Tropical Agr i cu l t u r a l Research and Extension 1 ( 1): 19-22, 1998 

Path analysis of fruit yield components ofcucurbita moschata Duch. 

C. Gwanama1, M.S. Mwala and K. Nichterlein2 

Department of Crop Science, University of Zambia, P.O. Box 32379, Lusaka, Zambia. 

Accepted 10 February 1998 

ABSTRACT 
Traits contributing to Cucurbita moschata fruit yield were investigated to identify selection aids. Twenty-
nine land races were grown at two sites in Zambia during the 1994/95 and 1995/96 wet seasons. The 
contribution of traits to fruit yield were partitioned by path analysis. Mid-season traits (internode with first 
female flowers, length of internode with first female flowers, length of primary axis, number of primary 
branches and number of leaves per plant) exhibited insignificant phenotypic (P) and genotypic (G) direct 
effects (P, G <0.17) on fruit yield. The length of the primary axis, number of primary branches and number 
of leaves per plant had genotypic indirect effects of intermediate magnitude on the fruit yield through the 
number of fruits per plant (-0.27,0.33,0.42; respectively). Late season traits (weight of first mature fruit 
and number of fruits per plant) had significant genotypic direct effects on fruit yield (G=0.67 for both 
characters). Therefore, in selecting for pumpkin yield, special attention should be given to number of fruits 
per plant and weight of first mature fruit of each plant. 

Key words: Cucurbita moschata, correlation, direct and indirect effects, fruit yield, pumpkin. 

INTRODUCTION 

The most important use of tropical pumpkin, 
Cucurbita moschata Duch. in Zambia is vegetable 
fruit consumption (Gwanama and Nichterlein 1995). 
This species is known locally as pumpkin and 
elsewhere as 'calabaza1, 'calabash', 'ayote', 'zapallo' 
or 'winter squash'. The yield of pumpkins is 
influenced by many traits individually and jointly 
(Whitaker and Davis 1962). Although there may be 
many traits with high correlation to the ultimate 
product, not all of them may be significantly and 
directly contributing to it. Path analysis separates the 
direct and indirect effects of yield contributing traits 
(Solanki and Shah 1989). Identification of traits with 
high direct effects on fruit yield would be useful in a 
breeding program. 

One major problem in C. moschata breeding is 
large size of the plants. Single vines can be as long as 
15 meters with profuse branching. Recommended 
production planting densities, of around 1,700 plants 
per hectare (Van Zijl et al. 1978) are not suitable for 
selection plots because plants form a web with 
neighbours and observation of individual characters 
becomes difficult. Therefore, substantial 
experimental space is required. This is seldom 
obtainable and when available, introduces high 
environmental variation and increases the cost of 

Preseant address: 
1. Department of Plant Breeding, University of Free State, 
P.O.Box 339, Bloemfontein 9300, South Africa. Fax: 27 SI 
4303692; Email: pteelt@landbou.uovs.ac.za 

2.Plant Breeding and Genetics Section, International Atomic 
Energy Agency, P.O.Box 100, A-1400 Vienna, Austria. 

trials. 
Therefore reduced number of replications and 

use of single plant experimental units have been 
practiced in C. moschata breeding (Wessel-Beaver 
and Flores 1996). Implementation of these solutions 
results in failure to completely separate 
environmental variance from treatment variance. 
Therefore, identification of highly correlated early 
season selectable traits would be helpful so that 
inferior plants can be discarded early. 

The objective of this study, therefore, was to 
identify the characters, which are important in 
selecting for higher fruit yield in tropical pumpkin. 

MATERIALS AND METHODS 

Twenty-nine land races of C. moschata were grown 
for two seasons at the University of Zambia (UNZA) 
and Zambia Seed Company (ZAMSEED) farms. 
Both farms were within 15 kilometers of Lusaka 
City. Nineteen of the varieties were S, genotypes 
obtained by self pollinating field collected 
accessions in the 1993/94 season. The remainder 
was half sib families obtained directly by collecting 
one pumpkin fruit per land race in 1994. 

