Gain efficiency in short-term testing: experimental results USDA 1 Forest Service Forestry Sciences Laboratory, Olympia. WA, U.S.A. AND c.c. LAMBETH Container Corporation, Callahan, FL. U.S.A. Received March 8, 1991 Accepted September 11, 1991 LI, B., WU.LJAMS. C.G., CARLSON. W.C., HARRINGTON, C.A., and LAMBETH, C.C. experimental results. Can. J. For. Res. 22: 290-297. 1992. Gain efficiency in short-term testing: Height growth of loblolly pine 1 Pinus raeda L.) was measured in trees subjected to one of five irrigation and fertilization regimes in a closely spaced genetic test for 3 years. Shoot components of 3rd-year annual height increment were measured over two contrasting treatments. Ju\'enile height and number of stem units in summer growth length in the fully irrigated and fertilized short-term test regime exhibited (i) the highest juvenile-mature correlations (family mean correlation= 0.41-0.68), (ii) high individual-tree heritabilities I 0.38-0.44), which were two- to three-fold higher than older tree values in a conventional genetic test of the same families. (iifl high genetic stability across two extreme short-term test treatments (genetic correlation= 0.61-0.80), and (iv) an efficiency in genetic gain per generation of 81-87% relative to selection on height at age 8 years. LI, B., WILLIA.\IS. C.G.. CARLSO:'\. W.C .. HARRINGTON, C.A., et LAMBETH. C.C. experimental results. Can. J. For. Res. 22: 290-297. 1992. Gain efficiency in short-term testing: La croissance en hauteur chez le pin a encens (Pinus taeda L.) a ete mesuree chez des arbres ayant subi un des cinq regimes d'irrigation et de fertilisation appliques durant 3 ans dans un test genetique a faible espacement. Les composantes de Ia pousse annuelle a Ia 3" annee de croissance Ont ete mesurees selon deux traitements differents. Dans le test a court terme ou l' irrigation et Ia fertilisation etaient maximales. les mesures de hauteur juvenile et de nombre d'unites de tige accumulees pendant Ia croissance d'ete en longueur ont montre (i) les plus fortes correlations entre les stades juvenile et mature (0.41-0,68); (ii) des heritabiJites individuelles e\evees I 0.38-0,44) qui etaient deux a trois fois plus grandes que (eS valeurs derivees d' arbres plus ages representant les memes families dans un test genetique conventionnel; (iii) une stabilite genetique elevee entre les deux traitements extremes du test a court terme (correlation genetique = 0.61-0.80); (iv) une efficacite en gain genetique par generation variant entre 81 et 87c-c par rapport a Ia selection sur Ia hauteur a 8 ans. Introduction Height growth of juvenile trees in short-term tests is used to evaluate future performance in the tield because it is rela­ tively easy and inexpensive to measure and it generally cor­ relates well with later tree performance (Squillace and Gansel 1974; Lambeth 1983; Franklin 1983: Campbell et a/. 1985; Magnussen and Yeatman 1986; Riitters and Perry 1987; Williams 1987; Carter et a/. 1990). However, there has been speculation that shoot components that contribute to juvenile height, such as numbers of stem units (NSU) and mean stem unit length (MSUL), might be better indicators of later growth than juvenile height itself. Shoot growth components NSU and ISUL are under mod­ erate genetic control in conifer species (Kremer and Larson 1983; Bongarten 1986; Li et al. 1991). NSU is primarily responsible for the variation in shoot length ( Lanner 1968; Boyer 1970). 1The use of trade names does not imply endorsement of products used nor criticism of products not used. zPresent address: University of Minnesota North Central Experi­ ment Station, Grand Rapids, MN 55744. C.S.A. 3Author to whom all correspondence should be addressed. Printed in Canada I ImprimC au Clllada [Traduit par Ia rt!daction) Studies relating these shoot components to later field performance have shown some interesting but inconsistent results. MSUL was highly correlated with tree height in blue spruce (Picea pungens Engelm.) (Bongarten 1986), and MSUL of fixed growth in seedlings of maritime pine (Pinus pinaster Ait.) was significantly correlated