A small but growing body of research suggests that retirement and cognitive decline are related. In fact, some have argued that retirement causes cognitive decline. The aim of this paper is to add to this literature using data of older women from The Irish Longitudinal Study on Ageing (TILDA). Ordinary least square (OLS) regressions show a significant negative association between retirement duration - defined as time elapsed since last job ended - and cognition. This association is, however, very small when compared to the association of age or education with cognition. As retirement is potentially endogenous to cognition, instrumental variable (IV) methods are also used. The instrument employed is the abolition of the so-called "Marriage Bar". In simple terms, the Marriage Bar was the requirement that women leave paid employment on getting married. It was established in the 1930s and removed in the 1970s. When IV estimations are used, the effect of retirement duration is statistically insignificant. Differences between OLS and IV estimates are compared with a standard test. OLS estimates are preferred as there is no statistical evidence that retirement duration is endogenous. As OLS estimates are very small, it is extremely unlikely that working into old age has much of an impact on slowing cognitive decline amongst older people.