Decision fatigue statistics come from several kinds of evidence: controlled experiments, field observations, clinical records, and measurement-scale studies. The results are not uniform. Some studies associate repeated choices or late-session timing with lower persistence or fewer tests, while others find no important shift effect or show substantial differences between professionals. The measurement period and setting matter throughout.
Contents
- Controlled experiments and self-control
- Choice volume, time, and persistence
- Shopping and consumer behavior
- Sequential decisions in judicial settings
- Time of day in outpatient medicine
- Emergency-department findings
- Measuring decision fatigue in nurses
Controlled experiments and self-control
The 2008 laboratory experiments reported in Making choices impairs subsequent self-control: A limited-resource account of decision making, self-regulation, and active initiative compared participants who selected options with participants who rated or read the same materials without choosing. The results describe short-term performance differences after a choice task; they do not establish that every real-world decision produces fatigue.
In one experiment, participants made product and occupation choices and then drank an unpleasant vinegar mixture. The choice group drank an average of 2.06 ounces, compared with 7.67 ounces among participants who rated the materials. In a matched-duration experiment, the corresponding averages were 1.89 and 6.87 ounces. The choice task involved a 60-item questionnaire containing 292 decisions. Vohs et al., 2008
The choice task averaged 210.32 seconds, versus 146.32 seconds for the frequency-rating task. Mean positive affect was 24.31 in the choice condition and 25.05 in the rating condition on the study’s PANAS measure. Mean negative affect was 13.19 and 11.89, respectively. These affect scores provide context for the task but are not a general population measure of tiredness.
Several physical-persistence tasks showed the same direction. After product choices, participants held an arm in 1°C water for 27.70 seconds on average, compared with 67.42 seconds after product ratings. Students who chose courses for degree requirements practiced mathematics for 8.39 minutes, compared with 11.40 minutes after reading course options without choosing. In that course-choice experiment, the manipulation did not significantly change the number of math problems attempted, completed, or answered correctly; each comparison had t less than 1.
After 35 consequential course-material choices, students persisted on unsolvable tracing puzzles for 9.11 minutes, versus 12.25 minutes after carefully reading the same material without choosing. In a solvable-problem experiment, persistence was 14.70 minutes after choices and 17.80 minutes in the no-choice group. The choice group also had a significantly higher math error rate per attempted problem, with F(1,38)=5.10.
Choice volume, time, and persistence
The same 2008 research separated actively choosing from deliberating without selecting and from implementing preset choices. In a Dell computer-configuration task, mean deliberation ratings were 5.64 for active choosers, 5.82 for deliberation-only participants, and 2.82 for participants implementing preset choices. Task duration averaged 273.24 seconds for active choice, 320.95 seconds for deliberation-only, and 223.18 seconds for implementation.
The subsequent anagram task measured persistence in seconds. Active choosers persisted for 379.24 seconds on average, compared with 571.29 seconds for implementers and 514.0 seconds for participants who deliberated but did not select. The overall difference among the three conditions was statistically significant, F(2,61)=3.99, p<.03. The planned contrast comparing active choice with both nonchoice conditions was also significant, t(61)=2.77, p<.01. Active choice differed from implementation at t(61)=2.66, p<.01 and from deliberation at t(61)=2.01, p<.05. Vohs et al., 2008
The study also used a threefold duration contrast: its high-choice condition lasted 12 minutes and its low-choice condition lasted 4 minutes. Duration and choice count therefore appear together in this comparison, so the result should not be read as an isolated estimate of the effect of one additional decision.
| 2008 Dell configuration condition | Mean task time | Mean anagram persistence |
|---|---|---|
| Active choice | 273.24 seconds | 379.24 seconds |
| Deliberation only | 320.95 seconds | 514.0 seconds |
| Preset implementation | 223.18 seconds | 571.29 seconds |
The pattern is useful for interpreting decision fatigue: the condition requiring active selection had the shortest later persistence, even though it was not the longest configuration task. That is an experimental association, not a universal rule for work or consumer decisions.
Shopping and consumer behavior
The 2008 shopping-mall field study examined choice-related measures, self-reported tiredness, and arithmetic performance. The choice-scale factor explained 49% of variance, while the tiredness factor explained an additional 17%. Among shoppers, the choice-factor score predicted fewer correctly completed addition problems, with standardized beta -0.32, t(50)=2.40, p=.02. The full regression predicting correctly solved addition problems was significant, F(6,50)=2.48, p<.04.
Self-reported tiredness behaved differently in that model. Its standardized beta was -0.04, and it did not significantly predict correctly solved addition problems. This distinction matters because a measured choice burden and a person’s stated tiredness were not interchangeable variables in the field analysis. Shoppers had reported shopping durations ranging from 1 minute to 4.5 hours before completing the arithmetic task.