with tree height at 6 years (Kremer and Xu 1989). However, 2nd-year annual height increment was found to rank 8-year height performance better than either MSUL or NSU in loblolly pine (Pinus taeda L.) (Bridgwater 1990). These studies did not quantify the genetic gain expected from selection based on juvenile height versus its compo­ nents. Since many studies have shown the value of juvenile selection in closely spaced short-term genetic tests (e.g., Franklin 1983; C arter et al. 1990; Williams 1986; Magnussen and Yeatman 1986; Bridgwater 1990), it is timely to (i) define optimum testing environment for closely spaced tests and (ii) evaluate shoot components on the basis of genetic param­ eter estimates and genetic gain efficiency rather than on the basis of family ranking or juvenile-mature correlations alone. The results will provide a basis for selection-age decisions in operational genetic improvement programs (Newman and Williams 1991). .93 LI ET AL. where X;k is the maternal half-sib family mean based on the ith maternal parent and the kth sample of the environment-time continuum. The casual components of the family mean correlation are as follows: I l ! ' l j • l! 1, · COVA1A2 are the variance and covariance components, respectively, among half-sibs for seedling and field growth where traits measured in different environments; \-E1, \-E2, cove1e2 are the variance and covariance components, respectively, for two random points in environment and time; Nt, N2 are the total number of individuals per family in the test(s). l ji l Thus rp r ---7 rg (genetic correlation) as N ---7 oo, since the environmental covariance, cove1e,. is assumed to be negli­ gible. Additive genetic variance -is the underlying genetic basis for progeny testing, and it is the casual component of interest to traditional conifer breeding programs that rely on sexual recombination as a means to sequester favorable alleles. , The association between the two ages depends on the mag­ nitude of the additive genetic covariance, covA A common p to the trait(s) at two ages. Thus the phenoty ic juvenile­ mature (family mean) correlation is a simple, elegant measure of progeny testing success in seedlings (Williams 1986). When the number of individuals in the families is large and testing locations and environments are random, the family mean correlation is close to the genetic correlation (Williams 1986). The product-moment correlation coefficients were estimated from phenotypic covariance and variance of family means. The significance of the difference between two corre­ lations was tested by the -test (Snedecor and Cochran 1967). Within the short-term test, the additive genetic correlation was estimated for each trait across the two treatments (1-F and NI-NF) to assess the stability of the juvenile traits; the genetic correlations between short-term treatments were esti­ ratio of the gains based on juvenile trait E = _ , = ­ 'F for the juvenile (J) and mature (m) traits, respectively; root of heritabilities of the juvenile U) and the mature (m) traits, respectively. The equation can be h1, hm are the square simplified with the conservative assumption that 7 [ ] E = l j \ j 1 li where is the gain at mature age based on selection for the juvenile trait; is the selection intensity in the juvenile trait; are the square root of heritabilities of the juvenile (j) hi, and mature (m) traits, respectively; rpr is the genetic corre­ lation of the juvenile (J) and maturl(m) traits; is the hm = im: hi r-pr,., hm Gain efficiency per generation (E) was compared between height at age 8 years and juvenile traits measured in the short-term test. Gain efficiency is considered the ultimate ''criterion of success" for early selection because it is the aggregate value of several other criteria: degree of additive genetic control, genetic stability across treatments, and degree of genetic commonality with the older tree selection goal. the least favorable treatment (NI-NF), juvenile-mature cor­ v"' ronmental effects was adjusted according to Yamada's (1962) methods. The genetic gain for mature performance from