Among the three stop codons in the standard genetic code, UAA has seven single-nucleotide nonsynonymous neighbors (SNNs, i.e., sense codons that differ by one nucleotide from the stop codon), and UAG and UGA each have eight. Just as a sense codon can be misread by a release factor, causing a premature translation termination, a stop codon can also be misread by a tRNA, for example, misreading of UGA by tRNATrp that decodes UGG, causing a readthrough error. In both prokaryotes and eukaryotes, UAA exhibits the lowest termination error rate, and UGA the highest. However, the translation termination signal is a tetranucleotide signal instead of just a stop codon. The nucleotide immediately downstream of the stop codon, known as the +4 site, is strongly biased and has been experimentally shown to have a strong effect on translation termination. The 12 possible tetranucleotide translation termination signals, i.e., three stop codons and four nucleotides at the +4 site, are used differently among different species and coevolve with the availability of release factors, especially in bacterial species with two class-I release factors, RF1 decoding UAG and UAA, and RF2 decoding UAA and UGA. An explanation is offered for the nearly universal rarity of UAG usage in bacterial species. Problems with optimizing the translation termination signal in mRNA medicine are highlighted, and optimal termination signals are suggested for different cellular environments.