The text in the Voynich manuscript was analysed with respect to transition frequencies between the various characters within words. Principal Component Analysis (PCA) was utilised to visualise the relationships graphically, which provided evidence of grouping of characters into two or more classes depending on their transition probabilities to other characters. Based on comparative analysis on texts in known languages, consonant and vowel-classes were suggested. Vowels and consonants were clearly separated in the PCA score plots of the characters in the known languages. The Voynich characters were generally close together in the PCA score plots if the characters looked similar. That is, aggregation based on visual similarity was observed. Furthermore, some other properties of the different characters related to their transition frequencies to other characters were revealed and discussed in terms of their visual layouts, ligatures or accents in the text.