This study focuses on the use of laboratory spectroscopy to recognize the presence and abundance of salts in soils. To create soil samples with large variations in soil salinity, six different saline solutions in various densities were prepared (containing either MgCl2, NaCl, KCl, K2SO4, MgSO4 or Na2SO4). The saline solution were used to sub-irrigate three sets of soil samples with silty clay loam (n(obs) =65), sandy loam (n(obs)=41) and sand textures (n(obs)=41). After sub-irrigation, soil spectra and soil salinity were measured for each of the 147 samples. Statistical techniques were used to detect and characterise diagnostic absorption bands and to develop predictive regression models between the spectral signature and soil salinity levels. Spectral analyses revealed that salt-affected soil samples do not exhibit all of the diagnostic absorption features that were found in the spectra of salt minerals used to treat them. It also showed that the number and clearness of diagnostic bands reduces as the salt concentration in the samples decreases. Amongst the investigated salts, MgCl2 and MgSO4 were estimated with the highest accuracies (cross-validated R-CV(2)>= 0.80 and PRMSECV < 40%). For NaCl, estimation of soil salinity gave an acceptable accuracy (R-CV(2) =0.65 with PRMSECV=50.3%). Only weak relations could be established to quantify KCl and Na2SO4. The results show that the observed absorption features at further than 1300 nm were broaden, the position of maximum reflectance were shifted toward shorter wavelengths and overall reflectance were changed proportionally as salts concentration were increased in soil. The continuum-removed (CR) spectra indicate a strong negative correlation between increase of soil EC and changes in absorption bands parameters (depth, width and area). (c) 2008 Elsevier B.V. All rights reserved.