Euphytica 54 : 2 2 1-229, 1991 .© 1991 Kluwer Academic Publishers . Printed in the Netherlands .

Growth analyses of oat lines with low and high groat-oil content*

H. Schipper' & K .J . FreyIowa State University, Department of Agronomy, Ames, IA 50011, USA ;1present address : Vanderhave Research, P.O. Box 1, 4410 AA Rilland, The Netherlands

Received 21 December 1990 ; accepted 28 March 1991

Key words: Avena saliva L., groat-oil content, growth analysis, oat, phenotypic recurrent selection

Summary

A considerable increase in groat-oil content of oat (Avena sativa L.) has been accomplished recently . Theobjective of this study was to determine whether physiological traits of oat plants with high groat-oil contentare changed in a way that would provide the energy needed for increased groat-oil content . Growth analyseswere conducted in 1987 and 1988 on 25 oat lines with low and on 25 with high groat-oil content . Threeharvests were made in 1987 and six in 1988. Plot biomass, plant weight, leaf area, and leaf weight weremeasured at each harvest and estimates for relative crop growth rate, unit leaf area, leaf area ratio, specificleaf area, and leaf weight ratio were derived. Grain yield, groat-oil content, and groat-protein content weremeasured at maturity . The data were analysed by using the stepwise multivariate analysis of variancetechnique . The results suggest that changes in growth characteristics and increases in photosynthetic capacityprovide the extra bioenergy required for synthesis of more groat oil . The groat-protein content did notchange as a result of increased groat-oil content in oat .

Abbreviations :BM - biomass, GF - groat fraction, GO - groat-oil content, GP - groat-protein content, GW- groat weight, GY - grain yield, HD - heading date, HI - harvest index, HO-lines - lines with high groat-oilcontent, HT - plant height, l - mean level, LA - leaf area, LAR - leaf area ratio, LO-lines - lines with lowgroat-oil content, LW - leaf weight ; LWR - leaf weight ratio, m - slope, PW - plant weight, q - curvature,RGR - relative crop growth rate, SLA - specific leaf area, SW - seed weight, ULR - unit leaf rate

Introduction

Oat cultivars currently grown in Midwestern U .S.A .have groat-oil contents that range from 40-60 gkg- ' (Frey & Hammond, 1975) . Branson & Frey(1989a) showed that, after three cycles of pheno-typic recurrent selection, the mean groat-oil con-tent (GO) of the 10 highest lines was 131 g kg -1 .The question arises whether oat plants with a high-

er GO have significantly altered plant type, devel-opment, and physiology .

The bioenergy required for synthesis of oil in aplant is more than that required for synthesis of anequal amount of protein or carbohydrate (Penningde Vries et al ., 1974) . Therefore, an increase in GOof oat requires either (a) more photosynthate, (b)reduced groat-protein content (GP) and/or grainyield (GY), or (c) increased translocation into thegroats. The objective of this study was to use

* Journal Paper No. J-14325 of the Iowa Agric . and Home Econ . Exp . Stn ., Ames, IA 50011, USA, Project No . 2447 .

222

growth analyses to determine whether an increasein GO of oat was associated with an increase inphotosynthetic capacity and/or with concomitantchanges in GP and GY.

Materials and methods

Materials and field experiments

The materials for the study in 1987 were 25 oat lineswith low and 25 oat lines with high GO . The oatlines were selected from the cycles C ° through C3 ofa recurrent selection program conducted to in-crease GO of oat by Branson & Frey (1989a) . Thematerials used in 1988 included 5 lines with low and5 lines with high GO from each of the C °, C1 , C2 , C3 ,and C4 of the same recurrent selection program . Alllines were S °-derived. Lines with low and high GOwill be referred to as LO- and HO-lines, respec-tively .

