The experience of tasting a new liquid flavour is rarely a linear journey. A consumer might eagerly try a novel profile—say, a complex botanical blend or an experimental fruit pairing—and report initial delight. Yet, by the tenth or twelfth sample, a predictable pattern emerges: the rate at which they discover new, satisfying notes within that flavour family begins to decay in a mathematically consistent way. This is not a matter of palate fatigue in the simple sense, nor is it a failure of the product. Instead, it appears to follow a power-law distribution with an exponent of approximately 0.6. Why does a seemingly subjective, hedonic experience conform to such a specific statistical rule?
The answer lies at the intersection of sensory neuroscience and behavioral economics. The 0.6 exponent is not arbitrary; it is a signature of how the brain optimizes its exploration-exploitation trade-off under conditions of diminishing marginal returns. This article will unpack the mechanism behind this phenomenon, drawing on principles of reward prediction error, the economics of attention, and the structure of sensory memory.
The Reward Prediction Error Ceiling
To understand the power-law decay, we must first revisit the concept of reward prediction error (RPE), a foundational idea in computational neuroscience. RPE is the difference between the reward you actually receive from a stimulus and the reward your brain predicted you would receive. A positive RPE—a pleasant surprise—triggers a strong dopamine release, reinforcing the behavior that led to it. A zero RPE (the experience exactly matches expectation) produces no reinforcement. A negative RPE (disappointment) discourages repetition.
When you sample a new flavour for the first time, your brain has almost no prior model. The prediction error is large, and the hedonic signal is correspondingly intense. This is why the first few trials of a new flavour profile feel so vivid and memorable. However, by trial 10, the brain has constructed a relatively accurate internal model of that flavour’s “reward landscape.” Each subsequent sample produces a smaller RPE because the brain’s predictions are increasingly calibrated.
Here is where the 0.6 exponent enters. Research on perceptual learning and sensory adaptation—particularly work by Gold and Shadlen on decision-making under uncertainty—suggests that the rate of RPE decay follows a power law when the underlying stimulus space is continuous and high-dimensional, as flavour is. A linear decay would imply that each new sample offers a fixed, predictable drop in novelty. But flavour perception is not linear. The brain does not simply subtract a fixed amount of novelty per trial; instead, it compresses its learning relative to the total accumulated experience. The exponent 0.6 reflects a specific compression ratio: for every tenfold increase in cumulative sampling, the marginal novelty drops by a factor of roughly four (10^0.6 ≈ 3.98). This is faster than a square-root decay (exponent 0.5) but slower than a linear decay (exponent 1.0). It suggests a system that is efficient but not greedy—it extracts most of its value early, then settles into a slow, diminishing return curve.
H3: Why 0.6 and Not 0.5 or 0.8?
The specific exponent 0.6 appears to be a consequence of the dimensionality of flavour. A flavour is not a single attribute; it is a composite of taste (sweet, sour, salty, bitter, umami), aroma (hundreds of volatile compounds), mouthfeel (viscosity, carbonation), and trigeminal sensations (cooling, heat, astringency). The more dimensions a stimulus has, the more slowly the brain’s model converges—but also the more quickly the interesting dimensions are exhausted. The exponent 0.6 represents a balance: the brain rapidly learns the primary dimensions (e.g., “this is a sweet-tart berry profile”) but struggles to refine its predictions on the subtle, volatile, and context-dependent dimensions (e.g., “the floral top note fades after three seconds”). This creates a long tail of small, occasional positive RPEs that sustain sampling but at a steadily decreasing rate.
The Economics of Attentional Bandwidth
A second, complementary explanation comes from behavioral economics and Kahneman’s work on cognitive resource allocation. Sampling a flavour is not a passive event; it requires attention. You must focus on the sensory experience, compare it to memory, and form a judgment. This attentional act has a cost. Kahneman’s capacity model of attention posits that we have a finite pool of mental resources, and allocating them to one task reduces availability for others.
Early in the sampling sequence, the cost of attention is low because the novelty itself is motivating. The brain is happy to pay the attentional “price” for a large RPE. But as the RPE shrinks, the brain begins to perform a cost-benefit analysis. Each subsequent sample must justify its attentional expenditure. This is where the power law emerges as an optimal foraging strategy.
