An attractor network in the hippocampus: theory and neurophysiology

A quantitative computational theory of the operation of the CA3 system as an attractor or autoassociation network is described. Based on the proposal that CA3-CA3 autoassociative networks are important for episodic or event memory in which space is a component (place in rodents and spatial view in primates), it has been shown behaviorally that the CA3 supports spatial rapid one-trial learning and learning of arbitrary associations and pattern completion where space is a component. Consistent with the theory, single neurons in the primate CA3 respond to combinations of spatial view and object, and spatial view and reward. Furthermore, single CA3 neurons reflect the recall of a place from an object in a one-trial object-place event memory task. CA3 neurons also reflect in their firing a memory of spatial view that is retained and updated by idiothetic information to implement path integration when the spatial view is obscured. Based on the computational proposal that the dentate gyrus produces sparse representations by competitive learning and via the mossy fiber pathway forces new representations on the CA3 during learning (encoding), it has been shown behaviorally that the dentate gyrus supports spatial pattern separation during learning, and that the mossy fiber system to CA3 connections are involved in learning but not in recall. The perforant path input to CA3 is quantitatively appropriate to provide the cue for recall in CA3. The concept that the CA1 recodes information from CA3 and sets up associatively learned back-projections to neocortex to allow subsequent retrieval of information to neocortex provides a quantitative account of the large number of hippocampo-neocortical back-projections.     

Behavioral functions of the CA3 subregion of the hippocampus

From a behavioral perspective, the CA3a,b subregion of the hippocampus plays an important role in the encoding of new spatial information within short-term memory with a duration of seconds and minutes. This can easily be observed in tasks that require rapid encoding, novelty detection, one-trial short-term or working memory, and one-trial cued recall primarily for spatial information. These are tasks that have been assumed to reflect the operations of episodic memory and require interactions between CA3a,b and the dentate gyrus via mossy fiber inputs into the CA3a,b. The CA3a,b is also important for encoding of spatial information requiring multiple trials including the acquisition of arbitrary and relational associations. These tasks tend to be non-episodic and can be mediated by arbitrary and conjunctive operations. All these tasks are assumed to operate within an autoassociative network function of the CA3 region. The output from CA3a,b via the fimbria and the medial and lateral perforant path inputs play a supporting role in the neural circuit that supports the operation of these tasks. The CA3a,b also plays a role in sequential processing of information in cooperation with CA1 based on the Schaffer collateral output from CA3a,b to CA1. The CA3a,b also supports retrieval of short-term memory information based on a spatial pattern completion process. Finally, CA3c may, in cooperation with the dentate gyrus, serve an important role in processing the geometry of the environment.     

Intrinsic and extrinsic wiring of CA3: indications for connectional heterogeneity

Within the framework of a special issue on CA3, it was deemed relevant to summarize what is known about the extrinsic and intrinsic wiring of CA3 as a basis for other contributions. Here, I have aimed to update already existing excellent reviews on the subject and to raise the issue whether or not the known architecture of the field supports the generally accepted notion that CA3 is particularly wired to function as an autoassociative network. The data reviewed strongly support this notion but in addition point to connectional heterogeneities that may point to functional specializations in CA3, on top of its role as an autoassociative network uniquely relevant to efficient encoding and recall of information.     

Effects of language dominance on item and order memory in free recall,serial recall and order reconstruction

Spanish–English bilinguals (N = 144) performed free recall, serial recall and order reconstruction tasks in both English and Spanish. Long-term memory for both item and order information was worse in the less fluent language (L2) than in the more fluent language (L1). Item scores exhibited a stronger disadvantage for the L2 in serial recall than in free recall. Relative order scores were lower in the L2 for all three tasks, but adjusted scores for free and serial recall were equivalent across languages. Performance of English-speaking monolinguals (N = 72) was comparable to bilingual performance in the L1, except that monolinguals had higher adjusted order scores in free recall. Bilingual performance patterns in the L2 were consistent with the established effects of concurrent task performance on these memory tests, suggesting that the cognitive resources required for processing words in the L2 encroach on resources needed to commit item and order information to memory. These findings are also consistent with a model in which item memory is connected to the language system, order information is processed by separate mechanisms and attention can be allocated differentially to these two systems.     

Maintenance of item and order information in verbal working memory

Although verbal recall of item and order information is well-researched in short-term memory paradigms, there is relatively little research concerning item and order recall from working memory. The following study examined whether manipulating the opportunity for attentional refreshing and articulatory rehearsal in a complex span task differently affected the recall of item- and order-specific information of the memoranda. Five experiments varied the opportunity for articulatory rehearsal and attentional refreshing in a complex span task, but the type of recall was manipulated between experiments (item and order, order only, and item only recall). The results showed that impairing attentional refreshing and articulatory rehearsal similarly affected recall regardless of whether the scoring procedure (Experiments 1 and 4) or recall requirements (Experiments 2, 3, and 5) reflected item- or order-specific recall. This implies that both mechanisms sustain the maintenance of item and order information, and suggests that the common cumulative functioning of these two mechanisms to maintain items could be at the root of order maintenance.     

