CHAPTER 4: DISCUSSION
4.1. GENERAL DISCUSSION
The aim of this study was to record in parallel a subject's performance during a learning task and changes in event related potentials ERPs. It has to said that two aspect of this work were unforeseen during planning phase
4.1.1. Learning and Memory:
Learning is acquisition of new information or knowledge, and memory is the retention of learned information for the future.
Psychologists have studied learning and memory extensively. As a result, they have distinguished what appear to be different types. Several schemes have been proposed distinguishing different categories of memory. Cognitive psychological studies have shown that there are at least two distinct types of memory storage: the conscious recall of information about people, places, and things (explicit or declarative forms of memory) and recall of information about motor skills and perceptual strategies (implicit or procedural forms of memory). These two forms of memory have been localised to different neural systems within the brain. Explicit memory requires regions within the temporal lobe of the cerebral cortex, including the hippocampus; implicit memory involves the specific sensory and motor systems recruited for the particular task.
Memory for facts and events is called "declarative memory" and it is usually what is meant by the word "memory" in every-day usage and can be assessed by conscious recollection. Memory for skills or behaviour is called procedural memory that cannot be assessed by conscious recollection and it is more like "habit". These procedures can be performed without conscious recollection. Working memory is a more general term for temporary information storage that allows several types of information to be held at the same time.
How do we learn? "Practice" in a continuous regular way and you will have much progress. "Study" practice study-practice and learning takes time, and effort. Learning of anything continues over the entire life span. There is no end to the learning of any topic.
How does one learn? There can be no definite answer, with intellectual activities one problem of learning is simply the size of the task, few ideas seem difficult when examined in isolation. The task is not easy; the difficulty seems to lie in the interrelationships among the ideas that must be acquired. It is the totality of the topic that requires time and mental effort.
How do we remember? Some events are easy to remember and some not. Sometimes remembrance comes only with difficulty. Learning a person’s new name, telephone number, vocabulary of foreign language, etc.
To remember is to have managed three stages successfully. The "acquisition" when a sensory impression is registered as a memory. The "retention" once a memory is registered, it retained and can be recalled. The "retrieval" recall what registered and retained of the information. Failure at managing one of these three stages means failure to remember. Learning is very close to understanding, you learn how to play a game of chess when you understand the moves.
Was the first formal enunciation of the biological stage of the concept of memory formation. The permanent memory is probably a form of passive store which involves relatively permanent alterations to the connectivity between synapses in a neural network, while a shorter term active representation must be present, involving transient electrochemical events in the network.
The most complex biological system known to us is the brain and among the most complex questions concerns the mechanisms of learning and memory. Since we believe that learning/memory mechanisms occur at synaptic connections (Bailey and Kandel, 1993; Jessel and Kandel, 1993), the site of information transfer between neurons has become an interesting area for research.
Learning occurs in many ways. At the most basic level, the level of the neuron, learning is associated with the production of more glial cells, better blood supply, more acetylcholinesterase and thus acetylcholine (Rosenzweig, 1984), as well as more synapses per neuron (Turner & Greenough, 1985) and more complex dendritic trees (Greenough & Volkmar, 1973). While these changes could result from a number of factors, such as better health, they must be related to learning (Carlson 1991). Further, while it may not be safe to say that these changes necessarily represent learning, it is safe to say that they may be associated with learning.
4.1.2. Long-term potentiation:
Long-term potentiation is operationally and typically expressed as a long-lasting increase in synaptic efficacy (lasting from hours to days) following brief tetanic (high-frequency) stimulation of an afferent pathway (fibers). Thus, following LTP induction, a fixed amount of presynaptic stimulation induces a " potentiated" post-synaptic response, e.g., an increase in excitatory post-synaptic potentials (EPSPs). In 1966 Terge Lomo, working in the laboratory of Per Anderson initially observed the phenomenon of LTP. In 1973, the first full article described LTP in the hippocampus of the rabbit, a collaborative effort between Lomo and Timothy Bliss (Bliss & Lomo 1973), (Bliss and Gardner-Medwin 1973) Exploration of the mechanisms underlying LTP induction has been one of the most active areas of research in neuroscience. LTP was discovered after it was found that electrical stimulation of the perforant path, which connects the entorhinal cortex with the dentate gyrus of the hippocampus, resulted in "stronger" excitatory post-synaptic potentials in the ipsilateral dentate gyrus. This effect may last several months and thus represents a relatively long term strengthening, or potentiation, of neurons and their multiple synapses.
By 1989, the U.S. National Library of Medicine listed 312 articles with the term "long-term potentiation" in the title. In the 1990's alone, over 700 additional articles have appeared. This search vastly underestimates the research effort, because many articles that address LTP do not contain LTP in the title phrase or refer to the same phenomenon with a different name (e.g., "long-term enhancement" (McNaughton et al. 1986).
The concerted attention that LTP attracted over time is perhaps of no surprise to those familiars with the search for the engram (a neural memory store) and the associated mechanism that could account for its formation. Prior to the observation of LTP, the search had produced virtually no viable candidate mechanisms, at least in the vertebrate nervous system (Kandel and Tauc 1965). In this regard, LTP was and still may be the best candidate. In several recent reviews, different authors have concluded that not only is LTP a viable mechanism for the induction and storage of memories, but also is the most promising candidate (Morris et al 1991)
The most well studied synaptic pathway exhibiting LTP is the termination of the Schaffer collateral system of axons, projecting from pyramidal cells in hippocampal area CA3 and synapsing on the dendrites of pyramidal cells in hippocampal area CA1. This pathway has been implicated in the storage of declarative perceptual memories in humans and spatial learning in rodents. The most extensively studied form of LTP requires the activation of postsynaptic N-methyl-D-aspartate (NMDA) type glutamate receptors. The influx of Ca2+ ions through NMDA receptors triggers a partly understood cascade of enzyme activity which results in a long-term strengthening of the synapse (Lisman, 1994). The form of NMDA receptor dependent LTP which is expressed immediately following induction is called early LTP (E-LTP), which after about an hour is replaced by a late form of LTP (L-LTP) which depends on postsynaptic protein synthesis. This paper focuses on NMDA receptor dependent, early LTP at CA3-CA1 synapses, and refers to this phenomenon simply as LTP unless mentioned otherwise.
