Radiographs are the main outcome measure in epidemiological studies of osteoarthritis (OA). Ultrasound imaging has unique advantages in that it involves no ionising radiation, is easy to use and visualises soft tissue structures. Our objective was to measure the inter-rater reliability and validity of ultrasound imaging in the detection of features of knee OA. Eighteen participants from a community cohort, had both knees scanned by two trained musculoskeletal sonographers, up to six weeks apart. Inter-rater reliability for osteophytes, effusion size and cartilage thickness was calculated by estimating Kappa (κ) and Intraclass correlation coefficients (ICC), as appropriate. A measure of construct validity was determined by estimating κ between the two imaging modalities in the detection of osteophytes. Reliability: κ for osteophyte presence was 0.77(right femur), 0.65(left femur) and 0.88 for both tibia. ICCs for effusion size were 0.70(right) and 0.85(left). Moderate to substantial agreement was found in cartilage thickness measurements. For osteophytes, κ was moderate to excellent at 0.52(right) and 0.75(left). Substantial to excellent agreement was found between ultrasound observers for the presence of osteophytes and measurement of effusion size; it was moderate to substantial for femoral cartilage thickness. Moderate to substantial agreement was observed between ultrasound and radiographs for osteophyte presence.