You are on page 1of 27

Questions of Law (1998), 114 Law Quarterly Review 292-321

Here is a question of law, if your Lordships please to treat it as such.


Lord Denning to the House of Lords1

And be it moon, or sun, or what you please:


An if you please to call it a rush-candle,
Henceforth I vow it shall be so for me.
Katharine to Petruchio2

Lord Denning did not take an analytical approach to identifying questions of law, any
more than Katharine took an analytical approach to identifying the moon. They both
sought a characterization of the phenomena that would do what they needed done--
they took a pragmatic approach.

A consensus has recently emerged among academic commentators on administrative


law: that the courts should take a pragmatic approach to identifying questions of law.
According to the consensus, it is impossible or useless or dangerous to ask what
questions are questions of law. Judges should ask, instead, what questions it would
be useful to treat as questions of law. The consensus is that a pragmatic approach to
identifying questions of law is preferable to an analytical approach. I will argue that a
sound analytical approach is possible, and is capable of being useful, and is incapable
of being dangerous. If the argument succeeds, it will mean that administrative law
scholars need to reconceive the distinction between analytical and pragmatic
reasoning.

The problem of identifying questions of law is introduced in section 1, using the


House of Lords decision in Cozens v. Brutus.3 The problem arises in several branches
of the law; it is worth pointing out some of those branches, if only to see the special
difficulties that the problem creates in administrative law (section 2). Part of the
argument of this article is that it is worth paying closer attention to the techniques that
the judges have developed to cope with the problem (section 3). Those techniques are
diverse and occasionally confused, but a common thread has emerged that supports
the main argument of this article. The common thread is the principle that a question
of application of statutory language is a question of law when the law requires one
answer to it. Section 4 describes the academic consensus in favour of a pragmatic
approach to the problem, and section 5 comments on the nature of analytical and
pragmatic approaches to problems. The main argument of the article is presented in
section 6, and summarised at the end of that section. The argument presents an
account of analysis as resting on normative judgments, and concludes that the

1 Griffiths (Inspector of Taxes) v. J.P. Harrison (Watford) Ltd. [1963] A.C. 1 (HL).
2 Shakespeare, The Taming of the Shrew IV.v.13-15.
3 [1973] A.C. 854 (HL).
2

common thread in the cases corresponds to a sound analytical approach to the


problem. Sections 7 and 8 comment on the extent to which the proposed solution is at
odds with the academic consensus, and is at one with the reasoning of the judges.

1. The Problem

On a June afternoon at Wimbledon, 1971, Dennis Brutus invaded Court Number 2.


Brutus was an anti-apartheid activist. He wanted to disrupt a doubles match featuring
a South African player, Cliff Drysdale. Brutus blew a whistle, and he threw leaflets,
and he sat down on the court. The police dragged him away, and charged him under
the Public Order Act 1936, with the offence of using “insulting” behaviour that was
likely to occasion a breach of the peace.

The magistrates dismissed the charges against Brutus. They decided that his
behaviour had not been insulting. The Divisional Court allowed an appeal by the
police, but the House of Lords restored the magistrates’ decision. Lord Reid did not
just agree with the magistrates. He thought the Divisional Court should not have
heard the appeal. They only had jurisdiction to hear an appeal on a question of law.
Lord Reid held that the question of whether Brutus’s behaviour was insulting was not
a question of law: “The meaning of an ordinary word of the English language,” he
said, “is not a question of law”.4

Was there a question of law at issue in the police appeal? We can see three questions
the magistrates had to answer:

1. Question of Fact: “What did Brutus do on Court Number 2?”

The magistrates had the testimony of the police witnesses to help them answer that
question.

2. Question of Law: “What is the offence?”

The magistrates had the Public Order Act to help them answer that question. When
they read section 5, they found that Brutus’s behaviour had to have been insulting.
And that raises the third question, which we can call a “question of application”:

3. Question of Application: “Was Brutus’s behaviour insulting?”

The magistrates dismissed the information against Brutus because their answer to
Question 3 was “no”. Then the judges in the Divisional Court had to address a new
question: did the Divisional Court have the power to reverse the magistrates’
decision? It could only interfere on questions of law. So look back at the three
questions the magistrates faced. Question 1 is not a question of law.5 Question 2 is a

4 [1973] A.C. 854 at p.861.


5 The court may interfere with the tribunal’s answer to question 1, but not simply on the ground that it
was mistaken: see below section 3.9. See also Timothy Jones, “Mistake of Fact in Administrative
Law” [1990] P.L. 507.
3

question of law, and if the magistrates had, e.g., said that the Act should be read to
prohibit any inconsiderate behaviour, the Divisional Court could have reversed their
decision. But the magistrates simply read it as it was written-- to say that there was
no offence unless Brutus’s behaviour was insulting. So they seem not to have erred
on a question of law, unless Question 3 is a question of law.

What about the vexed Question 3, the question of application? Is that a question of
fact, or a question of law? Or is it some other animal? The judges of the Divisional
Court had to solve that problem (I will call it “the problem”) in order to decide if they
could interfere with the magistrates’ decision.

2. The Scope of the Problem

The distinction between law and fact can be important in a variety of legal contexts:
mistakes of fact and mistakes of law may be treated differently in criminal law, and
mistakes or misrepresentations of fact and of law may be treated differently in
contract law. Rules of pleading typically require a party to plead facts, and prohibit
the pleading of law.

But the most common use of the notion of questions of law is to distribute decision-
making power and responsibility. It is the standard device that common law systems
have used to order relations between two decision-makers. The obvious example is
jury trials: judges answer questions of law, and juries answer questions of fact.6 But
the doctrine of precedent is an example too. The common law distributes decision-
making power by enabling the court that set the precedent to bind a later court on
questions of law, but not on questions of fact. The final example, and the most
important for our purposes, is that courts reviewing the decisions of other decision-
makers are often given power to interfere on questions of law, but not on questions of
fact. Arbitrators are one such decision-maker.7 But most cases arise in three areas on
the fringes of administrative law: appeals from magistrates (as in Brutus), and from
the Employment Appeals Tribunals,8 and (in vast numbers) from the tax
commissioners.9 The same distinction also governs the scope of review in the core
areas of administrative law: dozens of statutes provide for appeals on “a question of
law”.10

6 For qualifications on this generalisation and for a discussion of the problem in the context of jury
trials see A.A.S.Zuckerman, The Principles of Criminal Evidence (1989), 24-32.
7 Arbitration Act 1979 s.1(2) (appeal to a court on a “question of law arising out of an award”). See
Vitol SA v. Norelf Ltd. [1996] 3 W.L.R. 105 (HL).
8 Employee Protection (Consolidation) Act 1978 s.136; Trade Union and Labour Relations
(Consolidation) Act 1992 s.291(2).
9 Appeal may be brought from the General or Special Commissioners to a court on a question of law:
Taxes Management Act 1970, s.56.
10 E.g. Child Support Act 1991, s.25 (appeal from Child Support Commissioner to a court); Social
Security Administration Act 1992, s.24 (appeal from Commissioners to a court); Asylum and
Immigration Appeals Act 1993, s.9 (appeal from Immigration Appeal Tribunal to a court). Appeals
can be brought from twelve types of tribunals (including rent assessment tribunals and VAT tribunals)
under the Tribunals and Inquiries Act 1992 by a party “dissatisfied in point of law” (s.11 and Schedule
1).
4

Most dramatically, the judges have developed their own supervisory review of
administrative decision making so that they can reverse an administrative decision on
a question of law without a statute conferring the power to do so. Since Anisminic,11
the judges have refined jurisdictional review to such a pitch of sophistication that the
word “jurisdiction” is no longer needed, and a court will simply substitute its view for
that of the administrative decision-maker on any question of law. We can call this
form of review the “Page doctrine”, after its most unequivocal assertion in the House
of Lords:

...in general any error of law made by an administrative tribunal or inferior


court in reaching its decision can be quashed for error of law.12

Under the Page doctrine, review for error of law is its own rationale. Why is judicial
review available against an error of law? Because it is an error of law.