4.1.3. NMDA Receptors and Post-synaptic Calcium:
One defining feature of LTP is its dependence on high levels of postsynaptic calcium, a common feature of most learning-induced neuronal modifications. In and of itself, a definition which includes "calcium dependence" provides little insight since a wide range of cellular functions require calcium and still more are dependent on elevations of intracellular Ca2+ above basal levels. Although the exact role of calcium in LTP induction is a matter of debate, elevation of post-synaptic calcium is clearly necessary, and may even be sufficient for the induction of hippocampal LTP. Induction of LTP is prevented by a pre-tetanus injection of calcium chelators into the post-synaptic cell (Lynch et al. 1983; Malenka et al. 1988), and induction occurs when the postsynaptic cell is artificially loaded with the ion (Malenka et al. 1988). A great deal of evidence (Jahr & Stevens 1987) indicates that the primary source of calcium influx during the induction of hippocampal LTP occurs through an ion channel that is coupled to the NMDA subtype of glutamate receptor. This receptor is unique in that stimulation of the channel ionophore requires glutamate binding as well as a moderate level of depolarisation. At normal resting potentials (-70 mV), the channel is blocked by magnesium, and glutamate binding is insufficient to open it. However, at depolarised membrane potentials (> -40 mV), magnesium is expelled from the channel, which can then be opened by glutamate and which displays a high selectivity to calcium ions. Thus, the NMDA receptor complex is said to be dually regulated by two factors: ligand and voltage. These cofactors can be recruited through several means.
First, a relatively long, high intensity pre-synaptic burst of activity (such as a high-frequency train of stimulation) can induce LTP by releasing glutamate onto the postsynaptic receptor, while depolarising the postsynaptic cell through stimulation of the non-NMDA type of glutamate receptors (AMPA).
Second, shorter and more physiologically relevant levels of pre-synaptic activity can induce hippocampal LTP by stimulating the NMDA receptor with glutamate, while the postsynaptic cell is depolarised via an alternative means such as an input from a second afferent pathway.
Other forms of LTP, such as that induced in CA3 pyramidal cells following mossy fiber tetanization, occur independently of the NMDA receptor, and are instead dependent on Ca2+ influx through voltage-gated channels, with some debate as to whether the critical Ca2+ signal occurs pre- synaptically (Castillo et al. 1994), post-synaptically (Johnston et al. 1992). As alluded to earlier, even in area CA1, LTP can be induced without the participation of NMDA receptors, provided that the tetanus (or post-synaptic depolarisation) is of sufficient intensity to activate voltage-dependent calcium channels (Kullman et al. 1992). Thus, activation of the NMDA receptor is critical to many forms of LTP, but it is not necessary for all. In contrast, intracellular calcium appears to be a necessary element for the induction of LTP. A necessary role for calcium in LTP is consistent with LTP's presumed role in learning; calcium plays a critical role in many cellular modifications thought to underlie conditioned behavioural responses (Abrams & Kandel 1988, Matzel & Rogers 1993).
Long-term potentiation (LTP) is a long-lasting increase in synaptic strength (larger excitatory post-synaptic potentials or EPSPs) induced by high frequency, high intensity stimulation of the pre-synaptic neuron. LTP is exciting to memory researchers because it provides empirical support for the hypothetical strengthening of connection hypothesised by Hebb. It is a phenomenon most widely studied in vitro in the rat hippocampus slice, but it has also been recorded in cortex - both the cortex and hippocampus are considered to be possible sites for memory processing. In addition, LTP is popular because it has many of the characteristics expected of a memory mechanism, in that it:
Changes in the brain are recorded as plasticity whereas changes in behaviour or cognition are the result of learning. Psychologists have their own problems with learning, instrumental conditioning, impinting, recall, forgetting curves as well as learning curves.
Do both disciplines have the same aims? Not always, one wishes to understand behaviour and other the brain tissue. When they have come together in the past their joint findings have been most insightful. One might mention the much quoted case of HM (Scoville and Milner 1957) who underwent bilateral temporal lobe resection. In the context of these experiments HM learned how to solve several perceptual-motor puzzles but denied he had even seen the puzzle let alone solved it on previous occasion (quoted in Churchland P.S. 1986)
There are many interesting cases in the world literature beginning with Pheneas Gage who survived frontal lobe damaged which rendered a skilled worker into a feckless individual. Much work has been a careful psychological examination of individuals with specific and well documented brain lesions. This gives a psychological picture of deficits but there are pitfalls in integrating these observations into a theory of brain function.
Now things are changing - by the build up anatomical psychological maps of deficits, it is possible to build pictures of "facits". New techniques PET and fMRI give pictures of which parts of the brain are hard at work during a psychological task. Event related potentials give only poor spatial information but have very accurate temporal discrimination
It would have been very exciting and worth while to do these experiments with contemporary ERP and PET but that was not possible.
We divided the participants in each of the experimental groups into two subgroups according to their performance as learners and non-learners. In the second group (FR) all were learners. Males and females contributed equally to both groups. Due to selection of medical undergraduate and postgraduate students no real initial intelligence difference was present. As it happened all were right handed.
4.2.1. Levels of learner performance:
Learners can be described and classified according to their level of performance in the learning process.
The five levels identified by Pacific Crest are as follows (in increasing level of performance):
Trained Individuals, who have developed a specific knowledge base, with specific skills for a specific context.
Learned Individuals, who have acquired a broad base of general knowledge and can apply it to related contexts.
Lifelong Learners, who have developed the skills and motivation to self-facilitate their ongoing learning and can, apply it to a variety of contexts.
Enhanced Learners, who have developed a higher level of performance skills and actively seek new knowledge and contexts for application in a constantly changing environment.
Self-growers, who continually grow by using strong self-assessment skills to improve future performance
This highest level of learner performance is characterised by the following:
My subjects were all being trained to be lifelong learners and some would be expected to develop enhanced learners and self-growers. Reasons have been given earlier why all subjects tried and were attentive to the task. Debriefing gave some insights into their inner mental life during the task.
