People often find it difficult to make decisions, that is, to choose among available options. One difficulty is that risk or uncertainty is associated with decision outcomes. Another difficulty is the requirement to resolve value conflicts. We focus here on still another, less frequently stressed difficulty, namely that of making rational decisions when facing the outcomes of prior decisions. Such a difficulty is exemplified by the sunk cost effect, that is, the irrational effect of a previous loss on evaluations of future courses of action. Our approach is however not normative but descriptive. In the following paragraphs we review some descriptive accounts of how prior outcomes affect risky decisions.

Some descriptive theories aiming at explaining how people make decisions are similar to normative theories in that they assume that a decision maker first assigns a utility or value v(x) to each outcome x, then select the option with the highest sum of values across all outcomes. An example is prospect theory. A difference to normative decision theory is the assumption made in prospect theory that a decision maker edits options prior to assigning values to outcomes. Editing operations include framing of outcomes as gains or losses relative to a reference point. Framing also involves segregating or integrating prior outcomes. In this respect Tversky and Kahneman (1981) believe that prior outcomes are frequently segregated because this “..(i) simplifies evaluations and reduces cognitive strain, (ii) reflects the intuition that consequences should be causally linked to acts, and (iii) matches the properties of hedonic experience which is more sensitive to desirable and undesirable changes than to steady states” (p. 457).

To support the belief that outcomes are frequently segregated (also referred to as the isolation effect), Kahneman and Tversky (1979) and Tversky and Kahneman (1981) report an empirical demonstration. Subjects were asked to imagine that they had been given 1000, then to indicate whether they would choose a sure loss of 500 over an equal probability of losing 1000 or nothing. A majority of subjects chose the risky loss exactly as they would have done if there was no prior outcome (i.e., the amount they imagined to have been given). If subjects had edited the options to take the prior outcome into account, according to prospect theory they would instead have chosen the sure gain of 500 (i.e., the 1000 they were given minus the sure loss of 500) over a risky gain of 1000 or nothing.

Although subjects segregated a prior outcome in this demonstration, other researchers have found an influence of prior outcomes on subsequent choices. Therefore, the questions need to be asked when, how, and why subjects integrate instead of segregate prior outcomes? Integration of prior outcomes sometimes occurs despite that it requires cognitive effort. In order to account for this, several motivational explanations have therefore been proposed. In this vein Thaler and Johnson (1990) suggested that prospect theory should incorporate a hedonic editing rule. Whereas other editing rules are employed for the sake of cognitive simplification, the goal of the hedonic editing rule (i.e., integrating/segregating a prior outcome) is to maximize value. An alternative account is labeled the renewable resources model. The point of departure is still that a decision maker maximizes value. Since a gain is believed to buffer a loss, outcomes entailing gains and losses are integrated. However, being aware of his or her limited but renewable resources to cope with large losses, multiple losses are aversive to the decision maker who will therefore segregate them. Gain-savoring resources are also perceived to be limited but renewable. Therefore, multiple gains are also segregated.

As noted by Larrick (1993), in risky choices people are often more concerned about avoiding negative outcomes than attaining positive ones. This may reflect that anticipated negative events are more salient. A compatible editing rule of integration/segregation proposed by Gärling and Romanus (1997) as the loss-sensitivity principle is to only add the prior outcome to the expected loss of the current choice. Since a prior loss is integrated with an expected loss, the dissatisfaction with the expected loss will increase after a loss. Conversely, since a prior gain is integrated with an expected loss, the dissatisfaction will decrease after a gain. Satisfaction with an expected gain is however not affected by a prior outcome.