In the early seventies, Shelah proposed a model-theoretic construction, nowadays called “iteration”. This construction is an infinite replication in a tree-like manner where every vertex possesses its own copy of the original structure. Stupp proved that the decidability of the monadic second-order (MSO) theory is transferred from the original structure onto the iterated one. In its extended version discovered by Muchnik and introduced by Semenov, the iteration became popular in computer science logic thanks to a paper by Walukiewicz. Compared to the basic iteration, Muchnik’s iteration has an additional unary predicate which, in every copy, marks the vertex that is the clone of the possessor of the copy. A widely spread belief that this extension is crucial is formally confirmed in the paper. Two hierarchies of relational structures generated from finite structures by MSO interpretations and either Shelah-Stupp’s iteration or Muchnik’s iteration are compared. It turns out that the two hierarchies coincide at level 1. Every level of the latter hierarchy is closed under Shelah-Stupp’s interation. In particular, the former hierarchy collapses at level 1.

A word rewriting system is called monadic if each of its right hand sides is either a single letter or the empty word. We study the images of higher order indexed languages (defined by Maslov) under inverse derivations of infinite monadic systems. We show that the inverse derivations of deterministic level n indexed languages by confluent regular monadic systems are deterministic level n+1 languages, and that the inverse derivations of level n indexed monadic systems preserve level $n$ indexed languages. Both results are established using a fine structural study of classes of infinite automata accepting level $n$ indexed languages. Our work generalizes formerly known results about regular and context-free languages which form the first two levels of the indexed language hierarchy.

Several recent works have studied subfamilies of deterministic context-free languages with good closure properties, for instance the families of input-driven or visibly pushdown languages, or more generally families of languages accepted by pushdown automata whose stack height can be uniquely determined by the input word read so far. These ideas can be described as a notion of synchronization. In this paper we present an extension of synchronization to all context-free languages using graph grammars. This generalization allows one to define boolean algebras of non-deterministic but unambiguous context-free languages containing regular languages.

The monadic second-order theory of term rewritings is considered. It is shown that the monadic theory of the rewriting (or the suffix rewriting) of a ground rewrite system is undecidable. Furthermore, the first-order theory is undecidable for the prefix derivation according to a linear context-free grammar on linear terms. Nevertheless, a new notion on terms with variables is introduced: a term is entire if each of its subterms either is a variable, or is without variable or has the same variables as the term. It is shown that the monadic theory is decidable (respectively undecidable) for the prefix rewriting according to a rewrite system on entire terms, with an axiom (respectively without axiom).>