How is it that words – be they lines on a page or vocal utterances – come to have their meanings? To know this is to better know how to think clearly. Those who are ignorant of the language in which their own thoughts are worded, can commit errors in those same thoughts. How can we avoid said pitfalls? Semantics is one of the hardest parts of philosophy, and agreeing on the words for things can be fraught. After all, language is a social process; for a word’s definition to have authority, there needs to be a consensus about it between speakers.
Just to complicate things further, many words have multiple definitions in parallel use at any point in time, and in academic philosophy this can be especially problematic, where a word such as “realism” can have many different meanings, leading to information overload, ambiguity, and confusion. Socrates probably summarised the idea a few thousand years ago.
The beginning of wisdom is the definition of terms.
Definitions are relative; there will often be people who dissent and refuse to recognise a word’s authority. As individuals, we might accept a statement such as “a man is an adult male human”, and in doing so, we are choosing to regard that as being the case. In this way, definitions only become true to us when we adopt them, so no definition is objectively correct. Alas, “truth is correspondence with experience”, and words only correspond to the world when we decide so. Humans impose meaning upon the world, “carving nature at the joints”, in the same arbitrary way we impose nations and state boundaries upon continuous expanses of land and sea.
And yet despite being artificial constructs, some definitions are more useful than others, allowing us to communicate and think more clearly. When Galileo defined the solar system as having the sun at its centre, he significantly simplified the mathematics of planetary motion. And rather than dividing it into random pieces, we think of a train as being composed of a locomotive and its carriages, for they work as individual components.
Here I want to present a useful categorisation of definitions. Along two different axes, I see four major types: those that refer to either empirical or theoretical entities, and those with either descriptive or prescriptive motivations. It is important to note that all our abstractions are artificial; not only including definitions, but also categorisations such as those in this essay. I could have claimed that the two major types of definition are those made in the northern hemisphere and those made in the southern hemisphere, or those that include the word “ant” and those that don’t, but that would have been pointless. On the other hand, a consideration of a definition’s reference and motivation illuminates how and why we use them. Hopefully this distinction will assist those who have not previously considered such questions.
Definitions by Referent
1. Empirical referents
Any situation in which a physical, observable thing is given a label is an example of what we might call an empirical definition. We can give stuff names; either concrete or abstract ones. We might have a pet dog to which we assign the proper noun “Bob”, to indicate his concrete individuality. Equally, we could give him the common noun “dog” to indicate that he belongs to an abstract category of similar entities. Both are valid ways of defining observable stuff.
We can also provide other people with our empirical definitions; this is done through ostensive definition – “definition by pointing”.  This is where a physical object is used as a way of conveying the meaning of a word to another person. For instance, in teaching a small child to recognise individual entities, we might point to our dog and say, “Bob”. Or, to teach a more abstract concept, we might show them a picture of a few different canines, saying “dog” for each. This ensures they grasp the common elements; that all dogs have a certain look, for example. Ostensive definition is thus an important teaching tool in the process of socialisation.
2. Theoretical referents
When we define a term in writing or speech, it will usually be made up of various combinations of other words; we might call this a theoretical definition.  These can also be either concrete or abstract in form, like their empirical relatives. I might say that “Bob is a pet dog”, and the rules of grammar – themselves learnt similarly to vocabulary itself – would indicate that he is a concrete individual. On the other hand, I could say that “a dog is a domesticated canine”, using grammar to indicate that we are talking about an abstract class of things with common properties. Of course, a definition can never be purely theoretical; concepts must ultimately refer to the world, or we would not be able to comprehend them. For this reason, circular definitions in dictionaries are inescapable. A written work must assume in the reader a prior acquaintance with at least some of its words.
It should be noted that fictional concepts in particular are always defined theoretically, because by definition they cannot be observed in reality. This happens, for example, when we claim that “a unicorn is like a horse, but with a horn on its head“. We can use language and thought to explore such possibilities, imagining things which do not exist. Such imagination imbues us with great creative power; to consider different recombinations of things is an excellent strategy for invention and discovery. This is why Leonardo Da Vinci’s 15th-century sketches of flying machines symbolise much of the history of aviation; they represent a completely new synthesis, for which there was once not even a word.
Sometimes a theoretical definition can invoke causes and effects as part of its formula. This is known as an operational definition, where a word is defined as the result of an experimental or procedural “recipe”.  Operational definitions are particularly useful where phenomena are difficult to observe without certain conditions being met. For example, we might say that “anger is the result of dashed expectations”; anger is not the usual mental state for most people, so defining it as something which is only produced when expectations are dashed might be helpful. Most operational definitions are usually abstract rather than concrete; it is rare to find a situation in which it would be desirable to create an operational definition for an individual object.
Another type of theoretical definition is the recursive definition, an idea which is mainly applicable to computer science.  These are conceptually difficult compared with other definitions, as they reference themselves. There is usually a condition included to end the self-reference at a certain point, so as to avoid an infinite loop of regression. For example, we might say “an ancestor is either a parent or the parent of another ancestor” though in English this is a tedious way to define such a word.
Definitions by Motivation
1. Descriptive motives
A person who writes a dictionary is called a lexicographer, and when they are doing such a job they are aiming to describe – as accurately as possible – how a word is actually used using lexical definition.  Often this will involve listing multiple senses of the word which might be meant, depending on its context. Whilst there is always some degree of interpretation involved in this task, generally there is a correct and an incorrect answer to what a lexical definition should be. If a definition fails to describe how a word is really used, it is lexically incorrect. Of course, we can ask about how widespread the usage must be for the definition to be worthy of inclusion in a dictionary, and there is no definite threshold for this. But suffice it to say that the word has to be understood by the average speaker.