Field experiments were conducted on a Nicolletsilty loam (fine-loamy, mixed, mesic Aquic Haplu-doll) soil at the Agronomy and Agricultural Engi-neering Research Center near Ames, Iowa . Theexperiments, each with 50 entries and several har-vest dates, were grown in split-plot designs withthree replications in 1987 and four in 1988 . Harvestdates were applied to whole plots and entries wereassigned to subplots . In 1987 harvests were made atboot stage, 14-18 days post anthesis, and maturity,and in 1988 at three-leaf, first node visible, booting,anthesis, 14-18 days post anthesis, and maturity .Experimental areas received preplant broadcastapplications of 34, 22, and 28 kg ha -1 of N, P, andK, respectively . A subplot was a hill sown with 20seeds and the hills were spaced 30 cm apart in per-pendicular directions . Two rows of hills were sownaround each whole plot to ensure competition forperipheral subplots .

Growth analysis

Four traits were measured on each subplot :Biomass (BM) - dry weight of the bundle of

culms in Mg ha-1 .

Weight per plant (PW) - mean dry weight of fourrandom plants in g .

Leaf area per plant (LA) - mean green leaf areaof four random plants in cm 2 measured with anelectronic planimeter .

Leaf weight per plant (LW) -mean dry weight ofgreen leaves from four random plants in g .

PW, LA, and LW were measured at all harveststages except maturity .

To provide appropriate values for estimating rela-tive crop growth rate (RGR), unit leaf rate (ULR),leaf area ratio (LAR), specific leaf area (SLA), andleaf weight ratio (LWR), quadratic functions of timecentered around the mean harvest date (t) were fittedfor each subplot for each of the four traits measured :

1nBM(t) = a° + S°(t - t) + Z°(t - t)2 ,lnPW(t) = a + (3 (t - t) + 'r (t - t)2 ,1nLA(t) = a' + R'(t - t) + i'(t - t)2 ,1nLW(t) = a" + R"(t - t) + ti"(t - t) 2 .

Logarithmic transformation of the data was used toavoid heteroscedasticity . In 1987, only linear func-tions could be fitted for PW, LA, and LW. Func-tions of time for RGR, ULR, LAR, SLA, andLWR were derived by using the following equa-tions from Garretsen & Keuls (1986) :

RGR(t) = (1BM) (dBM(t)/dt)•

13° + 2x°(t - t),

ULR(t) _ (1/LA) (dW(t)/dt)_ ((3 + 2x(t - t)) x

e{(a- a')+(3- P') (t- f)+ (T- t') (t - 1) 2 )

LAR(t) = LA(t)/PW(t)•

e{(a' - a) + (0' - 3) (t - f) + (T' - Z) (t - f)2)

SLA(t) = LA(t)/LW(t)e{(a' - a") +(A'- 3") (t - , + (r'- t") (t - •/2)

LWR(t) = LW(t)/PW(t)•

e{(a"- a)+(P"- P) (t- f)+ (t" - T) (t - 1)2 )

Next, RGR, ULR, LAR, SLA, and LWR wereestimated on a subplot basis . RGR equals the dailyincrease in biomass per unit biomass . ULR equalsthe daily gain in plant weight per unit leaf area .LAR, which is an index of plant leafiness, is the

ratio of plant leaf area to plant weight . SLA, whichis an index of leaf density, is the ratio of plant leafarea to plant leaf weight . And LWR, which is theratio of plant leaf weight to total plant weight, is anindex of leafiness on a weight basis .

Data were analysed by using the stepwise multi-variate analysis of variance technique suggested byKeuls & Garretsen (1982) . For BM, PW, LA, andLW, orthogonal polynomials of the form f(t) =m.zo(t) + l .z,(t) + q •z2(t) were fitted for each sub-plot. The coefficients m, 1, and q are estimates forthe mean level, the slope, and curvature, respec-tively . Further, z o is a vector containing 1's, z, =(t - t ), and z2 is a vector such that z o 'z, = 0 andz,'z2 = 0 (for derivation see Keuls & Garretsen,1982). Garretsen & Keuls (1986) showed how ap-proximate quadratic functions of time centeredaround t can be derived for ULR, LAR, SLA, andLWR using Taylor series expansions: f(t) + r(t)(t - t) + fl(t)1/2(t - t) 2 . Estimates of f(t), F(t),and, where appropriate, f'(t) were computed foreach subplot .

