The test was run during the period 2009-2012 in the Plant Breeding Station of Cereal Research Company in Szeged. The previous crop was all year’s oil rape. The location of experimental fields [9] is several hundred meters from the GPS coordinates: 46˚11'42.15"; 20˚8'56.13". The field is in the Tisza river valley, it is alluvial with medium to high clay content, and has high humus content (3% - 4%), and a deep production depth of about 1 m with excellent water economy. The experimental field was eight ha, in the middle a road that divided it into two parts side by side, where wheat and oil rape alternated. The soil was fertilized at 60 + 60 + 60 kg a.i. for NPK (nitrogen, phosphorus, potassium) in the autumn before plowing [mid-September) and an additional 60 kg N was added at the beginning of April. The yield of grain each year also depends on the amount of rainfall, with the yearly means of grain between 4 and 10 t/ha. Sowing time was in the middle of October with a difference of several days between years, which was the optimal sowing time. The seeds were sown by Rowseed TC single rod seed machine from Wintersteiger GmbH, Ried, Austria.

In the artificially inoculated plots with two row plots were used, the row spacing was 20 cm, and the plot length was 1.8 m. For the spawn method plots of three rows were sown. All tests were replicated twice, in a randomized factorial block design (e.g. two plots per treatment). In one plot two groups of heads were inoculated with the same isolate, together 8 groups of heads were treated with the four isolates. the two factorial replicates within the same plot were averaged, and so two replicates from the two plots were the entries into the statistical analysis.

The years 2009, 2011, and 2012 were dry to very dry, with a May-June precipitation of 40 - 50 mm, but 2010 was extra wet with 290 mm precipitation in the same period. 2009 was hot in the inoculation period, 2011 and 2012 were rather cool with 18˚C - 20˚C daily maximums for one week-10 days.

20 Hungarian and 20 Austrian winter wheat heat genotypes were selected for the experiment considering a wide variability in resistance. As the only trait for selection was to provide a high variability, the other traits were not tested like quality, yielding ability etc. In the Hungarian genotypes earlier tests were considered from the very susceptible to highly resistant. The Austrian genotypes were selected based on earlier experimental data. The Hungarian materials were about 5 - 7 days earlier than the Austrian ones, however the two inoculation dates we applied gave similar results, except several highly resistant materials from the Hungarian breeding program. The name of the cultivars is printed in Table 2 etc. From the tables it is clear that the cultivars and lines had a large variability for FHB resistance, e.g. they represented a large variability that allowed to compare behavior of cultivars for resistance and their stability under different epidemic conditions. Otherwise the combination or the experimental entry number is given. Generally the Hungarian genotypes were earlier flowering, the Austrian were about one week later. The higher number of genotypes was necessary as sound conclusions can seldom be made based on a low number of genotypes.

The origin of the isolates (Table 1) is: F. graminearum: No. 12377 (from maize seed; Vésztő, Hungary, 1978), No. 46.06, [wheat grain, Kiszombor, Hungary), No. 13.05 (wheat grain Kiszombor, 2005), F. culmorum No. 12375 (wheat root; Szeged, Hungary, 1978) and No. 12551 (wheat stalk base; Szeged, Hungary, 1978) and No. 89.4 (INRA Rennes, France, 1986). All are of DON type; however 89.4 can also produce nivalenol. The F. graminearum isolates belonged to the F. graminearum strict senso of the F. graminearum species complex. The isolates were monosporic, and were stored in a freezer at −80˚C.

Inocula were produced by the bubble breeding method in 10 L heat stable glass balloons [1] [6] . After one week of aeration the suspension was checked for aggressiveness at original concentration and dilutions with ion changed water at 1:1, 1:2 and 1:4 dilution ratios [53] . The test results suggested us to use or discard the given inoculum. The seedling tests and head tests showed most cases close results and founded this way of doing [49] . The inocula were kept until use in a refrigerator at 4˚C. The isolates were monosporic, and were stored in a refrigerator at −80˚C. Inocula contained amixture of mycelium and conidia. As the flowering period was about 10 days long with some differences in the years tested, only such an amount was taken out from the inoculum that was necessary for the given day, while the rest was kept in the refrigerator. This way the identity of the inoculum and its aggressiveness could be kept stable at the original aggressiveness level. As surprises seldom occurred, at the end of the inoculation period the rest of the suspensions were tested again to check aggressiveness. In this test no loss of aggressiveness was found, so the data could be used for evaluation.