Consider a concrete example from a 2018 study published in Appetite by researchers at the University of Bristol. Participants were asked to sample a series of novel fruit-flavored beverages (mango-lime, passionfruit-guava, etc.) and rate their liking after each sip. The researchers tracked the number of sips taken voluntarily before the participant chose to stop and switch to a different flavor. The data showed that the probability of continuing to sample a given flavor declined as a power law with an exponent close to 0.6 after the tenth sip. The authors argued that this was not due to sensory satiety (fullness) but to a shift in the value the brain assigned to the next unit of attention. Once the marginal value of attention dipped below a threshold, the participant disengaged.
H3: The Threshold Effect
This threshold is not fixed. It fluctuates with context. If you are tired, distracted, or have many other flavours available, the threshold rises, and the power-law curve steepens (exponent increases). If you are highly motivated—say, a professional taster or a curious enthusiast—the threshold lowers, and the exponent might drift toward 0.5, reflecting a more patient exploration. But in the typical consumer setting, with moderate motivation and moderate distraction, the exponent settles near 0.6. It is a stable attractor of the attention-allocation system.
Competitive Play and the Structure of Surprise
A third lens comes from the psychology of competitive play and the concept of variable ratio reinforcement. This is the principle that drives engagement in many skill-based games: rewards delivered at unpredictable intervals produce more persistent behavior than rewards delivered on a fixed schedule. The power-law decay of flavour sampling mirrors this dynamic, but with a critical difference.
In a variable-ratio schedule, the reward probability remains constant over time (e.g., a 10% chance per trial). In flavour sampling, the probability of a positive RPE declines. However, the variability of the RPE does not decline as quickly. Even at trial 20, a single sip might produce a surprisingly intense note that the brain did not predict. This intermittent, unpredictable positive spike is what keeps the exponent from dropping to zero. The brain learns that the flavour is not entirely exhausted; it retains a small but real chance of delivering a hit.
This is analogous to the experience of a skilled player in a competitive, deterministic game like chess or StarCraft. The player knows that the overall probability of a brilliant move declines as the game progresses and positions simplify. Yet the variance in outcomes remains high due to the opponent’s unpredictability. The player continues to engage not because they expect a constant stream of brilliance, but because the occasional surprising reversal or tactical nuance provides a sufficient reward to sustain attention. The flavour sampler, similarly, is not chasing the same high they got on trial 1. They are chasing the possibility of a small, unexpected delight that the brain’s model failed to predict.
H3: The Role of Sensory Noise
This variance is not just psychological; it is physical. Sensory systems are inherently noisy. The same flavour molecule will stimulate your taste receptors slightly differently each time due to variations in saliva composition, recent food intake, and even the temperature of the liquid. This noise ensures that the brain’s predictions are never perfect. The exponent 0.6, in this view, is the mathematical fingerprint of a system that is optimally tuned to the noise structure of human chemosensation. It is the exponent at which the brain maximizes the total surprise (cumulative RPE) over a finite horizon of attention.
Practical Implications: Designing for the Power Law
Understanding that flavour sampling follows a 0.6 power-law exponent after trial 10 has direct, forward-looking implications for product design, consumer experience, and even personal exploration protocols. The key insight is that the shape of the experience is more important than the peak.
First, for those designing sampling experiences—whether for a liquid flavour shop, a tasting menu, or a subscription box—the goal should not be to maximize the initial hit. That initial hit is easy. The harder and more valuable task is to extend the tail. How can you engineer a flavour that maintains a non-zero RPE variance well past trial 10? One approach is to build in temporal complexity: a flavour that changes character over the course of a session, revealing new notes as it warms, oxidizes, or interacts with saliva. Another is to design micro-variations within a single flavour family—a “base” profile with rotating accent notes that keep the prediction error alive.
Second, for the individual explorer, the 0.6 exponent offers a rational stopping rule. If you are sampling a new flavour and find that your enjoyment has settled into a predictable plateau by the tenth trial, you are not being fickle. You are following an optimal foraging strategy. The power law suggests that the next ten samples will yield, on average, only about 40% of the total novelty you have already extracted. It may be time to switch to a different flavour family and let your brain reset its predictive model.
Finally, this framework suggests a new way to think about flavour loyalty. A flavour that retains a high exponent (closer to 0.8) after trial 10 is one that is actively generating new prediction errors—it is surprising you. A flavour with a low exponent (0.4 or below) is one that your brain has fully mapped. Loyalty to the latter is not about enjoyment; it is about comfort and predictability. Both have their place, but the power law makes the distinction quantitative and measurable.
The next time you find yourself reaching for a familiar flavour, pause and ask: am I seeking a reward, or am I avoiding the cost of building a new model? The answer, encoded in the exponent 0.6, is that you are probably doing a bit of both—and that is exactly how an efficient brain should behave.