To test for significance of the contrast LO-linesversus HO-lines, within-year analyses of variancewere computed for the coefficients m, 1, and qderived from the orthogonal polynomials of BM,PW, LA, and LW and for f(t), F(t), and F(t) de-rived from the Taylor expansions of RGR, ULR,LAR, SLA, and LWR. The contrast was testedagainst the contrast-replication interaction meansquare for each analysis of variance . Multivariate

Table 1 . Agronomic traits recorded at maturity

Trait

Description

Heading date (HD)Plant height (HT)Biomass (BM)Grain yield (GY)Harvest index (HI)Seed weight (SW)Groat weight (GW)Groat fraction (GF)Groat-oil content (GO)

Groat-protein content (GP)

223

significance tests for interaction over time betweenmeans for LO- and HO-lines for each trait wereobtained after polynomial transformation of thedata. Wilk's lambda was computed for each testand a test of significance was performed as suggest-ed by Rao (1951) .

In addition, simple linear regressions were com-puted using 1988 data from LO- and HO-lines . Thecoefficients m, 1, and q or f(t), F(t), and f°(t) of atrait were regressed on cycles of selection . Thisdetermined whether changes occurred in traits thatwere associated with selection for high GO .

Agronomic traits

The agronomic traits that were recorded or calcu-lated are shown in Table 1 . Groat weight (GW) wasnot recorded in 1987, and GP was measured onlyon two replications . An analysis of variance wascomputed for each year and for each trait . Repli-cations were random and genotypes fixed effects .The LO-lines versus HO-lines contrast was testedagainst the interaction mean square . By utilizingdata of LO- and HO-lines from 1988, each trait wasregressed upon cycles of selection . Regressioncoefficients were tested for significance via a t-test .Phenotypic correlations of GO with agronomic traitswere computed for each year, and evaluated forsignificance via Bonferonni t-tests (Bailey, 1977) .

Number of days from planting to 50% of panicles emerged .Distance from ground to tips of panicles in cm .Dry weight of bundle of culms in Mg ha - 'Dry weight of threshed grain in Mg ha' .(GY/BM) x 100 in % .Weight of 100 seeds in g .Weight of 100 groats in g .(GW/SW) x 1000 in g kg- ' .Measured on 4 .0-6 .0 g of groats by wide-line nuclear magnetic resonance spectroscopy(Conway & Earle, 1963) and expressed in g kg -1 .N-content measured on 4 .0-6 .Og of groats by the near infrared reflectance method(Williams, 1975) and multiplied by 6 .25 and expressed in g kg - ' .

224

-2 ,50 60 70 80 90 100 110

Number of days after planting

Ln LA(t), LA in cm 25 .0 -

4 .0

3 .0

Fig . 1(c)

2 .050

60

70

80

90Number of days after planting

RGR(t), Mg ha Mg -1 ha" day-1

0 .15fFig . 1(e)

0 .10

0 .05 ~I

0 .00

-0 .05

200

100

40 50 60Number of days after planting

LAR(t), cm '2 day"I

300Fig . 1(g)

LWR(t), g 9 1

1 .0Fig . 1(i)

. . . . . ... . . . . . . .. .. . .

70 80 90 100

0.0 50

60

70

80

90Number of days after planting

110

0

-1

-2

-3

-4 -50

60

70

80

90Number of days after planting

ULR(t) . g cm -2 day'

.003Fig . 1(f)

.002

.001

.000 50

60

Ln LW(t), LW in g

Fig. 1(d)

70

80

90Number of days after planting

SLA(t), cm -2 9,1

Number of days after planting

Fig. 1 . Growth curves for various traits of LO-lines and HO-lines regressed on number of days after planting (--=1987, LO-lines ;1987, HO-lines ; . . . = 1988, LO-lines ; - = 1988, HO-lines) .

