What is it that makes something "science"?

In ancient Roman times, people thought that horse hairs that fell on the ground turned into earthworms when it rained. They based this on some observations:

• horses were common, and shed hairs

• earthworms are found on the ground after rain

• some things swell up when they absorb water (e.g. dried fruit, which the Romans knew about)

• horse hairs might look like worms if they swelled up and absorbed water when they were wet.

Of course, we know they were wrong. The prediction 'horse hairs turn into worms when wet' hadn't been tested. Neither did they think of some logical flaws: you can find worms in places there are no horses; horse coats don't turn into a wriggling mass of worms when they are wet.

The Romans were doing something all humans do: trying to understand and make predictions about the world around them. However, their method was not scientific.

Modern Science

Modern science has come about as a way to get the best answers about the way we understand the world, using observations, experiments and logic.

Scientific theories and ideas have the following features:

• they are based on real (and repeated) observations, not just stories which haven't been checked out

• they makes useful predictions (not predictions which could never be observed)

• those predictions must be able to be tested in some way, through more observations or experiments

• they are supported by lots of different pieces of evidence, not just one

• they are consistent with what is already known and understood

• the theories can be modified to fit with new observations if they occur

• the theory doesn’t invent any more 'new ideas' than is absolutely necessary (e.g. it doesn’t have to keep on inventing new things to explain every little observation that happens)

For example, some people believe in alien abductions.

They believe aliens in UFOs abduct people and do horrible medical examinations which injure the abductee. When the believers are asked:

"Why is there no medical evidence for the personal examinations of any of the people who claim to have been abducted ?"

they say:

"Aliens have advanced technology which can remove this medical evidence before the abductees are returned."

The UFO believer has invented a new idea (advanced alien technology) to explain their lack of evidence. This fails the last bullet point - that new ideas shouldn't be put forward unless absolutely needed. Putting forward a new idea to explain a lack of evidence isn't needed, because the lack of evidence is itself a pointer to the original idea being wrong. Neither do the believers have a logical explanation for why 'aliens' would want to abduct people, or why they would torture then heal people and so on, "Alien abduction theory" is not considered to be science for all these and more reasons.

The Scientific Method

The scientific method is a way of investigating and asking questions to get the answer that is most likely to be correct.

The steps in this method can vary a bit depending on whom you ask, but generally they are something like this:

• Decide on the question you are going to ask. Not all questions can be answered by a scientific investigation (e.g. "do angels exist?" ), or they can't be answered at the moment because we don't have a way to investigate (e.g. "do any other planets in the Universe have chocolate?").

• Make a hypothesis. A hypothesis is a 'best guess' about what the answer might be. It should be reasonable. For example, the hypothesis that the Moon is made of cheese is not reasonable.

• Design an investigation. This could be an experiment, but it doesn't have to be one. Other ways to investigate could involve making observations of the natural world e.g. observing animal behaviour in the wild, or observing clouds or collecting rocks and so on. Investigations could involve making measurements, or doing a survey, or using data collected by a survey e.g. data about light from thousands of stars, collected by observatories and available online. Data is a term used to describe a set of observations from a series of experiments or survey.

• Repeat the investigation. Any experiment should be able to be repeated to get similar results; any observation should be able to be made multiple times to be reliable.

• Make some conclusions. This has to relate to your hypothesis: it might be right, it might be wrong, or you might not be able to tell one way or another. If possible, a conclusion should be quantitative, which means you should be able to give a quantity e.g. "each extra gram of mass in the impactor made the crater 5 mm bigger". However, this is often not possible.

• Evaluate what you have done. Were the results repeatable and consistent? Did they show a clear pattern or do they appear to be just random? Do they fit with existing scientific knowledge, and how? Are the observations of good quality? Are there 'unanswered questions' raised by your investigation? Are there limits on any predictions that could be made?

There are several types of observations. Two important categories are

• Quantitative observations, which can be measured and given a number. For example, saying that 'the mouse had a mass of 18 grams' is quantitative.

• Qualitative observations. for example 'this chemical smells of almonds'. These can be described but are (usually) difficult to measure.

Some observations, such as colour, can be measured with special instruments but require a lot of preparation and special conditions . To measure colour, for example, you need to standardise the colour of the light used to illuminate the object before you start.