Encoding, consolidation, and retrieval of contextual memory: differential involvement of dorsal CA3 and CA1 hippocampal subregions

Studies on human and animals shed light on the unique hippocampus contributions to relational memory. However, the particular role of each hippocampal subregion in memory processing is still not clear. Hippocampal computational models and theories have emphasized a unique function in memory for each hippocampal subregion, with the CA3 area acting as an autoassociative memory network and the CA1 area as a critical output structure. In order to understand the respective roles of the CA3- and CA1-hippocampal areas in the formation of contextual memory, we studied the effects of the reversible inactivation by lidocaine of the CA3 or CA1 areas of the dorsal hippocampus on acquisition, consolidation, and retrieval of a contextual fear conditioning. Whereas infusions of lidocaine never impaired elementary tone conditioning, their effects on contextual conditioning provided interesting clues about the role of these two hippocampal regions. They demonstrated first that the CA3 area is necessary for the rapid elaboration of a unified representation of the context. Secondly, they suggested that the CA1 area is rather involved in the consolidation process of contextual memory. Third, they showed that CA1 or CA3 inactivation during retention test has no effect on contextual fear retrieval when a recognition memory procedure is used. In conclusion, our findings point as evidence that CA1 and CA3 subregions of the dorsal hippocampus play important and different roles in the acquisition and consolidation of contextual fear memory, whereas they are not required for context recognition.     

Distinct pattern separation related transfer functions in human CA3/dentate and CA1 revealed using high-resolution fMRI and variable mnemonic similarity

Producing and maintaining distinct (orthogonal) neural representations for similar events is critical to avoiding interference in long-term memory. Recently, our laboratory provided the first evidence for separation-like signals in the human CA3/dentate. Here, we extended this by parametrically varying the change in input (similarity) while monitoring CA1 and CA3/dentate for separation and completion-like signals using high-resolution fMRI. In the CA1, activity varied in a graded fashion in response to increases in the change in input. In contrast, the CA3/dentate showed a stepwise transfer function that was highly sensitive to small changes in input.     

Time course of effects of emotion on item memory and source memory for Chinese words

Although many studies have investigated the effect of emotion on memory, it is unclear whether the effect of emotion extends to all aspects of an event. In addition, it is poorly understood how effects of emotion on item memory and source memory change over time. This study examined the time course of effects of emotion on item memory and source memory. Participants learned intentionally a list of neutral, positive, and negative Chinese words, which were presented twice, and then took test of free recall, followed by recognition and source memory tests, at one of eight delayed points of time. The main findings are (within the time frame of 2 weeks): (1) Negative emotion enhances free recall, whereas there is only a trend that positive emotion enhances free recall. In addition, negative and positive emotions have different points of time at which their effects on free recall reach the greatest magnitude. (2) Negative emotion reduces recognition, whereas positive emotion has no effect on recognition. (3) Neither positive nor negative emotion has any effect on source memory. The above findings indicate that effect of emotion does not necessarily extend to all aspects of an event and that valence is a critical modulating factor in effect of emotion on item memory. Furthermore, emotion does not affect the time course of item memory and source memory, at least with a time frame of 2 weeks. This study has implications for establishing the theoretical model regarding the effect of emotion on memory.     

The Influence of Long-term Memory Factors on Immediate Serial Recall: An Item and Order Analysis

In immediate serial recall, an error can occur because the presented item is not recalled (item error) or because it is recalled at the wrong serial position (order error). Even if these two types of information can be selectively influenced, in most current studies, a global performance measure confounding item and order information is used. Here, the issues associated with the measure of memory for item and order information are discussed. First, it is argued that in some circumstances it is very important that item information be controlled for when measuring order retention, by for example, conditionalizing order memory on memory for item information. Second, using such measures, it is shown that long-term memory factors recently investigated in immediate serial recall produce a different pattern of results than what is predicted by most current models: Semantic similarity, word frequency, and lexicality all influence item recall, but only lexicality affects order information. These findings are discussed in the light of a retrieval-based account suggesting that degraded phonological traces must undergo a reconstruction process calling upon long-term knowledge of the to-be-remembered items.     

