Introduction

The neuroMod group is sponsored by an NWO PIONIER grant to Prof. Dr. Jaap Murre, which runs from 1998 to 2003. The original grant proposal is available on request. Below, we describe the projects and outline any recent progress.

The projects of the Neural and Cognitive Modelling Group (neuroMod group) focus on some of the psychological consequences of the dynamic rearranging of connectivity in the brain by posing a number of inter-related questions, addressing first of all (1) whether and how memory and other cognitive functions retain their stability in a changing brain, and (2) how the brain is able to repair itself at all, (3) to what extent brain repair can be stimulated and guided using specific rehabilitation strategies, and finally (4) whether and how we can relate with greater reliability the connectivity of the brain to cognitive functioning. In their general form, these questions are:

1. Regarding the long-term stability of cognitive representations: The general question is: How can memory representations persist in a changing brain, and still be recalled forty or more years later? This question is elaborated in series of more specific questions: Is there indeed a process of long-term memory consolidation, as is often hypothesized? Is dream sleep involved in such a hypothesized process of consolidation, and if so, how does it operate? What makes the preservation mechanism favor consolidation of some memories over others? How can the characteristics of different types of amnesia and dementia be related to these processes? What brain structures underlie these processes?
2. Regarding brain repair: How is the brain able to repair itself? Can it use the same mechanisms of learning and long-term consolidation for brain repair as those that are hypothesized to be used in the preservation of memories? Or are perhaps special-purpose mechanisms operative in the brain? When can a cognitive representation still repair itself, and when is it ‘beyond repair’?
3. Regarding rehabilitation: In cases of chronic or acute brain damage, to what extent can memories and skills be repaired without rehabilitation or other ‘external cueing’? At what times and in what form is active rehabilitation (guided recovery) most likely to be successful?
4. Regarding brain connectivity: What is the connectivity structure of the adult human brain? To what extent does this global connectivity constrain the formation of memory representations? To what extent is the structure of the brain the product of specific psychological requirements (e.g., being able to continue to function even with small brain lesions) that have worked on the formation of the brain during millions of years of evolution?

To make these questions concrete, we investigate them through a series of specific computational modeling studies. In order to do so, we also develop software for neural network simulations and we study relevant software development and analysis methods.

These questions are addressed in detail in five programs, which in turn are divided into a total of 16 projects. We will discuss the progress made in each project separately.

Program 1: Consolidation of memories

Project 1.1. Mechanisms of long-term memory consolidation

This project aims to investigate whether a long-term consolidation mechanism can explain some of the major findings in the neuropsychological literature. A connectionist model has been developed, called the TraceLink model that implements an autonomous ‘off-line’ conslidation process. The model consists of three subsystems: (1) a trace system (neocortex), (2) a link system (hippocampus and other areas), and (3) a modulatory system (basal forebrain and other areas). The model is able to account for many of the characteristics of anterograde and retrograde amnesia, including Ribot gradients, transient global amnesia, patterns of shrinkage of retrograde amnesia, correlations between anterograde and retrograde amnesia or the absence thereof (e.g., in isolated retrograde amnesia). In addition, it produces normal forgetting curves and can exhibit permastore. It also offers an explanation for the advantages of learning under high arousal for long-term retention, and for the effects of depth of processing. We, furthermore, have derived testable predictions concerning implicit memory, forgetting gradients, and the neuroanatomy underlying memory and amnesia.

Project 1.2. Model of the neural linking process

Synchronized neural activity and spatio-temporal patterns may play an important role in cortical information processing, in particular to connect faraway sites on the cortex in a slow neural linking process. Synchrony of synaptic inputs enables reliable transmission of information in the cortex through the groups of neurons involved at different processing stages. It has been proposed that the so-called ‘synfire chains’, circuits in which synchronous firing is propagated from one group of neurons to the next set, are the substrate for precise transmission of synaptic signals (Abeles, 1982, 1991). With Antonino Raffone from the La Sapienza in Rome, we have developed a model in which long-range cortical chains can emerge by self-organization in an initially unstructured network model through competitive Hebbian plasticity. A new neural network learning rule with connection weights resulting from both afferent and efferent synaptic competition is introduced in this model. This learning rule reflects a ‘limited resource principle’ of brain plasticity. Given the appropriate connectivity and plasticity conditions, a cortical chain will grow from a repeatedly stimulated group of neurons. Depending on the relative refractory period of the cells, either feedforward or recurrent circuits, possessing a higher number of internal loops, may emerge in the cerebral cortex. This model is also used to explain certain limitations in the capacity of working memory.

