Distal vs Proximal Models

• Why vs How
• Optimality
• Learning & Adaptive Control
• Implications for the Medical Model

Why vs. How

The distinction between distal and proximal models is quite a big deal. Scientific investigators who make proximal models (models of proximal causes of biological phenomena) have fundamentally a different perspective than investigators who make distal models (models of distal causes). Many investigators fall between the extremes, but most will have a view on the issue -- so what is the issue?

The ‘proximal’ school attempts to understand the mechanisms by which the behaviour or structure is controlled or physically generated. For example, this school asks how a limb moves. Which brain structures are involved? What is the biochemistry and physics of muscle? How is bone deposited or absorbed? How does the brain make consciousness? What causes death? In short, how does the genotype lead to the phenotype?

The ‘distal’ school attempts to understand the evolutionary function of an animal’s (including human) behaviour or structure. Why does the animal behave that way? Why are bones the shape they are? Why do we have consciousness? Why do we die? In short, what are the selective pressures that have led to the evolution of the current phenotype? In this sense the genotype is merely a ‘depository’ of the information needed to make fit phenotypes.

Consider the question: why do birds fly? A proximalist approach might be to discover the detailed structure of a bird’s wing such as the bones, feathers, blood supply, and the neural circuits involved in controlling flight muscles etc. Ultimately, perhaps the proximalist might also want to know the genes involved in the embryonic development of wings and their control in flight. In essence, the proximalist really wants to know how birds fly the way they do.

In contrast, the distalist is interested in the evolutionary advantages of flight -- how does flying enhance survival? What were/are the selective pressures that led to flight? What is the foraging range? Which predators are avoided or encountered by flight? What are the trade-offs between size and energy needed for flight, and caloric intake, etc. The distalist wants to know why birds fly the way they do.

Taken to extremes, advances could be made in one approach without any consideration for the other. Both would like to know what bird flight is, that is they would like to observe the behaviour, but the proximalist may believe there is no real need to understand selective pressures on bird flight. The distalist may not need to know the detailed biology of a bird or its wings. However, what would be the point of understanding the structure of wings without some understanding of what wings are used for and why they are advantageous? Equally, it would be pointless to try to build a distal model of flight without knowing the physiological limitations imposed on wings.

We should not assume that models (which are really hypotheses) are purely proximal or purely distal (as unfortunately some do!). There are gradations in between. For example, instead of asking how flight contributes to survival as a whole, we might ask what is the cost of flying at a given speed. Here we believe that speed is important for survival (avoiding predators, chasing prey, competing with others, etc). That is, all other things being equal, it might be of benefit to be as fast as possible. We might then ask about the mechanisms that convert energy into speed (metabolic rate, body mass, wing shape, and so on). Here we need to know something about mechanisms as well as evolutionary goals. Of course all other things are not equal, and we would need to look at trade-offs between speed and other costs. It is self-evident that we would like to understand bird-flight both proximally and distally.

We have used bird-flight as an illustration because the relationship between the wing and flight is obvious (ignoring ostriches, of course). Let us consider a trickier example, which is part of the focus of our research. When we humans make fast reaching movements, our hands tend to follow similar trajectories that are roughly straight. The hand or finger accelerate to some maximum velocity and then decelerate to stop at the target we are aiming for. If the movement is brief the velocity profile has a more-or-less symmetrical bell shape. We all do this unless we consciously try otherwise. Given the myriad ways we could move our hands, why does our brain choose the bell-shaped velocity profile? Proximal models may explain how the neural circuitry moves the hands, but it cannot answer the why question. We need to understand the evolutionary pressures that make the observed trajectory the preferred strategy. We need a distal model (we shall return to this important issue).

