In search of an auditory engram
- Laboratory of Neuropsychology, National Institute of Mental Health, National Institutes of Health, Department of Health and Human Services, Bethesda, MD 20892
-
Contributed by Mortimer Mishkin, May 18, 2005
Abstract
Monkeys trained preoperatively on a task designed to assess auditory recognition memory were impaired after removal of either the rostral superior temporal gyrus or the medial temporal lobe but were unaffected by lesions of the rhinal cortex. Behavioral analysis indicated that this result occurred because the monkeys did not or could not use long-term auditory recognition, and so depended instead on short-term working memory, which is unaffected by rhinal lesions. The findings suggest that monkeys may be unable to place representations of auditory stimuli into a long-term store and thus question whether the monkey's cerebral memory mechanisms in audition are intrinsically different from those in other sensory modalities. Furthermore, it raises the possibility that language is unique to humans not only because it depends on speech but also because it requires long-term auditory memory.
Both visual and tactile recognition memory in the monkey are severely impaired after bilateral ablation of the medial temporal lobe (1, 2), particularly if the damage is to the perirhinal/entorhinal, or rhinal, cortices (3–8). Similarly severe impairment after lesions or inactivation of the rhinal cortices has been demonstrated in olfactory, visual, and gustatory recognition memory in rats (9–11). The results thus suggest that the rhinal cortices are essential multimodal processing areas linking each of the cortical sensory streams to those deeper limbic and diencephalic structures that are also critical for the formation of stimulus memories (e.g., see refs. 12 and 13). The present study was aimed at extending this model of memory formation to the monkey's auditory modality, and for this purpose we examined the effects in monkeys of rostral superior temporal, complete medial temporal, and selective rhinal lesions on delayed matching-to-sample (DMS) with trial-unique sounds, a putative test of one-trial auditory recognition. Unexpectedly, although each of the first two types of lesion produced significant impairment, the rhinal lesions did not (see also ref. 14). A tentative explanation for this lack of impairment is provided by a comparison of the preoperative data gathered here in audition with those commonly obtained in other modalities. The comparison suggests the surprising possibility that, unlike other sensory stimuli, the auditory stimuli failed to engage the rhinal cortices (or their functional analogues in audition) and so were not stored as lasting stimulus representations. If so, then the auditory deficits observed after the rostral superior temporal and complete medial temporal ablations reflect impairments not in long-term auditory recognition but only in auditory working memory. Together, the results imply that the monkey's cerebral mechanisms for memory in audition are fundamentally different from those in other sensory modalities.
Methods
Subjects. The subjects were nine experimentally naive rhesus monkeys (Macaca mulatta), five males and four females, ranging in weight from 4 to 8 kg. They were fed a diet of Purina monkey chow (No. 5038, PMI Feeds, St. Louis) supplemented with fruit; water was available ad lib. All procedures were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Animal Care and Use Committee of the National Institute of Mental Health.
Apparatus. Automated testing of DMS took place inside a sound-attenuated chamber. The monkeys were trained to sit in a primate chair and respond to acoustic stimuli by touching one of two touch-sensitive copper plates (10 cm square), each mounted on the front of a speaker (Audiotex 30-5121, Rockford, IL). Each speaker was located 40 cm from the monkey, one directly in front (F speaker) and the other at an angle of ≈30° (20 cm) to the right (R speaker). Below each speaker/touch plate was a food well into which a single banana pellet (190 mg) was dispensed after a correct response. A library of 964 distinct acoustic stimuli, each ≈2 s long and presented at a sound pressure level (SPL) of 70–75 dB, included animal sounds, musical segments, and FM sweeps, as well as both natural and man-made environmental sounds. The order of the stimuli in the library was randomized weekly, and the library was divided into daily testing sets. The monkeys typically received 60 trials per day, five or six days per week; because 90 different stimuli were required for each session of DMS (60 sample sounds, 30 of which were presented again as the matching sounds, and 30 different or nonmatching sounds), a given stimulus was repeated once every 10 days on average.
