13 May 2009

Feinberg, Why the Mind is Not a Radically Emergent Feature of the Brain


by Corry Shores
[Search Blog Here. Index-tags are found on the bottom of the left column.]

[Central Entry Directory]
[Emergentism, Entry Directory]
[Computation Entry Directory]




[Some argue that we can continue our lives in software that perfectly replicates our brain dynamics. This position relies heavily on emergence theory: the mind is constituted by lower level physical structures and dynamics of the brain; however, the higher level exhibits properties not found on the lower levels. In the case of our brains, the highest emergent level is our consciousness. So no one neuron is conscious. Yet out of the complex dynamics of all neurons working chaotically in parallel, our consciousness emerges. William Hasker likens such an emergent consciousness to the magnetic field around a magnet. The field is different than the magnet. But the field would not be there were it not for its physical substratum.

Todd E. Feinberg is critical of such emergence theories of mind. His primary objection is that consciousness is not in any way 'higher' than the lower levels, even though it is made-up of them. We will examine one of his articles. Unfortunately I do not have any access to it except for a limited preview on google books, so parts will be missing from the following summary.]



Todd E. Feinberg

Why the Mind is Not a Radically Emergent Feature of the Brain

in

The Emergence of Consciousness


Abstract


Many claim that the mind is a "radically emergent feature of the brain.' Feinberg disagrees. He will examine how the ideas emergence, reducibility, and constraint relate to each other and to hierarchical biological systems. Then he explores such emergentist theories as the one Roger Sperry proposes. Feinberg argues that Sperry's model fails. Consciousness does not arise 'at the top command' of a 'non-nested neurological hierarchy.' Feinberg offers a nested hierarchy model without a top or summit. It avoids dualism. Finally, consciousness's non-reducibility is not a result of emergence, but rather of the mutual irreducibility of the subjective and objective points of view. Consciousness is not mysterious. We may explain it in terms of brain evolution and normal neural functioning. (123)


I
Introduction


The brain is made up of billions of neurons. How does consciousness unify? John Horgan presents the question using Humpty Dumpty. We may take apart the brain to analyze its individual pieces. But we cannot put Humpty back together again to explain the mind's unity. (123d)

One potential solution to the 'Humpty Dumpty Dilemma' is that the mind, somewhat mysteriously, emerges from the brain. In this line of reasoning, consciousness, as an emergent feature of the brain, extends beyond, or is 'more than the sum of the parts' of the brain. And since the mind transcends the brain, while the brain my be divisible, consciousness can nonetheless emerge unified from the brain like the eye that emerges from the top of the pyramid on a dollar bill. (124)



Feinberg does not like how consciousness lies atop an organizational hierarchy in this model. He does not dispute their being a hierarchy. His objection is that the hierarchy cannot be said to have a top. (125a)

To explain emergence, Feinberg quotes from Jaegwon Kim:

... although the fundamental entities of the world and their properties are material, when material processes reach a certain level of complexity, genuinely novel and unpredictable properties emerge, and that this process of emergence is cumulative, generating a hierarchy of increasingly more complex novel properties. Thus emergentism presents the world not only as an evolutionary process but also as a layered structure a hierarchically organized system of levels of properties, each level emergent from and dependent on the one below (Kim, 1992)
Consider water for example. The H2O molecule does not itself bear the properties of liquidity, wetness, and transparency. However, an aggregate of them does. (125)

At each hierarchical level of the organism, novel properties emerge from the levels below it.

C. Lyod Morgan held the view that consciousness lies at the top of an emergent pyramid. At the lower levels are atoms. They combine and exhibit the emergent quality of molecularity. Molecules combine into living forms with the emergent property of vitality; "yet higher, a new kind of natural relatedness supervenes and to its expression the word 'mentality' may be applied" (Morgan, 1923, p.35 qt. 125d)

Searle distinguishes two types of emergence.

Emergence1:
A system's higher order properties can be understood through a complete explication of its parts and their interactions. In this case, the emergent qualities can be said to be caused by the lower level. Hence Searle calls them 'causally emergent system features.' Consciousness causally emerges from the nervous system in the same way that liquidity causally emerges from water molecules.

Emergence2:
Here the emergent property has causally properties that cannot be explained by the lower level.
Searle writes:
If consciousness were emergent2, then consciousness could cause things that could not be explained by the causal behavior of the neurons. The naïve idea here is that consciousness gets squirted out by the behavior of the neurons in the brain, but once it has been squirted out, then it has a life of its own (Searle, 1992) (qt. 126)

Emergent features go beyond what we may expect from the lower level. Hence the higher levels are greater than the sum of their parts.

