A column about curious stories from the worlds of science, technology, and film. Today — a case in which a court had to seriously decide whether a jar of yeast paste could interfere with detection dogs. There is still no evidence that it can: only the word “potentially,” 20 years of practice, and one spectacular defeat for a dog at a festival.
On 26 October, an AP news item flashed through my feed: an Australian court had denied a life prisoner the right to eat Vegemite — the famous black yeast paste Australians have spread on toast since childhood. The grounds for the suit sounded anecdotal; the nature of the ban did not: for twenty years, a government prohibition has rested on a claim about dogs that nobody has ever tested. This is the spine of the investigation: a chronology of state-level bans, the chemistry of yeast, and a classic experiment on false alerts by detection dogs.
On Monday, 26 October 2026, the Supreme Court of Victoria denied 55-year-old prisoner Andrew McKechnie the right to eat Vegemite in prison. McKechnie is serving life for murder, and the dispute was over whether denying him a sandwich spread violated his right under the human rights charter to “enjoy his culture as an Australian in community with other persons of Australian origin.”
Sounds like a joke. But the principle is always what is at stake in such cases: how far can a government ban justified by the word “potentially,” rather than by measurement, extend?
Vegemite has been banned in Victorian prisons since 2006 — that is how it appears in Corrections Victoria’s list of contraband items. The official wording is that the paste “may interfere with drug detection dogs,” while its yeast-based nature gives it the “potential” to be used for brewing alcohol.
The Age reported back in 2007 that the first ban predated the official lists: authorities began confiscating Vegemite in the late 90s, after “prisoners discovered ways to process it.” And during the 2026 proceedings, prison manager Kevin White confirmed that the ban had been in force across all prisons in the state for 20 years.
The picture in the other states is even more interesting. In Queensland, South Australia, and Tasmania, Vegemite is also banned in prisons. But neighbouring New South Wales — no. In NSW, you can eat it. For twenty straight years. Without any noticeable collapse in drug control that could plausibly be attributed to the presence of the paste.
McKechnie was convicted in Queensland more than 30 years ago of the murder of developer Otto Kuhne, then transferred 10 years later to Victoria’s maximum-security Barwon Prison. In court, he represented himself: one of those rare cases where the defendant is at once the accused, the sole lawyer, and the chief expert on his own diet.
He built his case on the Charter of Human Rights and Responsibilities Act, Victoria’s human rights charter. At first he pursued a thiamine-deficiency argument: as a kosher-observant Jew, he did not eat pork, another source of B1, and Vegemite had been his daily supplement. He abandoned that line after objections from the prison administration’s lawyers, who called it irrelevant. What remained was the cultural right.
Justice Claire Harris delivered a 47-page ruling, and its key sentence contains both sides of the coin at once. She acknowledged that the ban limited McKechnie’s cultural rights under the charter. And immediately added: the limitation was “reasonable and demonstrably justified” by the prison administration’s statutory duties — the management, security, and good order of the institution. She put the specific source of McKechnie’s distress more bluntly: “it is difficult to accept that being unable to eat Vegemite, rather than the fact of his imprisonment, is a material cause of his separation from family and community.”
On the substance of the “dogs” argument, the judge wrote that McKechnie had presented no reliable basis for his view that the paste does NOT interfere with dogs. In other words, the burden of proof remained with the prisoner. The null hypothesis — “Vegemite interferes with dogs” — had survived on contraband lists for twenty years, and not once required the state to produce a single measurement in its support. Because on a contraband list, it is not a hypothesis but a starting assumption.
Now, the brewing of alcohol. In his prison testimony, Kevin White referred to “publicly available research” showing that Vegemite helps homebrew ferment faster. He is almost being honest: such research does exist. It simply says the opposite of what the average reader hears.
In 2016, University of Queensland biochemist Ben Schulz dissected the myth of “cheap Vegemite grog” — at the time, it was being attributed to Indigenous communities. The result: you cannot produce alcohol from the paste alone at all. Vegemite is a cooked yeast autolysate, a by-product of brewing, in which no live yeast remains — it has been killed by heat treatment. For anything to ferment, you need two ingredients in addition to the paste: sugar and separate live yeast. Only then do you get what Schulz described with rare poetry: a flat liquid with no foam head, which “looks like orange dishwater,” has a persistent “vegemitey” aftertaste, and costs 9 cents per bottle.
