3 :: golem-theory

What does the results of Penfield's Montreal Procedure tell us about the distribution of brain function around the brain? Before exploring the implications of the Montreal Procedure, note that current approaches for identifying regions of interest (ROIs) in brain mapping can be broadly classified into four categories (Li et al. 2009a) [1].

1. The first is manual labeling by experts based on their domain knowledge (Sobel et al. 1993). While widely used, this method is vulnerable to inter-subject and intra-subject variation and its reproducibility may be low.

2. The second method is to cluster ROIs from the brain image itself and is data-driven (Zang et al. 2004; Hyvärinen and Oja 2000; Beckmann et al. 2005; Calhoun et al. 2004). However, these data-driven approaches are typically sensitive to the clustering parameters used, and their neuroscience interpretation is not clear.

3. The third one is to predefine ROIs in a template brain, and warp them to the individual space using image registration algorithms (Shen and Davatzikos 2002). The accuracy of these atlas-based warping methods is limited due to the remarkable variability of neuroanatomy across different brains.

4. The fourth method uses task-based fMRI paradigms to identify activated brain regions as ROIs. This methodology is regarded as the benchmark approach for ROI identification.

Unfortunately, all of the above techniques are purely inductive/inferential , or 'bottom-up' methods, and are therefore in principia unable to map complex ROI's. The bottom-up mapping problem is inherently either redundant or ill-conditioned, or usually both. In The Logic of Scientific Discovery [2], Karl Popper rejected induction as a legitimate form of logic in the practice of science. Theory is more important than the facts, Einstein told Heisenberg, because theory tells us what the facts mean. Others argue that induction (ie observations) is the first step to the crafting of a suitable hypothesis, a step which is as inescapable as it is informal. The exclusive use of induction is called inductivism. It claims that (we) "can arrive at objective unbiased conclusions only by... recording, measuring, and describing what we encounter without any root hypothesis....".[3]

It is time to return to Penfield's Montreal Procedure, which revealed the existence of matched pairs of motor and sensor homunculi in left and right cerebral cortices. These images, which have achieved the historical status of modern or pop art, are reproduced in the left hand figure above. Yet their very familiarity belies the fact that they still have much to offer the cognitive theorist. They represent a more accurate map of one subject's structure:function data than fMRI images generally, since Penfield's electrodes were only a few millimetres in diameter, and were accurately located on the cortical surfaces by external verniers.

If one imagines the left half of the image being folded over upon the right half, then one would have constructed a 'paper' model of the right half of the central sulcus, the fissure that separates the input and output channels at the cortical level. This situation is depicted on the GOLEM framework in the right hand figure above, which depicts one (left or right) half of the GOLEM. Increasingly predictive 'motor plans' can be found 'upstream' of the motor homunculus. By the same token, increasingly historical 'sensory states' can be found 'upstream' of the sensory homunculus [5]. Note that there is a two-channel GOLEM in each cerebral hemisphere.

In the figure above, the sensory and motor homunculi revealed by the Montreal procedure are appropriately labelled.

The most common (and historically fruitful according to [4]) solution is to use a combined top-down/bottom-up method which is formally equivalent to the so-called 'scientific method'. This well-understood technique involves (1) proposing a theory, hypothesis or model (all these terms mean essentially the same thing) (2) constructing a proof by disproving one exemplar of the null hypothesis (deductive method) or searching for observations that seem to most unremarkably support the model (retroductive method). This is the reason for the (axiomatic) need to iteratively assume a model, in this case a pair of models - a software model (GOLEM) and a hardware model (NEODE). GOLEM theory claims that the separate adaptational requirements for software and hardware in computers apply to brains, because, quite literally, the brain is a computer. Conversely, computers are (extensions of our) brains. We are able to identify consciousness analogs in all modern computers, namely their graphical user interfaces, or GUIs. The modern GUI enables the computer user to easily and flexibly change the target focus of their current task by tracking the progress of all important tasks in a single slice-of-time 'snapshot'. This multi-tasking snapshot was given the name 'Situation Image' or SI in one of the first websites in this series www.tde-r.webs.com

Computer consciousness is one half of the bioplausibility requirement of the model, the other half being emotionality. While digital computers with GUIs have a non-living consciousness component (ie the GUI), they have no non-living analog of emotions. Instead, the computer user must supplant their own emotions. The GOLEM model possesses non-living analogs of both consciousness and emotionality. It relies on a concept of emotionality recently investigated by Elizabeth Feldman-Barrett. In her research, she identifies basic emotions as having the function of linking up two previously unconnected memory regions, and allocates this task (unsurprisingly) to the neurorgans at either end of the so-called limbic axis, namely the amygdala and the hippocampus. The tragic case of patient H.M. illustrates the vital need for the hippocampus, without which formation of new (anterograde) declarative memories is impossible. GOLEM uses the <'what' ; 'where'> dyadic memory paradigm, in which the amygdala locates the 'what' part of any new (declarative) memory, while the hippocampus locates the 'where' part.