Preoperative Training and Memory Assessment. The auditory DMS task is a variant of one originally described by Konorski (15), and the monkeys were trained on it by an approximation procedure adapted from the method used by Wright et al. (16). The monkeys' memory performance was then assessed in sessions with variable delay intervals. Altogether, there were six stages of training and memory assessment. Animals progressed through stages 1–5 after obtaining a predetermined performance criterion; however, in some cases when performance in the ensuing stage decreased to chance levels, the animals were retrained on the previous stage.
Stage 1. Sounds were first presented one at a time pseudorandomly through the F and R speakers, each presentation being signaled by dimming of the light in the testing apparatus followed by brightening of the light at sound offset. The monkeys tended to orient to the speaker from which the sound emanated, and they were gradually shaped to touch that speaker for reward delivery. Responses during the variable intertrial interval (3–30 s, ITI) reset the interval. This training stage continued until the monkeys responded quickly and correctly to each stimulus presentation and suppressed responding during the ITIs.
Stage 2. They were next trained to touch the speaker only to the presentation of a second sound, which followed the first one after a delay interval of 2 s. Half the sound pairs were “match” trials (i.e., the two stimuli were identical), and half were “nonmatch” (the two stimuli were different from each other), with the two trial types presented pseudorandomly. On match trials, both stimuli in the pair (sample and matching sound) emanated from the F speaker, and the animal was rewarded for touching it. On nonmatch trials, both stimuli in the pair (sample and nonmatching sound) emanated initially from the R speaker, and the animal was rewarded for touching that speaker. After a few such sessions, the pair of nonmatching sounds was presented through both the F and R speakers simultaneously, with the amplitude of the F speaker set very low at first and then increased stepwise within a few more sessions until it matched that of the R speaker (i.e., 70–75 dB sound pressure level). On these nonmatch trials, just as before (indeed, on nonmatch trials throughout testing), the animal continued to be rewarded for touching the R speaker. Training continued with equal sound settings for the two speakers on nonmatch trials until the monkeys achieved a single-session score of 90% correct responses. At this point, the amplitude of the sound presented through the R speaker only was attenuated gradually until it was eliminated completely. In this part of stage 2, sound attenuation of the R speaker proceeded by approximation depending on the animal's performance: A score >90% correct led to a decrease of the sound level on the following session; a score <75% led to a subsequent increase of the sound level; and scores between these two led to a session with the same sound level as before. Throughout this and later stages, a response either to the first sound or before 75 ms after onset of the second sound ended the trial with no reward; trials in which the monkey responded too soon or failed to respond at all were not scored. The intertrial interval was 20 s.
Stage 3. Once the two sounds for both trial types were presented through the F speaker exclusively, the animals were continued in training with the 2-s delay interval until they reached the criterion of 85% correct responses, with a minimum of 80% correct on each trial type, match and nonmatch, for two consecutive sessions.
Stage 4. The monkeys then advanced through a series of increasing delay intervals from the original 2 s, to 3, 4, and finally, 5 s. The criterion for advancing to the next longer delay was the same as that for advancing from stage 3 to stage 4.
Stage 5. When the monkeys reached criterion at the 5-s delay, they were switched to sessions with a variable delay schedule. These sessions included an equal number of trials at pseudorandomly presented delays of 4, 5, and 6 s, followed by sessions of 5, 7.5, and 10 s, and then by sessions of 5, 10, and 15 s, with the monkey advancing to the next set of longer delays after reaching the same criterion as before (85% correct overall for two consecutive sessions) on the earlier set.
Stage 6. On the final preoperative tests, the monkeys were not trained to a criterion but were instead assessed for a minimum of 15 sessions on each of two sets of relatively longer delays, again with each set containing an equal number of trials at each interval and with the different intervals presented pseudorandomly. The first set included delays of 5, 10, 15, 20, and 25 s, and the second set, delays of 10, 20, 30, 40, and 50 s. For each set, testing of a given monkey continued beyond 15 sessions if necessary, i.e., until that monkey reached a stable level of performance, defined as overall scores across three blocks of five sessions each that differed by no more than 5 percentage points. Statistical analyses were based on the data obtained during the 15 sessions of stable performance.
