In system identi cation, estimates of the unknown system model orders are often required. An algorithm for estimating model orders is described which looks at input/output data covariance matrix eigenvectors. When model orders are overestimated, zeros appear in the noise subspace eigenvectors. The number of zeros present can be used to estimate model orders.

Daeman and Rijmen had derived the distributions of correlations between linear combinations of the input and output of uniform random functions and uniform random permutations. We generalise their results by deriving the distributions of correlations between arbitrary combinations of the input and the output of uniform random functions and uniform random permutations.

Models of both speech perception and speech production typically postulate a processing level that involves some form of phonological processing. There is disagreement, however, on the question of whether there are separate phonological systems for speech input versus speech output. We review a range of neuroscientific data that indicate that input and output phonological systems partially over...