Perceptual learning

Discovery and fluency effects

Perceptual learning effects can be organized into two broad categories: discovery effects and fluency effects.^[1] Discovery effects involve some change in the bases of response such as in selecting new information relevant for the task, amplifying relevant information or suppressing irrelevant information. Experts extract larger "chunks" of information and discover high-order relations and structures in their domains of expertise that are invisible to novices. Fluency effects involve changes in the ease of extraction. Not only can experts process high-order information, they do so with great speed and low attentional load. Discovery and fluency effects work together so that as the discovery structures becomes more automatic, attentional resources are conserved for discovery of new relations and for high-level thinking and problem-solving.

The role of attention

William James (Principles of Psychology, 1890) asserted that "My experience is what I agree to attend to. Only those items which I notice shape my mind - without selective interest, experience is an utter chaos.".^[33] His view was extreme, yet its gist was largely supported by subsequent behavioral and physiological studies. Mere exposure does not seem to suffice for acquiring expertise.

Indeed, a relevant signal in a given behavioral condition may be considered noise in another. For example, when presented with two similar stimuli, one might endeavor to study the differences between their representations in order to improve one's ability to discriminate between them, or one may instead concentrate on the similarities to improve one's ability to identify both as belonging to the same category. A specific difference between them could be considered 'signal' in the first case and 'noise' in the second case. Thus, as we adapt to tasks and environments, we pay increasingly more attention to the perceptual features that are relevant and important for the task at hand, and at the same time, less attention to the irrelevant features. This mechanism is called attentional weighting.^[37]

However, recent studies suggest that perceptual learning occurs without selective attention.^[38] Studies of such task-irrelevant perceptual learning (TIPL) show that the degree of TIPL is similar to that found through direct training procedures.^[39] TIPL for a stimulus depends on the relationship between that stimulus and important task events^[40] or upon stimulus reward contingencies.^[41] It has thus been suggested that learning (of task irrelevant stimuli) is contingent upon spatially diffusive learning signals.^[42] Similar effects, but upon a shorter time scale, have been found for memory processes and in some cases is called attentional boosting.^[43] Thus, when an important (alerting) event occurs, learning may also affect concurrent, non-attended and non-salient stimuli.^[44]

Time course of perceptual learning

The time course of perceptual learning varies from one participant to another.^[11] Perceptual learning occurs not only within the first training session but also between sessions.^[45] Fast learning (i.e., within-first-session learning) and slow learning (i.e., between-session learning) involves different changes in the human adult brain. While the fast learning effects can only be retained for a short term of several days, the slow learning effects can be preserved for a long term over several months.^[46]

Receptive field modification

Research on basic sensory discriminations often show that perceptual learning effects are specific to the trained task or stimulus.^[47] Many researchers take this to suggest that perceptual learning may work by modifying the receptive fields of the cells (e.g., V1 and V2 cells) that initially encode the stimulus. For example, individual cells could adapt to become more sensitive to important features, effectively recruiting more cells for a particular purpose, making some cells more specifically tuned for the task at hand.^[48] Evidence for receptive field change has been found using single-cell recording techniques in primates in both tactile and auditory domains.^[49]

However, not all perceptual learning tasks are specific to the trained stimuli or tasks. Sireteanu and Rettenback^[50] discussed discrimination learning effects that generalize across eyes, retinal locations and tasks. Ahissar and Hochstein^[51] used visual search to show that learning to detect a single line element hidden in an array of differently-oriented line segments could generalize to positions at which the target was never presented. In human vision, not enough receptive field modification has been found in early visual areas to explain perceptual learning.^[52] Training that produces large behavioral changes such as improvements in discrimination does not produce changes in receptive fields. In studies where changes have been found, the changes are too small to explain changes in behavior.^[53]

Reverse hierarchy theory

The Reverse Hierarchy Theory (RHT), proposed by Ahissar & Hochstein, aims to link between learning dynamics and specificity and the underlying neuronal sites.^[54] RHT proposes that naïve performance is based on responses at high-level cortical areas, where crude, categorical level representations of the environment are represented. Hence initial learning stages involve understanding global aspects of the task. Subsequent practice may yield better perceptual resolution as a consequence of accessing lower-level information via the feedback connections going from high to low levels. Accessing the relevant low-level representations requires a backward search during which informative input populations of neurons in the low level are allocated. Hence, subsequent learning and its specificity reflect the resolution of lower levels. RHT thus proposes that initial performance is limited by the high-level resolution whereas post-training performance is limited by the resolution at low levels. Since high-level representations of different individuals differ due to their prior experience, their initial learning patterns may differ. Several imaging studies are in line with this interpretation, finding that initial performance is correlated with average (BOLD) responses at higher-level areas whereas subsequent performance is more correlated with activity at lower-level areas^{[citation needed]}. RHT proposes that modifications at low levels will occur only when the backward search (from high to low levels of processing) is successful. Such success requires that the backward search will "know" which neurons in the lower level are informative. This "knowledge" is gained by training repeatedly on a limited set of stimuli, such that the same lower-level neuronal populations are informative during several trials. Recent studies found that mixing a broad range of stimuli may also yield effective learning if these stimuli are clearly perceived as different, or are explicitly tagged as different. These findings further support the requirement for top-down guidance in order to obtain effective learning.

