A substring recognizer for a language L determines whether a string s is a substring of a sentence in L, i.e., substring-recognize(s) succeeds if and only if ∃ v,w: vsw ∈ L. The algorithm for substring recognition presented here accepts general context-free grammars and uses the same parse tables as the parsing algorithm from which it was derived. Substring recognition is useful for non-correcting syntax error recovery and for incremental parsing. By extending the substring recognizer with the ability to generate trees for the possible contextual completions of the substring, we obtain a substring parser, which can be used in a syntax-directed editor to complete fragments of sentences.