Can You Measure Temperature With A Turnip?
This essay might equally well be called 'Can you measure genes with a turnip?' or any number of other things. In it I will attempt to show that any theoretical term must be understood and well defined before we can consider the truth of the theory it appears in - namely that testing a theory is to see whether the theoretical terms involved fit together as the theory prescribes, and does not involve questions about what we mean by the theoretical terms. This would probably be termed either an operationalist or an instrumentalist account of the meaning of theoretical terms, it that it is largely concerned with how a particular thing operates in the world, and what the direct observational consequences of such a thing are. I shall argue against a 'double language' model of the meaning of theoretical terms - one which holds that a term can be defined in more than one way, it is only the interaction of theoretical and observational ideas that gives a term its meaning. I feel that the 'double language' model can only serve to create something of a muddle, and will not engender useful or efficient science, nor do I feel that its accusations about more purely empirical scientific definitions are fair, and do not amount to an adequate rebuttal.
>A model that holds that we must have already defined the terms we are discussing does not mean that we cannot have theories which debate aspects of a theoretical term, so long as we are not defining the term according to that particular aspect. To take a hopefully non-theoretical example to start with, supposing there were a serious scientific investigation into whether turnips were blue, we must already know what we mean by a turnip, and this knowledge cannot be in terms of colour - we can't say a turnip is a blue root vegetable, though we could say a turnip is a rounded root vegetable with a particular type of leaf, for example. The important thing is that when we are discussing a class in terms of a theory we have some way of deciding whether a given thing fits into that class - it's no use looking at every plant in a garden and seeing if it has blue roots if we later discover that half the plants in the garden are not in fact turnip plants - nor is it useful if the plants in the garden have something special due to their existence in the garden (such as that they have always been watered with a blue dye).<>Papineau discusses the question of how we might define 'temperature', arguing that an operationalist account is an oversimplification, and one that simply will not work. He claims that the word 'temperature' is used in a variety of different ways, and that its ambiguities are frequently what we are discussing about a theory involving temperature, and he states that any attempt to define and tie down the word will fail to take into account all the ways in which it is used. I would argue that 'temperature' could be tied down, and that we do need to define it in a particular way in order to get anywhere in a discussion of temperature - perhaps defining it crudely in terms of the change in volume of mercury in a thermometer, or more complexly as the relative levels of agitation of atoms necessary to turn H2O from solid to liquid or from liquid to gas.
In fact, I would argue that it doesn't really matter how we choose to define temperature so long as we are consistent. So long as we are talking about the same thing it won't matter how we define it, since any other equally valid definition will form the same class-boundaries. Papineau argues that we will have to be far too precise in our definition, saying, for example, that 'alcohol boils at 78.3°C at standard atmospheric pressure', but I think that surely we must be this precise, and that this by no means reduces the scope of our definition of temperature. If we define what I shall call temperature1 in terms of alcohol boiling at 78.3°C at standard atmospheric pressure we should discover that temperature2 which is defined in terms of a certain increase in the volume of mercury in a thermometer measures up exactly to it, or at least can be calibrated to do so. Surely this is a more useful as a discovery than simply being assumed by definition as it would be according to the 'double language' model, which holds that a large number of different aspects of temperature can be encapsulated by the word - yet Papineau holds just this as evidence against operationalism, claiming that a term cannot be defined in multiple ways.
The greatest problem for the 'double language' model in my opinion, however, seems to be that the model offers no way of choosing what sort of things we might allow to be part of our conception of a term, and which we would refuse. This leads back to the title of my essay - 'Can you measure temperature with a turnip?' - the 'double language' model seems to answer 'Yes', since it need only be part of our conception of temperature that it is theoretically bound to turnips (and the sort of interaction between the observation and theoretical which is required is not defined - possibly being true to the model itself!) and we are able to say that the workings of a turnip can tell us the prevailing temperature. I do not feel this is simply a facile example - the results of this model seem absurd, at least without a tightening of the interaction which we require. It may be that turnips expand and contract with heat, but an operationalist would most likely discover this as a consequence of temperature, not due to its definition - he would measure the size of turnips at the same time as the expansion of mercury in a thermometer and thus calculate this correlation.