Since 1972 the researchers have studied the subject’s performance and ERPs accompanying feedback. They conclude that the feedback provides information relevant to past behaviour which may be used to modify behaviour in the future trials (Jenness 1972; Poon et al. 1974; Johnson and Dochin 1978, 1982, 1985; Stuss and Picton 1978; Perrault and Picton 1980; Ruchkin et al. 1980, 1981, 1982; Stuss et al. 1980; DeSwart et al. 1981; DeLisle et al. 1986; Papkostopoulos et al.1986; Warren and McDonough 1995). During debriefing, motivation and frustration seemed to contribute to performance. Certain individuals appear to be over competitive or to try too hard, resulting in incorrect answers provoking frustration. The ideal subject seemed to be a moderately motivated and competitive individual. No formal psychological testing was employed and these subjective views are based on observation and questioning. A further group seemed less motivated and therefore contributed to the non-learning group. These people perhaps didn’t want to be there and were thinking of other things and couldn’t focus on the task. Thus they were in a situation not conclusive to learning.
Task performance was virtually perfect with fewer than 2% of target trials missed across the learners in general. Less than 10% were missed across the non-learners and they reported that they could not make their decisions in the allowed time (2000msec). The subjects included some comments about their strategies for performing the task: Some divided the image on the screen into four quadrants and then tried to concentrate on one of these four quadrants and follow it from one image to the next trying to find the differences between them. Others tried to do the same but with modification like choosing another quadrant and if they gave incorrect answers, they moved directly to another quadrant.
Another group of the subjects did not divide the image and dealt with it as one part, by having a quick look at it searching out for the differences. Then by the time they become more able and much quicker in making an accurate decision. The learners reported that they have been confident about their judgement and decision.
Some of them divided the image into centre and periphery and they concentrated at the beginning on the periphery but they found no differences at all. Then they directed their concentration around the centre of the image and found differences.
The subjects get a clue after few trials. Some of them get the idea of the distributed elements. They understand that these features are distributed in different percentage, from one image to another. They found as well that these elements (features) look like a tuning fork facing different directions.
Few of them reported that they could not make a decision within the allowed time in the very first trials. Another group of the subjects reported that they were pressing the button without intention and they got the feeling of pressing the button automatically (unconsciously).
The non-learners reported that the task was very difficult and they got the feeling before they knew the result of their performance that they had not reached the cut off percentage to be classified as learners.
An interesting observation was into our subject’s strategies and beliefs. Only two out of eighteen learners were close in their explanations of the pattern of the correct difference between patterns. A further two explained left to right or waterfall effects, although were unsure. The rest of them claimed to ‘ just know’, and one subject reported to hearing a voice that coincided with pattern presentation calling ‘left’, or ‘right’. The overall impression was that the classification task was at subconscious level or could not be expressed in words.
Only two of the learners developed explicit knowledge of the task; they could use this knowledge to form attentional experiences regarding the next image in the task. This developed explicit knowledge did not improve the procedural knowledge per se but facilitated the behaviour used to measure procedural knowledge. Mishkin and Appenzeller (1987) reported that "behaviour could be a blend of automatic responses to stimuli and actions guided by knowledge and expectation"
Non-learners also suggested guesses about orientation and employed a multitude of tactics including tilting the head. Some attempted to visualise three dimensionalities to the image as in the ‘magic eye’.
Because the subjects were unable to declare the differences between the patterns the learning task is considered "non-declarative" or " procedural skill learning).
The learning curves for the learners and their correct performance throughout the all trials for each group shows that, in the first group (LFr) there was a steady increase in average performance. Although they started between 50% to 60% correct responses their curve during the first 100 trials shows that they have been moving up and down, but finished above 80%. The feedback gives good understanding about the performance and everyone of the learners tried to modify his/her performance according to the feedback. Over the two hundred they were able to develop a method for success. This is learning.
The explicitly instructed subjects with the given feedback in this study performed better than did those given the neutral instruction. Reber (1989) concluded that the explicitly instructed perform poorly compared with those given the neutral instruction and he claimed that his conclusion was related to the explicitly instructed subjects who took longer to memorise the exemplars, and they were poorer at determining well-formedness of test string. I suggest that the explicit processing of complex materials has an advantage over implicit processing
The "learners" in group LFR were strictly not learners at all as they had been told the rule of operation. However they had never performed the task and needed to develop their skill. From the start they achieved learner status (>70% correct answers). There was a slight drop in the middle of the task but they finished with the highest score of all. They were not perfect however, indicating the difficulty of the task even when the rule is known.
It would have been of interest to take subject from Fr and FR groups and put them through the task again to seek further improvement. This would have taken the experiments away from the learning task into the realms of skill development and I decided against this.
The third group learners (Lfr) performed the task with neither information nor feedback. Their start was very interesting as expected for someone guessing, about 50% correct answers. During the middle of the first hundred trials their curve wavered between 50% and 80% giving a zigzag shape. At the end they finished in the learning category very close to the subjects who had no feedback but had a rule. I relate that performance to the strong challenge they had and their hard and serious effort to get the clue. They seemed even surer than others of their responses did as their learning was unaided and totally self generated.
Finally the fourth group (LfR) which I gave the subjects all information and the feedback was not given after they made decision. After a little while they showed an up and down curve which indicates that they have been getting correct responses and then they lost the clue again and that was repeated at the end they finished well within the learning group.
Generally all learners achieved the criterion by the middle of the task. Improvement was rapid during the first 100 trials and continued at a lesser rate thereafter.
As expected those who knew the rules in advance performed best and were performing well from the beginning of the task. The other groups had a similar performance to one another. I did not expect group (fr) to do good performance as well but there were only 8 learners out of the 24 subjects in the group.
The learning curves of the non-learners have around 50% correct, which is to be expected. Group fR are interesting in that they know the rule, and for the first 100 trials (first two quartiles) or so they perform better than chance. Thereafter performance declines to chance level.
The learning process appeared to go through three phases, with individual subjects taking varied times on each stage. Firstly subjects were guessing resulting in 50% performance with strings of three or four similar answers at most.
