A hierarchy of data structures, called the hypertree hierarchy, is presented which has strings and trees as its smallest two elements. A generalized frontiering operation is presented. Grammars, automata and regular expressions are extended to this hierarchy. These are shown to be equivalent and the resulting languages are called regular. This leads to a hierarchy of regular languages on the hypertree hierarchy. These are projected onto the set of strings by the frontier operation resulting in a true hierarchy of string languages. This hierarchy is called the (IO) algebraic language hierarchy. It has regular languages, context free languages and macro languages as its first three levels. It is contained in the set of context sensitive languages but is not equal to it. Other characterizations are presented.