Semantic collision matrix

Type: Comparative view
Status: Active derived view
Canonical for: Cross-system lexical collisions within the controlled Button sample
Depends on: Variant naming matrix, linked first-party sources, and evidence boundaries

Counts below are counts within the 15-system Button sample, not market prevalence. A term is counted only when exposed in the public mechanism sampled. The final column is an inference from observed semantic conflict, not a measured agent-error rating.

TermSample frequencyDocumented meanings observedAgreementHypothesized prior-conflict condition; not measured
primary8highest emphasis; preferred action; filled/accent treatment; category of emphasismoderate, relationalconflict is plausible when a local system assigns a domain-specific meaning
secondary8lower emphasis; secondary call to action; outline-like treatment; USWDS secondary colorconflictobserved cross-system meanings create a direct ambiguity to test
tertiary4lowest/lower emphasis; borderless/minimal treatmentmoderateinterpretation depends on relation and layout scope
default3ordinary action; default appearance; non-confirm/non-danger action typeweak/conflictingabsence of an override does not communicate purpose
neutral1 Button sampleSalesforce default white treatmentinsufficient sample; feedback systems use “no status”sample is insufficient for a directional hypothesis
brand1Salesforce brand treatmentinsufficient; generally presentation/identitymay cue identity or presentation rather than use
subtle2low visual prominence/minimal chromefairly stable appearance/emphasisappearance may be interpretable while selection remains unstated
ghost2chrome-free/minimal visual treatmentstable appearanceappearance is more explicit than decision purpose
danger5destructive or high-consequence action; sometimes also most prominent treatmentstable consequence, variable prominenceconsequence and prominence may still be conflated
destructive1destructive treatment/action, plus text formstable locallyseverity and confirmation stage remain unstated
success2positive/confirmation color or action stylingbroad tone, not functional roletone does not establish functional role

Collision details

secondary

  • Carbon uses it relationally in flows: the forward action is primary, Back secondary and Cancel ghost. Carbon
  • GOV.UK says it is for secondary calls to action. GOV.UK
  • PatternFly explicitly says a secondary button does not have to be second; the term describes visual hierarchy. PatternFly
  • USWDS defines the modifier as a secondary-color button. USWDS
  • Spectrum describes its variant axis as visual emphasis. Spectrum design data

Documented: the same lexical value denotes at least relative prominence and palette choice. An uninformed agent cannot recover the local meaning from the word alone.

danger / destructive

Consequence is more stable than secondary, but presentation is not. GitLab notes that destructive actions in lists may retain a default treatment so they do not overwhelm the hierarchy, while final confirmation should use danger. GitLab GOV.UK similarly reserves warning styling for the final serious action rather than the initial entry point. GOV.UK

Documented: consequence and prominence are orthogonal even when an API bundles them.

default

Primer uses default as a general visual variant; GitLab uses it as the non-confirm/non-danger action meaning; Atlassian uses it alongside prominence, consequence and experience-specific appearances. A default is also version- and context-dependent. Inferred: it is a poor canonical concept identifier because its extension changes when the surrounding system changes.

Interpretation rule

The matrix does not prove these names are bad for humans or machines. It demonstrates that cross-system lexical priors are insufficient. Correct interpretation requires system identity, property dimension, definition and local decision rules.