Abstract
There is controversy regarding the effect of isolated fornix damage on human memory.We report a patient who suffered a traumatic penetrating head injury that resulted in a significant and persistent anterograde amnesia. CT revealed a lesion that involved the region of the proximal, posterior portion of both fornices without evidence of damage to other hippocampal pathways or to other structures known to be critical for memory, such as the hippocampus, thalamus, or basal forebrain. The unique location of the lesion in this patient provides evidence supporting the role of isolated fornix lesions in amnesia.
NEUROLOGY 1995;45: 1546-1550
There is controversy about the effect of isolated fornix damage on human memory. [1] Reports documenting amnesia following fornix lesions have postulated damage to pathways that interrupt critical interconnections between the hippocampal formation and other structures important for memory: the septum, the anterior thalamus, and the mamillary bodies. The majority of efferent pathways from the hippocampal formation travel through the fornices. Several reported cases have implicated transection of the fornical pathways alone as the basis for memory impairment. [2-5] However, there is also a smaller contingent of pathways interconnecting the hippocampus with the septum and diencephalon that travel within the supracallosal longitudinal striae and the cingulate bundle.
In a single case report, von Cramon and Schuri [6] postulated that complete disruption of all three pathways (the proximal fornix, the supracallosal longitudinal striae, and the cingulate bundle) is necessary for persistent amnesia. Several cases already reported in the literature provide support for this claim. The patient reported by Heilman and Sypert, [3] with a space-occupying splenial tumor, was originally believed to have amnesia due to isolated bilateral proximal fornix damage, but damage to the supracallosal longitudinal striae and posterior cingulate bundles also likely occurred. The patient with ``retrosplenial amnesia'' reported by Valenstein et al [7] had a lesion involving predominantly the retrosplenial cortex and posterior cingulate bundle, but damage to the fornix could not be excluded. Finally, a second case reported by von Cramon [8] provided evidence that damage to the retrosplenial area involving the left posterior cingulum alone, without damage to the fornix, causes only temporary memory dysfunction. Thus, these case reports suggest that the likelihood of severe and persistent memory dysfunction increases with complete disruption of all three hippocampal pathways. [8]
We report a case of severe and persistent anterograde amnesia that we postulate was due to traumatic bilateral fornix transection without evidence of damage to these other hippocampal pathways (the supracallosal longitudinal striae and the posterior cingulate bundles). The unusual location of the lesion in our patient provides a unique opportunity for discovering additional evidence regarding the role of these structures in memory.
Case report.
Patient history.
The patient is a 32-year-old right-handed woman who suffered a penetrating head injury from a gunshot on October 9, 1991. She had attended one year of college and was a homemaker at the time of the accident. She had no history of drug or alcohol abuse. She had no cognitive impairments prior to her injury. She was not on any medications during her evaluation. The bullet entry was in the left parietal area, and no exit wound was found. On admission to the hospital, her vital signs were stable and she had a Glasgow Goma Scale score of 10. Her pupils were equal and reactive and she was described as responding only to painful stimuli. She underwent an emergent left parietal craniotomy for debridement of the wound. She was conscious shortly after surgery and there were no significant postoperative complications. Two months later, she required placement of a ventriculoperitoneal shunt for hydrocephalus, and at the time of that surgery further bullet fragments were removed through her previous craniotomy site. She was subsequently discharged to a rehabilitation hospital and initially evaluated 4 months after her injury. On elemental neurologic examination, she had a slightly larger but normally reactive left pupil and a left homonymous hemianopsia. She also had a dense right hemiparesis and severe proprioceptive loss on her left side.
General neuropsychological evaluation.
She was tested 4 months after injury. During testing, she was alert and cooperative. She had a full range of affect and was aware of her deficits. However, she often was giddy during examinations and seemed unconcerned about the severity of her deficits. Her attention was very good and she worked diligently on all the tasks for extended periods of time. On language evaluation, she produced fluent, grammatical speech that contained abundant phonemic paraphasias and utterances that were empty in content. She had impaired naming and repetition, but auditory and reading comprehension were normal. Formal testing with the Western Aphasia Battery [9] indicated that she fit the taxonomic category of conduction aphasia. There was no evidence of buccofacial or limb apraxia. She had difficulty performing the WAIS-R (Wechsler Adult Intelligence Scale-Revised) Block Design subtest. [10] She could copy simple figures accurately, but copying of complex figures showed mild left neglect and was disorganized. Line bisection showed mild left neglect, but she was able to cancel all stimuli on a letter-cancellation test. On the Benton Facial Recognition test, [11] she was borderline impaired.