The remarkable feature of the Page doctrine is the way in which it assimilates
supervisory review to appellate review, at least in this respect: that supervisory
review (generally) and statutory appeals (typically) are both available to correct errors
of law.13 It is a commonplace in the lore of administrative law that an appeal on a
question of law is to be distinguished from supervisory review;14 the House of Lords
has partly abolished the distinction.15

This article will focus on the problem in the context of administrative law. The
argument may have implications for jury trial and for the doctrine of precedent. But
in those contexts matters are somewhat more straightforward. In both, the decision-
maker with power to decide questions of law acts first. In jury trials, the judge directs
the trier of fact. Difficult issues arise as to how much the judge should tell the jury,
but those issues get resolved without conceptual headaches, by rules governing

11 Anisminic Ltd. v. Foreign Compensation Commission [1969] 2 A.C. 147 (HL).


12 R. v. Lord President of the Privy Council ex p. Page [1993] A.C. 682 (HL), per Lord Browne-
Wilkinson at p.702, drawing on remarks by Lord Diplock in Re Racal Communications Ltd. [1981]
A.C. 374 (HL) and O'Reilly v. Mackman [1983] 2 A.C. 237 (HL). Cf. Lord Slynn in Page: “certiorari
is now available to quash errors of law in a decision” (at p.706). None of those decisions turns on the
courts’ general power to review for error of law, but there is no doubting the authority of such
straightforward, consistent statements in the House of Lords.
13 Thus Lord Griffiths supported the majority view in Page, that university visitors have a special
exemption from review for error of law, by saying, “...to admit certiorari to challenge the visitor’s
decision on the ground of error of law will in practice prove to be the introduction of an appeal by
another name.” [1993] A.C. 682 at p.674. The implication is that the House of Lords should not
introduce such a form of appeal. Lord Griffiths did not comment on the propriety of introducing “an
appeal by another name” from decision-makers other than university visitors.
14 Sir William Wade and Christopher Forsyth (Administrative Law (7th ed., 1994) [“Wade and
Forsyth”]) retain the claim that appellate review concerns the merits of the tribunal’s decision while
supervisory review concerns its legality (at pp.38, 954). But appellate review on a question of law only
concerns the decision’s legal merits; and under the Page doctrine, reviewing the legal merits of a
decision is reviewing its legality.
15 Remains of the distinction between supervisory review and appeal are: (i) procedures differ; (ii) it
may be that a party entitled to an appeal on a question of law is sometimes unable to raise jurisdictional
issues (but see Wade and Forsyth at pp.954-955 for an argument that this is never the case); (iii)
statutory provisions may give the court powers that it does not have in supervisory review to substitute
its own remedy for that of the tribunal (see Peter Cane, An Introduction to Administrative Law (2d ed.,
1992) [“Cane”] at pp.8-10); and (iv) although the Page doctrine sets the same standard of review for
both, judges may incline to exercise more self-restraint in supervisory review.
5

directions (indeed, the problem is often addressed by talking of “questions for the
judge” and “questions for the jury” rather than “questions of law” and “questions of
fact”16). In the doctrine of precedent, as in jury trial, the decision-maker with power
to bind on questions of law acts first. And in this context the need to interpret
previous cases and the power to distinguish them provide very flexible techniques for
addressing the problem: it is the decision-maker that is bound on questions of law
that decides what answers the first decision-maker has given to questions of law.

In judicial review,17 the process is reversed: the decision-maker with power to bind on
questions of law makes its decision after the other decision-maker has already given a
decision. Imagine that a jury gave its verdict first, without instruction, and then a
judge had to decide whether they had erred in law. Or imagine that the doctrine of
precedent worked backwards, with the second court passing judgment on the legal
validity of the previous decision. These topsy-turvy scenarios point out what is
difficult about distribution of decision-making power in judicial review: it is an
institutionalised form of second-guessing.18

So in what follows I will concentrate on the problem in administrative law, for two
reasons: because that is where the interesting difficulties arise, and because the
problem has such far-reaching practical implications for the scope of judicial review,
especially under the Page doctrine. For convenience I will call the decision-maker
whose decision is under review the “tribunal”, and the reviewing decision-maker the
“court”.19

3. The Judges

Looking at the devices courts have used to address the problem is like looking into the
average toolbox. There is a lot of clutter that could have been cleared out long ago.
There are one or two baffling gadgets with no readily identifiable function. And there
are a few old, sturdy, and serviceable tools that do all the work. It would be a mistake
to think that the toolbox is useless, just because it is messy.

16 So, for instance, whether particular words are “defamatory” is a question for the jury, and the courts
reach that conclusion without agonizing over whether that is a question of law or of fact. See e.g.
Botham v. Khan, The Times, 15 July 1996 (CA).
17 Taken in the widest sense to include review of arbitrators’ decisions and statutory appeals on
questions of law as well as supervisory review under the Page doctrine.
18 The closest analogue of the judicial review situation is appellate review within the hierarchy of
superior courts. But there the notion of questions of law is not used to determine whether the appeal
court can interfere: although the Court of Appeal is circumspect about interfering with conclusions of
fact, it has the power to do so. As a result, there is no pressing need to identify questions of law. See
R.S.C.Ord.59, r.10(3).
19 In fact, the decision-maker whose decision is under review may be a court, or an administrative
agency other than a tribunal, and the reviewing decision-maker may be a tribunal (e.g., the
Employment Appeals Tribunal can hear appeals from Industrial Tribunals on questions of law), or a
commissioner, or even the Secretary of State.
6

At first glance the cases look inconsistent, as if the judges were addressing the
problem ad lib. Some suggest that questions of application are questions of law,20
some suggest that questions of application are questions of fact,21 and a much more
numerous class of cases concludes that a question of application may be a question of
fact or a question of law, depending on a perplexing variety of flexible distinguishing
criteria.

We can simplify matters first by pointing out that cases of the first kind are few and
not recent.22 And no case of the second kind is consistent with itself. Puhlhofer is an
example. The applicants were refused housing assistance on the ground that they
were not homeless, because they already had “accommodation” (for their family of
four, the applicants had one room in a bed and breakfast, with no cooking or washing
facilities). Lord Brightman said that “What is properly to be regarded as
accommodation is a question of fact to be decided by the local authority. There are
no rules.”23 But that statement was too general: Lord Brightman admitted in his next
breath that the court could interfere in some conceivable cases -- on the very question
of what is properly to be regarded as accommodation:

Clearly some places in which a person might choose or be constrained to live


could not properly be regarded as accommodation at all; it would be a misuse
of language to describe Diogenes as having occupied accommodation within
the meaning of the Act.24

That, at least, is a rule. Suppose that Diogenes applied to his local authority for
housing assistance, and the local authority said, “you are not homeless, because you
have a barrel available to you as accommodation.” The local authority’s decision
would be reversed for error on a question of law.

The question of the application of the word “accommodation” was a question of fact
in Puhlhofer. But the question of application of the same word would be a question
of law if the applicants lived in a barrel. Similar conclusions provide the explicit
common thread in the numerous cases that hold that a question of application may be
a question of law or a question of fact. According to the courts, the question of the
application of a single statutory term can be a question of fact in one case, and a
question of law in another. This common thread in the cases needs explaining.

It might be thought that we have to identify the question whether behaviour is


insulting as a question of law or a question of fact, so that either every decision of the
magistrates applying s.5 of the Public Order Act 1936 was subject to review, or none
was. But whether behaviour is insulting is not actually a question; it is a fragment of

20 Farmer v. Cotton’s Trustees [1915] A.C. 922, Woodhouse v. Peter Brotherhood Ltd. [1972] 2 Q.B.
520 at p.536, Pearlman v. Keepers and Governors of Harrow School [1979] Q.B. 56 per Lord
Denning, A.C.T. Construction Ltd. v. Customs and Excise Commissioners [1981] 1 W.L.R. 49 (CA)
per Lord Denning at p.54.
21 E.g. Simmons v. Heath Laundry Co. [1910] 1 K.B. 543 (CA) per Cozens-Hardy M.R., Fletcher
Moulton L.J., and Buckley L.J. (doubted by Ackner L.J. in O’Kelly v. Trusthouse Forte P.L.C. [1984]
1 Q.B. 90 at pp.113, 114; R. v. Hillingdon L.B.C. ex p. Puhlhofer [1986] 1 A.C. 484 (HL) at p.517.
22 But see the discussion of Lord Templeman’s speech in Smith v. Abbott, at n.33 below.
23 [1986] 1 A.C. 484 at p.517.
24 [1986] 1 A.C. 484 at p.517.
7

a question (try asking yourself, “is behaviour insulting?”). The question that had to
be identified as a question of fact or a question of law in Brutus v. Cozens was this:
“was Brutus’s behaviour insulting?” (See “Question 3”, above)

Judges (like Lord Reid) and academic commentators sometimes claim or assume that
(analytically) either every question of the application of, e.g., s.5 is a question of law,
or every question of its application is a question of fact. But we will see that the
mainstream of judgments support the view (i) that the question of the application of
s.5 may be a question of fact in one case, and a question of law in another case, and
(ii) that the cases in which the question of application is a question of law are cases in
which the law requires one answer. This common feature will show itself again and
again when we examine the variety of devices in the judicial toolbox.

3.2 Questions only a trained lawyer can answer

One of several tools that Lord Denning has devised is to ask whether “the correct
conclusion to be drawn from the primary facts requires, for its correctness,
determination by a trained lawyer”.25 If the conclusion “can as well be drawn by a
layman”, it is a conclusion on a question of fact. What can Lord Denning have
expected of trained lawyers? He gives instances of what would make a trained lawyer
necessary, but they do not help: “because it involves the interpretation of documents
or because the law and the facts cannot be separated, or because the law on the point
cannot properly be understood or applied except by a trained lawyer”. The first is
irrelevant to the problem (unless statutes are “documents”), the second is baffling, and
the third is redundant. The problem is that most of the cases on questions of law
involve the application of straightforward (though vague) statutory expressions, in
contexts that the tribunal is familiar with: “insulting”, “accommodation”, “structural
alterations”, “maintenance”, “trade”, “plant”, and so on. The application of such
terms is difficult in some cases, but the difficulties do not call for a trained lawyer.
They call for good sense and attention to the context. At least concerning questions
of the application of statutory provisions such as these, what is at stake is not whether
a trained lawyer is better at applying the expression than a layperson, but whether the
court ought to substitute its judgment for that of the tribunal. In any case, the courts
have not followed Lord Denning’s lead in this case: the trained lawyer test is defunct.

