Genomic tiling arrays, cDNA sequencing and, more recently, RNA-Seq have provided initial insights into the extent and depth of transcribed sequence across human and other genomes. These methods have led to greatly improved annotations of protein-coding genes, but have also identified transcription outside of annotated exons. One resultant issue that has aroused dispute is the balance of transcription of known exons against transcription outside of known exons. While non-genic ‘dark matter’ transcription was found by tiling arrays to be pervasive, it was seen to contribute only a small percentage of the polyadenylated transcriptome in some RNA-Seq experiments. This apparent contradiction has been compounded by a lack of clarity about what exactly constitutes a protein-coding gene. It remains unclear, for example, whether or not all transcripts that overlap on either strand within a genomic locus should be assigned to a single gene locus, including those that fail to share promoters, exons and splice junctions. The inability of tiling arrays and RNA-Seq to count transcripts, rather than exons or exon pairs, adds to these difficulties. While there is agreement that thousands of apparently non-coding loci are present outside of protein-coding genes in the human genome, there is vigorous debate of what constitutes evidence for their functionality. These issues will only be resolved upon the demonstration, or otherwise, that organismal or cellular phenotypes frequently result when non-coding RNA loci are disrupted.