During the second stage a seesawing of collections or strings of right or wrong answers was seen. The resultant runs extended to as many as 15 patterns correct in some cases. At this point the subject's tactic was to compare a past sensory image with the present stimuli. Subjects were only able to distinguish between the on-screen pattern and the previous image. A fragile model of the classification rule had been formulated relying on comparison to a working memory presentation. When a string of answers ended it was usually followed by an opposite run. For example, having got one wrong after a long correct string the classification rule seemed to break down for a while. Then another correct string appeared indicating a modification rule had been formulated.
In the third stage the learners appeared to construct a more robust model. The subject’s response was now independent of previous answers. This represents an understanding of which pattern corresponds to each button, subject have established a classification rule which works well. The non-learners never achieved this. They stayed in stage two or reverted to guessing. This May be due to frustration or surrender.
All subjects were motivated and attentive throughout. I observed them for signs of inattentiveness or drowsiness. Boredom provokes fidgeting; inattention would result in missing responses. All subjects were volunteers and interested to volunteer. They were not there just to earn money as no money was offered. While expecting that one can never know the contents of their mental life, the outward signs and the results of debriefing indicate that they were trying and attentive throughout the test. Some, no doubt, felt frustration but none gave up.
Barrett concluded (1996) that the ERPs appear to have a role to play in the assessment of the effect of used subject's strategies when performing cognitive task. An understanding of the relationship between strategy and ERPs morphology will be essential to an understanding the electrophysiology of complex cognitive function.
4.3. DECISION TIME
Decision time is the time taken from the onset of a stimulus or signal to the initiation of response. Since nineteenth century the decision time has been recognised as a potentially powerful means of relating mental events to physical measures.
The learners in debriefing reported that the task at the beginning was very hard and then became much easier by the time when they started to learn the differences between the two patterns. One of the most firmly established findings in some studies of discrimination, is that as some measure of the difference between two stimuli is decreased (hard task). The time taken to discriminate between them increases.
Barrett (1996) found that there are some subjects considered the task complex and others considered it as a piece of trivia, and when the complex cognitive function would be hard to achieve the increased complexity leads to increased work load and increased response time.
Naturally it takes more time to distinguish two very similar stimuli than two which differ greatly. The time to react in a situation in which any one of several signals may occur, each calling for a different response, must include four processes (Decision time includes four processes):
First and fourth steps are relatively short taking a few tens of milliseconds to receive the signal and give the response action like a simple button press. Second and third steps take longer for identification and choice and it depends on each subject performance.
The subjects were asked to respond within 2000msec. There was no encouragement for them to be quick or as quick as possible. The prize or goal was correctness, not speed. Most decision time experiments in the literature encouraged subjects to respond as fast as possible.
4.3.1. Learners and non-learners decision time:
The decision time was quicker for the learners than non-learners. The reduced decision times for the learners during the last quartile of the attempt and during the correct answers suggested that knowing the answers and responding is a quicker process than guessing answers.
The non-learners showed no improvement in the decision times from the beginning to the end of the trials. The non-learners (nLfr & nLfR) showed minimal improvement in the decision time for the correct answers which was significantly shorter than the decision time for the incorrect answers. Other two groups (Fr) did not show statistical significant changes between the correct and the incorrect trials answers, variance reinforced the hypothesis that, once learning mechanisms were effective, this increased processing speed.
Corresponding with a start and end ERPs difference, learners showed a significantly shorter decision time to non-learners at the beginning and the end of the trial. The decrease in the learners decision time is related to my explanation that the decision or the response time is affected by the self-confidence which subjects reported, after making a particular judgement.
There were no significant differences between the subjects decision time either in the learners group or the non-learners group for type A and Type B patterns as expected. The decision times for the learners during type A & type B were quicker than the non-learners resulted in from the differences in each of these groups' strategies and tactics.
For the correct and incorrect answers decision time were significantly different, for the learners but not for the non-learners. That finding is in line with the findings of Fernandez, et al 1998, that the reaction time is significantly different between the correct and incorrect answers for the group with good performance. Fujihara, 1998, found that the decision time for the target category was quicker than the non-target. Thus learning results from tactic employed within this extra exposure time of the stimuli.
On the level of each group separately, group one (Fr) the learners as a general conclusion I found were quicker than the non-learners. During the last fifty trials in particular the learners were quicker than the non-learners were.
Group two (FR) who got the all instructions for good performance had unfortunately no non-learners to compare. We can compare the learners in that group (LFR) with the rest of the learners. I found that the learners decision time during the last quartile in this group (LFR) were not quicker than the decision time of the learners of group (LFr) and they are quite similar to group three (Lfr) and four (LfR). I believe the decrease in the decision time was according to the given feedback to the group one subjects, which helped them to make a quicker decision. In the same time the more instruction the slower decision time, perhaps they take more time to be surer that they made accurate decisions.
Group three learners (Lfr) were quicker than the non-learners were. This group performed the task without either feedback or instructions. The subjects decision time for the learners, and even for the non-learners were slower than group (LFr) who performed the task with given feedback and no instructions given as well.
Group four (fR) the learners were quicker than the non-learners and I believe that it is related as mentioned before in the previous paragraph to the given feedback and could be related as well to the instructions and the information which have been given before involved in the task.
The learners built up their own explicit knowledge and by using the given feedback in this study they improved their performance and the time taken to make their decision. Willingham et al. (1989) found that at least some participants with high amount of explicit knowledge used this knowledge to anticipate the following stimulus, and as a result had reaction time much faster than could have been achieved if they had waited for the stimulus before beginning the finger movement.
4.3.2. The subjects co-operation:
It might be argued that those who did not learn were not trying. Perhaps they decided the task was too difficult and they were mentally "switched off" perhaps their attention wandered or drowsiness set in.
According to the debriefing this did not happen but subjects may deny the experimental access to their inner mental life. There were several reasons against non-co-operation.
All these features suggest that the subjects were co-operative and entering fully into the challenge of the task.
Although the decision time for non-sense stimuli in the present study was not available, a measure of accuracy was. Since the accuracy of responses must depend on the availability of sufficient information about the stimuli, one might expect to find a relation between response accuracy and the timing of response relative to the P300/P600.