Lesion Groups. The nine animals were divided into three lesion groups. The rostral superior temporal gyrus (rSTG) was removed in two of the monkeys, the medial temporal lobe (MT) was removed in three, and the rhinal cortex (Rh) was removed in four. All lesions were bilateral, performed in a single stage. rSTG lesion. This removal covered the rostral 15–17 mm of the superior temporal gyrus, including within these limits the ventral bank of the lateral sulcus and dorsal bank of the superior temporal sulcus. The removal thus encompassed cytoarchitectonic areas TPO, PGa, IPa, TAa, Ts1, Ts2, and most of Ts3, of Seltzer and Pandya (17), as well as the dorsolateral part of their area Pro in the temporal pole. Region rSTG, which provides the majority of the superior temporal projections to the perirhinal and entorhinal cortices (18–20), is activated by auditory stimuli (21, 22), and its removal together with other nonprimary auditory cortex in the superior temporal gyrus leads to auditory memory deficits (23, 24).
MT lesion. The MT ablation included the amygdala, hippocampal formation, and subjacent portions of the entorhinal cortex [area 28 of Brodmann (25)] and parahippocampal cortex [areas TH/TF of Bonin and Bailey (26)]. Although the perirhinal cortex [areas 35 and 36 of Brodmann (25)] was not included in this ablation, it is now known that such lesions commonly disconnect perirhinal cortex from its downstream projection targets (39).
Rh lesion. The Rh removal included the entorhinal and perirhinal cortices [comprising areas 28, 35, and 36 of Brodmann (25)]. The intended removal was similar to the one described by Meunier et al. (5), except that it extended more rostrally to include the perirhinal extension into the dorsomedial temporal pole [area 36p-dm of Insausti et al. (27)]. This region is the major perirhinal recipient of auditory projections from rSTG (20, 28) and, like rSTG, is activated by acoustic stimuli (21, 22). In an attempt to clarify the role of the hippocampus and parahippocampal cortex in auditory memory, one animal in the Rh group (case Rh-2) had both the hippocampus and parahippocampal tissue removed bilaterally in a second surgical procedure (case Rh-2/H+PH) after the behavioral effect of its first lesion was assessed.
Surgery. Monkeys were fasted for 12 h before surgery and pretreated with an antibiotic (Ditrim, 24% solution, 0.1 ml/kg i.m.). On the day of surgery, the animal was sedated with ketamine (10 mg/kg i.m.) and anesthetized with isoflurane (1–4% to effect). Throughout surgery, the animal was kept warm with a heating pad and hydrated (Ringer's solution i.v.). Mannitol (30%, 30 cc i.v. over 20 min) was infused to reduce brain volume and increase accessibility to temporal lobe areas. Temporal lobe lesions were made by using sterile neurosurgical procedures. In all cases, the zygomatic arch was removed and a large frontotemporal bone flap was turned, extending from the orbit ventrally toward the temporal fossa and caudally to the auditory meatus. The opening was enlarged ventrally to the infratemporal crest to maximize space for the temporal lobe retraction, and the dura was then opened over the posterior frontal and anterior temporal lobes. Tissue in the intended area of the lesion was viewed through an operating microscope and removed by aspiration through a 20-gauge stainless steel tube. In the case of the MT and Rh lesions, in which the intended removals included medial temporal tissue that could not be approached rostrally, a second pair of dural openings was made over the occipitotemporal junction, thereby allowing access by means of a caudal approach. In case Rh-2, which received a second surgery (see Lesion Groups), the region of the occipitotemporal junction was reopened to permit ablation of the hippocampus and parahippocampal cortex (case Rh-2/H+PH). On completion of the lesion, the dura was sewn, the bone flap was reattached, and the scalp wound was closed in anatomical layers. In each case, the lesion was made bilaterally in a single stage. Postsurgical analgesics were provided as required, in consultation with the National Institute of Mental Health veterinarian.