Although this project can be called completed in as far as the actual project description is concerned, we are working on several models that further explore the behavior of the linking process, for example, how the hippocampus can guide the neocortex during learning of sequences (some successful simulations have already been carried out). We are also working on a model of working memory, which we plan to integrate with a hippocampus-neocortex model of long-term memory. Work on a more detailed model of the hippocampus-neocortex has also started. All of this work will be done in cooperation with Antonino Raffone (La Sapienza, Rome).

Project 1.3. ‘Dreaming neural networks’: Combining slow-wave sleep and REM sleep

On this project, which extends Project 1.1 by adding specific sleep stages to the consolidation process, Martijn Meeter has carried out some simulations. We are currently not certain that this project will be continued, because the original Crick-Mitchison hypothesis (i.e., about the assumed neural cleaning-up operation by REM sleep) may not work in practice.

Project 1.4. A model of semantic dementia

Semantic dementia is a recently documented syndrome associated with non-Alzheimer degenerative pathology of the inferolateral temporal neocortex, with relative sparing (at least in the early stages) of the hippocampal complex. Patients typically show a progressive deterioration in their semantic knowledge about people, objects, facts and the meanings of words, yet, at least clinically, seem to possess relatively preserved day-to-day (episodic) memory. In this project, the following issues are addressed: 1) the relative preservation of category-level (superordinate) compared to fine-graded (subordinate) semantic knowledge as the disease progresses; 2) the preservation of new learning, as measured by recognition memory, early in the disease; 3) increased long-term forgetting of newly learned material; and 3) impaired implicit memory. On the basis of simulations thus far with the TraceLink model we conclude that recent findings from semantic dementia offer strong support for the view that memory consolidation in humans is dependent upon interactions between the hippocampal complex and neocortex. In this project, we cooperate closely with a research group at the MRC Cognition and Brain Sciences Unit in Cambridge (UK).

The future of this project will mainly concern further detailing of the model architecture and model behavior. In particular, we are in the process of adding a category layer to the model, placed between the trace and link system. In this layer, new representations must be formed in a process of new learning. New patient data are still being recorded actively in the semantic dementia field, notably through patient testing, fMRI, and post-mortem neuroanatomical analysis. we will strive to accommodate these new data or—preferably—to predict them.

Project 1.5. Implementation of the modulatory system

Preliminary simulations have been carried out, in support of the other projects. Extensive single-neuron modelling is currently undertaken, investigating the modulatory role of specific neuromodulators.

Project 1.6. Localization of memories in the brain

This project has temporarily been redirected and combined with a new project, aimed at investigating the role of the hippocampal reactivation of memories in sleeping rats (see Project A.3 below). We constantly monitoring research into the localization of memories in the literature. A paper on this topic has been planned with Dr. Kopelman.

It is not certain we will return to the original aim, namely to write a paper on where memories are located, because it seems less and less clear that such an undertaking could be successful. In this rapidly developing field, the sheer amount of data, the fragmentation of the data in different subdomains, as well as the lack of precision in the data hamper a clear sight on the locations of memories in the brain. We will reconsider this issue in the Fall of 2001.

Project 1.7. Integration with other models

We have implemented a working version of Shiffrin and Raaijmakers’ successor to their SAM model, called REM. Integration of REM with the TraceLink model has not yet been attempted, although some insights into what roads to follow have been gained in this implementation process and through discussions with Raaijmakers and Shiffrin, who has visited our group several times in the past two years.

The work on Project 1.7 took an unexpected turn when integration with a hitherto unrelated project was suddenly achieved in the Spring of 2000. Dr. Tony Chessa (applied mathematician) is postdoc with Prof. Dr. Murre on an ‘Aandachtsgebied’, working on models of language learning. Mathematical abstraction has in the past year yielded a completely new mathematical framework for modeling learning and forgetting. This framework is based on the mathematical theory of point processes and thus far we have obtained very good fits to over 280 different forgetting and learning curves (since 1885). The model is in many ways a ‘natural’ abstraction of the TraceLink model, which has two systems or ‘stores’ (hippocampus and neocortex). This idea has been extended in the point process model to an arbitrary number of stores. Under some simplifying assumptions, without much loss of generality, we are able to derive learning and forgetting equations analytically.

