Perceptual Learning Research Paper

2. Terms Related To Perceptual Learning

Quite a number of processes in the nervous system are able to change their responses to sensory stimulation and there is quite a number of terms related to the term ‘learning.’ To understand the concept of perceptual learning clearly these terms have to be defined in order to clarify differences and similarities. These related terms include ‘plasticity,’ ‘adaptation,’ ‘habituation,’ ‘after-effects,’ ‘priming,’ ‘development,’ and ‘maturation’ as well as ‘improvement through insight.’

‘Plasticity’ is defined here as modifiability of functional and anatomical organization of the central nervous system leading to more appropriate function as a result of sensory experience, or to overcome limitations caused by lesions. The term ‘adaptation’ is most often used in relation to adjustments of information processing within a predefined working range, as a result of stimulation. A classic example is luminance adaptation—adjusting the working range of the visual system to ambient light levels, with no long-term consequences. ‘Habituation,’ or satiation’ in the context of rewards, seems to be a special case of adaptation, namely a shift of working range towards lower sensitivity, as in the case of decreased reflex response after repeated stimulation. ‘After-effects’ can be considered, in many cases, as the result of selective short-term adaptation in cases where perception is the result of antagonistic neuronal channels, as with many after-images. ‘Priming’ describes the effect of a (some- times subliminal) stimulus on a subsequent stimulus– reaction pair, or more generally, the effect of initiating a certain type of behavior. The effect of priming is usually short. ‘Development’ and ‘maturation,’ unlike learning, ascribe the thrust of the changes in behavior to genetics, not the environment; hence the mechanisms underlying changes of behavior differ radically between these phenomena on one side and perceptual learning on the other. ‘Improvement through insight,’ as the name indicates, is a term that should be reserved for positive changes of information processing based on cognitive processes, such as one-shot learning.

So what is so special about perceptual learning? I would like to stress the following difference between ‘ordinary’ vs. ‘perceptual’ learning. Most forms of learning to collect better information about the outside world are generally considered to rely on relatively intelligent or complex or cognitive levels of information processing. Hence, learning leading to improved discrimination between sensory stimuli would have been considered only a decade ago to be more of the declarative type of learning (as in learning a poem) than of the procedural type of learning (as in learning to play the piano). Perceptual learning, on the other hand, is a form of learning leading to better use of sensory information which is relatively independent of conscious or declarative forms of learning but relies partly on rather low-level modifications in the central nervous system. Perceptual learning hence resembles, in many respects, procedural forms of learning that are common in motor learning, for example learning to ride a bicycle.

3. Perceptual Learning And Plasticity Of Primary Sensory Cortices

This type of low-level modification would have been considered impossible in the 1980s. At this time, the primary visual cortex of adults was considered to be a rather ‘hard-wired’ first stage of visual information processing, which served to detect and extract certain ‘elementary features’ from the complex scenes surrounding us (cf. Marr 1982). It was considered to lack plasticity since changing preprocessing during the training of one task might have disadvantageous consequences for solving other perceptual tasks. Indeed, there is direct experimental evidence for a decrease in plasticity of the primary visual cortex during maturation. Hubel and Wiesel (1965) found that covering one eye in young cats leads to changes in their cortical wiring patterns that are reversible during their kitten hood but not thereafter. Similarly, children suffering from a squint may achieve only low visual acuity in the deviating eye; therapy can reverse this loss in small children only. Therefore it seems that a critical development phase exists during which the primary visual cortex is still plastic, but that it loses this plasticity at a later stage of development. Improvement through training in perceptual tasks by adults was therefore supposed to take place on higher levels of cortical information processing.

However, during the last decade of the twentieth century, a number of electrophysiological and psychophysical experiments cast some doubts on this interpretation and suggested that even adult primary sensory cortices showed much more plasticity than was hitherto believed. Some of this evidence will be presented here. A first example of this evidence was the orientation specificity of improvement through learning. For example, Fiorentini and Berardi (1981) found that discrimination between complex types of grating improved through practice, but that the improvement did not transfer to stimuli rotated by 90 degrees. The same was true for a vernier discrimination task where observers had to indicate whether the lower segment of a vertically oriented vernier stimulus consisting of two almost collinear bars was offset to the right or to the left relative to the upper segment, or whether the right segment of a horizontal vernier was above or below the left segment. Observers improved their performance very significantly during training, but their performance returned to base level when the stimulus was rotated by 90 degrees. Hence, the improvement obtained through training was specific for stimulus orientation.

In a control experiment, the pushbuttons by which observers had to indicate the direction of offset were switched between hands. Observers had to push the left button when a vernier was offset to the right, and vice versa. Results did not deteriorate, indicating that not every change of experimental procedure leads to a drop in performance. Another group of observers were trained in vernier discriminations at eight different positions in the periphery of the visual field for one hour at each position. During that time, they improved performance, but returned to baseline at the transition from one visual field position to the next. The same was true for training vernier discriminations with one eye covered: observers improved their performance through training, but when they performed the same discriminations with the cover moved to the other eye, performance dropped to pretraining levels. The same specificity of improvement for the eye used during training had been observed earlier in a texture discrimination task in which observers had to discriminate a figure from its surround based on the orientation of the stimulus elements (Karni and Sagi 1991). Moreover, improvement did not transfer between a three-dot vernier and a three-dot bisection task, although the stimuli of both tasks differed by approximately one photoreceptor diameter.

