A comparison of methodologies in the measurement of olfactory sensitivity under conditions of non-adaptation and co-adaptation.

An investigation into the relative efficiency of Threshold and Signal Detectability measures of olfactory Sensitivity was undertaken using isopropyl alcohol as stimulus. Intensive testing of seven subjects under adaptive and non-adaptive conditions revealed that the Signal Detectability paradigm, although theoretically desirable because of its allowance for the subject's response bias, was difficult to implement because of the prolonged testing required. A variance of the rating technique involving multiple stimulus concentration presentations in a three-hour testing session was attempted. Results in the nonadapting environment indicated that the method was more effective than single-stimulus concentration presentations. However reliable results under adapting conditions were obtained In the case of one subject only. As a comparison, the constant stimulus method was used to obtain threshold using a procedure similar to Cheesman's and Mayne's group threshold determinations, but modified for individual subject testing. Practice and learning effects were noted and their relevance discussed. The Cheesman hypothesis viz, that adapting odour concentration and threshold elevation obey a linear logarithmic relation which is characteristic of the adapting and test stimulus compounds was confirmed in two subjects only. Reasons for non-confirmation include the extended adapting stimulus concentration range, subject boredom and inadequate control of the stimulus; the latter factors being a consequence of prolonged testing. The sniff-bottle and air-dilution forms of stimulus presentation were employed, although not with the same subjects. Thus a direct contrast of these presentation methods was not possible. However results were generally more consistent with the air-dilution technique. The feasibility of an odour classification based on subject responses during adaptation rather than on molecular parameters was evaluated and the difficulties likely to be encountered in employing Signal Detectability measures of sensitivity was discussed.