Long-term retention of real-world experiences in a patient with profound amnesia

Abstract

The medial temporal lobe (MTL) is known to be critical for healthy memory function, but patients with MTL damage can, under certain circumstances, demonstrate successful learning of novel information learned outside the laboratory. Here, we describe a patient, D.C., with extensive but focal bilateral MTL damage centering primarily on his hippocampus, whose memory for real-world experiences was assessed. Tests of remote memory indicated at least some capacity to retrieve specific details. To test his anterograde memory, he was taken on a tour of the NIH Clinical Center, with unique events occurring at each of ten specific locations. His memory for these events was tested after one hour, and again after fifteen months. Initially, D.C. could not recall having participated in the tour, even when cued with photographs of specific places he had visited. However, he achieved 90% accuracy on a forced choice recognition test of old and new objects he encountered on the tour, and his recognition of these objects remained intact over a year later when he was tested once again. Subsequent recognition memory tests using novel picture stimuli in a standard laboratory-style computer task resulted in chance-level performance across multiple test formats and stimulus categories. These findings suggest a potentially privileged role for natural learning for long-term retention in a patient with severely damaged medial temporal lobes.

1. INTRODUCTION

The medial temporal lobe (MTL) is known to support healthy memory function, and damage to structures within the MTL can produce profound memory impairments (Scoville & Milner, 1957; Squire & Zola-Morgan, 1991; Nadel & Moscovitch, 1997; Eichenbaum, 2000). Memory deficits following MTL damage are frequently identified and dissociated from other cognitive impairments using standard neuropsychology tests and computerized lab-based experiments, and these can effectively determine if deficits are selective or part of a broader decline in cognitive function.

However, studying memories acquired in the lab may not necessarily reflect patients’ capacity to learn in the real world. As has been discussed in the autobiographical memory literature (see e.g., Maguire, 2001; McDermott et al., 2009; Chen et al., 2017), real-life experiences are rich, multi-modal affairs that differ substantially from laboratory-based experiments and typical encoding/retrieval paradigms both in terms of task demands and the neural correlates with which they are associated. A logical question is therefore if (or how) memory-impaired patients might benefit from the “natural” encoding of everyday experiences because they engage additional neural circuitry outside the (damaged) MTL.

An example of natural learning in patients was observed in H.M., who suffered from dense amnesia, yet was able to learn the floorplan of a house he lived in after the onset of his amnesia following a long period of repeated natural encoding experiences (Corkin, 2002; but see also Bayley, 2005). It was hypothesized that the sensorimotor nature of encoding assisted his learning. Other researchers have measured natural learning in a more controlled environment, and following only a single exposure, that nevertheless occurred outside the laboratory. Dede et al. (2016) led hippocampal and MTL patients on a tour of the UC San Diego campus, testing their memory immediately after the walk. Although patients recollected fewer details of the tour than controls, they nevertheless retrieved and described accurate references to specific location details, perceptual details, or specific happenings (despite their hippocampal damage). An important question emphasized by the results of this experiment is therefore not what may be lost, but rather, what memory functions may remain relatively preserved in hippocampal amnesic patients under natural encoding conditions.

With this in mind, we present a newly described patient, D.C. A Caucasian male in his mid 50s, D.C. became profoundly amnesic as a result of secondary CNS lymphoma. The lymphoma resulted in severe damage to his hippocampus bilaterally, with some additional loss of tissue in surrounding MTL structures. The nature and location of D.C.’s MTL damage suggests that his case (including both his impaired memory abilities and those that are intact) may be of theoretical interest to students of memory. Here, we describe tests that probed D.C.’s ability to retrieve information acquired prior to the onset of his amnesia as well as his ability to encode and retain memories for information acquired following its onset. We ask to what extent memories acquired naturally, as well as in laboratory conditions, might be preserved and retrievable.

2. MATERIALS AND METHODS

2.1. PATIENT DESCRIPTION

The patient, to who we refer using the arbitrary initials D.C., is a right-handed Caucasian male who was 55 years old at the time of his initial participation. Prior to the onset of his amnesia, he had a stable job in the manufacturing sector. He attained a high school education but did not attend college. Consultations with family suggest that he had normal memory and cognitive function before he became amnesic. He had a history of drug/alcohol abuse but became sober in 2012 without any episodes of relapse.

In early 2019, D.C. was diagnosed with secondary CNS lymphoma, which initially responded to standard of care chemotherapy at a community hospital. Several months later, the CNS lymphoma recurred, presenting as large masses centered in the temporal horns and choroid plexus of the lateral ventricles bilaterally, with evidence of CSF dissemination. Despite further therapy, the masses continued to progress including extensive edema and invasion of the medial temporal lobe. In September of 2019, he presented to the NIH for enrollment into the TEDDI-R clinical trial for CNS lymphoma (Roschewski et al., 2018). At this time, progressive memory loss and agitation were observed. Following implantation of an Ommaya reservoir (Ratcheson & Ommaya, 1968) he underwent TEDDI-R treatment resulting in complete remission, which has been sustained as of Feb 2022 (Figure 1). Although the tumor resolved, residual gliosis and encephalomalacia were present at the sites of tumor involvement. Prior to his participation in this study, we obtained informed consent from D.C., and he participated under the NIH IRB approved Clinical Study Protocol 10-M-0047 (clinicaltrials.gov ID: NCT01087281).

Figure 1.

MRI imaging depicting course of lymphoma and end stage sequelae. Hyperintense signal on initial axial FLAIR imaging (A) represents a combination of lymphoma and surrounding vasogenic edema, initially progressing and evolving (B) to stable encephalomalacia and gliosis (C-E). The scan in panel B was collected just before beginning TEDDI-R treatment. The Clinical Center walkthrough described later in this report occurred ~5 months before the image in panel D was taken. R: right hemisphere.

2.2. VOLUMETRIC ASSESSMENT OF MEDIAL TEMPORAL LOBE REGIONS

The impact and extent of lesion damage within the MTL was assessed in the patient using a group of 20 age- and sex-matched controls taken from the HCP Aging cohort (Bookheimer et al., 2019). Controls had an average age of 55.8±2.9 years (range: 51.5 to 60.8 years). Control participants’ T1-weighted anatomical images were collected at 3T using a 3D MPRAGE sequence (TR = 2400 ms, TE = 2.14 ms, TI = 1000 ms, flip angle = 8°, 256 0.7 mm slice thickness, 320 × 320 matrix, 0.7 × 0.7 mm resolution in-plane), and were taken from the “minimally processed” HCP pipeline (Glasser et al., 2013; Harms et al., 2018), which included a downsampling step to 1 mm isotropic resolution (see Glasser et al., 2013). Whole-brain volumes for each control subject were estimated using the brainmask image generated using recon-all in Freesurfer v6.0.0.

D.C.’s structural scans were collected on a Siemens Magnetom 7T scanner (Siemens, Erlangen, Germany). Two high-resolution T1 scans were acquired (MP2RAGE: TR = 4300 ms, TE = 2.27 ms, TIs = 1000 and 3200 ms, flip angle = 4°, 192 0.75 mm thick slices, 320 × 300 matrix, 0.75 × 0.75 mm resolution in-plane), coregistered, and averaged to improve overall image quality. As with the control participants, D.C.’s whole-brain volume was estimated using the brainmask image generated using recon-all in Freesurfer v7.2.0. In addition, two high-resolution T2 scans were acquired and used to identify and exclude hypointensive voxels, corresponding to abnormal tissue from D.C.’s volume measurements (spcR: TR = 4090 ms, TE = 281 ms, flip angle = 120°, 288 0.67 mm thick slices, 636 × 488 matrix, 0.33 × 0.33 mm resolution in-plane). The T2 scans were coregistered and averaged with one another, and then to the T1 average scan.