Words can have many different senses to them, depending on the context in which they are used. Sometimes we might want to clarify the meaning of one of these word-senses; we might call this a philosophical definition. The difference from lexicography is that we are not aiming to note all possible senses of a word, but rather to understand how people use a particular one. For example, “What is the meaning of life?” is a question of philosophical definition, for it is essentially a word game, where the concept of “meaning” – in the sense and context it is used here – requires careful consideration. Dictionaries often include vague, circular or otherwise unhelpful word-sense definitions, such as “meaning: 1. the significance of something” that leave much to be desired in terms of explanatory depth. But little more can be expected, for such detail is too much work for even a team of dictionary writers.
Compared to lexicography, philosophical definition focuses more on testing the conceptual implications of a potential definition against both personal intuition and broader social acceptance and consensus. However, it should still be properly understood as a part of lexicography, because well-researched word sense definitions can be integrated into dictionaries. To use another example, the question “what is causation?” has been asked many times through history, and throughout this time, dictionaries have stated such wishy-washy definitions as “causation: 1. the act of producing an effect“. David Hume tried to improve on these when he claimed that causation is a state of “constant conjunction” between entities. Others found this to be insufficient, for despite their constant conjunction, we do not say that a rooster causes the sun to rise. There are now a few different theories which attempt to distinguish causation from mere correlation, and if one of these were to accord with intuition and gain consensus, the new definition would surely make it into dictionaries.
2. Prescriptive motives
Languages are constantly evolving in an ongoing process of word creation, modification and deprecation. Whenever we as individuals change our idiolects – our own personal usage of language – we take part in that process, by offering a new way to use said language. If others accept the changes and start using the new word themselves, it can quite rapidly spread throughout entire communities.
Occasionally it is useful to coin a new term, or to impose a single, clear definition upon a pre-existing word as a premise for argument. Either way, what results is known as a stipulative definition.  These can add new words (so-called “neologisms”) to our idiolect and are quite common at the forefront of scientific research; where new phenomena are discovered, people invariably give names to them. The concept of quantum physics did not exist before Einstein and Bohr, for example, but once the mechanics of subatomic particles were determined sufficiently, it became desirable to label the new theory. Stipulative definitions have a significant role in philosophy too; for example when Auguste Compte coined the word “altruisme”, he gave a label to a concept that has since become a common way to frame diverse theories, from ethics to evolutionary biology.
When there is a desire to stipulate a new term, doing some research to see if a word already exists for the concept will prevent “reinventing the wheel”. Indeed that is what I have done in this essay; I got many of these terms from Wikipedia, as the citations indicate, which got them in turn from other sources. However I did need to create two new terms, “empirical definition” and “philosophical definition“, due to my finding no existing concepts. If – like here – a new term does need constructing after all, then making it as easily memorable as possible will help it stick with people and get used more often. This is why existing words are often built from prefabricated building blocks; so-called morphemes.  The same applies to the construction of compound terms such as “empirical definition“above; its components already existed. “Neologism” is itself also an example, built using the prefix “neo”, base morpheme “logos”, and suffix “ism”. The end result can be interpreted as something like “new utterance” to anybody with a basic knowledge of English morphemes.
When we want to redefine a pre-existing term to suit our application, we are using a precising definition; a specific type of stipulative definition.  This is done without the intention of accurately describing word usage. Precising definitions are common in government and business; for example a transport department might claim that “for the purposes of driving in Australia, “intoxicated” means having a blood alcohol content at or above 0.05 percent.” The definitions that such claims give rise to are known as jargon and can serve to isolate the community that uses them from the broader language community, as everyday people often use the words differently.
Occasionally people define concepts to incorporate their moral biases through what is known as an emotive definition; sometimes also called a persuasive definition.  These are characterised by a lack of neutrality regarding the subject of the definition and often result from the use of morally charged words. A person might claim that “abortion is the slaughter of fetuses”, but such a definition obstructs rational argument by starting with loaded terms. If we want to think clearly, it is best to avoid these definitions entirely and instead use unbiased terminology.
There are situations where people want to be deliberately offensive or imprecise, but these are exceptions. Usually people want to be understood clearly and politely, and if they believe a term is ambiguous, loaded, or associated with negative connotations, they may replace it with a more precise term or phrase. For example, I have seen debaters, when asked whether they are feminists, retort with “that is a loaded question; I believe in equality for women”. Similarly, many people will avoid the use of the word “bitch”, because despite a harmless original meaning, it is now usually offensive. If people believe a term is likely to bring up unpleasant thoughts, people will usually replace it with a euphemism. For example, people are less likely to say “I’m going to defecate” and more likely to say “I’m going to the bathroom”. And if new information makes people realise that a word refers to a fictitious entity, it can fall out of favour. For example, after the Michelson-Morley experiment proved that light does not travel through an “aether”, the concept was no longer taken seriously.
What can we take from all this? Whether our definitions refer to phenomena or to other words; whether they are intended to describe the language or to prescribe it; a single requirement determines whether they enter common parlance: they must be useful for communication. They must be able to convey those impressions we wish to share. We all play a role in shaping our language, and while no definition is objectively correct, some are more useful than others, and as discussed, those are the definitions that persist. Keeping these ideas in mind should empower us to use our words more strategically and effectively.