3.3 Questions of application of ordinary English words are questions of fact

Related to the trained lawyer test is the technique that Lord Reid contributed in
Brutus’s case: “The meaning of an ordinary word of the English language is not a
question of law.”26 This is an unstable test --not because there is no such thing as
ordinary words with ordinary meanings, but because ordinary words can be used in
special senses, and the question can always arise whether that is what Parliament has

25 British Launderers’ Research Association v. Borough of Hendon Rating Authority [1949] 1 K.B.
462 (CA) at p.472. Primary facts are “facts which are observed by witnesses and proved by oral
testimony” (at p.471).
26 [1973] A.C. 854 at p.861.
8

done. Lord Reid himself made that clear.27 Thus the ordinary word test has been
shaken by Lord Scarman’s speech in R. v. Barnet LBC ex p. Shah:

Though the meaning of ordinary words is, as Lord Reid observed in Cozens v.
Brutus..., a question of fact, the meaning to be attributed to enacted words is a
question of law, being a matter of statutory interpretation.28

Moreover, consider what would have happened if the magistrates had convicted
Brutus. Lord Reid had good reason for his view that Brutus’s behaviour was not
insulting, and it seems (though he does not discuss it) that he would have reversed the
magistrates if they had convicted. Yet if they had convicted, they would have done so
simply by erring on a question of the application of an ordinary English word.
What’s more, Lord Reid did not say that the magistrates’ application of an ordinary
word is unreviewable: he actually offered a reasonableness test.29 That test would
have given him a technique for reversing the magistrates if they had convicted. Which
goes to show that, even in cases where ordinary words are used in an ordinary sense,
we cannot accept Lord Reid’s statement that the meaning of ordinary words is not a
question of law, without the qualification that the tribunal errs on a question of law if
it applies the term unreasonably. So Brutus itself suggests that we can dispense with
the notion of ordinary words altogether and use (something like) a reasonableness test
(see section 3.8, below).

3.4 Mixed Questions of Fact and Law

Here is one of the baffling gadgets in the judicial toolbox: questions of application
are often called mixed questions of fact and law.30 The notion has even crept into the
statute books, though only as part of an extravagant device for creating a power to
hear appeals on any question.31 The nature of the mixture is unexplained, and it
seems that “mixture” is actually a rather unhelpful low-voltage metaphor: a question
of application does not mix fact and law, it asks the decision-maker to apply the law to
the facts.

It has sometimes been supposed that a court cannot entertain mixed questions of fact
and law:

...the appeal tribunal has no jurisdiction to consider any question of mixed fact
and law until it has purified or distilled the mixture and extracted a question of
pure law.32

27 Immediately after the statement that the meaning of an ordinary word is not a question of law, he
wrote that “The proper construction of a statute is a question of law. If the context shows that a word is
used in an unusual sense the court will determine in other words what that unusual sense is.”
28 [1983] 2 A.C. 309 (HL) at p.341. Cf. R. v. Poplar Coroner, ex p. Thomas [1993] Q.B. 610 (CA)
per Simon Brown LJ at p.630: “Cozens v. Brutus..., in short, seems to me of limited value in this case:
even ordinary words can have more than one usual sense and be capable of differing applications
depending upon the particular context in which they are found.”
29 [1973] A.C. 854 at pp.861-862. Reasonableness tests are discussed below, section 3.8.
30 Wade and Forsyth trace the phrase back as far as Johnstone v. Sutton (1785) 1 T.R. 510 at p.545.
31 Criminal Appeal Act 1968, s.1(2).
32 O’Kelly v. Trusthouse Forte P.L.C. [1984] Q.B. 90 per Lord Donaldson M.R. at p.123.
9

But the contrary view is current in the House of Lords. Considering whether
journalists read newspapers “in the performance of” their duties, Lord Mustill simply
held that the question was “one of mixed fact and law which the court is entitled to
review”.33

The most charitable interpretation can make nothing more of the phrase “mixed
question of fact and law” than an unexplained report of a conclusion: that a decision
can be reviewed, or cannot be reviewed, depending on the effect taken to attach to the
phrase.

3.5 Questions of Fact and Degree

Like “mixed question of fact and law”, this is a baffling turn of phrase. But at least its
effect is clear: calling a question a “question of fact and degree” (or a “question of
degree”) always implies that the tribunal’s judgment is protected. The phrase trips
readily off the tongues of administrative lawyers and judges, 34 and has done so since
the early tax commissioner cases a century ago. Yet it sounds inept.35 There are clear
questions of law that are questions of degree (“when can the House of Lords overrule
itself?” is an example). It might seem that a question of application is a question of
degree if the term to be applied is imprecise --yet we will see that imprecise terms can
apply as a matter of law.

To understand what is going on when judges talk of questions of degree, we need to


look at the single most important case on questions of law: Edwards v. Bairstow. 36
Lord Radcliffe gave the inept phrase an undying prestige by expressly making the
relation between questions of degree and questions of fact an entailment: he spoke of
questions which “can be described as questions of degree and therefore as questions
of fact”.37 But look at the questions that he called questions of degree: “All these
cases in which the facts warrant a determination either way can be described as
questions of degree and therefore as questions of fact.”38 He did not say that a
question of application is a question of fact. In fact, he said that he would
“deprecate... too much abbreviation in stating the question, as by asserting that it is

33 Smith (Inspector of Taxes) v. Abbott [1994] 1 All E.R. 673 (HL) at p.691; Lord Jauncey concurred
in this characterization of the question. Lord Browne-Wilkinson, dissenting, relied on Lord Radcliffe’s
speech in Edwards v. Bairstow (see below at n.36). The most unusual approach was Lord
Templeman’s: he simply called the question of application a question of law (at p.685). His speech
hints at (but does not avow) the radical view that all questions of application are questions of law; if
that was his view it was not adopted by the other Law Lords.
34 See e.g. A.C.T. Construction Ltd. [1981] 1 W.L.R. 1542 (HL) per Lord Roskill at p.1547, Cole
Bros.Ltd. v. Phillips (Inspector of Taxes) [1982] 1 W.L.R. 1450 (HL) per Lord Hailsham at p.1455,
and Ransom (Inspector of Taxes) v. Higgs [1974] 1 W.L.R. 1594 (HL), per Lord Wilberforce and Lord
Simon. Cf. Currie v. Commissioners of Inland Revenue [1921] K.B.D. 332 (CA) per Lord Sterndale at
p.336; see also Scrutton LJ at p.341; Octavius Jepson v. Frederick Wynne Gribble (Surveyor of Taxes)
[1895] 1 Tax Cas. 78 (Exch.).
35 Paul Craig calls it “a confusing tag”: Administrative Law (3rd ed. 1994) [“Craig”] at p.159.
36 [1956] A.C. 14 (HL).
37 [1956] A.C. 14 at p.33.
38 Ibid., emphasis added. Cf. Octavius Jepson v. Frederick Wynne Gribble (Surveyor of Taxes) [1895]
1 Tax Cas. 78 (Exch.): “There it was a matter of degree; here it is no matter of degree at all.”
10

simply a question of fact whether or not a trade exists”;39 it is a question of fact only
in cases in which the tribunal cannot be said to be wrong:

...the law does not supply a precise definition of “trade”... In effect it lays
down the limits within which it would be permissible to say that a “trade”...
does or does not exist.40

So “question of degree” is an unfortunate phrase for a technique which is based on


whether the object in question is a borderline case for the application of a vague
statutory term such as “trade”. The question is whether the facts warrant a decision
either way.

We can draw the same conclusion from cases decided decades before Edwards v.
Bairstow: in Currie v. Commissioners of Inland Revenue, Lord Sterndale MR had
said that a question of application is a question of law in a case “in which nobody
could arrive at any other conclusion. ...That reduces it to a question of law....” So a
question of application is a question of law in cases he calls “extremes”. “But
between those two extremes there is a very large tract of country in which the matter
becomes a question of degree; and where that is the case the question is undoubtedly,
in my opinion, one of fact”.41

“Question of degree” is simply a label for a question of application in a case in which


the law permits a decision either way. I propose that the important tests for
identifying questions of law are all very similar to this test, and are united by the
theme we see in Currie and Edwards v. Bairstow: they all amount to asking whether
the law requires one answer to the question of application. The phrase “question of
degree” is a label for that technique and adds nothing to it, except a faint air of
confusion. But the technique itself is very important, and the courts have found
several other ways of referring to it, which are identified in sections 3.6 to 3.9.

