Single-trial Learning Research Articles

Chapter 3 Episodic memory in semantic dementia: a computational approach based on the TraceLink model

This chapter discusses the recent empirical studies of long-term memory in semantic dementia. The chapter describes how one model of long-term memory, TraceLink, can account for semantic dementia. The chapter discusses a number of predictions related to the disorder, which can be tested using computational simulations. The chapter shows a TraceLink model and its three main components are (1) a trace system, (2) a link system, and (3) a modulatory system. The main purpose of TraceLink model is to explain some of the most prominent phenomena in the neuropsychology of amnesia and semantic dementia. The trace system (neocortex) holds the bulk of the memory representations. It has a large capacity but is not able to store a new episodic memory in a single learning trial: it suffers from slow learning. The chapter also focuses on the effects of lesions of the trace system, which are able to explain several of the observed characteristics of semantic dementia.

Rapid, Labile, and Protein Synthesis– Independent Short-Term Memory in Conditioned Taste Aversion

Short-term memory is a rapid, labile, and protein-synthesis-independent phase of memory. The existence of short-term memory in conditioned taste aversion (CTA) learning has not been demonstrated formally. To determine the earliest time at which a CTA is expressed, we measured intraoral intake of sucrose at 15 min, 1 hr, 6 hr, or 48 h after contingent pairing of an intraoral infusion of 5% sucrose (6.6 ml over 6 min) and toxic lithium chloride injection (76 mg/kg). Rats were implanted with intraoral catheters to allow presentation of taste solutions at arbitrary times. Intraoral intake was measured under conditions of long-delay, single-trial learning typical of CTA. Rats decreased intraoral intake of sucrose at 15 min after contingent pairing of sucrose and LiCl, but not after noncontingent LiCl or sucrose. Thus CTA learning can be expressed rapidly. To determine if short-term CTA memory is labile and decays in the absence of long-term memory, we measured intraoral intake of sucrose after pairing sucrose with low doses of LiCl. Rats received an intraoral infusion of 5% sucrose (6 ml/6 min); 30 min later LiCl was injected at three different doses (19, 38, or 76 mg/kg). A second intraoral infusion of sucrose was administered 15 min, 1 hr, 3 hr, 4.5 hr, 6 hr, or 48 hr later. The formation of long-term CTA memory was dependent on the dose of LiCl paired with sucrose during acquisition. Low doses of LiCl induced a CTA that decayed within 6 hr after pairing. Central administration of the protein synthesis inhibitor cycloheximide prior to LiCl injection blocked long-term CTA expression at 6 and 48 hr, but not short-term CTA expression at 1 hr. Thus, short-term memory for CTA learning exists that is acquired rapidly and independent of protein synthesis, but labile in the absence of long-term memory formation.

Evidence for a specific short-term memory in the cuttlefish, Sepia

Cuttlefish quickly learn to inhibit their predatory motor pattern when shown prawns in a glass tube. The available literature suggests that cuttlefish show an excellent retention between 2 and 8 min, a recovery of the predatory responses around 20 min and good retention after 1 h of the training phase. These results have been considered as the product of two separate short- and long-term memory stores. In this study, we have investigated the fact that the retention seen after a brief delay of the training phase corresponds to a true effect of learning. We compared animals under three experimental conditions. In two, there was a unique training trial of different duration (5 or 20 min), a third group served as controls. Our results demonstrate that the control situation failed to reduce the level of attack; in contrast the short-term retention, obtained after a single learning trial, is related to a specific short-term memory process.

