This issue of the Machine Learning journal is devoted to inductive logic programming (ILP). The Ninth International Workshop on ILP (ILP'99) was held in June 1999 in Bled (Slovenia). After the workshop, we invited authors who presented their work at ILP'99 to consider submitting a suitably extended version of their paper to the Machine Learning journal. As ILP workshops are not (indeed, should not be) the only place where work on ILP can be found, we scrutinised other proceedings such as ICML, ECML, COLT, KDD, and UAI for ILP-related papers and wrote to their authors with a similar request. By the end of March 2000 we had received 12 submissions, each of which was reviewed by 3 reviewers. On the basis of the reviews, we selected the four papers that are included in this issue. In its most common form, ILP is concerned with inducing rules from examples and background knowledge, all of which are expressed as Prolog programs. This uniformity of representation is relatively unique within the diverse field of machine learning, and has contributed significantly to the identity and coherence of inductive logic programming as a field of research. However, one should not confuse the contingencies of syntax with the essentials of representation. What is crucial about ILP is not that rules are written with the conclusion preceding the conditions, or that variables in rules are to begin with an uppercase character, but that the underlying logic is first-order predicate calculus, which means that the objects classified by these rules can have a deeply nested yet flexible structure. Hierarchical structures are required whenever the objects to be classified have more structure than can be expressed by an attribute-value vector. Flexible structures are required, e.g., whenever only part of an object is responsible for its classification, but it is unknown in advance which part (as in the multiple instance problem itemulti-instance), or when the objects are sequences (as in bio-informatics or natural language domains). So, in a more general sense, ILP encompasses the application of machine learning methods to domains with flexible nested structures. It should therefore come as no surprise that one of the papers in this issue is clearly addressing ILP issues, even though it does not present a single line of Prolog. (...) ILP started to claim its place in the world as a separate branch of machine learning when the first ILP workshop was organised in 1991. We think it is appropriate to characterise the 10 years that have elapsed as ILP's adolescence, and we are happy to say that the papers in this issue show that ILP is coming of age. (...) In our view, these papers demonstrate that inductive logic programming is firmly embedded in machine learning, and that we can look forward to more exciting work exploring the connections between ILP and other machine learning approaches.