortfeasor is conferred. These two different relationships may very well support two discrete forms of strict liability for the tort of another, each supported by distinct substantive reasons. 126 A close examination of the relationships giving rise to strict liability for the tort of another also explains the caution with which the term "agency" has been used in this article. When used in a contractual sense, an agency relationship more closely resembles the conferral of authority cases as the agent has been vested with authority by the principal to create a contractual relationship with the claimant. 127 In contrast, there is no need for a relationship between the defendant and the claimant in the employment cases, and the claimant may very well be a stranger to both the employer and the employee. 128 125 Note 75 above. 126 See more generally Beuermann, Reconceptualising Strict Liability. 127 Ibid., ch. 6. 128 Ibid., ch. 5. C.L.J. 523 Discerning the Form at the Second Stage of Vicarious