4. Neuronal Basis Of Perceptual Learning

Hence, the changes of the nervous system underlying these forms of perceptual learning should occur on a quite early level of cortical information processing, in which the neurons are already selective for different stimulus orientations—unlike in the retina—but still partly monocularly activated—unlike in all cortical areas beyond the primary visual cortex. In particular, the eye specificity of improvement points to the primary visual cortex as the most probable site for at least a large part of the changes underlying these forms of perceptual learning.

Recent electrophysiological evidence supports this conclusion of a relatively early site of parts of the neuronal changes underlying perceptual learning. Single-cell recordings in adult monkeys have demonstrated that receptive fields of neurons in the primary visual cortex of adult primates can change position after the parts of the retina that were supplying the input to these cells have been destroyed (Gilbert and Wiesel 1992), that the microcircuitry of the adult visual cortex can change (Eysel 2001), and that the distribution of mass potentials evoked by visual stimulation in humans changes as a result of training, especially pronounced for short latencies and over the primary visual cortex.

5. Different Cortical Levels Of Perceptual Learning

Marr’s insight is still true—changing the very front end of information processing as a result of learning one perceptual task would necessarily change the processing of many, if not all, other stimuli presented to the same sensors. The speed and amount of learning depend strongly on attentional control, that is on top-down influences within the brain. Hence, present models of perceptual learning increasingly emphasize that learning will occur at quite a number of different levels of information processing, and that top-down influences from ‘higher’ levels will play a crucial role in adjusting, in a task-dependent way, processing on the ‘lower’ levels. Recent experiments have indeed provided direct evidence for strong top-down influences on perceptual learning as a result of error feedback and attentional processes (Ahissar and Hochstein 1993, Herzog and Fahle 1998).

It seems that the high specificity of perceptual learning is partly lost if relatively easy tasks are learned, while specificity of improvement is highest for very difficult tasks. A possible explanation is that ‘easy’ tasks are learned on a relatively ‘higher’ level of information processing, at which the information extracted from the visual scene is better used after training than was possible before. Difficult tasks, on the other hand, may require changes at lower levels of processing that are specific, for example, to exact visual field position as well as stimulus orientation.

6. Conclusions

To conclude, perceptual learning differs from other forms of learning, especially declarative forms, in that it can be very task-and stimulus-specific, and probably involves functional and anatomical changes even in primary sensory cortices. Though at first sight perceptual learning seems to resemble declarative forms of learning, and to rely on relatively complex cognitive processes, the specificity of improvement for quite low-level attributes such as visual field position, stimulus orientation, and the eye used for training a visual task indicates a strong and crucial involvement of the primary visual cortex, where neurons are still partly monocularly activated. Dependence on attention, error feedback, and ‘insight,’ on the other hand, demonstrates that strong top-down influences play a major role in perceptual learning and that perceptual learning also involves more cognitive levels of the brain. Hence, the study of perceptual learning processes not only shows us the amazing amount of plasticity even in adult sensory information processing at a relatively peripheral level, but also leads to a view of cortical information processing not as a feed-forward system of subsequent neuronal layers but as a complex and plastic feedback system with strong and important top-down influences that shape ‘lower’ or ‘early’ parts of information processing.

Bibliography:

1. Ahissar M, Hochstein S 1993 Attentional control of early perceptual learning. Proceedings of the National Academy of Sciences of the USA 90: 5718–22
2. Eysel U 2001 Plasticity of reactive fields on early stages of the adult visual system. In: Fahle M, Poggio T (eds.) Perceptual Learning. MIT Press, Cambridge, MA
3. Fahle M, Poggio T (eds.) 2001 Perceptual Learning. MIT Press, Cambridge, MA
4. Fiorentini A, Berardi N 1981 Perceptual learning specific for orientation and spatial frequency. Nature 287: 43–4
5. Gibson E J 1963 Perceptual learning. Annual Review of Psychology 14: 29–56
6. Gilbert C D, Wiesel T N 1992 Receptive field dynamics in adult primary visual cortex. Nature 356: 150–2
7. Herzog M H, Fahle M 1998 Modeling perceptual learning: Difficulties and how they can be overcome. Biological Cybernetics 78: 107–17
8. Hubel D H, Wiesel T N 1965 Binocular interaction in striate cortex of kittens reared with artificial squint. Journal of Neurophysiology 28: 1041–59
9. Karni A, Sagi D 1991 Where practice makes perfect in texture discrimination: Evidence for primary visual cortex plasticity. Proceedings of the National Academy of Sciences of the USA 88: 4966–70
10. Marr D 1982 Vision. Freeman, San Francisco
11. McKee S P, Westheimer G 1978 Improvement in vernier acuity with practice. Perception & Psychophysics 24: 258–62
12. Ramachandran V S, Braddick O 1973 Orientation-specific learning in stereopsis. Perception 2: 371–6
13. Vogels R, Orban G A 1985 The effect of practice on the oblique effect in line orientation judgments. Vision Research 25: 1679–87