Data analysis
For the sensory panel experiments, a complete randomized block design was used. All samples were coded with three digit random numbers and the serving order was also completely randomized. The Duncan's New Multiple Range Test was used as the multiple comparison test for determining statistical significance between treatment means. The Spearman's rank order correlation analysis was also generated and followed the criteria outlined by Ott (1988). The analysis of variance tables and multiple comparison test values were verified using SAS software (SAS Institute, Inc., 1991).
For rheological data, the Bohlin VOR software executed the power law model to determine consistency coefficients and flow behaviour indices of all treatment samples. Data generated from the instrumental rheology and firmness measurements were analysed using the Number Crunching Statistical Analysis System (J. L. Hintze, Kaysville, Utah, 1987).
Upon initial analysis, there were significant trial effects and interactions between trials and other parameters the data from the two processing trials. This was believed to result from the fact that for the second trial with the fat free ice cream (FF), a modified starch that, although originating from the same source, was from a different batch than the starch used for other samples. This sample was difficult to work with and appeared to alter the physical properties of the ice cream. As a result, for the data from the second FF trial have not been used in any of the statistical analyses. Comments on the unusual behaviour of this sample, however, are included in the discussion.