Academics love a definition. Many journal articles, book chapters, even whole books include definitions of various terms. This is sensible because words are slippery. Defining terms helps us to use language with more precision. This makes it more likely that our readers will take from our writing what we had intended it to say, rather than attributing some other meaning to our work.
People attribute meanings to our work which we did not put there because everyone is always influenced by their own past experiences and present emotions. These lead some readers to reach different conclusions from other readers about the same piece of writing. Even though they have read the same words in the same order, they have had different experiences.
Also, human memory is notoriously fallible. Think of something you read a few months ago. How much of it can you remember? Unless you are one of the rare people with a photographic or eidetic memory, I bet the answer is ‘not much’. This means that even when we define our terms, the impact of that definition will fade with time.
This has happened even to key research terms such as ‘statistical significance’. That term was originally intended to indicate how likely or unlikely a result was to have occurred by chance. However, lay people might read a newspaper article reporting that a research finding is ‘statistically significant’ and attribute the everyday meaning of ‘significance’, i.e. important or meaningful, to the findings. Indeed the journalist who wrote the article may have made the same attribution. Yet statistical significance was never intended to imply that a result was important or meaningful in everyday terms.
The statisticians who devised tests of statistical significance were careful to define their terms. Unfortunately the care they took was diluted over time, and the conflation between the use of ‘significance’ as both a technical and an everyday term caused a multitude of errors and conflicts, ultimately leading to mass calls for its retirement as a technical term.
There are examples of this from other fields too. Emergency medical dispatchers in the US used to ask callers whether the patient was alert. This caused confusion and delay, which is not what anyone wants in a crisis. Enquiries revealed that ‘alert’ has a specific clinical meaning which is not understood in the same way by members of the public. Now dispatchers ask callers whether the patient is responding normally, which is much easier for most people to answer and still tells the dispatcher what they need to know.
There is a symbiotic relationship between language and thought. Language helps us to think; many of us think in language, at least some of the time; the language we use, hear, and read influences the thoughts we have. When we need to name something, such as a new research method we have devised, it is tempting to reach for a name with pizazz, a name that will be eye-catching and memorable. Perhaps those early statisticians chose ‘statistical significance’ for that very reason. However, experience shows that it makes more sense to choose a term which offers a description that is as simple, clear, and accurate as possible. Even then there are no guarantees that everyone will understand – but at least we have given it our best shot.














