When equipped with motion and force sensors, box-trainers can be good alternatives for relatively expensive Virtual Reality (VR) trainers. As in VR trainers, the sensors in a box trainer could provide the trainee with objective information about his performance. Recently, multiple tracking systems were developed for classification of participants based on motion and time parameters. The aim of this study is the development of force parameters that reflect the trainee's performance in a suture task. Our second goal is to investigate if the level of the participant's skills can be classified as experts or novice level. In the experiment, experts ({\rm n} = 11) and novices ({\rm n} = 21) performed a two-handed needle driving and knot tying task on artificial tissue inside a box trainer. The tissue was mounted on the Force platform that was used to measure the force, which the subject applied on the tissue in three directions. We evaluated the potential of 16 different performance parameters, related to the magnitude, direction, and variability of applied forces, to distinguish between different levels of surgical expertise. Nine of the parameters showed significant differences between experts and novices. Principal Component Analysis was used to convert these nine partly correlating parameters, such as peak force, mean force, and main direction of force, into two uncorrelated variables. By performing a Leave-One-Out-Cross Validation with Linear Discriminant Analysis on each participants' score on these two variables, it was possible to correctly classify 84 percent of all participants as an expert or novice. We conclude that force measurements in a box trainer can be used to classify the level of performance of trainees and can contribute to objective assessment of suture skills.