Associative Learning - an overview | ScienceDirect Topics
Excerpt
You might find these chapters and articles relevant to this topic.
Chapters and Articles
You might find these chapters and articles relevant to this topic.
Cerebellum: Associative Learning
K.M. Christian, in Encyclopedia of Behavioral Neuroscience, 2010
Introduction
Associative learning is the process through which organisms acquire information about relationships between events or entities in their environment. It is expressed as the modification of existing behaviors, or the development of novel behaviors, that reflects the conscious or unconscious recognition of a contingency. It is the contingent, and contiguous, relationship among stimuli that is a hallmark of associative learning â a meaningful temporal or spatial proximity of A and B and the perceived consequent occurrence of B if A. As such, it is fundamental to our sense of causality and is the basis of much of our understanding of the external world. Associative learning also underlies the majority of our adaptive behavior when the association is recognized to have either positive or negative consequences. Adaptive changes in behavior can be triggered by both aversive and appetitive stimuli and can thus enable the organism to avoid negative outcomes or to increase the probability of obtaining a reward. This type of associative learning depends on the presence of signaled reinforcement.
However, there are many forms of associative learning, and the brain regions that support the acquisition and expression of these learned behaviors are determined by the nature of the information acquired and the response itself. Certainly, sensorimotor systems are required, first, to transduce the information related to the critical stimuli and then, subsequently, to perform the sequence of actions reflective of the memory, but it is the integrative neural trace of the association itself that is the focus of learning-related research. Several structures have emerged as being the critical loci of plasticity underlying specific types of associative memory. For example, emotion-based memory often involves the amygdala and the amygdalar nuclei are critical for the acquisition of conditioned fear â a process wherein neutral stimuli become associated with a fearful event. Similarly, the medial temporal lobe is involved in some forms of associative learning as well as long-term storage of associative information. Structures within this region show changes in neural activity at time points that precede and/or coincide with the behavioral expression of newly acquired associations between visual stimuli, as well as novel spatial and temporal relationships between task-relevant environmental features. But, perhaps, the most exhaustive and complete description of the neural instantiation of memory in the mammalian brain is that involving a simple form of associative memory localized to the cerebellum â namely, classical conditioning of discrete reflexes to behaviorally neutral stimuli. That the cerebellum is involved in cognitive processes at all was met with some resistance originally, although it is now widely accepted that it is the essential neural substrate for the acquisition and expression of this associative memory.
Neurocognitive Development: Normative Development
Marc Philippe Lafontaine, ⊠Sarah Lippé, in Handbook of Clinical Neurology, 2020
Associative learning in language acquisition
Associative learning is an essential mechanism in early language development, since it allows for the ability to pair a concept to a word (Tsui et al., 2019). Infants are first able to detect individually presented words by the age of 6 months, as assessed in an experiment by Bergelson and colleagues in which infants were presented with a set of pictures while their parent named one of the depicted items. Infants between 6 and 9 months of age directed their gaze toward the named picture, indicating successful association of the word with the picture (Bergelson and Swingley, 2012). Though extensive variability exists in early language development, infants are generally able to understand dozens of words and produce a few by 1 year of age (Fenson et al., 1994). A popular task to investigate associative word learning is the switch task, in which toddlers habituate to novel object-word pairings, followed by a test phase in which their capacity to notice a switch in the pairings is evaluated (Werker et al., 1998). In a recent meta-analysis, Tsui and colleagues found that associative word learning, as evaluated by the switch task, did not improve between 12 and 21 months of age, suggesting that the ability is in place around 1 year of age (Tsui et al., 2019). The theory of fast mapping accounts for rapid word learning in the second year of life. It describes the capacity of toddlers to learn a new word through a single, brief exposure (Carey and Bartlett, 1978), starting at around 14 months of age (Werker et al., 1998). Though the validity of fast mapping has been debated (McMurray et al., 2012), it has been suggested as a more general learning mechanism, going beyond language acquisition (Rovee-Collier and Giles, 2010). Rovee-Collier and Giles suggest that fast mapping explains the period of exuberant learning and fast forgetting in the first third of infancy, which is replaced by more mature declarative memory systems toward the end of infancy (Rovee-Collier and Giles, 2010).
