Scanning Behaviour and Pattern Recognition
Lishman, John Rowland
MetadataShow full item record
Two basic models of human pattern recognition have been advanced: feature analysis and hypothesis testing. These can only be discriminated by looking at behaviour before recognition. This is studied here by having the subject scan with a pen that writes only where the (invisible) picture is black. Although a computer simulation shows that it is possible and efficient, subjects scanning capital letters and simple shapes with this technique rarely use hypothesis testing.