This model can also directly be compared with the recent (formal) model by Nadel and Moscovitch (1997), who hypothesize that the role of the hippocampus is not time-limited (contrary to what we believe). We expect very interesting results and debates in this line of research in the next two years.

Program 2. Brain repair and the recovery of cognitive abilities

We also obtained some interesting and solid results on self-repairing neural networks. We investigated a repair mechanism and demonstrated that the effects of gradual destruction of connections can be fully undone if repair cycles follow each lesion. These restore the redundancy in the neural networks, in which the memories are stored. We showed that neural repair may be driven by specific input, for example, from undamaged brain areas, or by random processes. If lesions were too large, memory representations were found to be beyond repair. We analyzed the conditions for repair using the theory of random graphs and concluded that for representations with many neurons, reparability is nearly a step function of the average number of intact connections per neuron.

We will continue to explore self-repair in other types of neural networks, such as Kohonen networks and backpropagation. We will also look at more engineering type applications, for example, how to ‘empty’ a Hopfield networks by removing one by one all stored representations. We currently have indications that this may be possible and that even in some cases all memories may be removed, leaving only any noise that was added lateron.

Project 2.1. Self-repair in the somato-sensory cortex

Simulations on an implementation of a model of somato-sensory cortex have started, the basic architecture of the network having been completed. Some work is currently being undertaken to speed up the simulations, as the complexity of the network causes very long calculation times and slows down progress. We plan to work on this model for at least a year, investigating how it reacts to simulated limb lesions and whether an additional repair mechanisms leads to predictions of qualitative changes in receptive fields that may possibly be detected in experimental animals. We are planning with Prof. Robertson’s group at Trinity College Dublin to carry out animal studies to verify any such predictions.

Project 2.2. Recovery from spatial hemi-neglect

This project has just started as the Ph.D. student (Jehee) on this started work on August 1^st 2000. A literature search has been completed. We currently have a clear idea of the target architecture. The plan is to complete implementation of the architecture in the Fall of 2000 and than continue to examine its behavior through 2001. We will specifically look at possible time-courses of recovery and ways to influence them (speed them up).

Project 2.3. Postulate on a common mechanism for brain repair and memory consolidation

This project has not yet started. It is not planned to start until 2002.

Program 3. Rehabilitation from brain damage

Since this program is mainly carried out by a Ph.D. student who just started in August 2000 (Jehee), not much work has been done on this project. It should be remarked that this late start is according to the original (final) proposal, because it builds on work done in Program 2.

Project 3.1. The role of arousal in rehabilitation

The main goal of this project is to explore the role of arousal and executive function in recovery from brain damage

Arousal has been a pivotal concept in our models since 1988. The CALM model, for example, has an Arousal node (Murre, Phaf, and Wolters, 1989, 1992). The TraceLink model of amnesia by Prof. Dr. Murre has a modulatory system, which has the same role as the arousal system in the CALM module: control of brain plasticity and of search processes (i.e., to find a fitting representation for an input pattern). Since 1988, several other researchers have published models and data that corroborate ours. For example, Singer (1991) stresses the importance of control of brain plasticity by central [arousal] states in ontogenesis of the brain. Recently, Hasselmo (e.g., 1995) has done extensive modeling of modulatory systems, which he locates in the basal forebrain, and which have as a main role the control of plasticity and regulation of recall versus encoding processes. These modulatory functions of arousal are mediated by acetylcholine and other neuromodulators. Arousal-induced modulation is associated with general alertness and wakefulness. In addition, there is reason to believe that the actual increase in plasticity under the influence of modulatory systems only takes place when the brain is in a state of focused attention (e.g., Singer, 1991). One of the control centers of arousal and attention is located in the right hemisphere. Consequently Robertson and Murre (submitted) predict that recovery will be impaired in cases of right hemisphere damage (e.g., in the parietal cortex) as compared with similar damage in the left hemisphere, which leaves much of the arousal system intact. Preliminary experiments have confirmed this prediction (Robertson, personal communication).

