Tuck devised a function Q(x), associated with a function D(x), whose positivity guarantees the absence of complex zeros of D(x) close to the real x axis, and observed that large values of Q are very rare if D is associated with the Riemann zeros. In an unusual and challenging application of random-matrix theory with P Shukla, this is explained by studying the probability distribution P(Q) for functions D with N zeros corresponding to eigenvalues of the Gaussian unitary ensemble (GUE).