Why Group Words by Meaning Instead of Alphabet? A Map Rather Than a List

A dictionary starts from a word and supplies its meaning, which helps somebody who has encountered a word and does not know it. A writer has the opposite problem, starting from a meaning and needing the word, and no alphabetical arrangement helps with that, since the word being sought is precisely what is unknown. Grouping words by the ideas they express solves it, allowing a search to begin from an approximate word and arrive at a set of related ones among which the right one may be recognised. That is a genuinely different tool serving a different need, which is why the two kinds of book coexist rather than one replacing the other.

Peter Mark Roget published his in 1852 after decades of private work:

• •He was a physician and scientist who compiled lists as a personal habit
• •The arrangement divides all concepts into six classes
• •Those subdivide through sections into about a thousand numbered heads
• •Opposites are placed side by side, so ideas and their contraries sit together
• •Words appear under every head where they apply rather than in one place
• •An alphabetical index points into the classification, and was added later

The structure was not a convenience but an argument, since Roget was attempting a systematic classification of ideas in the tradition of natural history, the way a botanist classifies plants. That ambition explains features that look eccentric, including the placement of opposites together, which reflects a view that concepts come in contrasting pairs, and the numbered hierarchical arrangement, which asserts that ideas have a natural order. Later editions have retained the framework while adding enormously to the contents. Most modern books sold under the name abandon the classification entirely and list words alphabetically with synonyms beneath, which is easier to use and discards exactly the structure that made the original interesting.

Modern reference works of this kind are increasingly built from evidence about how words actually behave rather than from an editor's judgement about what means what. Enormous searchable collections of real text allow the words that habitually occur together to be identified statistically, so a writer can see that one candidate is normally used of people and another of institutions, or that one carries a negative colouring in almost every recorded use. Collocation dictionaries present exactly that information and are considerably more useful to a learner than a list of synonyms. Distributional methods in language technology take the same idea further, representing a word by the company it keeps, which is the principle behind the systems that now suggest words as people type.

The tool is notoriously dangerous in inexperienced hands and the reason is specific. Words grouped under one heading are related in meaning and are almost never interchangeable, differing in register, in connotation, in the company they keep and in the grammatical constructions they fit. Substituting an unfamiliar word for a familiar one because both appear under the same head produces prose that is wrong in ways the writer cannot see, and the results are recognisable to any reader and to any teacher marking essays. The correct use is to recover a word already known but not recalled, or to find a candidate and then check it in a dictionary and in real examples before using it.

A dictionary goes from word to meaning while this goes from meaning to word, which is the opposite problem and is what a writer actually has. Roget published his in 1852 as a systematic classification of ideas in six classes and about a thousand heads, placing opposites together. Most modern versions discard the classification for an alphabetical list. Words under one head differ in register and connotation and are rarely interchangeable.