3.6 Questions “capable of decision either way”

Lord Lowry formulated the test for questions of law in this way in Inland Revenue
Commissioners v. Scottish & Newcastle Breweries Ltd.42 The Special Commissioners
held that decorations in the taxpayer’s hotel, including sculptures of seagulls in flight,
were “plant” under the Finance Act 1971. Lord Lowry held that a case “capable of
decision either way” raises “a question of fact and degree” on which the tribunal’s
judgment is protected.

The same technique was at work in Edwards v. Bairstow: “the facts fairly admitting
of the determination come to, there is no error which justifies the court’s
intervention.” 43 It is an approach that is at least as old as a 1914 judgment of

39 [1956] A.C. 14 at p. 34.


40 [1956] A.C. 14 at p. 33.
41 Currie, [1921] K.B.D. 332 at p.336. (Cf. Scrutton LJ at p.341: “All these cases which involve
questions of degree seem to me to be eminently questions of fact.”)
42 [1982] 1 W.L.R. 322 at p.327.
43 Lord Radcliffe at p.34; cf. “If the facts of any particular case are fairly capable of being so
described, it seems to me that it necessarily follows that the determination... is not ‘erroneous in point
11

Scrutton J. on the power to overturn decisions in an appeal by case stated: “...if the
facts stated are such that you may come to one conclusion or the other, the fact that
the Court itself would have come to a different conclusion from that which the
Commissioners have come to is no reason for disturbing the decision of the
Commissioners.”44

3.7 A Matter of Law

Another related way of deciding whether the question of application in a case is a


question of law is by asking whether it is answered “as a matter of law”:

...the interpretation of the language of an Act of Parliament often involves


declaring that certain conduct must as a matter of law fall within the statutory
language (as was the actual decision in Edwards v. Bairstow); that other
conduct must as a matter of law fall outside the statutory language; but that
whether yet a third category of conduct falls within the statutory language or
outside it depends on the evaluation of such conduct by the tribunal of fact.
This last question is often appropriately described as one of “fact and
degree”.45

Lord Simon was right to suggest that this scheme is a formulation of the Edward v.
Bairstow doctrine: if as a matter of law it is not insulting merely to disrupt other
people's lawful activities, then the law does not permit the tribunal to find the
behaviour insulting.

The same technique is also used in a recent House of Lords decision on a statutory
appeal from an arbitrator.46 The arbitrator had held that the buyers of a shipment of
propane had repudiated the contract, and that the sellers had communicated their
acceptance of the repudiation by taking no further steps to perform. Lord Steyn held
that “The only question is whether the relevant holding of the arbitrator was wrong in
law.”47 The issue for the court was “whether non-performance of an obligation is
ever as a matter of law capable of constituting an act of acceptance”.48

This technique fits the pattern that we see emerging: cases in which it is permissible
to say that the statutory term applies or does not, cases capable of decision either way,
cases in which the tribunal’s decision is not wrong as a matter of law, are cases in
which the tribunal has not erred on a question of law.

of law’” (at p.33). See also Viscount Simonds at p.29, and Lord Roskill’s similar approach in A.C.T.
Construction Ltd. v. Customs and Excise Commissioners [1981] 1 W.L.R. 1542 (HL) at p.1547.
44 Smith v. Incorporated Council of Law Reporting for England and Wales [1914] 3 K.B. 674 at
p.683. Cf.: “to quote again from the judgment of Pearson LJ [in Jarrold v. John Good & Sons Ltd.
[1963] 1 W.L.R. 214 at p.225]: ‘Either view could have been taken’; in other words the question was
one of fact.” Cole Bros.Ltd. v. Phillips (Inspector of Taxes) [1982] 1 W.L.R. 1450 (HL) per Lord
Hailsham at p.1457.
45 Ransom (Inspector of Taxes) v. Higgs [1974] 1 W.L.R. 1594 (HL), per Lord Simon of Glaisdale at
p.1618 [emphasis added].
46 Vitol SA v. Norelf Ltd. [1996] 3 W.L.R. 105 (HL) per Lord Steyn.
47 [1996] 3 W.L.R. 105 at p.113.
48 [1996] 3 W.L.R. 105 at p.113 [emphasis added]. Lord Morris also took the “matter of law”
approach in Brutus: [1973] A.C. 854 at pp.863-864.
12

3.8 Reasonableness?

Can we simply sum up this pattern by saying that a question of application is a


question of fact and degree in a case in which it would be reasonable to answer the
question of application “yes” or “no”? We can certainly phrase the test for questions
of law in those terms, because reasonableness is such a flexible tool. But doing so
will not actually simplify anything.

The courts have often taken the reasonableness resort49: we saw that Lord Reid did so
in Brutus v. Cozens. And of the possible ways of stating the doctrine in Edwards v.
Bairstow, Lord Radcliffe’s own preferred formulation was that a tribunal errs in law if
“the true and only reasonable conclusion contradicts the determination”.50 This
technique seems to explain the pattern we have seen: if a decision either way is
reasonable, the court will not interfere, but if only one decision is reasonable, the
court will hold the tribunal to it.

But a reasonableness test leaves us with some explaining to do. In identifying


questions of law, the court is trying to decide whether it has the power to substitute
judgment. How can the existence of the power to apply a test of correctness be a
matter of reasonableness?

Here is a possible answer to this puzzle: the reasonableness test used in identifying
questions of law amounts to a test of clear cases, with the same effect as the “capable
of decision either way” test. There is a silent premise in the use of a reasonableness
test to address the problem: that it would be unreasonable for a tribunal to hold that
the statutory language applies (or does not apply) when it clearly does not (or does).
If you come to the conclusion that Brutus’s behaviour was not insulting, you might
still conclude that it is reasonable to say that it was insulting. But if Brutus’s
behaviour was clearly not insulting, it would be unreasonable for someone to say that
it was insulting. Notice that Lord Reid, in stating the reasonableness test he proposed
to use in cases on the application of ordinary words, tied the test to the use of
language:

The question would normally be whether their decision was unreasonable in


the sense that no tribunal acquainted with the ordinary use of language could
reasonably reach that decision.51

It seems that the reasonableness test for questions of law is simply another way of
putting the “capable of decision either way” test. We might say that any case is
capable of decision either way, in the sense that it is conceivable that a magistrate (or
even a Divisional Court) could call behaviour “insulting” when it is clearly not. But
that would be unreasonable. So we would not be tampering with Lord Lowry’s
phrase if we changed it to “reasonably capable of decision either way”. The point to

49 See e.g. Griffiths (Inspector of Taxes) v. J.P Harrison (Watford) Ltd. [1963] A.C. 1 (HL) per
Viscount Simonds at p.11, Lord Reid at p.16, Lord Morris at p.23.
50 [1956] A.C. 14 at p.36. Lord Radcliffe borrowed the phrase “true and only reasonable conclusion”
from Lord Cooper’s speech in I.R.C. v. Toll Property Co.Ltd. [1952] S.C. 387 (HL), at p.393.
51 [1973] A.C. 854 at p.861. And cf. Lord Brightman in Puhlhofer [1986] 1 A.C. 484 (HL) at p.517:
“it would be a misuse of language to describe Diogenes as having occupied accommodation within the
meaning of the Act”.
13

bear in mind is that “reasonableness” can be used as a way of referring to the test that
we have seen described in other terms (...the tribunal cannot be said to be wrong;
cases capable of decision either way; facts warrant a decision either way; the decision
is not wrong as a matter of law...). It is not an alternative to that test.

3.9 No evidence?

A tribunal errs in law when it makes a finding of primary fact which there is no
evidence to support.52 Can this rule solve the problem? Why not simply say that a
tribunal’s answer to a question of application can be set aside if there is no evidence
to support it?53 Lord Radcliffe considered the possibility: “I do not think that it much
matters whether this state of affairs is described as one in which there is no evidence
to support the determination... or as one in which the true and only reasonable
conclusion contradicts the determination”.54 But he preferred the latter description.
He said that it was “rather misleading to speak of there being no evidence”. It is
misleading because lack of evidence is a problem of proof, and the problem in a case
like Brutus is not a problem of proof. The problem in Brutus was not whether one
party had proven the facts it alleged, but whether the facts alleged and proved by the
police supported the application of the statutory term. If a court said that there is no
evidence to support the finding that Brutus’s behaviour was insulting, that could mean
nothing more or less than that the term “insulting” cannot be applied to Brutus’s
behaviour. It is not that some piece of evidence was missing-- that we do not know
enough about what Brutus did to form a judgment that his behaviour was insulting. It
is that there is nothing in what he did that amounts to insulting behaviour. If that is
right, the no evidence rule has no part to play in solving the problem: it applies only
to findings of primary fact.

