14 Conclusions

We have successfully developed a robust and accurate automatic rotation curve folder, and the crucial resources underwriting this success were the PS data base of 900 folded rotation curves and the corresponding raw data of MFB. Additionally, there were five key points in the development:

• The knowledge, derived from an analysis of the PS sample (Roscoe 1999A) that the exteriors of optical rotation curves conform to power laws, $V/R^\alpha = {\rm const}$, to a very high statistical precision. This information is fundamental to the mode-choosing strategy (see Sect. 7.1) and the hole-cutting process;
• The availability of a measure of internal consistency in velocity measurements in MFB's data base of unfolded rotation curves;
• The identification of an appropriate, Fourier-based, minimization principle;
• The recognition that the hole-cutting algorithm effectively partitions rotation curves into two dynamically distinct regions - the interior one of which has been identified as a dynamical transition region;
• The recognition that the class of transition regions on rotation curve interiors appears to be partitioned into intrinsically quiet and intrinsically noisy subclasses.

Furthermore, the intrinsically quiet and intrinsically noisy transition regions referred to above appear to be preferentially associated with distinct regions along the $\ln A$ axis of Fig. 6; this suggests that a study which compares the archived CCD images of the galaxies observed by MFB with the positions of those galaxies in Fig. 6 might be a very worthwhile undertaking.

Finally, Appendix E shows that, in practice, there is no advantage to be gained by using more than five Fourier modes for the folding process, whilst the final Appendix makes various miscelleneous comments of a detailed nature.

Acknowledgements

I would like to thank Vince Ford, of MSO, for his help in making available the original unreduced MFB rotation curve sample without which this work would not have been possible.