Feeling the Past Investment: Two Different Questions in Sunk-Cost Decisions

Time already spent on an activity cannot be recovered by spending another hour on it. Yet stopping can feel like losing that earlier time now. Two questions then compete: which option offers the better future outcome, and how does it feel to leave the previous investment behind? A decision about what comes next can carry an emotional response to what has already happened. Sunk-cost research investigates that influence without making persistence itself a universal sign of error.

Koen A. Dijkstra and Ying-yi Hong examined context-specific feelings in a 2019 paper. Across four studies, participants read hypothetical situations and reported both their affective responses and their preferred actions. The researchers distinguished feelings elicited by the particular decision from a generally unpleasant mood. Their findings connected scenario-related negative affect with continued commitment, but the pattern of group means in the cognitive-load study complicates any simple practical message. Reading the measurements separately is essential to understanding what the paper actually established. [1]

An irrecoverable cost is not all information about the past

A sunk cost is a resource that cannot be recovered whichever option is chosen next. Ignoring it in the relevant comparison does not mean that all past experience should be discarded. Experience may change a forecast of future success. Switching may introduce a new cost. Continuing may genuinely offer greater future benefits. Those are different considerations from continuing solely because something has already been spent.

The action of persisting cannot therefore be labeled a mistake without examining the alternatives. In the scenarios of interest, changing the earlier investment can change the preference even when the future-facing comparison would not justify that change. The research asks how an irretrievable commitment enters the decision. It is not an assessment of a particular person's investments or a recommendation about a real financial transaction. [1]

A feeling attached to the decision rather than the day

The paper separates integral affect, elicited by the option or situation being considered, from incidental affect arising elsewhere. Feeling bad about abandoning a particular project is not the same measurement target as being in a bad mood before reading about it. The two can coexist, but should not be treated as interchangeable explanations. Asking about feelings after each scenario offered a way to investigate the response attached to that specific context.

The first study recruited eighty-nine Chinese University of Hong Kong students. Three were excluded under an affect-score outlier criterion. Participants considered ten hypothetical scenarios, reporting affect before indicating their preferences. Low affect scores meant more negative feelings. Low choice scores meant a stronger preference for the sunk-cost option. Higher choice scores therefore indicated less, not more, susceptibility according to the study's normative scoring. [1]

The two scales were positively correlated: more negative feelings went with a stronger preference for the sunk-cost alternative. That association does not, by itself, demonstrate that emotion independently caused the decision. Both variables were self-reports about the same situations, and choices did not require participants to commit actual resources. The result identifies a relationship worth explaining rather than completing the causal explanation in a single step. [1]

Adding a prior investment to a similar situation

The second study presented paired scenarios to the same participants. One member of each pair described a previous investment; its counterpart described no investment or a smaller investment. For a boring television movie, the pair differed in whether a payment had been made. For a hobby project, the comparison was five years of work versus one month. The latter was not literally an investment-versus-zero comparison. It varied the amount of prior commitment. [1]

Sixty online participants were recruited and fifty-seven remained after the reported outlier exclusions. Investment versions elicited more negative affect and a stronger preference to continue. A within-person statistical mediation analysis also supported an indirect relationship connecting the scenario difference, the affective-response difference and the choice difference.

Statistical mediation is not the same as independently randomizing the mediator. The feelings were measured, not separately assigned while every other feature was held constant. The analysis is consistent with the proposed emotional account, but cannot by itself exclude every alternative process connecting the responses. The distinction matters especially when a paper's language moves from describing an association to claiming that one psychological event drives another. [1]

What did generating a reason change?

In the third study, one group was asked to spend a minute thinking about reasons for its decision and write one down. The comparison group was instructed not to think too much before responding. Choices in the justification group were, on average, less sunk-cost-oriented, with statistical evidence at the conventional boundary. The groups also differed in reported affect. However, the manipulation did not isolate reason-writing against an equal-duration condition with otherwise matched instructions. [1]

A reporting detail further limits the emotional interpretation. The methods describe measuring affect before the instruction to think of reasons, whereas the discussion emphasizes justification making affect less negative. Without resolving that temporal ambiguity, it would be unsafe to state that the one-minute intervention demonstrably changed a feeling already measured. Reporting a group difference and establishing the sequence that produced it are separate tasks. The published description does not close that gap. [1]

Cognitive load: an average and an association answer different questions

The fourth study required participants to maintain a pattern of check marks while considering a scenario. Both conditions used a three-by-three grid. The easier and harder patterns differed, not the dimensions of the grid. Participants saw the pattern, read the scenario, rated their feelings, selected an action and finally reconstructed the pattern. This imposed an additional storage demand during the decision. [1]

The relationship between affect and choice was stronger in the high-load condition. Yet the mean choices in that condition were less sunk-cost-oriented than those in the low-load condition. These observations do not have to contradict one another. The strength of an association among people within a group is a different property from that group's average score. A steeper relationship does not establish that the group generally made more biased choices.

Consequently, the discussion's suggestion that reducing load could protect against the fallacy should not be repeated as a universal rule directly demonstrated by the group means. The study offers evidence about affect's relationship with preferences under load. Its average-choice pattern does not support the simpler statement that greater load necessarily produced greater sunk-cost bias. Keeping the two questions apart prevents a psychologically appealing account from replacing the actual result. [1]

A bounded conclusion about commitment

Together, the studies connect feelings about past investment with preferences about continuing. They do not establish that hypothetical responses reproduce every feature of consequential, long-term behavior. Self-report, the ordering of questions, outlier exclusions and sample size all affect interpretation. In the second study, the recruited sample was also below the target identified in its own power calculation. Those details belong alongside the findings, not outside the account of their reliability. [1]

The useful distinction is between feeling the loss of an earlier investment and evaluating the value of the next step. Neither suppressing every feeling nor condemning every act of persistence resolves that distinction. A more informative question is whether the reason for continuing concerns a future outcome or the wish not to count an irrecoverable investment as lost. With the evidence kept in proportion, sunk cost becomes less a label for thoughtlessness than a concrete problem about how different information and feelings enter the same decision. [1]

Sources

[1] Dijkstra, K. A., and Hong, Y.-y. (2019). The feeling of throwing good money after bad: The role of affective reaction in the sunk-cost fallacy. PLOS ONE, 14(1), e0209900. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0209900

The source study is distributed under Creative Commons Attribution: https://creativecommons.org/licenses/by/4.0/ This is an original explanatory and critical account, not an official translation of the paper.