Project 3.2. Equilibrium of activation

The brain is a highly structured and extremely interactive system. Many areas in the brain, especially contra-lateral areas (i.e., on opposite sides of the brain), actively inhibit each other. When one of these systems is damaged, the other system may effectively suppress it, even if the damage is moderate. In such a case, we observe the paradoxical effect that additional damage to the brain (namely to the non-lesioned area contra-lateral to the damaged one) will cause an improvement in behavior because the equilibrium is restored (see Kapur, 1996, for an extensive review). We propose to explore this approach by modeling a number of these effects in more detail.

Dr. Robertson from the MRC Applied Psychology Unit has pioneered several rehabilitation methods that address the hypothesized disturbance of the inhibition-activation equilibrium in patients that suffer from spatial hemi-neglect. In particular, he observed an immediate improvement in the ability to attend to the neglected ‘side of the world’ when patients moved their arm about on the neglected side. Robertson and Murre (submitted) modeled this effect and explained it as caused by a stimulation of an actively inhibited area that had incurred moderate damage. The effect disappears when both arms are moved, leading further credence to the inhibition-stimulation hypothesis. More detailed modeling and concomitant data gathering will help refine this rehabilitation strategy.

Our first goal in this project is to investigate in some detail the effect of various stimulation and cueing strategies in the model of hemi-neglect developed in Project 2.2. We will focus on the effects of the various stimulation techniques mentioned (e.g., moving of the left arm) on the speed of recovery from neglect. Our aim here is to reproduce the behavioral effects observed in the experimental rehabilitation studies by Dr. Robertson and co-workers.

As a second goal, we also propose to model the role of diminished inhibition in collicular recovery after contra-lateral lesioning. The lesions concern the superior colliculus, which is an area of the brain involved in certain low-level aspects of vision. It is implicated in mediating visual attention and control of saccadic eye movements. Studying the recovery of the lesioning, as we propose to do here, is an elaboration of the so called Sprague effect (Sprague,1966). He showed that certain visual impairments in cats could be ameliorated by destroying the superior colliculus on the side opposite to initial visual input, thereby freeing the lesioned hemisphere from the collicular inhibition of the intact hemisphere and allowing circuits on the same side as the lesion to operate. In other words, the original lesion caused a cognitive impairment (hemianopia) which was subsequently diminished by an additional lesion to the brain. The important aspect that is illustrated by the Sprague effect is that brain lesions do not just cause a cognitive impairment because they directly weaken the cortical basis of certain cognitive process, but they may do just as much harm (or more) by upsetting the balance of activation and inhibition in the brain. A relatively small lesion to an area can have very strong effects because the small lesion may undermine the ability of that area to resist inhibition from certain other areas (e.g., from similar ones, located on the opposite area of the brain). As a result it will be inhibited to the point where it becomes silent. This in turn may cause a deterioration of the neurons and their interconnections over time. The upset equilibrium may, thus, not just impair recovery but it may in fact cause a continuing deterioration of the disorder. It is therefore, imperative, for any model of rehabilitation that issue is addressed in some detail (also see Kapur, 1996). To be sure, we are not suggesting that rehabilitation theory must include additional lesioning of the brain. This seems far too drastic a technique to explore as a rehabilitation strategy in humans, but it can nevertheless give us very valuable insights into the workings of these process, including into its neuroanatomical basis. Moreover, in some cases areas of the brain can be inhibited (rather than lesioned) by certain behavioral measures. An example would be taping of the healthy arm to allow recovery of the improperly functioning arm due to brain damage. This prevents the good arm from being used to the exclusion of the bad arm. Other examples can be adduced here as well.

Project 3.3. Spacing of stimuli in guided recovery

From our preliminary modeling studies (Murre and Robertson, 1995; Robertson and Murre, submitted), it has become clear that in our models the effect of external cues varies greatly, depending on the time at which they are administered. When cues are presented too early or too late they can even be maladaptive. Moreover, the frequency and the completeness of the cues necessary to guide recovery of the representations depends on the stage of recovery. The goal of this project is to explore systematically the effects of the spacing of cues on rehabilitation.

