Apple (Malus × domestica Borkh) is one of the major fruits produced around the world. As the production and consumption of apple fruit continue to increase, improved quality control has become a main focus of postharvest science. The absence of blemishes, flesh firmness, sweetness, and acidity are closely related to consumer preferences. As an internal fruit quality attribute, sweetness can determine consumer purchases. Although sweetness can be determined by quantifying sugars via several chemical methods such as ion-chromatography, the most broadly used method to quickly determine sugars is through measuring soluble solid content (SSC) by refractometer. In apples soluble solids consist mainly of fructose, glucose, and sucrose. Fruit dry matter concentration (DMC) is comprised of all components except water, and in recent years DMCs have been studied in relation to fruit maturity as well as consumer preferences. At harvest, DMC can be regarded as a predictor of SSC after storage as the DMC correlates strongly with the SSC if no starch is present. However, no effects of 1-MCP and storage period on this relationship were detected. The traditional measurements of fruit SSC and DMC are laborious, time-consuming as well as destructive. Near-infrared (NIR) spectroscopy is a newly available method to predict fruit DMC and SSC nondestructively. Models of SSC and DMC were successfully built for individual- and multiple- cultivars, and both internal and external validation were applied to test the accuracy and precision of all the models. Under similar calibration performances, the individual-cultivar models had higher slope values of regression lines, which may indicate more accurate predictions in internal validation. However, the individual-cultivar models revealed issues of model over-fitting reference and value distribution in external validation. The multi-cultivar models were able to predict SSC and DMC. To improve the robustness of the model, variability among-trees (e.g. crop load), within-orchard variability, orchard variability, and seasonal variability have to be taken into consideration. Apples with an internal flesh browning disorder were examined by interactance NIR spectroscopy. Linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and stepwise variable selection (SVS) were applied to training and validation sets. QDA model with SVS produced the lowest misclassification rates (around 17%). These results indicated that QDA outperformed than LDA, and SVS could be used to improve the identification of defected apples. Overall NIR spectroscopy has a great potential to predict fruit internal quality non-destructively. However, more studies on model robustness and related chemometrics are required.