Neurocognitive Development: Normative Development
Marc Philippe Lafontaine, ⊠Sarah Lippé, in Handbook of Clinical Neurology, 2020
Associative learning
Classical and operant conditioning
If a major function of the brain is to allow efficient processing of the environment using models such as Bayesian inference and PC, then encoding of unitary items and novelty detection through repetition are insufficient. Indeed, in order to generate expectations or predictions of incoming stimuli, an ability to construct reliable and sensible models of complex environments is necessary. In turn, this ability must rely on mechanisms enabling the encoding of associations or relationships between items and events. Associative learning is defined as learning about the relationship between two separate stimuli, where the stimuli might range from concrete objects and events to abstract concepts, such as time, location, context, or categories. The most basic form of associative learning is classical conditioning, in which an unconditioned stimulus (US) leading to an unconditioned response (UR) (e.g., a simple reflex response such as salivation) is paired with a conditioned stimulus (CS). After several (or even just one) pairing of US and CS, the presentation of the CS alone leads to a conditioned response (CS) that resembles the UR, though it is a learned rather than a reflexive response. Similarly, operant conditioning involves the pairing of a certain behavior with a reward or punishment to strengthen or decrease said behavior. In humans, the scope of associative learning can be broadened to include a vast array of relational connections. In fact, in contemplating everyday life, one realizes that the majority of what we are able to accomplish, understand, and remember relies on an ability to associate several items in space and time, allowing us to construct and store complex scenes and sequences of events.
Developmental perspective
As mentioned earlier, classical and operant conditioning can be found across species, even in relatively simple-level organisms. Considering this, and knowing that the human infantâs brain is already very complex when it enters the world, it is unsurprising that classical conditioning has been found in newborn humans as early as 2Â h after birth (Blass et al., 1984). Despite the simplicity of the classical conditioning mechanism, it appears to be a building block of child development. Reeb-Sutherland and colleagues found learning rate during a delayed eyeblink conditioning paradigm at 1 month of age to predict social behaviors, but not cognitive abilities, at 5, 9, and 12 months of age (Reeb-Sutherland et al., 2012). Operant conditioning has a scope of application in pediatrics for the treatment of swallowing and feeding disorders (Gosa et al., 2017). For example, Chorna and colleagues successfully used a pacifier-activated music player to improve oral feeding in preterm infants (Chorna et al., 2014). Operant conditioning remains an important learning mechanism throughout infancy and the entire lifespan and some even argue that it is central to language acquisition, though this has been heavily debated (Sturdy and Nicoladis, 2017). In contrast to other, more complex forms of learning, operant conditioning in infants under 4 months of age has been found to be unaffected by socioeconomic status (Gerhardstein et al., 2012).
More complex forms of associative learning
The questions of how more complex associative and relational learning develops and when it can first be observed have been given much attention in infant research. Of course, to relate different stimuli to one another, infants must first develop an ability to reliably parse different items in their visual field. Such an ability develops over the first few months of life and relies at first on innate mechanisms of figure-ground segregation (Arcand et al., 2007) and later integrates more experience-dependent abilities (e.g., Johnson, 2001). Along with rapid brain maturation and the development of perceptual prerequisites, more complex forms of associative learning arise during the first years of life.
Associative learning in categorization
In very young infants, research has focused on processes of categorization. It is commonly accepted that associative learning contributes to infantsâ category acquisition, though it is debated how much it contributes, as opposed to an innate understanding of certain concepts as proposed by psychologic essentialism (Gelman and Meyer, 2011). Either way, categories are an essential building block of learning that are refined throughout infancy and especially in the first year of life. By adapting preferential looking paradigms, such as the familiarization-novelty preference, from habituation to association, researchers were able to investigate categorization in infants as young as 3 months of age. After infants aged 3â4 months were habituated to a series of cat photographs, the infants thereafter preferred to look at the photo of a dog when presented with the choice between a novel cat or dog photograph (Quinn et al., 1993), leading researchers to believe that by 3 months of age infants are able to form perceptual categories. However, when the order of the experiment was reversed, i.e., infants were habituated to dogs, they showed no novelty preference for the cat picture in the test phase, showing that categories in young infants are asymmetric and differ from those in adults (Mareschal et al., 2000). In their connectionist account, Mareschal and colleagues expand on the comparator model to explain asymmetries in infant categorization. In short, the extent of looking time to a novel stimulus decreases the more similar the stimulus is to the existing internal representation of a familiar object, explaining why categorization is easier for more similar (cats) than heterogeneous (dogs) stimuli (Mareschal et al., 2000). Soon infants move on from feature-specific categories to representations of feature combinations (Younger, 1985).
