36 
PACIFIC SCIENCE, Vol. IX, January, 1955 
not responsible for the variation in egg count 
between sampling days. It was apparent, 
without statistical analysis, that there was no 
significant relationship. 
Attention was next directed to the possi- 
bility of lunar effects. On the basis of one 
series of samples taken in Ala Wai Canal dur- 
ing July, 1949, Tester and Yamashita (1950: 
1) suggested that spawning activity might be 
related to the lunar cycle, with maximum 
spawning during the first quarter and mini- 
mum spawning during new moon. To investi- 
gate a possible lunar relationship, geometric 
means of the adjusted egg counts were cal- 
culated according to lunar days for the years 
1950- 51 and 1951-52. The data were then 
grouped according to phases of the moon. 
The geometric means for dark, first quarter, 
full, and third quarter were respectively 13.63, 
18.20, 13.55, and 12.04 in 1950-51, and 4.53, 
3.54, 2.61, and 2.66 in 1951-52. The differ- 
ences in mean count between phases were 
neither statistically significant nor were they 
consistent during the two years. The present 
data yield no evidence of lunar periodicity in 
spawning. The cause or causes of the large 
variation between successive sampling days 
remains unknown. 
Despite the large variation between suc- 
cessive sampling days it is evident from the 
data that spawning is seasonal in nature with 
a peak production of eggs during the summer 
months and a low production during the 
winter months. This is best illustrated by the 
geometric means per 100 cubic meters accord- 
ing to successive months as given in Table 2 
and portrayed on a logarithmic ordinate scale 
in Figure 3a. Average production of eggs was 
highest during July, 1951, and next highest 
during August, 1952. It was relatively low 
during the winter months of 1950-51, and 
still lower during the winter months of 
1951- 52. 
The seasonal march of egg production fol- 
lows closely the seasonal march of tempera- 
ture (Table 2), but this does not necessarily 
imply a dependent relationship. There is a 
Fig. 3. Seasonal variation in the abundance (lo- 
garithm of the geometric mean plus one) of (a) eggs 
and (b) larvae. 
marked difference in egg production between 
1950-51 and 1951-52 (September to August, 
inclusive) with a geometric mean for the 
former of 14.12 and for the latter 3.27 eggs 
per 100 cubic meters. For the corresponding 
periods, mean water temperatures were 25.00° 
C. and 24.38° C. and mean chlorinities were 
18.91 p.p.m. and 19-10 p.p.m., suggesting 
that higher temperatures and/or lower chlor- 
inities were favorable for greater egg produc- 
tion. This possibility was investigated further 
using as variates the difference (with due re- 
gard to sign) between the means for the same 
months in the two years for temperature, 
chlorinity, and log egg count. Neither of the 
two partial regression coefficients (0.22 for 
log egg count and temperature; 0.06 for log 
egg count and chlorinity) was significant. The 
multiple correlation coefficient (R = 0.41) 
indicated that 83 per cent (1-R 2 ) of the total 
variation in egg count differences was un- 
explained by variation in the differences for 
temperature and chlorinity. It may be pointed 
out that this method of handling the data 
nullifies the seasonal trends. Although the 
results suggest that temperature may influence 