It has been argued that the judges must be operating a no evidence test, because if
they treat a question of application as a question of fact in a borderline case they are
committed to treating a question of application as a question of fact in clear cases too-
- and then the only legal ground for interfering in a clear case is lack of evidence. 55
This argument refuses to accept the common thread in the Edwards v. Bairstow line
of cases: it makes nonsense of the judges’ consistent practice of treating a question of
application of the same statutory term as a question of fact in one case and a question
of law in another.

3.10 Conclusion

Now we can tidy up the toolbox. Talk of questions a layperson could answer and
questions that need a trained lawyer is obsolete, and talk of ordinary versus technical
language is obsolescent. Nothing is gained by talking of mixed questions of law and

52 For one of many House of Lords statements of the rule, see I.R.C. v. Scottish & Newcastle
Breweries Ltd. [1982] 1 W.L.R. 322 per Lord Lowry at p.327.
53 This was done in Bean v. Doncaster Amalgamated Collieries Ltd. [1944] 2 All E.R. 279 (CA) per
DuParcq LJ at p.284.
54 [1956] A.C. 14 at p.36. Viscount Simonds treated the question of application in Edwards v.
Bairstow as a question of fact, on which the tax commissioners could be reversed “if it appears that the
commissioners have acted without any evidence or upon a view of the facts which could not reasonably
be entertained” (at p.29).
55 Etienne Mureinik, “The Application of Rules: Law or Fact?” (1982), 98 L.Q.R. 587 at pp.615-616.
14

fact (though the courts do so). The no evidence rule plays no role in solving the
problem. Even the ubiquitous technique of characterising questions as questions of
degree does no work: it only states a conclusion.

One surprisingly consistent56 set of techniques does the work. We can call it the
Edwards v. Bairstow doctrine (since Lord Radcliffe’s speech is invoked so often),
though Lord Sterndale and Lord Justice Scrutton had set it out as clearly 35 years
earlier in Currie.57 These techniques treat a question of application as a question of
law in a clear case of the application of the statutory term (a case in which the law
requires a particular answer to the question of application), and as a question of fact in
an unclear case (a case capable of decision either way). We can call this a
reasonableness test, as the courts sometimes do, as long as we see that the sense in
which the tribunal must act reasonably (if it is not to be held to have erred on a
question of law) is that it must not decide that the statutory term applies when it
clearly does not, or that it does not apply when it clearly does. The House of Lords
has endorsed the technique repeatedly, in Toll Property (1952), Edwards v. Bairstow
(1956), Ransom (1982), Scottish and Newcastle Breweries (1982), A.C.T.
Construction (1982), Cole (1982), and Smith v. Abbott (1994).58

The test is vague and flexible (its effect depends, as it should, on the statutory
language in question), and there are many cases in which it would be impossible to
predict how a court will apply it. So it can be manipulated and misappropriated, as
Lord Justice Scrutton recognised in 1921:

...there has been a very strong tendency, arising from the infirmities of human
nature, in a judge to say, if he agrees with the decision of the Commissioners,
that the question is one of fact, and if he disagrees with them that it is one of
law, in order that he may express his own opinion the opposite way.59

But the infirmities of human nature respond in a similarly wayward fashion to all the
vague standards of administrative law. There is nothing particularly inconsistent in
the court’s use of the question of law standard.

4. The Academic Consensus

What have legal scholars done with the problem? Recently they have developed a
surprisingly strong consensus, which we can paraphrase roughly as follows:

On an analytical or “logical” or a priori approach, either (i) all questions of


application are questions of law, or (ii) there is no answer to the problem. On
a pragmatic or functional or a posteriori approach, questions of application
are treated as questions of law when it is appropriate for the court to interfere.

56 Pace Wade and Forsyth: “there can hardly be a subject on which the courts act with such total lack
of consistency as the difference between fact and law.” (948).
57 And in Edwards v. Bairstow, Lord Radcliffe thought that he was expressing “settled law” (at p.35).
58 Per Lord Browne-Wilkinson, dissenting; the majority did not disagree on the test for questions of
law.
59 Currie [1921] K.B.D. 332 at p.339.
15

Under the Page doctrine, the analytical approach is intolerable: either (i) it
would lead the court to substitute judgment on all questions of application
(because all questions of application are questions of law), or (ii) it would be a
sham for the judge’s whim (because there is no analytical answer to the
problem).

I will call this view “the consensus”. It unites the leading administrative law scholars
in Britain.60 Some writers have dissented from the consensus,61 and some of those
who support it would qualify it in important respects. But we can fairly say that the
common opinion is that judges could follow either an analytical or a pragmatic
approach to the problem, and that they would do well to choose the pragmatic
approach.

Is there any judicial support for the consensus? We might say that it cannot lack for
judicial support: if all questions of application are questions of law, then every case
in the Edwards v. Bairstow line takes a pragmatic approach; if there is no analytical
solution to the problem, then every case on the problem must be taking a pragmatic
approach. In either case, judges might hesitate to own up to what they are doing: they
like to portray themselves as applying the law. Nevertheless, a few judges have
deliberately adopted the approach that the consensus recommends. Lord Denning is
their champion: he followed an unswerving rule of calling a question “a question of
law” when he wanted to.62 But avowed pragmatists are few enough that the consensus
has to treat most judges as taking either an unjustifiable analytical approach, or a
clandestine pragmatic approach.

5. Analytical and Pragmatic Approaches

What is an analysis of a distinction (or a concept, or a doctrine...)? It is an attempt to


understand the distinction: to clarify the meaning of the terms being distinguished.
Even if an analysis enables a judge to identify a question of application as a question
of law, it will not dictate how to decide the case. Judicial reasoning is practical
reasoning, and analytical reasoning cannot exhaust practical reasoning. If a question
of application is identified as a question of law, other questions remain, such as
whether the tribunal erred in answering the question of application, and whether the
law requires substitution of judgment on a question of law, and whether to follow the

60 Cane at pp.108-110; Craig at pp.157-160, 347-349; Wade and Forsyth at pp.947-950; Beatson, “The
Scope of Judicial Review for Error of Law” (1984) 4 O.J.L.S. 22, at pp.40-45. The view taken in de
Smith, Woolf & Jowell Judicial Review of Administrative Action (5th ed. 1995) is complex but seems
to fit the consensus: the question of application “could as well be styled one of law as one of fact” (at
p.279). See also S.H.Bailey, B.L.Jones and A.R.Mowbray, Cases and Materials on Administrative
Law (2d ed. 1992) at p.640, and Gwyneth Pitt, “Law, Fact and Casual Workers” (1985) 101 L.Q.R. 217
at pp.223-231.
61 Mureinik, supra n.55, arguing that all questions of application are analytically questions of law, and
that courts ought to treat them as such; C.T.Emery and B.Smythe, Judicial Review (1986) ch.3-4, esp.
at pp.117-127, arguing that questions of application are sui generis, and that an unreasonable decision
on a question of application indicates that the tribunal has made an unstated error of law.
62 “I find that when a tribunal has gone wrong, the High Court is usually able to find that it has made
an ‘error of law’.” The Discipline of Law (1979) at p. 66. See the judgments cited above at notes 20
and 25.
16

law, or to change it, or to ignore it. An analytical approach to the problem is an


approach that treats the analytical question, “what questions are questions of law?” as
a step in reaching a practical conclusion-- the decision in a case. That approach seeks
grounds for distinguishing questions of law, and treats those grounds as part of the
judge’s grounds for decision in a case.

A pragmatic approach (as the consensus views it, and as I will use the term)
concludes that analytical considerations cannot usefully function as a step in judicial
reasoning. On that view, analysis provides no grounds for distinguishing questions of
law that can be adequate to the task of identifying questions on which the court ought
to substitute its judgment for that of the tribunal. That conclusion is justified if
analysis yields no grounds for deciding whether a question of application is a question
of law, and it may arguably be justified if analysis yields inadequate grounds for
identifying questions of law-- grounds that expose error of law as an unsatisfactory
standard for the scope of judicial review. It may turn out that there is no sense to the
distinction, and that the terms being distinguished are meaningless; then a successful
analysis will point out the nonsense, and a pragmatic approach will find other grounds
for deciding whether the court should substitute judgment. Or analysis may show that
all questions of application are questions of law, and then a pragmatic approach may
conclude that the criteria the distinction provides are inadequate to the task of
dividing power between tribunal and court, so that other grounds for judicial
intervention are needed.

“Give me a pragmatic approach in any case,” you may be thinking. “It’s more
important that judges do something useful with this distinction, than that they give a
tidy formulation of it.” There is something to this. You do not actually need an
analytical approach in order to make good decisions. Your instincts may be sound.
There is every reason to think that a judge can make good decisions without being
able (or simply without having the time) to come up with a satisfactory analysis.