300Fig . 1(h)

200

100 50

60

70

80

90

Table 2 . Means of coefficients m, 1, and q for LO- and HO-lines for the traits BM, PW, LA, and LW in 1987 and 1988

Results and discussion

Growth analysis

The time functions for the means of LO- and HO-lines are plotted by trait and year in Fig. la-i . Themeans for LO- and HO-lines for coefficients m, 1,and q of BM, PW, LA, and LW and for parametersf(t), F(t), and f"(t), of RGR, ULR, LAR, SLA,and LWR are presented in Table 2 and 3, respec-tively . The linear regressions are shown in Table 4

q

HO-lines

LO-lines

HO-lines

0.028

-0.0009

-0.00110.0530.000570 .0021

0.044 -0.0012 -0.00110.067 -0.0016 -0.00160.0065 -0.0015 -0.00150.010

-0.0014

-0.0015

* Contrast between LO- and HO-lines is significant at the probability level 0 .05 and 0 .01, respectively .

225

and the significance tests for interaction over timebetween LO- and HO-lines are given in Table 5 .

The coefficient m for BM was significantly high-er for HO- than for LO-lines in 1988 and 1 for thistrait was significantly higher for LO-lines in 1987(Table 2 and Fig . la). In both years LO- and HO-lines interacted over time (Table 5) . When m, 1,and q for BM were regressed upon cycles of selec-tion, the regression coefficients for all three weresignificant . The m and q increased whereas 1 de-creased over cycles. Thus, biomass increased but

Table 3. Means of parameters f(i), f' (t) and F(i) for LO- and HO-lines for the traits RGR, ULR, LAR, SLA, and LWR in 1987 and 1988

* ** Contrast between LO- and HO-lines is significant at the probability level 0 .05 and 0.01, respectively .aThe values for ULR were all multiplied by 1000 .

Trait f(i) f'(t) r(t)

LO-lines HO-linesLO-lines HO-linesLO-lines HO-lines

1987 :RGR 0.029* 0.028 -0.0018 -0.0021ULR* 1 .04 1 .13* 0.054 0.060LAR 49 .9 47 .5 -2.51 -2.49SLA 242 246 -0.70 -0.44LWR 0.21 0.19 -0.0098 -0.0099

1988 :RGR 0.054 0.052 -0.0024 -0.0023ULR* 1 .22** 1 .13 0 .021** 0.017 -0.0026 -0.0025LAR 61 .6 64.2** -3.77 -3.89 0 .25 0.25SLA 223 232** - 1 .11 -0.90 -0.020 0.006LWR 0.28 0.28 -0.016 -0.016 0.00098 0.00096

Trait m I

LO-linesLO-lines HO-lines

1987:BM 1 .85 1 .80 0.029*PW 0.70* 0 .61 0 .051LA 4.60** 4 .46 0.00072LW - 0.88** -1.04 0.0038

1988 :BM 0.87 0 .98** 0.045PW -0.21 -0.23 0.069**LA 3.93 3 .95* 0.0076*LW -1.47 -1 .49 0 .013

226

Table 4. Regression coefficients form, l, and q and f(t), V(i) and f'(i) regressed on cycles of selection for the traits BM, PW, LA, LW,RGR, ULR, LAR, SLA, and LWR in 1988

*, ** Indicate significance at the probability level 0 .05 and 0 .01, respectively.a The values for ULR were all multiplied by 1000 .

the rate of biomass accumulation decreased withselection for GO . These results suggest that in-creasing the GO resulted in lines with greater vigorduring early growth but with a lessening of thiseffect during later growth .

The m for PW was higher for LO- than for HO-lines during both years, with the difference beingsignificant only in 1987 (Table 2) . The growthcurves for LO- and HO-lines in 1988 coincidedduring early growth but diverged during latergrowth (Fig. lb), causing a significant differencefor 1 in 1988 . For PW, m increased significantly overcycles of selection and 1 decreased .

In 1987, LO-lines had a significantly greatermean for LA than did HO-lines . In 1988, however,the reverse occurred (Table 2) because the curvesfor LO- and HO-lines were dispersed during earlygrowth but coincided during later growth (Fig . 1c) .The 1 was significantly higher for LO-lines in 1988 .There was no interaction between LO- and HO-lines for LA over time in both years (Table 5) . Theregression was significantly negative for 1 over se-lection cycles, but was not significant for m . Thus,mean LA over time is not affected by selection forhigh GO but the rate of increase in LA sloweddown .

