SUMMARY: Although translation initiation has been well studied, many questions remain in elucidating its mechanisms. An ongoing challenge is to understand how ribosomes choose a translation initiation site (TIS). To gain new insights, we analyzed large sets of TISs with the aim of identifying common characteristics that are potentially of functional importance. Nucleotide sequence context has previously been demonstrated to play an important role in the ribosome's selection of a TIS, and mRNA secondary structure is also emerging as a contributing factor. Here, we analyze mRNA secondary structure using the folding predictions of the RNAfold algorithm. We present a method for analyzing these results using a rank-ordering approach to assess the overall degree of predicted secondary structure in a given region of mRNA. In addition, we used a modified version of the algorithm that makes use of only a subset of the standard version's output to incorporate base-pairing polarity constraints suggested by the ribosome scanning process. These methods were employed to study the TISs of 1735 genes in Saccharomyces cerevisiae. Trends in base composition and base-pairing probabilities suggest that efficient translation initiation and high protein expression are aided by reduced secondary structure upstream and downstream of the TIS. However, the downstream reduction is not observed for sets of TISs with nucleotide sequence contexts unfavorable for translation initiation, consistent with previous suggestions that secondary structure downstream of the ribosome can facilitate TIS recognition.