We will use the hemi-neglect model developed in Project 2.2 to first study what the effects of different forms of spacing input stimuli are on the speed and quality of recovery. The input stimuli are selected such that they will indeed help guide recovery. The modeling results will then be compared systematically with behavioral data. If we are able to derive any guidelines for spacing of rehabilitation, this will be tested systematically with patients by Dr. Robertson’s group. This project will benefit from partial integration with Project 1.7.a because the Memtool program and underlying mathematical model addresses spacing effects in normal learning, memory, and forgetting in great detail. It is likely that these results apply to rehabilitation as well, but this has not been explored systematically. We will be in a position to test this hypothesis.

The main focus of this project is on speed and quality of recovery as a function of the presentation schedule of patterned input. Behavioral data will mainly focus on spatial hemi-neglect, but other suitable disorders will be considered if data of sufficient quality is made available to us.

We anticipate to arrive at a set of validated guidelines for optimal presentation schedules of patterned input in rehabilitation therapy. The results will mainly apply to recovery from hemi-neglect, but we will attempt to generalize the results based on the computational model to other disorders as well.

Program 4. Brain connectivity

Project 4.1. Brain connectivity, volume, and cognitive functions

We have carried out another review of the quantitative neuroanatomical literature and extended and refined the estimates and models by Murre and Sturdy (1995), studying in particular the effects on brain volume of internal cavities in the brain, effects of neuron soma, of tapering connection structures (dendrites, axons), and the effects of glia. In other words, we have removed many of the simplifying assumptions of Murre and Sturdy (1995), making the model more biologically plausible (and more detailed). We have also looked at other connectivity schemes, notably Gaussian connectivity and combinations of random and k-nearest neighbor connectivity. In addition, we have reviewed all literature that we could find on ‘the remarkable constancy of number of neurons under 1 mm² of pial cortical surface across animals and brain areas’. We conclude that there is good evidence for this constancy in the literature (with some caveats) and we have developed a new hypothesis to explain it (the so called ‘window hypothesis’).

This work has been completed, although several alleys remain open for further exploration. In addition, we have made a start with a detailed neuroanatomical analysis of connectivity in the hippocampus with the aim of supporting a detailed connectionist model of this structure.

Program 5. Walnut neurosimulator

Project 5.1. The Walnut project

The first phase of this project is nearly completed. The Nutshell neurosimulator has been tested thoroughly, both by us (with serious simulation work) and by students in connectionist laboratory courses. Several usability studies have already been carried out. The software engineer (Eric Maryniak) and assistant programmer (Robert Berg) have added documentation (not yet complete) and an extensive infrastructure for OpenSource distribution (web pages, version control system CVS, mailing lists, installation software, bug and issue tracking, etc.). Members of the group have written new Walnut neural network paradigms, though more need to be added. The system is nearly ready for public release, which is planned for November 2000 from our web site (www.neuromod.org).

Nutshell currently runs on Windows only. Portability has been studied extensively and we have decided to redevelop parts of Nutshell under Python, a very flexible and portable scripting language. The Python Nutshell (pyNutshell) will work on top of existing (and newly added) neural network paradigm components, written in C++. We are still considering whether we want to propose and register an official XML standard for our Neural Network Markup Language (NNML).

Added and related projects

Since writing the original proposal, the following projects have been added to the PIONIER grant. Some of them are separate projects that are strongly tied to the projects in the PIONIER program.

A.1 Point processes model of learning, forgetting, and amnesia

This project is the direct result of a project ‘Computational modeling of human memory’ in the NWO Aandachtsgebied entitled ‘Learning and automation of a second language’ (by Prof. A. de Groot). A postdoc on this project was awarded to Prof. Dr. Murre, which aims at the development of a computational model of learning and forgetting of a second language. In addition to a theoretical component, it has the practical aim of developing a computer program that allows students to learn more in the same amount of study time and to reduce forgetting afterwards. Modeling in this project is purely mathematical and in the original project proposal we did not even hint at the possibility of integrating this research directly with any of the projects of the PIONEER grant.