Memory testing.
The most significant feature of the patient's impaired cognition noticed by family and staff was a severe anterograde memory deficit. Because of her impairments in language, she was tested on a battery of memory measures that required only a verbal yes/no response or pointing-responses that we determined she could reliably perform. To examine the potential influence of perceptual impairments on memory performance, we directly examined her ability to match stimuli used in the nonverbal memory measures. When presented with stimuli from the Kimura recurring figures test [12] or the Warrington Recognition Memory Test (for facial recognition), [13] she was consistently able to select the target from among four alternatives (10/10 correct for both tests), suggesting that she did not have difficulty perceptually identifying these stimuli.
Her performance on a variety of memory tasks is detailed below. When possible, her performance was compared with normative data from the literature. For those experimental measures for which normative data were not available in the literature, 10 age- and education-matched controls were tested.
Working memory.
Working memory was impaired in both the verbal and the visual modalities on the Wechsler Memory Scale-Revised (WMS-R) [14]; her digit span, tested by having her point to numbers, was three forward and two backward, and her visual memory span was four forward and three backward. Each of these scores fell below the lowest score reported as normative in control data. She was also tested on a delayed recognition memory span test [15] in which she had to identify a new stimulus among an increasing set of previously presented familiar stimuli. She had markedly reduced verbal (7, versus 10.7 for controls) and spatial (4, versus 10.4 for controls) spans.
Episodic new learning.
There were numerous accounts by family and staff of the patient exhibiting an impairment in episodic memory. She was oriented to place, but not consistently to the date. Throughout her hospitalization she could not recall visits by family members even though they occurred quite frequently. She also consistently could not remember details about her therapy sessions, not even when asked about them on the same day that they took place. For example, she often forgot having been in the swimming pool for therapy even though she still had her bathing suit on during questioning. It would often take numerous contacts before she would learn the names of her therapists. Our neuropsychologist reported that 5 minutes after his examination of her she could not remember his having been with her. On another occasion, she was examined for more than 20 minutes at a medical conference with approximately 40 persons in attendance. When asked about this event 30 minutes later, she had no recollection of its having occurred.
On formal testing, she was impaired on several different measures of both verbal and visual memory. On the Warrington Recognition Memory Test, her memory for a list of recently studied words was significantly impaired (29/50 correct; less than 5th percentile). On a modified version of the California Verbal Learning Test [16] consisting of nine items, her recognition memory was still very poor, even after five presentations. She recognized eight of nine targets with two false-positive responses on immediate testing and five of nine targets with four false-positive responses after a 20-minute delay. The large number of false-positive responses compared with correct responses suggests that she was performing at chance level after a delay. She was also impaired in recognizing previously presented words on the Kimura recurring words test [17] (20, versus 35.5 plus minus 2.9 for controls). To examine the effect of different encoding manipulations on memory, a levels-of-processing task [18] was administered in which she was asked to judge a set of 36 words for either orthographic (ie, Is it printed in uppercase letters?), phonologic (ie, Does it rhyme with ``match''?), or semantic (ie, Is it a type of animal?) characteristics. Despite answering all questions correctly, she displayed impairment in delayed recognition of these words at 10 minutes (15/36 correct, versus 25.2/36 plus minus 5.2 for controls). Further, unlike in normal subjects, there was no benefit for recognizing words that were encoded semantically (6/12, versus 10.5/12 plus minus 1.6 for controls) compared with phonologically (3/12, versus 8.2/12 plus minus 2.0 for controls) or orthographically (6/12, versus 6.5/12 plus minus 2.6 for controls), suggesting deficient encoding of information in memory.
She had similar impairments in visual memory. On the Visual Paired Associates subtest of the WMS-R she was severely impaired at learning figure-color pairs during three consecutive trials (1/18 correct, versus 14.8/18 plus minus 3.7 for controls). On the Warrington Recognition Memory Test, her memory for recently studied faces was significantly impaired (32/50; less than 5th percentile). She also had difficulty recognizing previously presented figures on the Kimura recurring figures test [12] (8, versus 20.8 plus minus 5.5 for controls). In summary, she had evidence of significant impairments in learning new information on several different tasks in both the verbal and visual modalities.
