This paper proposes that a root should be defined as a contentful morph that can occur as part of a free form without another contentful morph. This definition can be applied to all languages using the same criteria and is very largely in line with existing usage. Roots are concrete forms with a shape consisting of a contiguous string of segments, so that consonantal skeletons of the Semitic type do not fall under the definition. They differ from affixes in that they may occur freely and have contentful meaning, i.e. denote an action, an object or a property. These three root classes are closely related to the word class notions verb, noun and adjective. The definition of root proposed here is largely intuitive, but it must be noted that heterosemous root pairs (such as English hammer (noun) and hammer (verb)) cannot be seen as having “the same root”, but must be treated as sister roots.