The aim of this study was to analyse to what extent the use of different loads modifies freestyle stroke and coordination parameters during semi-tethered swimming, and to examine whether those changes are positive or negative to swimming performance. First, behaviour of swimming speed (v), stroke rate (SR) and stroke length (SL) with increasing loads was examined. Secondly, mean and peak speed of propulsive phases (propvmean and propvpeak) were analysed, as well as the relative difference between them (%v). Finally, index of coordination (IdC) was assessed. Eighteen male swimmers (22.10±4.31years, 1.79±0.07m, 76.74±9.00kg) performed 12.5m maximal sprints, pulling a different load each trial (0, 1.59, 2.21, 2.84, 3.46, 4.09, 4.71, 5.34, 5.96, 6.59, 7.21 and 7.84kg). Rest between repetitions was five minutes. Their feet were tied together, keeping a pull-buoy between legs and isolating the upper limb action. A speedometer was used to measure intra-cycle speed and the test was recorded by a frontal and a lateral underwater cameras. Variables v and SL decreased significantly when load increased, while SR remained constant (p<0.05). Propvmean and propvpeak decreased significantly with increasing loads (p<0.05). In contrast, %v grew when load rose (r = 0.922, p<0.01), being significantly different from free swimming above 4.71kg. For higher loads, swimmers did not manage to keep a constant velocity during a complete trial. IdC was found to increase with loads, significantly from 2.84kg (p<0.05). It was concluded that semi-tethered swimming is one training method useful to enhance swimmers' performance, but load needs to be individually determined and carefully controlled.