This thesis is concerned with quantitative visual assessments of plant disease severity and with food tasting studies. Judgements made by humans are prone to bias from several sources. Thus, the designs used for such experiments and the models used for analysis of such data need to account for this bias. In studies where human assessors make judgements of long sequences of varying levels of a stimulus, sequential effects, such as carry-over from the previous stimulus and order of the stimulus in the sequence, are likely to arise. Designs which are balanced for order and carry-over have therefore been studied here, and a program which searches for sequences of such design was written. The sequences generated by this program were then grouped according to certain invariance and optimality properties. Calibration consists of comparing the performance of different measuring instruments that are used to measure similar samples of interest and correcting for biases of some of the instruments. Here, humans were used as measuring instruments in a visual assessment experiment. A test experiment was carried out for which true stimulus intensities were known, and then calibration of responses from a subsequent similar experiment was done. This kind of calibration is known as absolute calibration because the true stimulus intensities in the test experiment were known. It was based on a Bayesian predictive method applied to a regression model of the responses on the true stimulus levels, with carry-over and order effects, as well as first order auto-correlation in the errors. A method to select the best assessors was based on the Shannon information criterion. Data were analysed from a series of food tasting experiments, in which a panel of assessors made judgements based on a number of attributes. Data from these experiments were combined in order to study assessor performance over time, and to use information about the assessors to improve analysis of their future performances.