The 44 sounds of English, decomposed into binary features.
Every phoneme is a bundle of [±] distinctive features. This tool lets you sort the full feature matrix, watch how a single feature (or combination) carves the inventory into natural classes, and—given any set of sounds—compute the minimal feature specification that picks out exactly that class and nothing else.
The feature matrix
Every row is a phoneme; every column a feature. + means the feature is present, − absent, · unspecified. Click any column header to sort.
Feature values are a simplified, internally consistent pedagogical analysis of the common “44 phonemes” school inventory. +/− mark a feature as present/absent; · marks it unspecified (not relevant for that sound). Central vowels are treated as [+back]; schwa is the only [+reduced] vowel. Theoretical and dialectal treatments differ, especially for vowels, /r/, /h/, and the affricates.
Phonemes on a sphere
The 44 phonemes projected by feature similarity onto a 3D sphere. Each arc threads through every sound sharing a feature value — where arcs cross, phonemes share multiple natural-class memberships.
The sphere is a feature-distance projection: nearby phonemes share more features. Feature arcs pass through the phonemes that carry a selected value, so intersections reveal shared natural-class membership. A 2D sphere cannot exactly represent all 21 binary dimensions, but it reveals the clustering structure of the feature space.
How features group sounds
Toggle feature values on the left to build a feature formula. Sounds matching every constraint form a natural class—highlighted below.
Find the minimal feature set
Select a set of sounds. The solver searches for the smallest bundle of feature-values that defines exactly that set to the exclusion of every other phoneme in English.
Your sound set
Minimal defining features
Select sounds and run the solver to see the minimal feature specification.
The solver returns the globally smallest set of feature-values shared by all selected sounds whose extension (the set of all phonemes satisfying them) equals your selection exactly. If no combination of features isolates your set, it reports that the set is not a natural class under this feature inventory.