To our (pleasant) surprise, we were able to develop a model that both abstracts from the TraceLink model and extends it (also see Project 1.7). The resulting mathematical model is psychologically and neuroanatomically transparent, meaning that we can identify all parameters with psychological and neurobiological processes. Being firmly embedded in the theory of point processes (so far we have limited ourselves to Poisson processes), it is also a rigorous framework in which all of the important phenomena (forgetting, learning, recognition, savings, effects of repetition, rehearsal, and certain lesions, etc.) can be derived analytically. We have tested the model against over 280 different forgetting, learning, and amnesia experiments in the literature since 1885, achieving superior fits. It is a mathematical model that integrates learning, (simultaneous) forgetting, and neuroanatomy in a single model. It can even predict learning and forgetting of advertising without any modification whatsoever.

A.2 Schizophrenia model

In this project we cooperate with Terry Goldberg and Brita Elvevaag from the NIMH in Washington D.C. We aim to base a schizophrenia model on the TraceLink model. The reason for this is that memory impairment is one of the most consistent aspects of the cognitive disorders associated with schizophrenia. We have now completed the model architecture and have started to run simulations. The NIMH group has the capacity to carry out experiments with schizophrenic patients as suggested by our model. Two such suggestions are in the process of being checked.

A.3 The role of sleep in consolidation

A crucial question in the so called standard model of amnesia is whether transfer of memories from hippocampus to neocortex takes place during sleep. For nearly fifty years, the evidence for this has been accumulating, but the crucial test has never been carried out. This test would entail monitoring memory performance for up to three weeks in animals in which the hippocampus is deactivated during sleep only. Since the sleep-consolidation hypothesis is at the heart of the PIONIER grant, we decided to try to realize this experiment. The University of Amsterdam has awarded matching funds for this experiment. Informal cooperation has been established with several labs in Amsterdam, including that of Prof. Wadman, Prof. Witter, and a group at the NIH, including Prof. Uylings and Prof. de Bruin. We plan to carry out the actual experiments at the NIH.

We have been preparing this experiment for about one and half years now and it has become clear that it is quite hard to do. After an exhaustive review of all possible methods for repeatedly and reliably deactivating the hippocampus in freely behaving rats, we have settled on developing an entirely new instrument, which we call the cryotrode (see next project). After design and testing of this instrument has been completed, we will carry out the experiment in the three phases of increasing technical difficulty. These experiments are planned for 2001 and 2002.

A.4 Cryocybernetics

Development of the experimental technique necessary for project A.3 was so complex that we have designated a separate project for this. We needed a method for repeatedly, rapidly, and reversibly deactivating deep brain structures (e.g., the hippocampus) in freely behaving rats. No current method comes even close. After an exhaustive review of the literature we decided to develop an instrument based on cooling. Neural tissue cooled below 15-20 °C ceases to function. After cooling has been terminated, the neural tissue starts functioning again without any detectable permanent modifications. As far as known, this process can be repeated indefinitely. In 1999, we contacted cooling engineers at Fokker Space (Leiden) and commissioned a report on the best cooling technique for this purpose.

This research has been completed (by Ir. Coesel and Ir. Huizinga from Fokker Space) and we are now in the process of developing a so called cryotrode: an instrument with four or six pins that is inserted into the brain of a rat. Only the tips of the pins reach a low temperature. The cryotrode can remain inserted permanently. We have written a grant proposal to STW for additional funds, so that we can develop the cryotrode into a commercially viable instrument. When we reach this stage, we (with Fokker Space) will create a spin-off to market and distribute the instrument.

We plan to develop the technique with the cryotrode to a more advanced level by hooking it up to a permanent EEG (ECG/EMG) recorder with ‘brain state detection software’. (Other brain states indicators are possible as well, e.g., through measuring certain hormone levels.) Depending on the detected brain state (in our case ‘sleep’, or ‘REM sleep’ versus ‘slow-wave sleep’), the cryotrode will be activated or deactivated. In this way, we are developing a new experimental technique, which we call cryocybernetics in which brain structures are activated or deactivated depending on the detected brain state.

A.5 Neuromod.org site

To support the various internet activities of our group (also see Project A.7 below), an internet site was created that has been made public under the neuromod.org domain. This site serves to support the OpenSource distribution of the Walnut/Nutshell neurosimulator and in general it supports distribution and publicity of our modeling efforts. The site is run with Linux, Apache, Zope, mySQL, WinCVS, Bugzilla, Python, Perl and other OpenSource software and is managed by Eric Maryniak (software engineer, PIONIER grant).