A randomised complete block design with three 
replications was used at each site. Each plot 
consisted of four plants sown on 15 cm high beds. 
The spacing between beds was 2 m and between 
plants in a bed also 2 m. The four plants per plot 
sample size was chosen as the best compromise 
between single plant, single harvest and large plot, 
multiple harvest trials (Wehner and Miller 1984). 
Large plot testing was reserved for advanced trials 
after the number of genotypes has been trimmed. 

mailto:pteelt@landbou.uovs.ac.za


20 GWANAMA ETAL.: PATH ANALYSIS IN CUCURBITA MOSCHATA. 

One meter wide borders were left between blocks 
while no borders were left between plots in a block. 
Sowing dates were 17 (UNZA) and 23 (ZAMSEED) 
December 1994 and 21 and 27 December 1995. 
respectively. An equivalent of 400 kg ha" 
'Compound D' fertiliser (N:P,0,:KaO = 10:20:1) was 
applied. Standard cultural practices were observed. 
Natural rain was the source of moisture throughout 
the experiment. The total amount of rainfall received 
at UNZA was 570 mm in 1994/95 and 790 mm in 
1995/96. The ZAMSEED trial received 650 mm and 
815 mm, respectively. An abundance of honeybees 
for adequate pollination was assured from insect 
pollinated crops grown at the stations for seed 
production. 

Two sets of characters were observed. The first 
group were mid-season traits observed at 50% 
flowering (48-65 days after planting), viz. internode 
with first female flowers, its length (cm), the length 
of the primary axis (m), the number of primary 
branches and the number of leaves per plant. The 
second group were late season characters taken at 
harvest, namely, the weight of the first mature fruit 
(kg), the number of fruits per plant and the total fruit 
yield per plant (kg), while the mean fruit weight (kg) 
was recorded as a derived variable. Harvesting of all 
mature fruits was done once for all plots when vines 
began to wither (at 120 days after planting in first 
season and 135 days in second season). Maturity of 
fruits was ascertained by pricking with a forefinger 
nail. All immature fruits were discarded. 

Data analysis 

The procedure suggested by Singh and Chaudhary 
(1985) was employed in the calculation of 
phenotypic (rp) and genotypic (rG) correlations as 
well as path coefficients at the phenotypic (P) and 
genotypic (G) levels. This procedure uses plot 
means in the calculation of phenotypic correlations 
and genotype means pooled for seasons and 
locations for genotypic correlations. Phenotypic and 
genotypic direct and indirect effects were obtained 
from operations on the respective correlation 
matrices. Effects of yield components were 
considered to be: 

(1) significant if rP or r c > 0.6 and significant at p< 
0.05 andP/r PorG/r c>0.9 

(2) intermediate if rP or rG was significant at p < 0.05 
and 0.6<P/r P orG/r o <0.9 

(3) low in all other cases 
RESULTS 

The first year experienced a drought at the time of 
pollination which reduced the number of fruits set 
and the mean fruit weight. The second year had 

sufficient rain and comparatively better yield. Data 
for both years was included because drought is a 
fairly common occurrence in the country. There was 
considerable variation in the yield performance of 
genotypes. Some genotypes yielded very small fruits 
(< 1 kg) while others had large fruits (5 kg). The 
mean fruit yield on hectare basis varied from 2.5 Mg 
to over 23 Mg (Table 1). 

Correlations between the yield components 
were In most cases low and insignificant both at 
phenotypic (-0.11 <r p < 0.22) and genotypic (-0.21. 
< r 0 < 0.37) levels. No significant negative 
correlations were observed either between the yield 
components or between the yield components and 
the fruit yield. Phenotypic correlations of mid-
season traits to the fruit yield were insignificant, 
with the exception of number primary branches 
which was highly significant. However, at the 
genotypic level, both the length of the primary axis 
and the number of primary branches had significant 
correlations to fruit yield. On the other hand late 
season traits had high and highly significant 
phenotypic and genotypic correlations (0.54 < rp 

<0.71; 0.60 < rG < 0.70) (Table 2). 
Most of the direct and indirect effects of traits on 

fruit yield were very low. Mean fruit weight and 
number of fruits per plant had intermediate and 
significant direct phenotypic effects (P=0.50,0.68), 
respectively. However, the intermediate direct 
phenotypic effect of mean fruit weight was not 
associated with a corresponding genotypic effect (G 
= -0.02). At the genotypic level there were 
significant direct effects of weight of first mature 
fruit and number of fruits per plant (G=0.67 for both 
characters) on fruit yield. Additionally, the mean 
fruit weight per plant exerted a significant genotypic 
effect through the weight of the first mature fruit (G 
= 0.65). The length of the primary axis, the number 
of primary branches and the number of leaves per 
plant had intermediate indirect genotypic effects (G 
= 0.27,0.33,0.42 respectively) via number of fruits 
per plant (Table 2). 