A separate gift-registry experiment measured anticipated enjoyment, which averaged 4.21 on a 1-to-7 scale. That figure describes anticipated enjoyment in the experiment; it is not a prevalence estimate for decision fatigue or a measure of shopping exhaustion. Vohs et al., 2008
Sequential decisions in judicial settings
The Israeli parole-board study, Extraneous factors in judicial decisions, analyzed 1,112 rulings collected over 50 days within a 10-month period in 2010–2011. Judges had mean professional experience of 22.5 years, with a standard deviation of 2.5 years. The sample included 727 Jewish-Israeli men, 326 Arab-Israeli men, 50 Jewish-Israeli women, and 9 Arab-Israeli women. Parole requests represented 78.2% of the decisions.
Judges considered 14 to 35 cases per day, averaging 22.58 cases with a standard deviation of 4.67. Each deliberation lasted about 6 minutes on average, with a reported mean of 5.98 minutes and standard deviation 5.13. The morning food break began between 9:49 and 10:27 a.m. and lasted 38.48 minutes on average. The afternoon break began between 12:46 and 2:10 p.m. and lasted 57.37 minutes.
Breaks occurred after an average of 7.8 cases in the morning session and 11.4 cases in the post-snack, pre-lunch session. Overall, 64.2% of prisoner requests were rejected. Favorable rulings declined from approximately 65% at the beginning of each decision session to nearly zero later in the session, then returned to about 65% after a food break. Danziger, Levav, and Avnaim-Pesso, 2010–2011
These figures describe a judicial dataset and its session structure. They do not by themselves identify the psychological mechanism behind the pattern, and they should not be generalized to every judge, parole system, or legal decision.
Time of day in outpatient medicine
The retrospective outpatient study covered January through December 2017 and analyzed 3,342 appointments by 20 attending physicians at five clinics. It compared activity in the first hour and last hour of clinic sessions. Four of the 20 physicians showed statistically significant first-versus-last-hour differences in diagnostic tests ordered per visit, while six showed significant differences in diagnoses assessed per visit.
Sixteen of 20 physicians ordered more diagnostic tests per patient on average in the first hour than in the last hour; the p-value for equal directional probability was .012. Four of five clinics showed substantial within-clinic heterogeneity in the two decision-process outcomes, with p<.01. The analytic sample contained 2,013 first-hour appointments and 1,329 last-hour appointments. In the broader source data, 39 physicians and 5,354 unique appointments were initially identified before eligibility filtering.
The physician-level results were not uniform. One physician averaged 3.3 diagnostic tests per first-hour visit and 2.95 per last-hour visit, a difference of -0.35 tests per visit. The same physician averaged 3.12 diagnoses assessed per first-hour visit and 2.66 per last-hour visit, a difference of -0.46 diagnoses per visit. Another physician showed the reverse test-ordering pattern: 14.86 tests in the first hour versus 21.10 in the last hour, with adjusted p<.001. Time-of-day changes in physician clinical decision making, 2017
Emergency-department findings
An emergency-department study examined 87,752 patient visits over a two-year administrative period; the years were not stated in the abstract. It found no important shift-related difference in consultation rates. Consultations that did not result in admission also did not differ importantly across the eight hourly periods of the physician shift. Emergency-department return visits within 72 hours did not differ importantly by the hour at which patients were seen.
Some ordering outcomes did change over the shift. CT-head ordering fell from 15.8% in the first shift hour to 12.2% in the last hour, with OR 0.73 and 95% CI 0.66–0.80. CT-abdomen ordering fell from 9.6% to 7.6%, with OR 0.72 and 95% CI 0.64–0.80. Emergency-department length of stay fell from 5.5 hours in the first shift hour to 4.9 hours in the last, a relative difference of 0.83 with 95% CI 0.81–0.85. Decision fatigue in the Emergency Department
Taken together, the emergency findings support a narrower interpretation than “all decisions worsen with time.” Consultation and return-visit outcomes showed no important hourly difference in this study, while CT ordering and length of stay moved in the opposite direction from a simple assumption that later care always means more testing or slower decisions.
Measuring decision fatigue in nurses
The U.S. clinical-nurse study used a nine-item Decision Fatigue Scale. Each item was scored from 0 to 3, producing a total range of 0–27. The U.S. sample’s mean score was 9.6, compared with previously reported means of 5.2 at baseline and 6.8 eight weeks later in the original caregiver work. The measurement period is reported as beginning 21 July 2020 onward, during the COVID-19 pandemic context.
The scale had Cronbach’s alpha of .95, and removing any of the nine items did not significantly improve reliability. A single factor had an eigenvalue of 6.4 and explained 70.9% of variance. All nine items had factor loadings above .40, ranging from .70 to .91. These are scale-development and validation statistics; they do not convert the score into a percentage probability of impaired judgment.
Decision-fatigue scores correlated negatively with practice-environment scores at rho=-.32, p<.01, and positively with traumatic-stress scores at rho=.55, p<.01. Scores were not significantly correlated with practice location, rho=.11 and p=.16, or with whether respondents or someone they knew had COVID-19, rho=.09 and p=.24. Decision fatigue among clinical nurses during the COVID-19 pandemic
The nursing results show why decision fatigue statistics need both an instrument and a setting. A 0–27 scale can quantify reported symptoms or experiences within a study, while correlations describe relationships in that sample. Neither result alone establishes that fatigue caused a particular clinical decision or predicts performance for every nurse.