Abstract: 【Objective】 The aim of this study is to screen scientific evaluation indexes for apple taste. 【Method】 Six indexes of 132 apple samples were determined, including pH determined by acidity meter, total soluble solid (TSS) determined by brix meter, soluble sugar (SS) determined with the fehling reagent titration, titratable acidity (TA) determined with the indicator titration method, the taste of each sample was determined by tasting, the ratio of total soluble solid to titratable acidity (RTT), and the ratio of soluble sugar to titratable acidity (RST). Statistical methods (such as level analysis，regression analysis, cluster analysis, and factor analysis) were used to clarify the level of each index and its distribution, to clarify the variation tendency between indexes and the quantitative relationships between/among indexes, and to screen appropriate indexes for evaluating and classifing apple taste. 【Result】 The variances of three indexes (pH, TSS and SS) were all small, with the variable coefficient of 8.5%, 10.7% and 10.6%, respectively, while the variances of other three indexes (TA, RTT, and RST) were all big, with the variable coefficient of 45.1%, 64.9% and 66.5%. Both TSS and SS distributed normally, with the probability values of 0.6241 and 0.6298. pH, TA, RTT and RST distributed unnormally. pH, RTT and RST all had a significant negative correlation with TA, with the correlation coefficient of -0.8810, -0.8117 and -0.8116. Both RTT and RST had a significant positive correlation with pH, so did SS with TSS, and RST with RTT. pH had a significant logarithmic-function concerning TA with a determination coefficient of 0.8522. Between TSS and SS, as well as between RTT and RST, there was a significant linear function. RTT as well as RST had a significant power function concerning TA. Each of the six indexes mentioned above, except TA (with low fitting precision and low prediction accuracy), had a multivariate linear model concerning the other five indexes with the mean fitting error of 1.88%-2.74% and the mean prediction error of 1.83%-2.59%. From the six indexes studied, TA, TSS and RTT were screened as evaluation indexes of apple taste, which represented sour index, sweet index and comprehensive taste index, respectively. With use of these indexes, 132 apple samples can be divided into five classes that were sweet, sour-sweet, sweet-sour moderately, sweet-sour, and sour. 【Conclusion】 Apple taste can be scientifically evaluated and classified by three indexes, including TSS, TA and RTT.

Key words: apple, taste, evaluation index, cluster analysis, factor analysis, regression analysis