Liking of food textures and relationship with oral physiological parameters

Abstract

The first two objectives of this study were to examine the relationships among liking ratings of a wide variety of food textures and to group people based on their liking ratings. In Part 1, 288 participants rated their liking of 106 texture attributes. Principal components analysis (PCA) of the liking ratings produced a 34-component solution; none of the components explained more than 3% of the variation in liking ratings. Agglomerative hierarchical clustering (AHC) of participants on their liking ratings produced a 5-cluster solution that explained only 5.6% of the variation in liking ratings. Most individuals fell into clusters without distinct texture liking profiles. Another objective of this study was to examine relationships among individuals’ food texture liking ratings, mouth behavior group, and measurements of four oral physiological parameters (saliva flow rate, chewing efficiency, biting force, and particle size sensitivity). In Part 2, 100 participants completed the survey on food texture liking and then classified themselves into one of the four mouth behavior groups (Chewers, Crunchers, Smooshers, Suckers) proposed by Jeltema et al. (2015). Measurements of the four oral physiological parameters were also recorded for each participant. AHC of participants on their oral physiological measurements produced a 4-cluster solution consisting of a ‘low particle size sensitivity’ cluster, a ‘high biting force’ cluster, a ‘high saliva flow rate’ cluster, and a ‘low saliva flow rate and low chewing efficiency’ cluster. These clusters accounted for 52% of the variation in the oral physiological measurements. Again, most individuals fell into clusters without distinct texture liking profiles. Analyses of variance (ANOVA) of liking ratings and oral physiological measurements by mouth behavior group and regression analyses of liking ratings, oral physiological measurements, and mouth behavior group revealed only a few relationships among these three types of variables. All of these relationships would have been statistically insignificant if we had applied Bonferroni corrections to adjust for the very large number of statistical tests conducted. Complete R and SAS codes used for data analysis are included as a supplementary file.