Felsic tuffs are common throughout the Late Permian Newcastle and Wollombi Coal Measures (NCM, WCM) of the northern Sydney Basin. Petrographic studies reveal that they are composed of fragmented crystals of plagioclase, quartz, and less common degraded biotite and K-feldspar, together with lithic fragments and relict, altered glass shards. Alteration during burial at T < 100 C has resulted in the formation of illite/smectite (I/S; I62-78), kaolinite, siderite/ankerite and, in extreme circumstances, tonsteins dominated by I/S. The latter is formed by the alteration of vitric tuffs. Zr/TiO2 and Nb/Y ratios, and rock-primordial and chondrite-normalised REE patterns (La/Yb = 4.05 to 11.37) indicate that the tuffs have been derived from rhyodacitic to dacitic, continental arc, calc-alkaline magmas. The accuracy of a recent lithostratigraphic correlations between the NCM and the WCM has been tested by determining the chemical composition of four stratigraphically well defined tuffs in the NCM and comparing them with those obtained from tuffs of the WCM which are thought to be stratigraphically equivalent. The comparison was carried out using multivariate, statistical analysis. The analysis revealed that two of the tuffs (Awaba, Nobbys) in the NCM could be distinguished; the others (Mt. Hutton, Warners Bay) showed considerable scatter. This contrasted with the tonsteins formed from the tuffs, which, apart from the Mt. Hutton Tuff, were able to be separated. In the WCM, three of the tuffs (Nalleen, unnamed tuff equivalent to the Mt. Hutton Tuff, Monkey Place Creek) could be separated; however, the fourth (unnamed tuff equivalent to the Warners Bay Tuff) exhibited some scatter. Treating tuffs in the WCM separately, as unknowns in the discriminant model to determine possible correlatives in the NCM, revealed that the Nalleen Tuff is equivalent to the Awaba Tuff in the NCM, confirming the lithostratigraphic correlation. The remainder of the tuffs in the WCM, however, show moderate to poor correlation with the proposed stratigraphically equivalents in the NCM. Thus, the analysis shows that correlation based on lithostratigraphic grounds may be incorrect.