A second comment on "The CRAVE scale to measure motivational states for physical activity and sedentary behavior: Psychometric properties and psychophysical correlates": Evidence from the released data

After our initial comment documented internal inconsistencies in Samadieh et al. (2025), the authors released the data and AMOS model files. In a written response dated 9 October 2026, they acknowledged reporting errors and inappropriate statistical comparisons while maintaining their main conclusions. The released materials reproduce much of the published statistical output but reveal more consequential problems. Most important is the extensive duplication of observations within and across studies. Among the reported samples of 313, 346, and 244 participants, only 98, 206, and 83 distinct 26-item CRAVE response profiles appear. Moreover, 238 of 244 Study 3 rows exactly reproduce a Study 1 CRAVE profile, and 274 of 346 Study 2 rows differ on at most three CRAVE items from a Study 1 row with the same gender, marital status, education, employment status, and BMI. We also identify discrepancies between stored outcome totals and their constituent items, which favor the hypothesized Right Now associations. In Study 3, 26 stored physical-dysfunction or vitality totals differ from the sums of their released constituent items; the discrepancies occur in 23 rows representing 16 distinct underlying records, and each shifts the stored outcome in the direction predicted by that row's Right Now MOVE and REST scores. Separately, 19 Study 3 rows matching Study 1 profiles have BMI values 9.9-10.0 units higher. Substituting the released item sums and corresponding Study 1 BMI values substantially weakens or eliminates several criterion-validity associations. Finally, the AMOS specifications that reproduce the published fit statistics contain 19 residual covariances not described in the article, and parallel analysis does not support two components for the 10-item Right Now version. These discrepancies occur in the released records themselves and are not resolved by manuscript corrections. Resolving the provenance of the released records requires the original survey exports, timestamps, participant linkage, and processing code.