Writing down at least five predictions at a fixed time each day and checking them against daily events can train and observe the functioning of perception.

The main trial recruits approximately 600 participants; after a 2-week baseline, they are randomly assigned to three groups: specific prediction practice, matched-duration probability calibration training, and daily observation diary; intervention lasts 8 weeks, followed by 4 weeks of follow-up. The practice group records at least five predictions at a fixed time each day; the primary analysis uses the first five fixed predictions, and the event window is pre-specified as 48–72 hours after recording and cannot be extended post hoc. All three groups complete the same independent daily-event predictions during a uniform testing week; probabilities, hits, and misses are recorded; blind raters score according to pre-registered rules; the primary endpoint is change in Brier score relative to baseline and to controls, and the secondary endpoint is hit rate. With individuals and events as clustering units, and controlling for event base rates, self-fulfillment, and attrition, the project takes about 12–18 months. An emotional sub-trial separately recruits about 400 participants; anxious and neutral films are presented in randomized crossover; after manipulation checks, participants complete ordinary event predictions and a two-alternative target task generated randomly only after responding, with about 200 trials per condition; the pre-registered emotion × task interaction is tested; approximately 9–12 months. If early records are obtainable, first verify original timestamps, complete failure records, and independent emotion coding, and conduct blinded exploratory analyses.

Daily Forecasting Practice, Delayed Feedback, and Emotional State: A Pre-registered Randomized Trial

Only sustained above-chance prediction of targets generated independently and randomly after responding would modify the assumption that predictive advantage comes from information available at the time and from learning.

Barbara Ann Mellers，宾夕法尼亚大学; Philip Eyrikson Tetlock，宾夕法尼亚大学; BIAL Foundation，心理生理学及超心理学研究资助机构

The daily number and fixed time are more specific than general practice suggestions and are replicable; the two-to-three day buffer is fixed according to the proposed scheme for this item, but no minimum improvement magnitude exists; to demonstrate that five predictions or this interval has a special advantage, dose and feedback-lag factorial experiments are still needed.

No directly peer-reviewed experiment was found for the exact combination of Items 1 and 5, and no independently auditable complete prediction–emotion raw data from Jane and Rob were obtained here; for research directions, see [Barbara Ann Mellers](https://marketing.wharton.upenn.edu/profile/mellers/) and [Philip Eyrikson Tetlock](https://mgmt.wharton.upenn.edu/profile/tetlock/).