Background
•Story-recall tests are among the most sensitive tools for detecting early memory decline in aging [1] and prodromal dementia [2]
•The most-validated instruments (Logical Memory, ADAS-Cog) carry per-subject licensing fees and distribution restrictions that exclude community clinics, low-resource settings, and international sites
•The Craft Story 21 (CS) became the non-proprietary standard in the NACC dataset but still needs developer permission, lacks automated scoring, and stays tied to the Alzheimer's research-center network [3]
•No existing free task combined open licensing, both manual and automated scoring, and validation across the normal-to-MCI range
Objective
•Test whether the Puppy Escape (PE), a freely available story-recall task, matches CS for detecting mild cognitive impairment (MCI)
•Derive and validate an 18-item, pen-and-paper short form (PE-18)
Methods
Participants: 388 adults from a community brain-health cohort:
153 cognitively normal (CN), 106 subjective cognitive impairment, 87 MCI
Procedure: Each participant completed both PE and CS on separate visits. PE scored automatically (transcription plus AI rubric scoring, 57 items); CS scored by trained examiners per the standard protocol
Analysis: Compared the two tasks on agreement, correlation with a 12-test neuropsychological battery, CN-vs-MCI discrimination, and added diagnostic value
Results
•PE and CS agreed moderately: r = .42–.45 (all p < .001)
•PE matched CS on 10 of 12 neuropsych assessments
•PE delayed separated CN from MCI at Cohen's d = 1.03, above the d = 1.0 benchmark; AUC = 0.776 vs CS AUC = 0.802, no sig. difference (DeLong p = .51)
•Automated scoring revealed selective MCI impairment of location, action, and name content (Group × Category interaction, p = .010)
•PE-18 retained the full test's accuracy (d = 1.02, imm.) using just 18 of 57 items
Discussion
•A free, openly licensed task matched the CS across every validation domain. Documents freely provided, requires simple agreement with University of Miami
•The pen-and-paper PE-18 needs no licensing fees, kits, or digitization, while automated scoring adds dimensional content that total-score tasks cannot resolve.
•Limitation: the short form was derived and validated in the same sample, so its accuracy is an upper bound pending independent replication.
•Next step: validate PE and PE-18 in an independent cohort and examine longitudinal decline.
