KurisutinaIdentity and access. Self-Knowledge of an Amnesic Patient: Toward a Neuropsychology of Personality and Social Psychology. Journal of Experimental Psychology: General 125(3), 250–260. Peer-reviewed single-case human study. DOI; author-hosted full text; author publication record. Accessed 2026-09-22. Local PDF, text, provenance.
Factual core
W.J., an eighteen-year-old student, developed retrograde amnesia after a concussion; CT revealed no abnormality. Five days later, she recalled no event from the preceding six months on twenty-four constrained autobiographical cues, although older episodes and general personal facts remained available. Three female memory controls recalled recent events on 96% of cues. Four weeks later, her episodic recall had substantially recovered. W.J. rated eighty traits describing her college self at both sessions: the cross-session correlation was .74, compared with .78 for two female personality controls. Her initial self-ratings correlated .65 with her boyfriend’s ratings, matching the control couples’ reported correlation. Some endorsements opposed social desirability. Her high-school self-rating correlated .53 with the later college rating; the authors interpreted the larger college retest correlation as evidence of context-specific retained knowledge. There was no preinjury trait assessment, objective behavioral criterion, or identified focal lesion. The authors acknowledged that older memories or episodic access triggered specifically by trait cues could contribute. Full methods and results, pp.252–256.
Original inference audit
A dissociation in accessible reports, with a bounded scope. The constrained recall condition usefully addresses a simple sampling alternative: failure to volunteer recent events cannot be explained solely by preferring older memories. However, failing these cues does not establish absence of every relevant trace or of all episodic contributions during trait judgment. The observed contrast supports the possibility that a person can access an established self-description when recent autobiographical recall is severely impaired. It does not establish that acquiring, updating, or using all self-knowledge is independent of episodes. Recovery concerns accessibility and does not show that a destroyed representation was rebuilt.
Reliability is not exact agreement or behavioral accuracy. A profile correlation measures ordering across trait items and can remain high despite systematic rating shifts. An r of .74 therefore does not justify the discussion’s stronger suggestion of no change at all. Agreement with a close informant is useful convergent evidence, but two people can share a flattering, culturally typical, or otherwise mistaken view. Endorsing some undesirable traits argues against an extreme rule of always choosing favorable answers; it does not quantify how much agreement exceeds a shared normative trait profile. The relevant missing comparison is person-specific agreement after accounting for item norms and comparison-person ratings. Nothing here measures whether the reported traits predict actual future behavior.
Case and measurement dependence matter. Eighty adjectives are repeated measurements of one person, not eighty independent clinical cases. Trait meanings and valence also overlap. Similar descriptive correlations to two controls do not establish statistical equivalence to a population. The .74-versus-.53 comparison shares both the same items and the later college ratings. A valid inferential comparison must handle that dependence; the reported t(158)=1.71, one-tailed, does not document how it was handled. Without the ratings and their full correlation structure, the claimed college-specific increment cannot be verified. This qualification does not erase the observed profiles.
Recovery is not a randomized manipulation. There was no preinjury baseline or reversal of condition order. The same questionnaire was repeated, and new learning was not globally abolished. Prior answers, continued experience, or evolving interpretations of the questionnaire could therefore contribute to retest stability. No consecutive case series or case-selection denominator is reported, limiting prevalence claims. A normal CT does not identify the affected neural process. Finally, the measured object is a semantic self-description: continuity of preferences, relationships, skills, motives, action policies, lived experience, or personal identity was not tested.
Original implication for extraction and transfer
Trait summaries should be treated as potentially useful information that can survive reduced access to their supporting episodes. They should not automatically be treated as either a complete person model or a substitute for episode records. A transferred summary might reproduce what someone says about themselves while missing when they behave differently, how they revise that belief, or what an event means to them.
A proposed recipient test would separate three targets: later self-description, independent informant judgment, and prospectively observed choices. Compare population item norms, a person’s trait summary, episodic records, and their combination under explicitly matched information budgets. For the incremental release comparison, hold the source’s elicitation history fixed; estimate any acquisition-policy effect separately. Freeze the model before new occasions and include occasions where competing interpretations predict different behavior. Distinctive person-level prediction should improve over matched generic and mismatched-person controls, with uncertainty clustered by person and attention to correlated items. No autobiographical loss needs to be induced, and no identity-continuity claim follows from succeeding at these limited targets. This is an original proposal, not a demonstrated transfer result.
Read audit and limits
Read the complete introduction, Methods, Results, Discussion, and Conclusions (pp.250–257), including the discussion’s wider neuropsychological examples. Visually checked Figure1, Figure2A–B, and Table1 against the retained scan. The references through p.260 were retained but not independently followed; the final-page editorial notice is not study content. The eleven-page author-hosted PDF is a publisher-formatted scan with an existing text layer; title/authors/journal/pages match the author record and Crossref DOI metadata. No supplement, complete item-level ratings, analysis code, or registration was identified on the checked primary sources. No graph digitization, participant sessions, or experiments were performed.