Procedural learning.
Because of the hemiparesis affecting her right hand, she was tested with her nondominant hand on the rotary pursuit task. Due to significant proprioceptive loss in her nondominant hand, her baseline performance was quite impaired. Nonetheless, she showed some learning over five trials (trials 1 to 5: 0.46, 0.61, 0.85, 1.1, 3.6 seconds), and there was no loss of learning after a 20-minute delay (3.7 seconds). Similar motor learning was found on the Luria ``fist-edge-palm'' test [19] during both the immediate and delayed conditions.
Remote memory.
She was able to recall the past three presidents and several past world events, such as the Persian Gulf War and the J.F.K. assassination. On the famous faces test [20] she showed mild impairment and a flat curve in recall across decades (1950s, 4/8; 1960s, 5/8; 1970s, 3/8; 1980s, 5/10). On a fame judgment test in which she was asked to judge whether or not 24 photographs or names of persons were famous, she was impaired in both the visual domain (71% correct, versus 88.7% plus minus 6.8 for controls) and the verbal domain (88% correct, versus 98.0% plus minus 3.1 for controls). All errors consisted of failure to recognize famous people. Autobiographical memory based on an informal interview seemed to be adequate, as verified by her husband.
Semantic knowledge.
Using a multiple-choice format, she performed in the low average range (17/29 correct) on the WAIS-R Information subtest. [21] She was able to sort photographs from the famous faces test into appropriate semantic categories (ie, politicians, actors, and so forth).
Follow-up.
At a routine 1-year follow-up examination, she continued to display significant impairment on several memory measures. She obtained identical scores on the Warrington Recognition Memory Test in both the verbal and visual conditions. Her family reported that her episodic memory remained markedly impaired in her day-to-day activities.
Neuroimaging.
Head CT performed 8 months after injury showed a lesion that tracks the pathway of the bullet wound Figure 1. MRI could not be performed because of remaining bullet fragments within the brain.
Figure 1. Brain CT showing contiguous sections through the full extent of the lesion. Entry of the bullet was in the left posterior parietal lobe (panel F), causing a lesion track that followed a straight path through the right lateral ventricle into the region of both proximal fornices and the hippocampal commissure (small arrowheads) and ended in the right posterior temporal lobe (panel A). Critical memory structures such as the hippocampus (large arrow-heads), basal forebrain, dorsomedial and anterior nucleus of thalamus, and mamillary bodies were not damaged. In addition, the supracallosal longitudinal striae, which lie just posterior to the splenium of the corpus callosum (curved arrow), and the cingulate bundle within the retrosplenial region (straight arrow), were not damaged. A ventriculoperitoneal shunt can be seen in panels C through F, and bullet fragments can be seen in panels B and C.
Discussion.
Our patient has a severe anterograde amnesia that has persisted for at least 1 year after her injury. Even though impairments in other cognitive domains such as language and perception were also evident, her amnesia is clearly disproportionate. While these nonmemory impairments can lead to difficulty with encoding new information, [22,23] our methods of measurement indicated that she could sufficiently process the stimuli we asked her to remember. Her deficits were severe across a wide range of memory measures. Day-today observation and interactions with our patient made it clear to our staff and her family that she had a severe episodic memory impairment that affected her functional activities. Her anterograde amnesia was out of proportion to her limited retrograde amnesia, a finding consistent with another reported case of amnesia following fornix damage. [4] Similar to amnesics with bilateral hippocampal system damage, she retained her ability for procedural learning. [24]
Her language impairment limited our ability to investigate the underlying mechanisms of her amnesia, such as possible encoding or retrieval difficulties. However, we did determine that she did not benefit from deeper semantic encoding of information in later recognition. Her amnesia seemed to be severe in both visual and verbal modalities. This would be consistent with bilateral damage to hippocampal pathways. Unlike other reported patients with fornix damage, she has a significant impairment of working memory. We postulate that this is due to damage to her left posterior parietal region, which is implicated in auditory-verbal short-term memory. [25] This finding in itself cannot explain her difficulties in learning new information, since patients with damage restricted to the left parietal lobe have relatively intact long-term memory. [25]
The damage from the bullet wound that is evident on the CT likely involves both fornices and the region of the hippocampal commissure. There is no evidence of damage to other areas involved in memory processes, such as the hippocampus, the amygdala, the anterior or dorsomedial thalamus, the mammillary bodies, or the medial septal/diagonal band complex. Of course, microscopic damage to these structures cannot be ruled out without neuropathologic analysis. With this caveat in mind, what seems critical in this case is the apparent lack of damage to other hippocampal interconnecting pathways besides the fornix. First, because the lesion track goes under the lateral ventricles, it cannot involve the supracallosal longitudinal striae. Second, the lesion track is anterior to the splenium of the corpus callosum and therefore cannot involve the retrosplenial and posterior cingulate cortex. Therefore, our case provides evidence that bilateral damage to the proximal fornix is sufficient to cause amnesia. The lesion in this case may be sufficiently proximal to prevent fibers from traveling between the hippocampal formation and the supracallosal longitudinal striae or the cingulate bundle, but this seems unlikely since the lesion track goes above the temporal horns, which is above the area of the proximal posterior hippocampus.
