Several human hereditary neuromuscular and neurodegenerative diseases are caused by abnormal expansion of triplet repeat sequences (TRSs) CAG/CTG, CGG/CCG, or GAA/TTC on certain chromosomes. It is generally accepted that multiple slippage synthesis accounts for the instabilities of TRS. Earlier in vitro experiments by Behn-Krappa and Doerfler showed that TRS with high GC content can be expanded. In contrast, here we demonstrated that certain AT-rich TRSs, (TTC)17, (GAA)10/(TTC)10 and (GAA)17/(TTC)17, were also expansion-prone in PCR. With respect to the sequence of TRS, surprisingly, we found that the AT-rich (GAA)17/(TTC)17 extended more efficiently than the GC-rich (CAG)17/(CTG)17. This strongly suggested that the AT content of the repeat may influence TRS expansion. Furthermore, to examine the expansion of single-stranded TRS, we showed that only (TTC)17, but not the complementary (GAA)17, can be expanded. This suggested that a T-T mismatch may stabilize compatible secondary structures, most likely hairpins, for slippage synthesis. However, another poly-pyrimidine TRS, (CCT)17, is not amplification-prone in PCR. Due to the high C-content, this TRS is unlikely to adopt hairpin structures at the high pH used for PCR. Thus, the single-stranded PCR experiment may serve as an indirect assay for the ability of a sequence to adopt a hairpin conformation. When amplification was performed in reactions using Klenow DNA polymerase, only the double-stranded TRSs can be expanded. The reaction rate for (GAA)10/(TTC)10 was slower than for (GAA)17/(TTC)17, suggesting that the length of the repeat may be important for the amplification of TRS. The findings of these in vitro experiments may aid in understanding TRS expansion in vivo.