Review of Neuro Linguistic

Synthetic event-Related potentials: A computational bridge between neurolinguistics models and experiments

An article was chosen from a peer-reviewed journal known as Neural Networks on the basis that it not only has various implications for the future of neurolinguistics research, but I also found it interesting and related to my intended future career profession. There have been different methods of mapping actual brain activity that occur in normal human cognition and such efforts have also led to various models of cognition and schema networks that continually become more refined as the scientific research progresses. This research began a few generations ago, and at the time the technology was fairly primitive, at least by comparison to the tools and technics that are available to today's researchers. Today, researchers are able to construct 3D representations of the brains activity that show the activity in real-time. This allows for possibilities that were never before imagined as science begins to unlock the secrets of the mind.

The researchers in the study have apparently produced methods that allow them to develop Synthetic Brain Imaging to link neural and schema network models of cognition and behavior to PET and fMRI studies of brain function in previous research efforts. As a follow up to the methodology that have already created, in this research study the researchers have extended this approach to Synthetic Event-Related Potentials (Synthetic ERP), and although the method is of general applicability, they focused on ERP correlates of language processing in the human brain (Barres, Simons, & Arbib, 2013). The different pathways that are available to individuals in their brains must somehow be related to how the process of cognition works, as well as the process of language that allows for it, and thus examining the neural networks and neural activities can help science better understand how these processes work.

This article review will briefly outline the methodology the researchers created and the results of their attempts to implement their network mappings of language processing. This article was published fairly recently and provides a good overview of different ways in which researchers have tried to reconcile different data regarding neural network maps and actual data collected from various brain scanning devices. This technology has already gone through several generations of development and has made substantial progress. New technologies such as fMRI and PET scans have offered even more dynamic data sets to researchers regarding activity in the brain. Much of the article is pretty technical and can be hard to fully understand by anyone who is not a specialist in the field. However, simply bearing witness to the types of methodologies that are being constructed to help researchers map these processes is enlightening and can offer insights as to where the future of the field of neurolinguistics is heading.

Article Review

Much of the research progress that has been made in this field has benefited from the advent of tomographic brain imaging, which allows researchers to witness activity in the brain that occurs during normal cognitive processes. Researchers have been able to identify which parts of the brain are typically involved in processing different types of brain activities by scanning the brains of participants as they perform various actions. For example, many of the higher cognitive tasks are going to be located in the prefrontal cortex while vision process is located in other parts of the brain. However, as the science progresses, researchers are attempting to build a more detailed and accurate assessment that is even more specific.

Following these initial breakthroughs, the lab expanded this simulation method to develop Synthetic Brain Imaging. Synthetic PET which was a technic that also facilitated modeling and comparison of different neural processes that were found in primates and humans, and a similar approach was used to associate a synthetic BOLD signal with a model of primate imitation. Using such mapping technics, researchers can try to determine which processes in the brain are related to different functions and begin to map these connections.

The researchers argue that the key idea that has been the basis of these research interests is to:

"start with a biologically grounded neural network for executing a task set that matches a range of neurophysiological and behavioral data. A spatial and temporal average over the simulated absolute value of all synaptic activations across a region provides a viable prediction of the activation of that region for brain imaging -- thus enabling the use of simulations of biologically grounded neural networks to yield predictions to be tested against brain imaging studies (Barres, Simons, & Arbib, 2013)."

Thus, by building maps of different neural processes, and then comparing these maps against the actual processes that are recording during scanning sessions, researchers can continually refine their models so that they represent a more accurate reflection of the various networks that exist and their functions. There have been three sources of data for neurolinguistics: lesions, fMRI (or PET), and ERPs, however there has historically been a disconnect between processing models of neurolinguistics which seek to derive linguistic phenomena from empirical data and biological constraints and representational approaches which employ a top -- down approach seeking to assign theories of language structure to biological systems (Barres, Simons, & Arbib, 2013).

Figure 1 - Realistic Brain Model (Barres, Simons, & Arbib, 2013)