The aim of this study was to apply a novel sensitive technique, involving reflectance spectroscopy in the 400-2500-nm region, to assess dust loads. A spectral library was created to enable identification of the possible sources of settled dust in indoor samples -- mineral versus organic-anthropogenic. Two field experiments were carried out at different dates, the first in dwellings and the second in office environments. Two main spectral patterns were found. Type A spectra indicate a high proportion of minerals in the sample and are characteristic of dust samples taken from the dwelling environment during April (when there were 5 dust storm events). Type B spectra denote a high proportion of organic matter in the sample and are characteristic of the dust samples taken from the offices during March (when there were only 2 dust storm events). The spectral shape within the visible range can be used to estimate the relative amount of mineral and organic components in the sample. Multivariate data analysis, based on Partial Least Squares (PLS) regression, was utilized to predict the relationship between the reflectance of a dust sample and its mass. The relative Root Mean Square Error of Predictions (%RMSEP) generated for the dust sampled in dwellings (6.5%) and offices (7.0%) are quite impressive considering the relatively small amounts of settled dust and its precise gravimetric weight accurate to +/-0.01 mg (min and max values are 0.1-3.2 mg). In addition, PLS regression analysis was used to identify which variables influence dust load. Possible applications of the proposed method are discussed.