Sticking of peptides (as demonstrated by peak distortion) containing a high proportion of glycine and proline residues to the capillary wall was investigated. At acid pH where the carboxy terminal proline (e.g. Gly-Pro) is protonated, distinct sticking of these peptides occurred. On the contrary, the Pro-Gly peptide yielded a well shaped peak. At alkaline pH, where the situation was reversed, the peptide sticking most to the wall was Gly-Gly, while the behaviour of the dipeptides possessing N-or C-terminal proline did not suffer from sticking to the wall.It was concluded that the separation of simple proline-and glycine-containing dipeptides can be partly optimized by manipulating the pH of the background electrolyte; particularly if cetyltrimethylammonium bromide, methanol or acetonitrile is added to the run buffer (reversed polarity mode with cetyltrimethylammonium bromide). However, all attempts to resolve more complex mixtures of proline-(and hydroxyproline-) containing di-and tripeptides failed; addition of an organic modifier, i.e. methanol and acetonitrile, hexylamine, triethylamine and cetyltrimethylammonium bromide was unsuccessful. Such separations can be materialized by using simple 25 mmol / l phosphate buffer at pH 10.5 provided that the sample is dissolved in (aqueous) 17.5 mmol / l Brij or 33 mmol / l sodium dodecyl sulfate. These systems are also applicable to large polypeptides rich in proline, hydroxyproline and glycine: typically parent collagen a-chains, their dimers, trimers and higher polymers were successfully separated.