So why try to work out a decent analytical approach? An analytical approach offers
an articulate understanding of what you are doing. And that is valuable in two ways:

(1) It might contribute to good practical conclusions. This is not because


analytical considerations might weigh against any practical considerations. Judicial
reasoning has nothing to fear from an analytical approach. But an analytical approach
might contribute to practical conclusions because it might point out pragmatic
contradictions-- situations in which some value or policy or principle is promoted in
one area of the law and flouted in another. An analytical approach might also point
out pragmatic tautologies-- so that purported reasons will be exposed as not being
reasons at all.63

(2) Aside from its potential usefulness in reaching decisions, judges need to take
an analytical approach if they are to give reasons. If their only duty were to give good
decisions, an analytical approach would merely be potentially helpful. But it is
essential, if they are to give an account of what they are doing. Note, however, that
the judges’ need to analyse what they are doing leaves open the possibility that the
distinction between questions of law and questions of fact is meaningless, and that the

63 There is a pragmatic tautology in the passage from the Page case cited above at n.12.
17

problem has no solution. Then all a judge could do would be to adopt a pragmatic
approach explicitly.

What will happen if we ask what questions are questions of law? If we conclude that
all questions of application are questions of law, then we will conclude that, whenever
error of law is the standard of review, the law requires the court to substitute its
judgment for that of the tribunal on all questions of application. Then there will be
reason to think that the Page doctrine is misguided, and that the panoply of statutes
giving appellate review on questions of law are all too invasive. If we conclude that
an analytical approach never tells us whether a question of application is a question of
law, then we will conclude that the law in this area is nonsensical.

In fact, I propose that a sound analytical approach will conclude that a question of
application is a question of law when the law requires a particular answer to it. The
law is not particularly absurd, and it does not require universal substitution of
judgment by a court on questions of application.

6. An Analytical Approach to the Problem

How would a sound analytical approach deal with the problem? It might start by
shaking off the notion that a question can only be unequivocally a question of law or
not a question of law. Think of the question, “Is Brutus guilty?” Is that a question of
law? Well, of course it is-- because it asks what Brutus’s legal position is. But of
course it is not a question of law-- it might be answered by considerations of the
credibility of witnesses and so on that are pure questions of fact for the magistrates. If
it were a question of law, magistrates’ decisions could be reversed for any error. The
distinction between fact and law would disappear.64

Well then, is the question, “Is Brutus guilty?” sui generis? Of course it is, because it
is categorically distinct from the paradigm question of law, “what does the law
prohibit?” and the paradigm question of fact, “what did Brutus do?” But of course it
is not an independent genus of question. It is just a question of applying the answer to
one question to the answer to the other.

If a question can be a question of law in one sense, and not a question of law in
another sense, we need to look at the senses in which a question of application might
be a question of law.

6.1 Senses in which questions of application are questions of law or questions


of fact

With the question we just looked at, (“is Brutus guilty?”), an analytical approach to
the distinction between law and fact simply points out relations it bears to the obvious
cases of questions of law and questions of fact. An analytical approach to the
64 For instance, if the Divisional Court thought that the magistrates had reached the wrong decision by
giving too much credence to a witness, then the Divisional Court would disagree with the magistrates
on the question, “Is Brutus guilty?” So, if that question is a question of law, the Divisional Court can
overturn decisions of magistrates even for (operative) errors in assessing witness credibility.
18

problem (“was Brutus’s behaviour insulting?” --is that a question of fact or a question
of law?) would start out by doing the same thing: pointing to ways in which “was
Brutus’s behaviour insulting?” resembles (senses in which it is) a question of fact and
ways in which it resembles (senses in which it is) a question of law.

question of application = question of fact

The question of application is like a question of fact because it is particular. Brutus’s


name crops up in both questions. The paradigm question of law (“what is the
offence?”), by contrast, has the same answer in every case: it is a general question.

Also, the question whether Brutus’s behaviour was insulting has the ring of a question
that would be put to the jury, if Brutus were tried before a jury. And note this: if the
Divisional Court in Brutus’s case had been right, and the magistrates ought to have
asked themselves a different question (something like whether Brutus’s behaviour
showed contempt for other people’s rights), that question would have to be remitted
to the trier of fact. Triers of fact typically answer questions of application, though the
judge will typically tell the trier of fact what question of application has to be
answered. “What question of application is the tribunal to answer?” --That is a
question of law. Whatever answer is given, there always remains some question of
application for the tribunal to answer.65

Finally, we should note that, if Brutus’s behaviour was insulting, that is undoubtedly a
fact. A social fact, no doubt,66 but not a juridical fact. If “insulting” were not used as
a standard in a statute, we could decide whether Brutus’s behaviour was insulting
without any reference to the law.

question of application = question of law

So it starts to appear that this question of application is obviously, clearly a question


of fact. Except that “insulting” was used in a statute --used to define a criminal
offence. The answer to this question of application determines Brutus’s legal
position: a negative answer entails that he is not guilty. So of course it is a legal
question-- a question of law. This is a very simple point that may underlie the
consensus that, analytically, questions of application are questions of law.67 The

65 Bodies that merely identify primary facts are conceivable (for a discussion see Zuckerman, supra
n.6, at p.27). They would have a function like that of a commission of inquiry, and unlike that of any
adjudicative body.
66 I call it a social fact because what is insulting (like what is polite, complimentary, rude...) depends
largely on the conventions of a community (although those conventions often make sense to outsiders,
and perhaps some actions would qualify as insulting in any community). The question whether
behaviour is insulting can also be an evaluative question (because calling behaviour “insulting” can be
a way of condemning it). That point affects nothing in my argument. We could make the point with
any of the evaluative terms that the law uses (e.g. “dishonest”, “malice”...): if they are non-juridical
their application is in that sense not a question of law, though we would not want to call it a question of
fact if we were occupied with the philosopher’s task of distinguishing evaluative statements from
statements of fact.
67 See e.g. Mureinik at p.618: “...the court’s answer to the question attaches legal consequences to
that occurrence or situation. Why, then, should we decline to call that answer the judgment of the
law?...And if the decision is the judgment of the law, it says what the law is. That is all that we mean
by an answer to a question of law.” The court’s answer to the question “Is Brutus guilty?” also
19

question of application is legal in such a clearly important sense that, forced to


choose, it is tempting to plump for calling it a question of law. This simple point is
deceptive, and we need to loosen its hold.

“Is this tie fashionable?” is obviously a question of fashion. But suppose that fashion
demands yellow ties. Then what about the question, “is this tie yellow?” It has
fashion implications: a negative answer entails that the tie is not fashionable. But
that is the only sense in which it is a question of fashion. Fashion does not answer the
question “is this tie yellow?” Similarly, the question “was Brutus’s behaviour
insulting?” is not unequivocally a question of law, merely by virtue of the fact that it
has legal implications.

So while the question of application is a legal question in the obvious sense that its
answer determines Brutus’s legal position, a good analytical approach will not lose
sight of the distinct senses in which it is a question of fact, and will not lose sight of
the fact that it is a question of law in the special sense that it is the question that the
law poses. The paradigm question of law (“what does the law prohibit?”) asks what
standard the law sets; the question of application is the standard.

6.2 What more can an analytical approach do? --Find the relevant sense

What conclusion should an analytical approach reach? That a question of application


is neither a question of fact nor a question of law? or both? No; again, it is only
neither in that there are respects in which it is not a question of law and other respects
in which it is not a question of fact. And it is only both in that there are respects in
which it is a question of law and other respects in which it is a question of fact.
Should an analytical approach plump for one alternative or the other? No; plumping
in this context would be analytically disastrous, because it would commit us to
discounting the facts that do not fit our conclusion.

An analytical approach does not have to plump. Given an expression that can be
applied differently in different senses, an analytical approach should ask what the
relevant sense is. That means asking, “why draw this distinction?” An analytical
approach cannot get off the ground without addressing that question. And look at the
consequence: “why distinguish questions of fact from questions of law?” --that is a
pragmatic question. It is impossible to come up with an analytical approach without
taking account of pragmatic considerations. No analytical solution to the problem can
avoid asking, “what is the relevant sense of ‘question of law’?” This is a pragmatic
inquiry, but note how general it is: it does not ask, “what answer to this question
would be convenient?” but “what would make sense of this question?”

This point is central to the argument, and it is worth elaborating.68

attaches legal consequences to the occurrence, and is the judgment of the law; but it cannot be a
question of law in the sense required of a test for judicial review (see above, n.64).
68 The claim that any analysis of the concepts used to describe human life must rest on the analyst’s
evaluations is familiar from John Finnis, Natural Law and Natural Rights (1980), chapter 1. For a
recent discussion of the role of normative premises in conceptual analysis see Joseph Raz, “Intention in
Interpretation”, in Robert P.George, ed., The Autonomy of Law (1996) at pp.260-262. Raz discusses
the normative premises on which an explanation of the nature of law or of legal interpretation must be
based, and rejects the notion that a theory of interpretation must be value-free. These views enlarge on
20

The flute with no holes

Suppose you bring home a prop from the theatre where you work. You tell your son,
“here’s the flute we used in ‘Twelfth Night’”. He takes it to school and asks if he can
join the band. The band leader says no. “That’s not a flute”, she says. “Flutes have
holes.”