Given his amnestic condition, specific medial temporal lobe structure volumes were of particular interest to compare between D.C. and the control participants. These included the amygdala, hippocampus, entorhinal cortex, perirhinal cortex, and parahippocampal cortex bilaterally. Each was manually identified following the methods of Insausti et al. (1998) for the amygdala, hippocampus, and rhinal regions, and Frankó et al. (2014) for the parahippocampal cortex. Spatial volumes for each region in each hemisphere were calculated by multiplying the voxel size by the number of voxels. The distribution of spatial volumes of each structure in the 20 control participants were used to generate volume z-scores for each structure in D.C. Structure size differences were considered significant if their corresponding p-values (two-tailed) survived FDR correction q < .05 (Benjamini & Hochberg, 1995). Although D.C.’s whole brain volume did not significantly differ from the control group’s (z = −0.59, p = .559), we conducted a follow-up analysis in which structure sizes were converted to a percentage of whole-brain volume in control participants and in D.C. Volume loss z-scores were recalculated and the results were once again FDR-corrected (q < .05)

2.3. NEUROPSYCHOLOGICAL ASSESSMENT

The cognitive impacts of D.C.’s lesions were evaluated using a standard battery of neuropsychology tests. The cognitive domains tested included verbal and visuospatial memory, visuospatial processing, language, attention, processing speed, and executive function, as well as estimated IQ.

2.4. TESTING REMOTE MEMORY: AUTOBIOGRAPHICAL INTERVIEW

In addition to the above neuropsychological tests, D.C. completed a variant of the Autobiographical Interview (Levine et al., 2002), using the memory detail categories described in Gilmore et al. (2021), which are more granular than those in the original Levine method. Testing occurred in November of 2022. The Autobiographical Interview tested D.C.’s ability to remember events from the recent and remote past, as well as to imagine events that might occur in his near or distant future (a related capacity also thought to rely on the hippocampus; see e.g., Addis & Schacter, 2012; Maguire & Mullally, 2013). The recent past encompassed the last year, whereas the remote past was when he was aged 25 or younger. For imagining future events, the near future included the next year, whereas the remote future event had to occur 10 or more years away. The event cues were the same as those used by Levine et al. (2002) and one event was described in each condition. D.C. freely and overtly described each event until he reached a natural stopping point, at which point additional prompts for specific details were provided as described by Levine et al. (2002). Verbal descriptions were recorded, transcribed, and scored for content. Words and phrases in each description were identified as “internal details” (which were event-specific and episodic, such as perceptual details associated with the event) or “external details” (including off-topic statements, non-specific statements, or repetitions). The total number of internal and external details was tabulated for each event description. To ensure accurate scoring, the rater (JMW) had been trained to a high level of inter-rater reliability (ICCs > .95 for both the internal and external detail categories) with one other lab member prior to seeing or scoring the events related to this report.

2.5. TESTING REMOTE MEMORY: CHILDHOOD HOME

D.C. was asked to draw, from his memory, the façade of his childhood home. Testing occurred approximately two years after his treatment at the NIH, and he has not visited the house during that time. As such, although the time since his last exposure is not known, it would be at least several years (likely longer, and necessarily before the onset of his amnesia). There was no specific time limit on the drawing and the patient determined when he completed the picture. Drawing was done with a pencil to allow the patient to make modifications. Specific details in the patient’s drawing were then compared to a recent photograph of the house. Family members stated that the exterior of the structure had not changed (barring a new coat of paint) since he lived there during childhood.

2.6. HOSPITAL WALKTHROUGH

D.C.’s ability to learn novel information during a natural experience was also assessed. He was taken on a walk through the first floor of the NIH Clinical Center in July of 2021. Two experimenters (AG, SA) took D.C. to 10 specific locations, with a specific and unique event occurring in each. The tour lasted approximately 20 minutes total. The stops, and what occurred in each location, were as follows:

1. The atrium café: asking the patient if he wanted a particular bag of chips.

2. Scale model of the Clinical Center: pointing out and discussing where the patient had stayed during his cancer treatment and where he had been earlier in his visit that day.

3. Presidential visits wall: discussing the photographs that captured moments particular U.S. presidents had visited the NIH campus.

4. The fish tank: watching, pointing to, and discussing the various fish in the tank.

5. The NIH Bookstore: looking at and chatting about coffee mugs that depicted amusing expressions.

6. Nobel laureate relief sculpture: asking the patient to describe the process of shaping and sculpting metal, with which he had prior experience.

7. A corridor: an experimenter dropped a clipboard and scattered a stack of papers.

8. A locked door: a research assistant unfamiliar to the patient delivered a book to one of the experimenters and explained that the next door down the hallway was locked.

9. The Lasker award wall: describing what the Lasker awards are and highlighting Anthony Fauci’s photograph as someone D.C. may have seen in the news recently.

10. The elevator: The experimenters and D.C. were joined in their elevator ride by a confederate wearing a tiger costume and holding a giant stuffed tiger.

2.7. HOSPITAL WALKTHROUGH MEMORY TEST

D.C.’s memory for the walkthrough was tested approximately 1 hour later. He was given 3 successive tests, each offering more retrieval support than the last. The first test used free recall, where the patient was given the prompt “Please tell me about the walk we took through the clinical center.” D.C. was allotted up to 6 minutes to verbally describe his experience, although he could stop earlier if he did not need this entire time. Following the Autobiographical Interview procedure (Levine et al., 2002), D.C. was given a general probe if he ended his description before the maximum time was reached (“Is there anything else you can tell me? Please tell me more about the walk.”), but this probe did not specify what kind of information should be described. Scoring was designed such that each specific (or “internal”) detail would be tabulated.

Free recall was followed by a cued recall test. In this test, the patient was provided with pictures of each visited location, was told it was a place he had stopped at earlier in the day, and was asked “Can you tell me about what happened here?” He had up to 1 minute to verbally describe what occurred at that location. Each picture remained visible for the duration of the recall period. The scoring procedure was identical to that used in free recall.

After the completion of all cued recall testing, a final testing phase began. This final test used a two-alternate forced choice procedure. The patient was shown 10 pairs of object photographs. In each pair, one was an object D.C. had interacted with or talked about on the tour, whereas the other was a related foil image, which was an object from the same or similar context in which the patient did not interact with on the tour (see Supplementary Table S1). D.C.’s task was to identify the image containing the object with which he had interacted. To help D.C. remain oriented to the task, a prompt was written above each set of images that reiterated the instruction. His score reflects the total number of correctly selected pictures.

2.8. DELAYED HOSPITAL WALKTHROUGH TESTING

In a session that occurred 15 months after D.C.’s walk through the clinical center, he was given the same tests, using the same stimuli and prompts, to determine his ability to retrieve information about the tour over a year after it occurred. Scoring was performed as described previously.

