If you have ever worked at a laboratory bench with a Bunsen burner lit beside you, you were probably told why. The flame warms the air above it, the warm air rises, and the rising column is supposed to carry dust, skin flakes and whatever bacteria they are riding on up and away from your open plates. Work close to the burner, the instruction goes, and you are working inside a small protected zone.

A six-author team at Providence College in Rhode Island has now measured that zone. Across three sets of paired trials, Petri dishes sitting 10 centimetres from a lit burner collected no fewer settling bacteria than dishes sitting 10 centimetres from an unlit one. In none of the three conditions was the difference statistically distinguishable from nothing at all.

The study appears in Microbiology Spectrum, published online on 17 August 2026 and open access under a CC BY licence. Kate DeVincent, Ariana Carretta, Vanessa Colina, Imani Laporte, Tyler Oung and Hannah E. Gavin set out expecting the opposite result. Their stated hypothesis was that the flame would help.

The zone students are taught to trust

The convection story is specific and testable. An open flame heats surrounding air; heated air becomes less dense and rises; a rising column has to draw replacement air in from the sides. In the version taught to students, that circulation lifts airborne particles away from the bench surface, leaving a pocket of cleaner air around the burner in which plates can be opened and cultures transferred.

The authors found this claim in textbooks, in peer-reviewed publications and on university-affiliated websites. One protocol they quote instructs researchers to perform all work “close to the Bunsen burner flame where air currents are drawn upward”. What they could not find was a number. None of the resources describing a burner-proximal aseptic zone cited experimental data, and the dimensions those resources give for the zone disagree with each other, its reported radius ranging more than twofold even among sources that endorse it.

That gap is not newly noticed. In a 1996 commentary in the journal Resonance, titled “On Bunsen burners, bacteria and the Bible”, Watve wrote that “there is no authentic report that quantitatively studies the microbial load of the so-called aseptic zone”. The Providence College team says that to their knowledge this remained true in 2026, and the hedge is worth keeping: they are reporting the absence of a study they could find, not proving that none exists. Their own paper describes Watve’s complaint as coming “two decades ago”, which understates it by a decade.

The paired-plate experiment

The design is deliberately plain. Two mirrored workstations sat on one lab bench with the burners about 80 centimetres apart. One burner was lit and one was not. At each station a Petri plate of tryptic soy agar was placed with its centre 10 centimetres from the centre of the burner base, opened for exactly one minute, then covered. Plates were incubated at 30 degrees Celsius for two to three days and the colonies counted.

The bacterium was Gordonia rubripertincta, strain NRRL B-16540, revived from frozen glycerol stocks on tryptic soy agar and then grown up in tryptic soy broth. Counting was restricted to its colonies: a pair of plates was thrown out if a stray non-Gordonia contaminant made the count unreliable. To load the air, the researchers soaked sterilised hardware-store shop towels in dense liquid culture, wrung them out, dried them overnight and then shook them vigorously over the bench for a minute while circling it. For the wet condition they diluted culture and sprayed it from ordinary spray bottles. Both researchers wore lab coats, gloves, goggles, face masks and shower caps, and sat still while plates were open.

Two flame sizes were used: a standard 18 millimetre classroom burner and a wider 39 millimetre one, each calibrated to an inner blue cone of roughly four inches. Pairs were also discarded if either plate came in under 15 colonies or over 400. That left 43 pairs for the 18 millimetre burner with dry particles, 11 pairs for the 18 millimetre burner with wet particles, and 49 pairs for the 39 millimetre burner with dry particles.

The plates beside the lit burner

On the headline comparison, nothing happened. Wilcoxon matched-pairs signed-rank tests returned P values of 0.52 for the 18 millimetre dry condition, 0.53 for the 18 millimetre wet condition and 0.69 for the 39 millimetre dry condition. Median counts were 89 colonies on control plates against 73 on flame-side plates in the first condition, 43 against 49 in the second, and 58 against 54 in the third.

The authors then asked whether the flame’s effect depends on how dirty the air is. Plotting log-transformed flame-side counts against log-transformed control counts, they got a slope of 0.65 for the 18 millimetre dry condition, with a 95 per cent confidence interval of 0.44 to 0.86. That excludes a slope of 1, which indicates a contamination-dependent effect. The wet condition gave a slope that cannot be told apart from 1. The 39 millimetre burner produced almost no relationship, with contamination level explaining 6 per cent of the variation in flame-side counts.

