A parser must be able to continue parsing after encountering a syntactic error to check the remainder of the input. To achieve this, it is not necessary to perform corrections on either the input text or the stack contents. A formal framework is provided in which noncorrecting syntax error recovery concepts are defined and investigated. The simplicity of these concepts allows the statement of provable properties, such as the absence of spurious error messages or the avoidance of skipping input text. These properties are due to the fact that no assumptions about the nature of the errors need be made to continue parsing.