Phenolic compounds play an intrinsic role in maturation, aging and stylizing of a wine. They also contribute significantly to the sensory properties of the wine. The primary objective of this study was to determine if any correlation existed between the phenolic composition of twelve, 1998 red wines and the results obtained from a sensory evaluation that focused on the meaning of "hard" and "soft" tannin, as used by a group of winemakers, using a "hardness" scale and quantifiable sensory terms of bitterness, sourness and astringency by a trained panel. Wines were analyzed using reverse phase high performance liquid chromatography (HPLC), spectrophotometry and protein precipitation assays. Using Partial Least Squares (PLS) regression, the collected data was then compared to the sensory evaluation results. The PLS regression models constructed for "hardness" versus chemical analysis highlighted the importance of protein precipitable tannin concentration (TAN) on influencing "hardness". Both the complete and simplified models were able to predict "hardness" in these twelve wines with good confidence (p<0.001). The PLS regression model constructed for HPLC phenolic results had the highest percentage of explained variance for the chemical variables - 88%. The PLS regression model constructed for tannin and polymeric pigment results had the highest percentage of explained variance for "hardness" - 80%. In addition to predicting the sensory term "hardness," TAN content was also the most influential chemical variable in predicting the maximum intensity of bitterness (Imaxl B) and astringency (Imaxl A) for the first sip. Titratable acidity was the most important chemical variable influencing the prediction of the maximum intensity of sourness for the first sip (Imax1 S). From score plots, only the simplified PLS regression model of "hardness" versus all chemical analysis demonstrated a trend within the four wineries; each of which donated two wine samples each, one "hard" and one "soft." The "hard" wine was consistently place to the right of the "soft" wine. All other regression models did not clearly exhibit this trend. Moreover, collectively the twelve wines were not clearly separated into "hard" and "soft" groupings. Thus, winemakers did not universally apply the same definition or criteria for determining if a wine was "hard" or "soft."