We frequently hear expressions such as `scattered showers' and `sunny intervals' on weather forecasts, yet few of us realise that such terms are precisely defined. `Cold' in winter is probably colder than `very cold' in summer ...

Wittgenstein would have been fascinated by the weather forecast. It provides a fine test-ground for both his theories of the philosophy of language.

According to Ludwig Wittgenstein's first theory, as laid out in his Tractatus Logico-Philosophicus in 1922, a sentence can be broken down into "atomic propositions" whose elements stand for things in the real world. Social convention gives words their meaning, and we all agree that the sun is the sun, and warm is warm. These are real-world concepts to which we can all relate.

Later, however, he rejected this "picture theory" approach to language, replacing it with one of "language games" in which usage, rather than convention, took the dominant role. Under the new theory, words acquire meanings through their accumulated usage in different contexts. We acquire a feel for the difference between "drizzle" and "rain" not through dictionary definitions, but through our experience of hearing those words used in poor weather conditions.

So where does the weather forecast fit into all this? Well, the language game played between forecaster and listener is a perfect controlled experiment to test the merits of both Wittgenstein's theories. For in general, the forecasters are playing by the rules of his first theory, while the listener is stuck with the rules of the later one.

When forecasters use terms such as "scattered showers" and "sunny intervals", they know the precise definitions of those terms. They know whether "sunny intervals" promise more or less sunshine than "sunny periods", and they even know exactly how long a "shower" has to last before it becomes "rain". But in the true fashion of well-conducted research, they never seem to have told the experimental subjects - the listeners - what these definitions are.

So we are left in the position of Wittgenstein's second theory: we have to deduce the meanings of words and phrases from usage. This is a language game in which the experimenters know the rules, and we do not. If we can work them out, it would support the first theory: words have precise meanings relating to things in the physical world. If we cannot do so, and all end up with different vague impressions of the meaning of a particular expression, then the second theory would be validated.

All the evidence - in my experience, at any rate - comes down heavily in favour of the second theory. Just pause a moment and think about "sunny periods" and "sunny intervals". What's the difference? Then turn your mind to "scattered showers" and ask yourself whether they are scattered in time or place, and how scattered they have to be. Then think about the current "mild" weather we've been having (apart from the occasional storm or tornado) - and ask yourself what the difference is between "mild" and "very mild".

You will find all the answers below in the Weather Wise column. Quite apart from validating Wittgenstein, they may also help you to understand precisely what the weather forecasters are trying to tell us.