Histology. On completion of postoperative testing, the monkeys were injected with a lethal dose of sodium pentobarbital and perfused transcardially with saline followed by 10% formalin. The brains were blocked in the coronal plane, removed, photographed, cryoprotected through a series of glycerol solutions (29), frozen, and subsequently cut in 50-μm slices on a freezing microtome. Every fifth section was mounted on slides, stained with thionine, coverslipped, and examined under the microscope. The extent of the damage in each case was then plotted on drawings of a standard rhesus monkey brain at 1-mm intervals, and the lesions were reconstructed (Figs. 5–10, which are published as supporting information on the PNAS web site). Also, the volume of the lesioned areas relative to the volume of those areas in the standard brain was measured by using a digit-tablet in conjunction with the nih image program. (Because in case Rh-2/H+RH only the combined damage after both surgeries could be evaluated histologically, the extent of its first lesion was estimated by plotting the damage visible on magnetic resonance images taken before the second surgery.) The quantitative results of the histological analysis are shown in Tables 2 and 3, which are published as supporting information on the PNAS web site.
Postoperative Retraining and Memory Assessment. Two weeks after surgery, the monkeys were retrained to criterion at delay intervals of 2, 3, 4, and 5 s, in succession (as in preoperative stages 3 and 4). They were then given 15 sessions on each of the two final preoperative sets of variable delays, i.e., 5–25 s and 10–50 s (as in preoperative stage 6). Because the performance of the monkey with ventrocaudal neostriatal infarcts (case MT-1) remained at chance on delays >15 s, resulting in an overall score of 58% correct on the first set of variable delays (5–25 s), it was not tested on the second set (10–50 s).
Results
As indicated in Methods, nine monkeys were given approximation training through six stages of auditory DMS, after which they received lesions of either the rSTG, MT, or Rh, and then retested on three of the stages (stages 3, 4, and 6). The lesions were generally as intended (see Figs. 1 and 6 and Tables 2 and 3), with little adventitious damage except in case MT-1, which sustained bilateral infarcts to the tail of the caudate nucleus and ventral putamen. Complete lesion reconstructions for each monkey with example photomicrographs of the lesions are shown in Figs. 5–10.
Summary of the three types of intended lesions shown on a series of coronal sections and on the ventral and lateral surface views of a standard rhesus monkey brain. Numbers indicate distance in mm relative to the interaural plane. Rh, horizontal lines; MT, blue; rSTG, green.
The mean number of trials taken by the nine animals preoperatively to reach criterion on DMS at a delay of 5 s (training stages 1–4) is presented in Fig. 2. As the figure illustrates, the task proved extremely difficult, with the animals requiring a total of ≈15,000 trials (or ≈250 sessions) to achieve a score of 85% correct responses at the 5-s delay. They then required a mean of nearly 3,000 additional trials preoperatively to complete the first series of variable delay schedules (stage 5). Also, their final preoperative performance on a delay of 50 s, the longest one included in the second series of variable delay schedules (10–50 s, stage 6), fell to ≈70% correct (see Fig. 3b, Preop). Statistical analyses indicated that the three groups into which the animals were later divided did not differ reliably on any of these six stages of preoperative training and memory assessment (see Table 4 and Fig. 11, which are published as supporting information on the PNAS web site).
Preoperative learning. Shown is the mean number of trials (± SE) taken by all nine monkeys to complete each of the first four stages of training (see Methods), culminating in attainment of criterion at the 5-s delay. Stage 1, sound from either the front or the right-side speaker, response to location of sound source; stage 2, sounds shifted gradually from (i) either speaker to (ii) both speakers to (iii) front speaker only, with the correct response now being to front speaker for match trials and to right-side speaker for nonmatch trials; stage 3, training to criterion at delay of 2 s; stage 4, training to criterion at delays of 3, then 4, and then 5 s.
Postoperative performance. Shown are mean scores (±SE) on randomized, variable delay schedules. (a) Shown are 5- to 25-s delays. (b) Shown are 10- to 50-s delays. ▪, Monkeys with complete lesions of the medial temporal lobe (a, n = 3; b, n = 2). ○, Preoperative scores of all animals (n = 9). •, Monkeys with selective lesions of the rhinal cortices (n = 4). ▴, Monkeys with lesions of the rostral superior temporal gyrus (n = 2).
