Testing Method and the Correlation Between Fruit Yield and Yield Components in Cucumber

Cucurbit Genetics Cooperative Report 23:16-20 (article 5) 2000

Christopher S. Cramer
Department of Agronomy and Horticulture, Box 30003, New Mexico State University, Las Cruces, NB 88003-8003

Todd C. Wehner
Department of Horticultural Science, Box 7609, North Carolina State University, Raleigh, NC 27695-7609

Introduction. An alternative to improving fruit yield per se in cucumber (Cucumis sativus L.) would be to select for characters which were highly correlated with yield and that had a higher heritability than yield. Recently, Cramer and Wehner (2, 5) identified certain yield components that were correlated with fruit yield of pickling and slicing cucumber populations. The potential exists for the selection of those components to improve yield in those populations.

The test method used by the plant breeder influences yield and yield components. Cramer and Wehner (2, 3, 5) observed that cucumber populations grown in hills (spaced plants) at low density (6,450 plants/ha) produced more branches, and total, marketable and early fruit per plant, pistillate nodes, and nodes per branch than populations grown in plots at a normal density (6,500 plants/ha). Wehner (8) also observed an increase in the number of fruit per plant as the plant density decreased from 123,500 to 10,300 plants/ha. In addition, gynoecious hybrids produced fewer pistillate nodes as the plant density increased from 84,000 to 256,000 plants/ha (7). Plot measurement of yield and yield components can be labor-intensive, time-consuming, difficult, and inefficient. Testing in single-plant hills would be less labor-intensive, faster, easier, and more efficient.

The objective of this study was to determine the effects of two testing methods on the 1) correlation among yield components, 2) correlation among fruit yield traits, and 3) the correlation between yield components and total fruit yield of cucumbers. We were interested in selecting yield components using the easier method of hills (low density) but only if the strong correlations were maintained.

Methods. The four pickling cucumber populations used were the North Carolina wide base pickle (NCWBP), medium base pickle (NCMBP), elite pickle 1 (NCEP1), and hardwickii 1 (NCH1). The four slicing cucumber populations used were the wide base slicer (NCWBS), medium base slicer (NCMBS), elite slicer 1 (NCES1), and Beit Alpha 1 (NCBA1). The populations differed in their genetic diversity and mean yield performance (9, 10, 11). Populations were developed using modified half-sib recurrent selection to improve fruit yield, earliness, and shape of the population (12, 13). Three cycles of selection were chosen from each population (0, 3, 5) for NCWBP; 0. 5. 10 for NCH1; 0, 5, 9 for NCEP1; 1, 5, 10 for NCES1; 0, 4, 8 for NCBA1) to represent early, intermediate and late cycles of selection (2, 5). Eight families were chosen at random from each population-cycle combination (four in 1995 and four in 1996) and self-pollinated in the greenhouse.

The experiment was a split-split plot treatment arrangement in a randomized complete block design with four replications in each of two seasons (spring, summer) in each of two years (1995, 1996) with two testing methods [plot (64,500) or hill (6,450 plants/ha)] (2, 3, 5). Whole plots were the right cucumber populations, subplot were three cycles of recurrent selection (early, intermediate, late) and sub-subplots were testing method [hill (3) or plot (30 plants per 3.1 m)]. The experimental factors, planting and harvesting dates, plot size, border plots, soil type, and cultural practices were identical to those reported by Cramer and Wehner (2, 4, 5).

Each test plot was evaluated for number of branches, leaves, pistillate flowers, and total, early (oversized), and culled (crooked and nubbined) fruit. Plants which had fewer than five leaves, no flowers, and a stem length less than 40 cm were considered weak and were eliminated from the plot. Plots were corrected to three plants per plot for the hill method, or to 30 plants per plot for the plot method if they had two plants for hills or 16 to 34 plants for plots (2, 3, 4, 5). Plots with fewer than two plants for hills, or 16 plants for plot were considered missing. PathSAS (6) was used to determine correlations among yield components, among yield traits, and between yield components and total fruit yield per plant (2, 3, 5). Correlations of 0.70 and higher (positive or negative) were considered to be strong correlations while correlations between -0.69 and 0.69 were considered weak (2, 3, 5).

Results.The total correlations between yield components and fruit yield, among yield components, and among yield traits discussed for each population, season, and cycle combination for plot (1, 2, 5) and hill (1, 3) testing method have been published and will not be presented in this paper. Of interest here is the proportion of instances in which changes in correlation strength occurred when the testing method was changed. Correlations for certain population-cycle-season combinations will be presented when changes in correlation strength occurred with a change from plot to hill testing method.

