We acquired more than 4 million useful sequences using a high-throughput method from a library for miRNA identification, which is constructed from a mixture of 14 RNA samples from different developmental stages. We mapped 247,410 reads to known silkworm miRNAs in miRBase (13.0), 701,913 reads to other RNA molecules based on sequence homology, and 3,219,395 reads to the silkworm genome. Our analysis identified 54 silkworm known miRNAs. A striking strand bias between miRNAs and their corresponding miRNA*s was found, and was speculated to reflect that transcripts from the passenger strand of pre-miRNAs may have important biological roles. Using an elaborate screening protocol, we predicted 287 candidate novel miRNAs (represent 116,494 short reads), and 59 of them have both miRNA and miRNA* sequences. Most of the previously identified silkworm miRNAs are cross-species conserved with a high abundance, while those predicted candidates tend to be species-specific miRNAs. Our discovery of SNPs among miRNAs implied within-species functional diversity. Target prediction uncovers that considerable silkworm miRNAs may aim at modulating more than one hormone signaling pathway components and/or hormone biosynthesis-related proteins implying their important roles in silkworm development.