csplines

The smooth csplines option connects consecutive points by natural cubic splines after rendering the data monotonic on x (see smooth unique). The smoothed curve always passes through the data points, so closely-spaced points may generate local bumps and excursions in the smoothed curve.

splot ... smooth csplines with lines fits splines to the x, y, and z coordinates of successive data points. Unlike 2D csplines, the points are not sorted first so it is possible to fit splines to a trajectory containing loops. In the general case three separate sets of spline coefficients are generated, each treating one coordinate x, y, or z as a function of a shared implicit trajectory path parameter. This is equivalent to the 2D plot ... smooth path option.

In the special case that the curve lies in the xz, yz, or xy plane then only a single set of spline coefficients is generated. This allows you to generate a stack of smoothed curves in 3D where each one replicates the spline fit you would have obtained from a 2D plot of the coordinates in projection.