Four reciprocal backcrosses were obtained in 1978 between the susceptible cultivar « Roazon » and the resistant cultivar « Carala », the former being on the cytoplasm of Aegilops ventricosa Tausch. From each backcross two lines were selected, one with a short incubation time or LPI (R line), the other with a long one (L line). These eight lines and the two parents were grown in Versailles during the 1981-1982 season on small plots with four replications. The first replication was used as the check (T), the second (plots A. EPI) was infected with a suspension of spores before the heading stage, and the third (plots EPI) at the heading stage. On each plant of these three replications the following characters were scored : length of stem (H), thousand kernel weight (TKW) and number of seeds per ear (NSE). On the second and third replications the diseased area was evaluated regularly. On the fourth replication which was kept uninfected, small leaf fragments were collected and set on benzimidazole medium in controlled conditions. Two parameters were obtained after four and six days respectively : LPI as the percentage of visible symptoms and Pmax as the average length of the necrotic area. In all the calculations these parameters have been transformed respectively into Arc sin LPI and (2 Pmax) to avoid bias in the variances. Disease progress on the foliage in the field was characterized on replications 2 & 3 by two parameters : - the area S limited by the curve Y = ( formula see attached document ) (Y being diseased leaf area) and by the abcissa, where x is biological time tB in accordance with the SHEARER and ZADOKS formula for the latent period (see diagrams 4A, 4B, 5A, 5B), - the slope sr of the regression lines between tB and Log ( formula see attached document ). Under controlled conditions, significant differences between lines were observed for LPI and Pmax. A highly significant correlation coefficient was found (table 3) between LPI and Pmax (r = 0.79) in 1982. The same was true for the pooled observations of 1981 and 1982 (r = 0.87). Depending of both these parameters the best lines were L12, L22, R67. In the field, the check plots showed no significant correlation between H or TKW and LPI or Pmax values. The TKW values were higher for L than for R lines. In the A. EPI plots, significant differences were observed between lines for S (table 3) and sr (table 9). The correlation coefficient between S and Pmax (table 3) was highly significant (r = 0.88). TKW values of these plots were partially explained by LPI, Pmax and TKW values of the check plots (multiple correlation coefficient r = 0.87). The relative reduction of TKW (RTKW) was partially explained by LPI and Pmax (multiple correlation coefficient r = 0.83). The percentages of contaminated seeds was significantly correlated with Pmax (r = 0.83) and with S (r = 0.88). The seedling test confirmed the value of the lines L 12, L22 et R62. In EPI plots, significant differences were also observed between lines for S (table 3) and sr (table 10). There were significant correlations between S and LPI (r = 0.74), Pmax (r = 0.87) or H (r = - 0.76). But for H the correlation coefficient was different according to cytoplasm (table 11). The only significant correlation for TKW was with TKW from the check plots. RTKW was partially explained by LPI, Pmax and H (multiple correlation coefficient r = 0.84). The best line was L22 with a RTKW value of 0.95. On the ears the symptoms spread in the opposite way to that on foliage, their appearance being faster for the L than for the R lines. The effect of artificial infection in the field depended to a large extent on the growth stage of the plants at infection. This was particularly true for the relation between H and S. The parameters LPI and Pmax obtained under controlled conditions were good predictors of epidemic development in the field ; under natural condition, a slow progression of the disease on the leaves involves a low reduction of yield. Three of the lines selected in these backcrosses accordingly seemed resistant enough to avoid consistent losses of yield. Moreover a cytoplasmic effect was detected on the relation between Pmax and the development of disease on leaves (fig. 8). Finally a combination of tests under controlled conditions and in the field is suggested in breeding for partial resistance to glume blotch.