When considering a simple hypertext model, having only nodes and links, the possible paths through the hyperdocument are nothing more than sequences of nodes. When we associate names with nodes, the sequences form a simple language which we can analyze formally. Incidentally, the language of a hyperdocument is accepted by the finite state automaton which has states and transitions that correspond (one to one) to the nodes and links of the hyperdocument. The state of the automaton represents the "current" node.
Navigational problems in a hyperdocument can thus easily be reduced to properties of the automaton. A deadlock in the automaton corresponds to a node with no outgoing links for instance. Loops correspond to infinite strings accepted by the automaton.
When the hypertext system is considered as well as the hyperdocument the situation becomes more complicated: the navigation-aids offered by the system generate more possible paths through the hyperdocument. Some systems also offer the possibility to display more than one node at a time, and to follow links from either of the simultaneously displayed nodes.
In a complex and open-ended hypertext model, such as the Tower Model, it is not possible to describe the browsing semantics without placing some restrictions on the open-endedness of the model. In [ADBNK93] some restrictions are placed on the Tower Model to enable a study of the browsing semantics using Petri-nets.