Another developmental shift is found between 3 and 6 months of age, showing that even though the younger group was already capable of detecting and remembering feature correlations, only the 6-month-old infants used correlated attributes to categorize new objects (Bhatt et al., 2004). Entering the second year of life, categories become central to language acquisition. Though category learning is involved in nearly all aspects of cognition, it is particularly entwined with language acquisition, in such a way that language shapes category learning and vice versa (Gelman and Meyer, 2011). Specifically, words children learn imply a vast number of categories; language allows us to access their understanding of categories; and, finally, language can be used to transfer category-relevant information (Gelman and Meyer, 2011). Throughout toddlerhood, perceptual and verbal categories become more refined. With growing understanding of the world, concepts become more important than percepts in the formation of categories and categories become more specific (Mandler and McDonough, 1998). Starting at around 2 years of age, toddlers categorize based on function rather than appearance if the function is plausible and distinctive (Kemler Nelson et al., 2000). At preschool age, children finally display as much range and flexibility in their categories as adults (Gelman and Kalish, 2006).
Generalization of associative learning
Starting in the second half of the first year of life, infants are able to learn associations across categories, i.e., to form relational memories. Using a visual associative learning procedure in which they presented scenes of colored shapes in different configurations (context) in combination with a cartoon icon at a specific target location, Bertels and colleagues showed that 8â12-month-old infants were able to learn the association between target location and the specific context of the presentation (Bertels et al., 2017). At 9 months of age, infants were able to successfully associate faces to simultaneously presented landscapes; however, the associations were only maintained for a short period of time (Richmond and Nelson, 2009). One-year-olds were able to associate a person with the object they were looking at, but could not anticipate future actions based on their object-directed gaze (Paulus, 2011). Generalizing associations learned in one context to another is an important aspect of associative learning. In a visual habituation-novelty preference task focused on element pair groupings, Kangas et al. (2011) found a developmental shift between infants aged 3â4 and 6â7 months. Whereas the older group was able to generalize grouping based on common region to grouping based on proximity, younger infants failed to do so, even though they were able to process correlations (Kangas et al., 2011). Along with the development of declarative memory and corresponding changes in hippocampal functioning during the second year of life, the ability to remember and generalize relational associations improves (Hayne et al., 2000; Sluzenski et al., 2004). From 12 to 21 months, childrenâs capacity to generalize cues when accessing memories for a prior event substantially increases (Hayne et al., 1997). Relational memory continues to develop throughout preschool years and reaches adult levels at around 6 years of age (Sluzenski et al., 2006). However, some forms of associative learning, such as acquired equivalence, a feedback-based paradigm in which the equivalence of two or more stimuli has to be learned in terms of outcomes, appear to develop well into adulthood (Braunitzer et al., 2017). Specifically, the efficiency of pair learning retrieval developed further into adulthood, whereas generalization appeared adult-like at age 6 years (Braunitzer et al., 2017).
Associative learning in language acquisition
Associative learning is an essential mechanism in early language development, since it allows for the ability to pair a concept to a word (Tsui et al., 2019). Infants are first able to detect individually presented words by the age of 6 months, as assessed in an experiment by Bergelson and colleagues in which infants were presented with a set of pictures while their parent named one of the depicted items. Infants between 6 and 9 months of age directed their gaze toward the named picture, indicating successful association of the word with the picture (Bergelson and Swingley, 2012). Though extensive variability exists in early language development, infants are generally able to understand dozens of words and produce a few by 1 year of age (Fenson et al., 1994). A popular task to investigate associative word learning is the switch task, in which toddlers habituate to novel object-word pairings, followed by a test phase in which their capacity to notice a switch in the pairings is evaluated (Werker et al., 1998). In a recent meta-analysis, Tsui and colleagues found that associative word learning, as evaluated by the switch task, did not improve between 12 and 21 months of age, suggesting that the ability is in place around 1 year of age (Tsui et al., 2019). The theory of fast mapping accounts for rapid word learning in the second year of life. It describes the capacity of toddlers to learn a new word through a single, brief exposure (Carey and Bartlett, 1978), starting at around 14 months of age (Werker et al., 1998). Though the validity of fast mapping has been debated (McMurray et al., 2012), it has been suggested as a more general learning mechanism, going beyond language acquisition (Rovee-Collier and Giles, 2010). Rovee-Collier and Giles suggest that fast mapping explains the period of exuberant learning and fast forgetting in the first third of infancy, which is replaced by more mature declarative memory systems toward the end of infancy (Rovee-Collier and Giles, 2010).
The impact of context for associative learning
Varying with age, a number of factors influence the efficiency of associative learning. Associative potentiation describes a phenomenon in which infants learn a new association better in the presence of a prior association (Rovee-Collier et al., 2013). In a number of experiments, Barr and colleagues showed that memory strength for a usually rapidly forgotten task in a 6-month-old infant could be substantially increased by combining it with a task that is remembered for longer (Barr et al., 2001, 2002, 2011). The authors suggested that the prior knowledge effect found in adults originates in associative potentiation found during early infancy (Barr et al., 2011). It has been suggested that associative potentiation counteracts rapid forgetting of younger infants by increasing associative learning that can be connected to prior experience (Rovee-Collier et al., 2013).
