I saw a discussion somewhere recently about the pitfalls of using autorads of blots, etc., to estimate quantities of things, eg message levels. I dont think it was a regular publication, but it had references. I have lost this reference and will appreciate it if anyone can put me bace onto it. Thank you.