The mean for LW in both years was greater forLO- than for HO-lines, but the difference was sig-nificant only in 1988. The regression for m was not

respectively .

significant over cycles of selection but for 1 it wassignificantly negative . Thus, as with other traits,the rate of increase in LW decreased with selectionfor high GO .

Biomass, PW, LA, and LW provide informationabout plant growth. RGR, ULR, LAR, SLA, andLWR, however, tend to give more insight intogrowth patterns . The parameter f(t) for RGR washigher for LO-lines than for HO-lines, but signif-icantly so only in 1987 (Table 3) . Interaction be-tween LO- and HO-lines for RGR was significantin 1987 but not in 1988 (Table 5) . There was a

Table 5. Significance tests for interaction between LO- andHO-lines over time in 1987 and 1988

*, * * Indicates significance at the probability level 0 .05 and 0 .01,

Trait b ± SE

m 1 qBM 0.033** ± 0 .0087 - 0 .0010** ± 0 .00025 0 .000030** ± 0 .000011PW 0.013** ± 0 .0062 - 0.00081** ± 0 .00027 0 .000023 ± 0 .000022LA 0.0044 ± 0 .0067 - 0 .00071 ± 0 .00035 0 .000045 ± 0 .000023LW -0.0045 ± 0 .0071 - 0.0010* ± 0 .00041 0.000022 ± 0 .000029

f(i) f'(t) f'(t)RGR - 0.0013** ± 0 .0003 0.0000059** ± 0 .000022ULRa 0.0011 ± 0 .0089 -0.000057 ± 0.00072 -0.000075 ± 0 .00013LAR - 0.70 ± 0 .44 0.048 ± 0 .026 -0.00063 ± 0 .0025SLA 1.04 ± 0 .80 0.042 ± 0 .055 0 .0098 ± 0 .0090LWR - 0.0044** ± 0 .0017 0.00020* ± 0 .000093 -0.000011 ± 0 .000007

Trait Wilks' lambda

1987 1988

BM 0.933** 0.899**PW 0.989 0 .950*LA 1 .000 0 .981LW 0.992 0 .963RGR 0.971* 0.982ULR 0.965* 0 .951LAR 0.999 0 .982SLA 0.995 0 .985LWR 0.998 0 .974

significant decrease in f(t) and an increase in f' (t)over cycles of selection . Therefore, selection forhigh GO caused a reduction in the relative dailyincrease in crop biomass . This indicates that in latercycles of selection, oat plants diverted relativelyless bioenergy to produce biomass, which shouldmake more bioenergy available for groat-oil syn-thesis. Over cycles of selection, there was a signif-icant positive regression for F(t) of RGR . Thisindicates a change in the growth pattern of oat lineswith selection for high GO .

The ULR showed little variation over time (Fig .lf) . The parameter f(t) for ULR was significantlygreater for HO-lines then for LO-lines in 1987, butthe reverse occurred in 1988 . Parameter f'(t) wassignificantly higher for the LO-lines in 1988 (Table3) . Significant changes did not occur in f(t), F(i) orF(t) of ULR over cycles of selection (Table 4) .Therefore, selection for greater GO did not causeany change in the daily increase in plant biomassrelative to the amount of leaf area . That is, a similaramount of bioenergy per unit leaf area is used forproduction of plant biomass . Historically, ULRhas been viewed as being synonymous with netassimilation rate . However, ULR is based solely onleaf area, neglecting thickness, density, or weightof leaves and all genotypes are assumed to haveequally efficient photosynthesis .

The f(t) values of LAR were similar for LO- and

Table 6. Means of traits measured on LO- and HO-lines in 1987and 1988 and regressions of traits on cycles of selection

Trait 1987

1988

LO-lines HO-lines LO-lines HO-lines b ± SE

*, * * Contrast between LO- and HO-lines is significantly differ-ent at the probability level 0 .05 and 0 .01, respectively .