DISCUSSION 

The range of fruit yield averaged for two seasons in 
this study (2.5 - 23.3 Mg ha'1) is low compared to the 
findings of other investigators. For instance, vine 
and bush hybrids and inbreds developed in Florida, 
USA, have been reported to yield between 20-60 Mg 
ha"1 (Maynard 1996). The mean fruit weight in our 
trial was 1.8 kg and mean number of fruits per plant 
was 2.4, while in the latter case corresponding values 
were 3.1 kg and 3.8 fruits per plant. However, direct 
comparisons may be misleading because only 
mature (marketable) fruits were considered in our 
study and land race varieties were employed as the 
study material. The effect of the drought in the first 
season also contributed to the low yield. 
Nonetheless, our results are indicative of the 



Tropical {Agricultural Research ami Extension 1 ( 1): 19-22,1998. 21 

Trait 1994/95 1995/96 Pooled 

Range' Mean" Range" Mean" Range" . Mean' CV(%) 

1 14.6 -22.6 19.0 16.32 
2 13.4 - 28.2 22.3 20.46 
3 2.5 - 4.5 3.4 2.7 - 5.2 4.2 2.8 - 4.8 3.8 28.45 
4 4.9 -16.5 11.7 5.1- 18.6 13.3 5.8- 18.2 12.5 26.11 
5 90.7-205.3 162.3. 102.9-251.7 186.3 112.7-225.8 174.3 39.01 
6 0.6 - 4.5 1.5 . 0.6 - 6.2 2.5 0.6 - 5.1 2.0 39.20 
7 0.5 - 2.0 1.4 0.7 - 4.8 2.2 0.5 - 2.4 1.8 42.95 
8 0.3 - 5.0 1.8 1.2 - 5.5 3.0 0.7 - 5.0 2.4 62.14 
9a 0.6- 3.3 2.4 2.4- 23.1 6.7 1.0 - 9.3 4.3 59.65 
9b 1.5- 8.2 6.0 6.0- 57.8 16.7 2.5 -23.3 10.6 

#: Range of genotype means across locations 

1 = Internode with first female flower, 2 - length of internode with first female flower (cm), 3 = length of 
primary vine (m), 4 = number of primary branches, 5 = number of leaves per plant, 6 = weight of first 
mature fruit (kg), 7 = mean fruit weight per plan (kg)t, 8 = number of fruits per plant, 9a = fruit yield per 
plant (kg), 9b - fruit yield per hectare (Mg). 

situation in unimproved land races in a low yield, 
semi-arid environment.The significant phenotypic 
correlations of mean fruit weight per plant and of 

number of fruits per plant with yield agree with the 
findings of Gopalakrishnan et al. (1980) and 
Chigwe (1991). However, Chigwe (1991) 

Table 2 Effects of fruit yield components on C. moschata fruit yield and correlation of yield traits. (Phenotypic (P) and 
genotypic (G) direct effects on diagonal, indirect effects above diagonal, phenotypic (rP) and genotypic (r c) 
correlation coefficients below diagonal.) 

Trait Trait Trait 

1 2 3 4 5 6 7 8 
1 0.02 0.00 0.00 0.02 0.01 0.00 -0.06 0.07 IP 

0.03 0.00 0.00 0.04 • -0.04 -0.14 0.00 0.01 G 
2r r 

0.22 0.00 0.02 0.01 0.01 -0.01 -0.05 0.14 2P 
rn 

0.01 -0.02 0.02 0.02 -0.03 -0.09 0.00 -0.11 G 
3r r 0.05 0.34** 0J1S 0.02 0.01 0.01 0.03 0.19 3P 

rn 
-0.20 0.44* 0.04 0.04 -0.05 0.12 0.00 0.27 G 

4r, 0.30** 0.22 0.37** 0.06 -0.01 0.01 0.05 0.29 4P 
U 0.36 0.16 0.40* f U i -0.10 0.11 0.00 0.33 G 

5rP 0.08 - 0.06 -0.01 0.13 - 0.09 0.01 0.03 0.07 5P 
to 0.23 0.17 0.33 0.65** -0.16 -0.07 0.00 0.42 G 

6r, -0.07 - 0.08 0.09 0.15 0.08 0.06 0.48 0.07 6G 
ro -0.21 - 0.13 0.18 0.17 -0.11 0.67 -0.02 0.01 G 

7r, •0.11 -0.10 0.05 0.09 0.06 0.97** 0.50 -0.01 7P 
'o 0.18 - 0.16 0.17 0.12 -0.16 0.98** -0.02 -0.08 G 

it. 0.11 0.21 0.28* 0.42** 0.10 0.10 -0.02 0.68 8P 
r0 

0.01 0.23 0.40* 0.49** 0.63** 0.02 -0.11 0.67 G 
9r, 0.05 0.13 0.29 0.41** 0.02 0.61** 0.54** 0.71** 

to -0.12 0.06 0.41* 0.47* 0.29 0.70** 0.60** 0.66** 

Residual effect: Phenotypic=0.413, genotypic 0.294 

*=signiftcant at p-<0.05, **.=significant at p-<0.05,1 = Internode with first female flower, 2= length of 
internode with first female flower, 3= length of primary vine, 4= number of primary branches, 5= 
number of leaves per plant, 6= weight of first mature fruit, 7= mean fruit weight per plant, 8= number 
of fruits per plant, 9= fruit yield per plant. 