Psychoneuroendocrine immunology: site of recognition, learning and memory in the immune system and the brain

How the interaction between the brain and immune system takes place has not been clearly defined. Because multiple changes are occurring simultaneously in all organ systems (e.g., cardiovascular, gastrointestinal, reproductive, renal, respiratory, immune, CNS), how many single systems interacts with the brain becomes extraordinarily difficult to understand. The problem boils down to developing an approach that not only allows one to study the whole organism and define the mediators of the interacting systems, but also permit one to establish the connection and physiologic relevance of the responses that are being evaluated. Conditioning, a phenomenon made popular by the work of Pavlov (1906, 1927), may provide insight into the pathways of communication between the brain and possibly any organ system of the body. Conditioning allows one to separate the afferent from the efferent circuits. That is, signals from the immune system to the CNS (IS→CNS) can be effectively separated from signals from the CNS to immune system (CNS → IS). This permits one to study each pathway individually. Simple, single association trial models to condition fever, natural killer (NK) cell and cytotoxic lymphocyte (CTL) activities have been developed to evaluate the pathways. Single trial learning is not new. Pavlov has observed that "The electric buzzer set going before administration of food established a conditioned alimentary reflex after only a single combination," whereas the reverse order of presentation failed to condition the animal (Pavlov 1927 p. 27). Thus, conditioning can be used to train the brain to activate the immune system and other organ systems participating in the response. During the course of the conditioned response, presumably the CNS via the hypothalamus integrates in a cohesive orderly fashion all input and output signals and coordinates the responses made by the brain to the organ systems. The odor of camphor, the conditioned stimulus (CS) can be associated with the response produced by an unconditioned stimulus (US). The unconditioned stimuli used are poly I:C to raise fever and nonimmunospecific NK cell activity or alloantigens to raise immunospecific CTL activity. The unconditioned stimulus serves only as a means to activate the immune system and unbalance the homeostasis so that a transient but new bidirectional communication loop can be established between the immune system and the CNS (IS↔CNS). The expression of the conditioned response (i.e., elevation of fever, NK cell, or CTL activity) induced with the CS (odor stimulus) is an outcome of neural activity (CNS → IS). This infers that during conditioning, the signals generated by the CS and US imprints a neural pathway located within the central nervous system and leaves behind a CS/US memory of the association. The immune activity (NK cell or CTL activity) which is modulated indicate that the memory pathway was activated in the brain of the animal expressing the conditioned response. The immune cells that are modulated can be considered to be casual bystander cells. These cells however must be in the proper (ready) state of activation to receive salient signals from the brain. Along with changes in the indicator cell population, other complex physiological processes are altered by the brain via sympathetic and neuroendocrine pathways to raise the fever response. These observations suggest that the physiological changes which are being evaluated such as fever, NK cell or CTL activities or perhaps blood pressure, heart rate, fat metabolism, oxygen consumption serve only as indicators (readouts), and infer that the CNS has made a coordinated reply in response to the CS signal.

The Effects of Response Repetitions on Sight Word Acquisition for Students with Mild Disabilities

Single response learning trials, defined in terms of the three-term contingency model, were compared with repeated response learning trials. The purpose of Study 1 was to assess the effectiveness and efficiency of single and repeated response learning trials on sight word remediation. In Study 1, one single response learning trial having one response opportunity per word was compared to one repeated response learning trial having five response opportunities. In Study 2, three single response trials were compared to three repeated response trials. In both studies, the training conditions were compared using an alternating treatments design. Measuring the cumulative number of sight words mastered showed the training conditions equally effective. However, measuring (a) the mean duration of training sessions, (b) the cumulative frequency of words mastered per session duration, (c) the mean number of training sessions per word required for mastery, and (d) the training time per mastered word showed that the single response condition was more efficient. Discussion focuses on the need to include training data when determining best practices for teaching.

Modelling human memory: Connectionism and convolution

The mathematical operation of convolution is used as an associative mechanism by several recent influential models of human memory. Convolution can be used to associate two vectors (representing items to be remembered) into a memory trace vector in one operation. An approximation to either of the input vectors can then be retrieved, using the other vector as a probe. Recent convolution-based memory models have accounted for a wide range of data. Connectionist models may have greater potential for providing developmental accounts, but the architectures that have been most widely used to account for developmental phenomena cannot perform one-trial learning and this has limited their use as models of human memory. We show that a connectionist-like architecture can learn, using a gradient-descent algorithm, to perform single-trial learning in a similar manner to convolution. The solution that the network finds leads to less variable retrieval than does convolution. Furthermore, the network can learn to carry out the convolution operation itself. This provides a link between connectionist and convolution approaches, and a basis for models with many of the attractions of both connectionist and convolution approaches.