What is claimed to be a greater problem for the operationalist or instrumentalist is that the required distinction between the observational and the theoretical just does not hold up and cannot be formulated in any systematic way. Any 'direct' observation, it is argued, is actually theory-laden, with processes already going on in the background. So when we see a square, we recognise it as a square not because we see it having four equal lines and four right-angles (we rarely do see this, since it only occurs when we look at squares exactly face on), but because we are able to interpret the figure we actually do see, and work out that if measured we would find these factors to be present. If, for example, we hear a sentence, we do not simply interpret at the point of language, but even before we hear the words - we can hear various phoneme-distinctions that the Chinese, for example, cannot (and vice versa), so it becomes very difficult to argue that there is an objective, 'true' thing that is being heard, away from the language model we are working in.
I cannot, however, see why this lack of a distinction should be a particular problem for anyone taking an empiricist view of theoretic terms. Even if instrumentalism holds that theoretical terms should not be construed as true, but simply as useful models, we can take observations that involve theory as useful rather than true. If this is the case then the instrumentalist is still construing theories in just the same way as if they were combinations of 'real' observations and 'useful model' theories by using 'useful model' observations and 'useful model' theories, relying on the Popperian idea that so long as the theory-laden observations are not laden with the theory we are investigating it can be considered to be accurate. It still seems to me to be a perfectly valid instrumentalist position that there are very few (or even no) 'facts', and that everything is simply a useful fiction. All that we need is that at any time when we are asked whether, for example, a turnip is blue we are able to give a simple answer to this, without a dependence on our observations being literally true, and I think the instrumentalist will be happy to make this claim, even if the observation cannot be taken out of context - we are not going to have a problem with someone claiming that the reason turnips look blue is only because that is the model we are working in, if all we want to know is whether turnips look blue.
Feyerabend, for example, argues that science merely aims at consistent sets of theories, and that it will never be able to decide on the basis of evidence that a single set of theories is literally correct, and all the others are merely fictions - there will always be a number of different ways of explaining incomplete data, and it is arguable that even with all the possible data (which clearly will not be possible to amass without a time machine) there will still be any number of equally valid interpretations of that data. Feyerabend is sometimes rather unfairly characterised as proposing 'anything goes' in science, where in fact what he was saying was that we can never choose between different theoretical frameworks and terms which fit the data and fit together equally well, and I think this is a perfectly reasonable thing to assert.
Quine suggested that we could start from any claim which we define to be true and produce a set of theories about that - and that everything else is just a 'web of beliefs' spreading out around a central core. If we were to do this then we could incorporate any theory simply by shifting around different ideas further out in the web, and no amount of experimentation could falsify our central belief. As well as agreeing with Quine's claim, I would assert that when considering theoretical terms we must do this.
Consider gravity - if we take it that all masses attract one another there can be no amount of data or experimentation that could prove this not to be the case. It has recently been suggested that the expansion of the universe is accelerating, yet this should not be possible if gravity is the only force acting at a large distance, since gravity always tries to make things come together, and whilst they can expand due to another force (such as an explosion during the big bang) this expansion should be resisted by gravity, and be slowing down. The response to this, however, is not to abandon the concept of gravity, nor to redefine it as a force that sometimes attracts and sometimes fails to (or repulses), but to suggest that there could be another force, previously unknown, which is counteracting the effects of gravity and pushing things away from one another. All the while, however, we have to have a firm grasp of what we mean by gravity, or we would not be able to say that it appears to be having limited effect. A further point in favour of an instrumentalist or operationalist account would be to question whether gravity could be anything more than the observational effect that it has been defined as. Children frequently wonder what was so impressive about Newton formulating a theory of gravity when they consider it to be so obvious, and it does seem to be in favour of an instrumentalist position that there does not seem to be an easy way to explain why gravity is more than the fact that everything falls when you let it go. What Newton realised was that rather than everything just going downwards, it was a matter of any mass going towards any other, with the level of attraction in proportion to their size and proximity, yet there doesn't seem to be any intrinsic truth about what gravity is except for that it is a coming-together of all masses. As badly defined as our conception of what a law is may be, it seems to make more sense to say that what gravity consists in and should be defined as is an observable effect rather than a whole host of ideas making up a complicated conception.
Overall, therefore, I cannot see why we should refuse to define theoretical terms specifically, and feel that diffferent aspects of a term are more useful as discoveries and extensions resulting from a definition, rather than written into the definition itself. If we decide what we are talking about before we talk about it we will get a good deal further than if we use vague, badly defined terms - which I don't believe can ever be protected from the introduction of rogue definitions; I can't see how we would avoid allowing the measurement of temperature with turnips if we let people use whatever conceptions they might have of temperature to discuss it.