The world is models all the way down. You can’t make sense of anything without models. Your entire perception of the world is models. Models, models, models, all the way down.

Some magazines have little games in which a picture of an everyday object is shown from an unusual angle. At first, the picture is simply an arrangement of shapes and shades. Then there is the a-ha! moment when you work out what it is and suddenly that assemblage of lines and areas turns into something. Photons are still bouncing off the page and streaming into your eyes in exactly the same way as before, but what you see has changed utterly. The model you have of that object fills in the gaps.

Our brains are spectacularly good at this gap filling, at imposing models on the world. Sometimes the model is so strong it can overwhelm what’s actually happening. I experience this most vividly when trying to walk up a stationary escalator. The feeling that it ought to be moving makes a stomach lurching mismatch with what the eye is seeing.

Even the simple act of looking out into the everyday world is possible only because of models. In a sense the only what there can be an everyday world is because of models.

When we look at the world in more complex ways, for example, using scientific instruments, then models become even more important, although it can be tempting to see measurements and observations as something which we use to test or validate models and which must therefore somehow be model free. That’s not to say that there isn’t anything more than models, because crucially a measurement is sort of where all our theories eventually crash into the real world, which is famously unkind to models and theories of all sorts.


Take temperature. Everyone has a sense of warmer and colder, but each person feels the heat in a different way. In order to impose some objectivity on this, we use thermometers.

Most household thermometers work (or used to, the world’s getting fancier) on the principle that certain liquids expand when they get warmer. This principle can be used to convert a change in volume of the liquid into a change in the length of a column of liquid, which can then be used to read the temperature off a calibrated scale. The model here looks like a simple one, but there is a model.

Crucially, how we measure temperature depends very much on what we think temperature is. We know that the fancy new thermometers which you point like a ray gun at your bubbling pan full of stew measures the same thing as the old-fashioned thermometer you actually have to stick in the stew. We know this because we have some underlying theory of temperature.

That’s all a bit theoretical, albeit essential, but the model of a bog-standard liquid in glass thermometer, which is already theoretically complex, needs to be somewhat more practically complex too if we are to understand the measurements we get out of it.

For example, what does a mercury thermometer measure the temperature of? Answer: itself. To register a change of temperature, the temperature of the mercury in the bulb has to change. A cold thermometer plunged into a hot stew will quickly warm up to the temperature of the stew. But if the sample is small the thermometer will change the temperature of the sample as much as the sample changes the temperature of the thermometer. In situations where the temperature changes rapidly, the thermometer never quite catches up with the thing you are trying to measure. We also need to know that the thermometer is calibrated so that we read the temperature off from the top of the meniscus and that the small separation of the scale from the pin-thin mercury column means that you have to get your eye at the right level to ensure that you read off the right value.

Then we need to ask if there is anything else that might affect the measurement. Take a simple air temperature measurement. When taking an air temperature measurement, we need to be sure that the thermometer is not in direct sunlight because the sun can heat the thermometer up above the temperature of the air. One way to do that is to build an enclosure for the thermometer. That then leads to the problem that the air is no longer quite so free to circulate past the thermometer bulb and the air trapped in the enclosure might not be the same temperature as the air just outside. So, we puts angled slats on the sides of the enclosure and we paint it white. And so on…

Understanding a temperature measurement (and understanding why a measurement goes wrong when it goes wrong) relies on all these models. Understanding those models, requires more models, and those models are… you get the idea.

Even for something as simple as a temperature measurement it’s models, it’s models all the way down.