Inhibition of Nitric Oxide Synthase Impairs a Distinct Form of Long-Term Memory in the Honeybee, Apis mellifera

Nitric oxide has been shown to be implicated in neural plasticity that underlies processes of learning and memory. In the honeybee, studies on the role of nitric oxide in associative olfactory learning reveal its specific function in memory formation. Inhibition of nitric oxide synthase during olfactory conditioning impairs a distinct long-term memory that is formed as a consequence of multiple learning trials. Acquisition or retrieval of memory or memory formation induced by a single learning trial is not affected by blocking of nitric oxide synthase. This finding provides a first step toward dissection of molecular mechanisms involved in memory formation, in general, and the special function of nitric oxide synthase, in particular.

Effects of task demands on implicit memory for object-location associations.

Abstract The experiments examined the effect of task demands on implicit and explicit memory for object - location associations. Objects (letters or symbols) were probed one at a time in one of nine locations in a rectangle, and the required task was either location or object naming. Each object was probed in the same location across ten trial blocks and then all the objects changed locations. A decrease in naming times across the first 10 trial blocks followed by an increase in naming times when the objects change locations would indicate priming for the associations. The experiments demonstrated that priming could be obtained when the task engaged, but did not overtax attentional capacity. Explicit memory performance was best when the task did not require much attentional capacity. This paper focuses on implicit and explicit memory for new nonverbal associations. Explicit memory refers to memory for an event which relies on conscious or intentional recollection. Implicit memory refers to memory for an event that is evidenced by a change in performance as a result of prior experience, but which need not be accompanied by conscious or intentional recollection of the prior episode. Although most forms of implicit memory occur across multiple trials, one type of implicit memory, the priming effect, is observed after a single learning trial. Priming refers to a change in the facility with which a stimulus is processed as a result of its recent exposure. The priming effect occurs for both familiar and novel material as long as the stimuli are presented for study singly (e.g., a single word, nonword, or nonverbal stimulus; see Bowers & Schacter, 1993, for a review). When two stimuli are presented simultaneously (e.g., when two words are presented together), such that an association must be formed, associative priming effects have not consistently observed. With normal populations priming of word pair associations has been observed in word stem completion tests after elaborative study tasks, but not after nonelaborative tasks (e.g., Graf & Schacter, 1985; Schacter & Graf, 1986). Goshen - Gottstein and Moscovitch (1995a; 1995b) reported evidence for associative priming using a lexical decision task when there was a perceptual match between study and test environments. However, Musen and Squire (1993a) reported no associative priming in a perceptual identification task. Associative priming effects with memory impaired populations have also been inconsistent. Moscovitch, Winocur, and McLachlan (1986) reported that associative priming was possible using a reading speed paradigm, while other researchers did not find evidence for associative priming (Mayes & Gooding, 1989; Shimamura & Squire, 1989; Musen & Squire, 1993a). The difficulty in obtaining priming for new associations may stem from the fact that the two elements must be unitized into a single item (Graf & Schacter, 1989), and the likelihood of successful unitization may depend on specific stimulus and task situations. Several findings suggest that associations can be more readily learned if elements to be associated are part of the same object (e.g., an association between a color and a shape could appear together as a colored shape). Musen and O'Neill (1994) reported that participants could learn associations between colors and words, and between colors and shapes, when a) the two elements appeared as a single item (i.e., a colored word or colored shape); and b) attention was paid to the association between the two. In contrast, in a study by Treisman (1992), in which the same colored line patterns were studied as in the Musen & O'Neill (1994) study, but color was irrelevant, only pattern priming was observed. Thus, it is possible that priming of the associations requires that attention be paid to both elements, and this unitization process may be facilitated when the two elements to be associated are present in a single object. …

Oxytocin mediates acquisition of maternally associated odor preferences in preweanling rat pups.

