Abstract
Jazayeri, Ogden and Rounds have shown that the high time complexity of Knuth's algorithm for testing attribute grammars for circularity is no accident. It was proved that there is a constant c>0 such that any deterministic Turing Machine which correctly tests for circularity must run for more than 2 ^(cn/log n) steps on infinitely many attribute grammars (AGs) (the size of an AG is the number of symbols required to write it down). The proof was rather complex; the purpose of this note is to provide a simpler one.