Ask most practitioners whether “burden of proof” differs from “onus of proof” and you will be told they are twins. Onus is the Latin, burden the English, and there the matter ends. Our law reports mostly agree with them. Open almost any reported judgment and you will find the compound expression “burden or onus of proof,” the disjunctive doing no work at all.[1]
It looks like a quibble about vocabulary. I think it has a cost, and that the cost falls almost entirely on people who will never read a law report.
Two burdens, not one
Two questions arise in every contested trial, and they are not the same question.
The first is what happens if the court remains undecided once all the evidence is in. Somebody has to lose, and the rule identifying who that is gives us the legal or persuasive burden. It is fixed at the outset by the substantive law or by the pleadings, and it does not move. Think of it as a tie-breaking rule held in reserve for the end of the trial rather than as something a party discharges along the way.
The second question is more immediate. On the evidence as it stands right now, who must produce something or lose the issue? That is the evidential burden, and it shifts back and forth continuously as the trial proceeds. Strictly it is not a burden of proof at all, since it can be discharged by material falling well short of proving anything.
The distinction is a tool, not a taxonomy
The common law did not keep these two apart because scholars enjoy fine distinctions. It kept them apart because the distinction turned out to be working constitutional machinery.
When an English court meets a statute that seems to require an accused to prove his own innocence, it has a third option beyond enforcing the provision or striking it down. It can read the burden down, from persuasive to evidential. That is what happened in R v Lambert, where the reverse burden under the Misuse of Drugs Act was softened in precisely this way.[2] In Sheldrake v DPP Lord Bingham did both things in a single judgment, upholding one reverse burden as genuinely persuasive while reading a Terrorism Act provision down to an evidential one.[3] Two provisions, two results, one instrument. Both statutes survived. What changed was the character of what the accused had to do.
I should not overstate this. Vocabulary on its own produces nothing. England had the distinction from Woolmington in 1935 and did not read a reverse burden down until 2001, and what intervened in the meantime was not a word but the Human Rights Act. Australia makes the same point from the other direction: its Criminal Code separates legal from evidential burdens in express statutory terms, yet, lacking any national bill of rights, it has produced correspondingly little constitutional review of reverse onuses.[4] Structural review seems to want two things at once. A vocabulary in which the argument can be framed, and a rights instrument the courts are willing to treat as operative. We acquired the second in 2010. The first is still missing.
The same inheritance, three destinations
Our own Order reproduced the Evidence Act 1872 almost exactly: Article 117 for section 101, Article 121 for section 105.[5] Stephen, drafting in 1872, used “burden of proof” to render onus probandi and never troubled to separate its two senses. Section 101 speaks of a party who “must prove.” Section 102 speaks of the party who “would fail if no evidence at all were given.” We can see now that these are the persuasive and the evidential burdens respectively. The statute itself nowhere says so.
Ceylon inherited the identical provision, and it was on appeal from Ceylon that Lord Devlin drew the distinction precisely, in Jayasena v The Queen.[6] That case is awkward for my argument in two ways, and I would rather concede them than have them put to me. It shows that the distinction was always available on our text; the silence was never a prohibition. It also holds that the burden the provision places on the accused is a persuasive one, which largely closes off reading-down as a matter of construction. Whatever argument we make will have to be constitutional rather than interpretive.
India shows what that route looks like when built on the same inheritance. Its courts began with the taxonomy: Raghavamma is canonical, holding that burden of proof never shifts while onus of proof does.[7] Having built it, they then used it. In Noor Aga v State of Punjab the Supreme Court of India heard a challenge to the reverse-burden provisions of the NDPS Act and discussed Lambert expressly, treating the English doctrine of compatibility as analogous to its own doctrine of constitutionality, before holding that the impugned sections had to be read in the light of Articles 14 and 21 of the Constitution.[8] The provisions survived. They survived calibrated: the presumption bites only where the prosecution has first proved the foundational facts beyond reasonable doubt.
That is structural review. Same statute, same silence, and a court willing to reach for a constitutional standard rather than dispose of the matter case by case. I have not found a Pakistani judgment that does comparable work. I would be glad to be shown one.
The anaesthetic
What we built instead is a benefit-of-doubt jurisprudence of striking generosity. Tariq Pervez holds that a single circumstance creating reasonable doubt entitles the accused to acquittal, “not as a matter of grace and concession but as a matter of right.”[9] Nothing that follows is a criticism of that rule. My complaint is that it has been doing a job it was never built for.
Consider what the rule is, structurally. It operates at the very end of the process, on one completed record, and its remedy is the acquittal of one appellant. About the allocation of the burden it says nothing whatever. It does not ask whether the legislature was entitled to require this accused to prove anything, or to what standard, or whether pointing to some evidence might have been enough. It asks a single question: on this record, is there a doubt?
Each acquittal is therefore a transaction with one customer, and the provision that produced the conviction is never itself put in the dock. Whether the easy availability of that remedy has actively discouraged structural challenges, or whether such challenges have simply never occurred to anybody, I cannot say without a survey of the reported record. That survey is worth doing. But the pattern seems clear enough: the symptom is relieved so reliably that nobody gets round to diagnosing the condition.