2.9. SCENE/OBJECT RECOGNITION TESTING

Additional tests of D.C.’s recognition memory ability were conducted using a more traditional approach that involved studying images and subsequently being tested with a mix of studied items and novel lures. Testing occurred in November of 2022. This testing took 4 distinct formats, which effectively crossed two types of stimuli with two types of recognition memory assessments. The two types of stimuli were pictures of scenes and pictures of objects, and the two types of tests were two-alternative forced choice and single-item old/new recognition. All object images and some scene images were taken from the Mnemonic Similarity Task github page curated by Craig Stark (https://github.com/celstark/MST). Additional scene images were identified and downloaded using Google images to match the total number of stimuli of each category. The tasks were all programmed and presented in Psychopy v2022.2.4 (Peirce, 2007) on a 15” Macbook Pro laptop (Apple, Cupertino, CA).

Irrespective of the type of stimulus presented or type of recognition test used, the encoding format followed the same overall design. First, 48 pictures were studied for 2 s each, with an ITI of 1 s during which a crosshair was centrally presented. The encoding task involved making a binary semantic judgment about each picture (larger or smaller than a shoebox for objects; indoor or outdoor for scenes). Responses were made on a keyboard using the 1 and 0 keys, and text at the bottom of the screen reminded D.C. which key corresponded to each judgment. To assist D.C. in remaining oriented to the task, he was presented with a question above each picture (for objects, “Would it fit in a shoebox?”; for scenes “Is it indoor or outdoor?”). The delay between study and recognition testing consisted only of the time it took to explain the recognition memory task instructions to the patient after the final encoding trial. Recognition instructions were written on the screen while the experimenter also spoke this information aloud.

Recognition trials were self-paced and were followed by a 1 s ITI. For the two-alternative forced choice format, 48 pairs of images were presented to the left and right of a central fixation cross. One of the images was always seen during the encoding phase. The other was novel and had not been viewed at encoding. Of these, half were visually similar to the encoded stimulus (“similar lures”), whereas the other half were visually distinct (“novel foils”). D.C. had to indicate, for each pair, which image he had seen previously. To keep him oriented, a basic question “Which did you see” was written at the top of the screen, and on the bottom the mapping of the key to the image was written to be visually consistent with other conditions. For the single item old/new test format, a single image was presented centrally. 24 previously studied stimuli, 24 novel foils, and 24 similar lures were pseudorandomly presented. The question at the top of the screen was “Did you see this today?” and along the bottom each specific response was on the side of the appropriate response key. Performance was calculated using a binominal probability distribution by comparing D.C.’s correct responses (hits + correct rejection of lures) to the total number of trials. Performance was considered above chance if the probability of that many or more successes was p < .05 and if the result survived FDR correction to account for the 4 separate tests being conducted (Benjamini & Hochberg, 1995).

Testing was separated by task type, with the two old/new tests preceding the two forced choice tests. Test types were separated by a period of approximately 20 minutes. In addition, stimulus orders were counterbalanced such that scene stimuli were used in the first of the old/new recognition tests but the second forced-choice recognition test.

2.10. SCENE/OBJECT RECOGNIITON TESTING CONTROL PARTICIPANTS

We additionally administered the same four recognition tests to healthy control participants on Pavlovia (Psychopy’s online testing platform; https://pavlovia.org/), recruited through Amazon’s Mechanical Turk. The task instructions and format were kept the same, except 6 catch trials per test format were inserted to help ensure quality performance. In these catch trials, participants were instructed to press a space bar when they saw a blue square instead of object or scene stimuli. The task was restricted to “Masters” participants (i.e. participants deemed high performing by Mechanical Turk), participants aged 45+, and to residents of the United States. Participants were compensated at a rate of $5/hour, and indicated informed consent via button press under the under the NIH IRB approved protocol number 000589.

Of the 118 number of controls that participated, 25 were excluded for failing to make the correct response on one or more of the catch trials, and 8 were excluded for failing to complete the task. Demographics information for the remaining 85 control participants spread across the four tests can be viewed in Table 1. Patient D.C.’s age and education was comparable to controls, never differing by more than 0.15 standard deviations in age or 1.73 standard deviations in education from controls on all four versions of the task.

Table 1.

Demographic information of control participants for each recognition test format

	N	Sex	Age (years)	Education (years)	Patient Age z-score	Patient Education z-score
2 AFC Scenes	25	11M/11F/3NR	56.92 ±6.99	15.08 ±1.78	−0.13	−1.73
2 AFC Objects	21	9M/12F	57.14 ±7.51	14.52 ±2.94	−0.15	−0.86
Old/New Scenes	21	8M/12F/1NR	56.33 ±6.81	14.62 ±1.96	−0.05	−1.34
Old/New Objects	18	6M/11F/1NR	56.33 ±7.24	14.78 ±3.08	−0.05	−0.90

Means are accompanied by ±1 standard deviation. 2 AFC: 2 alternative forced choice ; F: female, M: male, NR: not reported

We calculated D.C.’s z-score relative to mean control performance for correct (hits + correct rejection of lures) and incorrect (misses + false alarms to lures) responses, as well as separately for hits, false alarms, and misses, and compared his performance to that of controls for each type of response using t-tests computed using the Crawford method (Crawford & Garthwaite, 2012). A corrected recognition score (% hits - % False Alarms) was also calculated and compared to that of controls using the same method.

3. RESULTS

3.1. LESION LOCATION AND EXTENT

Manual volumetric quantification D.C.’s T1 images indicated extensive atrophy of the hippocampus bilaterally as well as surrounding structures (see Table 2 and Figure 2). When compared to an age- and gender-matched control sample, D.C. displayed significant reductions in the hippocampus bilaterally (ps <.001), in parahippocampal cortex bilaterally (ps < .001), and in right entorhinal cortex (p = .01), all of which remained significant following FDR correction. Differences in the size of right perirhinal cortex approached, but were not, significant (p = .05) and did not survive FDR correction. Other structure size differences did not approach significance. Adjusting for total brain volume did not alter this pattern of significance.

Table 2.

Volume estimation of MTL structures in patient D.C. and matched controls.

ANATOMICAL REGION	PATIENT VOLUME (mm3)	CONTROL VOLUME (AVG) (mm3 ±SD)	PATIENT Z-SCORE (raw vol)	PATIENT Z-SCORE (vol adjusted)	ESTIMATED VOLUME LOSS (raw vol)	ESTIMATED VOLUME LOSS (vol adjusted)
Left hippocampus	1323	3729 (±389.1)	−6.19*	−6.16*	−65%	−63%
Right hippocampus	1322	3911 (±437.5)	−5.92*	−5.95*	−66%	−65%
Left amygdala	1385	1741 (±287)	−1.24	−0.98	−20%	−17%
Right amygdala	1197	1655 (±335.6)	−1.36	−1.09	−28%	−25%
Left ERC	1516	2094 (±499.1)	−1.16	−1.04	−28%	−24%
Right ERC	1126	2194 (±418)	−2.55*	−2.35*	−49%	−46%
Left PRC	3073	2631 (±684.5)	0.65	0.75	+17%	+21%
Right PRC	1286	2535 (±637.7)	−1.96	−1.73	−49%	−47%
Left PHC	670	1816 (±296.8)	−3.86*	−4.41*	−63%	−61%
Right PHC	786	1531 (±224.7)	−3.32*	−3.01*	−49%	−46%
Whole brain	1,597,493	1,662,080 (±110,655)	−0.58	n/a	−4%	n/a

Notes:

^*indicates significance after FDR correction; ERC = entorhinal cortex; PRC = perirhinal cortex; PHC = parahippocampal cortex; vol = volume.

