Purpose: This work provides a quantitative assessment of the positional tracking accuracy of the NDI Wave Speech Research System. Method: Three experiments were completed: (a) static rigid-body tracking across different locations in the electromagnetic field volume, (b) dynamic rigid-body tracking across different locations within the electromagnetic field volume, and (c) human jaw-movement tracking during speech. Rigid-body experiments were completed for 4 different instrumentation settings, permuting 2 electromagnetic field volume sizes with and without automated reference sensor processing. Results: Within the anthropometrically pertinent "near field" (less than 200 mm) of the NDI Wave field generator, at the 300-mm[superscript 3] volume setting, 88% of dynamic positional errors were less than 0.5 mm and 98% were less than 1.0 mm. Extreme tracking errors (greater than 2 mm) occurred within the near field for less than 1% of position samples. For human jaw-movement tracking, 95% of position samples had less than 0.5 mm errors for 9 out of 10 subjects. Conclusions: Static tracking accuracy is modestly superior to dynamic tracking accuracy. Dynamic tracking accuracy is best for the 300-mm[superscript 3] field setting in the 200-mm near field. The use of automated head correction has no deleterious effect on tracking. Tracking errors for jaw movements during speech are typically less than 0.5 mm.