mass selection based on a juvenile trait is as follows (Falconer 1981): G i1 Family mean correlations between juvenile traits and 8-year height were highest for the full irrigation and fertilizer treat­ ment (1-F), which favored growth (Table 3; Fig. 1). Under where rg is the genetic correlation; 'f is the genetic variance among half-sib families; VFxr is the family x treatment inter­ action; t is the number of treatments. Bias in VFxr due to heterogeneous variance and fi xed envi­ l "' Optimum testing environment = ------ ( v} I Results VFxT rg and mature trait G· ijhJhm r-pc O'pf Gm imh O'pf,., where G1, Gm are the gains at mature age based on selection (6] mated for fixed effects (Yamada 1962): [4) j m gives the equation for genetic efficiency: i1 O'pr phenotypic standard deviation of the mature (m) tn'i'it. The relations were low (Tables 3 and 4), although individual-tree was slightly higher = 0.38 versus 0.55; heritability (h2) (h2 Table 5). Family performance averaged across all treatments showed lower juvenile-mature correlations than the best single treatment, 1-F (Tables 3 and 4). Juvenile-mature correlations Shoot growth components were measured on the third annual increment only; in comparison, the composite trait was limited to 3rd-year height (HT3) and its annual increment (HT3 - HT2). Summer NSU and annual increment measured on trees in treatment 1-F had equally high family mean cor­ relations with 8-year height (0.56 versus 0.57 for HT3 - HT2). All other component traits exhibited lower correlations with 8-year height in the North Carolina coastal progeny tests (Tables 3 and 4). Juvenile height and height increments were highly corre­ lated with age 8 height; the 2nd-year increment exhibited the highest correlation (HT2 - HTl, rpr = 0.68; Table 3; Fig. 2). 294 CAN. I. FOR. RES. V OL . 22. 1992 TABLE 3. Loblolly pine juvenile-mature family mean correlations of (i) 1-. 2-. and 3-year height and height increments in a short-term test with (ii) 8-year height Treatmenr' I-F Composite traita NI-NF I-NF I-lf2F Nl-VzF NCC progeny tests HTl 0.22 0.14 HT2- HTI HT3- HT2 0.68*** 0.57*** HT3- HTl 0.37* 0.36* 0.35* 0.48** 0.16 -0.08 0.70*** 0.38* 0.22 0.18 0.07 0.68*** 0.69*** HT2 HT3 0.34* 0.54*** 0.05 0.48** 0.33 0.34* 0.41 ** 0.35* 0.25 0.28 0.06 0.34* 0.20 0.45 0.56** 0.25 0.33 Jones County test HTl HT2 0.30 0.52** HT3 HT2- HTl 0.17 0.37 0.52** 0.52** HT3- HT2 HT3- HTI 0.42 0.44 0.29 0.38 0.45 0.52** 0.37 0.41 0.20 0.19 0.29 0.34 0.53** 0.54** 0.47** 0.52** 0.20 0.32 -0.03 0.18 NoTE: Eight·year height is based on North Carolina coastal (NCC) progeny tests and on a Jones County open· pollinated test. •, ••. •••. I. 2. and 3 years, respectively. Significantly different from zero at 10. 5. and l "HTI, HT2, HT3. height at le1·eis, respectively. b See Table I for a description of each treatment. 30,- 1: '*' c 0 j "" c OS a: 25 20 r r " I ;-( I l I [x l 0 0 )( : )( ! 15 i-10 3rd-year height increment. Unlike the North Carolina coastal progeny test results, other shoot growth components (total NSU and summer growth length) were strongly correlated with 8-year height (Table 4). Several shoot components in trees of treatment NI-NF also showed significant correlations with height performance. For example. fixed growth NSU for trees in treatment NI-NF was strongly, positively correlated (rpr= 0.74. p < 0.01) with 8-year height for trees in the Jones County test. ; :< )( :.: Genetic comrol ofjuvenile traits )( riil = 0.68*- ;.: * * 5 10 15 20 25 30 Rank based on a-Year Height FIG. 1. Plot of family ranks for 2-year height increment in treatment I-F (irrigation and fertilization) and 8 -year height in the progeny tests of 25 open-pollinated loblolly pine families. For example, the top five of the 25 families could have been correctly identified at age 2 and this also held true for correctly selecting the top 50% (Fig. 1). The height increment for the 1st year was higher than total height at age 1 year (Table 3), but after age 1 there was no statistical difference between total height and height incre­ ment. Also, juvenile-mature correlations reached a plateau after