Talk:Freeman's mass action
This article by Freeman and Kozma gives an excellent summary of the theory of neural mass action, originally proposed by Freeman almost 35 years ago. While longer than regular Scholarpedia articles, this entry contains the main features and concepts from the original theory. Still, the text is somewhat difficult to grasp, and requires knowledge and insight from a number of scientific fields in order to be fully appreciated. It is clearly more difficult to read than ordinary articles in Scientific American, as suggested for Scholarpedia articles. The text also ends a bit abruptly, without a clear message. Some kind of summarizing sentences, or concluding remarks could be helpful even for this type of text.
Although well founded in neurobiology, the article – as well as the theory presented – is full of concepts and ideas from physics, which might seem confusing. For example, many different types of energies, forces, pressures and actions are being used in the text, sometimes in a confusing combination. The authors are advised to reexamine the terminology to be consistent throughout the text.
I think it would be helpful to the general reader if the terminology chosen would be better anchored and explained in terms and concepts more commonly used in neuroscience. Even as a physicist, I imagine that the extensive use of physics could confuse more than clarify for a non-physicist reader, who might miss the main messages of the FMA theory. While the authors most certainly have tried hard to make the text comprehensible to the general reader, they could consider simplifying and explaining even more (without making the text longer).
A general comment concerning the terminology is the current usage of “ECoG” (electrocorticograms), which partly corresponds to a previous usage of the term “EEG” (electroencephalograms). This discrepancy should be made explicit in the text. In fact, it seems that Freeman and Kozma have divided the previously used term EEG (as in e.g. Freeman, 2005 and earlier work) into two more specific terms: “EEG” (for macro-dynamics) and ECoG (for current meso-dynamics). This shift in terminology also has bearing on the shift in usage of “mesoscopic brain dynamics”, which in a bulk of other work, including earlier work by Freeman (2000, 2005) is related to EEG recordings.
In the initial paragraph, the “abstract”, it seems more appropriate to write “neural population” rather than “brain population”.
Par. 1: What does it mean that "brains sustain spontaneous neural activity at and near Self-organized Criticality"? Maybe it would be better to say that the neural activity shows characteristics akin to self-organized critical (SOC) phenomena (or something similar). Even though SOC is an interesting theory, it is not self-evident that it applies to the brain. A short discussion on this could be appropriate.
Par. 1 and elsewhere: Force (e.g. neural force), energy and pressure is used in a way that is not quite clear. For example, it says that FMA is energy in many forms, but it is not obvious which forms of energy (except for metabolic energy).
Par. 2: Why is an open thermodynamic system contrasted with a neural network or a chemical system, which at least in some cases also can be considered open thermodynamic systems? Further, the second law of thermodynamics is formulated for closed, isolated systems, and since the brain is considered an open system, this law is not relevant here.
Par. 2: Energy can be dissipated but one usually does not talk about dissipation of entropy (which, however, can increase through dissipation of energy).
Par. 2: The sentence that starts, “The brain requires these energy sources and sinks…” looks a bit strange and should be rewritten.
Par. 3: The (traditional) law of mass action concerns states in equilibrium, whereas the brain, as all biological systems, are open systems far away from equilibrium. This difference in usage of the concept (law of mass action) should be commented.
Par. 3: Perhaps there could be some references to work by Baars and Changeux on the global/neural work space, which seems to relate to “the collective neural action that executes intentional behaviors”, discussed here.
Par. 4: Insert a “practically” (or something similar) in front of “infinite dimensional dynamical system”
Par. 4: The sentence “The three levels relevant here are micro-, meso-, macro“ can be omitted, since these are only relative terms and provide no specific information. (The following sentences bring up these terms with their descriptions).
Par. 5: What is meant by “condensing pressures of FMA”?
Box: Neural Networks
The text in the “Box section” on neural networks versus neurodynamics, should probably be put in a bordered box to set it apart from the rest of the text.
I think the term “neural networks” is too general to use in this context, and even if it is clear from the text that it concerns artificial rather than biological networks, it is not obvious which types of artificial neural networks (ANN) are considered. It is a too crude generalization to say that neural networks are Newtonian models, even though some (perhaps most) of them might indeed be considered so. Several neural network models proposed share the “Maxwellian” properties of the FMA models.
Mesoscopic state variables:
Again, the usage of "mesoscopic" has shifted from previous work, and this should be mentioned.
Par 1: It is stated that the “summation and smoothing of microscopic activities of neurons and glia” create the mesoscopic FMA. While the neural activities have been extensively discussed, there is no further reference to any glia activities in this article. This should at least be commented upon.
Fig. 1, legend: The figure does not appear to be the same as Fig. 7.17 p. 442 in [Freeman 1975], as stated, but rather as Fig. 7.1, p. 404. In addition, the ECoG in Fig. 1 (current article) is referred to as EEG in the original figure legend. (There is, in fact, no reference to electrocorticograms, or ECoG recordings anywhere in that book [Freeman 1975]).
The quantities and units used in the discussion of ECoG signals are somewhat confusing and should be looked over. For example, “extracellular tissue resistance” is given in [ohm.cm/cm^2], which seems incorrect, and “synaptic pressure” should be more explained. Also note that the unit for force in the CGS system is dyne, but abbreviated as dyn (in correspondence with cm, s, etc.). In addition, the CGS unit for power is erg/s, and not watts, which is an SI unit. In general, it is better to use the internationally more accepted SI units instead of CGS units, which would remove some of the confusion.
