Statistics for discovery

The question is discussed of why investigators in engineering and the physical sciences rarely use statistical methods. It is argued that statistics has in the past been overly influenced by the needs of mathematics rather than those of scientific learning and discovery. Remedies are suggested.