Zinc is a good indicator of metallicity in damped Ly alpha (DLA) systems be
cause it is almost unaffected by dust depletion. However, the use of zinc a
s a tracer of metallicity evolution has been hampered by the difficulty of
detecting the Zn II resonance lines at high redshift. The measurement of zi
nc abundance in a DLA system at z(abs) > 3 obtained by means of the UVES sp
ectrograph at the VLT prompted us to reanalyze the full sample of zinc abun
dances present in the literature to search for a metallicity-redshift relat
ion in DLA systems. A study of the metallicities of individual systems show
s evidence for an anticorrelation between [Zn/H] and redshift, supported by
different types of statistical tools. The zinc metallicity decreases by -0
.3 +/- 0.1 dex per unit redshift interval in the range 0.5 less than or sim
ilar to z(abs) less than or similar to 3.5. This rate is in good agreement
with that found by Savaglio, Panagia, & Stiavelli in their recent study of
DLA abundances corrected for dust depletion. The present result does not re
quire a knowledge of the dust depletion pattern(s) in DLA systems. On the o
ther hand, analysis of the column-density-weighted metallicity of the sampl
e, [Z], does not show a clear evidence for redshift evolution, consistent w
ith previous studies of zinc abundances. We propose that the apparent lack
of evolution of [Z] is due to the combination of selection bias effects tog
ether with the extreme sensitivity of [Z] to low-number statistics.