Inoculation was performed accordingly at the full flowering stage (Feekes scale 10.51). This is the most sensitive phase of the plant, and therefore chosen as the time of the inoculations [1] [50] [51] independently which inoculation method is used. They all have different variants in their methodology, some art of spraying, the point inoculation and the spawn method. Since FHB pathogens are not specialized [24] [33] [44] [45] [52] , most researchers use single isolates or a mixture of isolates to balance possible variation between isolates [35] [51] . We have found that the aggressiveness of a given isolate varies strongly from year to year [6] [24] [25] [33] . Prediction based on conidium concentration was not possible except when the same inoculum was diluted. The results were rather different, even showing a reduction in values [43] . For this reason the aggressiveness test and not the conidium concentration was used [43] [53] to decide whether to use a specific inoculum or not. Since resistance expression depends also on the level of aggressiveness, the mean values of more epidemics generated from the mean of several independently used isolates (inocula) always gives more precise information on the level of resistance than a single isolate with its more or less accidental aggressiveness level or a mixture that represents only one aggressiveness level. The more epidemic time periods modeled by the use of more (in our praxis four) isolates, has proved their scientific value (more exact phenotyping for mapping populations and resistance screening of genotypes). So it was useful both in practical field work and also in genetic research. Since no environmental interaction was found between epidemics caused by different isolates (the same conditions applied for all treatments), the influence of aggressiveness differences could be studied much more precisely than with a mixture of isolates where values were observed for only one aggressiveness level.

Suitable plots were selected for next day inoculation in the afternoon before The necessary amount of inoculum was counted and prepared for the inoculation step next day and was kept until next morning in a cooler at 4˚C. The suspensions were stored on the field in cooled bags to avoid warming.

1) Spraying method combined with polyethylene bag coverage [1] [24] [25] [33] . The inoculation was performed with pure isolates, without any mixing. This way four epidemics were tested at the same time since the inocula exhibited different levels of aggressiveness. No environmental interaction was registered, as the different isolates were exposed to the same environment. In a single plot, the four isolates collected into two bunch of 15 - 20 heads apiece were sprayed from all sides to allow all spikelets to be covered. A group of heads was sprayed by 15 - 20 ml suspension since the head sizes of the cultivars were different. In the hot year of 2000 it became clear that 24 hr. time periods were not nearly enough to cause severe disease. On one occasion we tested 18 hrs. of coverage, but the results were discouraging (unpublished), therefore no further test was made in this direction. 48 hrs. coverage since 2000 by polyethylene bags secured appropriate infection also under dry and warm conditions [see Figure 1(c) in Mesterházy et al. [6] . After removing the bags the sprayed heads remained loosely bound until harvest. Grausgruber et al. [54] tested 9 genotypes with the spraying and bag method compared with the cotton method. He received a fairly good correlation between the two methods, but FDK and DON were not tested.

2) Spraying method combined with mist irrigation. Inoculation was applied in the same manner; however following inoculation mist irrigation was used in the next two days after each inoculation event according to Buerstmayr et al. [12] . The labeling was the same as in Method 1.

3) Spawn method. The modified versions of the method are used, especially by breeders. This was the reason we included is test. Earlier we tested the spawn method using F. graminearum infected maize or other cereal grains [35] [51] . We had similar experience in several instances as de Villier [55] . Lemmens (pers. communication) in dry winter observed that no perithecia production was found; therefore the inoculum for inoculation in spite of misting was very sparse. Dill-Macky [35] was disappointed with the use of infected grains. However, Thomson [56] compared a spawn method with the spray and bag method and found significant, but not very high Spearman correlations between r = 0.41 and 0.83 in 2008, and r = 0.22 to 0.54 in 2009. Lemmens et al. [39] also came to a similar conclusion. Here more severe symptoms were achieved with the bagging method. This version did not have much success also in Szeged; therefore the maize stalk residue was used in this test (Figure 1). At the end of March corn debris was collected on another field and deposited into the experimental field, the rate about 20 - 30 stalk pieces of 15 - 30 cm length per square meter followed by application of misting 2 - 3 times a week after the stand had closed. In 2010 the perithecium production on the residue was good and wide spread (right picture of Figure 1), but a right inoculum density control was not made as no literature sources reported about it. In 2011 and 2012 the number of perithecia in spite of the watering was sparse, but visible infection

developed. 2015 was very interesting (not this test), where larger amount of residues were on the soil surface, perithecia in Mai were very sparsely, but following a week humid weather 90% of the heads had very severe infection. It seems that a direct and close correlation between perithecia number in infections severity is not always the case. At flowering the spawn test received twice about 30 - 30 mm was given to enhance infection for early and late genotypes, respectively. The two days misting protocol thereafter was also used here two weeks. About two weeks after flowering, two groups of heads were labeled the same way as that the sprayed heads were.