Is the prop a flute? Everyone at the theatre called it a “flute”, even when they were
speaking out of character. Everyone in the band refused to call it a “flute”. Consider
two accounts of whether the prop is a flute, based on two attempts at giving an
analysis of the meaning of the word “flute”:

(1) Value-free analysis: The meaning of a word is a matter of how people use
the word. An analysis of meaning is an a priori inquiry: it asks how people use the
word without inquiring into their purposes, intentions, etc. So to ask if the meaning
of the word “flute” determines whether the prop is a flute, we look at the objects to
which people apply the word “flute”. Then we will have a test for the application of
“flute”: it applies to objects with the same properties as the objects to which people
apply the word. This analytical approach reveals that some people call the prop a
“flute” and some refuse to do so. We can conclude either (i) that the prop is a flute
because some people call it one, or (ii) that there is no analytical answer to the
question, because there are conflicting facts about the use of the word “flute”. On
conclusion (i), the musicians are wrong to deny that the prop is a flute (but there are
good pragmatic reasons not to call the prop a “flute” at band practice: people will
laugh at you). On conclusion (ii), there is no answer to the question whether the prop
is a flute, so you might as well just do what achieves good results. Call the prop a
“flute” at the theatre, so that you will fit in. Don’t call it a “flute” at band practice,
because people will laugh at you.

(2) Analysis based on normative premises: The actors who called the prop a
“flute” were using the word for something that is not a flute. The actors were not
actually concerned whether it was a flute, but only whether it looked like one. They
did not need a real flute; musicians do need real flutes. Something that is not a real
flute is not a flute. So the prop is not a flute. The actors, of course, speak English.
When they called the prop a “flute”, they were using the word in a secondary,
analogical sense that is akin to a metaphorical use (the dawn does not have rosy
fingers, and the prop is not a flute, but we can call the dawn “rosy-fingered” and we
can call the prop a “flute”). In that sense, the prop really is a flute (the actors weren’t
lying when they called it a “flute”).

Approach (2) is preferable. It can explain how musicians and actors use the word
“flute”. Approach (1) makes the meaning of the word into a guessing game; approach
(2) makes sense of it. Approach (1) is an analytical approach as the consensus
conceives it. Approach (2) is an analytical approach that asks normative questions.
We cannot say what the object is, unless we know the point of calling something a
“flute” (i.e., in this case, whether the point is to credit it with the attributes that are

H.L.A.Hart’s notion of the “internal point of view” (The Concept of Law (2d ed. 1994) at pp.89-91,
98), though Finnis and Raz depart from Hart’s account and from each other’s in important respects.
21

important to actors, or the attributes that are important to musicians). Then an


analytical approach to the meaning of the word “flute” will not be able to answer the
question “is this a flute?”, unless the analyst knows which set of purposes is
appropriate. And appropriateness is a normative notion. So an analytical approach
needs to make a normative decision (or just needs to proceed on a normative
presupposition), in answer to the question, “what purposes should I count as
relevant?”

Why would an analytical approach say that the sense in which the musicians use the
word “flute” is primary, and that the actors are borrowing the word and applying it to
something that is not a flute? Why not say that the musicians are reserving the word
for a particular kind of flute-- the kind with holes? Because of a decision (or
presupposition) that the purposes of the musicians provide the norm for the use of the
word, and that the actors’ purposes derive from the musicians’. That is a normative
judgment. It is patently correct, and it explains the way the actors use the word
“flute” as well as the way the musicians use it. The actors were pretending to do what
the musicians do. But that is not a fact that we could read off from information about
which objects people have applied the word to. Your son can only grasp the meaning
of the word “flute” if he grasps this “internal” aspect of its use: that musicians have
reason to use the word to refer to things that are good for making music.

Notice that we need to understand why the word is used, what matters about the
things people call flutes, what is proper to flutes, the point of talking about flutes,
what you should call a flute-- all of which are pragmatic, or prescriptive, or normative
matters. But they are very general normative matters. We do not need to know what
some particular person might want to do with a flute, or what ought to be done with a
flute in a particular situation-- only what people characteristically want to do with
flutes. Suppose your two-year-old daughter hears the word being used, and decides
that “flute” means “tool for bashing things with”. She asks for a flute when she wants
to bash something, and when she sees a hammer she says, “nice flute!” Your
daughter misunderstands the word “flute”. And if you told her so, it would be no
answer for her to say, “No, I know what it means: ‘flute’ means ‘bashing
implement’. I have answered the normative question, ‘what are flutes for?’ (which
you say an analytical approach must answer), by reference to the best purposes I
could think of.” The fact that she wants to bash things with the flute does not make
the word “flute” mean “bashing implement”. That is so even if she has good reason
to bash things with it-- even if in some situation it is right or even heroic to bash
things with it. What she will be bashing things with is a musical instrument. To
understand that a flute is a musical instrument, she will not have to understand any
particular person's purposes, nor what it would be good to do with a flute in any
particular circumstances; she will have to understand the purposes of people in
general who use flutes, and what a flute is good for in general.

Why this palaver about flutes? Just to make two points about analysis: (i) An
analytical answer to the question “is a question of application a question of law?”
cannot get off the ground without asking a normative question. You cannot even
decide whether an object is a flute without making (or taking for granted) normative
judgments. And you cannot decide whether a question is a question of law without
doing so, either. But the normative question that an analysis needs to address is very
general: it is not, “what solution to the problem would give the best result in the case
22

before the court?” but “why ask this question?” (ii) As long as that is the normative
question that we are asking, it makes sense to say that we are still giving an analysis,
clarifying the meaning of the terms in question, asking what questions are questions
of law, and so on. It makes sense because our approach will be distinct from the
approach of Katharine and Lord Denning, the purely pragmatic approach that asks,
“what characterisation of the phenomena will achieve the right outcome in this case?”

6.3 The point of distinguishing between questions of law and questions of fact

Why ask whether a question of application is a question of law? The function of the
error of law standard, remember, is to distribute decision-making power between the
court and the tribunal. The purpose must be to do so in a way that could improve
administrative decision-making (this is the normative judgment that an analytical
approach must make). The use of error of law as a standard of review must be treated
as a way of distributing decision-making power that is calculated to promote the
interest of the community in good administration, and the interest of the applicant in
fair and just administration. The purpose cannot be, e.g., to enhance the prestige or
augment the territory of the judges, because that would not be a good way of fulfilling
the function of distributing decision-making power.

How can we make the doctrine of substituting judgment on questions of law


comprehensible as a technique for accomplishing the purpose of improving decision-
making? It would not be a comprehensible technique for improving decision-making
if we said, “the court must substitute its judgment when the answer to the question has
legal consequences”. That would treat the law as the preserve of judges, rather than
their responsibility.

Similarly, review for error of law would not make sense as a technique for improving
decision-making if we said, “the court should only substitute its judgment on general
questions, and every question of application is particular”. That would not make
sense because an improvement in decision-making in a particular case is just as truly
an improvement (though it may not be as great an improvement) as is a general
improvement.

What would make sense of review for error of law as a technique for improving
decision-making? Review for error of law must be taken as aimed at improving
decision-making by substituting the answer that the law requires (if it requires one)
for the view of the tribunal.69 The notion that the court must substitute its judgment
for that of the tribunal when the law demands a particular answer would make sense
of error of law as a standard of review. And then the rationale for the standard of
review would be to make the court’s view of what the law requires into a limit on the

69 Dividing decision-making power can improve decision-making if review provides consistency,


prevents abuse of power, replaces inept decisions with better decisions.... None of those results is
guaranteed to result from judicial intervention: there may already be consistency in a hierarchical
system of tribunals, a court may abuse its power, a tribunal may be more skilful.... And review for
error of law may not be the best way to achieve these results. But substitution of judgment on
questions of law must be understood as taken to provide such results, or it would not achieve the
purpose of improving decision-making.
23

autonomy of tribunals. This may be a deficient standard.70 But it should be


understood as a standard, as something intelligibly (if wrongly) directed to fulfil a
function.

6.4 The proposed solution

When should a question of application be treated as a question of law? When doing


so will give effect to the court’s view of what the law requires. That is, when the
court decides that the law requires one answer to the question of application. So I
propose this solution to the problem:

a question of application is a question of law in the relevant sense when the


law requires a particular answer.

6.5 Applying the Solution

To apply the solution that I have proposed, we need to know when the law requires
one answer to a question. This is simpler than it might sound:

(1) Clear cases: The law requires one answer to the question of application in
a clear case of the application of the statutory language. So if Brutus had
hurled racist taunts at the spectators, the law would have required the
magistrates to hold that his behaviour was insulting.

(2) Judicial elaboration: The courts have the legal power to elaborate the
law. When it is unclear what the language of the statute requires, the law is
unclear. But the law gives the courts the legal power to make it clear by their
decision. When the courts decide that the law treats certain behaviour as not
insulting, then, the law does treat it as not insulting. Lord Reid held that
behaviour like Brutus’s does not amount to insulting behaviour for the
purposes of the Public Order Act. After Cozens v. Brutus, it did not. The
House of Lords gave the law’s answer to the question.