There is no doubt that our current era is dominated by proximalism. Indeed, the standard medical model is essentially proximal. We devise proximal models about how mechanisms go awry, and try to find ways to correct these abnormal mechanisms in order to cure the disease. Do distal models have a role to play in medicine? We think so, but we need to develop our argument further. [top of page]

Optimality

Local optimality is a logical consequence of natural selection. Through small random genetic variations, natural selection (i.e. competition) will favour ever fitter phenotypes. Eventually (we surmise), some limit must be reached where fitness cannot be improved any more. Any variation can only lead to a less fit phenotype, and in the jargon, we say that a locally optimum (in fitness) has been reached.

The assumption of optimality is a powerful tool in understanding the control of behaviour and allows us to build quantitative distal models. We can postulate factors that affect fitness (e.g. speed of movement, caloric intake etc.) and then, using rather sophisticated computational and/or mathematical techniques, we can find the theoretical optimum and compare it to observed behaviour. Various cost function can be explored and refined to find plausible explanations of behaviour. Such models can then be used to make predictions in unusual or artificial situations. Ideally, the behaviour can also be examined proximally to see how the organism achieves the optimal behaviour.

Optimality does not mean the ‘best of all worlds’! It means that costs and benefits of any variation are balanced. For example moving quickly may be a benefit (increased fitness) but it may be lead to greater inaccuracy or require more energy, which are costs (decreased fitness). Depending on the organisms ‘niche’, there will be some optimal trade-off between speed, accuracy and energy. This trade-off may vary with different environments depending on the organisms life style and current tasks (such as grazing vs. escaping a predator).

Optimality only means that a local optimum is found. Random genetic variations tend to be small, so that improvement in fitness will be gradual. This means that only a local peak in fitness can be found. To find a higher peak would require first going downhill. This would lead to a decrease in fitness and so would be unlikely to occur. Also, optimality is not teleology. Evolution does not ‘know’ what it is trying to achieve and does not even ‘know’ if it is at a local optimum. It is simply ‘trial and error’.

Although the assumption of local optimality is extremely useful for creating distal models, it has some caveats, and some antagonists. There are a plethora of popular science books dealing with all these issues and we will not discuss them here. [top of page]

Learning & Adaptive Control

Evolution occurs gradually over many generations (hence the word ‘distal’) and cannot respond to sudden changes in environment. Yet individual organisms encounter different environments during their lifetimes, particularly if they are mobile. To cope with such changes, organisms have greater chances of survival if they can modify their own behaviour to suit local environments. We call this development, learning, adaptive behaviour, etc. The more flexible the behaviour is, the greater the chance of surviving more extreme or uncertain environments.

Some evolutionary biologists have considered the existence of behavioural plasticity as an argument against optimality because the optimal trade-off is not the same for all individuals (e.g. frequency selection). In fact plasticity allows individuals to maximise their fitness by attaining behavioural objectives by different behavioural means in different environments. Indeed, except for some infantile reflexes and fixed action patterns (‘instincts’), most behaviours are not directly genetically determined but are the product of learning. It is the learning mechanisms and the learning objectives that are genetically coded.

This leads us back to the proximal vs. distal issue. At one extreme, we can ask how does learning take place (what is the role of memory, what are the learning rules, what is the neural organization that permits learning, etc.) -- the proximal approach. At the other extreme, we can ask what is the objective of the learning: what is being learnt? – this is the distal approach.

Let us return to our previous reaching example because, as it turns out, it is also an example of adaptive control. As mentioned earlier, normally, fast reaching movements have a stereotyped bell-shaped velocity profile. If we now change the physics of the arm (add weights to it etc) then, not surprisingly, the trajectory changes. However, after some trials the former bell-shaped trajectory is gradually re-instated, even though the dynamics of the arm have been changed. We say (using jargon) that the trajectory is under ‘adaptive control’, and the brain gradually learns a new motor command to send to the muscles in order to generate the desired bell-shaped trajectory.

There are numerous proximal questions about this fascinating phenomenon. How does the brain knows the dynamics have changed? How does it learn a new motor command? Which parts of the brain are involved in this adaptation? Nevertheless, these questions do not address the distal problem: what is the objective of the adaptive control? That is, there must be an adaptive control process in the brain whose objective, when reached, turns out to be a bell-shaped velocity profile.

