Plant Breeding and Drought in C₃ Cereals: What Should We Breed For?

Abstract

Drought is the main abiotic constraint on cereal yield. Analysing physiological determinants of yield responses to water may help in breeding for higher yield and stability under drought conditions. The traits to select (either for stress escape, avoidance or tolerance) and the framework where breeding for drought stress is addressed will depend on the level and timing of stress in the targeted area. If the stress is severe, breeding under stress‐free conditions may be unsuccessful and traits that confer survival may become a priority. However, selecting for yield itself under stress‐alleviated conditions appears to produce superior cultivars, not only for optimum environments, but also for those characterized by frequent mild and moderate stress conditions. This implies that broad avoidance/tolerance to mild–moderate stresses is given by constitutive traits also expressed under stress‐free conditions. In this paper, we focus on physiological traits that contribute to improved productivity under mild–moderate drought. Increased crop performance may be achieved through improvements in water use, water‐use efficiency and harvest index. The first factor is relevant when soil water remains available at maturity or when deep‐rooted genotypes access water in the soil profile that is not normally available; the two latter conditions become more important when all available water is exhausted by the end of the crop cycle. Independent of the mechanism operating, a canopy able to use more water than another would have more open stomata and therefore higher canopy temperature depression, and ¹³C discrimination (Δ¹³C) in plant matter. The same traits would also seem to be relevant when breeding for hot, irrigated environments. Where additional water is not available to the crop, higher water‐use efficiency (WUE) appears to be an alternative strategy to improve crop performance. In this context Δ¹³C constitutes a simple but reliable measure of WUE. However, in contrast to lines performing better because of increased access to water, lines producing greater biomass due to superior WUE will have lower Δ¹³C values. WUE may be modified not only through a decrease in stomatal conductance, but also through an increase in photosynthetic capacity. Harvest index is strongly reduced by terminal drought (i.e. drought during grain filling). Thus, phenological traits increasing the relative amount of water used during grain filling, or adjusting the crop cycle to the seasonal pattern of rainfall may be useful. Augmenting the contribution of carbohydrate reserves accumulated during vegetative growth to grain filling may also be worthwhile in harsh environments. Alternatively, extending the duration of stem elongation without changing the timing of anthesis would increase the number of grains per spike and the harvest index without changing the amount of water utilized by the crop.

INTRODUCTION

Historical perspective

Agriculture in the Old World started about 10 000 years ago, coinciding with the beginning of the Holocene. From this time up to the present, C₃ cereals such as bread (Triticum aestivum L.) and durum (Triticum turgidum L. var. durum) wheat as well as barley (Hordeum vulgare L.) have remained the outstanding crops in terms of area cultivated and food source (Evans, 1998 and references therein). Early agriculture seems to have been performed under favourable moisture conditions (Araus et al., 1997c, 1999b). Nevertheless, as agriculture originated in the eastern Mediterranean region (the Fertile Crescent), cereal crops have, from the beginning, experienced drought as a main yield‐limiting factor. On the other hand, the increase in atmospheric CO₂ concentrations from approx. 270 ppm before the beginning of the industrial revolution 250 years ago to the current level of approx. 370 ppm appears to have affected long‐term net assimilation and water‐use efficiency (WUE; Araus and Buxó, 1993). This would explain increases in yield of nearly 50 % under present‐day conditions (Manderscheid and Weigel, 1995; Mayeux et al., 1997), independent of the yield increase due to the doubling in harvest index (HI) in the current cultivars (Loss and Siddique, 1994; Araus et al., 1999a; Calderini et al., 1999). Future scenarios are expected to involve a further steady increase in atmospheric CO₂ concentrations, which is likely to have a positive effect on yields (counterbalanced somewhat by increases in temperature). In a study of well‐managed crops in Germany, it was concluded that barley yield would increase by 0·35 % per ppm increase in CO₂, whereas increases in wheat would be about 25 % lower (Manderscheid and Weigel, 1995). Of course, in harsh environments or where management practices (particularly nitrogen fertilization) are less than optimal, the effect of increasing CO₂ on yield can be much lower or nil.

To date, increases in cereal productivity have kept pace with population growth (Evans, 1998), even in the 20th century when the population rose more rapidly than ever before. During the first half of the 20th century, demands for higher wheat and barley production were satisfied mostly by increases in harvested area (Slafer et al., 1994, 1996), while yields per unit area rose only slightly (Calderini and Slafer, 1998). This pattern changed during the second half of the century (Fig. 1), coinciding with the ‘Green Revolution’, when yields per unit area climbed steadily thanks to the enormous success of crop breeding (primarily the introduction of dwarfing genes), combined with intensive agronomic practices (mainly the use of increased nitrogen fertilizer). Yield stability also increased during this period (Smale and McBride, 1996).

Fig. 1. Pattern of increases in mean wheat and barley (inset) yields during the 20th century in countries characterized by good water conditions or prone to water deficits. Data derived from FAO’s web‐site (www.fao.org).

Current scenario and future prospects

Breeding success during the second half of the last century (particularly since the 1970s) has been largely due to the introduction of dwarfing genes and the consequent increase in HI (Richards, 1996a, 2000), breeding gains being largest under the most favourable growing conditions. Reciprocally, the progress in yield attributed to breeding is inversely proportional to the stress in the growing environment (Richards, 1996a). It has, for example, been ten times smaller (0·4 Mg ha^–1 vs. 4 Mg ha^–1) in Australia compared with the UK. Nevertheless, when yield increases are expressed as a percentage no differences between extreme environments emerge (Fig. 2; Calderini and Slafer, 1998).

Fig. 2. Relative increases in wheat yield due to breeding under favourable (UK) and drought (Australia) conditions.

To reach the forecasted global demand for wheat in the year 2020 global production will need to increase by 1·6 to 2·6 % annually; that is, to increase from the current average grain yield of 2·5 Mg ha^–1 to 3·8 Mg ha^–1 (Braun et al., 1998). Future production increases will continue to be strongly dependent on our ability to achieve higher yields rather than to augment the harvested area (Slafer et al., 1996; Slafer and Araus, 1998; Blum, 2000). However, genetic increases in wheat yields since 1960, based for example on harvest index, are becoming harder to achieve (Reynolds et al., 1996; Sayre, 1996; Mann, 1999). This scenario is further complicated when we consider that in many regions abiotic (mostly water and temperature) stresses strongly limit current yields (and yield stability) as well as the genetic gains associated with crop breeding (Richards, 1996a). Indeed, the spread of modern cultivars into the drier areas has been much slower and their impact on yields far weaker than for favourable areas (Evans, 1998): wheat yield gains over traditional cultivars have usually been below 20 %, and often less than 10 %, and have even been negligible in extremely harsh environments (Byerlee and Morris, 1993). Broadly the same scenario may be presented for barley. This is the case in some Mediterranean regions with very severe droughts, such as the Near East and north Africa, where wheat (mostly durum wheat, followed by bread wheat) and barley are by far the dominant staples.