Indeed Coles et al (1985) found such a relationship and suggested that the P300 latency is dependent on response accuracy. In a related vein, Kutas et al (1977) found that the correlation between P300 latency and the decision time is higher when subjects are told to concern themselves with accuracy rather than speed. Both these findings suggest an association between the accuracy and latency of P300.
4.4. Event Related Potentials (ERPs)
4.4.1. Event Related Potentials
Generally the brain ERP results reflect several findings related to the impact of learning on the underlying brain circuitry.
Establishing the functional significance of cognitive ERPs requires the identification of both their cognitive correlates and their neural origins. Progress on each of these fronts is accelerating rapidly and it is important to note that work on event related potential is beginning to connect psychological ideas with brain activity. At present the cognitive event related potential research is more closely integrated with midstream cognitive psychology than before. Although the absence of comprehensive information about the neural basis of an ERP places strong constraints on the conclusions that could otherwise be drawn, the importance of ERP recording in normal subjects and patients is now clearly recognised and much research is being devoted to elucidation of the processes underlying it. Consequently, when a significant body of knowledge concerned with the cognitive and neural origins of event related potential has been accumulated, ERP studies of human will have come of age.
Studies have demonstrated that P300 amplitude and latency can be used as indices of the nature and timing of a subject’s cognitive response to stimulus since P300 discovery by Sutton and his colleagues (Sutton, et al 1965 and Sutton et al 1967).
It is well established that P300 is elicited by unexpected events, and that the lower the subjective probability of an event the larger will be the P3 it elicits (Duncan-Johnson and Donchin 1977). A larger positive wave occurring sometime after 300msec post-stimulus onset is variously termed P300, P3, or Late Positive Component LPC (Sutton, et al 1965 and Sutton et al 1967; and Donchin et al 1987).
Reports of visual and auditory modalities in normal subjects show that there are two types of P300. These two positive component are sometimes seen in simple targets detection tasks consisting of a fronto-central P3a and P3b. Correctly detected targets elicit a large P3b and non-targets elicit fronto-central P3a. Picton et al 1984 reported age-related changes in the ERPs that there were decreases in P3 component amplitudes (-0.25 micro-volt/year) and P3 latency (1.41 msec/year). This small change will not affect my results, as the subjects were all young adults.
Its amplitude (size) and latency (timing) measure the positive peak. Amplitude (m V) is defined as the voltage difference between a pre-stimulus baseline and the largest going positive or negative peaks of the ERP waveform.
Latency (msec) is defined as the time from stimulus onset to the point of maximum positive amplitude within the latency window.
P3 scalp distribution is defined as the changes in component amplitude across the midline recording sites from the frontal (FZ), central (CZ), parietal (PZ) locations. Amplitude and latency are very important and effective on the scalp distribution, since variation in P3 measures from the manipulation of task or subject variables has been used to infer information about the underlying neural generators (Johnson 1993; Polich and Heine 1996)
The P100, N100, N200, P200, and P250 peaks are considered as early sensory components, which is consistent with findings in a variety of studies using visual stimuli (Mangun and Hillyard 1988).
4.4.2. Learning process and ERP positivity
The idea was to see if the event related potentials were associated with learning or not
Through my research many experiments were performed to explore and study the morphology and characteristics of positive components of cognitive event related potentials in order to investigate the neurophysiological and neuropsychological processes underlying cognition.
Is there any ERP components related to the task?
It is apparent that subjects attempting the task show a positive shift over frontal areas. Those merely observing the patterns have no positive shift.
The positivity is a long duration change with no clearly defined onset at about 200msec. It decreases towards the end of the time window. Pattern A and B produced the same effect. There was no reason to suspect the two patterns should produce anything different.
When subjects who learned or who made correct responses are compared it was clear that there is no unique waveform associated with learning. The difference between the groups of subjects is one of degree. To measure this degree the mean amplitude during two time windows was determined. The positivity has been referred to as P300 and P3 because it clearly developed at 300msec. It is not intended to imply that it is the same as the classical P300 associated with oddball stimuli.
Perhaps a better term for it would be "Positivity Associated with Learning" or (PAL).
In the present study I selected two time windows 250msec to 550msec and 550msec to 850msec. These cover positive peaks occurring during both time windows. The group FR was the most positive in the learning groups during the last fifty trials. Others groups of learners waveforms showed similar peaks of less amplitude. P300 has been associated with increased memory effect, depth of processing and comparisons of stimuli (Fiztgerald and Picton, 1981, Donchin 1981). This is a representation of the processing involved in comparing the stimuli to previous patterns. The increased P300 after learning is agreement with previous studies (Karis et al, 1984). The ease of pattern to category assignment, an initial categorical perception step for learning will contribute to the increased P300 amplitude (Duncan-Johnson, 1981). Mecklinger and Ullsperger (1995) showed that P300 increases in amplitude with ease of stimulus categorisation.
Further experimentation is needed to see if the positivity decreases when the category can be easily distinguished, as hypothesised. A parallel can be drawn with Friedman and Sutton’s 1987 experiment, which used different task hardness, showing increased P300 to categorical memory task.
Positivity is present at the start of the task when there has been no learning but gets bigger and stronger in learners, so part of it may be due to learning.
Positivity is still present in non-learners so parts of it are related to attention and trying.
I am trying a logical and synthetic approach. The striking differences between the passive (observers) and all these who were trying to learn is a late positivity seen most strongly over the frontal areas. There is no unique waveform associated with learning and giving correct answers, but there is a change in the quality of the underlying wave.
The beginning of the trial for most of learners showed different ERP’s morphology from the non-learners. The unique feature of this ERP is the negative peak; the learner’s waves had more negativity than the non-learners waves. This initial negativity could be hypothesised to be the result of a process that is essential for later learning. For example a foundation block for a robust model of learning. This would insinuate that a prediction could be made from an ERP at the beginning of the trial as to whether a subject will learn, possibly reflecting a difference in tactics involved.
The learners and the non-learners in all groups in my research experiments showed more positive activity compared with the passive group (observers), but the positivity was more still for the learners than non-learners. The learners in all four groups showed nearly the same trace (waveform) morphology.
