ABSTRACT.Thesholds for vanillin dissolved in water were determined first by 32 panelists, then by 100 additional panelists. Extreme‐value analysis was applied to the results from the first 32 determinations to estimate the likelihood of thresholds later encountered being even lower than those already detected. No threshold among the 100 additional determinations was below the 5% confidence limit for the regression curve based on the 32 determinations originally made. There was evidence that some panelists were ageusic to vanillin in aqueous solution. Although some investigators have suggested that the frequency distribution for threshold values of vanillin is bimodal, application of the Kolmogorov‐smirnov test to the distribution of the 132 values, as natural logarithms, gave little cause to reject the hypothesis that the distribution was unimo