Enrichment versus differentiation

In some complex perceptual tasks, all humans are experts. We are all very sophisticated, but not infallible at scene identification, face identification and speech perception. Traditional explanations attribute this expertise to some holistic, somewhat specialized, mechanisms. Perhaps such quick identifications are achieved by more specific and complex perceptual detectors which gradually "chunk" (i.e., unitize) features that tend to concur, making it easier to pull a whole set of information. Whether any concurrence of features can gradually be chunked with practice or chunking can only be obtained with some pre-disposition (e.g. faces, phonological categories) is an open question. Current findings suggest that such expertise is correlated with a significant increase in the cortical volume involved in these processes. Thus, we all have somewhat specialized face areas, which may reveal an innate property, but we also develop somewhat specialized areas for written words as opposed to single letters or strings of letter-like symbols. Moreover, special experts in a given domain have larger cortical areas involved in that domain. Thus, expert musicians have larger auditory areas.^[55] These observations are in line with traditional theories of enrichment proposing that improved performance involves an increase in cortical representation. For this expertise, basic categorical identification may be based on enriched and detailed representations, located to some extent in specialized brain areas. Physiological evidence suggests that training for refined discrimination along basic dimensions (e.g. frequency in the auditory modality) also increases the representation of the trained parameters, though in these cases the increase may mainly involve lower-level sensory areas.^[56]

Selective reweighting

In 2005, Petrov, Dosher and Lu pointed out that perceptual learning may be explained in terms of the selection of which analyzers best perform the classification, even in simple discrimination tasks. They explain that the some part of the neural system responsible for particular decisions have specificity^{[clarification needed]}, while low-level perceptual units do not.^[37] In their model, encodings at the lowest level do not change. Rather, changes that occur in perceptual learning arise from changes in higher-level, abstract representations of the relevant stimuli. Because specificity can come from differentially selecting information, this "selective reweighting theory" allows for learning of complex, abstract representation. This corresponds to Gibson's earlier account of perceptual learning as selection and learning of distinguishing features. Selection may be the unifying principles of perceptual learning at all levels.^[57]

Consolidation and sleep

Several studies asked whether learning takes place during practice sessions or in between, for example, during subsequent sleep. The dynamics of learning are hard to evaluate since the directly measured parameter is performance, which is affected by both learning, inducing improvement, and fatigue, which hampers performance. Current studies suggest that sleep contributes to improved and durable learning effects, by further strengthening connections in the absence of continued practice.^[45][58][59] Both slow-wave and REM (rapid eye movement) stages of sleep may contribute to this process, via not-yet-understood mechanisms.

Comparison and contrast

Practice with comparison and contrast of instances that belong to the same or different categories allow for the pick-up of the distinguishing features—features that are important for the classification task—and the filter of the irrelevant features.^[60]

Task difficulty

Learning easy examples first may lead to better transfer and better learning of more difficult cases.^[61] By recording ERPs from human adults, Ding and Colleagues investigated the influence of task difficulty on the brain mechanisms of visual perceptual learning. Results showed that difficult task training affected earlier visual processing stage and broader visual cortical regions than easy task training.^[62]

Active classification and attention

Active classification effort and attention are often necessary to produce perceptual learning effects.^[59] However, in some cases, mere exposure to certain stimulus variations can produce improved discriminations.

Feedback

In many cases, perceptual learning does not require feedback (whether or not the classification is correct).^[56] Other studies suggest that block feedback (feedback only after a block of trials) produces more learning effects than no feedback at all.^[63]

Limits

Despite the marked perceptual learning demonstrated in different sensory systems and under varied training paradigms, it is clear that perceptual learning must face certain unsurpassable limits imposed by the physical characteristics of the sensory system. For instance, in tactile spatial acuity tasks, experiments suggest that the extent of learning is limited by fingertip surface area, which may constrain the underlying density of mechanoreceptors.^[11]

Declarative & procedural learning

In many domains of expertise in the real world, perceptual learning interacts with other forms of learning. Declarative knowledge tends to occur with perceptual learning. As we learn to distinguish between an array of wine flavors, we also develop a wide range of vocabularies to describe the intricacy of each flavor.