From the 18 millimetre dry regression they calculated the crossover point at which flame-side and control deposition come out equal: about 83 colony-forming units per minute. Above that settling rate, the flame was associated with fewer colonies landing. Below it, the flame was associated with more. The arithmetic yields two further crossover points, 148 for the wet condition and 52 for the 39 millimetre burner, but neither means much, because in the first case there is no real crossover for the number to mark and in the second the regression barely explains anything. Eighty-three is one condition’s figure, not a universal threshold.

It is also a much softer figure than a single number looks. Its own 95 per cent confidence interval runs from 16 colonies a minute to 62,300, spanning more than three and a half orders of magnitude, so the crossover is located far less precisely than the headline suggests. And the direction of the effect below it rests on a regression slope, not on a direct comparison of the two sets of plates. That direct comparison found nothing in any condition.

Several other limits are the authors’ own. They tested one working distance, 10 centimetres, and their argument from it is geometric: any continuous sphere, cone or cylinder of larger radius would have to contain the tested area, so the area within a 10 centimetre radius can be excluded from the claimed sterile zone. What happens further out was not tested. They never observed or measured convection currents; the turbulence they propose to explain the 39 millimetre burner’s noisier behaviour is a suggestion. They did not control particle size, and settle plates preferentially capture small and large particles rather than mid-sized ones. The contamination levels at the top of their range exceeded their own uninoculated laboratory air by several orders of magnitude and could be sustained only for minutes after deliberate, aggressive inoculation.

They also tested whether the flame needs time to establish its currents. Comparing plates exposed immediately after lighting against plates exposed after two minutes of burning, a Kruskal-Wallis test returned P = 0.997. Duration made no measurable difference under any combination of flame size and particle type.

One number in the discussion does not match the paper’s own reference list. The text attributes a finding about airborne microbes in occupied classrooms to “a 2015 study by Qian et al.”; the reference it points to is Qian and colleagues in Indoor Air, dated 2012.

A precedent from 1972

The most striking thing in the paper is a citation. In 1972, in Applied Microbiology, Brunker and Fernandez tested a different piece of the same ritual: flaming the mouth of a reusable test tube during culture transfers. They concluded that flaming a tube “does not reduce contamination significantly and, hence, that this heretofore unquestioned ritual may represent only a waste of time”.

Textbooks and laboratory manuals went on recommending it into the 2010s. The 2020 edition of Current Protocols in Microbiology gets closest to acknowledging the problem and then declines to finish the thought, conceding that “the value of ‘flaming’ has been questioned” and that “the practice of flaming is not necessary”, before instructing that if users choose to include flaming, a brief pass of opened tubes through the flame should be added after opening and before closing. Watve, in 1996, recorded colleagues answering the evidence with a shrug: “working between the burners may be ineffective, but what can we do if we can’t afford laminar flow systems?”

That objection is the real one, and the Providence College team takes it seriously. Biosafety cabinets remain what the paper calls the leading evidence-based solution for reducing airborne contamination, and it accepts they are not financially or physically viable in many situations. So the paper points at cheaper middle ground: freestanding HEPA filters, which it puts at 10 to 100 times less than a cabinet; circulation-free “dead air” PCR cabinets at several thousand dollars; and still air boxes assembled from hardware-store materials, popular among amateur mushroom growers and, the authors note, untested in laboratory asepsis.

Their concrete recommendation is cheaper still: run the settle plate assay in situ, which needs agar plates, a timer and an incubator. If a room routinely shows contaminant settling above roughly 50 to 100 colonies a minute, they say, a burner may be justified. If it does not, the gas line, the consumable gas, the fire-code overhead and the open flame standing next to trainees and flammable solvents are all being paid for by habit.

The comfort of a visible flame

What makes this study unusual is not that it overturned a belief. The belief had already been dented, in print, twice, and carried on regardless. A 1972 result was absorbed and then ignored. A 1996 classroom paper named the missing prior evidence, reported its own thirty-pair null, and changed nothing in the teaching. A 2020 protocol conceded the point and then reinstated the practice in the next sentence.

The authors offer a human explanation, and it is the most quotable line in the paper: the incongruity of those recommendations, they write, “feels like an acknowledgment that what researchers know intellectually might conflict with what they feel about losing a facet of agency over experimental outcomes.” A lit flame is visible, controllable and reassuring. Its absence is none of those things.

Thirty years after a classroom paper reported a thirty-pair null and noted that no authentic prior report had quantitatively studied the zone, six people at a college in Rhode Island counted the colonies again. The flame helped nowhere the direct comparison could detect. And on one regression, with a standard classroom burner and dry airborne particles, it was working against the people beside it in the low-contamination conditions where the authors expect most American teaching and research laboratories to sit.