The logarithmic hexadecane-air partitioning constant (L) is one of the most common measures for solute-solvent nonspecific interaction. It is thus an essential compound descriptor for Linear Solute Energy Relationships (LSERs) that are used for predicting partitioning of organic chemicals. In this work L values from 2.3 to 13.7- thus covering 11 orders of magnitude of the linear partitioning coefficient - were measured with standard deviations of <0.28 (average = 0.10) using inverse gas chromatography with non-polar capillary columns. The 104 studied compounds were all multifunctional, 52 of them were environmentally relevant pesticides, 6 hormones, 6 drugs and 4 phthalates. In addition, the predictive performance of 3 software tools (ABSOLVE, SPARC, COSMOtherm) for the L value was evaluated. To this end the data set was divided into 2 subsets: one with compounds containing maximally 2 functional groups and the other with more than 2 functional groups. None of the software tools reached the desired accuracy, defined as root mean square error over all compared compounds, of +/-0.3 units. ABSOLVE performed well for the bifunctional compounds but failed for many pesticides and drugs. SPARC had major problems with highly fluorinated and phosphate containing compounds, but showed a good performance for all other compounds. COSMOtherm worked best for the predictions of the pesticides and drugs while its performance for the bifunctional compounds was somewhat poorer than the other tested models. Lastly, the experimental L values from this work were used to refine the LSER descriptors S, A and B of a suite of multifunctional and mostly polar compounds including pesticides and pharmaceuticals. The performance of the refined descriptors was significantly better for the prediction of the water-air partitioning, equal for the heptane-methanol partitioning but slightly poorer for the octanol-water partitioning than that of the former values. (C) 2010 Elsevier B.V. All rights reserved.