To develop a new solvent-impregnated resin (SIR) system for the removal of phenols and thiophenols from water, complex formation by hydrogen bonding of phosphine oxides and phosphates is studied using isothermal titration calorimetry (ITC) and quantum chemical modeling. Six different computational methods are used: B3LYP, M06-2X, MP2, spin component-scaled (SCS) MP2 [all four with 6-311+G(d,p) basis set], a complete basis set extrapolation at the MP2 level (MP2/CBS), and the composite CBS-Q model. This reveals a range of binding enthalpies (DeltaH) for phenol-phosphine oxide and phenol-phosphate complexes and their thio analogues. Both structural (bond lengths/angles) and electronic elements (charges, bond orders) are studied. Furthermore, solvent effects are investigated theoretically by the PCM solvent model and experimentally via ITC. From our calculations, a trialkylphosphine oxide is found to be the most promising extractant for phenol in SIRs, yielding DeltaH=-14.5 and -9.8 kcal mol(-1) with phenol and thiophenol, respectively (MP2/CBS), without dimer formation that would hamper the phenol complexation. In ITC measurements, the DeltaH of this complex was most negative in the noncoordinating solvent cyclohexane, and slightly less so in pi-pi interacting solvents such as benzene. The strongest binding is found for the dimethyl phosphate-phenol complex [-15.1 kcal mol(-1) (MP2/CBS)], due to the formation of two H-bonds (P=OH-O- and P-O-HO-H); however, dimer formation of these phosphates competes with complexation of phenol, and would thus hamper their use in industrial extractions. CBS-Q calculations display erroneous trends for sulfur compounds, and are found to be unsuitable. Computationally relatively cheap SCS-MP2 and M06-2X calculations did accurately agree with the much more elaborate MP2/CBS method, with an average deviation of less than 1 kcal mol(-1).