Winter triticale grain yield, a comparative study of 15 genotypes

Abstract

In the paper, we studied genotypic differences in the process of winter triticale grain yield formation, in terms of yield determination by its components. We applied yield component analysis on four dwarf and 11 traditional (conventional stature) genotypes; an ontogenetic approach was employed to provide the appropriate view of this process. Using the k-means clustering procedure, the genotypes were grouped subject to similar yield structure (i.e., mean values and coefficients of variation of grain yield and its components, first clustering), similar correlation matrices among grain yield and its components (second clustering), and similar results of yield component analysis (third clustering). Based on the analysis, we attempted to provide an overall general view of determining winter triticale grain yield by its components. Thus, we elucidate that the pattern of influencing winter triticale grain yield by its components, as well as the pattern of co-relationships among these traits, cannot be explained by yield structure, and hence these two processes are determined by genotypic factors. The correlation among the traits (grain yield and its components) and the results of yield component analysis, on the other hand, were significantly related in the study, so they have similar genotypic sources.

Keywords:

• Cluster analysis
• genotypic differences
• sequential development
• yield component analysis

Inrtoduction

Yields of winter triticale cultivars are still improving; new cultivars provide higher grain yield and of better quality. In COBORU (Research Centre for Cultivar Testing in Slupia Wielka, Poland) investigations, the average yield of all 26 registered cultivars ranged from 6.1 to 9.3 t ha⁻¹, depending on agronomic and environmental (location and year) conditions, but on plots with higher levels of agricultural inputs it exceeded 10 t ha⁻¹ (COBORU, 2005). The main direct criteria in plant breeding of winter triticale, as well as of other cereals, are potential yield level and stability. On the other hand, triticale cultivar yield formation has not been widely investigated in terms of contribution of yield components to yield determination. It has been proven by Giunta et al. (1999) that this process has also some genetic background. Taking this into account, we may assume that morphologically different triticale cultivars, because of various wheat and rye gene contribution, may differ in level of final grain yield determination by its components. Rozbicki (1997), in investigations on winter triticale (cv. Presto), found that grain yield was determined by number of kernels per spike and, to a much lesser extent, by kernel weight. In the opinion of Kozdój (2003), these two components are positively correlated with each other and strongly determined by genotypic factors, and hence are specific for triticale genotypes. Considering the feature of triticale spike fertility, Giunta et al. (1999) reported that this trait is more dependent on number of spikelets per spike than on spikelet fertility. Number of mature florets in triticale makes up about 40% and number of grains about 25–30% of total floret initiations. The genetic variation in morphogenetic spike potential (number of grains at harvest) arises from differences in number of floret initiations at advanced stem elongation stage as well as from number of fertile florets at anthesis (Kozdój, 2003). According to Giunta et al. (1999), the pre-anthesis period (when a large number of floret primordia become fertile florets) without drought stress during spring, is essential to obtain high triticale yields as an effect of both number of grains per m² (sink formation) and grain filling rate (source capacity) in the post-anthesis period.

An important question, probably not discussed in the literature until now, is whether the genotypes with similar average grain yield and characteristics of canopy (first component), spike (second component), and grain (third component), as well as variability of these traits, are characterized by a similar pattern of influencing grain yield by its components. In the following, by ‘yield structure’ we will understand the average values and variability (i.e., the coefficients of variation) of yield components and grain yield. If not, it would indicate that this process is determined by genotypic factors (since genotypes with similar or even the same yield structure would be characterized by different yield components contribution to grain yield determination). From this point of view, it is also interesting to compare two groups of winter triticale genotypes – traditional (conventional stature) and dwarf ones. Is the process of influencing grain yield by its components similar for them and for the genotypes belonging to these groups? Are dwarf genotypes able to fill a similar number of kernels as the conventional cultivars, since they produce more chaff and grain per biomass unit (Brancourt-Hulmel et al., 2003) with the increased availability of assimilates for developing spikes (Otegui & Slafer, 2004)?

The aim of the paper is to investigate genotypic variability contribution of yield components to grain yield determination. Fifteen winter triticale genotypes – four dwarf and 11 traditional – were studied. Moreover, we attempt to provide an overall general view of determining winter triticale grain yield by its components, and search for possible sources of this relationship. In particular, we answer the question whether the pattern of influencing yield by its components can be explained by general yield structure (i.e., average values and variability of grain yield and its components) of winter triticale.

