Frederic, thanks for the response. Let’s use this as an opportunity to compare theories. I will try to address aspects briefly covered by my theory and perhaps you could cover the same aspects from your perspective. The original 2006 article I wrote is available on my contribution page and it has details the 2012 article did not repeat. I will not cite the articles supporting my statements since these are included in those two published articles. If the cortical column is the level at which cortical processing and memory occur, let’s discuss how this fits with other levels of investigation. With the thousands of neuroscience-related articles a year being published in multiple journals, it is possible to find discussions from molecules to modules as keys to memory and processing.

When all is said and done, and there is conclusive evidence as to how our brains code and process information related to higher cortical functions, it appears likely that there will be only one level involved. This is based on the assumption that in evolution, the most efficient and effective solutions are those that remain. Thus I believe simpler is going to win out over complex as a coding strategy in the cortex since it is less costly in terms of energy expenditure and simpler also decreases the likelihood of dysfunction and damage susceptibility. If true, then all of the complex computational theories of cortical functioning will prove to be a function of human ingenuity as opposed to what I think is a basic binary coding system. I will briefly discuss how a columnar bit would fit with some of the different levels of assessment. This includes single cell recording, receptors, imaging, EEG, and modules/locations. The attached brief article discusses hippocampal and thalamic influences on cortical memory storage, which was also discussed in more detail in the 2012 article.

As I have noted in the articles, not all neurons within the area of the column need to be involved. Cell firing rate in the gamma frequency range allows dynamic column formation. Only the outermost neurons of the column are theorized to be synchronized by gamma frequency input which allows a large (0.4 to 1.0 mm) area of representation which would be highly resistant to damage. However, overlapping columns means that within that 0.4 to 1.0 mm area, the borders of a large number of columns exist. If single cell recordings are done at the cortical level and it is found that there is activation with two distinct stimuli, then it depends on where in the column the neuron is located and the complexity of the column’s representation as to what that activity means. For example, the stimuli could lead to two different overlapping columns activating which share the neuron being recorded. An alternative is that the two stimuli activate columns that feed into a single higher-order column where the recorded cell is located. Even if one stimulus is presented numerous times and the recorded neuron activates, the pattern of the neuron firing rate would vary depending if it is on the boundary of the column or on the interior. If on the boundary, it is part of the information bit of that column and should maintain the same gamma frequency rate each time it is activated. If on the interior, the neuron will be asynchronous in relation to the boundary neurons and the firing rate would slow with repeated stimulus presentations.

It is known that glutamatergic NMDA receptors are involved with gamma frequency activity. This means they would be associated with EEG waveforms in the gamma range. These receptors are voltage dependent which would be affected by higher rates of cell firing (i.e., increased neurotransmitter release into synapses from upstream neurons leading to hypopolarization via influx of calcium ions of the downstream neurons). GABA is implicated as the neurotransmitter involved in limiting the column-sized spread of activity at the gamma frequency range. As I noted in my 2006 and 2013 article it appears that increased metabolic activity as measured by fMRI occurs with “new learning” but decreases over time with improved performance indicative of learning. I suggested the initial generalized activity allows general activation (noise) around a column, with there being a strong inhibitory surround defining the column boundary (signal). Once a columnar circuit becomes consolidated (i.e., upstream columns reliably activate downstream columns due to strengthened synaptic connections among columns), there is no longer a need to have general activation and fMRI no longer shows any signs of increase metabolic activity.

In relation to “modules,” these do not exist. The columnar theory indicates such areas as the fusiform face area, parahippocampal place area, motion detection areas, object recognition, etc., are simply the locations of the columns representing the highest level of integration in that specific information stream which originates in the primary sensory receiving area. Association memories involve the highest level columns in more than one information stream projecting to a common column in the medial temporal lobe cortex. That temporal cortical column represents information from the highly organized information of the columns projecting to it. This common column in turn projects to the hippocampus and involves the hippocapmpo-thalamo-cortical circuit described in the attached paper and the 2012 article.

Article Gamma-band Synchronization & Cortical Columns

There is more to this apparently simple cortical unit story.

- For one, not all areas of the brain contains similar cortical column, hence their definition is not that simple. Early sensory areas have much better defined columns that do latter areas.

- Then, it is not a binary unit of storage. Neurons within a column encode information in a continuous fashion. Information theory does provide the tools to estimate the amount of information (in bits) carried by a given column. For instance, it can be computed from the signal-to-noise ratio.