For economic and environmental reasons, we cannot expect a contribution from improved management as significant as that corresponding to the last half century. This statement does not imply that improved management can not still make a contribution (for instance by improving the efficiency of the irrigation process, or by designing better rotation systems thereby not only improving sustainability but also reducing crop losses from pests that may be controlled with rotations). Although these management improvements may have a large impact locally, the dominant factor in the large effect that improvement in management produced in the last half‐century was the increase in the use of inputs, particularly agrochemicals (fertilizers, insecticides, herbicides and fungicides). Globally, the level of usage of these inputs will hardly increase significantly in the future. Thus, strong reliance on future genetic gains in yield of wheat and other cereals will gain importance for the required production increases. However, this task is not simple: future genetic improvement will have to be even more efficient than in the past, but this time with a crop that has already received an intense breeding effort to produce large increases in yield (Slafer et al., 1999). In fact, this relatively high yield level of current cultivars is likely to be responsible, at least in part, for genetic increases in wheat yields becoming harder to achieve (Reynolds et al., 1996; Sayre, 1996).

Empirical vs. analytical breeding

To date, cereal breeding in Mediterranean environments has been based principally on empirical selection for yield per se (Evans, 1993a; Loss and Siddique, 1994). However, this approach is far from being optimal, since yield is characterized by a low heritability and a high genotype × environment (G × E) interaction (Jackson et al., 1996).

In the above context, there is now a strong argument that an indirect (or analytical) approach, based on an understanding of the crop at the physiological and molecular biology levels, may help target the key traits that are currently limiting yield. Such an approach may therefore complement conventional (i.e. empirical) breeding programmes and hasten yield improvement (Araus, 1996; Richards, 1996a; Slafer and Araus, 1998). Moreover, even the potential power of molecular biology techniques for locating gene sequences and introgressing quantitative trait loci (QTL) or even for selecting or genetically transforming economically important QTL (such as those directly determining potential yield or the response to abiotic stresses) will depend strongly upon our understanding of yield‐determining physiological processes.

An integrated approach to plant breeding

By combining information on the physiological basis of yield limitation with new physiological selection tools, the probability of accelerating the rate of genetic progress through plant breeding should be significantly increased. Parents can be selected for improved physiological traits and crossed to high‐yielding agronomically elite materials. Furthermore, identification of progeny phenotypes with favourable interactions among genes permitting the expression of higher yield can be enhanced by: (1) eliminating inferior agronomic phenotypes visually in early generations; (2) selecting superior physiological phenotypes using rapid detection techniques in intermediate generations; and (3) selecting for higher performance in yield trials in advanced generations (Reynolds et al., 2000).

From a breeding perspective there are, of course, specific requirements that any secondary or physiological trait should fulfil before it is included in a breeding programme. Namely, the selected trait must exhibit enough genetic variability, a high genetic correlation with yield and a higher heritability than yield itself in genetic populations representative of those being evaluated (Jackson et al., 1996). Moreover, evaluation of these traits must be fast, easy and cheap (Araus, 1996; Slafer and Araus, 1998; Araus et al., 2001). In relation to the physiological scope of this work, these requirements may be summarized in question form: how do we choose and evaluate a secondary trait? Answering this question will be the aim of our paper.

HOW DO WE CHOOSE A SECONDARY TRAIT?

Any trait to be considered must be directly related to yield. The literature is filled with proposed traits dealing with the lower levels of organization (i.e. molecular, biochemical), which frequently show only poor and inconsistent relationships with crop yield in the field (Richards, 1996a; Araus et al., 2001). In this category we can include metabolic traits such as enzyme activities (e.g. Rubisco or nitrate reductase), levels of substrates (e.g. proline or sugars) and growth regulators (e.g. abscisic acid). Yield is by nature a very integrative trait: through the plant cycle it takes in many levels of plant organization, from the molecular level to the canopy. Therefore, any trait consistently related to yield should also be integrative, either in time (by being determined through part, if not all, of the crop cycle), or in level of organization (by representing a level close, if not identical, to that of yield); or both (Araus, 1996; Slafer and Araus, 1998; Araus et al., 2001). Moreover, any attempt to improve the breeding gain through an analytical approach depends upon our ecophysiological understanding of the crop in question. There are different approaches to identify potential secondary traits. The soundest is probably that relying on the identification of the main physiological processes determining yield. Retrospective studies, comparing physiological bases of the differences in yielding capacity among genotypes released at different periods, may also complement this approach. Finally, the kind of trait selected will depend strongly on the target environment where the selection is addressed. Nevertheless, since there are common elements of physiological response to a range of environmental stresses (see Araus, 2002 and references herein), it may be worth selecting those traits that confer a wider adaptation.

One approach: identifying the main physiological processes determining yield

It may be rewarding to analyse the physiological processes behind the determination of yield to help identify opportunities for future breeding programmes. Since, in many circumstances, cereal crops experience the effect of abiotic stresses during at least part of the growing season, these key breeding traits must somehow help to improve the crop response.

Yield as a function of incoming radiation.

The yield of the harvestable part (grain yield for cereals, GY) of a crop throughout a given period of time can be expressed as follows (see also Hay, 1999):

GY = RAD × %RI × RUE × HI(1)

where RAD is the total quantity of incident solar radiation received by the crop throughout the growing period; %RI is the fraction of RAD intercepted by the canopy averaged across the crop cycle; RUE is the radiation‐use efficiency, i.e. the overall photosynthetic efficiency of the crop: the total dry matter produced per unit of intercepted photosynthetically active radiation (PAR); and HI is the harvest index or fraction of the total dry matter harvested as yield. This is normally expressed in terms of above‐ground production, excluding the root system.

Total biomass, the result of RAD × %RI × RUE, can be understood in a physiological sense as the result of accumulated net photosynthesis of the crop. Thus, it is well accepted that total canopy photosynthesis during growth is closely related to yield, as reported in several species (Zelitch, 1982; Ashley and Boerma, 1989). Furthermore, a strong relationship between biomass at anthesis and yield has been demonstrated in bread (van den Boogaard et al., 1996) and durum (Villegas et al., 2001) wheat.