Three studies assessed the role of central oxytocin systems in maternal affiliation in preweanling rats. Fifteen-day-old rats were found to approach on odor faster and spend more time in contact with an odor when it had been paired with the mother on the previous day than when the odor was not paired with the mother, provided that the pairing occurred after either intracerebral administration of oxytocin (0.5 microgram) or saline. However, if rats were administered an oxytocin antagonist (OTA) prior to odor-mother pairing, neither the decreased odor approach latency nor the increased odor preference was apparent in the odor-mother group. OTA treatment did not attenuate a single-trial odor aversion learning task, nor did it disrupt interaction with the anesthetized mother. Results are discussed in terms of a possible reward-satiety role for oxytocin in the infant-mother context.

Mathematical and Connectionist Models of Human Memory: A Comparison

Recent convolution-based models of human memory (e.g. Lewandowsky & Murdock, 1989), have accounted for a wide range of data. However such models require the relevant mathematical operations to be provided to the network. Connectionist models, in contrast, have generally addressed different data, and not all architectures are appropriate for modelling single-trial learning. Furthermore, they tend to exhibit catastrophic interference in multiple list learning. In this paper we compare the ability of convolution-based models and DARNET (Developmental Associative Recall NETwork), to account for human memory data. DARNET is a connectionist approach to human memory in which the system gradually learns to associate vectors, in one trial, into a memory trace vector. Either of the vectors can than be retrieved. It is shown that the new associative mechanism can be used to account for a wide range of relevant experimental data as successfully as can convolution-based models with the same higher-level architectures. Limitations of the models are also addressed.

Behavioural impairment induced by calcitonin gene-related peptide (CGRP) antiserum in passive avoidance reflex in rats

The effects of CGRP antiserum on passive avoidance behaviour were tested in rats. Intracerebroventricular (i.c.v.) administration of CGRP antiserum immediately after a single learning trial attenuated the passive avoidance responses when the animals were tested for 24 h retention. Significant decreases in avoidance latency were found at CGRP antiserum dilutions of 1:20 and 1:40. The data supported the assumption that endogenous brain CGRP may be a modulator in learning-associated memory processes.

Influence of manipulated category knowledge on prototype classification and recognition

The recognition and classification of category members was explored, following a variable number of learning trials. In Experiment 1, subjects received 1 or 9 learning trials, followed by a recognition-then-classification test containing old, new, prototype, and foil patterns. In Experiment 2, subjects received 1, 6, or 12 trials, and made either classification or recognition judgments. In each experiment, classification accuracy for all item types was at near-chance performance after a single trial but steadily increased with increased learning trials. On the transfer test, oldness judgments were highest for the category prototype after a single trial. However, with increased learning trials, oldness judgments increased for old instances and decreased for the category prototype and new instances. We suggest that false recognition of the category prototype, especially after a single learning trial, need not reflect an abstraction process. We discuss the possibility that an abstracted prototype may emerge with additional learning as an unfamiliar, ideal point.

The central amygdala is involved in conditioning but not in retention of active and passive shock avoidance in male rats

The present study was designed to further characterize the involvement of the central amygdaloid nucleus (CEA) in the conditioning and the retention of both active and passive stress-related behaviors. Active shock avoidance was investigated in a shuttle-box. CEA-lesioned and sham-operated animals received 30 trials on 2 consecutive days. During the first day of conditioning, the CEA-lesioned animals failed to learn to avoid, but on the second day these rats showed avoidance behavior that was comparable to the sham-lesioned ones. When lesioned animals received a total of 60 trials on 1 day, they failed to show avoidance, suggesting deficits in acquisition performance but presence of latent learning following CEA lesioning. However, the retention response, which was studied by lesioning the CEA between a 2-day avoidance training and a retention test, was unaffected. Stress-related learning was also studied in a test of passive avoidance of a previously received single, inescapable electric footshock. When the lesion was placed before the single learning trial, the retention of the response was absent. However, CEA-lesioning after the learning trial revealed no differences between the CEA-lesioned and sham-lesioned animals. In conclusion, these results demonstrate that the CEA is involved in the conditioning rather than retention of both active and passive shock avoidance responses.

