The effects of random and nonrandom errors were analyzed. Four types of nonrandom errors were simulated and examined in detail: 1. misidentification of compounds; 2. unresolved components; 3. errors in the quantification of individual compounds; and 4. errors in the quantification of a series of analyses. These cases were compared using data from studies inJuniperus. Misidentification of 1 of 40 components had only a slight effect on the resulting classification even when that error accounted for 20% of the total character weights. Unresolved components also had only a slight effect. Nonrandom errors in the quantification of individual compounds (throughout a study) were shown to have no effect on the similarity measure used in this study. Errors in the quantification of a series of analyses (such as encountered in seasonal variations) appeared to be the most serious source of error and were shown to be of considerable importance in chemosystematic studies. Several preventative procedures are suggested.