In vitro hippocampal gamma oscillation power as an index of in vivo CA3 gamma oscillation strength and spatial reference memory

Neuronal synchronisation at gamma frequencies (30-100 Hz) has been implicated in cognition and memory. Gamma oscillations can be studied in various in vitro models, but their in vivo validity and their relationship with reference memory remains to be proven. By using the natural variation of wild type C57bl/6J mice, we assessed the relationships between reference memory and gamma oscillations recorded in hippocampal area CA3 in vivo and in vitro. Local field potentials (LFPs) were recorded from area CA3 in behaviourally-characterised freely moving mice, after which hippocampal slices were prepared for recordings in vitro of spontaneous gamma oscillations and kainate-induced gamma oscillations in CA3. The gamma-band power of spontaneous oscillations in vitro correlated with that of CA3 LFP oscillations during inactive behavioural states. The gamma-band power of kainate-induced oscillations correlated with the activity-dependent increase in CA3 LFP gamma-band power in vivo. Kainate-induced gamma-band power correlated with Barnes circular platform performance and object location recognition, but not with object novelty recognition. Kainate-induced gamma-band power was larger in mice that recognised the aversive context, but did not correlate with passive avoidance delay. The correlations between behavioural and electrophysiological measures obtained from the same animals show that the gamma-generating capacity of the CA3 network in vitro is a useful index of in vivo gamma strength and supports an important role of CA3 gamma oscillations in spatial reference memory.     

Short-term memory for serial order: a recurrent neural network model

Despite a century of research, the mechanisms underlying short-term or working memory for serial order remain uncertain. Recent theoretical models have converged on a particular account, based on transient associations between independent item and context representations. In the present article, the authors present an alternative model, according to which sequence information is encoded through sustained patterns of activation within a recurrent neural network architecture. As demonstrated through a series of computer simulations, the model provides a parsimonious account for numerous benchmark characteristics of immediate serial recall, including data that have been considered to preclude the application of recurrent neural networks in this domain. Unlike most competing accounts, the model deals naturally with findings concerning the role of background knowledge in serial recall and makes contact with relevant neuroscientific data. Furthermore, the model gives rise to numerous testable predictions that differentiate it from competing theories. Taken together, the results presented indicate that recurrent neural networks may offer a useful framework for understanding short-term memory for serial order.     

Time-limited effects of emotional arousal on item and source memory

Two experiments investigated the time-limited effects of emotional arousal on consolidation of item and source memory. In Experiment 1, participants memorized words (items) and the corresponding speakers (sources) and then took an immediate free recall test. Then they watched a neutral, positive, or negative video 5, 35, or 50?min after learning, and 24 hours later they took surprise memory tests. Experiment 2 was similar to Experiment 1 except that (a) a reality monitoring task was used; (b) elicitation delays of 5, 30, and 45?min were used; and (c) delayed memory tests were given 60?min after learning. Both experiments showed that, regardless of elicitation delay, emotional arousal did not enhance item recall memory. Second, both experiments showed that negative arousal enhanced delayed item recognition memory only at the medium elicitation delay, but not in the shorter or longer delays. Positive arousal enhanced performance only in Experiment 1. Third, regardless of elicitation delay, emotional arousal had little effect on source memory. These findings have implications for theories of emotion and memory, suggesting that emotion effects are contingent upon the nature of the memory task and elicitation delay.     

Multiple mechanisms for recency with vowels and consonants

Three experiments were conducted to test the hypothesis that the difference in recency effect between vowel-contrasting and stop-contrasting lists of syllables in immediate ordered recall can be explained by item discriminability and regular short-term memory mechanisms, without any recourse to echoic memory or precategorical acoustic storage (PAS). In Experiment 1, the short-term memory mechanisms were manipulated by reducing amount of output interference and length of retention interval. The partial-report technique was used. The most important finding was the usual final-position recency effect (difference in recall between the fifth and sixth serial positions) for the vowel lists but not for the stop lists, regardless of the type of report. Thus the PAS theory could not be rejected. In Experiments 2 and 3, the last item was differentiated from the other items of the list, either by lengthening the interstimulus interval between the last and the next-to-last (Experiment 2) or by increasing the intensity of the last item (Experiment 3). In both cases, an increase of the final-position recency effect was found even for stop lists. Since a drop in recall errors was also obtained for the fourth item when its intensity was increased (von Restorff effect), this final-position recency effect for stop lists is likely to be due to item discriminability, and not to echoic memory for the last item. Item discriminability appeared to be the critical factor.     

The beneficial effect of contextual emotion on memory: the role of integration

This study investigates the effects of emotion on the integration mechanism which binds together the components of an event and the relations between these components and encodes them within a memory trace [Versace, R., Vallet, G. T., Riou, B., Lesourd, M., Labeye, É, & Brunel, L. (2014). Act-In: An integrated view of memory mechanisms. Journal of Cognitive Psychology, 26(3), 280–306. doi:10.1080/20445911.2014.892113]. Based on the literature, the authors argue that, in a memory task, contextual emotion could strengthen the integration mechanism and, more specifically, the relations between a target item and its contextual features. To test this hypothesis, the authors used two odorants (neutral and negative) to compare the effects of a negative context with those of a neutral one on three different types of recall: item recall (memory for pictures objects), source recall (spatial position of the pictures in a matrix) and recall of the association between an item and its location. The results showed that, in the negative odour context, association recall and source recall – but not item recall – were better than in the neutral odour context thus confirming the effect of emotion on integration. The results lead to the hypothesis that the effects of emotion on memory are linked to the way emotion is introduced into the experimental settings: via the items to be memorised or via the context.     