For the majority of yield components and populations, correlation strength among yield components remained unchanged when testing method was changed (Table 1). The proportion of instances in which correlation strength did not change with testing method was greater than 0.75 for most yield component-population combinations. When the number of instances was averaged over populations for each crop type, the correlation strength among yield components was more stable over testing methods for the pickle populations than for the slicer populations (Table 1). When averaged over all eight populations, the proportion of instances in which the correlation strength did not change with a change in testing method was high for each yield component. However, several instances existed in which the change in testing method directly influenced correlations among yield components (Table 1). Both the NCMBS and NCES1 populations exhibited a proportion of correlation strength between the number of branches per plant and other yield components when the testing method was changed from plot to hill. In addition, the correlation strength of nodes per branch and other yield components varied with testing method in several instances for the NCWBS and NCMBS populations. For the NCEP1 and NCMBS populations, the correlation strength between percentage fruit set and other yield components varied with testing method in several instances.

Of the populations examined, the NCMBS population exhibited the lowest proportion of instances in which the correlation strength among yield components did not change with a change in testing method. The differences in correlation strength among yield components with a change in testing method may have resulted from similar differences in yield component means between plot and hill testing methods (1, 2, 3, 5). When averaged over all populations, the number of branches per plant, the number of nodes per branch, and the percentage of pistillate nodes were greater when plants were grown in hills than when plants were grown in pots. Those changes in component means would alter the correlations among yield components and would result in changes in correlation strength with a change in testing method.

Correlation strength between yield components and total fruit yield per plant remained unchanged when testing method changed for a majority of the population-yield component combinations (Table 2). The proportion of instances in which the correlation strength did not change with testing method was 0.67 or greater for 75% of the population-yield component combinations. Most of the correlations between yield components and total yield per plant at either testing method were considered weak because they ranged from -0.69 to 0.69 (1, 2, 3, 5). With this wide range of correlation values, correlations could change with testing method without a change in correlation strength. This wide range of correlation values for the weak correlation classification could explain the high number of population-yield component combinations where the correlation strength did not change. When the number of instances was averaged over populations for each crop type, the correlation strength between yield components and total fruit yield per plant was more stable over testing methods for the pickle populations than for the slicer populations (Table 1). When averaged over all eight populations, the proportion of instances in which the correlation strength did not change with a change in testing method was high for each yield component.

However, several population-yield component combinations were observed where the proportion of instances in which the correlation strength did not change with testing method was 0.50 or less (Table 2). More of these population-yield component combinations were observed for slicer populations than for pickle populations. For both the NCES1 and NCBA1 population, the correlation between the number of branches per plant and total yield per plant changed strength with a change in testing method in a number of instances. The NCBA1 population also exhibited two instances in which the correlations of total fruit yield per plant with the number of nodes per branch and percentage of pistillate nodes changed strength from plots to hills (Table 2). For both the NCWBS and NCES1 population, the correlation between percentage of pistillate nodes and total fruit number per plant changed strength with a change in testing method in 50% of the instances. the correlation between the percentage of fruit set and total fruit yield per plant changed strength from plots to hills for the NCMBP and NCWBS populations. The changes in correlation strength between total fruit yield per plant and yield components with a change in testing method may have resulted from similar differences in yield and yield component mean values between plot and hill testing methods (1, 2, 3, 5). When averaged over all populations, the total number of fruit per plant, the number of branches per plant, the number of nodes per branch, and the percentage of pistillate nodes were greater when plants were grown using the hill method than when plants were grown using the plot method. These changes in yield and yield component means would alter the correlations between total yield and yield components and would result in changes in correlation strength with a change in testing method.

With regard to the correlations of total fruit yield per plant with marketable and early fruit yield, the proportion of instances in which correlation strength did not change with a change in testing method was 0.67 or greater for a majority of population-yield trait combinations (Table 3). Several population-yield trait combinations existed in which changes in correlation strength occurred for 50% of the cycle-season combinations (Table 3). For the NCBA1 population, a change in testing method changed the correlation strength between total and marketable fruit number per plant when the population was tested in both seasons (Table 2). For both the NCMBP and NCH1 populations, the correlation between total and early fruit yield per plant changed strength from the plot to the hill testing method for 50% of the instances observed (Table 3).

Table 1. The percentage of instancesz in which the correlation between the yield component of interest and the other three yield components did not change when testing method was changed from plot to hill for eight cucumber populations.

Population Branches/plant Nodes/branch Pistillate nodes (%) Fruit set (%) Average
NCWBP 0.89 0.72 0.72 0.89 0.81
NCMBP 0.83 0.83 0.89 0.89 0.86
NCEP1 0.94 0.78 0.78 0.61 0.78
NCH1 0.72 0.89 0.83 0.78 0.81
All Pickles 0.82 0.81 0.81 0.79 0.81
NCWBS 0.72 0.61 0.78 0.89 0.75
NCMBS 0.56 0.44 0.72 0.50 0.56
NCES1 0.50 0.67 0.83 0.67 0.67
NCBA1 0.78 0.78 0.78 0.78 0.78
All Slicers 0.64 0.63 0.78 0.71 0.69
Average 0.74 0.72 0.79 0.75 0.75

z The number of observations for correlation among yield components is 18 (population-yield component), 72 (population, crop-yield component), 144 (yield component), 288 (crop) and 576 (overall).