Crop management practices, such as changing planting date, or crop breeding strategies modifying plant duration (by changing crop phenology or by improving stress resistance and then leaf area duration) will affect RAD. %RI by the canopy may be increased by a faster approach to full cover (i.e. early vigour) and a higher leaf extension (e.g. by improving its stress resistance). RUE may be increased by improving the distribution of PAR among the various leaves (i.e. identifying the basis of canopy development) as well as enhancing the photosynthetic performance under stress conditions. When considering the effect of drought on yield, it is worthwhile determining the relative sensitivity of each of the components of the equation set out above in relation to this stress. Thus, whereas much work has been devoted to studying how drought stress affects photosynthetic performance, it is widely accepted that drought affects leaf expansion earlier than photosynthesis (Hsiao, 1973). Therefore, the effect of drought limiting yield through lower leaf growth (and even through earlier leaf senescence) may actually prove more important than the effect on photosynthetic rates in determining yield in many crops. Of course the harvestable component (i.e. grains) and the monocarpic nature of cereals (with no vegetative development during grain filling) balance the above picture somewhat in favour of photosynthetic performance. Therefore, genetic variation in leaf area growth, leaf area duration or leaf photosynthesis may become very significant factors under stress conditions (Richards, 2000).

Potential avenues for increasing RUE under abiotic stress.

Under stress, the assimilation rate is generally more limiting to yield than it is under optimal conditions, as indicated by the higher association commonly observed between yield and above‐ground biomass at maturity. While selection for higher rates of leaf photosynthesis has not generally resulted in improved yield under temperate conditions (Evans, 1993b) (probably because the source is less limiting than the sinks), greater success might be expected under abiotic stress. At high temperatures, genetic gains for yield have been reported in response to selection for leaf photosynthetic rate in early generations (Gutiérrez‐Rodríguez et al., 2000). Another way to increase RUE might be to reduce photorespiration, perhaps by increasing the affinity of Rubisco for CO₂, thereby decreasing its oxygenase activity. Modest variation for CO₂ specificity has been found in land plants (Parry et al., 1989; Delgado et al., 1995), with wheat having one of the highest values for a crop species. However, much higher values are reported in marine algae (Read and Tabita, 1994). Molecular biology techniques may offer the possibility of genetically transforming wheat Rubisco from its current specificity of 95 to values of 195 corresponding to that of the thermophilic alga Galderia partita. However, Rubisco also plays a protective role in dissipating excess energy, with O₂ uptake in the light playing a significant role in preventing chronic photoinhibition under field conditions (Osmond and Grace, 1995). Thus, it may not make sense to alter Rubisco’s oxygenase specificity especially in crops subjected to a high degree of abiotic stress. It may be more productive to look to genetic manipulation of other mechanisms involved in dissipating excess reducing power, for example, through the Mehler–ascorbate peroxidase reaction (Osmond and Grace, 1995) and by cycling carotenoid pigments (Gilmore, 1997). Under well‐watered conditions, the photosynthetic rate of the whole canopy can be enhanced by manipulation of leaf angle, which is under relatively simple genetic control (Carvalho and Qualset, 1978), and possibly by manipulating leaf‐N distribution throughout the canopy. However, abiotic stress occurring during canopy development will modify many of the canopy characteristics in comparison with a well‐watered crop. Leaves are often smaller, creating a more erectophile canopy than when unstressed (Araus et al., 1986). In addition there may be fewer tillers, and this, together with the lower leaf area, reduces the leaf area index (LAI). Finally, due to accelerated senescence, green area duration may reduce the potential for assimilation. ‘Ideal’ canopy structures under stress have yet to be defined and will vary depending on the stress profile of the target environment. The main consideration should be to maximize productive interception of light while avoiding excess radiation which may cause photoinhibition and wilting. Rapid canopy establishment may be beneficial for increasing radiation interception and reducing evaporation of soil moisture if conditions are favourable in the early part of the cycle (Richards, 1996a). Under severe drought the opposite may be true if excess leaf area leads to a premature depletion of soil water before the life cycle is complete. However, under hot, irrigated conditions, early vigour of genotypes was neither positively nor negatively associated with yield and seemed to be mainly a function of precociousness (Reynolds et al., 1994). Low LAI and traits such as leaf rolling may be beneficial under stress to reduce light interception and hence avoid damage due to excess radiation. The same may be said for a low leaf chlorophyll content in the upper leaves (Tardy et al., 1998). Loss of chlorophyll may even be beneficial in the later stages of grain filling if indicative of remobilization of stored assimilates to the grain (Blum, 1998) and if no more soil water is available. Traits like leaf waxiness and pubescence could also be potentially useful in reducing radiation load (Richards, 1996a). In summary, optimization of crop photosynthesis under stress requires a balance between maximizing assimilation at critical growth stages and when conditions are favourable on the one hand, and avoiding the destructive effects of excess radiation when stress is most intense on the other. The traits to select for will depend heavily on the specific stress profile likely to be encountered.

Yield as a function of available water.

As water stress is the main environmental factor limiting yield, Passioura (1977, 1996) proposed a parallel way of considering grain yield in a water‐limited situation:

GY = W × WUE × HI(2)

where W is the water absorbed and further transpired by the crop plus direct evaporation from the soil; and WUE is the ability of the crop to produce biomass per unit of water evapotranspired.

Even when the terms of the above equation are not purely independent—their sign may change depending on the target environment—it remains as a pedagogical model. Accordingly, with the component analysis proposed by Passioura (1977, 1996) the traits to select for would be those increasing: (1) the capacity to capture more water; (2) the efficiency for producing dry matter per unit of absorbed water; and (3) the ability to allocate an increased proportion of the biomass into grains (Fig. 3A).

Fig. 3. A, Several scenarios derived from Passioura’s equation. B, Schematic diagram showing a common negative G × E interaction, implying a crossover at relatively low levels of environmental index (arrow), when cultivars were selected for yield per se in near optimum conditions (1) or under severe stress (2). C and D, Comparative performance of old (closed circles) and new (open circles) wheat cultivars in favourable (C) and drought (D) environments.