The visual symptoms were rated according to Mesterházy [24] and Mesterházy et al. [6] . For spawn method the groups of heads were evaluated at dates like the other groups of heads in artificial inoculation to receive comparable data. The flowering dates corresponded as between the artificial an spawn method test was 3 m distance. The difference between flowering dates maximum one day difference was found. The data show the percentage of visually diseased spikelets, which correspond to the mean disease index. At harvest time the groups of heads were cut with a short stem that allows easier threshing [Seed Boy, Wintersteiger AG., Ried, Austria). All heads in the group were threshed. The air regulator was fully open to retain all small, infected or healthy, but shriveled grains. A fine cleaning was made using an EtsPlaut-Aubry air separator [41,290 Conan-Oucques, France). Wind speed was finely regulated to keep all infected and shriveled light grains. From the cleaned grains FDK was visually estimated as a percentage value (0% - 100%). Only the rose, white colored grains were considered together with the light infected whitish powdered grains. The normally colored, but shriveled grains were not included. This latter symptom is an indirect effect of the bleaching, meaning that the head part above the infection point of axis dies, and healthy, but shriveled grains are harvested.

For DON analysis the two samples of an isolate or the two groups of heads of spawn method from a given plot were pooled and mixed. Of this 6 g was separated, and milled with a Perten Laboratory mill (Laboratory Mill 3310, Perten Instruments, 126 53 Hägersten, Sweden). The details of the DON analysis are given in Mesterházy et al. [25] . The only difference is that the measurements were made by an Agilent Infinity 1260 HPLC analyzer (Agilent Technologies Santa Clara, California, USA).

The data of the two groups of heads for an isolate were averaged for FHB and FDK; they served as entries for the ANOVA analysis. Since for FHB a bunch was evaluated 4 - 5 times depending on the season, they represent the mean of 8 - 10 data. In FDK two data play the same role. For DON the grains of the two samples per isolates from a plot were pooled, so their DON content serves as entries. By this way 3840 data points were statistically evaluated for FHB, FDK and DON.

The stability of the behavior of genotypes was calculated according to Jalaluddin and Harrison [57] , and Eberhart and Russel [58] . These methods were developed to measure yield stability during variety testing. We applied it for the first time to test disease resistance stability [Mesterházy [24] . The procedure is as follows: the columns represent the cultivars [40] , while the rows contain the data for the individual epidemics [36] including years, epidemic severity, and methodical variants. So the data set is large enough to allow solid conclusions. The regression for each cultivar and trait will be counted between the mean of cultivars for 36 epidemic situations and the individual variety data. The “b” value of the linear regression is the stability index. When it is lower than one, it is positive, when higher, it mean very sharp responses under different conditions. The mean of all genotypes is one.

Correlation and regression analyses (Pearson) were performed using the built-in functions of Microsoft Excel. The four-way analyses were conducted via Microsoft Excel with the functions given by Sváb [59] and Weber [60] . When comparing the slopes for significance, the functions from Sváb [59] were used. In several cases, the statistical program SPSS (SPSS Hungary 1115 Budapest Bartók Béla street 105 - 113) was used.

The template is used to format your paper and style the text. All margins, column widths, line spaces, and text fonts are prescribed; please do not alter them. You may note peculiarities. For example, the head margin in this template measures proportionately more than is customary. This measurement and others are deliberate, using specifications that anticipate your paper as one part of the entire journals, and not as an independent document. Please do not revise any of the current designations.

The FHB data reveal very large and highly significant variety differences (Table 2), such as FHB index values ranging from 3.62 to 16.74 (Incidence*severity). The spam and spray+bag method gave similar results [11.6% and 10.2%); the spray + misting gave about 50% of these (6.78%).