You could infer how far away the earthquake is from the seismic observations, because different earthquake waves travel at different speeds so the time-lag between them depends on the distance. Inferences are made using things that are known to be highly likely, and are therefore likely to be true.

In this case, you can't directly observe the distance to the earthquake but you can still work it out. That is what makes it an inference.

If you are working something out from an observation using ideas that are less certain, it would be a hypothesis. For example, if the earthquake I detected was 150 km under Taupo, I could make a hypothesis that it had occurred on the subducting Pacific Plate.

• the variable you (the experimenter) are changing to find out what is going to happen. This is called the independent variable. In this case, it would be the temperature of the water you were dissolving the tablet into.

• the variable you observe or measure to see what is going to happen. This is called the dependent variable. In this case you would be observing how long it takes for the tablet to dissolve.

• things you have to keep the same to make the comparison fair are called controlled variables. In this case, you would want to make sure that you used the same sized tablet (because a bigger one might take more time to dissolve) and the same amount of water (because using more or less water might change things). You would probably also want to use the same size and shape container to hold the water and so on.

Data

Data is the results of an experiment, usually numbers. Since you do the experiment multiple times to see if your answers are consistent, you can wind up with an lot of numbers to deal with .

For example, the following data is from an experiment melting an ice cube in a cup of water and measuring the temperature of the water: Several students have collected the data for different trials on different bits of paper (this is not a good idea because they could easily get lost):

The students have made a few other mistakes as well.

• They don't have a consistent way of writing their results down,

• they haven't labelled which numbers are the minutes and which is the temperature (mostly). We can guess, in this case - but there are other experiments where guessing would be far too hard.

• One of the trials is labelled "Fifth trial" but there only appear to be four trials; this is confusing.

A better way would be to put the results into a single table prepared in advance. When you plan out your experiments, you should try to draw up such a table in advance if you can. Below is an example

Here we have assumed the results of the 'fifth trial' are really the 4th. There are still a few problems.

We need somehow to show the pattern of results. To do this, the best way is a graph. But which set of results should be graphed?

The answer is all of them, but not as separate points. Instead we average the results. To do this, we add them together and divide but the number of results.

• Be careful you average the correct direction. We want to average all the 1 minute temperatures for four trials, not all the temperatures of Trial 1.

The average of the one minute temperatures is: (17 + 18 + 17 + 17) divided by four = 17.25 degrees.

However, you might not want to graph to this many decimal places and might round this to 17 degrees.

We have a problem with the 5 minute result for Trial 4: should we include it in the average? The answer is usually no, as it appears to be a mistake. However, it should be checked out in case it isn't a mistake. A result like this is called an outlier and outliers are not included when we average results for analysis.

Below is the table of averaged data. We call this the processed data.

Now we can graph the results. Some graph rules for Science:

• Science graphs are mostly line graphs - one quantitative variable against another.

• A bar graph or pie graph could be used if one of the data sets is not quantitative

• Science graphs should have a title explaining what the graph is about

• Usually, the independent variable goes along the horizontal axis, which is called the x-axis

• Both axes must be labelled and with units

• not all graphs have to start at zero, but graphs should usually not have a 'break' in them

• the numbers on the axes have to go up in even amounts

• the data points are best plotted as a small cross (x)

• a smooth line or curve should show the trend of your results.

A graph of results for the experiment above might look something like what is shown below:

Some important points about this graph

• the title says what the graph is about

• the label on the axes says what quantity is being measured; it must have units if they exist

• the points are plotted at the co-ordinates given in your summary table

• a graph should start at zero if the number 0 is important for the graph; otherwise either axis can start and finish where convenient

• the scale of the axes should be big enough to show a pattern and not leave too much unused space on either axis

• numbers on the scale have to go up in regular jumps ( e.g. 3, 6, 9)

• the trend line should be a straight line or a smooth curve that comes as close to the points as possible. Ideally, equal numbers of points should be on either side of the line. This is because it is assumed that the fact they aren't exactly on the line is caused by random errors in the experiment. Do NOT just join the points up..

Below is a copy of the graph above on which I have annotated the bullet points above:

When you write up a method to do an experiment, you have to say what the aim of the experiment is, what the variables are, how you are going to control the variables and how you are going to measure or observe your results. Below is an example:

A point to note above that many students forget: I converted 1 minute 8 seconds to 68 seconds to graph. I wrote the header for the time in the original table as minutes.seconds because there is no standard metric way of writing that in short form.