The ability of the targeted quantitative structure-property relationships (TQSPR) method to predict properties for groups of congeneric compounds was tested with T-c and p(c) data for five homologous series: n-alkanes, 1-alkenes, 1-alkanols, n-alkylbenzenes, and n-alkanoic acids. Training sets were identified from a database of 326 hydrocarbon and oxygen compounds with different structures, described with 1664 descriptors, or from the respective series only. It has been established that the TQSPR method can identify descriptors collinear with the property studied and develop linear equations for the series from measured data. In most cases, the respective collinear descriptors could be identified with the controls imbedded in the TQSPR program. Comparison with presently available methods shows that TQSPR achieves deviations from measured data in most cases within the average experimental uncertainties, like the best ABC methods, but it needs smaller amounts of measured data and provides higher statistical confidence in long-range prediction. The method has been tested with only five homologous series, but the existence of descriptors collinear with properties found in the present work is relevant to all homologous series. When applied to simple molecules, TQSPR can also provide insight into the way compounds are selected by structural similarity and outline eventual inefficiencies in this selection.