Final Report Summary - FAST MAPPING (Fast Mapping: How to acquire new declarative memories independently from the Hippocampus?)
Current theories of declarative memory suggest that it is supported by two complementary memory systems: the medial temporal lobe (MTL) which specializes in rapid acquisition of novel associations and the neocortex, which slowly learns through environmental regularities. Gilboa et al. recently reported an extraordinary phenomenon of rapid acquisition of novel arbitrary associations by adults with severe anterograde amnesia due to extensive MTL lesions. These data challenge the accepted view of dual declarative memory systems and offer a radically different mechanism for directly acquiring semantics without enlisting the hippocampus: the Fast Mapping. Fast Mapping refers to a process that supports vocabulary acquisition among children who do not have a mature hippocampus. They actively infer the meaning of novel words by generating hypotheses based on contextual cues or on the way the words are used in a sentence. By using a multidisciplinary approach, this project aims at highlighting the neural substrates and the cognitive mechanisms underlying this type of learning.
In the first experiment, we explored the neural basis of FM by a clinical approach. We designed a Fast Mapping task and tested patients suffering from Alzheimer’s disease (AD), as well as Healthy older adults (OA). Results showed that AD patients were impaired compared to OA on Explicit Encoding (EE) but performed equivalently on Fast Mapping (FM) both after 10 minutes and after one week. Furthermore, there was no significant decrease in performance after a week delay showing that patients and controls were able to retain what they learnt through FM.
All participants also benefited from a global neuropsychological assessment. Regression analysis showed that EE scores were correlated with clinical tests of Episodic memory and FM scores were correlated with semantic memory tasks.
Half of the participants were also enrolled in an imaging protocol including different types of acquisition (FLAIR, MPRAGE, DTI, Resting State...). A multi disciplinary approach including Hippocampal segmentation, Voxel Based Morphometry, Diffusion tensor Imaging, MultiVoxel Pattern Analysis and Cortical Thickness measurement was conducted. Multiple brain-behavior correlations were performed. Results showed that EE scores were correlated with Hippocampal volumes and with clinical tests of episodic memory whereas FM scores were correlated with neocortical regions such as ATL and specific Frontal and Parietal regions and with semantic memory tasks. The Multi Voxel Pattern Analysis highlighted areas of the GM in the frontal lobe, the parietal lobe and the lateral temporal neocortex that effectively discriminated good from poor fast-mappers.
In line with Sharon et al.’s study, these converging results suggest that FM is less sensitive to hippocampal atrophy and more sensitive to anterior and posterior neocortical degeneration that is also part of AD. The ATL might be involved in the Fast Mapping learning probably due to its role in representing semantic associative networks.
In a second experiment we investigated the neurocognitive mechanisms underlying FM. We designed a new paradigm in order to demonstrate how previous knowledge and inference are used in combination with repeated exposures to guide people in learning new words. Twenty four young adults were presented with 4 animals at a time along with a sentence describing a fact about one of the animal. They were asked to select the animal corresponding to the description. Each animal was presented 12 times in different context. Participants were tested 24 hours later on the animals’ name and facts. A Bayesian analysis showed that a significant improvement in accuracy occurred as a step-function rather than as a cumulative gradual increase due to statistical regularity. As trials proceeded, item-related experience influenced behavior more and the contribution of initial general-knowledge was diluted.
In order to determine the locus of Fast Mapping and to highlight the patterns’ changes of neocortical activity when participant learn through FM, we adapted the previous study to the MEG constraints and tested twenty young participants. We also added a categorization task where participants were asked to make a simple mammal/bird decision about the 20 targets and 20 lures. This categorization task was presented at four different time points. All items (targets and lures) were viewed an equal number of times prior to categorization, but only targets were associated with a name and facts, so the difference reflects semantic knowledge acquired about the items prior to categorization. MEG data were analyzed with different approaches. The results showed that, by the fourth categorization block, there is a clear left lateralized positivity over left anterior areas. The only difference between blocks is the amount of semantic knowledge acquired, so, in concordance with our hypotheses and previous lesion studies, anterior neocortical regions seem to be involved in the FM process.
The final purpose of this project was to create a rehabilitation tool for patients suffering from memory disorders who cannot learn new information. The results obtained in the first study were encouraging, since AD patients were able to learn new associations through FM, however, their performance was not optimal. In order to design the most efficient rehabilitation tool possible, we decided to adjust some parameters of the FM task in order to improve the memory performance. The main changes included the use of real objects instead of pictures and the addition of visual or verbal familiarization task prior to the FM task. Twenty two participants were tested with this new version of the paradigm. Results showed that the memory performance was better when participants were exposed to the items prior to the FM task itself. Moreover the familiarization task effect revealed that items were best recalled when they were previously visually displayed. During the visual familiarization task, all physical aspects of the item were activated, creating a new pattern of activation in the existing semantic network. During the FM task, participants only had to add the label of the item to this pattern of activation to integrate the new item in the pre-existing semantic knowledge. These results suggest that FM could result from the incorporation of novel information into existing knowledge structures. Embedding novel learning within an existing schema accounts for a shortened consolidation.
Another study arose from the first experiment. FM seems to be efficient to learn new association in patients with memory disorders. One of the main complaints of those patients is to remember names of people. Thus, we designed a new FM task, no longer associating animals or fruits with names but, in a more pragmatic way, associating faces with names. Results are in line with the previous ones and show that FM can be an alternative way of learning new associations for people facing episodic memory disorders.
In summary, the results obtained in the different tasks show that patients with episodic memory disorders and MTL damage can learn new associations through FM. This finding challenges the accepted view of episodic memory and open a brand new field of research.