age 2 years and correlations did not increase in the 3rd year (Fig. 2). Correlations with the Jones County open-pollinated test provided corroborating results. Third-year increment and summer NSU exhibited the highest juvenile-mature correla­ tions;· here, summer NSU had a higher correlation than Total height and height increments of loblolly pine seed­ lings exhibited higher heritabilities relative to older tree height in conventional genetic tests (Table 5). For example. heritability in treatment I-F was 0.38 for 2nd-year height and 0.44 for its height increment. This is two to three times higher 2 than heritability at age 8 in the Jones County test (h = 0.16). Summer shoot NSU and 3rd-year annual increment had slightly lower heritability estimates (h 2 = 0.28 and 0.20) than 2 3rd-year total height (h = 0.39). Stability ofjuvenile traits Height at age 2 years was the most stable composite trait across extreme treatments (rg = 0.78; Table 5). Total 3rd-year height and its height increment were also consistently stable across treatment components (Table 5). For those components initiated and elongated in the same season (i.e., summer growth components), MSUL and NSU were both stable across extreme treatments (Table 5). Only fixed growth length and NSU were unstable across treatments (Table 5). For these shoot components, the genetic correlation measuring F x T interaction was low and negative (Table 5). Gain efficiency of 8-year height versus shoot components Compared with selection at age 8 years, selection on 2nd­ and 3rd-year height and corresponding annual increments would have doubled genetic gain efficiency per generation (Table 6). Selection based on the best shoot component. Ll ET AL. 0.8 TABLE 4. Loblolly pine juvenile-mature family mean correlations of (1) 3 y ­ ear shoot composite tmits and shoot component tmits with (ii) 8-year height 0.6 Treatmentb Trait" I-F NI-NF 0 NCC progeny tests Composite: age 3 HT3 HT3 - HT2 Component: age 3 NSU MSUL Fixed growth Length NSU MSUL Summer growth Length NSU MSUL 0.69*** 0.57*** 8 0.07 -0.08 -, 0.54*** 0.56*** -0.24 -0.09 0.08 -0.23 0.52*;;: 0.45 .t .r e e .) r y :s n ). !- iS n t, ,l ' ! 0.55** -0.5 1** -0.25 -0. 12 -0.07 0.53** 0.74*** -0.34 0.52** 0.65*** -0.4 1 0. 13 0.40 -0.48 :-ISU, ••, •••. number of The use of a fully irrigated and fertilized short-term test treatment appears to be best for maximizing juvenile-mature correlations at Fort Towson, Oklahoma. In this short-term test, irrigation seemed to be the limiting factor; the second and third choices were both fully watered with no or partial fer­ tilizer treatments (I-NF and I-Y2F; Tables 3 and 4 ). The short­ l term test was located west of the loblolly pine range, where I drought occurs frequently, so supplemental water may be Jl more important here than if the test were conducted along the higher rainfall areas of the Atlantic Coast. If so, there may not be a single short-term test treatment that is best across t esting sites. Before using short-term testing, it may be 1 J prudent to develop a protocol for optimum short-term test 1 • conditions in each physiographic region. l j - HT2 HT3 Height(cm) The differences between short-term test treatments were mainly due to magnitude of the variance and covariance com­ ponents rather than to family rank change. The genetic cor­ relations across treatments, adjusted for scale (eq. 4: Table 5), were highly positive for height traits and summer NSU: this suggested that differences in juvenile-mature correlations among treatments could not have been due to family rank change alone. Choice of juvenile traits for early selection Discussion I HTI 0.29 0. 19 summer growth NSU, was 87% as efficient as selection at age 8 years. The value is based on gain per generation so the potential time savings provides incentive for early selection in short-term testing. Optimum testing environment HTO FIG. 2. Changes in phenotypic family correlation (r:or) as loblolly pine seedlings grow. Correlations were calculated between juvenile heights and 8-year height in the North Carolina coastal (NCC) progeny tests (25 families) and in Jones County test ( 16 families). 