The process is generally agreed to begin before the entire word is heard, so more than one lexical entry will be contacted for each word heard. In the
Cohort model, the recognition system makes contact with words on the basis of the first one or two phonemes of the input.