A.6 Memory.uva.nl site

This site was created (by Prof. Dr. Murre in English and translated into Dutch by Robert Berg) to support a ‘Day for the Public’ that had as a theme ‘Memory’ (7 mei 1999). It currently contains an ‘Online memory improvement course’ in both English and Dutch. The site explains mnemonic techniques and also contains online training software written in JavaScript. Several memory tests will be added (see Project A.8). We plan to add more general information for the public on memory in the future, but this has a low priority.

A.7 Geppetto project

This project is funded by the University of Amsterdam ‘Matching Funds’. It aims to create an internet site where computational models in the cognitive neurosciences can be accessed. The proposal is complementary to project 5.1, where the development of an international standard for neural network models in cognitive neuroscience is proposed. The added value of the Geppetto internet site stems from a review process of the models that are submitted for inclusion. This is a technical review aiming to assess how reliable the model software is, how hard it is to get running, etc. We aim to extend the concept of a refereed journal publication to include the publication of models. Once a model has been accepted for publication on the internet site, it will be possible to access it by all researchers interested. Journal articles can refer to the downloadable models by their worldwide web address (URL).

The first phase of this project is nearly completed, namely a nearly exhaustive database of available neurosimulation software. This has been entered into a database. The first technical review has started (of the Genesis neurosimulator).

A.8 Memory online or an internet test panel

Current tests for retrograde amnesia have as a common problem that easier questions are used for the older time periods (e.g., on very well known actors, compared to less well known actors for the current period), so that the scores on these remote periods do not become too low for healthy control subjects. In general, not many very long term memory tests have been carried out and many of these have methodological limitations (e.g., n=1 design, and possibly unequal exposure for various time periods,). We decided to utilize the new possibilities of the internet by devising an internet based battery of tests (online memory tests), while at the same time trying to recruit subjects willing to take these tests via the internet (internet test panel).

We currently have developed the database and web pages to administer the tests and are about to pilot the current design. The database currently contains questions based on public events taken from newspaper headlines (de Volkskrant). We plan to update the database daily for a prolonged period of time, while regularly inviting subjects (both paid and unpaid) to take the tests. This will result in forgetting curves for each question (news fact). Questions can be pooled, so that we can obtain detailed and reliable forgetting data.

The forgetting curves can be compared with those in various patient populations with amnesic disorders, potentially resulting in a comparison of true forgetting with retrograde amnesia curves. We also hope to gain experience with recruiting and managing subjects via the internet in this way. The database design is quite general and can also accommodate other types of memory tests and most types of query based tests (multiple-choice, free form answers, etc.). We plan to add more tests in 2001.

A.9 Public events test

As an instrument for testing patients with retrograde amnesia, Martijn Meeter has developed a public events test with questions that go back several decades. This test (in Dutch) will first of all be used as a research instrument. Perhaps, we will make it available online (see Project A.8) to selected persons (e.g., neurologists).

A.10 AGI: Dutch translation of Autobiographical Memory Interview

There is currently no commercially available test for retrograde amnesia in Dutch. This is why we (Prof. Dr. Murre and Drs. Meeter) decided to officially translate the English AMI (Autobiographical Memory Interview) by Kopelman, Wilson, and Baddeley. Swets Publishers has agreed to publish the test. We have completed the translation and are about to collect reference data with patients and normal subjects.

Other scientific activities

Organization of workshops

The PIONIER group has been involved with organizing two international workshops: ‘New Directions in Memory and Amnesia Research’ (7 May 1999) and ‘Memory and Emotion: Consolidation and Distortion’ (10-11 May 1999). In addition on 8 May 1999, Prof. Dr. Murre helped organize a ‘Day for the public’ on Memory.

Working visits

Working visits were paid to research groups in Newcastle, Rome, Edinburgh, Washington D.C. (twice), New York (NYU), INRIA Nancy (supported by a Van Gogh grant to Prof. Dr. Murre), Trinity College Dublin, and Cambridge (UK).

Visiting researchers

We received a group from Nancy (leader is Dr. Alexandre; our group and his hold a Van Gogh grant), Dr. Elvevaag (NIMH, Washington, D.C.), and Dr. Graham (MRC-CBU, Cambridge).

Publications

The following publications have been produced under the PIONIER grant since 1999. Publications in preparation have not been listed below.

Scientific publications

Submitted scientific publications

Professional publications

General publications