0.52** -0.31 MSUL, mean stem unit length. b See Table I for a description of each treatment. I // 0.2 -0.2'--0 50 100 150 200 250 300 350 " HT2, HT3, height at 2 and 3 years, respectively; i / p <0.05 / ,___ L. 0.03 0. 14 -0.2 1 stem units: J ///, Jones County test -0.24 -0. 13 -0. 13 Significantly different from zero at 5 and l% levels. respectively. j I 0�----- progeny tests and on a Jones Coumy open-pollinated test. :i NCC Progeny tests 0. 1 1 -0.25 NoTE: Eight-yeur height is based on North Carolina coastal (NCC) i1 0 1 0.4 1** -0.23 Jones County test Composite: age 3 HT3 HT3 - HT2 Component: age 3 NSU MSUL Fixed growth Length NSU MSUL Summer growth Length NSU MSUL ,a, .:. 04 c 295 The summer NSU was superior to both summer and fixed­ growth MSUL as a selection criterion. This does not corrob­ orate reports for temperate conifer species (Bongarten 1986: Kremer and Xu 1989), where fixed-growth MSUL was the best seedling growth indicator. Species or taxa summer growth patterns may account for the incongruent results. Blue spruce produces only fixed growth, so there can be no comparison with summer NSU. Maritime pine is generally monocyclic, thus it infrequently produces summer growth. Loblolly pine, by contrast. pro­ duces 69% of the annual growth as summer growth length (Table 2). For other conifer taxa such as Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), summer growth is a distinctly juvenile trait, disappearing with advancing age (Kaya et al. 1989). In loblolly pine, summer growth contrib­ utes to annual increment until rotation age (Harrington 1991). Juvenile selection criteria may vary with provenances (Kaya et al. 1989), or in this case, with species. Juvenile-mature family mean correlations increased as seedlings increased in height in the 2nd year, then the corre­ lations plateaued (Fig. 2). There are two possible explanations for this. First, the third growing season experienced a severe drought, so that even trees in the irrigated portion of the study may not have received adequate irrigation. If so, the best treatment would have been more like the less-desirable non­ irrigated treatments (Tables 3 and 4) and juvenile-mature cor­ relations would have been unchanged or even lower in the 3rd year. CAN. J. FOR. RES. VOL. 22. 1992 296 TABLE 5. Heritability estimates (with their standard errors in parentheses) and genetic correlations for loblolly pine composite and component traits Heritability estimate Genetic correlation between treatments Treatment I-F" Treatment NI -NP I-F and NI-NF combined 0.12 (0.17) 0.38 (0.14) 0.39 (0.13) 0.40 (0.14) 0.20 (0.09) 0.36 (0.13) 0.32 (0.12) 0.55 (0.18) 0.59 (0.19) 0.54 (0.18) 0.43 (0.15) 0.61 (0.20) 0.10 (0.07) 0.35 (0.13) 0.31 (0.13) 0.36 (0.13) 0.14 (0.08) 0.30 (0.12) 0.29 0.78 0.61 0.80 0.41 0.59 0.23 (0.10) 0.36 (0.13) 0.44 (0.15) 0.44 (0.15) 0.24 (0.09) 0.34 (0.12) 0.75 0.82 Length NSU 0.28 (0.11) 0.30 (O.l l) 0.08 (0.06) 0.24 (0.10) 0.20 (0.09) 0.13 (0.07) 0.06 (0.07) 0.06 (0.08) 0.09 (0.05) -0.32 -0.40 0.85 Length NSU 0.24 (0.10) 0.28 (0.11) 0.28 (0.11) 0.38 (0.14) 0.43 (0.15) 0.38 (0.13) 0.09 (0.08) 0.20 (0.09) 0.31 (0.1 I) -0.06 0.53 0.53 Trait0 Composite HT I HT2 HT3 HT2- HTl HT3- HT2 HT3- HTI Component: age NSU 3 MSUL Fixed growth MSUL Summer growth MSUL "HTI, HTI, HT3, height at l, b See Table 1 2. and 3 years. respectively; The second explanation may be the static ratio of cyclic growth to total height growth in the 3rd year. Cyclic growth, a combination of fixed and summer growth cycles produced after the initial seedling's bud set, has been shown to be better correlated with mature height performance than total height in 1st- and 2nd-year loblolly pine seedlings in greenhouse and short-term studies (Williams 1987; Li et al. 1991). As cyclic growth increasingly contributes to total height, juvenile-mature correlations have been observed to increase (Williams 1987). After the 2nd year, the ratio of cyclic growth to height changes only slightly, and the plateau in this ratio may also result in a plateau in the juvenile-mature correlation in the 3rd year. On