KurisutinaAlmut I. Weike, Alfons O. Hamm, Harald T. Schupp, Uwe Runge, Henry W. S. Schroeder, and Christof Kessler. Fear Conditioning following Unilateral Temporal Lobectomy: Dissociation of Conditioned Startle Potentiation and Autonomic Learning. Journal of Neuroscience 25(48), 11117–11124, 30 November 2005. Peer-reviewed primary human experiment. DOI: 10.1523/JNEUROSCI.2032-05.2005. Full primary text.
This is an explicit substitution for Bechara et al. (1995), DOI 10.1126/science.7652558. That paper's metadata were verified, but an intact publicly retrievable full text was not obtained: the indexed UTD PDF redirected to an unrelated institutional page; other available records did not supply a usable full paper. No substantive conclusion here is taken from Bechara's abstract or its retelling in Weike. This substitute tests response-system dissociation after mixed unilateral lesions; it does not reproduce Bechara's reciprocal lesion design.
Evidence from this source
Thirty patients with unilateral amygdala–hippocampal resections and 32 healthy controls underwent face–shock conditioning, followed by unreinforced trials. Most resections included anterior temporal tissue. Startle eyeblinks and skin conductance were recorded. Contingency knowledge was assessed afterward by interview and four-choice recognition: 9/30 patients and 24/32 controls identified the reinforced face. Skin-conductance discrimination varied with this classification; startle potentiation did not show a significant awareness interaction. Patients had reduced startle potentiation relative to intertrial baseline. Crucially, during acquisition they still discriminated reinforced versus unreinforced faces in startle responses, without a significant group interaction for that contrast. Thus, “impaired conditioning” depends on the endpoint. Authors favor distinct emotional and contingency-learning processes. Baseline and sensitization checks addressed some response-capacity alternatives. However, patients were older, had epilepsy, and took anticonvulsants; lesions were not amygdala-selective. Posttest recognition cannot determine awareness during learning. Only eight controls were classified unaware, limiting claims based on absent interactions. Full study.
Original interpretation and competing accounts
The useful lesson for functional memory extraction is to treat a person's account of a learned relationship, physiological reaction, and behavioral disposition as different targets. One cannot infer that obtaining the account has obtained every response shaped by the experience. Conversely, a physiological change does not reveal a unique hidden narrative that can be reconstructed from it. A memory representation may need several response functions linked to common cues, without assuming that all are governed by one latent “fear strength.”
The evidence is narrower than proof of learning without consciousness. An unsuccessful delayed report can reflect forgetting, failure to retrieve, uncertainty, or difficulty mapping the experience onto the response alternatives. Correct recognition can also occur by guessing. Without a validated contemporaneous measure, posttest classification cannot distinguish these possibilities. An absent interaction is not equivalence between aware and unaware participants. Correlations near zero also do not establish independent mechanisms when measurement reliability is unknown.
The lesion comparison adds causal relevance beyond imaging correlation, but it does not isolate acquisition, storage, retrieval, and response expression. A deficit in producing one conditioned output could occur despite retention of information that would support another output. The surviving discrimination contrast makes this especially important. Broad lesions and nonrandom clinical groups also leave several anatomical and clinical explanations available. Normal-range responses on control checks weaken a global inability-to-respond account without eliminating every performance contribution.
The paper does not measure a person's lifetime dispositions, avoidance choices, subjective fear, or transfer to an artificial system. Those are additional targets requiring their own evidence. In particular, predicting a physiological endpoint from a textual model and preserving a disposition that changes real choices are different achievements.
Proposed extraction and transfer test
For naturally occurring, consented everyday cues, acquire descriptions and repeated behavioral responses separately. Compare a narrative-only model with one that also receives cue-specific response histories. Match observation time and information budgets, include simple subject-specific baseline models, and test new cues and contexts. Keep explicit outcome expectations, reported experience, attention or response latency, and actual choices as separate outcomes. Physiological measures can be an additional channel when justified; they should not be assigned a psychological interpretation merely because they are involuntary.
For a controlled future study, use a benign learning task with known histories. Evaluate both response acquisition and change after revaluation or reversal; hold out some cue combinations. Compare spoken versus recognition-based knowledge probes, and counterbalance whether explicit probing occurs before or after behavioral measurement because probing can itself alter attention and learning. Preserving the pattern across these measures would support limited functional transfer. It would not show transfer of bodily experience, consciousness, or a complete emotional memory system. No participant or model sessions were run for this research note.
Reading and acquisition record
Read the complete introduction, materials and methods, results, discussion, Table 1, and captions for Figures 1–3 from the retained public full-text HTML. This includes the exact awareness questions, exclusions, response scoring/standardization, missing-value replacement, conditioning and postconditioning contrasts, and awareness subgroup analyses. References were scanned, not individually verified. No separate supplement was identified. Figures were not independently remeasured; raw recordings, analysis code, and individual participant records were not accessed.
The PMC HTML contains the full primary article and matching bibliographic metadata; it is not the browser-check page returned by some web requests. The publisher PDF returned HTTP 403, the PMC PDF endpoint returned HTML, and the Europe PMC fullTextXML endpoint returned HTTP 500; none was retained as a paper. The HTML states Society for Neuroscience copyright and does not identify a Creative Commons license. Public access does not by itself establish redistribution rights.
Local sources: full HTML, readable text, provenance and SHA-256 hashes. Accessed 22 September 2026.