Higher crop performance may be achieved through improvements in water use, water‐use efficiency and harvest index. Water use is relevant when there is still soil water available at maturity or when deep‐rooted genotypes access water deep in the soil profile that is not normally available. The other two factors become more significant when all available water is normally used up by the end of the crop cycle. The single most important attribute tackled to date by breeding programmes, determining performance under water stress, has been phenology: matching crop development and seasonal rainfall patterns (see references in Richards, 1996a; Slafer and Araus, 1998; Villegas et al., 2000). In fact, this attribute may affect either water use, or the efficiency of its utilization. Other developmental traits, such as deeper root systems and early vigour, as well as osmoregulation, may also help the crop to use more water. Thus, traits such as early vigour or the pattern of phenological development, already important when considered in light of eqn (1), are again significant here as they increase both W and WUE.

A complementary approach: some clues through retrospective studies

The ways in which breeding has improved yield during past decades can be properly assessed through retrospective studies. Cultivars released at different times in the past were cultivated simultaneously under certain conditions, thereby eliminating the effects on yield of improved management practices (Slafer et al., 1994). These studies may afford some clues to the physiological changes underlying the genetic gains in yield achieved in the past.

Most of the traits identified in retrospective analyses have been shown to be constitutive in nature; that is to be expressed in the absence of stress. Retrospective analyses show that HI has risen in the more recent genotypes, not only under stress‐free conditions but also under drought (Calderini et al., 1999). The increase in HI of modern cultivars has principally been attained by diverting assimilates previously used for long stem growth to increasing the size of the grain‐bearing ears. Associated therefore with higher HI is a shorter stature (Slafer et al., 1999). However, this trait may have negative consequences under drought‐prone environments where growth of kernels is sustained by assimilates accumulated in the internodes before anthesis (Loss and Siddique, 1994). This may explain the lack of a negative relationship in durum wheat between stem length and yield under moderate drought (e.g. Villegas et al., 2000).

An alternative to raising HI without penalizing response to drought would rely on the selection of cultivars with less wasteful tillers, something which seems to result in substantially large stems and ears in wheat (Atsmon and Jacobs, 1977; Richards, 1988). A second way to increase assimilate supply to selected organs is to increase their growth duration. This can be acheived without changing the duration of total crop growth (Richards, 1996b, 2000) and thus water use (Slafer and Araus, 1998). An example in both wheat and barley is to increase the duration of the spike growth period to allow more assimilate supply to the growing florets, thus reducing the very high rate of floret abortion just before anthesis (Miralles and Richards, 2000; Miralles et al., 2000; Slafer et al., 2001). This longer stem‐elongation period would be compensated by a reduced earlier period so that there would be no differences in the amount of water extracted from the soil (Slafer and Araus, 1998). Additionally, there is evidence that the dry matter (see references in Slafer and Araus, 1998), and, more importantly, the nitrogen content of the spike at anthesis is an important determinant of grain number (Abbate et al., 1995; van Herwaarden, 1995; Richards, 1996b; Slafer et al., 1999).

As noted previously, developmental traits such as early vigour or phenology may be particularly significant in the water‐limiting conditions of the Mediterranean environment (Slafer et al., 1994; Royo et al., 1995, 1999; Bort et al., 1998; González et al., 1999). Thus, greater early vigour and/or faster phenological development (e.g. earlier date of anthesis or maturity) reduces the evaporative loss of water from the soil surface on the one hand, while also ensuring that more growth and transpiration occur when the vapour pressure deficit (VPD) is small. However, the suitability of adopting early vigour and phenological adjustment (earliness) as breeding traits will depend strongly on the nature of the water stress. This is illustrated for earliness in Fig. 4, where retrospective studies for western Australia and the Argentine Rolling Pampas are compared, both regions normally experiencing water deficits but with quite a different pattern of rainfall distribution. In fact, empirical selection (i.e. for yield itself) has led to a systematic shortening of the duration from sowing to anthesis in Australia, whereas no consistent trends were found in Argentina. Western Australia is characterized by a Mediterranean climate, where terminal stress (i.e. drought during grain filling) is a common event during cereal cultivation. A more isohygrous rainfall distribution is common in most of the Rolling Pampas. Lack of changes in earliness of cultivars progressively released has also been reported in cereal regions of Canada and the USA where drought occurs but not with a Mediterranean pattern (Slafer et al., 1994).

Fig. 4. Retrospective studies comparing the pattern of seasonal rainfall and the changes in earliness of wheat cultivars released in western Australia and Rolling Pampas, Argentina.

Shortening of the crop duration (i.e. escape from the stress) has been a very successful strategy when breeding for Mediterranean conditions, particularly in very poor environments. As a result of a shorter cycle the relative amount of water used during grain filling has increased in the more recent cultivars (Fig. 5). Such increase carries great weight if we consider that HI shows a hyperbolic relationship with the ratio of post‐ to pre‐anthesis water use (Sadras and Connor, 1991). This is again illustrated by data from Western Australia (Fig. 6). In fact timing is a basic strategy to raise yield and biomass in relation to water supply (Slafer et al., 1994; Bort et al., 1998; Richards, 2000). If water is a major limitation, then maximized growth coinciding with periods of cool weather and relatively low vapour pressure deficits will raise water use efficiency and biomass production (Richards, 1991; Gomez‐Macpherson and Richards, 1995). Manipulation of major genes responsible for sensitivity/insensitivity to photoperiod and vernalization enables temperate cereals to be grown in a very wide range of durations (Richards, 2000).

Fig. 5. Changes in pre‐anthesis and post‐anthesis water use as well as in the post/pre‐anthesis ratio due to the earlier anthesis time in the most recent cultivars of wheat. All lines were fitted by regression (r = 0·65–0·85; P < 0·05). Data from Siddique et al. (1990).

Fig. 6. Relationship between the ratio of post‐ to pre‐anthesis water use and harvest index. The year of release of each wheat cultivar is shown in parentheses. Data from Siddique et al. (1990).

No increases in the rate of photosynthesis per unit leaf area were reported in association with breeding until the beginning of the last decade (Austin, 1989). There was even a reduction due to increase in ploidy (Austin et al., 1982; but see Araus et al., 1989). However, recent studies have investigated traits associated with the yield increases achieved through breeding, and some photosynthetic processes appear to be important. In wheats bred at CIMMYT (Mexico), it was found that under well‐irrigated conditions stomatal conductance g_s, maximum photosynthesis rate and canopy temperature depression (CTD) were all associated with yield progress (Fischer et al., 1998). Among the photosynthetic characteristics, the correlation with leaf conductance was strongest (Fig. 7; see also Richards, 2000).