Postoperatively, the three lesion groups did differ reliably. Because preoperative vs. postoperative comparisons are inappropriate for the relearning results, retraining having begun directly on stage 3, i.e., at the 2-s delay, relearning analyses were limited to postoperative comparisons (Table 1). Mann–Whitney U tests indicated that group Rh reacquired the DMS rule at the 2-s delay significantly more quickly than both the MT and rSTG groups (U = 1 and U = 0, respectively; P = 0.05 for both), whereas the latter two groups did not differ from each other. The same trends were seen in stage 4, at the 3- to 5-s delays, but the group differences were not significant.
Postoperative memory assessment on stage 6 (Fig. 3) revealed an impairment in both the rSTG and MT groups, but not in group Rh. In the case of these memory performance tests, the data for each set of variable delays were analyzed by a repeated-measures ANOVA with three factors: Pre/Post, Delay, and Group. For the first set, 5–25 s, the interaction between Pre/Post and Group fell short of significance (F = 4.39; df = 2/6; P = 0.067), but for the second set, 10–50 s, this interaction did reach significance (F = 6.90; df = 2/5; P = 0.036). On this second set (Fig. 3b), the postoperative decreases in the overall scores of the rSTG and MT groups (12% and 13% correct, respectively) were reliable (post hoc Bonferroni tests, P < 0.05), whereas the decrease in the overall score of the Rh group (2 percentage points) was not. Moreover, the postoperative performance decreases in the rSTG and MT groups were each significantly greater than the decrease in the Rh group (post hoc Bonferroni tests, P < 0.05). Interestingly, although the Delay factor was highly significant (P < 0.001) in both sets of variable delays, this factor did not interact with Pre/Post, Group, or Pre/Post × Group, suggesting that the postoperative impairments in the rSTG and MT groups were not delay-dependent.
Because the impairment produced by complete medial temporal ablations was not reproduced by selective removal of the rhinal cortices, one of the animals in group Rh (case Rh-2) was given a second operation in which its lesion was enlarged to include the hippocampus and parahippocampal cortex (case Rh-2/H+PH; see Methods). This second operation likewise failed to yield impairment either in reacquisition of the DMS rule (a total of 399 trials on stages 3 and 4, compared with 813 trials after the first operation) or in memory performance (an average of 86% and 79% correct on 5–25 and 10–50 s, respectively, compared with 83% and 80% correct after the first operation; see also Fig. 12, which is published as supporting information on the PNAS web site).
Discussion
Given the critical role of the rhinal cortices for stimulus recognition in every other sensory modality that has been investigated (see introduction), the lack of an impairment in audition after rhinal lesions was not anticipated, particularly because impairments did result from ablation of both the rostral superior temporal gyrus and the medial temporal lobe. The difference between the effects of medial temporal and rhinal lesions becomes even more perplexing considering our failure to reproduce the medial temporal deficit by superimposing a combined ablation of the hippocampus and parahippocampal gyrus on a preexisting rhinal lesion (case Rh-2/H+PH). Similarly negative results in auditory memory after combined ablations of the rhinal cortices and hippocampus were obtained in the study on dogs (14). In attempting to account for these unexpected findings, it is useful to begin with an analysis of the learning and performance results we obtained in our animals preoperatively.
A large literature accumulated over decades underscores the difficulty investigators have had in training monkeys to discriminate and remember auditory as compared with other sensory stimuli (30–35). However, in most previous studies of auditory memory employing the DMS rule, there was the possibility that the training difficulty had been compounded by the use of small stimulus sets, often just a single pair (23, 24), because such sets require the monkey to make contradictory responses to the same stimulus repeatedly, an ambiguity that is resolved only when the animal successfully abstracts the cue of stimulus recency, a form of short-term or working memory (36). By contrast, in the present study, where the goal was to examine long-term recognition memory, we necessarily used a very large stimulus set, which incidentally minimizes both stimulus-response and stimulus-reward ambiguity. In addition, in the attempt to further ease the animals' learning of the DMS rule, we adopted the strategy of training the monkeys (i) by approximation through several stages, and (ii) by using delays of only 2 s initially, which were then gradually increased by successive increments of only 1 s. Despite these numerous aids, the animals required an average of nearly 15,000 trials to meet the criterion of 85% correct responses at a 5-s delay. This result is ≈30 times more than the number of trials (≈500) that naive monkeys need to learn the DMS rule in vision, which they achieve (i) without approximation training, (ii) with a longer initial, 10-s delay, and (iii) to the more stringent criterion of 90% correct (36). Because of the marked qualitative differences in stimulus materials and procedures used to train animals in visual and auditory DMS (e.g., 3D colored objects vs. fluctuating sound bursts, simultaneous vs. successive stimulus pairing, approach/avoid vs. go center/go right responses, etc.; see Methods), the above comparison is not intended as an accurate quantitative index of the difference in rate of rule learning in vision vs. audition. Nevertheless, it is clear that acquisition of the DMS rule in audition was extremely protracted, and this slow learning provided the first indication in the present study that auditory memory mechanisms in monkeys could well differ from their memory mechanisms in other modalities.