Materials and methods

Site and experimental design

In the years 2000–2003, at Warsaw Agricultural University experimental station in the Mazovia region of Poland (52°05′ N, 20°33′ E), a two-factor field experiment, on a fine soil with a light loam texture, was carried out. Average three-year annual rainfall and temperature were 551 mm and 8.1°C, respectively. The experimental design was a randomized complete block in a split-plot arrangement with four replications within the years. Winter triticale (x Triticosecale Wittm.) genotypes were the main plots, and nitrogen rates were the subplots. The genotypes used in the experiment were released by Polish breeding companies DANKO Breeding Co. Ltd. and Strzelce Breeding Co. Ltd. Their characteristics are presented in Table I. Seeding rate was adjusted for a density of 400 viable seeds m⁻², sown on 22 September in 2000 and 25 September in 2001 and 2002.

Table I. Characteristics of the investigated winter triticale genotypes.

Genotype	Registration year	Origin	Group of maturity^1	Plant height (cm)
Alzo	1997	Ugo×MAH 9531-6	198	110
Bogo	1993	triticale Mo 7810-501×triticale Mo 7686-2	195	106
DED 798^DG,R	–	(Vero×Presto)×Pinokio	198	104
Disco	1997	(KT 77×LT 363/75)×Dagro	197	115
Fidelio^DG	1997	(135/16/82×L 627/80)×Presto ,,"	200	99
Janko	2000	MAH 19202×MAH 19267	197	117
Kitaro	1999	(Moniko×Malno)×Presto	196	109
Lamberto	1998	(CT 929/84×Moniko)×Presto	196	112
Magnat^DG	2000	{[(wheat Lanca×rye L 506/79)×LAD 627/80]×Dagro}x Presto	199	103
Marko	1997	Almo×CHD 782	197	113
Prado	1998	MAH 7387–11×MAH 7226–1	196	117
Pronto	1999	(DAD 187×LAD 1272/85)×Prego	196	109
Tornado	1996	MAH 6353-540×Mo 7877–1	195	110
Ugo	1988	/ 35/76–10×AD 206/×CT 93/76	198	119
Woltario^DG	2000	{[(wheat Lanca×rye L 506/79)×Bolero]×LAD 285}x Presto	197	99
^1Days from the beginning of the year to dough stage; ^DG: dwarf genotype;^R: removed from COBORU investigations in 2000 without being registered.

The genotypes were grown at three different nitrogen rates, i.e., 0, 80 (45 + 35), and 170 (100 + 70) kg ha⁻¹ applied twice, i.e., at the beginning of vegetation (GS (growth stage) 25–29), and at the end of the boot stage (GS 49) (Zadoks et al., 1974), broadcasted by hand as ammonium nitrate (34% N). The early spring, soil mineral nitrogen content (N_min) in the years 2001, 2002, and 2003, determined at a depth of 0–90 cm, was 87.1, 33.7, and 50.0 kg ha⁻¹, respectively.

Prior to sowing, soil was analysed for P and K contents and pH level. Amounts of phosphorus and potassium used were based on the recommendations for Polish farmers. Weeds, pests and diseases were chemically controlled as required.

Crop measurements

The following procedure for determining grain yield and its components was used. Number of spikes (X ₁ ) were counted on plant samples taken by hand from 1 square metre of each plot; the total plot area was 20 m²=2.5 m×8 m, with rows 0.11 m apart. The same spikes were threshed, and grain yield was determined (t ha⁻¹), and then corrected to a 150 g kg⁻¹ moisture basis. Grain sample of about 100 g was taken from each plot, weighed, and oven-dried at 70°C to achieve a stable grain mass to determine grain moisture content. According to Polish standards (PN-68/R-74017), weight of 1000 kernels (g), X ₃, was calculated from weight of two sets of 500 randomly chosen kernels per plot. Consequently, in the manuscript the term ‘kernel weight’ is used for the third component. Average number of kernels per spike (X ₂) was calculated on the basis of number of spikes, grain yield per square metre, and kernel weight.

Statistical analysis

To provide an overall view of a process of winter triticale grain yield formation studied as an effect of its components, statistical analysis was applied for each investigated cultivar independently. In particular, the analysis comprised correlation analysis for components and grain yield of each cultivar. The sequential analysis of influencing grain yield by canopy and plant traits developing in sequential order (so-called sequential yield analysis), thoroughly presented by Mądry et al. (2005), was applied to provide the view of contribution of yield components in yield determination; these contributions are presented in percentages of the determination coefficient for the model for grain yield versus its components. The method employs an ontogenetic approach (e.g., Dofing & Knight, 1992; Mądry et al., 2005).