Fig. 7. Relationship between yield of eight CIMMYT wheat cultivars differing in yield potential and average flag leaf photosynthetic rate at light saturation (A_max, A and B), stomatal conductance (g_s, A) and canopy temperature depression (CTD, B). Adapted from Fischer et al. (1998).

Stable carbon isotope discrimination (Δ¹³C) has also been positively associated with yield progress in the same set of eight genotypes from CIMMYT (Fischer et al., 1998). Recently, the stable oxygen isotope composition (δ¹⁸O) of flag leaves has been found to be significantly negatively correlated with g_s, CTD and grain yield in this historic set of CIMMYT wheats for two of the three seasons studied (Barbour et al., 2000). However, correlations between grain δ¹⁸O and these traits were less clear (Barbour et al., 2000), especially under rainfed conditions (Mateo et al., submitted). Further research is necessary to assess the potential usefulness of δ¹⁸O in breeding.

Several factors have been proposed to contribute to the positive relationship between g_s and yield: under irrigation decreased stomatal sensitivity to VPD or to subtle water stress; extra cooling particularly at warmer temperatures; or increased sink strength (Reynolds et al., 1994; Fischer et al., 1998; Richards, 2000). Under moderate to mild stress, genotypes able to extract more water and show better water status during their growth, whatever the physiological adaptation involved (e.g. osmoregulation, deeper roots), would also be those showing higher g_s. Of course, other factors such as crop earliness or a lower LAI may also lead to a higher stomatal opening (Richards, 1996a; Araus et al., 1998).

Defining the target environment for selection: productivity vs. survival under drought

The traits to select for, the environmental conditions most convenient for selection and, in fact, the entire philosophy that underpins breeding strategies rely strongly on the kind of target environment(s) to which breeding effort is addressed. This is clearly illustrated by the controversy between breeding strategies adopted by CIMMYT and the International Rice Research Institute (IRRI) on the one hand, and the International Center for Agricultural Research in the Dry Areas (ICARDA) on the other. CIMMYT coined the term ‘megaenvironment’, selecting for wheat under stress‐free conditions (Reynolds et al., 1996) with the idea of further increasing yield in a wide range of conditions, ranging from stress‐free to moderate stress environments. The same philosophy holds for IRRI’s rice‐breeding programme. The breeding efforts of ICARDA on barley are more devoted to improving survival (i.e. tolerance to severe stress) and thus yield stability by selecting in drought‐prone, poor environments characterized by yields frequently below 1·0 Mg ha^–1 of small cereals, mostly barley and durum wheat (Ceccarelli and Grando, 1996). However, the above controversy appears fictitious in the sense that it is based on the premise that the goals to achieve both strategies for selection are correct. These strategies fit well with of Finley and Wilkinson’s (1963) model of the response of yield of particular genotypes to the environmental index, understood as the average yield of all participating genotypes. If the range of environmental (E) index and genotype (G) variation is large enough, there would virtually always be a kind of negative G × E interaction, represented by a crossover in the yield of the genotypes of high and low yield potential (Fig. 3B). In other words, the genotypes selected for low yield conditions will probably perform better than those released for high‐yielding environments when grown under very poor environments and vice versa.

Thus, crop selection performed in nurseries with good growing conditions is frequently translated to cultivars with higher water use (and even higher water use efficiency) and increased productivity in a wide range of growing conditions, from non‐limiting (e.g. with yields over 7·0 Mg ha^–1), to mild (approx. 4·5–7·0 Mg ha^–1) and moderate stress (approx. 2·0–4·5 Mg ha^–1) environments. However, in more stressed environments the situation may reverse, with genotypes selected in good environments performing less well than those already selected under the poor conditions of the target environment. Normally, such a crossover occurs at low yield levels (see arrow in the abscissa axis of Fig. 3B) and appears to be slightly higher for barley (2·0–2·5 Mg ha^–1; Ceccarelli and Grando, 1991) than for wheat (1·0–1·5 Mg ha^–1; Laing and Fischer, 1977; Fischer, 1993).

From an ecophysiological perspective, changes occurring as a response to abiotic stresses may be initially divided into two distinct categories: those conferring on the plants the ability to tolerate (i.e. survive) extreme stress events, and those maximizing productivity under a relatively wide range of less stressful conditions (by avoiding stress). The expression of the former has penalties in terms of yield under less‐severe conditions, while the latter traits are constitutive and are expressed independently of the degree of stress, thus representing virtually no penalties for yield potential because they are still essential in sustaining relatively high yields under mild to moderate stress (Blum, 1996). Possibly as a consequence of this, selection for higher yielding performance has frequently resulted in higher yields in a wide range of environments (see examples in Fig. 3C and D). In these examples, modern cultivars have consistently outyielded older cultivars, even in the lowest‐yielding conditions of each particular study (Perry and D’Antuono, 1989; Slafer and Andrade, 1989, 1993; Calderini et al., 1995).

A shorter crop duration would correspond to a different ecophysiological strategy: drought escape. It is a constitutive trait that confers on genotypes better performance (in terms of yield and stability) in severe to moderate drought environments (Table 1). However this trait also has negative implications in terms of yield potential [see eqn (1)].

Table 1.

Grain yield (kg ha^–1) of a set of 140 genotypes of the Durum Core Collection grown during the 1996–1997 season at Tel‐Hadya (headquarters of ICARDA), NW Syria

	Irrigated	Rainfed	Late planting
Mean yield	4096	2410	1579
Standard deviation	773	468	564
Correlation coefficient	–0·15	–0·43	–0·77
Significance	N.S.	***	***

Three different growing conditions were assayed: the first two were winter plantings under irrigated and rainfed conditions and the third a late planting trial. The correlation coefficients of the relationship between grain yield and number of days from planting to heading are shown for each growing condition (Bort, Araus, Asbati and Nachit, unpubl. res.).

N.S., not significant; *** significant at α = 0·001.

In what follows we will focus on traits to improve yield (i.e. by avoiding stress) on the right side of the G × E interaction, that is, under moderate and mild drought stress situations. Unlike traits concerned with survival (i.e. by tolerating cellular dehydration), those conferring the ability to produce relatively high yields under water stress are far from clear (Passioura, 1996). Nevertheless, as pointed out by Richards (2000, and references herein), selection under favourable conditions usually translates to higher yields in less favourable environments.

