With the human DNA sequence nearing completion, the search for complex disease genes is gaining momentum, as is the debate over gene-finding strategies. This overview contrasts two pivotal methods: linkage analysis and association mapping. Linkage analysis has been used successfully to identify the genes underlying rare mendelian disorders. It has also played a role in attempts to map genes for common non-mendelian (also known as 'complex' or 'multifactorial') diseases such as psychiatric disorders. However, despite extensive efforts progress has been slow, marred by inconsistent or ambiguous results. Uncertainties about the utility of the linkage approach for complex genetic traits has spurred interest in association studies with candidate genes, as an alternate strategy. Recently, with the advent of new molecular tools, in particular high-density, single-nucleotide polymorphisms (SNPs) maps, it has been argued that, while linkage analysis may retain some role, genome-wide association studies with SNPs offer a superior strategy for unraveling genetic complexity. In this paper I review these issues, stressing the pros and cons of the various strategies. I propose that: (1) the uncertainties in association studies may have been underestimated; (2) neither method is sufficient or optimal; and (3) a joint linkage and association approach, together with genomic, statistical and computational advances, may have greater promise for understanding the genetic underpinnings of complex disorders in the new millennium.