β-glucan, a soluble fiber found in oat (Avena sativa L.) grain, is good for human health, and selection for higher levels of this compound is regarded as an important breeding objective. Recent advances in oat DNA markers present an opportunity to investigate new selection methods for polygenic traits such as β-glucan concentration. Our objectives in this study were to compare genomic, marker-assisted, and best linear unbiased prediction (BLUP)–based phenotypic selection for short-term response to selection and ability to maintain genetic variance for β-glucan concentration. Starting with a collection of 446 elite oat lines from North America, each method was conducted for two cycles. The average β-glucan concentration increased from 4.57 g/100 g in Cycle 0 to between 6.66 and 6.88 g/100 g over the two cycles. The averages of marker-based selection methods in Cycle 2 were greater than those of phenotypic selection (P < 0.08). Progenies with the highest β-glucan came from the marker-based selection methods. Marker-assisted selection (MAS) for higher β-glucan concentration resulted in a later heading date. We also found that marker-based selection methods maintained greater genetic variance than did BLUP phenotypic selection, potentially enabling greater future selection gains. Overall, the results of these experiments suggest that genomic selection is a superior method for selecting a polygenic complex trait like β-glucan concentration.