HOW TO MEASURE A SECONDARY TRAIT

Physiological traits, in general, have seldom been used successfully as selection criteria in breeding programmes (Slafer et al., 1994). There are many reasons for this (Slafer et al., 1999). One reason was that there was no practical way to measure them in realistic breeding programmes. With modern apparatus and new analytical tools, it is now relatively simple to measure physiological parameters or their surrogates in the field (Araus, 1996). Below, we discuss some of the most promising techniques for evaluating secondary traits. This is not intended to be an exhaustive list and new, sounder technical approaches may well emerge in the near future.

Using traits related to stomatal aperture to select for yield

Stomatal conductance and canopy temperature depression, which are easier to measure than photosynthetic rates (A_n), have been shown to be associated with performance of irrigated wheat under high radiation levels (Araus et al., 1993; Reynolds et al., 1994, 1999; Fischer et al., 1998) (Fig. 7). In the same way, stable carbon isotope discrimination has also proved to be a potentially powerful approach (Farquhar and Richards, 1984; Farquhar et al., 1989; Acevedo, 1993; Araus et al., 1998) with the advantage of integrating not only the functioning of the crop at the canopy level (as does CTD) but also through at least part of the plant cycle. The interest in evaluating g_s, either directly at the leaf level or by means of more integrative approaches (CTD, Δ¹³C), is highlighted by studies suggesting that g_s is a better indicator of the plant water status than, for example, the water potential or the relative water content, with changes in photosynthetic metabolism during progressive drought being tracked well by g_s (Sharkey, 1990; Flexas et al., 2000).

Canopy temperature depression.

When water evaporates from the surface of a leaf it becomes cooler, and the rate of evaporative cooling is directly affected by stomatal conductance (besides VPD), which is itself affected by feedback mechanisms of other processes such as photosynthetic metabolism and vascular transport. Canopy temperature depression is therefore a good indicator of a genotype’s physiological fitness, since a high value is indicative of good expression of all of those traits under a given set of environmental conditions. Moreover, leaf cooling contributes to improvement of the photosynthetic activity of leaves and prevents premature ageing. The surface temperature of a canopy in a field plot can be measured easily, cheaply and quickly (within a few seconds), with a simple infrared thermometer. Since the reading integrates the temperatures of plant organs over a small area of the canopy, error associated with plant‐to‐plant variability is reduced (Blum, 1988; Blum et al., 1989).

CTD measured on irrigated yield trials showed a good association with plot performance, but in addition to being a good predictor of yield in situ, CTD showed a significant association with performance of the same lines grown at a number of target breeding locations (Reynolds et al., 1994). Further work confirmed the potential for making genetic gains in response to selection for CTD in recombinant inbred lines. Recently, CIMMYT breeders successfully used CTD measured on small plots in their heat tolerance nurseries to identify the highest yielding entries. Genetic correlation coefficients of 0·6–0·8 were observed between final yield and CTD measured during grain filling, indicating the potential of this technique to pre‐screen for physiological potential prior to the execution of expensive yield trials (van Ginkel et al., pers. comm.).

However, other authors have found no consistent relationships between CTD and yield under Mediterranean conditions (Royo et al., 2001). Regardless of methodological considerations, such as the requirement for no wind and clouds to measure the plots accurately, CTD has proved useful for predicting differences in yield only when crops had recently been irrigated and when the canopy covered all the soil, especially in hot, dry environments (i.e. with a high VPD). In addition, phenological differences among lines being evaluated in a programme may also play a disturbing role due, for example, to different rates of transpiration between leaves and reproductive structures, the former being naturally cooler than the latter, or to the lower transpiration rate of senescing organs. Thus, when measured on a particular date, the stage of development of the plots must be taken into account to reduce the likelihood of under‐ or over‐estimating the differential potential of lines varying in instantaneous CTD.

Aerial infrared imagery.

Recent results from NW Mexico showed that aerial infrared (IR) images collected at a height of 800 m had sufficient resolution to detect CTD differences on relatively small plots (1·6 m wide). The data, collected from an IR radiation sensor mounted on a light aircraft, were positively correlated with final grain yield for a set of random derived recombinant inbred lines as well as a set of advanced breeding lines (Table 2). The results indicate the potential of aerial IR imagery as a means of screening thousands of breeding lines in a few hours for CTD (Reynolds et al., 1999), and are in agreement with previous studies (Blum, 1988).

Table 2.

Comparison of canopy temperature depression (CTD) data from aerial infrared (IR) imagery with hand‐held IR thermometers, Obregon 1996–1997, NW Mexico (adapted from Reynolds et al., 1999)

		Correlation of CTD with yield
		Aerial		Hand‐held
Trial	n	Phenotypic	Genetic	Phenotypic	Genetic
RILs (Seri82*7C66) random derived sisters	81	0·40**	0·63**	0·50**	0·78**
Advanced lines bread wheat	58	0·34**	–	0·44**	–

–, Genetic correlations not calculated due to design restrictions.

** Statistical significance at 0·01 level of probability.

Stable carbon isotope discrimination.

A positive association between yield and Δ¹³C of the upper organs of the plant (mostly grains) has been reported previously for different cereal species growing under favourable conditions (Condon et al., 1987; Romagosa and Araus, 1991; Richards, 1996a; Araus et al., 1998; Voltas et al., 1999). Δ¹³C is also widely accepted as an indicator of crop WUE because both traits are negatively correlated for given VPD conditions (Farquhar and Richards, 1984; Farquhar et al., 1989; Hubick and Farquhar, 1989; see also Nilsen and Orcutt, 1996) provided there is constancy in the evaporative component of the evapotranspiration. Because of the positive relationship between yield and Δ¹³C, those genotypes showing the highest yields would be those actually having the lowest WUEs, which seems, at first sight, to contradict the Passioura model. Under well‐watered conditions this can arise if the higher yielding genotypes have a lower assimilation capacity per unit area than the lowest yielding lines with low Δ¹³C (Araus et al., 1997a, b), or if lines with a higher g_s have the highest yields, which appears to be the most common situation (Richards, 2000). Alternatively, it has been reported that increased water availability during growth under Mediterranean conditions frequently leads to an increase in WUE for wheat, mainly because the evaporative loss is reduced due to increased soil cover by the canopy (Zhang et al., 1998; Oweis et al., 2000). In these circumstances a higher kernel Δ¹³C might be associated with higher crop WUE.