another point, shoot components were more highly cor­ related with 8-year height in the open-pollinated Jones County test than with performance based on the North Carolina coastal progeny tests. The Jones County test received superior silvicultural care compared with the North Carolina coastal progeny tests; this was exhibited in mean tree height (8.3 m for the Jones County test versus 6.2 m for the North Carolina coastal progeny tests). If intensive silviculture results in more fully expressed genetic potential, then field test quality may be more limiting to successful early selection than is genetic control of juvenile shoot components. Genetic control NSU. number of stem units; MSUL. menn stem unit length. for a description of each treatment. Shoot growth components tended to exhibit lower herita­ bility estimates than height growth traits (Table 5). Large measurement error was likely to be the cause, because shoot components were more difficult to measure than total height. There may also be a developmental basis; some fixed-growth components exhibited negative genetic correlations across treatments, suggesting complex interrelationships between components as reported by Kremer and Larson (1983) for jack pine. The value of NSU may be underestimated in our stud: because components were not measured in more than on growing season. Averaging across years may be more impor tant for genetic stability than averaging across short-term te treatments. Genetic gain efficiency: applications and limitations Early selection on height increment or a combination o height increment and summer NSU appears to yield nearly a' much genetic gain per generation as selection on age 8 heigh· itself for the North Carolina coastal loblolly pine provenance There is considerable potential for time savings, but applica­ tion of these results will determine the time savings (e.g.. Carter et al. 1990). Two low-risk applications might be tl rogue a production seed orchard or to choose which unrelatec parents to infuse into a breeding population. Forward selec­ tion encumbers more risk of choosing incorrect selections, but the risk-adjusted returns could still be substantial (Newmar, and Williams 1991). Early selection on height· increment must be monitored each generation to ensure that breeders are selecting for the largest increase in height increment with the least change between bud break and final bud set (Bridgwater et at. 1985: Williams 1986). Conclusions Height growth traits and summer NSU were the best choices of juvenile traits for early selection in short-term testing. This was based on four criteria: juvenile-mature family mean correlations, heritability, and genetic stability across extreme treatments, as well as an aggregate measure. genetic gain efficiency. For these traits, juvenile-mature cor­ relations were more highly positive in one fully irrigated and fertilized treatment than when averaged across five short-term treatments. The best 8 out of 10 can be selected correctly on Ll ET AL. TABLE 6. Relative efficiency of genetic gain (%) for loblolly pine composite and compo­ nent traits measured in a farm-field test near Fort Towson. Oklahoma Trait" Gain e fficiency per generation (%) (treatment I-F)b 100 Base HT8 Composite HTl HT2 HT3 HT2 - HT1 HT3 - HT l HT3 - HT2 Component (total for season) NSU MSUL Fixed growth Length NSU MSUL Summer growth Length NSU MSUL 26 81 82 83 51 79 63 -47 -33 -17 -5 64 87 -55 NoTE: The basis for comparison is 8-year height in a Jones County open-pollinated test. a HTl. HT2. HT3, HTS, height at l, 2. 3. and 8 years. respectively; NSU, number of stem units; MSUL. mean stem unit length. b See Table I for a description of this treatment. the basis of 2-year height. Individual-tree heritability was two to three times higher in the short-term test relative to the older tield test, and genetic stability over two extreme short-term test treatments was high. Genetic gain efficiency per genera­ tion is over 80% for these traits relative to 8-year height. Juvenile height growth and NSU in summer growth in a well­ watered and fertilized short-term environment appear to be reliable early selection criteria. Acknowledgements We acknowledge the assistance of Leon Burris of Weyerhaeuser Company, who installed and measured the short-term test in Fort Towson, Oklahoma, and Dr. Antoine Kremer, l'Institut national de Ia recherche agronomique (France), for his valuable suggestions on earlier drafts of the manuscript. Bailey, D.B., and Feret, P.P. 1982. Short note: shoot elongation in Pinus rigida x taeda hybrids. Silvae Genet. 