Not yet reviewed by Pith; the record is open.

The MILC collaboration computation of heavy-light decay constants is described. Results for $f_B$, $f_{B_s}$, $f_D$, $f_{D_s}$ and their ratios are presented. These results are still preliminary, but the analysis is close to being completed. Sources of systematic error, both within the quenched approximation and from quenching itself, are estimated, although the latter estimate is rather crude. A sample of our results is: $f_B=153 \pm 10 {}^{+36}_{-13} {}^{+13}_{-0} MeV$, $f_{B_s}/f_B = 1.10 \pm 0.02 {}^{+0.05}_{-0.03} {}^{+0.03}_{-0.02}$, and $f_{B}/f_{D_s} = 0.76 \pm 0.03 {}^{+0.07}_{-0.04} {}^{+0.02}_{-0.01}$, where the errors are statistical, systematic (within the quenched approximation), and systematic (of quenching), respectively. The largest source of error comes from the extrapolation to the continuum. The second largest source is the chiral extrapolation. At present, the central values are based on linear chiral extrapolations; a shift to quadratic extrapolations would for example raise $f_B$ by $\approx 20$ MeV and thereby make the error within the quenched approximation more symmetric.