Evaluation of published methods reveals that existing methods for saliva sampling do not address the physical-chemical attributes of volatile organic compounds (VOCs). This study describes and presents evidence for adopting in situ sampling of salivary VOCs directly from the oral cavity using a polydimethylsiloxane (PDMS) based sampler. In vitro studies indicated that the vapour pressure of analytes was a factor in both the recovery of analytes and the precision of the recovery. The highest recoveries were observed for VOCs with the lowest vapour pressures, for example 5-nonanol (vapour pressure (P v ) Œ 14 Pa) recoveries were approximately 20 times greater than those observed for octane (P v Œ 1726 Pa). Similarly, relative standard deviations reduced with vapour pressure, with the RSD for 5-nonanol responses observed to be 2.7% when compared to RSD Œ 26% for octane. Evaluation of VOCs recovered from 6 in vivo samples indicated that VOC concentrations in saliva may follow lognormal distributions; log-normal RSDs falling between 4.4% and 18.2% across the range of volatilities encountered. Increasing sampling time from 1 to 30 minutes indicated that the recovery of VOC into the sampler was affected by interaction between different physical-chemical properties and biogenic flux. A sampling time of 10 min was found to offer an acceptable compromise that enabled a representative sample to be acquired for the widest range of observed VOC behaviours with the sampler. The potential to 'tune' the sampling protocol for targeted analysis based on these factors was also noted. Comparison with passive drool saliva collection revealed up to 10 5 enhancement with reduced variability compared to drooled samples. This approach to in situ saliva sampling appears to have significant analytical utility for studying volatile signatures in humans.