The reason the middle option wins when there are three

Three coffee sizes, and the medium takes the largest share. Add a fourth, larger size and the old medium loses its crown to what used to be the large. The drink did not change and neither did the customer’s thirst. The winning option is defined by where it sits, not by what it is.

What everyone sees

The pattern gets described as moderation, a preference for the sensible middle. That framing treats the middle as a property of the product when it is a property of the assortment. Whoever controls the range controls which item is moderate, which means the retailer selects the customer’s moderate choice by choosing the extremes.

What is actually happening

Simonson and Tversky showed that adding an extreme option to a set shifts share towards the option that becomes intermediate, and attributed it to extremeness aversion: intermediate positions are less vulnerable to criticism on any single attribute. Sharpe, Staelin and Huber applied the mechanism to portion sizes and found that adding a larger option pulled choices upward, so the mechanism can be aimed. The critical structural point is that this violates a basic assumption of stable preference, since the ranking between two unchanged options is reversed by the arrival of a third. The customer is not consulting an internal ordering. They are locating a position within whatever range is presented, and the range is an editorial decision made upstream.

Why it stays hidden

The dependency hides because the assortment is invisible as a choice. A menu arrives looking like a description of what exists rather than a selection from what could have been offered. Since the discarded alternatives are never seen, the customer cannot notice that their moderate was manufactured, and experiences a range-relative decision as a personal preference.