- The amount of information carried by a column is not a fixed value. Some carry more, other carry less.

- We also need to think in term of population code rather than isolated single column.

Here's the link to a series of two videos that explains several concept linking information theory and neural code. I hope it will help http://www.youtube.com/watch?v=MUpyx5_GXtg

Frederic, thanks for the response. Let’s use this as an opportunity to compare theories. I will try to address aspects briefly covered by my theory and perhaps you could cover the same aspects from your perspective. The original 2006 article I wrote is available on my contribution page and it has details the 2012 article did not repeat. I will not cite the articles supporting my statements since these are included in those two published articles. If the cortical column is the level at which cortical processing and memory occur, let’s discuss how this fits with other levels of investigation. With the thousands of neuroscience-related articles a year being published in multiple journals, it is possible to find discussions from molecules to modules as keys to memory and processing.

When all is said and done, and there is conclusive evidence as to how our brains code and process information related to higher cortical functions, it appears likely that there will be only one level involved. This is based on the assumption that in evolution, the most efficient and effective solutions are those that remain. Thus I believe simpler is going to win out over complex as a coding strategy in the cortex since it is less costly in terms of energy expenditure and simpler also decreases the likelihood of dysfunction and damage susceptibility. If true, then all of the complex computational theories of cortical functioning will prove to be a function of human ingenuity as opposed to what I think is a basic binary coding system. I will briefly discuss how a columnar bit would fit with some of the different levels of assessment. This includes single cell recording, receptors, imaging, EEG, and modules/locations. The attached brief article discusses hippocampal and thalamic influences on cortical memory storage, which was also discussed in more detail in the 2012 article.

As I have noted in the articles, not all neurons within the area of the column need to be involved. Cell firing rate in the gamma frequency range allows dynamic column formation. Only the outermost neurons of the column are theorized to be synchronized by gamma frequency input which allows a large (0.4 to 1.0 mm) area of representation which would be highly resistant to damage. However, overlapping columns means that within that 0.4 to 1.0 mm area, the borders of a large number of columns exist. If single cell recordings are done at the cortical level and it is found that there is activation with two distinct stimuli, then it depends on where in the column the neuron is located and the complexity of the column’s representation as to what that activity means. For example, the stimuli could lead to two different overlapping columns activating which share the neuron being recorded. An alternative is that the two stimuli activate columns that feed into a single higher-order column where the recorded cell is located. Even if one stimulus is presented numerous times and the recorded neuron activates, the pattern of the neuron firing rate would vary depending if it is on the boundary of the column or on the interior. If on the boundary, it is part of the information bit of that column and should maintain the same gamma frequency rate each time it is activated. If on the interior, the neuron will be asynchronous in relation to the boundary neurons and the firing rate would slow with repeated stimulus presentations.

It is known that glutamatergic NMDA receptors are involved with gamma frequency activity. This means they would be associated with EEG waveforms in the gamma range. These receptors are voltage dependent which would be affected by higher rates of cell firing (i.e., increased neurotransmitter release into synapses from upstream neurons leading to hypopolarization via influx of calcium ions of the downstream neurons). GABA is implicated as the neurotransmitter involved in limiting the column-sized spread of activity at the gamma frequency range. As I noted in my 2006 and 2013 article it appears that increased metabolic activity as measured by fMRI occurs with “new learning” but decreases over time with improved performance indicative of learning. I suggested the initial generalized activity allows general activation (noise) around a column, with there being a strong inhibitory surround defining the column boundary (signal). Once a columnar circuit becomes consolidated (i.e., upstream columns reliably activate downstream columns due to strengthened synaptic connections among columns), there is no longer a need to have general activation and fMRI no longer shows any signs of increase metabolic activity.

In relation to “modules,” these do not exist. The columnar theory indicates such areas as the fusiform face area, parahippocampal place area, motion detection areas, object recognition, etc., are simply the locations of the columns representing the highest level of integration in that specific information stream which originates in the primary sensory receiving area. Association memories involve the highest level columns in more than one information stream projecting to a common column in the medial temporal lobe cortex. That temporal cortical column represents information from the highly organized information of the columns projecting to it. This common column in turn projects to the hippocampus and involves the hippocapmpo-thalamo-cortical circuit described in the attached paper and the 2012 article.

Article Gamma-band Synchronization & Cortical Columns