Central Amygdaloid Involvement in Neuroendocrine Correlates of Conditioned Stress Responses

The purpose of this study was to examine the effects of bilateral electrolytic lesions of the central nucleus of the amygdala (CEA) in comparison with sham lesions on neuroendocrine responses during conditioned emotional stress in male Wistar rats. Lesions in the CEA, made either before or after the single learning trial of inescapable footshock, failed to affect the conditioned response of plasma epinephrine levels. Plasma levels of norepinephrine showed neither a conditioned stress effect nor were influenced by lesioning. Pre-training CEA lesions, but not post-training intervention, abolished the conditioned elevations of circulating plasma corticosterone and prolactin. These results suggest that the CEA is involved in the conditioning rather than the retention of neuroendocrine stress responses. The effects of pre-training lesioning of the CEA can possibly be explained by a reduced feedback of all these neuroendocrine factors during or shortly after acquisition. In addition, there is a remarkable differentiation between various hormonal correlates of conditioned stress following CEA lesioning. Only corticosterone and prolactin, that appear to be correlates of a passive behavioural stress response, were abolished. The lesions failed to affect the sympatho-active stress parameters (epinephrine and norepinephrine). Relations between coping strategy-active and passive behaviour-and physiology in connection with CEA functioning are discussed.

Forgetting curves in long-term memory: Evidence for a multistage model of retention

An experimental study with 20 normal healthy young adult subjects was performed to evaluate the interaction of type of memory tasks, type of learning modalities, and length of acquisition/recall interval. Four different tasks were employed (serial learning, paired learning, rote learning, and visuolinguistic transfer), some requiring a single trial learning modality others a multitrial learning modality. Acquisition/recall intervals were immediate, intermediate (3 min), and delayed. The experimental design allowed for the comparison of effects from five different delayed recall intervals (2, 8, 12, 24, and 48 hr). Results demonstrated a specific interaction on learning rates due to different ceiling effects for the different types of memory tasks. Forgetting rates, on the other hand, demonstrated a specific effect due to type of memory tasks and learning modalities only for differences between immediate and intermediate recall. These differences remained stable during the longer intervals and were not affected by length of interval. A multistage composition of long-term retention was suggested to explain these results, and a practical indication to build experimental procedures to study memory in the clinical field was evidenced.

The Olfactory Memory of the Honeybee Apis Mellifera: I. Odorant Modulation of Short- and Intermediate-Term Memory After Single-Trial Conditioning

ABSTRACT In the first 15 min after a single learning trial the olfactory memory of the honeybee, Apis mellifera, proceeds through different processing phases during which time the memory is differentially sensitive to a cooling treatment that causes amnesia. During learning about floral odours in a natural situation, several decisions would normally be made about floral choice within that period. In order to study these phenomena in more detail, single-trial proboscis extension conditioning to different odorants was used. Several stimulus-specific effects on memory consolidation in the honeybee are shown. From previous experiments it was predicted that certain odorants would be more salient conditioning stimuli. This result is confirmed. Second, generalization from the conditioned odorant to a different odorant depends on the conditioned odorant and the time postconditioning. In some combinations, responses to a novel odorant are significantly stronger than responses to the conditioning odorant after memory consolidation. These data indicate that memory recall in the honeybee, as it is evidenced by proboscis extension, is sensitive to several aspects of stimulus identity and presentation. The acquisition and recall processes are therefore much more dynamic processes than realized previously.

