At four separate flavour trials conducted between 2021 and 2023, panels evaluating e-liquid samples consistently sorted, ranked, and described products more accurately when working from taste descriptors than when working from nicotine concentration labels. The finding is modest on its face — of course a taster knows what "vanilla custard" means better than what "6 mg/mL" means — but it opens a real question about how consumers, retailers, and regulators actually parse the information printed on a bottle. If the chemical label is the least legible part of the package, what is it doing there, and what should replace the cognitive work it was supposed to do?
What the Trials Actually Measured
The four trials varied in design, but shared a common structure. Participants were handed unlabeled or partially labeled samples and asked to complete two tasks: a sorting task (group these by similarity) and a preference-ranking task (order these by how much you'd want to buy them again). Half the participants received a card listing nicotine strength and PG/VG ratio. The other half received a card listing taste descriptors — "bakery," "menthol-forward," "tart fruit," "tobacco-adjacent," and so on.
Across all four trials, the descriptor group produced tighter clustering. Participants in that condition agreed with each other more often about which samples belonged together, and their rankings correlated more strongly with subsequent re-purchase intent measured two weeks later. The nicotine-label group did not sort randomly — they weren't confused, exactly — but their groupings tracked concentration bands rather than flavour families. They were organizing the shelf by pharmacology when the task was to organize it by taste.
That mismatch is the interesting part. It suggests that the nicotine label functions as a kind of attentional magnet: it's the most prominent number on the package, so it becomes the dimension along which people think. The descriptor, meanwhile, is often set in smaller type or relegated to a flavour name that varies wildly across brands.
Variable-Ratio Reinforcement and the Flavour Label
Behavioural psychology offers a useful frame here, though it needs to be applied carefully. Variable-ratio reinforcement — the schedule in which a reward arrives after an unpredictable number of responses — is one of the most robustly documented phenomena in operant conditioning, dating back to B.F. Skinner's work in the 1950s. It produces high, steady response rates and remarkable resistance to extinction. The classic laboratory demonstration involves a pigeon pecking a key that delivers food on an unpredictable schedule; the pigeon keeps pecking long after the food stops.
Flavour products sit adjacent to this literature in a way that deserves honest scrutiny. A consumer trying a new e-liquid is, in a loose sense, sampling from a distribution: some bottles deliver exactly the taste promised, some deliver something close, and some deliver a chemical approximation that misses entirely. The variability is real, and it is partly a function of how poorly taste is communicated on the label. When the descriptor is vague or the flavour name is a marketing invention rather than a sensory description, the consumer's experience becomes genuinely unpredictable — not because the product is designed to be a game of chance, but because the information environment is noisy.
This is where the trial results become practically relevant. If taste descriptors improve sorting and prediction, they reduce the variance in the consumer's experience. A label that says "tart fruit, slightly cooling" sets a more accurate expectation than a label that says "Blue Razz Ice." The former is a sensory claim; the latter is a brand name that may or may not correspond to anything the tongue recognizes.
Loss Aversion and the Cost of a Bad Bottle
Kahneman and Tversky's work on loss aversion — the finding that losses loom larger than equivalent gains — has an obvious application to product trials. A consumer who buys a bottle that tastes nothing like its label doesn't just fail to gain a pleasant experience; they register a loss. They paid money, they waited for shipping or made a trip to a shop, and they now own something they won't use. The psychological weight of that loss is roughly twice the weight of an equivalent gain, according to the original 1979 prospect theory paper and decades of replication since.
Retailers in the flavour space feel this acutely. A customer who has one bad experience with a mislabeled product is less likely to experiment with the next unfamiliar descriptor, even if that descriptor is accurate. The category's growth depends on trial, and trial depends on trust in the label. When the label is a nicotine number, it answers a question the consumer already knows the answer to — they know their preferred strength before they walk in. When the label is a taste descriptor, it answers the question they're actually asking: what will this be like?
The four trials suggest that the descriptor is doing more work than the industry credits it for. It is not decoration. It is the primary decision aid.
What a Better Label Would Look Like
If taste descriptors outperform nicotine labels on sorting and preference tasks, the forward-looking question is what a descriptor-forward label would actually contain. Three elements recur in the trial data.
First, a sensory anchor rather than a brand name. "Vanilla custard" tells the consumer something. "Velvet Cloud" does not. The most effective descriptors in the trials combined a familiar food reference with a modifier: "bakery cinnamon," "tart green apple," "cool mint with a sweet finish."
Second, a strength cue that is qualitative rather than purely numeric. Consumers in the trials responded better to "bold" versus "mellow" than to "12 mg" versus "6 mg" when predicting whether they'd enjoy a sample, even though the numeric label is more precise. Precision and usefulness are not the same thing.
Third, a note on what the flavour is not. "Not tobacco" and "no cooling agent" were among the most helpful descriptors in the sorting task, because they eliminated whole categories of expectation. Negative descriptors are underused in labelling, and the trial data suggest they carry disproportionate weight.
None of this requires abandoning nicotine information. It requires demoting it from the headline to the fine print, where it belongs for a consumer who already knows their number. The headline should answer the question the tongue is asking.