Many emergentists also hold a physicalist position: everything can be reduced to and explained by physical reality. However according to Van Gulick, to argue for emergent consciousness posses problems with giving a physicalist account of the relation between the brain and the mind:
The notion that causal powers might exhibit radical kind emergence merits special attention since it poses perhaps the greatest threat to physicalism. If wholes or systems could have causal powers that were radically emergent from the powers of their parts in the sense that those system-level powers were not determined by the laws governing the powers of their parts, then that would seem to imply the existence of powers that could override or violate the laws governing the powers of the parts, i.e. genuine cases of what is called 'downward causation' . . . in which the macro powers of the whole 'reach down' and alter the course of events at the micro level from what they would be if determined entirely by the properties and laws at the lower level (Van Guilick, 2001, [this volume], p. 18-19) (qt. 126)
[Next page on Roger Sperry missing. More on Sperry can be found at this entry and this one.]

Feinberg argues that
consciousness has proved to be particularly resistant to a simple scientific reduction to the brain. If consciousness can be shown to be in fact be a radically emergent feature of the brain, this could in principle account for a failure, on the basis of neurological principles, to reduce consciousness to the brain. And this in turn would force one to endorse a form of dualism. (128)
Downward causation, or 'constraint,' is the "process by which higher levels in the hierarchy impose control over the lower levels." (128)
Hierarchical biological systems operate via constraint. The individual cells of the human body constrain the microscopic organelles of the cell to perform sub-cellular metabolic processes, and the organs of the body in turn constrain these cells to perform functions such as secretion or muscular contraction. The entire body of the person constrains the individual organs to breathe, digest, and perform the macroscopic functions necessary for life. For example, the mitochondria, the organelles responsible for generating energy from oxygen via cellular respiration, are a microscopic part of the cell that along with other cells make up the tissues that eventually give rise to the lung. The mitochondria contribute to the emergence of the lung at a higher level of the hierarchy of the body. The lung, at a higher level on the hierarchy, displays emergent features not possessed by mitochondria, for example breathing. If the lung did not breathe, the body would not have oxygen, and if we did not have oxygen, the mitochondria would not be able to carry on cellular respiration. The mitochondria contribute to the emergence of the lung, and the lung in turn constrains the mitochondria. (129)
Feinberg notes that this lung emergence is emergence1.
The operation of the higher levels of the hierarchy of the lung is clearly seen to be the result of the operation of the parts of the lung and their interactions. Therefore, it is not surprising or mysterious that the whole lung considered in its entirety can affect and constrain the individual cells and organelles within it. In contrast, in Sperry's view, the mind is an emergent2 or radically emergent feature of the brain. Thus although the mind emerges from the brain, and is not reducible to the brain, it can nonetheless cause material events to happen in the brain. In this way, Sperry suggested that '. . . subjective properties were seen to exert control over the biophysical and chemical activities at subordinate levels' (Sperry, 1977), and the immaterial mind possesses causal properties over the material brain and constrains it.
The causal power attributed to the subjective properties is nothing mystical. It is seen to reside in the hierarchical organization of the nervous system combined with the universal power of any whole over its parts. Any system that coheres as a whole, acting, reacting, and interacting as a unit, has systemic organizational properties of the system as a whole that determine its behavior as an entity, and control thereby at the same time the course and fate of its components. The whole has properties as a system that are not reducible to the properties of the parts, and the properties at higher levels exert causal control over those at lower levels. In the case of brain function, the conscious properties of high-order brain activity determine the course of the neural events at lower levels. (Sperry, 1977; reprinted in Trevarthen, 1990, p.384) (qt. 129)
Feinberg's objection is this. If the model of consciousness involves higher order mental features that are radically emergent, then this implies a dualism. Hence we should not posit emergent2 features. Moreover, we need a model that explains "the hierarchical constraint that guides abstract thought, perception, and volitional motor behaviors." (130)

Feinberg will now show that there are no hierarchies in brain organization. This will refute Sperry's model of the emergent mind at the summit or top level of neural strata. His example is our vision system.