The analyses were conducted for all the observations from plots and years (36 for each cultivar), treating them as a representative sample from the population generated by agronomic (treatments) and environmental (years) conditions. The scatter plots for grain yield versus each component (not included in the paper), carried out for each cultivar as the basic validation, showed that the pooled data are correct for regression analysis (Quinn & Keough, 2003).

Summary statistics (mean values and coefficients of variation for traits studied), correlation matrices, and results of yield component analyses laid the basis for comparative studies on influencing grain yield by its components for investigated genotypes. Using the k-means clustering procedure (Legendre & Legendre, 1998), the genotypes were grouped into three clusters subject to similar: 1) yield structure (summary statistics, i.e., mean values and coefficients of variation of grain yield and its components, were the grouping variables); 2) correlation matrices (correlation coefficients between the studied traits were the grouping variables); and 3) results of yield component analysis (contributions of three grain yield components to yield determination were the grouping variables). The first clustering groups genotypes that are similar subject to yield structure. The second groups genotypes that are similar subject to the co-relationships among the traits (provided that just a general but not ontogenetic view of the process of yield formation is taken into account). The third clustering groups genotypes which are similar, subject to the process of influencing grain yield by its components.

To investigate whether these three clusterings are similar, relationships between the cluster belongings of the genotypes were investigated. Three two-way contingency tables (i.e., one table for each combination of belongings of the clusters) were studied. The independence of cluster belongings was tested using the χ ²-test for two-way tables (Quinn & Keough, 2003). Since the sample size was very small (n = 15), Pearson's χ ² test could not be used. Hence, following suggestions of Quinn & Keough (2003), Monte Carlo resampling was used to generate a distribution of χ ² statistics and to assess a significance (p-value) of the statistic and the relationship.

We considered three indicator variables: clusterings I, II, and III. They refer to particular clusterings. Their possible values are 1, 2, and 3, a particular value indicating to which cluster (there are three clusters for each clustering) in a given clustering (I, II, or III) a particular genotype belongs. Dependence between the pair clustering I and clustering II would indicate that yield structure might be a cause of the correlation among the studied traits. Dependence between the pair clustering I and clustering III would indicate existence of a relationship between yield structure and the pattern of influencing yield by its components. Dependence between the pair clustering II and clustering III would indicate that the correlation pattern between the traits (grain yield and its components) determines the pattern of influencing grain yield by its components.

The computation was performed using R language and environment (R Development Core Team, 2005).

Results

Table II contains summary statistics, i.e., mean values and coefficients of variation, for grain yield and its components of 15 winter triticale genotypes investigated. Various yielding and average values of yield components were observed for the genotypes. The largest average grain yield was observed for Kitaro, the smallest for Janko. Average number of spikes per m² ranged from 361 (Disco) to ca. 464 (Fidelio and Pronto); average number of kernels per spike ranged from 33 (Janko) to more than 40 (Bogo); and average 1000-kernel weight ranged from 36.4 g (Bogo) to 46.1 g (Kitaro). In general, the genotypes studied differed noticeably in mean values of grain yield and its components (Table II).

Table II. Summary statistics of yield and its components for 15 winter triticale genotypes.

Cultivar	Grain yield g (Y)		Number of spikes m^−2 (X 1)		Number of kernels spike^−1 (X 2)		Weight of 1000 kernels g (X 3)
	MV 10^3	CV	MV	CV	MV	CV	MV	CV
Alzo	682.3	0.24	446	0.19	38	0.17	40.5	0.11
Bogo	649.8	0.24	440	0.15	40	0.13	36.4	0.09
DED798	652.5	0.19	432	0.16	36	0.16	42.4	0.14
Disco	641.7	0.20	361	0.16	39	0.14	45.6	0.13
Fidelio	666.8	0.25	465	0.15	34	0.16	42.8	0.12
Janko	583.0	0.26	417	0.18	33	0.17	42.5	0.14
Kitaro	704.6	0.20	416	0.15	37	0.10	46.1	0.12
Lamberto	657.1	0.19	458	0.13	37	0.12	39.2	0.09
Magnat	682.1	0.22	395	0.11	38	0.13	45.5	0.14
Marko	635.5	0.20	395	0.15	39	0.14	41.5	0.11
Prado	619.0	0.20	400	0.11	39	0.11	39.2	0.11
Pronto	664.3	0.26	464	0.14	37	0.13	38.7	0.09
Tornado	608.3	0.24	406	0.19	38	0.15	39.8	0.13
Ugo	666.5	0.23	443	0.13	38	0.16	39.8	0.14
Woltario	651.7	0.23	415	0.10	37	0.15	42.3	0.15