The functional relevance of the area postrema in drug-induced aversion learning

Research into the neural mechanisms involved in the acquisition of learned aversions induced by drug points toward the area postrema (AP) as one of the structures implicated in the detection of drug aversive consequences. The evidence suggest that although the AP is indeed involved in drug-induced learned aversions, its functional integrity is not always a necessary requisite for learning to take place. The aim in this study was to determine whether the AP is essentially or selectively involved in all learned aversions induced by scopolamine methyl nitrate (SMN) using different number of trials with the aversive stimulus. In Experiment 1, AP-lesioned rats were injected with SMN fifteen minutes after consuming a flavoured solution during three consecutive trials. A single-stimulus test failed to detect learned aversions, which were, however, evident in two subsequent choice-tests. In one-trial paradigms, however, choice-tests as well as single-stimulus tests failed to detect learned aversions in AP-lesioned rats, both when SMN was injected immediately after stimulus intake (Experiment 2) and when a fifteen-minute delay was introduced (Experiment 3). The results suggested that the AP is not essential for the acquisition of SMN-induced aversion learning with three consecutive trials if learning is detected with a choice-test, although effective single-trial learning does apparently require a functional AP.

Methyl β-carboline-3-carboxylate enhances performance in a multiple-trial learning task in mice

In contrast to diazepam, a benzodiazepine receptor (BZ-R) ligand, which impairs memory processing, methyl β-carboline-3-carboxylate (β-CCM), another BZ-R ligand, administered before a training session, enhances performance in a retention test. This action, however, has only been demonstrated in single trial or single session learning protocols. The present report extends these results to a multiple-trial learning procedure in mice (brightness discrimination in a T-maze with negative reinforcement). The animals were trained for sessions of ten trials per day for six consecutive days. In a first experiment, the sessions during the first three days took place after administration of β-CCM (0.3 mg/kg), diazepam (2.5 mg/kg) or saline. In a second experiment, especially designed to study effects of β-CCM, during the first three days animals received β-CCM (0.3 mg/kg), Ro 15-1788 (15 mg/kg), β-CCM + Ro 15-1788, vehicles of these drugs or saline. In the first experiment, performance was improved by β-CCM and impaired by diazepam in the first three sessions as well as in the final three. In the second experiment, β-CCM alone, as well as Ro 15-1788 improved performance, and the simultaneous administration of the two drugs suppressed these effects. These results suggest that the performance enhancing effects of β-CCM observed in single trial learning protocols, during the retention test, can already be observed during drug treatment. They confirm that β-CCM has an action on acquisition (learning). As the effects of β-CCM are suppressed by the simultaneous administration of Ro 15-1788, our results could suggest a role for benzodiazepine receptors in learning. This question is discussed.

Time-course of memory formation differs in honey bee lines selected for good and poor learning

Six lines of the honey bee Apis mellifera capensis were selected for good and poor learning, and compared with respect to the time-course of their memory after a single learning trial to an odour stimulus (geraniol). The good learners showed the well known bi-phasic time-course with a high rate of conditioned responses immediately after learning, followed by a reduction 1–3 min later and a consolidation into a long-term memory after an interval longer than 5 min. Poor learners differed significantly with respect to their consolidation period and the initial high response rate. The initial period (of less than 1 min) is controlled not only by the associative memory of the learning trial but also by a sensitizing effect of the sucrose stimulus which was used as an unconditioned stimulus. It is concluded that the lines selected for learning performance differ with respect to the establishment of a long-term memory trace. As in Drosophila learning mutants, differences between good and poor learners are also affected by non-associative processes involved in sensitization.

Modulation of the immune response by emotional stress

The influence of mild, emotional stress was investigated for its effect on the immune system by subjecting rats to the one-trial-learning passive avoidance test. The reactivity of the immune system was tested by determining the proliferative response after mitogenic stimulation in vitro as well as the capacity to generate a primary antibody response in vivo after immunization with sheep red blood cells. Our results demonstrate that exposure of rats to a single electric footshock (learning trial) or habituation to the passive avoidance apparatus, induces an increase of the immune response in vitro and in vivo . Thus, emotional stimuli seem to facilitate immunological responsiveness. However, when the animal is confronted with a conflict situation, as tested by the retention of the avoidance response after a single learning trial, the initially enhanced reactivity of the immune system decreases. It is concluded that the immune system is capable of reacting specifically and immediately to distinct psychological stimuli.