He explains that some research suggests there is a hierarchical filtering down of vision information. The individual eye neurons see very limited parts of the picture. These neurons converge into a more limited strata of neurons that processes more specific information.
Each cell of the retina responds to a particular area of the visual world called its receptor field. The receptor field of each retinal cell is quite small, and the brain must build up whole and unified mental representations from these discrete points of contact with the world. (130)
Single neurons in vision-processing part of the brain do not responds to particular signals from just one eye neuron. There needs to be a row of eye neurons that perceives a line before the brain's visual neuron will fire.
a single neuron in the brain could respond to a line in the world if a line of adjacent firing cells each responsive to an individual point of light converged upon a single cortical cell further along the processing stream. The single cortical cell could 'add up' the points of light that each lower order cell had responded and in this fashion a single higher order cell could respond to a line. (130)
The convergence progresses to layers of neurons processing more and more specific information [see this entry for an explanation of converging neural circuits].
Hubel and Wiesel also found that multiple simple cortical cells converged upon other single higher order cells further along the processing stream to create what they called complex cortical cells, and complex cells in turn converged upon single neurons to create hypercomplex cells with increasingly specific and complex response properties. The model of Hubel and Wiesel is hierarchical in that simple cells lower or earlier in the neural processing chain create cells of ever increasing complexity higher up on the neural hierarchy. (130)
This convergence continues to a high degree of specificity. It can explain how we recognize specific people's faces.
This hierarchical process ultimately produces advanced 'higher order' cells that possess amazingly specific response properties. For example, there are neurons in the inferior temporal cortex far 'down-stream' from the simple cells found early in primary visual region that respond preferentially to highly specific and complex stimuli such as hands or faces. Some of these neurons are even responsive to a frontal view of a face while others react best to a side view. Neurons of this type led to the notion of the 'grandmother' cell, a cell so specific that it fires only to the face of ones [sic] own grandmother (Barlow, 1995). While cells of that degree of specificity do not exist, it is in general true that the farther along a sensory processing stream that one looks, the more specific a cell's response characteristics become. (130-131)
Feinberg addresses a problem with this model. To recognize a face is very specific. It does not involve knowing where the face is in the visual field. However, convergence should not eliminate this information. So to both recognize a face and know where it is, as is usually the case, we need both the higher and the lower layers of cells equally and simultaneously.
However, in the process of creating higher order cells, the receptive fields of these cells increase in size. While a retinal cell early in the stream monitors a small and specific point in the visual field, a hypercomplex cell such as a face cell will react to a face that appears almost anywhere in the visual field. It follows that while the cells early in the visual stream with small receptive fields 'know' where each line of the face is, these early cells do not 'know' that a given line is part of a face. The face cells, on the other hand, 'know' there is a face, but due to the process of topical convergence, these cells don't know where the face is located in space. While it is true that cells of the brain project to successive levels in a hierarchical fashion in order to code for increasingly specific complex and abstract properties, the information coded by cells earlier in the process is not and cannot be lost in awareness.
Hence
cells comprising both the early and late stages of the visual processing stream must make a unique contribution to consciousness. (131)
From this, Feinberg concludes that Sperry's hierarchical emergent model of the mind does not hold. There are not higher levels of neural organization. Consciousness is distributed across many different areas at once.
Therefore, in contrast to Sperry's model, consciousness does not emerge at the 'top' of the visual hierarchy. Even though the brain seems to be creating 'pontifical neurons' at the 'top' of the visual hierarchy, the conscious mind is in reality 'spread out' across the activity of millions of neurons located in different regions and levels of the brain. A visual hierarchy exists, but many levels of the hierarchy make a contribution to conscious awareness. This is even truer when the brain must coordinate features of a stimulus from multiple sensory domains, such as an object having both visual and auditory features. When considering the perception of a honking red Corvette passing a brown barking dog, binding must occur in both visual and auditory centres that are located in even more widely distributed areas of the brain. The brain must bind the colour red and the honking horn to the car, and the colour brown and the barking to the dog it passes, and not the other way around. And both of these elements must be bound to each other within the overall perceptual field of awareness. (131-132)
From the standpoint of emergence theory, the point is that there is no 'top' of the visual pyramid where visual awareness may be said to radically emerge. All levels of the visual system lower and higher, contribute to consciousness (132)
Feinberg offers other examples from our motor systems.
If you say to yourself 'I will now move my arm' and you do so, you will experience an 'inner I' as the source of that action. Suppose I, as your neurologist, search for the source of that 'will' somewhere in your brain. I will find that there is no central, integrated and unified physical locus that is the source of that action. Just as in the perceptual hierarchy that appears to lead to the 'grandmother cell' in the visual system, there is no singular 'top' of the motor hierarchy, no 'ghost in the machine' as Gilbert Ryle (1949) put it, that can serve as the source of our unified 'will'. (132)
Consider also when we speak. We do not just use motor neurons to move our mouths and tongues. We also use parts of our brains that control fluency, word selection, continuous monitoring, emotional inflection, physical hand gestures and so forth.
It is clear that vast areas of the brain are involved in the complicated act of speech production, yet somehow when we 'will' to speak, the entire act is unified into a whole and coordinated behavior. But there is no unified region or hierarchical level in the brain that in and of itself constrains our actions and intentions. Therefore, there does not appear to be any material 'top' to either the perceptual or motor hierarchy. (133a)