What has gone un-audited
Section 29 of the Control of Narcotic Substances Act 1997 presumes the offence committed, “until the contrary is proved,” once possession is established.[10] Our courts have narrowed the trigger sensibly enough, insisting on proof of conscious possession. But narrowing a trigger is not the same thing as calibrating a burden. So far as I can find, nobody has put the Noor Aga question to section 29: once the presumption operates, is what it demands of the accused persuasive or evidential, and can either survive Article 10A? A further question has not been asked at all. A provision realistically dischargeable only by the accused entering the witness box sits uneasily with Article 13(b), and I can find nobody who has said so.
Section 14 of the National Accountability Ordinance 1999 was upheld in Khan Asfandyar Wali. The prosecution, the Court reasoned, must first establish preliminary facts “whereafter the onus shifts,” and a burden on the accused “is not an alien concept in jurisprudence.”[11] Credit where it is due: the Court did reason in two stages. What it never asked was what kind of burden section 14 imposes once it shifts, or to what standard, or whether something lighter would have served the purpose just as well. The structure was put to work validating the provision rather than calibrating it.
We also have no idea of the scale of the thing. The Indian Supreme Court noticed in Noor Aga that nobody had counted how many Indian offences place the legal burden on the accused, and set that against a United Kingdom estimate putting the figure at around forty per cent of offences triable in the Crown Court.[12] No such count exists here either. We are not merely failing to audit these provisions. We do not know how many of them there are.
Why this is not an academic complaint
A retail remedy is distributed by retail logic. The benefit of doubt is granted on appeal, on a developed record, by a superior court, to an appellant who managed to get there. Which is to say: to the accused who could afford trial counsel competent enough to build a record worth appealing on, and appellate counsel to bring the appeal, and who survived the years of custody it takes to be heard.
It does not describe the ordinary section 29 convict. He is poor. He was defended perfunctorily, and for him the presumption did its work at trial and was never revisited. I put this as a proposition to be tested rather than as a finding. We have no published figures on representation or appeal rates in narcotics prosecutions, and that absence is itself part of what I am complaining about.
Structural review runs the other way. Calibrate a burden once and the ruling operates on every future trial under that provision, including the trials of people who will never file an appeal at all. It reaches the class, not the customer. And if a remedy is in practice available only to those who can pay to reach it, the objection is not confined to the presumption of innocence under Article 10A. It is an objection about equality before the law under Article 25, and I have not seen it put that way.
The road not taken
The capacity is plainly there. In Raja Khurram Ali Khan v Tayyaba Bibi the Supreme Court used the vocabulary without hesitation: the prosecution’s obligation under Article 117 was the legal burden, unshiftable absent express legislative command, while the accused bore only the evidential burden, which meant producing evidence enough to create doubt rather than proving his innocence.[13] In Ali Ahmad v The State, Syed Mansoor Ali Shah J held that the burden shifts to the accused under Article 121 only once a prima facie case sufficient to justify conviction has been made out.[14]
Both are right, and both are welcome. But both apply the distinction inside an existing statutory scheme. Neither turns it outward to ask whether the scheme itself can stand. We have the tool in hand and we are using it to describe rather than to review.
Conclusion
England keeps the two burdens apart because that is what lets a court save a statute by softening it. India kept them apart and then, in Noor Aga, used the distinction to make a reverse-onus regime answer to the Constitution. We inherited the same ambiguous text as India, and in Jayasena the Privy Council showed us exactly how the distinction runs. We have done neither thing with it.
What we have instead is a rule of appellate grace so generous that the omission has never been felt. Every year our courts acquit men who arguably should never have borne the burden that convicted them. Every year the provision that imposed it survives the acquittal in perfect health, ready for the next accused, who will probably not be able to appeal.
That is not a mature law of proof. It is a very humane way of never having to build one.
Notes
[1] The compound usage is pervasive in reported Pakistani judgments and in commercial digests of the Qanun-e-Shahadat Order 1984.
[2] R v Lambert [2001] UKHL 37, [2002] 2 AC 545.
[3] Sheldrake v Director of Public Prosecutions; Attorney General’s Reference (No 4 of 2002) [2004] UKHL 43, [2005] 1 AC 264.
[4] Criminal Code Act 1995 (Cth), Part 2.6, ss 13.1–13.6. Contrast R v Oakes [1986] 1 SCR 103, where the Canadian Charter supplied the missing instrument.
[5] Qanun-e-Shahadat Order 1984 (President’s Order No 10 of 1984), arts 117–129, corresponding to the Evidence Act 1872, ss 101–114.
[6] Jayasena v The Queen [1970] AC 618 (PC) 624.
[7] Addagada Raghavamma v Chenchamma AIR 1964 SC 136. See also Rangammal v Kuppuswami (2011) 12 SCC 220.
[8] Noor Aga v State of Punjab (2008) 16 SCC 417.
[9] Tariq Pervez v The State 1995 SCMR 1345; reaffirmed in Muhammad Masha v The State 2018 SCMR 772.
[10] Control of Narcotic Substances Act 1997, s 29. On sentencing by quantity see Ameer Zeb v The State PLD 2012 SC 380.
[11] Khan Asfandyar Wali v Federation of Pakistan PLD 2001 SC 607.
[12] Noor Aga (n 8). The United Kingdom estimate derives from A Ashworth and M Blake, ‘The Presumption of Innocence in English Criminal Law’ [1996] Crim LR 306.
[13] Raja Khurram Ali Khan v Tayyaba Bibi PLD 2020 SC 146.
[14] Ali Ahmad v The State, Criminal Appeal No 154-L of 2013 (Supreme Court of Pakistan, 14 March 2019), per Syed Mansoor Ali Shah J.
