Although it is frequently assumed that translation does not occur in eukaryotic nuclei, recent evidence suggests that some translation can take place and that it is closely coupled to transcription. The first evidence concerns the destruction of nuclear mRNAs containing premature termination codons by nonsense-mediated decay (NMD). Only ribosomes can detect termination codons, and as some NMD occurs within the nuclear fraction, active nuclear ribosomes could perform the required detection. The second evidence is the demonstration that tagged amino acids are incorporated into nascent polypeptides in a nuclear process coupled to transcription. The third evidence is that components involved in translation, NMD and transcription colocalize, coimmunoprecipitate and co-purify. All these results are simply explained if nuclear ribosomes scan nascent transcripts for premature termination codons at the site of transcription. Alternatively, the scanning needed for NMD might take place at the nuclear membrane, and contaminating cytoplasmic ribosomes might give the appearance of some nuclear translation. We argue, however, that the balance of evidence favours bona fide nuclear translation.