Lychnophora ericoides is a Brazilian medicinal plant used in folk medicine as an anti-inflammatory and analgesic agent. The essential oils from leaves of two populations with and without scent, collected at 2-month intervals during an 1-year period, were analysed by GC-MS. The results were submitted to principal component and cluster analysis which allowed two groups of essential oils to be distinguished with respect to sampling site and scent: cluster I (Vianópolis site, with specimens exhibiting an aromatic scent) containing a high percentage of alpha-bisabolol (44.7-76.4%) and alpha-cadinol (10.9-23.5%), and cluster II (Cristalina site, with specimens without scent) characterised by a high content of (E)-nerolidol (31.3-47.1%) and ar-dihydro-turmerone (4.8-15.4%). The canonical discriminant analysis showed that using the data set of the seven sampling months and (E)-nerolidol and alpha-bisabolol as predictable variables, it was possible to distinguish between the samples harvested according to Cerrado seasons, dry winter (May-September) and humid summer (November-March). In addition, canonical correlation analysis between the soil sampling sites and the populations revealed a significant relationship between oil components and edaphic factors. Oxygenated sesquiterpenes and potential acidity, Al saturation, cationic exchange capacity, silt, and sand load as the first canonical variate were fairly strongly related to samples collected in Vianópolis site. On the other hand, monoterpenes and sesquiterpene hydrocarbons were strongly related to chemical balance in soils (organic matter, P and base saturation), which is related to samples at the Cristalina site. The chemovariation observed appears to be environmentally determined.