A generous-looking serving can contain less ice cream. Christopher Hsee’s 1998 experiment made that distinction visible by comparing a portion piled above a small cup’s rim with a larger portion sitting below the rim of a bigger cup.
In “Less Is Better: When Low-Value Options Are Valued More Highly Than High-Value Options,” participants evaluated hypothetical servings illustrated with drawings. One contained seven ounces in a five-ounce cup; the other contained eight ounces in a ten-ounce cup. The task asked how much they would pay, rather than asking them to taste or purchase the ice cream.
When evaluated separately, the smaller serving attracted an average valuation of $2.26, compared with $1.66 for the larger one. When participants evaluated both together, the order reversed: $1.85 for eight ounces and $1.56 for seven.
The cup supplies a comparison of its own
Hsee’s explanation centered on evaluability. An isolated quantity can be difficult to judge, while a cup’s fullness provides an immediately available reference. Comparing the two servings directly makes their quantities easier to evaluate.
This is a useful distinction between information being present and information being easy to use. A number on a page can be accurate without telling the reader whether it represents a lot, a little or good value. A visual boundary can supply an answer quickly, even when that boundary is not the most relevant comparison.
Consider a hypothetical storage display showing 70 units against a capacity of 50 and another showing 80 against 100. The first exceeds its displayed reference, while the second falls short. That relationship is different from the absolute quantity. A clear interface needs to make it apparent which of those two questions a user is answering.
The container becomes part of the judgment.
Separate evaluation means different viewing conditions
The result should not be described as everyone choosing the smaller serving after inspecting both. Separate evaluation means people assessed one option in isolation; joint evaluation allowed a direct comparison. The difference between those conditions is part of the finding.
For a product team, that makes the route to a decision worth investigating. A customer viewing a single offer on a landing page has a different comparison set from someone reading a table of competing plans. Identical product details can appear in both places while the available reference points differ.
A useful test would preserve the information and vary its presentation, then measure the decision of interest. It would also distinguish a stated valuation from an actual purchase. Those outcomes may be related, but evidence about one does not automatically establish the other.
A preregistered replication found the pattern again
A 2023 paper in Collabra: Psychology reported a preregistered close replication and extension of three studies from Hsee’s original article, using a sample of 403 participants. It examined scenarios involving gifts, ice cream and dish sets.
The researchers found support for the less-is-better pattern under separate evaluation across all three. The ice-cream effect was smaller than in the original research: the reported standardized effect size was 0.32, compared with 0.74 originally.
The ice-cream scenario also supported the reverse ordering when the servings were evaluated jointly. However, support for the broader more-is-better prediction was weaker across the replication’s scenarios, particularly for the dish sets.
That combination matters. A replication can support the direction of a finding without reproducing its size, and results can differ between tasks within the same research program. Reporting only that a study “replicated” would leave out information useful to anyone judging how broadly to apply it.
Side-by-side comparison is not always a cure
Other research shows why joint evaluation should not be treated as a universal solution. In a 2021 paper on efficiency and consumption judgments, Eyal Gamliel and Eyal Pe’er examined cases where people misjudge improvements in measures such as fuel efficiency.
Efficiency and resource savings do not always have a straight-line relationship. The authors found that, under some conditions, the reduced sensitivity to magnitude associated with separate evaluation happened to produce judgments closer to the actual savings. Joint comparisons could instead reinforce mistaken linear reasoning.
That is a boundary on the design lesson, rather than a contradiction of the ice-cream result. Comparing quantities of the same product is a different task from interpreting a rate whose practical meaning requires another calculation. The best display depends on what the reader needs to understand.
Show the reference that matters
For a business presenting an offer, the practical question is which comparison the page makes easiest. Package fullness, a discount percentage, a unit price and total quantity each answer different questions. Making one prominent can leave another harder to assess.
A transparent comparison should make the relevant quantities and costs accessible, while avoiding the assumption that visual fullness is equivalent to value. Testing should then use an outcome appropriate to the claim, with the context and comparison conditions recorded.
The experiment’s appeal lies in a simple mismatch: one serving looked more generous while containing less. Understanding that mismatch requires keeping both the amount of ice cream and the way it was judged in view.
