Variable viruses, such as human immunodeficiency virus (HIV) and hepatitis C virus (HCV), persist despite host immune responses directed against them. Numerous lines of evidence have suggested that antiviral CD8+ T-cell responses are key among these immune responses, but these vary widely in their ability to contain virus. We propose that only a proportion of responses may exert significant antiviral pressure ('driver' responses), leading to control over viral replication (protection) and/or, ultimately, selection of escape mutants. Another set of responses may exert only weak pressure on the virus ('passenger' responses): these neither protect nor select. To examine this we have analysed (using established databases of HIV and HCV sequences and cytotoxic T-lymphocyte (CTL) epitopes, and published experimental datasets) two important features--predicted binding of the epitope to major histocompatibility complex molecule and observed variability of the epitope--that might distinguish such responses. We find that a high predicted binding estimate could only explain a limited set of 'driver' responses associated with protection or selection. There is statistical evidence that readily defined (and non-protective) CTL responses target regions associated with lower levels of viral variability. Taken together, this suggests that a large number of well-documented responses may represent 'passengers' and we propose a mechanism that might explain their presence.