Errors can occur at any level during replication and expression of genetic information. Genetic mutations are derived mainly from replication errors and have been extensively studied in evolutionary research. However, many details of the transcription error remain largely unknown. Transcription errors can indirectly give rise to misfolded proteins and impose a load on cellular integrity. Same as the rate of genetic mutation, the rate of transcription error is shaped by the directional process of selection and the random process of drift. However, transcription errors are fundamentally different from genetic mutations in many aspects. Firstly, unlike mutations that reside in genomes, transcription errors in RNA transcripts have short life spans and will finally be degraded. Secondly, there are multiple RNA copies of one gene and only a minor portion of them may carry transcription errors. These transient natures would dilute the fitness effect of transcription errors and selection is less sufficient to push down the rate of transcription errors. To investigate the lower limit of transcription error rate in the bacterium Escherichia coli, we took a novel rolling-circle sequencing approach and have accurately identified transcriptome-wide transcription errors. Our results indicate that the lower limit of transcription error rate in the bacterium Escherichia coli is 105 x higher than the corresponding genetic mutation rate.