Having attained the 85% criterion at delays of 5 s, the normal monkeys maintained that level only up to delays of ≈10 s (stage 6), after which their scores began to fall. If the delay at which the animals averaged 75% correct is taken as the forgetting threshold for one-trial auditory memories, then this threshold in the present study seems to have been reached by ≈35 s (Fig. 4). In contrast, for one-trial visual memories, the forgetting threshold of normal monkeys seems to be ≈15–20 min (37, 38), or ≈30 times longer than the one found here in audition. Similarly, for one-trial tactile memories, the forgetting threshold exceeds 10 min (7). Interestingly, however, after a rhinal cortical lesion, the forgetting thresholds in vision and touch are seriously reduced, also to ≈35 s (5, 7), i.e., just the same as the normal threshold in audition (Fig. 4). [The marked differences between vision and audition in rule learning, forgetting thresholds, and effects of rhinal lesions in the monkey may be related to the striking differences in the serial position functions that Wright (35) demonstrated in the monkey's memory for visual and auditory lists.]
Comparison of auditory (a) and visual (b) memory performance. Scores are group means before and after lesions of the rhinal cortices. (a) Auditory delayed matching-to-sample, 10- to 50-s delays. •, Rh (n = 4). ○, Preop (n = 4). (b) Visual delayed nonmatching-to-sample, 10- to 60-s delays. •, Rh (n = 7). ○, N (n = 4). Horizontal dashed line, 75% correct. Vertical dashed-dotted line, approximate forgetting threshold (see Discussion).
The above comparisons indicate that a normal monkey performing our version of auditory DMS resembles a monkey with rhinal cortical lesions performing visual or tactual DMS, in the sense that neither seems able to store long-term stimulus representations. In the case of the normal monkey performing auditory DMS, this inability suggests that the auditory stimuli were processed without the participation of the rhinal cortices (or any other cortical tissue serving long-term memory in audition), and consequently ablating the rhinal cortices had no deleterious effect. The correlate of this proposal is that, like the residual mnemonic ability in vision and touch after rhinal lesions, the mnemonic ability in audition observed here in normal monkeys rests entirely on mechanisms mediating short-term or working memory.
The proposal that performance on auditory DMS was based on short-term or working memory implies that this is the form of memory that was impaired after both the rSTG and MT lesions. Such an implication is supported by the finding that the impairments were independent of delay interval, appearing at even the shortest interval tested. Moreover, the implication that the deficit in group rSTG was in short-term memory is consistent with the anatomical position occupied by rSTG in the cortical auditory system. That is, rSTG seems to be a late station in the ventral processing stream for audition (M. Munoz, M.M., and R.C.S., unpublished data) much like area TE is for vision, and so it is highly likely that rSTG serves an important role in short-term auditory memory just as area TE does in short-term visual memory (12). However, the implication that the impairment in Group MT likewise was in auditory short-term memory seems to contradict the well established view that the medial temporal lobe generally, and the hippocampal system in particular (i.e., hippocampus plus perirhinal, entorhinal, and parahippocampal cortices) is essential for long-term, not short-term, stimulus memory. The finding that the auditory memory of case Rh-2/H+PH was in fact preserved despite complete removal of the hippocampal system suggests a possible resolution of the contradiction, namely, that the impairment produced by the MT ablation resulted not from damage to the hippocampal system but from collateral damage, such as severing the projections of rSTG to downstream areas in the prefrontal cortex, medial thalamus, or both (e.g., see ref. 39). An anatomical study undertaken to examine this possibility (M. Munoz, M.M., and R.C.S., unpublished data) demonstrated that an MT removal does indeed disconnect rSTG anatomically from several prefrontal and medial thalamic areas. Whether one or more of those disconnected areas normally serve short-term auditory memory, however, has not yet been determined, and so the explanation for the impairment produced by the medial temporal ablation remains unsettled.
