Sir

As a former denizen of the British boarding-school system, I was delighted to find that simpler and mathematically rigorous methods exist for knotting a tie (Nature 398, 31–32; 1999).

I wondered, however, whether experimental testing might reveal unexpected flaws in the modelling process.

I began with the three-move reverse-start knot, which is one move shorter than the four-in-hand knot I learned as a schoolboy. All was well. Until⃛ as I went to remove the experimental apparatus (my old-boy tie), I discovered to my horror that the ‘simpler’ knot left me with an overhand loop. The four-in-hand (and all knots with an even number of moves, including the Windsor and half-Windsor) leaves no such lingering trace requiring further engineering solutions.

I have reached three major conclusions from my experiment. (1) I shall generally stick with the method I learned in my Dickensian past, unless (2) I am late for an appointment requiring a tie, and the fraction of a second gained at the outset outweighs the delayed cost of undoing an overhand loop at the end of the appointment. And (3) a model is rarely complete until it has been field-tested in a variety of contexts.

Incidentally, many inhabitants in this part of North America have circumvented the entire suite of theoretical and applied problems entailed in knot tying by adopting the bolo tie.