A problem central to sensory diference testing is response bias. lhere are t w experimental strategies f i r dealing with this problem. lhe jrst is to use jhrced choice procedures, like the common duetrio or triangle tests, while the second is to use signal detection measures like d', p0 and the Rindex. lhese strategies are explained and discussed. lhe relationship between the Rindex and the other signal detection measures is eqlained. lhe relationship betweeen R-index values obtained by rating and ranking is explored, as are the alternative compldations of the R-index by ranking: RIB and RMAT. SENSORY DIFFERENCE TESTS AND RESPONSE BIASSensory difference tests are used for the detection and measurement of fine sensory differences between easily confusable food samples. They are a vital tool for the sensory evaluation of food. The development of procedures for such tests necessitates an understanding of what is happening to the judge who is performing them. It requires a consideration of cognitive and physiological factors, which can affect the judge's performance. One such cognitive factor, which is central to the understanding of difference tests, is response bias. Response bias is a comparatively insignificant effect when differences are relatively large and the stimuli are not so similar as to be confusable; for practical purposes it can then be largely ignored. However, for fine differences, which are so small that sensory difference 'Telephone: (916) 752-6389. Journal of Sensory Studies 7 (1992) 1-47. All Rights Resewed. 0 Copyright I992 by Food & Nurition Press, h c . , i'h?nbuN, Omnecticur.

If a chi‐squared analysis were to be performed to determine whether preferences were significant in a paired preference test, the appropriate expected frequencies in the analysis would represent those that would occur should the consumers have no preference. One way of determining these ‘no preference’frequencies, for a particular test situation, would be to note the preference responses obtained when the stimuli were putatively identical. Over 2000 consumers were given paired preference tests in which the stimuli were putatively identical. Response conditions and consumer groups were varied and the proportions of preference and no preference responses were noted. In a preliminary experiment, further research was seen to be justified when for putatively identical stimuli, judges did not exclusively express lack of preference; many expressed a preference for one or other of the stimuli. Further experiments were conducted using a written response condition and putatively identical potato chips (potato crisps) as stimuli. Using a single ‘no preference’option, variation in the placement of this option at either first, second or third place on the response sheet had no significant effect on the response frequencies. The proportion of ‘no preference’responses was approximately 30% in all cases. A previously reported 40‐20‐40 distribution was not always confirmed. The experiment was repeated with Korean consumers; these had fewer ‘no preference’responses. Deriving preferences from hedonic scales, rather than having judges respond to preference options, increased the proportion of ‘no preference’responses, with American judges still having more than Korean judges. Yet there are logical objections to extracting preference data from hedonic scales. Increasing the number of ‘no preference’options to two or three, had the effect of increasing the number of ‘no preference’responses up to as much as approximately 60%. Extending the results to Koreans, using two ‘no preference’options, it was seen that only the judges in an anonymous response condition had significantly fewer ‘no preference’responses than Americans. The use of these response frequencies as expected frequencies in chi squared analyses was illustrated, after adjustments for counterbalancing.

Strawberries (Fra~aria ananassa Duch., cv. 'S&a') were stored 10 days in 1.0%; O.s%, or 0.25% O2 or air + 20% Cb,; or 6 days in air + 50% or 80% CO, at 0 or 5°C without detrimental effects on quality. Decay and sofkning were reduced by treatments. An untrained taste panel, under ordinary eating conditions, did not consistently differentiate 'Pajaro' strawberries kept in 0.25% 0, from those stored in air. A trained taste panel, under controlled conditions, perceived slight off-flavor in 'G3' strawberries kept in 0.25% or 0% 02. This correlated with ethanol, ethyl acetate, and acetaldehyde in juice.The 50% or 80% CO2 treatments caused injury after 8 to 10 days, while 20% CO2 treatments did not. All high CO2 treatments caused increase in pH of juice.

The greater sensitivity of the 'one strong, two weak stimuli' over the 'one weak, two strong stimuli' versions of the triangle and duotrio tests was confirmed using a model system of salt and water stimuli. The relative distinguishability of salt and water, in adaptation states varied by prerinses, was measured using an R-index procedure. The relative distinguishabilities were used as the basis for a predictive model, 'sequential sensitivity analysis', to explain the relative sensitivity of versions of the triangular and duo-trio protocols. Adaptation did not furnish a complete explanation of their relative sensitivity.

The 9-point hedonic scale has been used routinely in food science, the same way for 60 years. Now, with advances in technology, data from the scale are being used for more and more complex programs for statistical analysis and modeling. Accordingly, it is worth reconsidering the presentation protocols and the analyses associated with the scale, as well as some alternatives. How the brain generates numbers and the types of numbers it generates has relevance for the choice of measurement protocols. There are alternatives to the generally used serial monadic protocol, which can be more suitable. Traditionally, the 'words' on the 9-point hedonic scale are reassigned as 'numbers', while other '9-point hedonic scales' are purely numerical; the two are not interchangeable. Parametric statistical analysis of scaling data is examined critically and alternatives discussed. The potential of a promising alternative to scaling itself, simple ranking with a hedonic R-Index signal detection analysis, is explored in comparison with the 9-point hedonic scale.