Correlation matrices for studied traits are shown in Table III. The matrix for each genotype is presented as a vector in the corresponding row. From the correlation analysis it follows that the co-relationships among studied winter triticale traits differed for various genotypes. Nevertheless, some regularities can be given. Namely, apart from Ugo cultivar, grain yield was significantly positively correlated with number of spikes per m². For all studied genotypes, significant positive correlation between grain yield and number of kernels per spike was observed. Taking into account the kernel weight, only in the case of three genotypes (Fidelio, DED 798, and Alzo) was significant correlation between this component and grain yield was not observed. Significant correlation between number of spikes per m² and number of kernels per spike was detected only for Fidelio (positive correlation) and Ugo (negative); significant correlation between number of spikes per m² and kernel weight was observed for eight genotypes (negative correlation), and between number of kernels per spike and kernel weight for five genotypes (one negative and four positive correlation coefficients).

Table III. Correlation matrices of yield and its components for 15 winter triticale genotypes.

Cultivar	r(Y,X 1)	r(Y,X 2)	r(Y,X 3)	r(X 1,X 2)	r(X 1,X 3)	r(X 2,X 3)
Alzo	0.60**	0.71**	0.23	0.01	−0.37*	0.17
Bogo	0.70**	0.71**	0.31*	0.22	−0.14	−0.03
DED798	0.49**	0.41**	0.26	−0.26	−0.38*	−0.10
Disco	0.39**	0.64**	0.32*	−0.21	−0.43**	0.10
Fidelio	0.51**	0.87**	0.24	0.29*	−0.50**	0.06
Janko	0.32*	0.75**	0.54**	−0.20	−0.39**	0.45**
Kitaro	0.65**	0.58**	0.39**	0.20	−0.27	−0.06
Lamberto	0.49**	0.75**	0.49**	−0.03	−0.35*	0.44**
Magnat	0.43**	0.60**	0.67**	−0.07	−0.01	0.05
Marko	0.64**	0.36*	0.34*	−0.24	−0.02	−0.31*
Prado	0.65**	0.46**	0.68**	−0.11	0.22	0.01
Pronto	0.70**	0.76**	0.52**	0.26	0.02	0.24
Tornado	0.37*	0.75**	0.39**	−0.19	−0.60**	0.54**
Ugo	0.14	0.75**	0.71**	−0.30*	−0.32*	0.43**
Woltario	0.60**	0.47**	0.63**	0.01	0.21	−0.22
*, ** significant at p≤0.05 and p≤0.01, respectively.

Table IV contains results of yield component analysis for 15 studied genotypes. For each cultivar, the path coefficients p _i for orthogonal components (indicating the direction of the relative influence of components on yield) and squared path coefficients p _i ² are presented. The coefficient p _i ² indicates the relative contribution of the ith component to yield determination; the sum of these coefficients for i=1, … ,3 gives determination coefficient for the linear model grain yield versus its components. Using results from clustering subject to results of yield component analysis (clustering III, Table V), a view of influencing winter triticale grain yield by its components can be provided. Three various groups (clusters) of genotypes were constructed; they differed in the pattern of determining grain yield by its components. The groups were as follows.

1. The first cluster comprised genotypes Pronto, Bogo, and Alzo. They were characterized by a small contribution of kernel weight to yield determination, and a quite large contribution of number of spikes per m² and/or number of kernels per spike.

2. The second cluster comprised genotypes Kitaro, Magnat, Woltario, DED798, Prado, and Marko. More or less similar contribution to yield determination of all components (without noticeable domination of any component) was observed for these genotypes.

3. The third cluster comprised genotypes Lamberto, Disco, Fidelio, Tornado, Ugo, and Janko. Explicit domination of number of kernels per spike in yield determination was detected for the genotypes from this cluster.

The results of clustering of the genotypes are presented in Table V; the genotypes were clustered based on mean values and coefficients of variation of studied traits (the indicator variable clustering I), on the correlation matrices for the traits (clustering II), and on the results of yield component analysis for the genotypes (clustering III). The significance level to test the relationship between the cluster belongings was 0.01. Hypothesis tests for the independence of the variables clustering I and II were not rejected, since the value of χ ² statistics was 4.8, and the significance (p-value), computed by Monte Carlo method, was 0.57. In the case of the contingency table for the variables clustering I and clustering III, the hypothesis was not rejected, either; χ ² value was 4.0 and Monte Carlo significance 0.507. The hypothesis test for the independence of the variables clustering II and clustering III was rejected, since the χ ² value was 20.0, and the computed significance was smaller than 0.001.