Acting or listening: Adult age differences in source recall of enacted and nonenacted statements

This study explored age-related differences in item and source recall of enacted and nonenacted items in a sample of healthy adults between the ages of 35 and 80 years (N=1,000). Participants were screened on a variety of demographic, psychometric, and biological variables. They were presented with sentences (e.g., Lift the book) that they either read or enacted. They were then asked to recall the object of each stimulus sentence (item recall) and to identify its format of encoding (source recall). Age-related decrements were observed both in item and source memory, although age differences in source memory were more accentuated than in item memory. Further, the results indicated an overall impairment of source memory across age when individual differences in the demographic, psychometric, and biological variables were taken into consideration.     

The mnemonic effects of recall on immediate retention

In three experiments, we investigated the mnemonic effects of an initial recall on later recall in an immediate memory setting. Recall is generally assumed to interfere with the recall of subsequent items (output interference), but previous experiments have failed to control for the confounding effects of time. In the experiments reported here, the passage of time was held constant on all trials; what varied was whether an additional item was recalled (or simply presented) during the retention interval. The results revealed clear evidence of recall's mnemonic effects, but output interference seemed strongest when the initial recall was of an item that followed the target item in the memory list. When participants initially recalled an item immediately preceding the target, target recall improved. This pattern of results places constraints on current models of immediate retention.     

Adult age differences in temporal and item memory

Three experiments tested whether the relationship between age differences in temporal and item memory depends on the degree to which the item memory measure relies on memory for context. The authors predicted a stronger relationship of temporal memory to free recall than to recognition memory. Results showed that age differences in temporal memory could be eliminated after controlling for free recall but not recognition memory performance. Under some conditions recognition memory accounted for a significant portion of age-related variance in temporal memory. These results challenge past research that has interpreted age differences in temporal and item memory as independent and suggest that a generalized decline in context memory may underlie reduced performance in older adults on all types of memory tests.     

Memory for Related and Unrelated Words: Further Evidence on the Influence of Semantic Factors in Immediate Serial Recall

A number of recent studies have explored the role of long-term memory factors in memory span tasks. The effects of lexicality, frequency, imageability, and word class have been investigated. The work reported in this paper examined the effect of semantic organization on the recall of short lists of words. Specifically, the influence of semantic category on immediate serial recall and the interaction of this variable with articulatory suppression was investigated in three experiments. Experiment 1 compared immediate serial recall performance when lists comprising items from the same semantic category were used (homogeneous condition) with a situation where lists held items from different semantic categories. Experiment 2 examined the same conditions with and without articulatory suppression during item presentation, and Experiment 3 reproduced these conditions with suppression occurring throughout presentation and recall. Results of all three experiments showed a clear advantage for the homogeneous condition. Experiments 2 and 3 showed that the homogeneous category advantage did not depend on the articulatory loop. Furthermore, error analysis indicated that this effect was mainly attributable to better item information recall for the homogeneous condition. These results are interpreted as reflecting a long-term memory contribution to the recall stage of immediate serial recall tasks.     

Visual and phonological components of working memory in children

Previous studies have shown that young children's immediate memory for a short series of drawings of objects is mediated by a visual component of working memory, whereas older children rely chiefly upon a phonological component. Three experiments investigated the hypothesis that older children rely also, but to a lesser extent, on visual working memory. Experiment 1 confirmed previous evidence that 11-year-olds' memory is disrupted by phonemic similarity of object names, but is unaffected by visual similarity of the objects themselves. However, when articulatory suppression was used to prevent phonological coding, levels of recall were sensitive to visual rather than phonemic similarity. Experiment 2 compared the effects of interpolating an auditory-verbal or a visual postlist task on memory for drawings viewed either with or without suppression. The visual task had a clear disruptive effect only in the suppression condition, where it interfered selectively with recall of the most recent item. Experiment 3 compared the effects of interpolating an auditory-verbal or a mixed-modality (visual-auditory) postlist task when subjects were not required to suppress. There was greater interference from the mixed-modality task, and this effect was confined to the last item presented. These experiments are taken as confirming the presence of a small but reliable contribution from visual memory in 11-year-old children's recall. As in younger children, visual working memory in 11-year-olds is sensitive to visual similarity and is responsible for a final-item visual recency effect. The results also show that older children's use of visual working memory is usually masked by the more pervasive phonological component of recall. Some implications for the structure of working memory and its development are discussed.