Aroids corms and cormels are chemically very variable and there is significant variation of their major constituents (starch, sugars, cellulose, proteins, minerals) between genotypes. A fairly common difficulty for breeding programmes is to assess precisely these compounds as chemical analyses are too expensive for routine screening. These programmes are often based on mass recurrent selection and great numbers of hybrids have to be screened to achieve some progress. However, the wrong selection of a parent can cause a serious constraint to the development of new varieties for processing purposes. Likewise, if the table quality is not acceptable for consumers, years of intense and expensive breeding efforts can lead to complete failure. Low-cost methods for rapid evaluation of numerous hybrids are urgently needed. The present paper, assess the potential of NIRS (Near Infrared Reflectance Spectroscopy) as an alternative method for predicting these major constituents and evaluating corms and cormels quality. Models have been developed using data from 642 root crops accessions and their predictive potential has been tested on 100 varieties and hybrid lines of taro, Colocasia esculenta. The NIRS calibration results for major constituents, and their practical applications for aroids breeding and genetic improvement of quality traits, are discussed.