Figure 2.

High resolution T1 imaging of D.C.’s medial temporal lobe (MTL). Left column: Sagittal views centering on the right hemisphere’s hippocampal long axis. The top and bottom views are identical, with manual tracings of MTL regions visualized on the lower image. Orange lines represent slices corresponding to the locations of the coronal views on the right columns. Middle column: Lesions extend bilaterally from entorhinal cortex and the amygdala anteriorly to the parahippocampal cortex. Right column: MTL structure masks overlaid on the same T1 slices. R: right hemisphere. Hypointense voxels corresponding to abnormal tissue identified on D.C.’s T2-weighted image were excluded from the MTL masks. Outside the MTL, MR signal abnormalities were also identified in left periventricular white matter near the centrum semiovale (Supplementary Figure S1). Excess mineralization was present in the basal ganglia, but no atrophy or gliosis.

3.2. NEUROPSYCHOLOGICAL TESTING

D.C. presents as a cognitively intact individual, with a friendly and chatty demeanor. The content of his natural topics of conversation is rooted in the past, prior to the onset of his amnesia. For example, he often speaks of returning to his prior place of employment or of a previous romantic partner. During testing at the Clinical Center, he routinely asked the experimenters about the status of his medical condition and when he could return home, despite having been discharged more than 12 months prior. Although he can hold a conversation easily, he repeats himself frequently. This is presumably due to poor memory for past topics after the topic has changed. He did not remember the experimenters after an hour-long lunch break, or on subsequent test sessions. He is aware of having difficulties with his memory.

Standardized neuropsychological tests indicated that D.C. is typical with respect to IQ, reasoning, working memory, word fluency, processing speed, and language (Table 3). In contrast, essentially all measures of episodic memory indicated impairment, with D.C. often falling several standard deviations below average for both recall and recognition test formats.

Table 3.

D.C.’s neuropsychological performance

COGNITIVE DOMAIN	TEST NAME	RAW SCORE	NORMATIVE SCORE	QUALITATIVE DESCRIPTOR1
Intellectual Functioning	Test of Premorbid Functioning (TOPF)	59/70	116e	High average
	WASI-II Block Design	50	58f	Average
	WASI-II Vocabulary	40	51f	Average
	WASI-II Matrix Reasoning	14	40f	Average
	WASI-II Similarities	28	43f	Average
	WASI-II Verbal Comprehension Index	--	95e	Average
	WASI-II Perceptual Reasoning Index	--	98e	Average
	WASI-II (4 subtest) Full-Scale IQ	--	96e	Average
Attention	WMS-III Digit Span	16/30	9g	Average
Language	Boston Naming Test	59/60	64b,h	Average
	COWA Letter (FAS) Fluency	26	39h	Low Average
	COWA Semantic (Animals) Fluency	13	33h	Below Average
Visuospatial	RCF Copya	33/36	>16thk	Average
	RCF Copy Timea	233”	>16thk	Average
Verbal Memory	WMS-IV Logical Memory I Recalla	18/50	7g	Low Average
	WMS-IV Logical Memory II Recalla	0/50	1g	Exceptionally Low
	WMS-IV Logical Memory Recognitiona	17/30	≤2%k	Exceptionally low
	HVLT-R Total Recall	14/36	≤20f	Exceptionally Low
	HVLT-R Delayed Recall	0/12	≤20f	Exceptionally Low
	HVLT-R Percent Retained	0%	≤20f	Exceptionally Low
	HVLT-R Recognition Discrimination	−3	≤20f	Exceptionally Low
Visual Memory	WMS-III Faces Ia	30/48	7g	Low Average
	WMS-III Faces IIa	26/48	5g	Below Average
	BVMT-R Total Recall	8/36	20f	Exceptionally Low
	BVMT-R Delayed Recall	0/12	<20f	Exceptionally Low
	BVMT-R Percent Retained	0	<1stk	Exceptionally Low
	BVMT-R Recognition Discrimination	−1	<1stc,k	Exceptionally Low
	RCF Short Delay Recalla	1/36	<1stk	Exceptionally Low
	RCF Long Delay Recalla	0/36	<1stk	Exceptionally Low
	RCF Recognitiona	14/24	<1stk	Exceptionally Low
Information Processing/ Psychomotor Speed	Trail Making Test A	39”	40d,h	Low Average
	Symbol Digit Modalities Test	35/110	−1.28d,i	Low Average
	Grooved Pegboard - Dominant Hand	80	43d,h	Average
	Grooved Pegboard - Nondominant Hand	100	38h	Low Average
	WCST-64: Perseverative errors	42	<20j	Exceptionally Low
Executive Functioning	WCST-64: Categories completed	0	2nd-5thk	Below Average
	Trail Making Test B	106”	38d,h	Low Average
Mood	Beck Depression Inventory - 2	11/63	--	Minimal
Anxiety	Beck Anxiety Inventory	4/63	--	Minimal

Notes: Abbreviations: BVMT-R=Brief Visual Memory Test - Revised; COWA=Controlled Oral Word Association; HVLT-R=Hopkins Verbal Learning Test - Revised; RCF=Rey-Osterrieth Complex Figure; WASI-II=Wechsler Abbreviated Scales of Intelligence - Second Edition; WCST-64=Wisconsin Card Sorting Test (64-card version); WMS-III=Wechsler Memory Scale - Third Edition; WMS-IV=Wechsler Memory Scale - Fourth Edition. Test author/publisher norms were employed to derive normative scores with exception of using Heaton, Miller, Taylor, and Grant (2004) norms for Boston Naming Test, COWA, Trailmaking Test, and Grooved Pegboard. Normative values for the RCF were derived from Meyers and Meyers (1995).

^aindicates that test was administered on a later occasion than remaining tests of Table 3.

^bindicates that D.C. declined on a subsequent administration of that measure.

^cD.C.’s BVMT-R Recognition Discrimination performance improved on a subsequent administration but remained low at an 11–16^th percentile ranking.

^dindicates that D.C. improved on a subsequent administration of that measure, but remained within qualitative average rankings on each test.

^eStandard score (M=100, SD=15), calibrated for age.

^fT-score (M=50, SD=10), calibrated for age.

^gScaled score (M=10, SD=3), calibrated for age.

^hT-score (M=50, SD=10), calibrated for age, education, gender, and race.

ⁱZ-score (M=0, SD=1), calibrated for age and education.

^jT-score (M=50, SD=10), calibrated for age and education.

^kPercentile score.

¹Guilmette et al. (2020) qualitative descriptors for performance based tests were used. See Supplementary Table 3 for normative test scores for each neuropsychology test collected at each point of testing.

The Rey-Osterrieth Complex Figure task, which involves direct reproduction without a memory component as well as short and long memory delays, is a single task that efficiently summarizes D.C.’s memory impairment (Figure 3A). His initial copy time was normal, and he generated a detailed reproduction when the image was in front of him. However, after a short delay, the patient generated a rather prototypical brick-and-mortar style building, which he accompanied with the unprompted comment, “Was it a house?” After a 25-minute delay, the patient refused to draw anything at all, instead stating that he had “no clue” regarding what was presented earlier.

Figure 3.