The correlations between the mean of the three methods had medium closeness (between r = 0.55 and r = 0.73, p = 0.001), meaning that with some difference show more or less similar ranking of the genotypes. The lowest correlation (r = 0.55, p = 0.001) was calculated between spam and spray + bag; a somewhat closer correlation (r = 0.66, p = 0.001) between spray and bag and spray + misting and the closest correlation (r = 0.73, p = 0.001) between the two mist irrigated inoculation methods. It seems that based on visual assessment none of the possible testing methods should be excluded. A large variation was observed between the three inoculation methods. The most resistant cultivars generally have low variance except ones such as Gallus or RSt/NB a. RSt/NB b, which react the opposite way, the lower data are at the spray + bag method, the other methods give higher indices. Among the medium susceptible genotypes GK Hattyu has a low variance with very similar performance under different inoculation methods. The susceptible GK Kalász presents the same very low variance indicating similarly high disease reactivity for the different methods. Therefore, a higher susceptibility does not mean a higher variance even the correlation between them is r = 0.56 (p = 0.001).

The influence of epidemic severity (Table 3) on the resistance expression shows significant differences is mean disease index. The mean of the lowest severity is 5.78% and the highest 15.97%.

The correlations between the data of the three lower severities is similarly higher than r = 0.90, p = 0.001 (Table 3), only the very high severity data correlate less with the lower epidemic data, whereas the correlations between medium and high and high and very high are closer. For this reason the correlations and resistance expression data show an interaction with each other. The variance shows a rather high variability in disease reaction. The mean FHB performance correlates with the variance r = 0.39 that shows only a loose correlation. In every resistance class we find very stable reactions, but the probability is higher in the more resistant genotypes.

The FHB variety reactions in the four year were very divergent [not shown in detail), in 2009 the mean disease index was 2.48%, in 2010 it was 17.73%, in 2011 it was 11.94%, and 6% in 2012. The very low values meant a disease index of less than 1%. The correlation coefficients between the data of 2009 and other years were between r = −0.15 and 0.12, which were not significant. This indicates that with very low disease index data without high differentiation no effective differentiation is possible, therefore a positive correlation with the data for other years is not in sight.

The mean FDK values again show great variability, with a minimum value of 1.13, and a maximum of 19.26%, meaning a nearly 20-fold difference (Table 4). The most severe infection was recorded in the spray + bag method with a value of 19.41%. Here the differentiation was the most successful with values between 1.38% and 34.3%, a 30-fold difference. The LSD is low, with a value of only 1.41%. The mean value of spray + misting method was only 5.72; the spawn method was only 4.62%. It seems that for the FHB data the weight of the different methods diverges we found. The correlations show a medium (r = 0.55) closeness between the two inoculum spray methods (p = 0.001), but the correlation was not significant between spraying + bag and spawn method (r = 0.27), but higher (r = 0.63, p = 0.001) between the two mist application methods. The ranking of the cultivars is somewhat different, with the most resistant genotype being Nobeoka Bozu containing Ringo Star and two other lines combined with Zugoly on the third and fourth places. GK Fény, one of the most infected cultivars for FHB (but for FHB it was ranked as a moderately resistant cultivar), now ranks at the beginning of the table with 7.4%, with a worst value of 19.26%. The other two Nobeoka Bozu cultivar combinations rank in 3rd and 4th place. The variance varies between nearly zero to 215, with a mean of 81.9. It correlates with the average FDK performance of r = 0.43 (p = 0.01). The tendency is similarly apparent as with the disease index. The moderate correlation means that the more resistant genotypes tend to have significantly lower variance than susceptible ones, but a low variance genotype can be identified in all resistance classes. So the line RSt/NB b has a variance of 0.69, with very low FDK values at all inoculation methods. GK Petur, a medium susceptible variety, is also stable with a variance of 27. GK Csillag has a similar resistance with an even a lower variance of

^*Spray + bag, spray + misting, spawn + misting.

21 showing a very moderate variance to the general mean of 82. GK Kalász and GK Él et al. so show very low variance with a value of about 20, indicating a more uniform reaction until different inoculation is applied. On the other hand, SzD 5417 with its variance of 215 shows high susceptibility to the spray + bag method, but a reasonable good value for the spawn method. Therefore inoculation method may prefer or do not prefer given genotypes and may be a source of error in resistance tests.