227

HO-lines in 1987, but significantly higher for HO-lines in 1988 . Thus, in 1988, HO-lines had more leafarea relative to the plant biomass than did LO-lines. HO-lines may have a greater photosyntheticcapacity . The parameter f(t) for LWR was similarfor LO- and HO-lines in both years (Table 3) . ForSLA, however, the f(t) was greater for HO- thanfor LO-lines in both years but only significantly soin 1988. If plants from LO- and HO-lines haveequal biomass, our results for LAR, LWR, andSLA indicate that HO-lines have larger leaves thanLO-lines but leaves of both sets have similarweights. This may indicate that HO-lines have ahigher photosynthetic capacity . The regressionanalysis did not indicate significant changes in f(t),f' (t), and f'(t) over cycles of selection for LAR andSLA. The regression of f(t) for LWR, however,was significantly negative indicating that selectionfor high GO caused a reduction in leafiness . Thiscould result in a smaller leaf area relative to plantweight and therefore in a reduction of LAR overcycles of selection . It could also result in largerleaves relative to their weight, and, therefore, in anincrease in SLA over cycles of selection and a con-comitant increase in photosynthetic potential .Signs of the regression coefficients for f(t) of LARand SLA are in agreement with this hypothesis .

In summary, selection for higher GO in oat hasresulted in changes in growth patterns . Reductionshave occurred in the rates with which plant and leaf

Table 7. Phenotypic correlations of groat-oil content with othertraits of oats in 1987 and 1988

GO

*, ** Significant at the joint probability level of 0 .05 and 0 .01,respectively .

1987 1988HD 79* 77 69* 69 - 0 .5** ± 0 .1HT 86 86 67 67 -1 .4* * ± 0 .4 HD - 0.50** -0.27BM 8.0** 6 .8 6 .3 6 .7 0.06 ±0 .07 HT -0.05 -0.24GY 4.0** 3 .3 2 .9 3 .2* 0.08** ±0 .04 BM - 0.46** 0.26HI 50 49 45 47* 1 .0** ± 0 .3 GY -0.46** 0.45**SW 2.68* 2 .40 3 .04 2 .98 0.014 ± 0 .017 HI -0.13 0.55**GW - 2 .09 2 .08 0.007 ±0 .014 SW - 0.60** -0.14GF 721 720 689 698 -1 .4 ± 2 .6 GW - -0.09GO 79.7 111 .8** 84 .3 106 .4** GF -0.03 0 .06GP 158.4 160 .7 185 .6 185 .3 -0.50 ± 0.44 GP 0.10 -0.15

228

biomass and leaf area increase during plant devel-opment. According to analyses of RGR and ULR,HO-lines from advanced cycles of selection spendless bioenergy on production of biomass per unitbiomass, but the biomass produced per unit leafarea did not change over cycles of selection . Ana-lyses of LAR, LWR, and SLA indicated that HO-lines had greater photosynthetic capacity than LO-lines . The results from the growth analyses suggestthat the additional bioenergy required for synthesisof more groat-oil in HO-lines is obtained from bothaltered growth rates and increased photosynthesis .

Agronomic traits

The HO-lines headed earlier than LO-lines in bothyears, and the regression of heading date (HD) oncycles of selection was significant and negative (Ta-ble 6) . In contrast, Branson & Frey (1989b) foundno changes in HD over three cycles of selection forhigh GO. LO- and HO-lines did not differ for plantheight (HT), but there was a negative and signif-icant regression over cycles for this trait . In 1987,GY was significantly higher for LO- than for HO-lines, but, in 1988, the reverse happened . Grainyield increased significantly over cycles of selec-tion. Branson & Frey (1989b) also found an in-crease in GY over cycles but it was not significant .In 1988, LO- and HO-lines had similar means forseed weight (SW), GW, and groat fraction (GF),but harvest index (HI) was significantly higher forthe HO-lines. In 1987, SW was significantly higherfor the LO-lines . Means for GP were similar forLO- and HO-lines, and the regression over cyclesof selection was not significant . Thus, selection forhigher GO was not associated with a reduction inGP, but the effect on GY was not consistent .