Examples of rapid memory loss and long-term retention. A) The Rey-Osterrieth Complex Figure task revealed that the patient could easily copy a complex image but quickly lost his ability to retrieve this information from memory. After a short delay he produced an incorrect image, expressing uncertainty as he drew it, and by a longer delay he was unable to produce any image from memory. B) In contrast, the patient was able to draw details of a childhood home (see Methods for additional details). These include features such as the presence and shape of a 1-story addition to the house (circle 1); the presence and number of a small 2-panel basement window (circle 2); the number of glass panels and location of the handle on the main entry door (circles 3 & 4); and the number of steps leading up to the main door (circle 5). A grayscale photograph of the home marking these same features is present to the right of the drawing.

3.3. FAÇADE OF CHILDHOOD HOME

In a separate testing session, D.C. was asked to reproduce, from memory, the façade of his childhood home. Comparison of this image to a photograph of the home suggests that he retained some precise visual details of this home (Figure 3B). These included an addition on the left side that was only one story in height, a double-panel basement window at the front of the house, a 6-panel window on top of the front door (somewhat obscured in the photograph), the placement of the knob on the left side of the front door, and the 3 steps leading from the ground to the main level. He also included more general details such as clapboard-style siding and the correct number of stories. Other features are absent (e.g., the missing first story roofline; the shape of the main roof), or distorted (e.g., the specific configuration of first and second floor windows). Although not evident in the drawing, D.C. subsequently volunteered an anecdote from his teenage years in which he described scraping down and repainting the entirety of the house over the course of a summer. He went on to specify the color of the paint—a light tan that he referred to as “buff”. The anecdote, the color of the house, and the name of the color were independently corroborated by a member of D.C.’s family.

3.4. AUTOBIOGRAPHICAL INTERVIEW

D.C.’s episodic memory capacity was also tested using a variant of the Autobiographical Interview procedure (Levine et al., 2002) that involved verbal descriptions of recent and remote past or hypothetical future events. Critically, descriptions of when and where events occurred were corroborated by a member of D.C.’s family. Memory descriptions were scored for content and counts of Internal Details (i.e., those that appear specific to the episode being described) and External Details (nonspecific, general, or off-topic statements) were tabulated. Content scores suggest that D.C. could retrieve some specific (Internal) details of remote, but not recent, memories (Figure 4). However, his descriptions of both recent and remote memories included numerically more External details. Regarding hypothetical future episodes, D.C. could generate Internal details for a hypothetical near-future experience but was unable to think of a specific event that might occur in the distant future. As with past events, more External than Internal details accompanied his future event descriptions.

Figure 4.

Results of D.C.’s verbal descriptions of past and future events on an adapted Autobiographical Interview. Left: D.C. was able to generate descriptions of remote past, but not recent past, events, that contained specific, episodic (i.e., Internal) details. In contrast, he provided Internal details for near but not far future hypothetical events. Tabulation of External details, which describe semantic statements, digressions, etc., exceeded those for Internal details in all conditions.

3.5. HOSPITAL WALKTHROUGH

Having characterized his memory of events prior to the onset of his amnesia, D.C.’s ability to encode and retain novel information was also tested. He was taken on a walk through the 1^st floor of NIH Clinical Center. The walkthrough consisted of 10 stops (Figure 5A) at specific locations. Each location involved interaction with, or discussion about, a central object, such as the fish in a large, prominent fish tank or a dropped stack of papers in the hallway. The patient remained engaged for the duration of the walkthrough and attended to the events at each location. Of note, one of the stops on the tour was the wall of portraits of Lasker Awardees. The wall includes over 30 photographs, nearly all of which depict middle-aged or older Caucasian males, and one of which is Anthony Fauci. The experimenters had planned to point out Dr. Fauci’s portrait, specifically, as someone he may have seen in the news during the COVID-19 pandemic (i.e., after the onset of his amnesia). Before being prompted by the experimenters, D.C.—who spends lots of his waking hours watching TV—walked straight to the portrait of Dr. Fauci and said, “Hey, I know that guy. I think he’s my doctor!” It bears noting that Dr. Fauci was not involved in D.C.’s care, and D.C. would not have encountered Dr. Fauci outside of the COVID-19 media coverage.

Figure 5.

Summary of D.C.’s walk through the NIH Clinical Center. A) Map of the route taken on the 20-minute walk. The path and order of the stops is indicated by red lines and yellow markers, respectively. B) Sample locations from the walk (fish tank and a specific hallway), which were also used as photograph cues to assist in recall. C) Object pairs from the 2-alternative forced-choice recognition test. Each row depicts the target of an interaction on the walk (real fish on the top row, a stack of papers and a clipboard on the bottom) as well as a lure. D) Performance on Free Recall and Cued Recall tests. D.C. was unable to recall a specific detail of his experience in either condition, either later in the same day or after a 15-month delay. E) Recognition memory performance, on the other hand, was significantly above chance irrespective of the delay between the walkthrough and the test period.

His memory for the hospital walkthrough was tested an hour after it ended, first using free recall, then a cued recall test with pictures of each location (Figure 5B), and finally a 2-alternative forced choice recognition test of the central object or objects in each scene (each accompanied by a related lure image; Figure 5C). The patient was unable to recount any specific details of his walkthrough in either the free recall or cued recall conditions (Figure 5D), including highly salient events such as the meeting with a confederate wearing a tiger costume. In contrast, he correctly selected 9 of the 10 objects with which he had interacted during forced choice recognition testing (Figure 5E), which is more than would be expected by chance (binominal test p = .021, two-tailed). When asked, the patient could not provide clear reasons for his selections, indicating that he did not have awareness of his memory despite his performance being well above chance.

The same test of memory for the hospital walkthrough was administered 15 months later in a separate visit. This allowed us to characterize his memory abilities across a long, rather than within-session, delay. As in the initial testing, D.C. was unable to recall any specific details about the walk (or, indeed, that he had gone on such a walk at all). Strikingly, his recognition performance remained significantly above chance (10/10 correct responses, binomial test p = .002, two-tailed). Collectively, these results suggest that long-term retention of an experience that occurred after his MTL damage is possible in D.C., at least under very specific encoding and retrieval conditions.

3.6. SCENE/OBJECT RECOGNITION TESTING

D.C.’s walkthrough results raised the possibility that he had intact recognition that was somehow missed in earlier neuropsychological testing. Alternatively, it was possible that his performance may instead be attributable to a more specific factor such as the type of stimuli used in the recognition testing (object images) or the forced-choice nature of the testing itself. To address these possibilities, D.C. was given additional recognition tests on a lab computer. In separate tests, he studied pictures of objects or of scenes under deep encoding conditions (size judgments for objects, indoor/outdoor judgments for scenes; Figure 6A) and was tested either using 2-alternative forced choice or single-item old/new recognition (Figure 6B). This produced 4 distinct testing conditions (forced choice objects, old/new objects, forced choice scenes, and old/new scenes) that, collectively, could address possible explanations of D.C.’s recognition performance following his walkthrough. We additionally collected control data from healthy age- and education-matched participants completing the same recognition tasks online on Amazon’s Mechanical Turk.

Figure 6.

