Although spectral peculiarities were noted in some high-velocity stars going back to the original Morgan and Keenan atlas of stellar spectra, the first study to demonstrate a deficiency in the abundances of elements heavier than helium in some stars was Chamberlain and Aller in 1951 (ApJ, 114, 52). They published values of [Fe/H]=-0.8 and -1.0 for HD19445 and HD140283 respectively. Not long after this an indirect relationship between stellar space motions and metallicity was first established by Roman (1954) who identified a correlation between stars with very high velocity and strong "UV excess". The high-v stars also had weak metal-lines for their hydrogen balmer-line strengths. This ultraviolet excess is caused by a differential decrease in line blanketing due to lines from metals. There is a higher density of lines in the U spectral band than in the B and V bands. As the abundance of metals go down, the U-B color becomes bluer faster than the B-V color.
Although a strict age-metallicity relationship is not in evidence for Galactic stars, the general trend (with significant scatter) that older stars have lower abundances of elements heavier than helium is clear. The correlations between [Fe/H] and kinematics were fundamental to early efforts to reconstruct the Galactic formation epoch (e.g. Eggen, Lynden-bell \& Sandage 1962, ApJ, 136, 748). As chemical evolution models were developed in the 1960s, the metal-poor end of the metallicity distribution function was one of the most poorly known observational constraints because halo stars are extremely rare in the solar neighborhood. The early lists of metal-poor objects were drawn from catalogs of high-proper-motion stars which sampled, even for high-space-velocity objects, a very small local volume.
In the 1960s and 70s, as confidence in Big Bang nucleosynthesis grew, there were concerted efforts to locate stars with primordial composition. These proposed zero-metallicity stars were called "Pop III" objects and in the 1970s and 80s there were several unsuccessful searches to identify the first examples.
In addition to the observational searches for Pop III objects there has been much interest in various theoretical aspects of these stars. Papers in the 1960s, (e.g. Peebles \& Dicke; 1968) led the way for a large number of investigations into the expected nature of the first generation of stars and the influence of these objects on structure formation in the early universe and the cosmic microwave background. The evolution of massive Pop III stars, in particular nucleosynthesis in such objects, has also been the subject of many investigations over the years. There has clearly been of tremendous interest for the last four decades to identify and characterize the extremely metal-deficient stars.
Bond (1981) compiled the work through 1980, including much of his own, identifying metal-deficient giants in the Galactic halo. Although a number of field stars with [Fe/H] comparable to the most metal-poor globular clusters were listed, very few stars with [Fe/H] $< -2.5$ were known. The metal-poor end of the metallicity distribution was considered to be deficient with respect to simple chemical evolution models and there was speculation that the lack of Population III objects indicated that the initial mass function for primordial gas was strongly biased to high-mass, short-lived stars. No or very few long-lived (sub-1M$_\odot$) stars from the initial Galactic formation epoch remained to record that period. Alternatively, it was suggested that Pop III stars were formed before the Galaxy and these objects were responsible for pre-enrichment of the material that eventually was incorporated into the Galaxy. Several groups were active in measuring abundances for the most metal-poor field stars in this era including the names Carney, Bessell, Gratton, Wallerstein, Sneden, Kraft and Perterson. Gilroy et al. (1988) in some sense brought an era to a close with a study using analysis of high-dispersion spectra of twenty of the most metal-poor stars then known and was able to hint at several interesting trends. Most notable were the lack of s-process abundance signatures in stars with [Fe/H]$<-2$, the common presence of r-process elements in metal-poor stars and significant scatter (between stars) in the ratio of neutron-rich elements to Fe.
The number of known very-metal-poor stars slowly grew in the 1980s mostly throug h searching through lists of high proper motion stars (e.g. Carney et al. 1994; Ryan, Norris \& Bessell 1991). However, in the last decade, first Beers, Preston and collaborators (e.g. Beers, Preston \& Schectman, 1992; Norris, Ryan \& Beers, 2000), then groups using the Hamburg-ESO objective prism survey plates (Christlieb et al., 1999) identified a large number of candidate extremely metal-poor (EMP) stars. Followup high-spectral resolution studies began in earnest on these candidates beginning with McWilliam, et al. (1995) and continuing to the present. Although many of the EMP candidates proved to be not exceptionally metal-poor, the high-R studies have verified (at this writing) [Fe/H] < -3 for some 35 stars. However, because the EMP candidates are relatively faint (V >11.5), for R>40000 studies requiring high (>100) signal-to-noise, the progress has been slow. There is still much work to do and it is very likely there are still quite unexpected surprises remaining to be uncovered.
The Beers et al. candidates have come from two surveys based on objective-prism plates taken with the Curtis Schmidt in the southern hemisphere and the Burrell Schmidt in the north. The initial candidates are found by inspection of objective-prism plates. Rough [Fe/H] values have been made for many of the candidates based on R=1200 blue spectra and measurements of the Ca II H+K strength (Beers et al. 1992), stromgren photometry (Anthony-Twarog et al. 2000; Schuster et al. 1996) and UBV photometry (Bonifacio, Monai & Beers, 2000; Norris, Ryan & Beers 1999). The southern hemisphere EMP star candidate list and data were published in 1992, the northern hemisphere data are just now becoming available. The southern hemisphere data alone increased the number of known stars with [Fe/H] < -3 by almost an order of magnitude. The principal reason these stars were not known in 1980 when Bond made his list of EMP stars is that they tend to be faint: 12.5 < V < 14.8. This is also the reason that large numbers of the new EMP candidates have not yet been the subject of high-resolution abundance studies.