The non-learner's slow decision times were associated with much more prominent negative going peak at latency of 200msec whereas for the learner's fast decision time the negative going peak appears as much as smaller potential at the same latency.
The learners in group two (LFR) had more positive going ERPs and if we arranged according to the positive activity we could have that classification group two first (LFR), followed by group one (LFr), then group four (LfR), and at the end group three (Lfr), and I claim that the information given played a very important role. At the same time the feedback was not less important to induce positivity. When we mixed the rule with the feedback the learners have got more positive going waves than when we gave just the feedback without rule as group two which is more positive going as well as the waveform for the learners who have got just the rule without feedback and the less positive group waveform was the third group where they have no rule and no feedback.
The positive peaks were earlier for group two (LFR) and group four (LfR) than group one (LFr) and group three (Lfr) and I explained that the learning processes was quicker for the groups who have got the rule and slower for the groups did not. The group two positive peaks were earlier than group four which means that the feedback could be responsible for the differences and that is confirmed by looking at the group two which was earlier than group three who have not neither the feedback nor the rule.
The non-learners in group two (nLFr) and group four (nLfR) had the same positive peak with the same amplitude and both groups positive peaks were earlier than the less positive peak in case of group three (nLfr) and that confirms the effect of the feedback and rule in the processes of learning even in the non-learners which means by another means that the groups of non-learners had the similar processes but it was not complete for them to make them able to succeed.
The differences between the non-learning ERP traces at the beginning and end of the trial are due to many interactive factors. Firstly partial learning could be responsible. Some of the processes needed for learning could be in place but not perfected resulting in ERP changes. A different way of dealing with the stimuli will have evolved through the trial. Boredom and/or frustration are other contribution.
Standard components can be identified from event related potentials. This is especially true for the learning group traces figures showed P100, P300a/b, N100 and N200a/b (Czigler, 1995; Polich 1990). The pre-stimulus baseline and sensory portion, until 200 msec. are very similar in all subjects, suggesting pattern induced matching brain processes. The P200 was increased for the learners after learning due to correctly recognised target stimuli (Czigler, 1995). From 200msec to 300msec the learning groups end ERP, showed an increased N200b. A positive peak was visible around 300msec. For the learners group when compared with the observers group, the observers did not show any positive activity around the same latency. That changes in ERPs most probably related to the learning processes, and the decision-making processes in the learners and non-learners group. But not applicable to the observers group who did not ask to learn or to make any decision. This finding is in the agreement with the finding of (Cutmore and Muckert 1998), who found a relationship between P300 amplitude and word distance on the underlying metric, was found only for the decided group. This was interpreted in terms of previously documented relationship between P3 and the constructs of decision confidence task difficulty.
4.4.3. The Correct Answers:
The positivity is stronger for the correct answers, and there is a negative wave.
The correct answer trial potentials in the learners group were more positive from about 300msec and up to about 900msec post-stimulus than the incorrect answer trial potentials for the learners in all groups. The same comparison done for the non-learners in all experimental groups and there was no difference between the correct answer trials and incorrect answer trials for the non-learners. One of the most interesting finding when I compared between the learners incorrect answers and the non-learners incorrect answer was showed no statistically significant difference and with the non-learners correct answers showed no differences which gave me impression that the brain learning processes and the ERPs elicited by the brain in these three conditions were similar. I could conclude that the more correct answers the more early positive peaks, and was due to learner’s early correct decisions matching. The learning processes, is represented by positivity. The negative peaks within the selected time windows were larger for the non-learners than the learners and that as well related to the learning processes and the decision making for both groups. We might explain that by looking at the correct and incorrect answers waves, where we found that more negative activity the more incorrect answers, and vice versa the more correct answers the less negative activity.
The building block that is essential for a robust model of learning would be a direct association of pattern to the left or right mouse button, similar to Braida’s (1969) context parameter. This process is partly responsible for the longer reaction time at the start. This would imply that the comparison to previous stimuli and association to mouse button has components that are dependent on each other. The resolution of this secondary process produces positivity at 830 mescs.
Both fragile and robust learning models are acting concurrently and rely on categorical perception. Knowing which group each pattern is in, or at least discriminating between them is essential for any sort of success.
Subtraction brainmaps for the elicited ERPs during the four quartiles of the learning task showed over all voltage changes starting from 200msec to 1000msec. these are related to the learning processes and decision making. The learner's cartoon brainmaps of subtraction showed a major frontal and temporal positivity that was especially predominant on the right-hand side. The subtraction for the four quartiles showed that the positive activity increased gradually throughout the task to reach the maximum during the last fifty (the fourth quartile) due to cumulative changes. The role the parietal and tempro-parietal area brain activities remain unclear and it has been shown that the activity within this region can be modulated by voluntary attention. The unique findings in that comparison was the unremarkable differences between the brainmaps during the first fifty in group two (LFR) when compared with the last fifty for the experimental groups, and I conclude that the learners in group two performance was nearly the same from the beginning to the end because they have been able to learn the differences and get the task clue from the start point.
4.4.4. The Event Related Potentials Generators:
The brain structures generating the ERPs are unknown, despite efforts using extracerebral MEG topography, scalp recording EEG recording in brain-lesioned subjects, animal models, and recording directly from the depth of the human brain (Halgren et al 1986; Paller et al, 1992).
Efforts to locate the neural generators of the P300 component of the event-related potentials have focused on studies using intractable epileptic seizures. These experiments have consistently revealed the presence of P300-like potentials in medial temporal lobe structures (i.e., amygdal, uncus, and hippocampus) in response to stimuli in the oddball task (Halgren et al., 1980, Stapleton & Halgren. 1987). Such reports have prompted efforts to determine whether temporal lobectomy surgery has any effect on the lateral symmetry or overall amplitude of P300, because this operation results in the unilateral removal of these medial temporal lobe structures. Unilateral removal of one of two symmetrical P300 generators should result in either asymmetrical and/or reduced overall P300 amplitude. However a number of experiments have failed to find any statistically significant differences between these patients and normal controls (Stapleton, Halgren & Moreno, 1987). Unlike shorter-latency components' of the ERP, the P300 is considered to be sensitive to the cognitive processes elicited by a stimulus, but relatively insensitive to its physical properties. It is therefore generally accepted that the P300 arises from a modality-independent neural generator. Whereas the early intracranial electrode studies relied exclusively on auditory stimuli, more rescent studies have found that both auditory and visual P300-like components are generated bilaterally in the same medial temporal lobe structures (Stapleton & Halgren, 1987). Such results are consistent with the modality-independent generator hypothesis.
