Communication lexicon with do and don't words: meaning and application

Different teams often choose different words: one writes 'contribution', another 'costs' and a third 'deduction'. Each of those words can evoke something different in your target group. A communication lexicon records which words you use deliberately, which you avoid and why, so that the choice is no longer left to chance.

A communication lexicon is a substantiated word list per target group or theme. It records which words you use deliberately (do words), which you avoid (don't words) and which alternative you choose, with the reason and origin of each choice.

Neurofactor uses the lexicon as a working tool: every choice refers to researched associations of the target group. Framing research shows that word choice can shift judgements, but those effects differ by context and by group. A lexicon is therefore a set of testable hypotheses, not a list of words that always work.

A communication lexicon records, per target group or theme, the words that matter for the choice you want to support. In this article it is a Neurofactor working definition. Practice has comparable tools, such as style guides with dos and don'ts. We have not found an established scientific origin for the term 'communication lexicon'.

The lexicon has three types of entries:

• Do words: words that evoke the intended meaning in the target group and that you use deliberately and consistently.
• Don't words: words you avoid because they evoke an unwanted or confusing association, always with a preferred alternative.
• Caution words: words you cannot do without, such as fixed terms from policy conditions, but that need a standard explanation.

Unlike an ordinary style guide, a lexicon asks for each word: what does it evoke in this target group, in this choice context, and how do we know?

Logically equivalent wordings can lead to different choices. In the well-known experiment by Tversky and Kahneman (1981), participants chose between two programmes against a disease. When the certain option was described as '200 people will be saved', 72 per cent chose it (N = 152). When the same outcome was described as '400 people will die', 22 per cent chose it (N = 155). According to the authors, the frame depends partly on the wording and partly on the decision-maker's norms, habits and characteristics.

Closer to word choice is attribute framing. In a study by Levin and Gaeth (1988), consumers rated ground beef more favourably when it was labelled '75% lean' than '25% fat'. The effect shrank after tasting. Words therefore weigh most when people have little other information.

A meta-analysis by Kühberger (1998) of 136 studies found a small to moderate average effect, strongly dependent on the research design.

The linguist Lakoff (2010) argues that words evoke frames and that negating a frame actually activates it. That is a theoretical position, not an experiment. We do not adopt his description of frames as brain circuits, which a word list cannot establish.

Schuldt, Konrath and Schwarz (2011) asked an American survey sample (N = 2,267) about either 'global warming' or 'climate change'. Among respondents who identified as Republicans, 44.0 per cent considered the phenomenon real under 'global warming' and 60.2 per cent under 'climate change'. Among Democratic respondents, the term made hardly any difference.

Villar and Krosnick (2011) found no difference in perceived seriousness between the two terms in the full American samples, but opposite patterns by party. In a European survey across 31 countries, the term made no difference either.

The lesson for a lexicon: a word effect can occur in one group, not in another and sometimes even reverse. Work per target group for that reason.

What a word evokes depends on its connections for a person. An association is such a connection between a word or concept and other ideas, feelings or experiences. In a network model, each concept is a node that can evoke other concepts through its connections.

The Small World of Words project collected several associations per participant for more than 12,000 English cue words, from 83,864 participants after quality screening (De Deyne et al., 2019). A Dutch dataset also exists. Such norms show what a word evokes on average in a broad population. For a lexicon, that is a starting point: your target group may respond differently in a specific choice context, so research the key words with that group. Read how to measure associations for the methods available.

Include only words that matter for the target group's choice, with the same fields for each word.

A rule without substantiation is a matter of taste; mark it 'still to be researched'.

A lexicon touches on other tools but answers its own question.

The jargon section of a target group profile describes which words the target group uses; the lexicon decides which words you use, often with jargon as input.

Managing fixed terms is a distinct field in information science. The American standard ANSI/NISO Z39.19 provides guidelines for the content, display, construction, testing, maintenance and management of monolingual controlled vocabularies. These serve mainly to retrieve information, not to persuade a target group.

For a communication lexicon, we take from it the idea that a word list is built, tested, maintained and managed. As editorial advice, that means:

• choose one preferred word per meaning and record common synonyms with a reference to that word;
• appoint an owner who approves changes;
• keep a change log, so that you know later why a word changed status.

1. Define target group, theme and choice context. A lexicon for new customers making a first purchase differs from one for existing customers renewing a contract.
2. Collect the words currently in use. Check your own texts, customer questions and conversation notes, and note the target group's own words.
3. Select the key words. Choose the words that most shape how the choice is understood. A list that is too long is rarely applied consistently.
4. Research the associations. Use, for example, a free association task or interviews with the target group, not only your own team.
5. Set status and alternative. Give each word a status, a preferred alternative and the reason, with a reference to the source.
6. Record the testing. Mark each choice as a hypothesis and note how and when you will test it.

Test a choice by comparing text variants that differ only in that word. Decide in advance which outcome counts, such as comprehension, the associations the text evokes or the number of follow-up questions. Analyse results per target group: an average null effect can hide opposite effects in subgroups. From target group insights to testable messages explains how to set up such a test.

Review the lexicon periodically and when there is a reason, such as new customer questions, a changed product or new regulation. A don't word that no longer causes problems may move back to 'caution' or disappear; record the change with reason and date.

This example is fictional. No results are known.

Situation. A health insurer receives many questions after policyholders get a bill for their deductible (Dutch: 'eigen risico'). Its texts mix 'deductible', 'offset', 'outstanding amount' and 'pay yourself'.

Choice question. Which words does the insurer use around the deductible for policyholders incurring healthcare costs for the first time, and which for those who do so every year?

Available information. Existing texts, customer questions and the words policyholders themselves use in them.

Approach. The team selects five key words and asks both groups for free associations, supplemented by short interviews. 'Deductible' is a fixed term in the policy conditions and cannot be dropped. It therefore receives the status 'caution', with a standard explanation such as 'the amount you pay yourself first'.

Possible interpretation. If the first group mainly links 'offset' to an unexpected deduction, it becomes a candidate don't word, with 'we will send you a bill' as the alternative. If it raises no objection in the second group, the choice applies only to the first group.

Next step. The team tests two letter variants on comprehension and follow-up questions, per group. Only then does the status change from 'hypothesis' to 'tested'.

• No magic words. Framing effects are small to moderate on average and depend on context and design. A word that worked in one study does not automatically work for your target group.
• Experience counts. In Levin and Gaeth's study, the word effect became smaller after tasting. A better word does not fix a poor experience.
• No brain circuit per word. Words do not automatically activate a specific brain circuit; a lexicon makes no claims about that.

Common mistakes are:

• using one lexicon for all target groups;
• don't words without an alternative;
• repeating a don't word in a negation, such as 'no hidden costs';
• using general association norms as if they were target group insights;
• never updating the lexicon.

A communication lexicon makes word choice deliberate, consistent and verifiable, based on researched associations. Framing research shows that word choice can matter, but also that effects differ by context and group. Treat every entry as a hypothesis: test it, and update the lexicon when the target group or context changes.