Another contextual factor influencing associative learning is the socialâemotional value of the learning situation. Seven-month-old infants presented superior learning in an associative learning task when the stimuli were paired with videos of their own mother, rather than a stranger or a cartoon (Tummeltshammer et al., 2018).
Even though infants are able to integrate multimodal information to some extent right after birth, multimodal processing continues to develop throughout infancy and childhood (Lewkowicz, 2000). The influence of multimodality on associative learning in infancy is inconclusive, suggesting both negative and positive effects on learning (Ter Schure et al., 2014). These contradictory results are possibly due to differences in familiarity and complexity of the information in each modality (Plunkett, 2010). The intersensory redundancy hypothesis suggests that when auditory and visual modalities are linked by an amodal property, such as synchrony, which can be processed more easily than the modality-specific information, multimodality promotes learning (Bahrick et al., 2004; Bahrick and Lickliter, 2012).
In school-aged children ranging from 5 to 6 years, Imuta et al. (2018) found that the older group in particular was able to recall the same amount of scientific information from an interactive lesson, independent of context (i.e., field trip vs classroom) (Imuta et al., 2018).
Animal Models of Learning and Memory
R.R. Miller, in Encyclopedia of Behavioral Neuroscience, 2010
Factors that Differentiate Contemporary Models of Associative Learning
Stimulus Definition
As associative learning is the formation of links between the mental representations of stimuli (or responses), it is essential that the meaning of stimulus is well understood. An early and still prevalent view treats a stimulus as would the layman; for example, a tree is a single elemental stimulus. This position is taken by Rescorla and Wagnerâs 1972 model and many other more contemporary models. In contrast, some models deconstruct gross stimuli such as a tree into a large number of stimulus elements, thereby replacing the representation of the tree with individual branches, twigs, leaves, or even leaf fragments (i.e., microelements). Such models often concern themselves with the consequences of associations between the elements, which can provide powerful explanatory mechanisms. At the other end of the spectrum of what constitutes a stimulus is Pearceâs 1987 configural model, which treats the entire momentary perceptual field (except for the US) as a single integrated stimulus (i.e., a configured stimulus). Here, rather than perceiving the tree, subjects are assumed to perceive the entire forest including hearing the birds sing. As such, stimuli are not apt to ever exactly repeat from training trial to training trial, or training trial to test trial. Much explanatory power is based, with considerable success, upon generalization between stimuli on different trials being a function of differences in stimulus similarity.
Emphasis on Acquisition or Performance
The chain of events required for associative control of behavior obviously must include acquisition (sometimes called encoding or learning), retention (sometimes called storage), retrieval, and behavioral expression. Acquired behavior would be disrupted by a failure of any of these links in the chain. Thus, any serious model of learned behavior should speak to each of these steps. Surprisingly, most associative models focus their attention on either the acquisition phase or the retrieval/expression phase, paying little heed to the other phases. That is, to explain differences in stimulus control of behavior between treatment groups, most models emphasize differences in acquisition to the exclusion of differential retrieval of stored memories. Other models emphasize differential information processing after acquisition as a function of treatment, rather than differential acquisition. Few models seriously concern themselves with information processing during both acquisition and testing (i.e., retrieval and expression). Notably, neuroscientists have largely focused on acquisition, generally ignoring information processing that occurs during test.
Most acquisition-focused models follow the Rescorla-Wagner model by assuming that acquisition is driven by one or another form of total error reduction. The assumption here is that, on a given trial, learning about each cue present is proportional to the difference between the maximum associative strength that the outcome delivered on that trial can support and the expectation of the outcome based on all cues present on that trial. Models subscribing to this principle use it to account for stimulus interactions (e.g., conditioned inhibition and cue competition). Some models that subscribe to total error reduction assume that this principle results in cue competition by decreasing processing of US representations, whereas other models assume that the total error reduction principle leads to cue competition by decreasing processing of CS representations. Models that focus on processing at the time of performance typically reject the total error reduction principle and use a local error-correction rule to prevent unlimited acquisition across many CSâUS pairings (an approach popularized by Bush and Mosteller). Here, acquisition for each specific cue is assumed to be proportional to the difference between the maximum associative strength that the outcome delivered on that trial can support and the immediate expectation of the outcome based on that specific cue. Such models account for stimulus interactions through processes that occur at the time of testing.