Table IV. Yield component analysis for 15 winter triticale genotypes.

Cultivar (R^2)	Coefficient	X 1	X 2	X 3
Alzo	p i	0.60	0.71	0.36
	p i ^2100%	36.1	50.0	12.8
(R^2=98.9%)
Bogo	pI	0.7	0.57	0.41
	p i ^2100%	49.7	32.5	16.4
(R^2=98.6%)
DED798	pi	0.49	0.56	0.63
	p i ^2100%	24.0	31.1	39.7
(R^2=94.8%)
Disco	pi	0.39	0.74	0.54
	p i ^2100%	14.9	54.3	28.7
(R^2=97.8%)
Fidelio	pi	0.51	0.75	0.4
	p i ^2100%	26.3	56.4	16.2
(R^2=98.8%)
Janko	pi	0.32	0.84	0.42
	p i ^2100%	10.5	70.1	17.4
(R^2=98.0%)
Kitaro	pi	0.65	0.46	0.59
	p i ^2100%	42.4	21.0	35.0
(R^2=98.4%)
Lamberto	pi	0.49	0.77	0.4
	p i ^2100%	23.8	59.0	16.0
(R^2=98.8%)
Magnat	pi	0.43	0.63	0.64
	p i ^2100%	18.4	39.3	41.0
(R^2=98.7%)
Marko	pi	0.64	0.53	0.55
	p i ^2100%	40.6	27.9	30.1
(R^2=98.6%)
Prado	pi	0.65	0.53	0.53
	p i ^2100%	42.4	28.6	28.4
(R^2=99.3%)
Pronto	pi	0.7	0.6	0.37
	p i ^2100%	49.3	36.0	13.5
(R^2=98.9%)
Tornado	pi	0.37	0.84	0.36
	p i ^2100%	13.5	70.7	13.1
(R^2=97.3%)
Ugo	pi	0.14	0.83	0.52
	p i ^2100%	2.1	69.5	27.4
(R^2=99.0%)
Woltario	pi	0.6	0.46	0.64
	p i ^2100%	35.5	21.5	41.5
(R^2=98.5%)

Table V. Belonging of genotypes to clusters under three clusterings: based on MV and CV (clustering I) – cf. Table II; on correlation matrices (clustering II) – cf. Table III; and on results of yield component analysis (clustering III) – cf. Table IV.

Cultivar	Clustering I	Clustering II	Clustering III
Alzo	1	1	1
Bogo	2	1	1
DED798	1	3	2
Disco	3	2	3
Fidelio	1	1	3
Janko	1	2	3
Kitaro	2	1	2
Lamberto	2	2	3
Magnat	3	3	2
Marko	2	3	2
Prado	3	3	2
Pronto	2	1	1
Tornado	1	2	3
Ugo	3	2	3
Woltario	3	3	2

Discussion

From correlation analysis applied independently for all the genotypes studied, a general view of co-relationships within winter triticale grain yield and its components can be given. Grain yield is usually positively correlated with all its components; number of spikes per m² is not correlated with number of kernels per spike; number of spikes per m² is not or is negatively correlated with kernel weight; and there is no meaningful correlation between number of kernels per spike and kernel weight, or the correlation is positive. The last observation may suggest that the total amount of available assimilates was large enough for obtaining large and numerous kernels at the same time, and that there were no physiological constraints that could be easily observed in sparse stands. This, in turn, may mean that improvements in yield are related mainly to a greater ability to fill an increasing number of kernels (Del Blanco et al., 2001; Brancourt-Hulmel et al., 2003; Shearman et al., 2005). In other studies (García del Moral et al., 2003), however, increase in kernels number was partially compensated by reductions in 1000-kernel weight (negative correlation between these components). Positive but low correlation for these traits was obtained by Kociuba (2000). Giunta et al. (1999) found a positive correlation between triticale traits (sink limitation to yield when number of grains per m² was in a range 20–25,000, and a negative correlation (source limitation to yield) when the range was exceeded. The high productivity potential of triticale is connected, in the opinion of Kozdój (2003), with the structure of a mature spike, which consists of higher number of spikelets compared to wheat and with larger kernel mass per spike compared to rye.