Nevertheless, the positive relationship between Δ¹³C and yield has also been reported for environments exhibiting moderate to even severe drought (Richards, 1996a; Araus et al., 1998; Royo et al., 2001). The best correlations are usually attained when the latest developed parts of the plant are analysed (Table 3). In this case the genotypes exhibiting higher Δ¹³C (and therefore higher g_s) are those having better water status, either as a result of constitutive (e.g. phenology, plant size) or stress adaptive (e.g. osmotic adjustment) traits that allow the crop to escape or resist drought stress.

Table 3.

Relationship between grain yield and carbon isotope discrimination measured in dry matter of seedlings, in the penultimate leaf and in mature kernels for a set of 144 durum wheat genotypes of the Durum Core Collection grown under rainfed conditions at Tel‐Hadya (headquarters of ICARDA), NW Syria (Araus et al., 1998, 2001)

	Seedlings	Penultimate leaf	Mature kernel
Correlation coefficient	–0·14	+0·29	+0·50
Significance	N.S.	***	***

N.S., not significant; *** significant at α = 0·001.

Regardless of the physiological mechanism operating (e.g. root traits, osmotic adjustment, etc.), a cultivar able to use more water than another during its growth would have more open stomata and therefore a higher CTD and Δ¹³C in plant matter. The same traits would also appear relevant when breeding for hot, irrigated environments (Araus et al., 1993; Reynolds et al., 1994). Those crops having a larger yield at the expense of a higher water use will frequently have a lower WUE. In other words, for moderately to slightly stressful or stress‐free environments, selecting for a higher WUE may actually limit the potential yield of the cultivars if the condition for having high WUE is related to stomatal sensitivity to (temporarily) restricted water availability. Under these circumstances it is clear that the terms of Passioura’s equation are not completely independent (Araus, 1996; Richards, 1996a; Slafer and Araus, 1998).

Where additional water is not available to the crop (i.e. all the water available is exhausted during the crop cycle), increased WUE appears to be an alternative strategy for improving performance. Therefore, and in contrast to situations where lines perform better because of increased access to water, lines producing greater biomass and yield due to superior WUE will have lower Δ¹³C. This may be the situation in very poor environments (Slafer and Araus, 1998) where barley is frequently the only cereal cultivated (Voltas et al., 1998; but see also Acevedo, 1993).

Given the relatively high cost associated with isotopic analysis, several surrogate approaches which are much cheaper, faster and easier to handle have been proposed. The option most studied has been to use the mineral or just the total ash content of leaves (Masle et al., 1992; Mayland et al., 1993; Araus et al., 1998) or grains (Febrero et al., 1994; Araus et al., 1998; Voltas et al., 1998). Another promising alternative relies on estimation of Δ¹³C through the Near Infrared Spectroscopy (NIRS) technique (Clark et al., 1995; Ferrio et al., 2001), which has the further advantage of being non‐destructive.

Spectral reflectance indices

A technique that may have application in screening for physiologically superior progeny is spectral reflectance in the visible and near infrared regions. This can be used to estimate a range of physiological characteristics (Araus, 1996), including canopy chlorophyll content, absorbed PAR, leaf area index and plant water status (Fig. 8). Spectral reflectance indices are formulations based on simple operations between the reflectances at some given wavelengths, such as ratios, differences, etc. Originally used in remote sensing by aircraft and satellites, spectroradiometric indices (SRI) measured at ground level are becoming a very useful tool for the assessment of many agrophysiological traits. Using SRI, all these traits can be estimated simultaneously in each sample, at a rate of up to 1000 samples per day; other methods are much more tedious and time consuming. This makes SRI a useful tool in breeding programmes when screening either for potential yield or for resistance to different stresses (Elliott and Regan, 1993; Bellairs et al., 1996: Peñuelas et al., 1997; Peñuelas and Filella, 1998; Aparicio et al., 2000; Araus et al., 2001) (Fig. 9). Modern narrow‐band spectroradiometers measure the irradiance at different wavelengths with a band width of approx. 2 nm through the visible (400–700 nm) and near IR regions (700–1100 nm). Most spectral indices use some specific wavebands in the range between 400 and 900 nm and only a few indices use longer wavelengths, such as 970 nm in the Water Index (Peñuelas and Filella, 1998).

Fig. 8. Typical variation in the spectra due to differences in water regimes. The three lines correspond to plots of wheat measured around mid‐grain filling at ICARDA. a, Rainfed in a very dry environment; b, rainfed in a moderately dry environment; and c, irrigated. The main differences are: the magnitude of the increase in reflectance at around 700 nm, which indicates differences in biomass; the pattern in the PAR region, which indicates differences in pigment composition; and the decrease in reflectance beyond 930 nm which indicates differences in water content.

Fig. 9. Prediction of grain yield based on the combination of three different spectroradiometrical indices calculated from the spectra reflected by the canopy. Measurements were made (in early June 1998) at mid‐grain filling in a set of 74 genotypes of durum wheat corresponding to the WANADIN collection developed by the CIMMYT/ICARDA breeding programme. The three canopy reflectance indices used were SAVI = (R₉₀₀ – R₆₈₀)/(R₉₀₀ + R₆₈₀ + L) × (1 + L) with L = 0·5 for most crops; SIPI = (R₈₀₀ – R₄₃₅)/(R₄₁₅ + R₄₃₅); and NPQI = (R₄₁₅ – R₄₃₅)/(R₄₁₅ + R₄₃₅). The function of prediction was grain yield = 3616 – 4297SAVI – 1370SIPI – 3240NPQI. Its performance seems to be based on the differences across genotypes in the date of maturity. Plants were grown at Tel‐Hadya (headquarters of ICARDA), NW Syria (Casadesús, Araus and Nachit, unpubl. res.).

Perhaps the most widespread application of reflectance indices is in the assessment of any of a group of parameters related to the intensity of canopy greenness. These parameters are related to the photosynthetic size of a canopy and include green biomass and LAI. The most widespread vegetation indices are simple ratios (SR = R_NIR/R_R) and the normalized difference vegetation index (NDVI = R_NIR – R_R/R_NIR + R_R ). As pointed out above [see eqn (1)], the amount of green area of a canopy determines the absorption of photosynthetic active radiation by photosynthetic organs, which in turn determines potential production by the canopy. By periodic measurements of reflectance during the growing cycle of a crop, we can assess the cumulative absorption of PAR by the crop, which is one of the parameters determining total biomass and thus final yield. However, the reliability of such indices for the assessment of grain yield seems to be affected by the amount of green biomass and culm density of the canopy.