We are back to our original distal question: why do humans make bell-shaped reaching movements? There is now, however, a subtle but important difference! Previously, the argument was that bell-shaped trajectories must be optimal (maximised fitness) over evolutionary eons (many generations). In principle, it could have been argued that bell-shaped trajectories occurred because, through evolution, that is the only way muscles and skeletons can make movements. (Of course, the distal question of why this should be, would still have been valid). However, we are now saying that bell-shaped trajectories are the product of a control process that is active within the lifetime of an individual (in fact, often within a few minutes of a lifetime). It means that the distal objectives are actively maintained, via adaptive control, in each individual (we suggest that the reader contemplates whether it could be otherwise). This implies that there must be neural machinery which codes the objectives that lead to the bell-shaped trajectory. [top of page]

Implications for the Medical Model

So what happens in disease? Clearly if my arm is paralysed because the nerves have been severed, then I will not be making any bell-shaped trajectories! Many diseases, though, are more subtle and perplexing. For example I might develop a neurological problem that prevents me from making nice accurate bell-shaped trajectories. This might become so severe as to interfere with my everyday life and to warrant being labelled a ‘disorder’. Conventionally, we search for a proximal cause for this abnormality, such as brain damage, nutritional problem, family history. We might find a correlate, or as often occurs in medicine, we may never find a ‘cause’ and then we would call it ‘idiopathic’ (medical jargon for no known cause). If we are lucky (!?), we might find a genetic mutation that is inherited in the family. Whether finding any ‘cause’ will help us understand how the disorder occurs is another matter!

Let us now ask the unconventional distal question. We know that bell-shaped trajectories are under adaptive control in the healthy individual. Therefore, why (not how) can I not make them? The simple answer is that the adaptive control process must not be working properly. One possibility is that my adaptive control machinery has itself been damaged so that the reaching system is completely out of control. The second possibility is that the adaptive control is working but attaining an unusual control objective, either because neural coding of the control objective is disrupted or because the signals needed to drive adaptive control are abnormal.

This last possibility is particularly interesting because it is at least theoretically possible that abnormal feedback signals from muscles or vision could misdirect adaptive control towards an abnormal behaviour. We would not find anything wrong in the brain but we would be unable to stop the abnormal behaviour because the adaptive control would always try to re-instate the abnormal behaviour (because this is what it ‘thinks’ it is trying to achieve). This may explain why some human motor disorders are chronic and untreatable. However speculative, it also opens up the possibility that we could intervene and redirect the adaptive process to produce normal behaviour – rehabilitation.

Consider a real example. In one of our experiments we explored saccadic adaptive control in adolescents who had chronically inaccurate saccades (dysmetria). These patients had suffered from Dancing Eyes Syndrome in infancy. It is well known that saccade accuracy is under constant adaptive control, so why does inaccuracy persist in these children? We originally believed that brain damage (in the cerebellum) had destroyed the adaptive control mechanism. We therefore carried out an experiment to see if these children could adaptively control their saccades. To our surprise, we found that their adaptive control was functioning. This is just the beginning of our study, but at least we now know that abnormal behaviour is the product of adaptive control (rather than resulting from its absence). Our next question is why does adaptive control allow this dysmetria to persist? Clearly, we need to know much more about saccade adaptive control, both proximally and distally.

Although we have focussed on adaptive control, the basic idea is applicable to other processes in which behaviour and structures are modified during the lifetime of the individual. For example, in human development we can ask the proximal question of how does development take place, or we can ask the distal question of what the objectives of development are. This becomes important when we consider congenital and perinatal disorders

On this website you will find information and descriptions of theoretical and experimental studies whose goals are to understand movement control distally, as well as proximally. This information is quite wide ranging, from mathematics to neurology, but hopefully you will find some of it interesting and thought provoking. [top of page]