Similarly, perceptual learning also interacts flexibly with procedural knowledge. For example, the perceptual expertise of a baseball player at bat can detect early in the ball's flight whether the pitcher threw a curveball. However, the perceptual differentiation of the feel of swinging the bat in various ways may also have been involved in learning the motor commands that produce the required swing.^[1]

Implicit learning

Perceptual learning is often said to be implicit, such that learning occurs without awareness. It is not at all clear whether perceptual learning is always implicit. Changes in sensitivity that arise are often not conscious and do not involve conscious procedures, but perceptual information can be mapped onto various responses.^[1]

In complex perceptual learning tasks (e.g., sorting of newborn chicks by sex, playing chess), experts are often unable to explain what stimulus relationships they are using in classification. However, in less complex perceptual learning tasks, people can point out what information they're using to make classifications.

Category learning vs. perceptual learning

Perceptual learning is distinguished from category learning. Perceptual learning generally refers to the enhancement of detectability of a perceptual item or the discriminability between two or more items. In contrast, category learning involves labeling or categorizing an item into a particular group or category. However, in some cases, there is an overlap between perceptual learning and category learning. For instance, to discriminate between two items, a categorical difference between them may sometimes be utilized, in which case category learning, rather than perceptual learning, is thought to occur. Although perceptual learning and category learning are distinct forms of learning, they can interact. For example, category learning that groups multiple orientations into different categories can lead perceptual learning of one orientation to transfer across other orientations within the same category as the trained orientation. This is termed "category-induced perceptual learning".

Neuropsychology of perceptual category learning

Multiple different category learning systems may mediate the learning of different category structures. "Two systems that have received support are a frontal-based explicit system that uses logical reasoning, depends on working memory and executive attention, and is mediated primarily by the anterior cingulate, the prefrontal cortex and the associative striatum, including the head of the caudate. The second is a basal ganglia-mediated implicit system that uses procedural learning, requires a dopamine reward signal and is mediated primarily by the sensorimotor striatum"^[64] The studies showed that there was significant involvement of the striatum and less involvement of the medial temporal lobes in category learning. In people who have striatal damage, the need to ignore irrelevant information is more predictive of a rule-based category learning deficit. Whereas, the complexity of the rule is predictive of an information integration category learning deficit.

Improving perceptual skills

An important potential application of perceptual learning is the acquisition of skill for practical purposes. Thus it is important to understand whether training for increased resolution in lab conditions induces a general upgrade which transfers to other environmental contexts, or results from mechanisms which are context specific. Improving complex skills is typically gained by training under complex simulation conditions rather than one component at a time. Recent lab-based training protocols with complex action computer games have shown that such practice indeed modifies visual skills in a general way, which transfers to new visual contexts. In 2010, Achtman, Green, and Bavelier reviewed the research on video games to train visual skills.^[65] They cite a previous review by Green & Bavelier (2006)^[66] on using video games to enhance perceptual and cognitive abilities. A variety of skills were upgraded in video game players, including "improved hand-eye coordination,^[67] increased processing in the periphery,^[68] enhanced mental rotation skills,^[69] greater divided attention abilities,^[70] and faster reaction times,^[71] to name a few". An important characteristic is the functional increase in the size of the effective visual field (within which viewers can identify objects), which is trained in action games and transfers to new settings. Whether learning of simple discriminations, which are trained in separation, transfers to new stimulus contexts (e.g. complex stimulus conditions) is still an open question.

Like experimental procedures, other attempts to apply perceptual learning methods to basic and complex skills use training situations in which the learner receives many short classification trials. Tallal, Merzenich and their colleagues have successfully adapted auditory discrimination paradigms to address speech and language difficulties.^[72][73] They reported improvements in language learning-impaired children using specially enhanced and extended speech signals. The results applied not only to auditory discrimination performance but speech and language comprehension as well.

Technologies for perceptual learning

In educational domains, recent efforts by Philip Kellman and colleagues showed that perceptual learning can be systematically produced and accelerated using specific, computer-based technology. Their approach to perceptual learning methods take the form of perceptual learning modules (PLMs): sets of short, interactive trials that develop, in a particular domain, learners' pattern recognition, classification abilities, and their abilities to map across multiple representations. As a result of practice with mapping across transformations (e.g., algebra, fractions) and across multiple representations (e.g., graphs, equations, and word problems), students show dramatic gains in their structure recognition in fraction learning and algebra. They also demonstrated that when students practice classifying algebraic transformations using PLMs, the results show remarkable improvements in fluency at algebra problem solving.^[57][74][75] These results suggests that perceptual learning can offer a needed complement to conceptual and procedural instructions in the classroom.

Similar results have also been replicated in other domains with PLMs, including anatomic recognition in medical and surgical training,^[76] reading instrumental flight displays,^[77] and apprehending molecular structures in chemistry.^[78]