The existence of discontinuities has first been observed by travel
time studies (Gutenberg, 1926; Byerly, 1926). In these studies, a kink
within the travel time curve was interpreted as a discontinuous change
in the velocity-depth curve. Later, these early observations were
interpreted as a travel time triplication due to an abrupt change of velocity
(Hales, 1972).
For the observation of triplications a dense network of seismometers is
necessary. Additionally, an accurate knowledge of the seismic velocities
beneath the stations is essential to correct the travel times. The
prerequisities made these studies extremely difficult for a long time.
Later, the long range seismic experiments with chemical explosions
(e.g. Early-Rise experiment, (Green and Hales, 1968)) or nuclear
explosions (e.g. PNE program (USSR) (Egorkin and Pavlenkova, 1981);
GNOME explosion (US), (Herrin and Taggert, 1962)) gathered
information on the structure of the mantle down to depths of 
700
km. Observations of earthquakes (Lehmann, 1959) or explosions
(Willmore, 1949) at permanently installed seismic stations gave good
knowledge of the structure of the upper mantle on a more regional scale than
the large experiments.
Globally distributed stations reveal a smoothed average structure on a larger
scale. Due to the long horizontal distance the waves travel within the
heterogeneous upper mantle compared to the depth where the rays turn from
downgoing to upgoing (e.g. 2000 km horizontally versus 400 km vertically),
lateral inhomogeneities in the upper mantle reduce the resolution of this
method.
The sharpness of the discontinuity cannot be determined directly, but the
study of waveforms in addition to the pure travel time observations can give
estimates on its thickness (Walck, 1984).