But perhaps the more important issue raised by the present study was the apparent failure by normal monkeys to place central representations of auditory stimuli in a long-term store, despite protracted training. This failure precluded both rapid learning of the DMS rule based on the categorical distinction between familiarity and novelty as well as robust retention of a given stimulus in memory beyond a period of seconds. Moreover, the monkey's apparent deficiency in auditory memory is not specific to the behavioral methods that were used in the present study. Ongoing attempts to train normal monkeys by using different tasks, different incentives, and different training procedures from the ones used here have similarly failed to yield either rapid learning of an acoustic-based behavioral rule or retention of acoustic stimuli beyond the limits of working memory (A. Poremba, M. M. Malloy, R.C.S., E. Kang, R. E. Carson, P. Herscovitch, and M.M., unpublished data). Further attempts of this type are needed, but the results to date clearly indicate that auditory memory in monkeys is qualitatively different not only from their memory in other sensory modalities but also from auditory memory in humans, all of which include long-term stimulus recognition as one prominent type of memory.
The possibility that monkeys lack auditory recognition ability seems to be directly contradicted by ethological studies reporting that monkeys recognize alarm calls signaling predators (40) as well as the voices of peers and kin (41). These ethological findings could, however, reflect the utilization of nonauditory mnemonic abilities, such as the acquisition of stimulus-response associations that underlie habit formation (42) or recoding of the auditory stimuli in visual memory based on the earlier formation of auditory-visual associations (43). In light of the present results, these alternative possibilities merit careful study.
Because monkeys, like mammals generally, have long-term recognition memory in every other modality that has been examined, their apparent lack of this ability in audition is a mystery that raises many important questions, such as how fluctuating auditory stimuli are processed, whether encoding and storing them necessitate special neural mechanisms not required in other modalities, whether other primates such as baboons and apes have such neural mechanisms, and even whether oral language is unique to humans because it depends not only on articulate speech but also on long-term auditory memory.
Acknowledgments
We thank Newlin Morgan and George Dold for help with design of the testing apparatus; Jeff Moran and David Schoppik for assistance with programming; Danielle Becker, Danielle Chappell, Jenny Kim, Rajeev Mohan, Kendra Pyle, Danielle Rickrode, and David Schoppik for assistance with monkey training; and Becky Lowther and Jon Rueckemann for assistance with figures. The authors declare they have nocompeting financial interests. This work was supported by the Intramural Research Program/National Institute of Mental Health/National Institutes of Health/Department of Health and Human Services and also by grants to J.F. from the National Institute on Deafness and Other Communication Disorders, and the Multidisciplinary University Research Initiative.
Footnotes
-
↵ † To whom correspondence should be addressed at: Laboratory of Neuropsychology, National Institute of Mental Health, National Institutes of Health, Building 49, Room 1B80, Bethesda, MD 20892. E-mail: richardsaunders{at}mail.nih.gov.
-
↵ * Present address: Neural Systems Laboratory, Center for Acoustic and Auditory Research, Institute for Systems Research, University of Maryland, College Park, MD 20742.
-
Author contributions: J.F., M.M., and R.C.S. designed research; J.F. and R.C.S. performed research; J.F., M.M., and R.C.S. analyzed data; and J.F., M.M., and R.C.S. wrote the paper.
-
Abbreviations: DMS, delayed matching-to-sample; F, front; R, right; rSTG, rostral superior temporal gyrus; MT, medial temporal lobe; Rh, rhinal cortex.