When a court holds that the law requires a tribunal to answer a question of application
in one way, it is often unclear whether the court is giving effect to the statute in a
clear case of its application, or is elaborating the law. Did the House of Lords make
new law in Brutus? We might say that the House of Lords made new law (to the
effect that behaviour like Brutus’s does not count as insulting for the purposes of the
statute) by imposing a special sense of “insulting”. Or we might say that the issue
was unclear, and that the House of Lords made new law by resolving the unclarity.
Or we might say that the law was already in Brutus’s favour, because in the context of
the Public Order Act, “insulting” clearly does not apply to Brutus’s behaviour.

70 It is a deficient standard if there are convincing reasons for specialist tribunals to be able to develop
the law in their special areas. Whether there are such reasons is beyond the scope of this article; it
often seems to be a fundamental tenet amongst English judges that there can be no such reasons, an
attitude which has doubtless generated the Page doctrine. My argument here includes the claim that,
whether that attitude is misguided or not, it does not require substitution of judgment on all questions
of application.
24

I call the court’s power to resolve the issue a legal power to elaborate the law; we
could (as the courts do) call it statutory interpretation; moreover, because it can be
unclear whether a case is unclear, ordinarily the courts can quite reasonably portray
their decisions as straightforward applications of the statutory language. What we
have to see in any of these cases is that it is up to the court to decide whether the law
requires one answer to the question of application --and its decision makes the law.

So in English law, when do courts have legal power to interfere with tribunals’
decisions on questions of application? In two circumstances which cannot be sharply
distinguished: (1) when the tribunal applies the statute wrongly in a clear case, and
(2) when the courts exercise their legal power to elaborate the law, in a way that runs
contrary to the tribunal’s decision.

6.6 Summary of the Argument:

1. There are senses in which a question of application is a question of


fact: its particularity; “insult” is not a juridical concept. There are
senses in which a question of application is a question of law: the
answer to the question of application has legal consequences; the
question of application is the question the law poses.

2. What is the relevant sense of “question of law”? Step 1 cannot tell us.
We need to ask, “relevant for what purpose?” That means asking,
“Why distinguish questions of fact from questions of law?”

3. The point of distinguishing questions of fact from questions of law is:


to make the court’s view of what the law requires into a limit on the
autonomy of administrative tribunals.

4. So a question of application is a question of law in the relevant sense


when the law requires one answer to the question of application.

5. The law requires one answer to the question of application


(1) in a clear case of the application of the statutory
language, and
(2) when the court exercises its legal power to elaborate
the law so as to require (or interprets the statutory standard to require)
one answer.

Once the proposed approach incorporates pragmatic considerations, why not just call
it a pragmatic solution? The story of the flute with no holes was meant to answer this
objection: An analytical approach is not an approach at all if it stops at step 1 in the
argument. And I propose that it makes sense to say that we are still asking what a
flute is (as opposed to asking what it would be useful to call a flute), even if
normative considerations necessarily enter into answering that question. The
normative considerations essential to an analytical approach are general (“what is the
point of calling a question ‘a question of law’?”), while the normative considerations
25

that a purely pragmatic approach would take into account are particular (“what good
consequences can I achieve here by classifying the question as a question of law or a
question of fact?”).

Here is a simple test for the success of the proposed solution: does it make sense to
say that a question of application is a question of law, in the relevant sense, when the
law requires a particular answer to it (as opposed to saying that it would be
convenient to use the notion of a question of law in a particular way)? If that does
make sense, the solution is descriptive, and analytical, though it is based on a
normative premise.

The pragmatic approach asks “what questions would it be useful to treat as questions
of law?” The analytical approach that I propose does not ask that question at all. It
asks, “what is the point of treating questions as questions of law?” The difference
between those two questions is like the difference between the question “what do you
want to do with this?” and the question “what is this for?” The proposed solution
seeks to state the requirements of the Page doctrine (and of the statutes providing for
appeal on questions of law). It is clearly distinct from a seriously pragmatic solution,
which would not be troubled about that at all.

7. The Solution and the Consensus

It is very easy to reconcile the solution I propose with the consensus --to an extent.
The proposed solution and the consensus are both in favour of the Edwards v.
Bairstow approach. And both reach some similar conclusions as to what courts
should do: they should pay attention to the reasons for review, and they should not
use their power to interfere on questions of law to replace the judgment of tribunals
on all questions of application. Moreover, the proposed solution is able to account for
the attraction of a pragmatic approach. Recall the second circumstance in which the
proposed solution claims that courts have power to interfere with a question of
application: when the courts exercise their legal power to elaborate the law. Like all
law-making decisions, the decision whether (and how) to do that is a pragmatic
decision. So the proposed solution shares pragmatic features with the approach that
the consensus recommends.

There remain three significant distinctions between the proposed solution and the
consensus. First, they differ over the nature of analysis. The proposed solution
describes the Edwards v. Bairstow doctrine as the result of a sound analytical
approach, not as the result of abandoning analysis. It claims that the “analytical
approach” rejected by the consensus is simply an unsound analytical approach, which
either plumps for calling questions of application questions of law, or says that there
is no analytical solution to the problem.

Secondly, the proposed solution and the consensus actually diverge in effect, in so far
as the writers are proposing the pragmatic approach as a way to avoid, or to modify,
the Page doctrine. On my view, that doctrine is not actually as intrusive as it has been
made out to be, because not all questions of application are questions of law. Yet
there are still convincing reasons for abandoning the Page doctrine: Albert Venn
26

Dicey notwithstanding, courts should not necessarily have the last word on questions
of law. Should the courts follow the consensus, then, and lean toward treating
questions of application as questions of fact simply in order to substitute a sensible
standard of review for the standard of review that the House of Lords has developed?
On the view that I have proposed, to do so would be to change the law without saying
so. Lower courts cannot do that, and the House of Lords has committed itself not to
do so. There may, however, be good reason for the House of Lords to change the
doctrine openly and deliberately.

Thirdly, the solution rejects the view that there is no analytical solution to the
problem. If that were so-- if the scope of review were based on an incoherent
distinction-- then there would be no law to disregard. All bets would be off.
Necessity would justify a pragmatic approach, and the judges would do well to craft a
sensible standard of review out of chaos.

The comity of the judges forbids them to say that seven House of Lords decisions and
dozens of Acts of Parliament are talking nonsense. But a careful theorist cannot rule
it out in the same way: life gives no guarantee against widespread judicial and
legislative nonsense. A careful theorist will, no doubt, form a view that the law tends
not to be nonsensical even when it is bad. But a careful theorist will not apply this
view as a presumption. Courts use presumptions out of necessity; theorists should,
instead, try to understand what is going on. The proposed solution does not claim that
widespread legal nonsense is inconceivable-- just that nonsense is not as widespread
as people think in this area of the law.

8. The Proposed Solution and the Cases

On the consensus view, there is something wrong with the cases. The judges seldom
discuss the pragmatic concerns the consensus urges on them. They do not hold
consistently to an analytical approach (as the consensus conceives it) either, and the
best we can say is that they must be pursuing some sort of inarticulate pragmatic
approach, which leads to all sorts of inconsistency because its motivating principles
are silent and undeveloped.

The solution that I have proposed makes more sense of the cases. First, it is probably
the only way to make sense of the common thread that we saw in the cases: the
notion that a question of the application of the same statutory provision can be a
question of law in one case and a question of fact in another case. Secondly, the
proposed solution offers to explain the uncertainty and variability of the cases. The
distinction between fact and law is unclear simply because the distinction between
clear cases and unclear cases of the application of statutory language is unclear. The
distinction is variable because whether the courts exercise their power to elaborate the
law will depend on the individual demands of particular situations (on the demands,
in fact, that would motivate a good legislator).

Finally, the proposed solution recommends that judges apply the reasoning that has,
in fact, been predominant in the House of Lords. They should hold that a question of
application is a question of fact when it is capable of decision either way, and that it is
27

a question of law when the law requires a decision one way. (Whether the House of
Lords should reform the Page doctrine, and whether Parliament should find other
ways to divide decision-making power between courts and tribunals, are separate
questions.)

So the proposed solution is consistent with the most influential and sophisticated of
the decisions-- the speech of Lord Radcliffe in Edwards v. Bairstow, which is widely
regarded as the prime example of the pragmatic approach.71 What the law does is to
lay down limits within which it would be permissible to say that the standard applies
or that it does not apply. There are “many combinations of circumstances,” Lord
Radcliffe said, “in which it could not be said to be wrong to arrive at a conclusion one
way or the other.”72

Those are the cases in which the law does not require one answer to a question of
application. In those cases, there is nothing in a sound analytical approach that
requires us to treat a question of application as a question of law.

TIMOTHY ENDICOTT.*

71 Cf. Beatson, supra n.60, at p.40; Craig at p.158.


72 [1956] A.C. 14 at p.34.

*St.Catherine’s College, Oxford. I benefitted from discussions with Joseph Raz and Karen Yeung, and
from comments by the Editor.

You might also like