The Application of Log-Linear Techniques to Recursive Models: Comment on Reynolds