Osteoarthritis is a joint condition affecting an estimated eight million people in the UK. The kinematics of walking and the impact experienced are thought to play an important role in the initiation and progression of the disease. Previous studies have looked the effect of osteoarthritis on the kinematics of walking in a laboratory environment. This work is part of the Newcastle Thousand Families Study which has followed a cohort of 1142 members since birth in 1947. Optoelectronic gait analysis methods are unsuitable for this environment, so inertial measurement units are being used. This study focuses on the validation of a protocol using inertial sensors to assess gait in the clinical environment. The sensors measure orientation in three dimensions. Our hypothesis was that an attachment position that minimises the movement of the sensor relative to the segment during gait was more important than the proximity of the sensor to anatomical landmarks. The effect of sampling rate, fatty tissue movement and material type were also tested Seven sensors (Xsens, Netherlands) were attached to participants on top of the foot, on the tibial plateau, on the lateral surface of the femur 10cm proximal to the lateral epicondyle, and over the sacrum. Attachment is by Velcro straps over the top of clothing for the waist, thigh and shank sensors, and with double-sided hypoallergenic tape on the foot. Four calibration movements are performed followed by a walking trial of ten paces down a corridor at a self-selected speed. Data is recorded wirelessly at a sampling rate of 50Hz. The calibration movements and trials are repeated twice and the time taken is 20 minutes. Measurement of the joint angles in the sagittal plane was used to assess the effect of changing the sensor position, simulating fatty tissue movement, and variation of material type underneath the sensor. The foot and thigh sensors were displaced in the distal direction by up to 10cm, the shank and waist sensors were displaced in the proximal direction by 5cm. Material types of different elasticity were tested. Fatty tissue movement beneath the straps was simulated using hydration gel packs. Each attachment scenario was repeated five times on a single subject. A “normal” attachment scenario was used to establish a baseline for repeatability of hip, knee and ankle angle measurement (mean±standard deviation of 49±1.28°, 61.5±1.28° and 33.5±0.69° respectively). Repeatability is comparable to that reported for an opto-electronic system (45±1.8°, 63±1.9° and 36±1.5°). Displacement of the foot, shank and waist sensors had no effect on the repeatability. Displacement of the thigh sensor decreased the repeatability for the knee and hip joint angles (52±3.22° and 62.5±2.91°). As the thigh sensor moved closer to the knee the movement artefact experienced increased. Altering sampling rate and simulated fatty tissue did not decrease repeatability. Of the materials tested, denim had the greatest affect, decreasing hip and knee angle repeatability (50.0±2.04° and 61.0±1.75°). A sensor attachment position that minimises sensor movement relative to the segment has been shown to produce the greatest repeatability, irrespective of their proximity to bony landmarks. This is particularly true for the femur sensor.