31: 209-212. Bongarten, B. 1986. Relationships between shoot length components in Douglas-fir and blue spruce. Can. J. For. Res. 16: 373-380. Boyer, W.O. 1970. Shoot growth patterns of young loblolly pine. For. Sci. 16: 472-482. 297 Bridgwater, F.E. 1990. Shoot elongation patterns of loblolly pine families selected for contrasting growth potential. For. Sci. 36: 641-656. Bridgwater, F.E .. Williams. C.G.. and Campbell. R.G. 1985. Patterns ofleader elongation in loblolly pine families. For. Sci. 31: 933-944. Campbell, R.K Echols, R.M and Stonecypher, R.W. 1985. Genetic variances and interactions in 9-year-old Douglas-frr grown at narrow spacings. Silvae Genet. 35: 24--29. Carter, K.K., Adams. G.W., Greenwood, M.S., and Nitschke, P. 1990. Early family selection in jack pine. Can. J. For. Res. 20: 285-291. Falconer, D.S. 1981. Introduction to quantitative genetics. 2nd ed. Longman Inc., New York. Franklin, E.C. 1983. Patterns of genetic and environmental variance in short-term progeny tests of loblolly pine. Proc. 17th South. For. Tree Improv. Conf. pp. 332-343. Harrington, C.A. 1991. Retrospective shoot growth analysis for three seed sources of loblolly pine. Can. J. For. Res. 21: 306-317. Kaya, Z., Campbell, R.K., and Adams, W.T. 1989. Correlated responses of height increment and components of increment in 2-year-old Douglas-fir. Can. J. For. Res. 19: 1124-1130. Kremer, A.. and Larson, P.R. 1983. Genetic control of height growth components in jack pine seedlings. For. Sci. 29: 454-464. Kremer, A., and Xu, L.-A. 1989. Relationship between first-season free growth components and later field height growth in maritime pine (Pinus pinaster). Can. J. For. Res. 19: 690-699. Lambeth, C.C. 1983. Early testing-an overview with emphasis on loblolly pine. Proc. 17th South. For. Tree Improv. Con f. pp. 282-291. Lambeth, C.C., and Duke. S.D. 1981. First-generation genetic improvement of growth rate, straightness and rust resistance of loblolly pine in North Carolina. Weyerhaeuser Company Tech. Rep. 090-1204/81/23. Lanner, R.M. 1968. The pine shoot primary growth system. Ph.D. thesis, University of Minnesota, St. Paul. Li, B McKeand, S.E.. and Allen. H.L. 1991. Genetic variation in nitrogen use efficiency of loblolly pine seedlings. For. Sci. 37(2): 613-626. Magnussen, S., and Yeatman. C.W. 1986. Accelerated testing of jack pine progenies: a case study. In Proceedings IUFRO Conference on Breeding Theory, Progeny Testing and Seed Orchards. 12-17 Oct. 1986, Williamsburg, VA. Edited by R.J. Weir. North Carolina State University, Raleigh. pp. 107-121. Newman, D., and Williams. C.G. 1991. The incorporation of risk in optimal selection age determination. For. Sci. 37(5): 1350-1364. Riitters, K.H., and Perry, D.A. 1987. Early genetic evaluation of open-pollinated Douglas-fir families. For. Sci. 33: 577-582. Snedecor, G.W., and Cochran, W.G. 1967. Statistical methods. 6th ed, Iowa State University Press, Ames. Squillace. A.E., and Gansel. C.R. 1974. Juvenile:mature correlations in slash pine. For. Sci. 20: 225-229. Williams, C.G. 1986. Early genetic testing of loblolly pine. Ph.D. dissertation, North Carolina State University, Raleigh. Williams, C.G. 1987. The influence of shoot ontogeny on juvenile­ mature correlations in loblolly pine. For. Sci. 33: 411-422. Williams, C.G., Bridgwater, F.E., and Lambeth, C.C. 1983. Perfor­ mance of single family versus mixed family plantation blocks of loblolly pine. Proc. 17th South. For. Tree Improv. Conf. pp. 194-202. Yamada, Y. 1962. Genotype by environment interaction and genetic correlation of the same trait under different environments. J. J. Genet. 3: 498-509. .• .• .•