In the first website of this research program (www.tde-r.webs.com as named above) the two-channel GOLEM model had not yet been conceived. Instead, an equivalent multi-scale fractal model, using a similar notion to triune models, was introduced. Called the TDE, a recursive acronym for TDE Differential Engine, it refers to a Turing-friendly version of the GOLEM, incorporating certain formal resemblances to Turing's Universal Computer. For example, it used a circular tape memory which retained Turing's division into e-for-eraseable and f-for-fixed squares, representing write-only and read-only data positions, respectively.

The input and output channels which constitute the GOLEM are identifiable on the TDE as key functional features in the following manner-"There are two such buffer-archive pairs (BAP's) in every TDE, one spatial and one temporal. There is a BAP for cospatial features (Temporal-Parietal), and a BAP for synchronous features (Limbic-Frontal)."

The TDE was formulated to be language-ready 'out of the box', if you will. It includes as a basic design parameter, the most likely possible reason for gross (macro-level) cerebral lateralisation (the differential functions observed in left compared to right hemisphere cognition [6]). This (non-adhoc) reason is the 'convergent evolution' of the system as a whole, where it emulates (exapts) pre-existing sagittal (ie fore-aft) asymmetries within each hemisphere, by exhibiting similar, but global asymmetries, now between the hemispheres. This sagittal asymmetry was then named GOLEM, an acronym [9] which, because it reflected the theory's core (goal-oriented c.f. FETCH-EXECUTE) operational agenda, was later used instead of TDE to denote the entire research project. By doing this, both animal and human brains are explained by a similar mechanism, a change which is also more faithful to the evolutionary record. These lateralised observations have arisen, for the most part, from observations of functional and behavioural deficits in stroke/ trauma victims [8]. Within the brains of these patients, uncontrolled or inadequate coagulation within one or both of the cerebral arteries (via clots or bleeds respectively) cause selective death of the neurons in a particular region, with matching loss of cognitive and/or somatic function.

1. Tianming L. (2011) A few thoughts on brain ROIs. Brain Imaging Behav. 2011 Sep; 5(3): 189-202.

2. In the preface to his book Popper includes the following introductory quote by Novalis (a.k.a. Georg Philipp Friedrich Freiherr von Hardenberg)- Hypotheses are nets: only he who casts will catch.

3. Mayr E. (1982) The Growth of Biological Thought: Diversity, Evolution and Inheritance. Cambridge, MA: Belknap Press of Harvard U. Press

4. Rothchild, I. (2006) Induction, Deduction and the Scientific Method- an eclectic overview of the practice of science. Society for the Study of Reproduction, Inc.

5. Note that the GOLEM convention of placing the input channel at the right, and motor channel at the left is apposite to the sense of Penfield's illustration.

6. In Tooby & Cosmides, the theory that the human mind is densely multimodular - that it is not a general purpose computer, but a whole array of special purpose ones- is presented as if it were a fait accompli, a done deal, rather than what it really is- just a theory, one that opposes the one (GOLEM theory) presented here, which posits that our brains are general purpose computers which owe their amazing versatility to the general nature of their programming which lets each infant (with a lot of parental assistance) guide its own behavioural development.

7. Remember, there is one hippocampus for each hemisphere, at the temporal sulcus.

8. This region can now be readily located using BOLD imaging, a type of fMRI which measures the dynamic oxygen metabolism in any given region. The basic idea is that dead neurons use much less oxygen than living neurons. Modern imaging methods augment, rather than replace, older observational approaches.

9. The resemblance between the acronym GOLEM (Goal Oriented Linguistic Emulation of Mind) and the infamous mad-mud-man from Hebrew mythology is unintentional, a fact easily verified by inspection, since each letter of the acronym stands for an essential component of the underlying model.