It was previously shown that the structural ensemble of model peptides DDKG and GKDG (H. Ishii et al. Biopolymers 24, 2045–2056, 1985), DEKS (A. Otter et al. J. Biomol. Struct. Dyn. 7, 455–476, 1989) NPGQ (F. R. Carbone et al. Int. J. Pept. Protein. Res. 26, 498–508, 1985), SALN (H. Santa et al. J. Biomol. Struct. Dyn. 16, 1033–1041, 1999), SYPFDV and SYPYDV (J. Yao et al. J. Mol. Biol. 243, 736–753, 1994), VPDAH and VPDSH (B. Imperiali et al. J. Am. Chem. Soc. 114, 3182–3188, 1992) in solution contains a significant—or in some cases dominant—proportion of β-turn conformation. In this study, a protein database was searched for the above, unprotected sequences which incorporate only L-amino acid residues. Simulated annealing and 25 ns MD simulations of structures were also performed. The DSSP and STRIDE secondary structure-assigning algorithms and clustering were used to analyze trajectories and i, i+3 hydrogen bonds were also sought. The DSSP analysis showed a fluctuation between β-turn and random meander structure, although bend structures were not detected because of the insufficient length of peptide chains. This alternating trend was confirmed when the STRIDE algorithm was used to analyze trajectories, but STRIDE assigned more turn structures. The population of the strongest clusters was above 40% and the middle structures adopted β-turn structure for most sequences. These results are in good agreement with previous experimental results and support the idea of the ultra-marginal stability of turns in the absence of stabilizing long-range interactions of the neighboring segments of a polypeptide chain. However, interactions between the side-chains in tetrapeptides could also contribute to turn stability and result in unusual stability in some cases. Our observations suggest that such interactions are the consequence rather than the driving force of turn formation.