Trial-Wise or Real-Time Processing
Whether they emphasize information processing during acquisition or during performance, most models of learning are trial-wise models. That is, on a given trial on which stimuli are presented, these models assume that organisms first compute responding, then perceive the outcome of the trial, and finally engage the learning mechanism based on what transpired on that trial. Trial-wise models are to be contrasted with real-time models, which assume that organisms are encoding new information instant by instant throughout the treatment sessions, that is, during both the so-called trials and the intertrial intervals. In reality, there is no doubt that organisms are real-time processors of information, but real-time models are often mathematically intractable. Hence, in practice, researchers simulate them by assuming that treatment sessions have a trial-like structure, composed of very small temporal windows throughout the so-called trials and the intertrial intervals. Moreover, trial-wise models sometimes address learning about context that can occur during intertrial intervals by treating each intertrial interval as many trials on which the target cue and outcome are absent (the specifics of parsing intertrial intervals can be critical). When this is done, trial-wise models can closely resemble real-time models that are simulated by parsing the entire treatment session into many trials of short duration. Hence, in simulation, trial-wise models sometimes do not differ greatly from real-time models, despite the fundamental difference in the underlying assumptions.
Learning during test trials must be considered because early parts of a first test trial can influence behavior in later parts of the same test trial. This is fully expected in real-time models (but not trial-wise models) and is commonly observed in some preparations. For example, tests of fear conditioning often consist of a so-called flooding measure, which involves exposing the subject to a potentially fear-inducing cue (or context) for a single long period of time. Initially, the cue disrupts the behavior that was ongoing prior to presentation of the test cue, but as exposure to the cue continues, the ongoing behavior recovers. Clearly, the recovery of ongoing behavior reflects information processing that occurred earlier on that test trial. Alternatively stated, unlike experimenters, subjects do not differentiate between training and test trials; they are real-time information processors.
Context
Context usually refers to the total physical environment exclusive of nominal CSs and USs. The contexts of both treatment (i.e., acquisition, extinction, etc.) and testing can play several important roles in determining associative responding to a target cue. Both types of context can act as another punctate cue that competes with or facilitates responding to the target cue. Most models attempt to account for this by treating the training context as an additional long-duration cue that is present during both CS presentations and intertrial intervals. This approach is conceptually sound, but in simulations it does not perform well because the predicted behavior varies greatly based on the assumed duration of each context-only trial. Less readily addressed at the conceptual level by most contemporary models is the role that the context (or a punctate stimulus) can play as a discriminative stimulus (aka an occasion setter), in which the context at test can serve to direct the organism to respond using information from only one of two contradictory learning experiences regarding the same cue to which the organism has been exposed. For example, after X-US trials in context #1 and X-noUS trials in context #2, responding to X is observed in context #1 but not context #2. Only a few models of associative learning even begin to account for such occasion setting. Bouton does account for such phenomena, but does not provide a general model of learning.
Sensory Science and Consumer Perception âą Food Physics & materials Science
S Tempere, ⊠G Sicard, in Current Opinion in Food Science, 2019
Associative learning and training
Associative learning is defined as the process by which one event or item comes to be linked to another through experience [30]. Associative training couples perceptual training with formal information about the stimulus.
For example, it is known that the valence of odors affects discrimination in mixtures: Rabin and Cain [31] demonstrated that the detection of a minor component in a mixture was affected by pleasantness. Unpleasant stimuli are more detectable probably due to the experience of detecting impurities or spoilage in foods.
Several studies have reported the potency of associative learning to enhance olfactory capacities. According to Goldstone [16], attentional weighting may modify perceptual abilities, making perception relevant by increasing the attention paid to important perceptual dimensions and features and/or decreasing attention to irrelevant factors.
Looking for further proof of the capacity of associative training to modify olfactory abilities, a simple conditioning protocol to wine professionals was tested (Sophie Tempere, PhD, UniversitĂ© de Lyon 2, 2010). In the first session, the individual rejection thresholds for geosmin of 68 wine professionals were measured. The subjects were then distributed equally among the training groups on the basis of their initial levels of rejection thresholds. The control group was simply informed that their thresholds would be measured twice in separate sessions. The second group received pure geosmin samples and was asked to smell this odorant once daily for one month. The third group also received a geosmin sample to smell daily and was instructed to pay attention to its musty, earthy off-flavor. After the training period, variations in wine rejection thresholds in the training groups were evaluated. A significant drop in rejection threshold was found in the third group only. The fact that the above-threshold odorant, geosmin, was presented to the wine tasters as a defect was likely to focus their attention and enhance sensory training (topâdown effect), constituting an aversive signal. Previous studies revealed that attention level played a role during exposure to an odorant, changing in the hedonic perception and enhancing the effect of exposure [32]. Hedonic changes following exposure to supra-threshold levels are not the result of an implicit process, but require active cognitive mediation. Another study focused on the Brett character, described as an off-flavor in wines since the 1990s. The results confirmed the impact of intentional training and learning on the perceptive skills and decision-making processes of professionals [33].