It is possible that the extent of damage to all three hippocampal pathways correlates with the severity of amnesia. Although direct comparison with the patient reported by von Cramon and Schuri, [6] who had disruption of all three critical pathways, is difficult because the lesion was unilateral and different memory measures were used, our patient's amnesia is at least as severe as that of their patient. Thus, our patient provides evidence that fornix damage alone can cause amnesia comparable to that caused by damage to all three hippocampal pathways, but this issue clearly needs further investigation.
Isolated damage to the proximal fornix (posterior portion) as in our patient is extremely uncommon. Other cases of damage to the proximal fornix, due to tumor invasion, also had damage to other hippocampal pathways, including the supracallosal longitudinal striae and the cingulate bundle. [3,26] The only reported case with pathologic documentation of bilateral damage restricted to the proximal fornix was caused by a lung carcinoma metastasis, and the patient was not reported to have a memory impairment [27] although formal neuropsychological testing was not performed. Also, during the final 2 weeks of the patient's life he remained confused, lethargic, or in a coma. The rapidly growing metastasis may have caused most of the fornix damage during this terminal period, when memory functioning could not be adequately assessed.
Unlike the case we report here, most reported cases have involved the distal (anterior) portion of the fornix. There are several possible explanations for the uncertainty that these cases have generated regarding the role of the fornix in memory. First, many reports have not clearly described the extent of fornix damage. Therefore, it cannot be determined whether the fornix is partially or completely disrupted. Second, reports of cases without memory impairment following fornix damage do not provide sufficient objective memory testing to adequately determine whether patients are amnesic. Gaffan and Gaffan [1] reviewed 142 cases of surgical transection of the distal portions of the fornices in patients with epilepsy. Most of these cases had only unilateral fornix transection and amnesia was not evident and would not be expected. Only two cases of bilateral fornix transection had lack of amnesia, but neither study had enough behavioral or anatomic data to verify such a claim. [1] Also, Heilman and Sypert [3] point out that surgical section of the columns of the fornix may spare memory because projections from the hippocampus directly to the anterior thalamic nuclei and to the septal area leave the fornix before the site of the anterior surgical transection.
Much stronger evidence of isolated anterior fornix damage causing amnesia comes from cases of surgical transection of the fornix during colloid cyst removal. Several recent well-studied cases have clearly documented amnesia following surgery. [2,4,28] Thus, evidence of isolated damage to the distal fornix in these cases, as well as damage to the proximal fornix in our patient, supports the hypothesis that fornix damage alone can cause lasting amnesia. The original hypothesis of Delay and Brion, [29] that interruption of the pathways that interconnect the hippocampus and the mammillary bodies is as likely to result in amnesia as damage to either primary structure, appears to hold true. Further evidence supporting this hypothesis will have to be obtained from naturally occurring lesions of these critical structures. Since our patient's lesion location is unlikely to be replicated by damage from tumors, arteriovenous malformations, or infarctions, this report provides a unique opportunity to contribute to the growing body of evidence supporting the role of fornix lesions in human amnesia.
Acknowledgment
The authors thank Dr. Randall Otto for performing parts of the neuropsychological testing.
- Copyright 1995 by Advanstar Communications Inc.
REFERENCES
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.