VI: Non-nested and Nested Hierarchies


Sperry's emergent hierarchy is "non-nested:"
A non-nested hierarchy has a pyramidal structure with a clear-cut top and bottom in which higher levels control the operation of lower levels. A classic example of a non-nested hierarchy is a military command in which a general at the top controls his lieutenants, who control their sergeants, and so on, down the chain of command until we finally reach the level of the individual troops. (133b)
The non-nested hierarchy that Sperry envisioned is considered non-nested because while the successive levels of the hierarchy interact, each level of the hierarchy is physically independent from all higher and lower levels. The various levels of a non-nested hierarchy are not composed of each other. (133bc)
In contrast to non-nested hierarchies are compositional or nested hierarchies. In these forms of organization, the higher levels contain the lower ones, in fact, they are entirely composed of them. For example, organs are nested within organisms, tissues within organs, and tissues are made-up of cells, and so forth.

The important difference is that there is no top and bottom to a nested hierarchy.
We are physically composed of minute organelles that are hierarchically organized to create a human being. In the hierarchy of a living person, it is the complete person who sits at the top of the hierarchy, and that person is not separate from the parts of which he or she is composed. Individual elements of the body that make up the person simultaneously contribute to the life of that person. The 'parts' of the person in this way are nested within the totality of the human being. (133-134)
Feinberg argues that the brain and mind are likewise nested hierarchies. Even though there might be a small number of cells that fire when we see a particular face, many other neurons had to fire along with it in order for that one to fire. To know that someone is your grandmother is not anywhere near the sum total of our consciousness. The details, images, so called "lower level" neuronal behavior make-up no less a portion of our consciousness.
In the perception of a speeding red Corvette, the roof and trunk of the car are composed of tens of thousands of individual line segments. These individual line segments of the car's outline then are combined into longer segments that produce the car's overall shape and form. The lower order features, for instance, the exact position of a small line sequence in space, emerge in awareness as 'part of' something else, such as the outline of the car. A short line segment is 'bound' to a longer segment to create the outline of the car, just as a small patch of red is 'bound' to a larger red patch that is part of the door. The redness of the Corvette is 'bound' to its shape which is bound to its movement which is bound to the honking horn, until all representations are 'bound together' to create the entire image. The colour, shape and movement of the car are nested together within the image of the car and this image in turn is nested within the entire scene. (134)
When elements are bound in this way, they "are represented in awareness dependently and are nested together."
Colour and shape are represented in awareness as a nested totality. We do not experience the colour of the car independently from the experience of its shape. On the other hand, the experience of the dog that the car passes is bound to the car within the entire visual experience, but is not tightly bound to the colour or shape of the car. The higher level or complex neurons that code for car and dog as entities make greater independent contributions to conscious awareness within the specific colour or shape of the car. (134d)
[Hume makes a very similar point with his white globe demonstration in §61 of his Treatise.]

[page 135 not available.]


Constraining and Volition


When our eyes reflexively blink, this serves a purpose, but it is not the result of purposeful behavior, such as speaking beautifully. To do so, we must unconsciously activate our mouths, tongues and so forth. These neurons are nested within the higher levels of organization that focus on the ideas that the body is supposed to express, and not on the details of the physical expression. Feinberg argues that willful conscious behavior is not an emergent higher order that constrains a distinct lower order. Rather, it is a higher part of a system that constrains the inner parts of its working so to fulfill a given purpose.
When we speak, it is the idea we wish to express, the purpose of our speaking, that sits at the highest level of the action hierarchy, and the degree of constraint and purpose distinguishes intentional from non-intentional telenomic behaviors. Beating hearts do not act purposefully; toolmakers do. And the greater the constraint over nested parts within a hierarchy, the greater a behavior is purposive and therefore conscious. (138b)
[The rest of the text is incomplete, except for the conclusion:]

... By its very nature, the brain functions in a fashion that produces irreducibly personal mental states, and the failure of mental states to be completely reduced to neurological states is not based on any variety of radical emergence of the sort posited by Sperry and others. Rather, from the standpoint of neurology, the irreducibility of mental states to the brain, to what extent such irreducibility exists, depends solely on the inability to reduce the subjective to the objective. (143)

Feinberg, Todd E. "Why the Mind is Not a Radically Emergent Feature of the Brain." in The Emergence of Consciousness. Ed. Anthony Freeman, Thorverton, UK: Imprint Academic, 2001.
More information and partial preview available at:



No comments:

Post a Comment