Finally, associations between wines and other stimuli may enhance tasting efficiency. In a recent study combining original approaches, Latour and Deighton [34] found that, after prior analytical training (use of a lexicon to decompose the stimulus), holistic training enhanced wine recognition. This holistic approach (defined by the authors as non-verbal, imagery-based, and involving narrative process) combined a global description of wine and the use of visual images.
The Zebrafish: Cellular and Developmental Biology, Part B
Robert Gerlai, in Methods in Cell Biology, 2011
I Introduction
A Why Study Associative Learning?
Learning and memory is a ubiquitous feature of all animals. From single cell organisms through the nematode and Drosophila to higher order vertebrates, learning and memory allow the organism to respond plastically to the changing environment. Instead of having to wait for hundreds, perhaps thousands, of generations to make individuals genetically adjust (evolve), that is, to adapt properly to local environmental conditions, learning and memory can achieve the âadaptationâ within a fraction of the lifetime of the organism. Learning in a broad sense here is defined as the acquisition of experience-dependent change in behavior, and memory is thought of as the consolidation, storage, and recall of the acquired information. Here, learning and memory are used interchangeably in the sense that these processes represent the two sides of the same coin: a temporal series of neurobiological mechanisms that lead to the manifestation of experience-dependent behavioral change. Associative learning, the focus of the current chapter, is one of the most widely studied and also complex forms of learning and memory (e.g., a simple Medline (PubMed) search with the keyword âassociative learningâ returns more than 20,000 entries). Associative learning requires the acquisition of temporal and/or causal relationship between at least two stimuli. The different forms of associative learning are not reviewed here because this question has been extensively discussed by others in the literature (e.g., Eichenbaum, 2006; Squire, 2004; Tulving, 1987). Suffice it to say that the simplest form of associative learning is when the animal (human or nonhuman) learns the association between two stimuli, US (unconditioned stimulus) and CS (the conditioned stimulus). One of the most complex forms of associative learning is relational learning, in which the animal learns loose relationships of a potentially large number of cues or stimuli (Eichenbaum, 2004). An example of the latter is episodic memory and spatial learning (Squire, 2004).
Apart from the scientifically fascinating aspect of the question of how associative learning works, there is also an important human clinical relevance of these studies. Numerous CNS disorders, for example neurodegenerative disorders including Alzheimerâs disease, are associated with impaired associative learning, particularly of the relational type (Dickerson and Eichenbaum, 2010). Concerted efforts have been and are being made to understand the mechanisms of these learning processes and to develop treatment applications. Indeed, by now we know a lot about intricate molecular, synaptic electrophysiological, and neuroanatomical (microstructural as well as macrostructural) aspects of these processes. For example, the second edition of the book, Mechanisms of Memory (one of the most comprehensive treatments of the biochemistry of learning and memory) (Sweatt, 2010), enlists hundreds of already known molecular components involved in one way or another in processes associated with memory formation. Is there anything else left to discover?
The answer to this question is a resounding yes. According to conservative estimates, vertebrate genomes harbor approximately 30,000 to 40,000 genes and at least 50% of these genes are expressed in the brain (DĂaz, 2009; Fritzsch, 1998). A large proportion of these genes is believed to subserve neural plasticity, one of the fundamental and most crucial functions of the brain (Sweatt, 2010). Briefly, one may expect as many as 10â15 thousand genes involved in learning and memory processes, a staggering number compared to a few hundred genes so far known to underlie these phenomena.
Emotion: Neuroimaging
R.J. Dolan, in Encyclopedia of Neuroscience, 2009
Imaging the Acquisition of Emotion
In classical or Pavlovian conditioning, a neutral stimulus, through pairing with an emotional stimulus (e.g., an aversive noise in fear conditioning), acquires the ability to predict its future occurrence. Associative learning provides a highly conserved means by which organisms acquire knowledge regarding the causal structure of the environment. Such knowledge endows an organism with the ability to anticipate future events of value, such as the likelihood of food or danger, on the basis of predictive sensory cues. The amygdala plays a central role in this type of causal learning. Thus, amygdala activation is seen when a previously neutral item (the conditioned stimulus (CS+)) acquires predictive significance through its pairing with a biologically salient reinforcer (the unconditioned stimulus (UCS). This pattern of activation is seen both for aversive and appetitive events. Thus, across a wide range of experiments, an enhanced amygdala response is evident when subjects learn that a neutral stimulus predicts the occurrence of an emotion-inducing event, for example, an aversive event.