Table 1. Performance of C. moschata genotypes grown at the University of Zambia and ZAMSEED during the 
1994/95 and 1995/96 seasons. 



22 GWANAMA ETAL.: PATH ANALYSIS IN CUCURBITA MOSCHATA. 

suggested also that the mean fruit weight could be 
used directly in selecting for higher fruit yield. The 
results of this study show that, at the genotypic level, 
mean fruit weight only affected fruit yield through 
weight of first mature fruit. Hence restricted 
selection using primary vine, simultaneously with 
number of fruits per plant, would have to be applied. 
One cause for deviations from Chigwe's findings is 
due to the fact that only data of the 25 top yielding 
genotypes among his 121 test land races were 
reported. The mean yield of these selected land races 
was 25.5 Mg ha"1. Our best yielding genotype (57.8 
Mg ha"') during the second season out-yielded his 
best land race (41.9 Mg ha"1). His results, therefore, 
were biased for elite land races and favourable 
environments. Traits observed in the middle of the 
growth season failed to show significant direct 
effects on fruit yield. Reducing selection 
populations on the basis of the length of the primary 
axis, the number of primary branches and the 
number of leaves per plant would be helpful since 
they had intermediate indirect and direct effects 
through the number of fruits per plant. 

CONCLUSION 

In the improvement of C. moschata land races 
maximum importance should be accorded to the 
number of fruits per plant and weight of first mature 
fruit as they have high genotypic and phenotypic 
direct effects on fruit yield. The length of the 
primary axis, the number of primary branches and 
the number of leaves per plant selected 
simultaneously with number of fruits per plant 
would also be helpful as early season selection aids. 

ACKNOWLEDGEMENTS 

We are very grateful to the German Government 
and its Technical Cooperation Agency (GTZ) for 
funding this project and ZAMSEED for 
experimental space at theirfarm. 
REFERENCES 

Chigwe CFB 1991 Evaluation of Malawi pumpkin 
(Cucurbita moschata Duch ex Poir) germplasm 
collection. Progress Report, Project No. 
SRG/39/5. Bunda College of Agriculture, 
Lilongwe, Malawi. Pp 1-32. 

Gopalakrishnan TR, Gopalakrishnan PK and Peter 
KV 1980 Variability, heritability and correlation 
among some polygenic characters of pumpkin. 
Indian J.of Agric.l Sci. 50:925-30. 

Gwanama C and Nichterlein K 1995. Importance of 
cucurbits to small scale farmers in Zambia. 
Zambian J. of Agric. Sci. 5:5-9 

Maynard DN 1996 Performance of tropical 
pumpkin inbreds and hybrids. In: Gomez-
Guillamon ML, Soria C, Cuartero J, Tores JA 
and Fernandez-Munoz R (eds). Cucurbits 
Towards 2000. Proceedings of the Vlth 
EUCARPIA meeting on Cucurbit Genetics and 
Breeding, May 28-30,1996, Malaga, Spain, pp 
59-65. 

Singh RK and Chaudhary BD 1985 Biomedical 
methods in quantitative genetic analysis. 
Kalyani Publishers, New Delhi, pp 68-78. 

Solanki SS and Shah A 1989 Path analysis of fruit 
yield components in cucumber. Progressive 
Horticulture. 21:322-324. 

Van Zijl J, Oliver O and Stander H 1978 
Establishment of cucurbits. Farming in South 
Africa Vegetable Series, Pretoria, South Africa. 
pp 1-10. 

Wehner TC and Miller HC 1984. Efficiency of 
single-harvest methods for measurement of 
yield in fresh-market cucumbers. J. Amer. Soc. 
ofHort. Sci. 109:659-664. 

Wessel-Beaver L and Flores DH 1996 Heritability 
of greenhouse and field methods for evaluating 
powdery mildew resistance in tropical 
pumpkin. In: Gomez-Guillamon ML, Soria C, 
Cuartero J. Tores J A and Fernandez-Munoz R 
(eds.). Cucurbits Towards 2000. Proceedings of 
the Vlth EUCARPIA Meeting on Cucurbit 
Genetics and Breeding, May 28-30, Malaga, 
Spain. Pp 241-247. 

Whitaker TW and Davis GN 1962 Cucurbits -
Botany, Genetics and Utilisation. Interscience 
Publishers, New York, pp 69-162.