Additional recognition memory tests performed by D.C. A) In separate tasks, D.C. studied pictures of objects and of scenes. He made semantic judgments for each type of stimulus at encoding. B) Following the encoding period, D.C. was tested under either 2-alternative forced-choice (2 AFC) or single-item old/new recognition memory test conditions. Instructions were always present on the screen to keep D.C. oriented to the task. C) D.C.’s performance in all 4 test conditions was compared to controls. He performed at chance level and significantly below the level of controls for all tests except for the Old/New Scenes test, where is performance was borderline.

Overall, D.C. did not exceed chance levels of performance (Figure 6C). For the forced choice objects test, D.C. correctly selected the studied item 58% of the time (binomial test p = .312, two-tailed). In contrast, controls correctly selected the studied object 89.48% of the time. D.C. selected the correct response significantly less often than controls (z = 4.81, t(20) = −4.7, p = 0.0001). Breaking down his incorrect responses according to lure type, D.C. false alarmed to new foils significantly more often than controls (z = 9.37, t(20) = 9.15, p < 0.0001), and also exhibited a tendency to select related lures more often than controls (z = 1.80, t(20) = 1.77, p = .093). His corrected recognition score (% hits - % false alarms) was significantly lower than controls (z = −4.89, t(20) = −4.77, p = 0.0001).

D.C. was also at chance for the old/new objects test variant, where he was correct exactly 50% of the time (binomial test p = 1, two-tailed), whereas controls were correct 76.16% of the time. Again, D.C. responded correctly significantly less often (z = 4.44, t(17) = −8.33, p < 0.0001) than controls. Breaking trials down by type, he correctly identified target items (hits) less frequently (z = 4.32, t(17) = −4.21, p = 0.0005), correctly rejected novel foils less frequently (z = 19.11, t(17) = −18.6, p < 0.0001), but did not differ in his rejection of related lures (z = 0.12, t(17) = 0.12, p = .91). With respect to incorrect trials, he missed targets significantly more often than controls (z = 4.31, t(17) = 4.2, p = 0.0006) and false alarmed to novel foils significantly more often (z = 20.98, t(17) = 20.42, p < 0.0001), but false alarmed similarly to controls (z = −0.12, t(17) = −0.12, p = 0.91) for related lures. Overall, his corrected recognition score was significantly lower than controls (z = −4.83, t(17) = −04.7, p = 0.0002).

For forced choice scenes, D.C. correctly responded to 44% of presented stimuli (binomial test p = .471, two-tailed), which is not significantly below chance. Controls correctly responded to 86% of the presented scenes. D.C. correctly responded significantly less often than did controls (z = 5.86, t(24) = −5.75, p < .0001). When considering errors by lure type, D.C. false alarmed significantly more than controls to novel foils (z = 9.14, t(24) = 8.96, p < .0001) as well as to related lures (z = 3.66, t(24) = 3.59, p = .001). Accordingly, his corrected recognition score was significantly lower than controls (z = −6, t(24) = −5.88, p < 0.0001). As years of education of the control group for this version of recognition testing was on average >1.5 standard deviations above that of D.C.’s, we repeated the above tests excluding control participants with greater than 15 years of education (n = 9 controls remaining, mean education = 13 ± 1.11 years, D.C.’s education was −0.89 standard deviations below this). D.C. performed significantly worse than these tightly matched controls on all metrics, and thus his education level was not driving the observed severe memory deficits (hits: t(8) = −4.73, p = 0.001, false alarms to novel foils: t(8) = 7.25, p < 0.0001, false alarms to related foils: t(8) = 2.91, p = 0.02, hits-total false alarms: t(8) = −4.82, p = 0.001).

For old/new scene judgments, D.C. was correct 63% of the time. Although this, in isolation, would be above chance (binomial p = .0443, two-tailed), the result did not survive FDR correction if one corrects for all 4 types of recognition test taken by D.C. Controls were correct 75.20% of the time. D.C.’s total correct responses were significantly reduced compared to controls (z = 2.26, t(20) = −2.65, p = 0.015). Focusing first on correct responses, he did not differ from controls in his identification of targets (z = −0.61, t(20) = −0.6, p = 0.56) or rejection of related lures (z = 0.11, t(20) = −0.11, p = 0.91) but rejected novel foils significantly less often (z = 4.38, t(20) = −4.27, p = 0.0004). Regarding error trials, D.C. did not miss targets significantly more frequently than controls (z = 0.61, t(20) = 0.6, p = 0.56) and had a similar rate of false alarming to related lures (z = 0.11, t(20) = 0.11, p = 0.91), but false alarmed to unrelated foils significantly more often (z = 4.36, t(20) = 4.26, p = 0.0004). His corrected recognition rate was marginally lower than controls (z = −1.88, t(20)=−1.84, p = 0.08).

We note here that D.C.’s comparable performance to controls on similar lure trials across tests was likely due to poor performance on these trials in the control group rather than intact functioning in D.C. Indeed, as can be viewed in Supplementary Figure 2, his responses were approximately equal across conditions, consistent with chance responding. D.C.’s performance was far below that of controls on all tests except for the Old/New Scenes test where his overall performance was marginal. Thus, there was no strong evidence when using standard recognition memory paradigms that D.C.’s walkthrough performance could be attributable to a general sparing of recognition, the type of stimuli used in testing, or the test format.

4. DISCUSSION

D.C. is an amnesic patient with bilateral lesions to his medial temporal lobes, especially affecting the hippocampi, resulting from secondary CNS lymphoma. His damage has resulted in profound deficits in his memory abilities with a general sparing of other cognitive domains. However, despite his amnesic status, some aspects of his memory were relatively spared, including an ability to recover certain details from remote memories and an ability to recognize, but not recall, aspects of a walk he took through the NIH Clinical Center after becoming amnesic. Notably, his recognition performance could not be replicated using more standard laboratory-based memory tasks and was not captured by standard neuropsychological testing. D.C.’s performance raises theoretically important questions about the cause of his relatively spared memory abilities under certain circumstances.

4.1. A RARE CASE OF EXTENSIVE BILATERAL MEDIAL TEMPORAL LOBE AMNESIA DUE TO SECONDARY CNS LYMPHOMA

Secondary CNS lymphomas are often localized to the cranial nerves, corpus callosum, or areas near the ventricles, and recurrent cases are typically fatal (Haldorsen et al., 2011; Malikova et al., 2022). The early presentation within our patient prior to being treated at the NIH is not atypical, and initial radiological reports highlighted the spread of cancer to these areas. However, to the best of our knowledge, this is the first report of secondary CNS lymphoma producing extensive bilateral lesions to the hippocampus and surrounding medial temporal lobe structures. This has resulted in a fairly specific collection of deficits resulting in a “classic” amnestic presentation consisting of remarkable memory impairment and general sparing of other cognitive domains. That said, differences in patient etiology have been discussed as a possible source of diverging results in past work (e.g., Kim et al., 2015; Maguire et al., 2016; for related discussion in non-human primates, see Málková et al., 1997), and further characterization of this patient, with the unusual source of his amnesia, may speak to this broader issue.

4.2. LONG-TERM MEMORY FOR NATURAL EXPERIENCES

A striking feature of D.C.’s case is his intact recognition memory for information learned naturally. D.C. scored nearly perfectly on a forced choice recognition test for objects related to his walk through the Clinical Center, both on the day of the walkthrough and over a year later. He also, unprompted, recognized a portrait of Anthony Fauci among a display of over 30 portraits during the tour, despite incorrectly identifying the source of his recognition. Importantly, D.C. would not have encountered Dr. Fauci prior to the onset of his amnesia, as his frequent appearance on national television during the COVID-19 pandemic only occurred after the onset of his profound memory impairment. Thus, despite his amnesia, D.C. seems able to learn, and subsequently access, at least some new information in a long-term capacity. At the same time, his explicit recollection was severely impaired. D.C. was unable to recall—under either free or photo-cued conditions—any aspect of his walkthrough, including highly unexpected or novel events, such as a shared elevator ride with someone wearing a tiger costume and holding a stuffed tiger.

