Purchase this article with an account.
Z-L Lu, BA Dosher; External Noise Methods and Observer Models . Invest. Ophthalmol. Vis. Sci. 2002;43(13):2915.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: The external noise method and the linear amplifier model (LAM) reveal inefficiencies of the perceptual system.1 The LAM models threshold vs external noise contrast (TVC) functions at a single performance level, but generally fails to account for TVCs at different performance levels.2 We proposed a perceptual template model (PTM), an LAM elaborated with a non-linear transducer and internal multiplicative noise/contrast gain control, to model TVCs at multiple performance levels or full psychometric functions.2 A"triple-TVC" method ---measuring TVCs at three performance levels--- provides necessary constraints on observer models including the PTM. A different "double-pass" test3 measures performance twice with identical stimuli (signal + external noise). Response consistency assays the ratio of internal and external noise amplitude and therefore estimates total (multiplicative and additive) noise. Method: We analyze the mathematical properties of three observer models, the LAM, the PTM and the EAW4 (LAM+multiplicative noise + uncertainty) in relation to both the triple-TVC and double-pass paradigms. Results: We show (1) LAM can accommodate neither multiple-TVCs nor response consistency; (2) LAMs with uncertainty and a max decision rule can accommodate multiple-TVCs, but not response consistency; (3) Both PTM and EAW with substantial uncertainty can accommodate both multiple-TVCs and response consistency. Conclusion: The triple-TVC and double-pass methods provide complementary constraints on observer models. Both multiplicative noise and some form of nonlinearity (either substantial uncertainty or a nonlinear transducer function) are necessary to model existing data from triple-TVC and double-pass measurements. Implications for models of the mechanisms of attention and/or perceptual learning are discussed. 1Pelli, Dissertaion, 1981. 2Lu & Dosher, JOSA, 1999. 3Burgess & Colborne, JOSA, 1988. 4Eckstein, Ahumada & Watson, JOSA, 1997.
This PDF is available to Subscribers Only