These results not only have important implications for neurobiological theories of memory, but also have significant prospective clinical application. A better understanding of the neurocognitive mechanisms of FM would allow clinicians and cognitive scientists to harness it towards behavioral rehabilitation attempts of persons with memory disorders. This would be a welcomed addition to the sparse arsenal of behavioral interventions currently available. Creating a rehabilitation tool based on Fast Mapping might rekindle hope to thousands of patients.
In the first experiment, we explored the neural basis of FM by a clinical approach. We designed a Fast Mapping task and tested patients suffering from Alzheimer’s disease (AD), as well as Healthy older adults (OA). Results showed that AD patients were impaired compared to OA on Explicit Encoding (EE) but performed equivalently on Fast Mapping (FM) both after 10 minutes and after one week. Furthermore, there was no significant decrease in performance after a week delay showing that patients and controls were able to retain what they learnt through FM.
All participants also benefited from a global neuropsychological assessment. Regression analysis showed that EE scores were correlated with clinical tests of Episodic memory and FM scores were correlated with semantic memory tasks.
Half of the participants were also enrolled in an imaging protocol including different types of acquisition (FLAIR, MPRAGE, DTI, Resting State...). A multi disciplinary approach including Hippocampal segmentation, Voxel Based Morphometry, Diffusion tensor Imaging, MultiVoxel Pattern Analysis and Cortical Thickness measurement was conducted. Multiple brain-behavior correlations were performed. Results showed that EE scores were correlated with Hippocampal volumes and with clinical tests of episodic memory whereas FM scores were correlated with neocortical regions such as ATL and specific Frontal and Parietal regions and with semantic memory tasks. The Multi Voxel Pattern Analysis highlighted areas of the GM in the frontal lobe, the parietal lobe and the lateral temporal neocortex that effectively discriminated good from poor fast-mappers.
In line with Sharon et al.’s study, these converging results suggest that FM is less sensitive to hippocampal atrophy and more sensitive to anterior and posterior neocortical degeneration that is also part of AD. The ATL might be involved in the Fast Mapping learning probably due to its role in representing semantic associative networks.
In a second experiment we investigated the neurocognitive mechanisms underlying FM. We designed a new paradigm in order to demonstrate how previous knowledge and inference are used in combination with repeated exposures to guide people in learning new words. Twenty four young adults were presented with 4 animals at a time along with a sentence describing a fact about one of the animal. They were asked to select the animal corresponding to the description. Each animal was presented 12 times in different context. Participants were tested 24 hours later on the animals’ name and facts. A Bayesian analysis showed that a significant improvement in accuracy occurred as a step-function rather than as a cumulative gradual increase due to statistical regularity. As trials proceeded, item-related experience influenced behavior more and the contribution of initial general-knowledge was diluted.
In order to determine the locus of Fast Mapping and to highlight the patterns’ changes of neocortical activity when participant learn through FM, we adapted the previous study to the MEG constraints and tested twenty young participants. We also added a categorization task where participants were asked to make a simple mammal/bird decision about the 20 targets and 20 lures. This categorization task was presented at four different time points. All items (targets and lures) were viewed an equal number of times prior to categorization, but only targets were associated with a name and facts, so the difference reflects semantic knowledge acquired about the items prior to categorization. MEG data were analyzed with different approaches. The results showed that, by the fourth categorization block, there is a clear left lateralized positivity over left anterior areas. The only difference between blocks is the amount of semantic knowledge acquired, so, in concordance with our hypotheses and previous lesion studies, anterior neocortical regions seem to be involved in the FM process.
The final purpose of this project was to create a rehabilitation tool for patients suffering from memory disorders who cannot learn new information. The results obtained in the first study were encouraging, since AD patients were able to learn new associations through FM, however, their performance was not optimal. In order to design the most efficient rehabilitation tool possible, we decided to adjust some parameters of the FM task in order to improve the memory performance. The main changes included the use of real objects instead of pictures and the addition of visual or verbal familiarization task prior to the FM task. Twenty two participants were tested with this new version of the paradigm. Results showed that the memory performance was better when participants were exposed to the items prior to the FM task itself. Moreover the familiarization task effect revealed that items were best recalled when they were previously visually displayed. During the visual familiarization task, all physical aspects of the item were activated, creating a new pattern of activation in the existing semantic network. During the FM task, participants only had to add the label of the item to this pattern of activation to integrate the new item in the pre-existing semantic knowledge. These results suggest that FM could result from the incorporation of novel information into existing knowledge structures. Embedding novel learning within an existing schema accounts for a shortened consolidation.
Another study arose from the first experiment. FM seems to be efficient to learn new association in patients with memory disorders. One of the main complaints of those patients is to remember names of people. Thus, we designed a new FM task, no longer associating animals or fruits with names but, in a more pragmatic way, associating faces with names. Results are in line with the previous ones and show that FM can be an alternative way of learning new associations for people facing episodic memory disorders.
In summary, the results obtained in the different tasks show that patients with episodic memory disorders and MTL damage can learn new associations through FM. This finding challenges the accepted view of episodic memory and open a brand new field of research.
These results not only have important implications for neurobiological theories of memory, but also have significant prospective clinical application. A better understanding of the neurocognitive mechanisms of FM would allow clinicians and cognitive scientists to harness it towards behavioral rehabilitation attempts of persons with memory disorders. This would be a welcomed addition to the sparse arsenal of behavioral interventions currently available. Creating a rehabilitation tool based on Fast Mapping might rekindle hope to thousands of patients.