This paper describes Grammar Learning by Partition Search, a general method for automatically constructing grammars for a range of parsing tasks. Given a base grammar, a training corpus, and a parsing task, Partition Search constructs an optimised probabilistic context-free grammar by searching a space of nonterminal set partitions, looking for a partition that maximises parsing performance and minimises grammar size. The method can be used to optimise grammars in terms of size and performance, or to adapt existing grammars to new parsing tasks and new domains. This paper reports an example application to optimising a base grammar extracted from the Wall Street Journal Corpus. Partition Search improves parsing performance by up to 5.29%, and reduces grammar size by up to 16.89%. Parsing results are better than in existing treebank grammar research, and compared to other grammar compression methods, Partition Search has the advantage of achieving compression without loss of grammar coverage.