Theoretical accounts of associative learning based on the RescorlaâWagner rule and their real-time extensions, such as temporal difference (TD) reinforcement models, provide plausible descriptions of the computational processes underling associative learning. Crucial to these models is the expression of a teaching signal, referred to as the prediction error. A prediction error is used to direct acquisition and refine expectations relating to predictive cues, and it records a change in expected affective outcomes, being expressed whenever predictions are generated,updated, or violated. Functional neuroimaging studies have demonstrated that such a prediction error is expressed in the human brain in structures such as amygdala, striatum, and orbital-prefrontal cortex (OFC) during aversive and appetitive associative emotional learning.
Learning Theory and Behaviour
M. Giurfa, in Learning and Memory: A Comprehensive Reference, 2008
1.29.2 Elemental and Nonelemental Forms of Associative Learning
Because this chapter intends to present the different levels of complexity that honeybees can reach in mastering different learning tasks, it is worth starting with operational definitions that allow discerning the simple from the complex. I focus on associative learning and introduce the distinction between elemental and nonelemental learning, which may be useful as a boundary between simple and complex forms of learning.
Associative learning is a capacity that is widespread among living animals and that allows extracting the logical structure of the world. It consists in establishing predictive relationships between contingent events in the environment so that uncertainty is reduced and adaptive behavior results from individual experience with such events. Two major forms of associative learning are usually recognized: In classical conditioning (Pavlov, 1927), animals learn to associate an originally neutral stimulus (conditioned stimulus, CS) with a biologically relevant stimulus (unconditioned stimulus, US); in operant conditioning (Skinner, 1938), they learn to associate their own behavior with a reinforcer. Both forms of learning, therefore, reliably predict reinforcement, either appetitive or aversive, and admit different levels of complexity. In their most simple version, both rely on the establishment of elemental links connecting two specific and unambiguous events in the animalâs world. What has been learned for a given tone in terms of its outcome is valid for that tone but not necessarily for another stimulus, such as a light. The outcome of a given behavior, such as pressing a lever, is valid for that behavior but not for a different one such as pulling a chain (See Chapters 1.03, 1.07, 1.08). These forms of learning, which have been intensively studied by experimental psychologists, are also particularly interesting for neuroscientists interested in the neural bases of learning because they allow tracing to the level of neural circuits and single neurons the basis of associations underlying learning. Because these forms of learning rely on specific stimuli (e.g., a given CS and a given US), it is possible to study where and how in the central nervous system such stimuli are represented, where and how their neural pathways interact in order to facilitate association, and how experience modifies their respective neural representations. Both at the behavioral and neural level, these forms of learning have in common the univocal and unambiguous relationships established between events in the world. Because they can be characterized through specific links between unique events, simple forms of associative learning are termed elemental learning forms. Typical examples of elemental learning are absolute conditioning (A+), in which a single stimulus A is reinforced (+), and differential conditioning (A+ vs. Bâ), in which one stimulus, A, is reinforced (+), while another stimulus, B, is nonreinforced (â) (see Table 1). In the former, an animal has to learn to respond to A, which is unambiguously associated with reinforcement; in the latter, it has to learn to respond to A and not to B because both are unambiguously associated with reinforcement and with the absence of it, respectively.
Table 1. Examples of elemental and nonelemental conditioning protocolsa
| Conditioning task | Training | Processing |
|---|---|---|
| Absolute conditioning | A+ | Elemental |
| Differential conditioning | A+ vs. Bâ | Elemental |
| Feature positive discrimination | AB+ vs. Bâ | Elemental |
| Negative patterning | A+, B+ vs. ABâ | Nonelemental |
| Biconditional discrimination | AB+, CD+ vs. ACâ, BDâ | Nonelemental |
a
In absolute conditioning, the subject has to learn to respond to stimulus A, which is unambiguously associated with reinforcement (+); in differential conditioning, the subject has to learn to respond to stimulus A and not to B; A is unambiguously associated with reinforcement (+), whereas B is unambiguously associated with the absence of reinforcement (â); in feature-positive discrimination, the subject has to learn to respond to the compound AB, which is reinforced (+), and not to B (â), which is nonreinforced; although B is ambiguous because it appears as often reinforced as nonreinforced, the fact that A is unambiguously associated with the reinforcement allows solving the problem. Simple links between a stimulus and reinforcement allow solving these three elemental problems. Elemental solutions cannot account for negative patterning solving, in which the subject has to learn to respond to the single stimuli A+ and B+ but not to their compound ABâ, because elements are as often reinforced as nonreinforced. The same remark applies to biconditional discrimination, in which the subject has to learn to respond to the compounds AB+ and CD+ but not to ACâ and BDâ.