Another potential application of reflectance indices is remote detection of water status. In this regard the water index (WI = R₉₀₀/R₉₇₀) closely matches relative water content, Δ¹³C and g_s, through drought and salt stress (Peñuelas et al., 1997; Peñuelas and Filella, 1998). However, besides the decreased accuracy of ground spectroradiometers in this band of the IR, this index appears to be affected mostly by severe drought stress.

While the photosynthetic capacity of a canopy can easily be estimated by vegetation indices which correlate with the photosynthetic size of the canopy, actual photosynthesis may not match the photosynthetic capacity due to variability in photosynthetic use efficiency of the absorbed radiation, especially when plants are exposed to unfavourable conditions. The photochemical reflectance index (PRI = R₅₃₁ – R₅₇₀/R₅₃₁ + R₅₇₀) has been developed with the aim of detecting the pigment changes in the xanthophyll cycle associated with changes in photosynthetic radiation use efficiency (PRUE; Filella et al., 1996). PRI has been shown to track changes in PRUE induced by different factors such as nutritional state (Filella et al., 1996) and drought stress (Tambussi et al., 2000, 2002).

Leaf pigments can easily be detected and quantified from reflectance spectra. The usefulness of pigment remote sensing includes the assessment of the phenological stage of the crop and the occurrence of several stress factors (Blackburn, 1998; Peñuelas and Filella, 1998). For example, changes in the ratio between carotenoids (Car) and chlorophylls (Chl) can be associated with senescing processes either due to the natural pattern of ontogeny of the plant or triggered by different stresses. Also, phenological stages can be associated to different Car/Chl values. In addition, several indices have been developed that are related to changes in pigment composition and which can be used for the remote detection of nutrient deficiencies, environmental stresses, pest attacks, etc. (see Peñuelas and Filella, 1998; Araus et al., 2001). In such contexts by periodic assessment of leaf area, the leaf area duration can also be used as an indicator of resistance to several environmental stresses.

FACING THE FUTURE: POTENTIAL USE OF NEW GENETIC TECHNOLOGIES TO FURTHER INCREASE CEREAL YIELDS

In the preceding discussion we have stressed the need to look for alternatives to complement conventional breeding for yield per se. Integration of novel techniques and methodologies into conventional programmes is needed to facilitate the identification, characterization and manipulation of genetic variation for continued and accelerated progress (Sorrells and Wilson, 1997). There are currently some new techniques, at different stages of development and application, which may be used in an integrated approach to crop improvement to increase the efficiency of plant breeding. One of these is the doubled haploid system, which accelerates the breeding process by reducing drastically the time needed to reach a sufficient degree of homozygosity. In addition, it increases the response to selection since it eliminates the dominance effects for major genes and increases the proportion of additive genetic variation available for selection for quantitative traits (Snape, 1998). As a result, better discrimination power is achieved between crosses and between genotypes within crosses, and the selection response across generations is improved (Snape, 1982). Moreover, developmental genes have strong pleiotropic effects on a number of performance traits, including abiotic stress tolerance. Much of the genetic variation for improving stress tolerance has been lost during domestication, selection and modern breeding, and it is unlikely that large improvements for abiotic stress tolerance can be made by altering the composition of developmental genes. Exploiting the genetic variation among the wild relatives of modern crops could provide information on stress‐tolerant genes that are unlinked to essential agronomic traits. It is in this context that wide crosses acquire their significance.

There is no doubt that the use of DNA‐marker technology will increase greatly in the future. The recent development of techniques to score variation at the molecular level has expanded the variability available for selection. DNA‐based markers may be useful tools to improve the efficiency of selection, not only for Mendelian traits, for which individual phenotypes provide much information about the underlying genotypes, but also for most of the complex traits of agronomic interest, for which phenotypes provide little information about the underlying genotypes. The detection and location of loci affecting QTL enable the use of marker‐assisted selection (MAS) for attributes difficult to manage by conventional breeding approaches, leading to a potentially more reliable, quick and efficient selection. Detailed information is provided by Tuberosa et al. (2002).

The ability to transform wheat opens up opportunities to increase wheat yields using molecular biology. Although there is relatively abundant information demonstrating that the introduction of a new gene into a cereal genome can dramatically modify end‐use quality or change the tolerance to relatively simple stresses (particularly insects and diseases, and tolerance to herbicides), no compelling evidence exists that the introduction of a new gene may result in potential increases in yield or in tolerance to a wide and variable (in occurrence and intensity) range of stresses (for rice, see Sheehy et al., 2000). Indeed yield is a highly quantitative trait by nature, with many different levels of organization (from molecular to the whole canopy) interacting (Richards, 1996a). While some workers report that they have identified QTLs for yield, water use efficiency, and other complex quantitative characteristics [examples in Slinkard (1998) and in this issue], the G × E interactions on the expression of these QTLs are frequently large (e.g. Kjaer and Jensen, 1996). As cultivars used in any realistic breeding programme must be cultivated over a large area and during several seasons, it is likely that the cultivar will experience a wide range of soil and weather conditions (Cooper et al., 1997). In this scenario, what appears indispensable is better understanding of the causes underlying the G × E interaction through collaboration between crop physiologists, plant breeders and molecular biologists. These promising techniques may then be put to wider use to increase yield potential and tolerance to complex and largely unpredictable stresses (Slafer and Otegui, 2000).

Rapid genetic progress will be directly proportional to our ability to precisely target key traits and to identify and locate gene sequences controlling agronomically important traits (examples in Snape, 1998). In fact, genetic analysis is the first requirement for cloning economically important genes, and the extensive collinearity appearing between the genetic maps of different cereal species (Moore et al., 1995; Devos and Gale, 1997) will be of paramount importance for future yield gains. Identification of appropriate parents and efficient selection of progeny in early generations becomes easier once a key trait underlying yield is targeted (Austin, 1993). Increasing our understanding of physiological–ecological processes at the crop level of organization, and thus contributing unequivocally to greater yields, may help in identifying realistic opportunities for future breeding (Slafer et al., 1999), whether these traits are employed in breeding directly (e.g. Araus, 1996) or in improving the contributions of molecular biology in breeding for quantitative, complex traits.

ACKNOWLEDGEMENTS

This study was supported by the research CICYT projects PB97–0865 and AGF 99–0611‐C03 (Spain).

Received: 25 May 2001; Returned for revision: 16 October 2001; Accepted: 12 November 2001.