Intriguing phenomena occur when experts use computer-assisted design tools in electronics. It can be seen that tools must support information access, 'escape from formalism', 'secondary notation', and differences between individual users. This paper explores a new account for the data, relying on the idea of graphical implicature, generalised from Grice's conversational implicature. All communicative artefacts carry implicatures, significance beyond their literal meaning. The important thing is to control them systematically, so that a graphic avoids unwanted implicatures, and carries the desired ones. Whether or not the current account is useful has broader significance. If it is useful, it may prove possible to predict and explain the properties of complex graphical representations, by borrowing formal techniques from natural language pragmatics.