Eric Halgren et al (1995) recorded from 537 sites (121-left hemisphere, 416 right) in the superior temporal plane and parietal cortex of 41 patients. Depth electrodes were implanted to localise seizure origin prior to surgical treatment. Subjects received an auditory discrimination task with target and non-target rare stimuli (standard oddball paradigm).
They distinguished three response patterns:
(2) In the posterior cingulate and supramarginal gyri, a sharp 'triphasic' negative-positive-negative waveform, which peaks at about 210msec-300msec-400msec, was observed. This waveform was of relatively small amplitude and diffuse, and seldom inverted in polarity. It was multi-model but most prominent to auditory stimuli, appeared to remain when the stimuli were ignored, and was not apparent to repeated words and faces. The 'triphasic' pattern may embody a diffuse non-specific orienting response that is also reflected in the scalp P3a.
(3) A broad, often monophasic, waveform peaking at about 380msec was observed in the Superior parietal lobe, similar to that which has been recorded in the hippocampus. This waveform could be of large amplitude, often highly focalised, and could invert over short distance. It was equal to visual and auditory stimuli and was also evoked by repeating words and faces. This broad pattern may embody the cognitive closure that is also reflected in the scalp P3b or late positive component. In summary, depth recording supported by other data clearly demonstrated that a P3 is locally generated in the hippocampus. Furthermore, scalp recording after hippocampal lesions clearly implies the existence of other P3 generators. The location of these generators has been suggested by depth recording, but such recording has been limited in scope.
Event-related potentials were recorded by Baudena et al (1995) from 991 frontal and peri-rolandic sites (106) electrodes in 36 patients during a discrimination task with target and non-target (distracter) rare stimuli. Variants of this task explored the effects of attention, dishabituation and stimulus characteristics (including modality). Rare stimuli evoked a widespread triphasic waveform with negative, positive and negative peaks at about 210msec, 280msec, 390msec, respectively. This waveform was identified with the scalp ERP complex termed the N2a/P3a/SW (slow wave) and association with orienting. It was evoked with rare target and distracter auditory and visual stimuli, as well as by rare stimulus repetition or omissions. Across most frontal regions, N2a/P3a/SW amplitude changes only slowly with distance. In summary this study demonstrated an early P3a-like activity that polarity inverts over short distances in the medial frontal lobe, and that it has a significantly shorter latency than similar potentials recorded in the temporal and parietal cortices. It is clear from the results of several investigators that rare stimuli evoke multiple overlapping components over the 100-600msec latency range and that each component has many generators, so to search for exact correlation between depth and scalp peaks would be fruitless. It must be pointed out that even across different scalp sites, that components bearing the same name often do not have identical latencies or task correlates. Further research is necessary to determine the exact locations of ERPs' generators. With reference to the findings above about the generators of the ERPs so far, 1 could not confirm with confidence which parts of the brain were involved in generating ERPs obtained in
Ji et al. 1999 were examined the topographic relation of P3 between the visual and auditory modalities, especially to examine whether there were any modality-specific hemispheric differences of P3 in normal adults. They were used auditory oddball task and visual paradigm with novel stimuli. They concluded that the topographic similarities between P3s recorded in the visual and auditory modality out number the differences. Profile analysis of P3 topography support the hypothesis of multiple generators of P3 that were differentially active in processing stimuli for different sensory modalities and were not symmetrically distributed between the two hemispheres.
1 may suggest that the frontal lobe and also the temporal lobe have an important role in eliciting positive components. To validate these finding with more confidence 1 need to record ERPs from subjects with learning disabilities and a much bigger group of normal volunteers.
We found that the positive activity was not symmetrically distributed between the two brain hemispheres for all groups in who learned or even in who tried but did not. Different brain area were involved in the task from the beginning to the end of the task by different percentage especially the parietal lobe which involved during the early stages and these results supported by the findings of Walsh et al 1998. The learning of perceptual skills is thought to rely upon multiple regions in the cerebral cortex Poldrack et al 1998. Reber and his colleagues 1998 conclude that the decrease activation of visual cortex when categorical patterns were being evaluated suggest that these patterns could be processed in a more rapid processing categorical patterns could be related to any of several processes involved in retrieving information about the learned exemplars.
4.4.5. Regional differences
Dipole analysis assumes only a very limited number of localised dipoles. We are dealing with widespread brain activity so, it does not work because there is no dipole solution for the dipole fitting equation, and it will be misleading to give dipole, so we did not use it.
Comparing the hemisphere data showed major differences. The right-hand side hemisphere, in conjunction with other visual and categorical perception tasks, shows increased positivity from 300msec onwards (Berlteson, 1982). The positive event related potential (ERPs) is considered to be closely related to cognitive processes, scalp positive activity latencies increase, parietal positive activity scalp amplitude decrease and the scalp potential field shift to relatively more frontal distribution (Anderer et al. 1998). The ERPs to pictures, but not to words, also demonstrated frontally distributed Old/New effects, which shifted over time from a left to a right-sided maximum (Schloerscheeidt & Rugg, 1997). Mecklinger and Meinshausen, 1998, found that the effects in the second time intervals may play a functional role in post-retrieval processing, such as recollecting information from the study episode or other processes operating on the products of the retrieval process, and presumably were mediated by right frontal cortical areas. Thus cognitive learning processes are predominantly a right hemisphere.
Studies of children learning to read or learning a language with novel script find a right hemisphere advantage at early stages of learning that shift to a left hemisphere advantage as reading became skilled (Silverberg et al 1979), and 1980). Behavioural research indicates that the right hemisphere performs memory judgement about specific visual items more quickly and accurately than the left hemisphere (Marsolek et al 1994, and Metcalfe et al 1995).