For three dwarf genotypes (Magnat, Woltario, and DED 798), more or less similar contribution of each component to yield determination was recorded. Only for the dwarf cultivar Fidelio grain yield was determined mainly by number of kernels per spike. Kociuba (2000) observed negative, but not high correlation between kernel weight and plant height of triticale genotypes. In our studies, the average 1000-grain weight for traditional and dwarf genotypes was, respectively, 40.8 and 43.2 g. These observations confirm Kociuba's (2000) statement. On the other hand, Gruszecka (2004) suggested that productivity improvement of dwarf genotypes should be based on a selection aimed towards increased number of grains per spike, since in her study no significant relationships between this trait and kernel weight and plants of lower height were observed. This could be obtained at the expense of improved partitioning of dry matter to growing spikes (Otegui & Slafer, 2004).

The general view of winter increase grain yield formation by its components follows from our study. Under the conditions of the experiment there were no genotypes for which number of spikes per m² would have been the dominant component in the process of grain yield formation. In the case of the genotypes Lamberto, Disco, Fidelio, Tornado, Ugo, and Janko, the contribution of this component was very small and ranged between 2.1 and 26.3%. Higher significance in yield determination had two other components, i.e., number of kernels per spike and kernel weight.

If such results are observed in studies of the multiplicity of environmental and agronomical conditions, we would be able to conclude that increasing seeding rate (which determines number of plants and spikes per area unit, and hence first yield component) for some genotypes does not necessarily increase yield. On the contrary, considerably higher significance would be given to two other components (number of kernels per spike and kernel weight). Nevertheless, the role of these components in final yield formation may depend on very unstable weather conditions (García del Moral et al., 2003).

Not rejecting the hypothesis on independence of the variables clustering I and clustering II leads to a conclusion that correlation among studied traits is not related to yield structure (given by the variability and average values of yield and its components). Hence, the genotypes with similar yield structure are characterized by different correlation patterns (matrices). For this reason, the pattern of correlation among the traits studied appears to result from genotypic factors. A similar result was obtained for the second pair of variables (clustering I versus clustering III); from this we conclude that influencing winter triticale grain yield by its components is not related to yield structure. Therefore, it has to be hypothesized that this process is particular for winter triticale genotype and does not depend on yield structure of the genotype.

The hypothesis regarding independence of the variables cultivar II and cultivar III was rejected. Hence, determining grain yield by its components is related to the correlation between the traits. However, that normal correlation analysis is the method that does not take into account a form of a causal system under consideration (in which grain yield is studied as the effect of its components that develop sequentially during ontogenesis) (Shipley (2000)). This is why determination of grain yield should not be interpreted by its components via Pearson's correlation analysis. On the other hand we have shown that a correlation matrix is related to results of sequential yield component analysis. It could have been expected, since sequential yield component analysis, as a method employing standardized regression analysis, is based on a standardized variance-covariance matrix (which is, actually, the correlation matrix) for yield and its components.

It should be noted that our approach, which is based on applying cluster analysis, makes it possible to distinguish the genotypes that are characterized by different patterns of influencing grain yield by its components. It may be possible to achieve this based on the results from Table IV, but such an approach would be very subjective. Applying the clustering approach, on the other hand, provides objective results. This approach can be applied in any study in which it is necessary to group genotypes (or any other objects) subject to a similar pattern of a process that can be quantitatively represented (in our case, the process studied is influencing grain yield by its components, and it is quantitatively represented by results of yield component analysis).

Conclusions

From our study it can be concluded that there cannot be given any general pattern of influencing winter triticale grain yield by its components. This is so because we have found that for many winter triticale genotypes, even with similar yield structure, the process of determining grain yield by its components is different. Furthermore, number of spikes per unit area, usually deemed as the most important yield component, appeared not to be the most significant in the process of formation of yield of majority of investigated genotypes. This is to some extent an unexpected result, especially in the light of the literature. We have found that the structure of contribution of yield components to grain yield determination was related to correlation pattern among yield and its components, but was not related to crop structure of a particular winter triticale cultivar. Hence, grain yield of two genotypes’ canopies with similar average values and coefficients of variation of grain yield and its components can be determined by the components in a much different way, and it might be controlled by a genotypic factor. From this it follows that we cannot give any general advice on which component of winter triticale grain yield is the most important. Undoubtedly, more effort should be put into a good stand shaped by means of agronomic treatments, particularly at tillering and stem elongation when formation of two major yield components (number of spikes×number of kernels per spike = number of grains per unit area) takes place.