However, other forms of associative learning are possible, in which unique links connecting specific events are useless because ambiguity characterizes the events under consideration (see Table 1). For instance, in the so-called patterning problems, animals have to learn to discriminate a stimulus compound from its components, a task that is not necessarily trivial. Consider, for example, negative patterning, a problem in which an animal has to learn to discriminate two single components reinforced from their nonreinforced binary compound (A+, B+ vs. ABâ). This situation is challenging because each element A and B appears as often reinforced as nonreinforced. Relying on elemental links between A (or B) and reinforcement (or absence of reinforcement) cannot solve this problem. Different strategies, such as treating the binary compound in a nonlinear form (i.e., as being different from the simple sum of A and B) have to be implemented to solve this kind of problem. A profuse literature has shown that some vertebrates can solve this kind of nonlinear processes and has put the accent on the nervous circuits and brain structures required for this kind of cognitive processing (Rudy and Sutherland, 1995; OâReilly and Rudy, 2001; Bucci et al., 2002; Alvarado and Bachevalier, 2005; Moses et al., 2005; Borlikova et al., 2006; Jacobs, 2006; See Chapter 1.10).
Having introduced these two forms of learning, which define different levels of complexity in cognitive processing in a formalized and operational way, I present findings showing that it is possible to dissect and understand the basic mechanisms underlying these two levels of processing, using honeybees as a model system. I demonstrate that this insect exhibits elemental and nonelemental forms of learning that are relevant in its natural life and that are amenable to the laboratory, thus allowing controlled study and access to the underlying nervous system.
Neurocognitive Development: Normative Development
Marc Philippe Lafontaine, ⊠Sarah Lippé, in Handbook of Clinical Neurology, 2020
Classical and operant conditioning
If a major function of the brain is to allow efficient processing of the environment using models such as Bayesian inference and PC, then encoding of unitary items and novelty detection through repetition are insufficient. Indeed, in order to generate expectations or predictions of incoming stimuli, an ability to construct reliable and sensible models of complex environments is necessary. In turn, this ability must rely on mechanisms enabling the encoding of associations or relationships between items and events. Associative learning is defined as learning about the relationship between two separate stimuli, where the stimuli might range from concrete objects and events to abstract concepts, such as time, location, context, or categories. The most basic form of associative learning is classical conditioning, in which an unconditioned stimulus (US) leading to an unconditioned response (UR) (e.g., a simple reflex response such as salivation) is paired with a conditioned stimulus (CS). After several (or even just one) pairing of US and CS, the presentation of the CS alone leads to a conditioned response (CS) that resembles the UR, though it is a learned rather than a reflexive response. Similarly, operant conditioning involves the pairing of a certain behavior with a reward or punishment to strengthen or decrease said behavior. In humans, the scope of associative learning can be broadened to include a vast array of relational connections. In fact, in contemplating everyday life, one realizes that the majority of what we are able to accomplish, understand, and remember relies on an ability to associate several items in space and time, allowing us to construct and store complex scenes and sequences of events.
Animal Cognition
L. Castro, E.A. Wasserman, in Encyclopedia of Human Behavior (Second Edition), 2012
Glossary
Associative learning
Making the psychological connection between contiguous events in the environment.
Categorization
Placing different objects or events into separate classes or categories.
Conceptualization
The process of forming an abstract or generic idea from particular instances.
Discrimination
Responding differently to two or more stimuli differing in one or more respects.
Generalization
Transferring a learned response from one stimulus to another.
Matching-to-sample
Experimental procedure for studying memory in which reward is given for responding to a stimulus if it is the same as a prior sample stimulus.
Memory
Control of current behavior by past stimulation.
Metacognition
Thinking about oneâs own cognitive states and processes.
Numerical processing
Set of abilities related to understanding and manipulating numbers, such as: using symbols to denote quantities; discriminating, ordering, and comparing different quantities; and, combining quantities in order to perform arithmetic operations.
Reinforcer
Any consequence of a behavior that increases the probability that the organism will repeat that behavior.
Time-out
A brief period of time during which there is no stimulation; time-out is a form of punishment that is used to decrease the probability that the organism will repeat a behavior.