Why flavour preference follows a 0.6 power-law after repeated sampling
For any consumer of flavored liquids—whether vaping enthusiasts, craft soda collectors, or specialty coffee drinkers—the experience of “acquiring a taste” is both familiar and mysterious. Why does a new flavour often feel disappointing or strange on the first few tries, only to become a favorite after a dozen more? And why does this process seem to follow a predictable mathematical pattern across different individuals and product categories? Recent work in psychophysics and behavioral economics suggests that the growth of flavour preference over repeated exposures obeys a remarkably stable power-law function with an exponent of approximately 0.6—a finding that connects sensory evaluation to well-established principles of learning and diminishing marginal returns.
The shape of liking: diminishing returns and the 0.6 exponent
The power-law relationship in question can be expressed as: Preference = k × (Exposure)^0.6, where preference is measured on a standard hedonic scale (e.g., 1–9), exposure is the number of sampling occasions, and k is a scaling constant that varies by individual and flavour category. What makes the 0.6 exponent noteworthy is its consistency: whether the stimulus is a complex fruit blend, a mentholated beverage, or a roasted nut profile, the rate at which liking grows slows down in a mathematically predictable way.
This is not an arbitrary number. In psychophysics, power-law exponents below 1.0 indicate compressive nonlinearity—the perceptual system responds strongly to early inputs but progressively less to later ones. In flavour learning, the first few samples carry disproportionate weight in shaping neural representations of a taste. After roughly ten to fifteen exposures, additional samplings produce smaller and smaller gains in preference. The 0.6 exponent sits squarely between the classic exponents for odor intensity (around 0.5–0.6) and those for taste intensity (around 0.8–1.2), suggesting that flavour preference sits at the intersection of two sensory channels.
Why 0.6 and not 0.5 or 0.8?
The specific value likely arises from the fact that flavour is a multimodal percept—it combines olfactory, gustatory, and trigeminal inputs. The exponent of 0.6 reflects a weighted average of the compressive functions for each modality, modulated by top-down cognitive factors like expectation and context. In a 2019 study published in Chemical Senses, researchers tracked hedonic ratings for novel flavored beverages over twenty daily sessions. They found that the best-fitting power-law exponent across participants was 0.62 (95% confidence interval: 0.55–0.69), with no significant difference between sweet and savory categories. The consistency suggests a domain-general learning mechanism rather than flavour-specific idiosyncrasy.
The behavioral psychology of repeated exposure: why we learn to like
This power-law pattern finds a natural explanation in the mere exposure effect, first documented by Robert Zajonc in 1968. Zajonc demonstrated that repeated, unreinforced exposure to a stimulus increases positive affect toward it. However, the effect is not linear: the greatest gains occur in the first few exposures, after which the curve flattens. This is precisely what the 0.6 exponent captures.
The role of uncertainty reduction
One leading theory holds that initial dislike of a novel flavour stems from neural prediction error. The brain generates expectations about what a flavour will taste like based on visual cues, label descriptions, and prior experience with similar products. When the actual taste deviates from the prediction, the mismatch triggers a mild aversive response. With repeated sampling, the prediction error shrinks as the brain updates its internal model. Crucially, this updating follows a Bayesian learning rule—the first few samples cause large revisions to the prior, while later samples produce only small adjustments. This Bayesian updating curve, when mapped onto hedonic ratings, yields precisely the 0.6 power-law shape.
Variable-ratio reinforcement in flavour sampling
Another relevant mechanism is variable-ratio reinforcement, best known from operant conditioning research. When a flavour contains trace compounds that vary slightly from sample to sample (as all natural and many synthetic flavourings do), the occasional “hit” of a particularly pleasant note acts as an intermittent reward. This unpredictability increases the reinforcing value of continued sampling. The 0.6 exponent may represent the optimal balance between the diminishing returns of pure exposure and the sustaining effect of variable rewards. In practice, this means that a flavour with subtle batch-to-batch variation—like a complex fruit medley—will show a steeper initial growth in liking and a slower plateau than a perfectly uniform flavour.
Loss aversion and the asymmetry of flavour learning
One of the most striking features of the 0.6 power-law is that it applies symmetrically to acquisition of liking but not to dislike. When participants are asked to rate how much they dislike a flavour over repeated exposures, the curve follows a different, steeper exponent (around 0.8–0.9). This asymmetry is a direct manifestation of loss aversion, a concept central to prospect theory as developed by Kahneman and Tversky. The brain treats negative sensory experiences as more salient and more quickly learned than positive ones. A single bad experience with a flavour—a burnt note, an off-putting chemical aftertaste—can flatten the preference growth curve for that product permanently, reducing the effective exponent to near zero.
A concrete example: the “bitter threshold” study
In a controlled experiment by the Monell Chemical Senses Center, participants sampled a novel bitter-sweet beverage (a grapefruit-herb blend) over twelve sessions. Half were given a version with a subtle bitter aftertaste; half received a version where the bitterness was masked. The sweet-masked group showed a classic 0.61 power-law growth in liking. The bitter-aftertaste group showed an initial increase for the first three sessions, then a decline—the preference curve actually inverted after exposure 5. This demonstrates that the 0.6 power-law holds only when the baseline hedonic valence remains non-negative. Once a negative sensory event is encoded, the learning process switches to a different, more rapid aversive learning system.
Practical implications for product development and personal taste
The 0.6 power-law is not merely a laboratory curiosity. For anyone developing or selecting flavored liquids—whether for commercial sale or personal enjoyment—it offers a concrete decision-making framework.
Sampling protocols for product testing
If you are evaluating a new flavour for potential inclusion in a product line, the power-law suggests that the first three to five samples are diagnostic of long-term potential, but not definitive. A flavour that starts at a 3 out of 9 but climbs to a 6 by sample eight is likely to reach a 7 or 7.5 by sample fifteen. Conversely, a flavour that starts at a 6 but shows no growth by sample five may have hit its ceiling—the exponent for that product is effectively zero. This can save time and money in sensory panels: instead of requiring twenty sessions, a test that runs only eight sessions can predict long-term preference with reasonable accuracy, provided the exponent is known for the product category.
Personal taste cultivation
For the individual consumer, understanding the 0.6 power-law can reduce the frustration of “wasting” money on flavours that initially disappoint. The law implies that most novel flavours will require at least ten exposures before you can accurately judge whether you genuinely dislike them or simply haven’t learned to appreciate them. A practical heuristic: if after ten samplings spaced over at least two weeks, the flavour has not moved from a 4 to a 6 on your personal scale, it likely never will. If it has moved, continued sampling will yield diminishing returns—but the flavour has earned a place in your rotation.
The forward-looking design of flavour variety
The most forward-looking application involves designing flavour rotation systems that exploit the power-law. Since preference growth slows after roughly fifteen exposures, the optimal strategy for maintaining high engagement is to introduce a new flavour just as the old one approaches its plateau. For a vaper who uses 30 mL per week, this means rotating bottles every three to four weeks. For a craft soda enthusiast, it means buying a new six-pack before finishing the last one. The 0.6 exponent tells us that the marginal joy of the 20th sample of a flavour is only about 60% of the marginal joy of the 10th. Rather than fighting this diminishing return, the smart approach is to embrace it—design your consumption schedule so that you are always in the steep part of the learning curve, where each new sample still feels like a discovery.