Groat-oil content was not correlated with HT,GF, and GP either year or with GW in 1988 (Table7) . Youngs & Forsberg (1979) and Gullord (1980)did not find consistent correlations between GOand GP in oats . Forsberg et al . (1974) and Gullord(1986) reported negative correlations . In bothyears, HD was negatively correlated with GO butnot significantly so in 1988 . The correlation be-tween GO and SW was negative, and that between

GO and GY was inconsistent . Positive correlationsbetween GO and GY in oats have been reported byForsberg et al . (1974), Gullord (1980), and Thro &Frey (1984), but Branson & Frey (1989b) foundindependence .

To summarize, selection for higher GO in oat didnot result in lower GP and there was no clearevidence for reduced GYs. The extra bioenergyneeded to produce groats richer in oil may haveoriginated from increased production of photosyn-thate .

Acknowledgements

We are indebted to Dr. D.E. Alexander, Dept . ofAgronomy, University of Illinois, Urbana, Illinois,and Dr. D .N. Peterson, Dept . of Agronomy, Uni-versity of Wisconsin, Madison, Wisconsin . The oiland protein determinations were conducted intheir laboratories, respectively .

References

Bailey, B .J.R., 1977 . Tables of the Bonferonni t statistic . J .Am. Stat . Assoc . 72 : 469-478 .

Branson, C.V. & K .J . Frey, 1989a . Recurrent selection forgroat-oil content in oat . Crop Sci . 29 : 1382-1387 .

Branson, C .V. & K .J . Frey, 1989b . Correlated responses torecurrent selection for groat-oil content in oat . Euphytica 43 :21-28 .

Conway, T. F. & F. R . Earle, 1963 . Nuclear magnetic resonancefor determining oil content of seeds . J . Am. Oil Chem . Soc .40 : 265-268 .

Forsberg, R.A ., V.L. Youngs & H.L. Shands, 1974. Correla-tions among chemical and agronomic characteristics in certainoat cultivars and selections . Crop Sci . 14 : 221-224 .

Frey, K .J . & E.G. Hammond, 1975. Genetics, characteristics,and utilization of oil in caryopses of oat species . J. Am. OilChem. Soc . 52 : 358-362 .

Garretsen, F . & M . Keuls, 1986 . Functions of time for growthcharacters, their evaluation and approximation to examinedifferences between genotypes. Euphytica 35: 11-15 .

Gullord, M ., 1980 . Oil and protein content and its relation toother characters in oats (Avena sp .) . Acta Agric . Scand . 30 :216-218 .

Gullord, M . ,1986.Oil and protein content in oats (A vena sativaL .) . p. 210-213 . In : D .A. Lawes & H . Thomas (Eds) . Proc .2nd Int. Oats Conf . Martinus Nijhof Publishers, Dordrecht .

Keuls, M . & F . Garretsen, 1982 . Statistical analysis of growthcurves in plant breeding . Euphytica 31 : 51-64 .

Penning de Vries, F.W.T., A.H.M. Brunsting & H.H. van Laar,1974 . Products, requirements and efficiency of biosynthesis : aquantitative approach . J . Theor . Biol . 45 : 339-377 .

Rao, C.R ., 1951 . An asymptotic expansion of the distribution ofWilks' lambda criterion . Bull . Int . Stat . Inst . 33 : 177-180 .

Thro, A.M. & K .J . Frey, 1984 . Relationship between groat-oil

229

content and grain yield of oats (Avena sativa L.) . Proc. IowaAcad. Sci . 91 : 85-86 .

Williams, P.C ., 1975 . Application of infrared reflectance spec-troscopy to analysis of cereal grains and oilseeds . Cereal .Chem. 52 : 561-576 .

Youngs, V.L. & R.A. Forsberg, 1979 . Protein-oil relationshipsin oats . Crop Sci . 19 : 798-802 .