A learner obtains evidence about what is and is not possible in a language from the input he or she hears. Positive evidence comes from the words and structures used by speakers in various contexts, while negative evidence may come from an absence in the input (the learner notices that a particular form or structure does not occur), or through explicit correction by other speakers.

Negative evidence consists of information about what is not possible in a language, which learners can obtain directly, through negative feedback or indirectly, by noticing an absence in the input.

Because Universal Grammar is intended as a minimal specification of an underlying grammar, generativist theories of language acquisition seek to explain acquisition without recourse to direct negative evidence.  Direct negative evidence, or correction, may not be consistently available to all children.  (In many cultures, children are not corrected as they learn their first language; when they are corrected, there is evidence that this feedback is not taken up.)

The issue of direct negative evidence arises in cases of overgeneralisation errors. When learners make errors by applying rules in 'illegal' or inappropriate contexts (e.g., adding the English past 'ed' to irregular verbs like 'go' - *goed), they construct grammars which are more inclusive, or larger than the target grammar. If they are to retreat or recover from overgeneralisation, they are likely to require negative evidence in the form of corrective feedback, which poses theoretical problems for some theories such as generativism.

Generative theory proposes a Subset Principle such that children 'avoid overgeneralisation by always sticking to the most conservative grammar' (MacWhinney, 02).  Emergentists take a different approach, for example the Competition Model.

Activity

Read and reflect: the Competition Model