This may, on the surface, appear consistent with cases of preserved familiarity coinciding with impaired recollection with hippocampal damage (e.g., Addante et al., 2012). Such dissociations have been observed in aging or various patient groups previously (see e.g., Yonelinas, 2002; Koen & Yonelinas, 2014), and functional imaging suggests that the neural substrates supporting these different types of memory appear generally distinct such that recollection is supported by the hippocampus and familiarity by extrahippocampal structures (e.g., Yonelinas et al., 2005; Wagner et al., 2005; Vilberg & Rugg, 2008; but see Wais et al., 2006; Wixted et al., 2010 for discussions focusing on the hippocampus specifically). Computational modelling and empirical data further suggest that the forced-choice test format, in particular, can be supported by extra-hippocampal cortex, while the degree to which old/new test formats are sensitive to the hippocampus versus extra-hippocampal cortex depends on the similarity of lures employed (O’Reilly & Rudy, 2000; Holdstock et al., 2002; Norman & O’Reilly, 2003). Thus, while we cannot rule out that residual hippocampal tissue could support observed forced-choice recognition functioning in D.C., we posit his performance is more likely supported by extrahippocampal cortex. Hippocampal volume estimates in D.C. indicated a >60% reduction, and it is notable that volume reduction estimates based on MRI may understate the extent of hippocampal damage (as noted by Gold & Squire, 2005, MR-estimated hippocampal volume reductions exceeding 40% likely reflect a “nearly complete” loss of neurons in the hippocampus (p. 84)). Indeed, while we excluded hypointense tissue on T2-weighted imaging that was clearly abnormal, much of the remaining tissue included in our volume estimates, while of normal signal intensity, appeared structurally abnormal and fragmented (see Fig 3, slices 5–8). Nonetheless, the interpretation that D.C. has relatively preserved familiarity is somewhat complicated by the fact that D.C. performed at chance level under typical recognition testing conditions with neuropsychological tests and in-lab computerized tests, across multiple stimulus categories (objects/scenes) and test formats (old/new vs forced choice). The exception was his performance on the Old/New Scenes test, for which his performance was borderline. A detailed follow-up investigation would be required to determine if this finding occurred by chance or if D.C. truly is able to recognize scenes to some extent in an Old/New format. We note here that other medial temporal lobe structures, most notably the parahippocampal cortex, were also compromised, albeit to a lesser extent than his hippocampi (Table 2).

One possible explanation for his preserved recognition performance for objects encountered on the walkthrough may be that he was not basing his selection decision on familiarity, but rather on some other criteria such as personal preference. Although this cannot be ruled out, there does not appear to be strong a priori reason for this to explain the current results. For example, it is not clear why, when presented with a “target” image of a pile of papers on the ground or a “foil” image of a disposable plastic water bottle in the same location, he would arbitrarily prefer the former over the latter. Similarly, mere exposure resulting in greater preference for an image is unlikely to explain D.C.’s behavior at the 15-month delay test because he had, at that point, seen all targets and all lures before (i.e., during the initial forced choice testing session), yet remained highly accurate.

An alternative explanation is that the location photographs used during cued recall testing of the walkthrough could have helped D.C. to select the correct object through deductive reasoning. For example, seeing a picture of the café (location #1) would have prompted him to select the correct object (a bag of chips) over the lure object (a “chewy marshmallow” bar) when he began his recognition memory testing several minutes later. This seems unlikely, as the lures were chosen to be related to the target to prevent utilization of this strategy. As both the target and foil were equally congruous with the context, the context was typically not a helpful cue. Furthermore, we were careful to ensure that the target objects were not visible in the photographs used as cues during retrieval. Finally, D.C. would have had to encode a representation of a specific spatial context (e.g., the café) presented during cued recall period, reflect upon it at the time of forced-choice recognition testing, and use the retrieved information to guide his decision for each recognition test question. This, too, appears an unlikely explanation, because the evidence from the cued recall period itself suggests that D.C. cannot consciously recollect specific aspects of prior experiences when presented with photographic cues.

Another explanation may be that the walkthrough involved the encoding of a relatively small number of distinct events. This possibility could account for D.C.’s poor performance on the standard recognition tests in the laboratory, as they required the encoding of dozens of stimuli that were presented in rapid succession. These tests may therefore have exceeded whatever residual encoding capacity remains in D.C. (for related discussion, see Dede et al., 2016). However, stimulus counts alone seem insufficient to explain D.C.’s intact recognition performance for walkthrough objects, because he was also impaired on neuropsychological tests of recognition memory which have, in many cases, far fewer trials than were present in our follow up testing (e.g., 12 studied items and 12 lures in the HVLT, and 6 studied items and 6 lures in the BVMT). At the same time, the fact that D.C. encoded and retrieved the material for both old/new tests in succession (and ~20 minutes later, both forced choice tests), may have impacted his observed performance relative to controls who only completed one recognition test per session. However, given that his recognition performance was impaired on all tests, including the old/new scenes test, which was the first recognition test given to D.C. during the session, and the forced-choice objects test, which was the first conducted after a break, the data appear to favor concluding that D.C.’s recognition memory performance for standard laboratory materials is impaired.

It is therefore possible that D.C.’s success on the walkthrough recognition task at least partially results from a combination of factors afforded by learning in a natural environment. Specifically, he could have benefited from encoding a relatively small number of salient experiences, in a multimodal fashion, with a relatively extended temporal spacing between each experience to minimize interference and competition for internal resources. In addition, socially enriched, “deep” encoding conditions unique to each event that resulted from conversations with experimenters during the walk may also have contributed. Indeed, patient H.M., who presented a fairly analogous case to the current patient in terms of lesion location and behavioral impairment, was able to learn the floorplan of a house he lived in after the onset of his amnesia. It was suggested that this was a case of intact personal semantic memory for such information, aided by learning in a multimodal fashion (i.e. with locomotion cues) slowly over an extended period of time, and with much repetition, all of which could theoretically leverage intact neocortical structures to scaffold successful encoding (Corkin, 2002). H.M. was reported to learn the novel floorplan only after numerous exposures, and other patients were unable to produce floorplans of locations encountered after the onset of their amnesia (Bayley, 2005). The current case clearly differs from these prior reports in the number of times exposed to the target material. Systematically investigating the different aspects of D.C.’s experience in future experiments (e.g., amount of social interaction available) will be critically important in understanding the nature of D.C.’s intact performance for his walkthrough recognition testing.