So Vegemite’s real role in makeshift moonshining is as an accelerator: sugar and live yeast will ferment on their own; the paste merely speeds the process along as a nutrient medium. But that is not all. Here is testimony from The Age in 2007: prisoners at the Metropolitan Remand Centre were found heavily intoxicated — and what they had fermented was FRUIT, allegedly stolen from the prison kitchen. Not paste. And Charandev Singh, a representative of a legal rights centre, said in the same article what might be called the occupational deformation of prohibitionist logic: “In prison, you can turn anything into alcohol. Fruit, sugar, bread. The issue is not banning Vegemite. The issue is basic security.”
It makes for a funny inversion of the lines of argument. In court, McKechnie himself said Vegemite contained “no live yeast,” and that any fruit was more useful for making hooch. Chemically, he was right — here, a prisoner defending himself without a lawyer was operating with a more accurate model than the prison administrator with the contraband list.
The most interesting part of the spectrum is the scientific one. How reliable is the “detection dog plus handler” pairing in general? There is a paper that has become a classic: Lit, Schweitzer, Oberbauer, “Handler beliefs affect scent detection dog outcomes,” Animal Cognition, 2011.
Lisa Lit and her colleagues at UC Davis ran a field Clever Hans experiment. 18 “handler + dog” teams — narcotics and/or explosives, all certified and with real finds to their names. Each team went through eight search scenarios in which there were, openly and in advance, NOT A SINGLE gram of drugs or explosives: every alert would count as an error. Some scenarios included a paper location marker — about which handlers had been falsely told that it marked the target point — human influence. Others included bait in the form of food or a toy to stir up the dog — canine influence. The control condition had neither.
The result: 225 false alerts. What is more, the human marker threw the teams off more than the canine bait: in marker scenarios, handlers reported that dogs had “alerted” at the paper more often than anywhere else. The issue was human expectation: it turned any interested sniff into a “confirmed alert.” The conclusion reads like a verdict on everyone who cites dogs as infallible instruments: handler beliefs affect search outcomes. The classic Clever Hans trap, only at the level of the human-nose pairing.
In the same vein — Szetei et al., 2003: around half the dogs in the experiment went to an EMPTY bowl indicated by a human, ignoring the bowl where they had seen with their own eyes — and smelled with their own noses — food. A human pointing gesture can, in some configurations, overpower scent. A dog is not a biosensor with quartz-like stability. It is part of a sociotechnical system where the main source of noise is the expectations of the human at the other end of the leash.
What does this mean for the Vegemite case? The paste’s ability to “confuse” a dog is an open empirical question, one nobody has bothered to measure in 20 years. But failure of the “dog-handler” pairing out of nowhere — with no extraneous smells at all — produced 225 alerts across 18 teams in a single experiment. When the issue is the risk of contraband slipping through, science says this: the most common false alarm in this system is born not in the dog’s nose, but in the human head.
There is also a field case that is easy to find and hard to forget. September 2017, the Listen Out festival in Melbourne. A young man taped packets of MDMA to his penis and heavily smeared the whole construction with Vegemite — the logic being that the paste’s powerful smell would overwhelm the smell of amphetamines for the dogs. The dog found everything immediately. Headlines in the Australian press. An accidental field counterexample to masking: against a trained nose, a household aromatic bomb marks the forbidden item far louder than it conceals it.
Notice the asymmetry: no court required the administration to prove that Vegemite conceals the smell of drugs FROM A DOG — the accidental field experiment of 2017 showed the opposite for them. Yet the proposition “Vegemite interferes with dogs” was already functioning as a legal fact on contraband lists long before any festivals, and in 20 years not a single government laboratory saw fit to test it.
What I like about this story is the collision of two truths that the state resolved in its own favour without a single measurement. Prison administrations have the right to set rules for daily life and the duty to catch drugs within their institutions. The judge is right that the bargain of “cultural right versus security” in prison may lawfully be resolved against the paste. But after that bargain, the substantive question remains — one the state has had no answer to for twenty years: why does a rule for every prison in the state rest on the word “potentially,” one manager’s testimony, and zero measurements — when the neighbouring state lives perfectly calmly without the ban, when the paste contains no live yeast and cannot itself be brewed into alcohol, and when the classic experiment on false dog alerts has been freely available in Animal Cognition since 2011, for 15 years now.
This is a story about how a precaution entered into a register once (“perhaps it may interfere”) turns into fact simply because nobody is required to spend money testing their own fears. McKechnie lost lawfully. But nowhere in the ruling’s 47 pages did there appear the one paragraph that would have cost the state a single day of laboratory work: a dog sniffing paste beside a target, and the result — any result. Instead, Australian enforcement practice was left with a legend on the level of a bug report with 20 years of uptime — and an inadvertent refutation from the 2017 festival, where a trained dog exposed the Vegemite disguise in seconds.