The right hemisphere also is more responsive to novel stimuli than the left hemisphere (Bradshaw and Nettleton, 1983). Conversely, The left hemisphere performs judgements about prototypical examples of a visual concept more rapidly than the right hemisphere (Marsolek 1995)
There were no statistical significant differences for any of the individual electrodes for the recorded ERPs from the temporal lobe, anterior temporal and parieto-temporal lobe. But from the brain maps the over all brain activity showed that the temporal area especially the anterior part to the right-hand side was more active like the frontal area early and late during the task which lead to the conclusion that these areas were involved in the learning processes.
Discriminating the target from a standard stimulus processing could initiate right frontal engagement, because such a processing requires the consistent application of attentional focus a major attribute of frontal lobe function (Pardo et al. 1991 and Posner 1992)
Positivity continues after button press and the given feedback appearance.
Is this the same neural process or not- it may be something different.
The additional or continuing positivity in the ERPs waves after the feedback, which was given to the subjects, was seen especially in the learners giving correct answers. Result in using the information given to the subjects by the feedback to modify them coming responses, and it is more positive in the learning group due to the expected feedback which was correct for most of the trials. The differences between the learners and the non-learners groups were not statistically significant at that latency (after button press). Johnson 1986, in his review of the ERPs feedback studies, noted involvement of additional cognitive processes after the stimulus categorisation as a result of feedback.
The present observation strongly supports the view that the learning processes takes place in neural regions located in the human frontal and temporal areas of both hemisphere especially on the right-hand side. The results suggest that the elicited positive activity depends on the processes taken by every subject trying to learn the differences between images type A and type B and depends on the memory search necessary to obtain the meaning of the stimuli and the meaning of each different category.
This is supported by Pouthas et al (2000) where they assume that the right frontal area plays a specific role in the formation of temporal judgements. Monfort et al (2000) reported that both left and right hemispheres especially the frontal lobe involvement is necessary for recognition of temporal information.
Research into this should pose interesting questions. To do this the task will need to be modified. Further studies into ERPs changes after the learning task has been competently dealt with are needed. Changing the task will show whether the ERPs are faithful to learning or depend on the stimuli.
4.5. Comparing of my results with Seger
4.5.1. Seger C. et al 2000 studied brain activity while people learned to distinguish between two novel visual prototype stimuli taken from Fried and Holyoak (1984) using functional magnetic resonance imaging (fMRI). They displayed total of 98 categorisation trials and were broken into four quartiles (four groups of 24) for purposes of analysis. During the scan the stimuli were presented for 2500msec, during which participants responded by pressing optical switches, with feedback given "smith" or "Jones" appeared for 500msec.
They had 6 learners and 4 non-learners out of the subjects participating in the task. They decided the cut off score for the learners and non-learners was 83% correct answers. Learners got 64%, 82%, 79%, 79% during the four quartiles, and their mean classification score was 91% in the final quartile. The non-learners got a classification score 61% in the final quartile.
Functional MRI shows increased metabolic activity during the task and Seger et al's task was similar to mine. She is able to report better spatial information; table (4.1.) summarises those areas of cortex showing increased activity during the classification than baseline across trials. While this is interesting spatial data, she gave no indication of how much metabolism in these areas increased. Her subjects learned quickly and frontal activity fell away later in the test, learning was accomplished.
Table 4.1. Shows the learning related brain activity change during learning in Seger et al 2000 study
The only group of subjects similar to Seger experimental group is group one (Fr) with feedback without rule, 18 learners out of the 34 subjects.
4.5.2. Learning brain activity changes in my study:
During the first quartile the occipital lobe the first showed positive activity, the parietal lobe and the central involved from the beginning about 200msec post-stimulus and even after the subject made responses. The frontal lobe and temporal lobe showed negative activities pre-stimulus and then not involved as much as the other areas, around 400msec post-stimulus showed greater positivities bilaterally than other areas.
In the second quartile the areas did not show too much changes from the first quartile. All the brain areas were involved by different amount of positive and negative activities.
In the third quartile the occipital area first showed positive activity and all other areas showed negative activity. The all areas showed positive activity and the frontal lobe showed more positive activity bilaterally.
In the fourth quartile there was not much difference from the beginning of the third quartile and then around 500 msec post-stimulus the frontal lobe area showed greater positivity especially to the right-hand side.
My study results give interesting time data in that the learning associated with positivity begins at 200msec and continues, but in an attenuated form, even after feedback. This strongly supports the idea that learning involves anticipation and planning for the next trial. The performance differences between learners and non-learners were associated with one reliable brain activation difference.
The main positive amplitude appears at the occipital sites at 100 to 200msec, which represents the sensory portion of the ERPs and this did not increase across quartiles.
Anterior temporal (+++)
Anterior temporal (++)
Anterior temporal (+++)
Anterior temporal (++)
Anterior temporal (+++)
Anterior temporal (++)
Anterior temporal (++)
Anterior temporal (+++)
Table 4.2 shows the learning related brain activity changes during learning in my study
(+ = 2m n ) As many of + sign means how much the area was actively involved in the learning task.
The ERPs traces start to increase the positivity at 250msec and end nearly about 1000msec post-stimulus after the button press which may lead me to say that the brain activation appear to be related to the constant processes of visuo-spatial analysis of stimulus and categorical learning processes
There are significant difference between the learners and the non-learners. We have two peaks of positivities: the first one latency range 350msec to 480msecpost-stimulus; the second positive peak latency range 550msec to 750msec post-stimulus, which is more, delayed for the learners than the non-learners.
The temporal lobe area showed significant activities bilaterally, the parietal and the central areas did not reach the threshold of significance changes from the first quartile to last one, but both were present from the beginning and persist throughout the learning task. They did not differ reliably in learner and non-learners
For the learners there are differences from the start to end, and for the non-learners it stays constant.
I could conclude that the frontal lobe was significantly involved throughout the task and the right-hand side frontal hemisphere activity increased significantly across the quartiles. The left-hand side hemisphere activity was present over all the quartiles but did not increase significantly.