Table 2. The percentage of instancesz in which the correlation between yield component and total fruit yield per plant did not change when testing method was changed from plot to hill for eight cucumber populations.

Population Branches/plant Nodes/branch Pistillate nodes (%) Fruit set (%) Average
NCWBP 0.67 0.67 1.00 1.00 0.83
NCMBP 0.83 0.67 0.83 0.50 0.71
NCEP1 0.83 1.00 0.67 0.67 0.79
NCH1 1.00 0.67 0.67 0.83 0.79
All Pickles 0.83 0.75 0.79 0.75 0.78
NCWBS 0.83 0.83 0.50 0.50 0.67
NCWBS 0.67 0.83 0.83 0.67 0.75
NCES1 0.50 0.83 0.50 0.83 0.67
NCBA1 0.17 0.50 0.50 0.83 0.50
All Slicers 0.54 0.75 0.58 0.71 0.65
Average 0.69 0.75 0.69 0.73 0.71

z The number of observations for correlation between yield components and yield is 6 (population-yield component), 24 (population, crop-yield component), 48 (yield component), 96 (crop) and 192 (overall).

Table 3. The percentage of instancesz in wyhich the correlation between yield traits and total fruit yield per plant did not change when testing method was changed from plot to hill for eight cucumber populations.

Fruit yield per plant
Population Marketable Early Average
NCWBP 0.83 0.67 0.75
NCMBP 1.00 0.50 0.75
NCEP1 1.00 0.83 0.92
NCH1 1.00 0.50 0.75
All Pickles 0.96 0.63 0.79
NCWBS 0.83 0.83 0.83
NCMBS 0.67 0.67 0.67
NCES1 0.67 0.67 0.67
NCBA1 0.50 0.67 0.58
All Slicers 0.67 0.71 0.69
Average 0.81 0.67 0.74

z The number of observations for correlation between yield components and yield is 6 (population-yield trait), 12 (population), 24 (crop-yield trait), 48 (yield trait, crop), 96 (overall).

Literature Cited

  1. Cramer, C.S. 1997. Specific combining ability for fruit yield and shape, yield, and yield components of cucumber (Cucumis sativus L.) populations improved using recurrent selection. Ph.D., Dissertation, North Carolina State Univ., Raleigh, North Carolina.
  2. Cramer, C,S, and T.C.Wehner. 1998a. Fruit yield and yield component means and correlations of four slicing cucumber populations improved through six to ten cycles of recurrent selection. J. Amer. Soc. Hort. Sci. 123:388-395.
  3. Cramer, C.S. and T.C. Wehner. 1998b. Fruit yield and yield components of cucumber populations grown at low planting density. In: J. D. McCreight (Ed.), Cucurbitaceae ’98, evaluation and enhancement of cucurbit germplasm, pp. 277-285. ASHS Press, Alexandria, VA.
  4. Cramer, C.S. and T.C. Wehner. 1998c. Stand correction methods for cucumber fruit yield. cucurbit Genet. Coop. 21: 18-20.
  5. Cramer, C.S. and T.C. Wehner. 1999. Fruit yield and yield component correlations of four pickling cucumber populations improved through recurrent selection. Crop Sci. (in review).
  6. Cramer, C.S., T.C. Wehner, and S.B. Donaghy. 1999. PATHSAS: A SAS computer program for path coefficient analysis of quantitative data. J. Hered. 90: 2560-262.
  7. Lower, R.L., O.S. Smith, and A. Ghaderi. 1983. Effects of plant density, arrangement, and genotype on stability of sex expression in cucumbers. HortScience 18: 7370738,
  8. Wehner, T.C. 1986. Efficiency of 3 single-harvest tests for evaluation of yield in pickling cucumber. Euphytica 35: 493-501.
  9. Wehner, T.C. 1997. three pickling cucumber populations: NCWBP, NCMBP and NCEP1. HortScience 32: 941-944.
  10. Wehner, T.C. 1998a. Three slicing cucumber populations: NCWBS, NCMBS, and NCES1. HortScience 33: 168-170.
  11. Wehner, T.C. 1998b. Two special cucumber populations: NCH1 and NCBA1. HortScience 33: 766-768.
  12. Wehner, T.C. and C.S. Cramer. 1996a. Gain for pickling cucumber yield and fruit shape using recurrent selection. Crop Sci. 36: 1538-1544.
  13. Wehner, C.S. and C.S. Cramer. 1996b. Ten cycles of recurrent selection for fruit yield, earliness, and quality in three slicing cucumber populations. J. Amer. Soc. Hort. Sci. 134: 322-326.