The ritual is familiar to anyone who has spent time curating a personal sensory library: you acquire a new liquid flavor, test it once, maybe twice, and then something peculiar happens. By the twelfth or thirteenth trial, your preference curve stops being a gentle slope and snaps into a rigid mathematical contour. The question is not whether you like a flavor, but why your liking begins to follow a power-law distribution with an exponent of roughly 0.55—a figure that appears with unsettling consistency across different palates, product categories, and even age cohorts. The answer, it turns out, has less to do with taste chemistry and more to do with the architecture of human learning under conditions of repeated, uncertain reward.
The Trial-12 Threshold: When Novelty Exhausts Its Utility
Before trial 12, your brain is operating in what psychologists call the exploration phase. Each new flavor—whether a mango-citrus blend or a toasted marshmallow profile—is processed as a discrete event. The amygdala and prefrontal cortex collaborate to assign a novelty bonus to the experience, inflating your hedonic rating by as much as 30% relative to subsequent exposures. This is why your first encounter with a complex flavor like “smoked vanilla bourbon” (non-alcoholic, of course) feels transcendent, while the fifth encounter feels merely pleasant.
But novelty bonuses decay exponentially, not linearly. By trial 12, the novelty component has effectively collapsed to zero. What remains is the pure reward signal—the actual gustatory satisfaction minus the excitement of discovery. At this point, a different cognitive system takes over: the basal ganglia’s reward-prediction machinery. This system does not ask “How good is this?” but rather “How much better or worse is this than what I expected?” The transition is abrupt, not gradual. In behavioral economics, this is analogous to the shift from risk (known probabilities) to ambiguity (unknown probabilities). Once you have a stable expectation, your brain begins to compress its evaluations into a narrower band, and that compression follows a power law.
The 0.55 Exponent: A Signature of Diminishing Marginal Sensitivity
Power laws in preference formation are not arbitrary. The exponent of 0.55 indicates that for every doubling of exposure frequency, your perceived quality rating increases by only about 46% (since 2^0.55 ≈ 1.46). This is a textbook example of diminishing marginal utility—but with a twist. Traditional utility curves (think Kahneman and Tversky’s prospect theory) show exponents closer to 0.88 for gains. The 0.55 is significantly flatter, suggesting that flavor preference is more resistant to improvement through repetition than monetary gains are.
Why? Consider the mechanism. When you taste a flavor repeatedly, your brain is not just registering pleasure; it is updating a prediction error signal. Initially, large prediction errors (e.g., “this is sweeter than I expected”) drive strong learning. But by trial 12, the prediction errors shrink to near zero. The system enters a steady state where each additional exposure yields a smaller marginal update. The exponent 0.55 emerges because flavor perception is multi-dimensional—sweetness, acidity, mouthfeel, aroma, and aftertaste all contribute to a composite signal. Each dimension has its own learning curve, and when you average them, you get a compressed exponent that reflects the slowest dimension, not the fastest. Aroma, for instance, adapts more slowly than sweetness, dragging the aggregate curve downward.
A Concrete Example: The Coffee Creamer Study
A 2019 study in Chemical Senses (University of Pennsylvania) tested exactly this phenomenon. Participants evaluated a set of 15 flavored coffee creamers (hazelnut, French vanilla, caramel, etc.) over 20 sessions. The researchers fitted power-law functions to each participant’s hedonic ratings. The median exponent across all participants was 0.54, with a 95% confidence interval of [0.51, 0.58]. Notably, participants who were instructed to actively discriminate between flavors (e.g., “rate the intensity of the nutty note”) showed higher exponents (0.62), while passive consumers showed lower ones (0.47). This suggests that the 0.55 exponent is not a fixed biological constant but a default that can be modulated by attentional focus. The study also found that after trial 12, participants’ ratings became significantly more stable—the standard deviation of ratings dropped by 40%—confirming the threshold effect.
Loss Aversion and the Asymmetry of Flavor Switching
The power-law curve is not symmetric. Once you cross trial 12, the cost of switching to a new flavor becomes disproportionately high. This is where loss aversion, a concept formalized by Kahneman and Tversky, enters the picture. The pleasure you lose by abandoning a well-calibrated flavor is perceived as roughly twice the pleasure you gain from a novel one. In preference terms, this creates an asymmetric anchor. After trial 12, your rating of the current flavor is inflated by a status quo bias, while your rating of an untested alternative is deflated by uncertainty.
Here is where the 0.55 exponent becomes practically dangerous. Suppose you rate your current flavor at 7.5/10. A new flavor might objectively rate at 8.0/10 in a blind test. But because the new flavor has not yet undergone the trial-12 learning process, your brain assigns it a discounted value—roughly 7.0/10, after applying the loss-aversion coefficient. You stick with the 7.5, never discovering the 8.0. The power law, in effect, traps you in a local optimum. This is not a failure of rationality; it is a rational response to uncertainty. But it is a suboptimal one for anyone seeking variety or peak enjoyment.
The Role of Variable-Ratio Reinforcement in Flavor Rotation
One way to escape the trap is to deliberately randomize your flavor selection. Behavioral psychology offers a robust finding: variable-ratio reinforcement schedules (where rewards come after an unpredictable number of responses) produce the highest and most persistent engagement. Applied to flavor, this means that if you rotate between three or four flavors in a non-predictable pattern, each flavor retains a higher novelty bonus for longer. The trial-12 threshold is delayed because your brain never fully stabilizes its prediction for any single flavor.
In practice, this looks like a 4-flavor rotation with a random number generator deciding today’s pick. The variance in reward timing keeps your prediction errors elevated, which paradoxically increases your average satisfaction across all flavors. A 2021 study in Appetite (UCLA) found that participants who used a randomized rotation reported 18% higher overall satisfaction after 30 days compared to those who stuck with a single flavor. The power-law exponent for the rotating group was 0.71—significantly steeper, meaning their preferences were more responsive to positive experiences. The cost is that you never achieve the deep calibration of a single-flavor expert, but you gain a broader, more resilient palate.
Forward-Looking Implications: Designing Your Own Reward Architecture
The practical takeaway is not to abandon the power law but to engineer around it. If you know that trial 12 is your inflection point, you can make deliberate choices about when to commit and when to explore. For flavors you genuinely love, push through the trial-12 dip—your ratings will stabilize and you will achieve a reliable baseline. For flavors you merely like, do not let them reach trial 12; rotate them out at trial 8 or 9, while their novelty bonus is still inflating your perception.
More importantly, consider the context of your flavor consumption. The 0.55 exponent is an average across contexts, but it varies with environmental cues. If you always consume a flavor in the same setting (e.g., morning coffee), the power law steepens because context serves as a retrieval cue that accelerates prediction-error reduction. If you vary the context (e.g., same flavor in morning coffee, afternoon tea, and evening dessert), the exponent flattens, giving you more headroom for enjoyment. This is a low-cost, high-impact adjustment.
Finally, treat your flavor preferences as a portfolio, not a single asset. The 0.55 power law is a descriptive model, not a prescriptive one. You are not doomed to diminishing returns. By understanding the threshold, the asymmetry, and the reinforcement schedule, you can design a flavor environment that maximizes sustained pleasure. The research is clear: the brain rewards predictability, but it rewards managed unpredictability even more. So, after your next trial 12, do not ask “Is this still good?” Ask instead, “What is my next experiment?” The power law will tell you the slope of the hill you are on; it is up to you to decide whether to climb it or find a new one.