The mean epidemic severity means were 3.33%, 6.92%, 10.47% and 18.95% (Table 5). The epidemic severity also influences genotype ranking. In spite of this, the correlations are close and significant. However, the lower epidemic severity shows a less close relation with the very high severity (r = 0.68) compared to the high or medium level epidemics (r = 0.79, 0.89, p = 0.001, respectively). This is similar we found for FHB. Moreover, the correlations between the higher severity classes were closer, between 0.80 and 0.92. We see that the variation for genotypes at low severity was between 0.78% and 7.68%, for the highest severity they were between 1.85% and 31%. The mean is 51, and the range is between 0.25 and 141. The correlation between FDK mean and variance is nearly double, with a correlation of r = 0.77 (p = 0.001), which means a good agreement between resistance level and variance, e.g. the variance is lower for the more resistant genotypes than what we found for the more susceptible ones. The susceptible cultivars GK Kalász and GK Élet which have low variances as measured by the methods approximate to the average. SzD5417 also has a larger variance in this test, e.g. it reacts very differently under different disease pressures.

The year effect on the mean value of FDK is rather different. In 2009 it was 1.4%, in 2010 it was 11.03%, in 2011 it was 14.40% and in 2012 it was 12.83%, the last three years being at the same level (data not shown in detail). The LSD 5% is 0.44% with the yearly values significantly differing from each other. The yearly data show a correlation of r = 0.56 between 2009 and 2010 and r = 0.39 between 2011 and 2012. The other correlations are between 0.15 and 0.17, but are not significant. The variance mean is 65, but the variation width is between 2.86 and 279.4. The more resistant cultivars have generally low values, in the first 20 cultivars only one surpassed the average of 65.4. In the second part 9 genotypes had values lower than the mean, and eight had a variance of over 100. Within this group two have very high levels of 240 and 279.

The genotype differences are large and highly significant. The most resistant genotype has a value of 1.8 mg/kg; the most contaminated had 13.9 mg/kg across 36 epidemic situations. The highest DON contamination was found for the spray + bag inoculation with a 14.47 mg/kg for the spray + misting gave 4.33 and only 2.11 mg/kg was found for the spawn method. The differences are highly significant. The data clearly show that the worst differentiation was found for the spawn method and the situation is somewhat better for the spray + mist method. It is remarkable that the best lines bread in Szeged performed very well in the test and produced the least DON among all genotypes tested. The correlations show the closest correlation between spray + bag and spray + misting methods; the other relations are less close. The variance of the three methods show a close correlation

All are significant at p = 0.001.

with the means for genotypes (r = 0.81, p = 0.001), indicating that the most resistant genotypes have less DON and also lower variance values. It is remarkable that the variance correlated best with the DON data of the spray + bag method, medium with the spray + misting inoculation and significant relation was found with the spawn method. However, among highly susceptible cultivars a low variance may also occur such as with GK Kalász where the reaction to different methods shows relative high values (Table 6).

The epidemic levels differed greatly, with the general mean values of 2.41, 4.43. 7.04 and 13.99 mg/kg (Table 7). All correlation coefficients are significant at p = 0.001. The values were less close between low and very high epidemic severities than between low and medium, medium and high and high and very high classes. The variance correlated well with mean performance (r = 0.76, p = 0.001), but there are exceptions such as GK Kalász or GK Holló where the variance is low, but DON contamination is high in all epidemic situations. We must add that at high epidemic severities the correlations between variance and mean performance are significantly closer. However, there are highly susceptible genotypes with lower variance (GK Kalász, GK Piacos) that have relative high values at all epidemic severities. It seems that different epidemic severities across years provide useful differentiation also at lower epidemic level.

The year influenced the DON contamination significantly (Table 8). For the FHB visual data and FDK we did not provide detailed data in this respect, but we do so here, because only limited information is present for DON and its significance is large. In 2009 the mean was 0.98 mg/kg, in 2010 6.30 mg/kg, in 2011 15.12 mg/kg and in 2012 4.30 mg/kg were measured. The correlations were all significant, but they varied between r = 0.34 and r = 0.78., much larger that the epidemic severity data did. Looking at the correlations between the values for individual years and mean, 2011 gave the closest relation since this year produced the highest toxin concentration. Of the three traits we received the highest correlations with DON.

The comparison of the three inoculation methods (Table 9) shows the mean data for the three traits and methods that among inoculation methods the spray + bag method showed the best differentiation for all traits. Of the four years except 2009 all gave the best differentiation of the genotypes. The two other methods normally gave much lower values. The spray + misting method gave acceptable differentiation for only two years, but a much lower level than the spray + bag method did. The least effective was the spawn method, where some differentiation was found during one of the four years. The different response of the traits is also clear. In 2010 the spawn method surprisingly showed severe infection. However, the FDK correlated well with the 2011 value and the DON

^***p = 0.001.

^***p = 0.001, ^*p = 0.05.