The “multiplication” dots in the dimensions (e.g. dyn.cm/s) should be different from a period (“end of sentence”) dot.
The reference to [Freeman, 2004], should either be 2004a, or 2004b or 2004a,b, relating to the reference list.
There should be an end-of-sentence period before the next-to-last sentence in the next-to-last paragraph. +++++++++++
RESPONSE TO REFEREE A: ++++++++++++++++++++++
The commentary of Referee A described the main difficulties that readers would encounter in attempting to understand our presentation. The text is far more complex than that suited for the level of Scientific American articles. The inclusion of many concepts and terms from physics with inadequate development leads to confusion. There is no clear message or direction at the end. Our terminology is often inconsistent and not in common use in neuroscience. The use of "EEG" in place of "ECoG" raises a discrepancy from prior usage.
In response to these insightful comments we have substantially revised our article by simplifying our presentation. We have done this by our strong emphasis on describing verbally and pictorially five primary features that we find in patterns of FMA. We present these forms as the primary components of FMA, which can be estimated by decomposition of the ECoG.
We describe the five features in language that does not require specialized knowledge in physics, mathematics, statistics, or even advanced neuroscience. We have made extensive use of references instead of footnotes to make the connections of each of our main steps to the presentation of the theory in published articles, so that if readers wants to follow from our text to proper derivations and interpretations, they can do so either through related articles in Scholarpedia or the literature in neurobiology, mathematics and physics.
We have kept and updated the same basic four figures and cited DVD clips that illustrate the dynamics of the five features. We have responded to the request for a clear message at the end by describing FMA as the neural mechanism for the early stages of perception, setting the explanation of this vital process as our goal, and describing how FMA can serve as the basis for explanation.
In order to keep the length to a minimum while still retaining the salient features of FMA, we have restricted the mathematical descriptions to references.
In the revised version, all the questions raised in the detailed comments section by the reviewer have been addressed.
The paper assumes a clear position in the field of the dynamical system witout any reference to logic or Turing machine. The brain is a dynamical system that changes in time to realise all the fundamental functions that we know. Now this approach is very important and near the reality of the brain structure. The problem with these approaches is to clarify the meaning and fucntionality of the brain as instrument to convey voluntary actions. How is the brain controlled? How can the brain realize its aim or purpose? Which is the logic of the brain ?. May be the brain uses many value logics that realise high levels of inferential process.The physiological meaning of the brain is well described. We know ,for example ,that in the quantum computer we have a similar dynamical system by unitary matrix, but we have also description of boolean logic by which we can organise inference and forcast future actions. In the FMA I cannot see how action is organised in a concrete and mathematical or logic way. The name force of the mass action is a good intuitive name but we know that in the dissipative system force is associated to the flux in a network where we can put the source of the force and the source of the fluxes. Now I appreciate the analogy with FMA and Maxwell equation where the forces are the electromagnetic four potentials and fluxes are the currents of electrons. In Maxwell equation we locate the sources in the charge particles that are in the same time dyanmically conditioned by the global force of the electromagnetic field. We know that in the Maxwell equations we have a complete mathematical description of the foces, in FMA we cannot see a similar global and mathematical formalisation . In the paper an attempt to have a mathematical representation of the FMA is the Neuropergolation. This attempt is based on the probability cellular theory. Cellular automata takes a local rule as fundamental logic element and repeats in parallel the same rule for any place. The well known problem with cellular automata is how we can know the global behaviour of the FMA when we know the local behaviour. I think that the Neuropergolation can be improved in suhc a way to be the first step to a more geeneral and global rpresentation of the FMA in line with Maxwell equation. I think that the paper assumes previous knowledge in different fields even if a lot of explanations are introduced.
+++++++++++++++ RESPONSE TO REVIEWER B: +++++++++++++++++++++++
Referee B emphasizes that our article omits all reference to logic and the Turing machine, and raises the problem that " I cannot see how action is organised in a concrete and mathematical or logic way", and that while "We know that in the Maxwell equations we have a complete mathematical description of the forces, in FMA we cannot see a similar global and mathematical formalisation ".
These are very important insights concerning mathematical and physical issues related to FMA. Accordingly we seriouly revised the manuscipt. In the present form, FMA article contains the phenomenological description of five primary features that we find in patterns of FMA. We present these forms as the primary components of FMA, which can be measured in the ECoG. Accordingly, the present work does not attempt to provide a mathematical or physical formulation of the related problems, which is the task of future studies. We omit references to forces and fluxes which need to be elaborated later, in line with Maxwell equation formalism.
In developing the consistent mathematical and physical theory of FMA, various methods can be invoked. We consider Random Cellular Automata and the corresponding neurpercolation theory is a potentially useful tool for the proper formalization. Moreover, Quantum Field Theory approaches can provide useful insights as well. As such developments are at the very incipient phase at the moment, we omitted their detailed introduction in the present article. We did provide references to those relevant works on neuropercolation and QFT approach to cognition by the authors and co-authors.
Finally, we emphasize our belief, shared with John von Neumann among others, that brains are not Turing machines. The neural mechanisms that control intentional, goal-directed behaviors depend on causal dynamic processes based in ionic fluxes in response to electric, magnetic and chemical forces. Formal logic is a very recent achievement of human evolution, which is clearly based in the neurodynamics of cerebral cortex, and which can be used to devise our models of cortical dynamics, but which does not in itself provide the model for cortical operations. Perhaps one might say that our use of differential equations and random graph theory is a branch of logic, but we believe that it is inappropriate to group those tools in the same category as the tools of formal logic.