The reflection effect (Tversky & Kahneman, 1981) refers to having opposite preferences for gambles differing in the sign of the outcomes (i.e. whether the outcomes are gains or losses). Reflection effects involve gambles whose outcomes are opposite in sign, although they do have the same magnitude. For example, most people would choose a certain gain of $20 over a one-third chance of gaining $60. But they would choose a one-third chance of losing $60 (and two-thirds chance of losing nothing) over a certain loss of $20. The outcomes actually involve different domains (gain versus loss), that is, they differ in sign (+$20 versus -$20).

The difference between reflection and framing effect is that in the framing effect the actual domain does not change (Fagley, 1993); the same outcome is phrased to appear to involve the other domain. So a loss of $20 might be framed to seem like a gain (as when an even larger loss was expected). Framing may cause it to seem like a gain, but it remains, objectively, a loss.

Reflection and framing effects are both predicted in prospect theory by the S shape of the value function: concave for gains indicating risk aversion and convex for losses indicating risk seeking.