Given the similarity of the walkthrough D.C. took to the campus tours conducted by Dede et al. (2016) in a separate group of amnesic patients, convergent and divergent findings between the two studies should be considered. Dede et al. tested four amnesic patients with average bilateral hippocampal volume reductions ranging from 35%–49%, which was restricted to the hippocampus. Of note, hippocampal volume loss in D.C. was more substantial (>60%, with the specific value differing based on hemisphere and adjusted for total brain volume) and was accompanied by volume reductions in nearby MTL regions (Table 2). The patients in Dede et al. could recall at least some details from their real-world experience, whereas D.C. was unable to do so under either free or cued recall conditions. However, Dede et al.’s patients were impaired when compared to controls under the same testing conditions, and while the severity of D.C.’s amnesia may differentiate him in one sense, both results converge on difficulties in describing recent natural experiences. The patients of Dede et al. were impaired compared to controls under a matched delay in their forced choice testing, but like D.C., were nevertheless generally accurate in their responses. Based on their results, Dede et al. suggested that “global” information about an event will be impaired in hippocampal patients, but “local” information about individual event details may be residually encoded. The use of photographs in the current experiment as forced choice recognition cues could certainly be thought to query local, rather than global, aspects of D.C.’s walkthrough. Under this view, our current results converge with those of Dede et al., while extending them to a delay of over 1 year.

4.3. REMOTE MEMORY CAPACITY AND IMPLICATIONS FOR CURRENT THEORY

It is perhaps worth considering D.C.’s memory capacity in light of competing theories of hippocampal function to memory consolidation. It has been argued that the hippocampus is required for precise memory retrieval, such that that memories accessed without hippocampal involvement will necessarily be “schematic” and lack detail (Nadel & Moscovitch, 1997; Moscovitch et al., 2005; Winocur et al., 2007; Moscovitch, 2016; Robin & Moscovitch, 2017; Sekeres et al., 2018). Others have argued that, through processes of consolidation, memory contents that are initially dependent upon the hippocampus can eventually be retrieved through neocortically-mediated mechanisms, regardless of memory quality (Zola-Morgan & Squire, 1990; Alvarez & Squire, 1994; Kirwan et al., 2008; Broadbent et al., 2010; Squire et al., 2015). Although the two models agree that memory for recently acquired information should be impaired, they differ regarding the role the hippocampus plays in retrieving remote, detailed memories.

While D.C.’s drawing of his childhood home omitted or distorted certain features, it nonetheless contained precise details such as the number of windows on the front door, the number of steps leading up to the door, and the presence and configuration of a basement window. Despite seeing this home daily in his youth, it has been at least several years (and possibly longer) since he could have last visited the building. It seems reasonable to argue that he therefore had to rely on “remote” memory to draw the façade of the structure, although the degree to which D.C. would have relied on episodic, as compared to personal semantic memory, to achieve this is unclear. To the extent that information in personal semantic memory can be highly detailed, this result can be explained in terms of retrieval that is neocortically, rather than medial temporally, mediated. Alternatively, there have been at least some suggestions that hippocampal recruitment is more broadly necessary for the retrieval of precise information (Moscovitch, 2016; Ekstrom & Yonelinas, 2020), and under such a view it appears difficult to account for the presence of such details. Rather, if that were the case, one might instead have expected a more generic (i.e., schematic) building to be drawn, such as that produced by D.C. in the Rey-Osterrieth task’s short delay drawing.

D.C. also demonstrated an ability to retrieve internal details when recalling specific, and remote, episodic memories or when describing hypothetical near-future events. He had difficulty generating any specific details from a recent past memory, defined as having occurred in the past year, but at the time of testing this would have occurred after the onset of his amnesia and might best be interpreted as a sign that the patient was not freely confabulating in response to Autobiographical Interview task cues. A lack of confabulation was also evident in his complete inability to describe a hypothetical event that might occur in his far future (i.e., over a decade away). All memory descriptions were accompanied by numerically more External than Internal details, reminiscent of a basic pattern that has been observed in older adults since the Autobiographical Interview was initially introduced (Levine et al., 2002). The results of the interview need not imply that D.C.’s memory capacity should be considered “typical,” (indeed, the broader pattern of evidence suggests it is not) but do suggest that at least some specific details may be retrievable from memory in hippocampally compromised individuals under certain circumstances (for related results and discussions, see Beatty et al., 1987; Teng & Squire, 1999; Kirwan et al., 2008; but see also Steinvorth et al., 2005; Miller et al., 2020).

Although aspects of D.C.’s preserved abilities, as well as his impairments, can be described as fitting either the predictions of theories asserting indefinite or of a time-limited role of the hippocampus in memory retrieval, there is one pattern of behavior observed in D.C. that is clearly predicted by both types of theory: D.C. was incapable of explicitly recollecting or recalling any information about recent experiences, such as his walkthrough. This is a critical pattern of results which indicates that, irrespective of the outcomes of certain tests, D.C.’s behavior replicates a basic finding that would be expected in a patient with extensive hippocampal damage. Therefore, his deviations from theory-predicted behaviors may be all the more important to understand.

4.4. LIMITATIONS OF THE CURRENT STUDY

This report focuses extensively on a single case study, and as with all such reports, there are inherent limitations that accompany the potential insights that the case itself may provide. One such limitation is that the patient had a history of drug and alcohol use, but the severity of this use is not entirely clear. Given that the controls used to estimate D.C.’s volume reductions did not have similar histories, the precise values should be considered estimates (although, notably, D.C.’s total brain volume was not significantly reduced compared to the control group average; z = −.58, p = .56).

In addition, although the controls for the volume reduction analysis were matched for sex and age, they were not matched on secondary factors such as years of education. Differences also existed between acquisition parameters for the anatomical images of controls, which came from the HCP Aging dataset and which were collected at 3T, and for the patient, which were collected on an NIH scanner at 7T. These differences could also impact the precise estimates of volume loss in patient D.C. However, the aspect of the sequences that would affect volume estimates the most is the voxel size. When we re-sampled the patient’s brain structure masks to the same voxel size as controls, we obtained very similar volume estimates (see Supplementary Table 2). In addition, the total brain volume estimated for each participant would have been impacted by differences in voxel sizes, yet this adjustment did not alter the pattern of observed results, suggesting that the reported findings are robust to sequence and voxel size differences between patient and control datasets.

One could consider the lack of control participants in the walkthrough a limitation of the study. However, given that the patient is at floor with zero details produced during free or cued recall, but at ceiling with the subsequent two-alternative forced-choice recognition, it is not clear that controls would provide nuance to this pattern: the D.C. was clearly impaired under recall-based testing conditions, and was not under recognition conditions. The same is true of the Autobiographical Interview. While controls could help to quantify the extent of his impairment retrieving remote memories, it is still notable that D.C. could retrieve nothing pertaining to his recent experiences but some specific details from remote memory. We additionally note here that D.C. completed a single event description for each time period. Although this is consistent with prior reports using densely amnesic participants (e.g., Steinvorth et al., 2005; see also Miller et al., 2020), it also limits the degree to which the interview scores reflect the range of underlying memory ability.

4.5. CONCLUDING REMARKS

D.C. is an amnesic patient who, through a secondary CNS lymphoma, suffered severe damage to his hippocampus bilaterally that was accompanied by additional bilateral volume reduction in the parahippocampal cortex and the right entorhinal cortex. Neuropsychologically, his lesion resulted in significant memory impairments but otherwise intact cognition. Despite severe memory impairment, D.C. was able to reproduce detailed features of his childhood home, produce some remote autobiographical memory details, and could recognize new information learned in a “natural” manner. This patient may prove particularly informative in the context of the current shift in cognitive neuroscience toward emphasizing more naturalistic forms of learning and memory, both with respect to the hippocampus and to the functional systems with which it interacts.