Mode declarations are a successful form of language bias in explanatory ILP. But, while they are heavily used in Horn systems, they have yet to be similarly exploited in more expressive clausal settings. This paper presents a mode-directed ILP procedure for full clausal logic. It employs a first-order inference engine to abductively and inductively explain a set of examples with respect to a background theory. Each stage of hypothesis formation is guided by mode declarations using a generalisation of efficient Horn clause techniques for inverting entailment. Our approach exploits language bias more effectively than previous non- Horn ILP methods and avoids the need for interactive user assistance.