Myhill graphs also generalize to the SLk case. The k-factors, however, cannot simply denote edges. Rather the string σ1σ2 ....... σk-1σk asserts, in essence, that if we have just scanned σ1σ2 ....... σk-1 the next symbol is permitted to be σk. The question of whether a given symbol causes the computation to reject or not depends on the preceding k - 1 symbols. Thus, we will take the vertices of the graph to be labeled with strings of length less than or equal to k - 1 over Σ plus one vertex labeled ‘x' and one labeled ‘x'.
We can interpret a k-factor σ1σ2 σk-1σk, then, as denoting an edge between the node labeled σ1σ2 ........σk-1 and that labeled σ2.......σk (the last k - 1 symbols of the string obtained by adding σk to the end of σ1σ2 ........σk-1). While the symbol responsible for the transition along an edge can be determined by looking at the last symbol of the label of the node the edge leads to, for clarity we will label the edges with that symbol as well.
Each of the factors of form xσ2 ........ σk will be interpreted as a path from the vertex labeled x through the vertices labeled with successive pre?xes of σ2 ........ σk, to the vertex labeled σ2 ........ σk with the edges labeled σ2, . . . , σk in sequence. Those of the form σ1 ...... σk-1x will be interpreted as an edge from the vertex labeled σ1 ...... σk-1 to that labeled ‘x', with the edge labeled ‘ε'.
Finally, those of the form xσ1.......σix, for 0 ≤ i < k - 1, (where the substring σ1 ........ σi may be empty) will be interpreted as a path through vertices labeled with successive pre?xes of σ σ (possibly no intermediate vertices) from the vertex labeled ‘x' to that labeled ‘x', with the edges labeled with σ1, . . . , σi (possibly ε) in sequence.