THE RELATION BETWEEN J-IGHT AND riGMENT-FOKMATiON. 547 



were immersed in bell-jars filled respectively with Larainaria 

 saccharin a, Nitophy Hum, and Ulva. Similar batches 

 of wrasses were also placed in dark-bottomed and porcelain 

 vessels, and in complete darkness. A double circulation was 

 maintained, and the weeds were renewed twice a week. The 

 young' fish were fed with tow-nettings and with amphipods. 

 The bell-jars stood on the south side of the laboratory, and 

 i-eceived diffuse daylight on all sides. 



Table I, pp. 574-575, gives the result of this experiment, 

 which lasted for about thi*ee weeks. 



The lio-ht reflected from the weed backsfrounds is a most 

 important factor in the case, and in previous experiments has 

 not received sufficient attention. Different as the three weeds 

 are to the naked eye, their spectroscopic examination reveals 

 little diversity; indeed, the important differences in the light 

 i-eflected from their surfaces (or transmitted through them) is 

 the preponderance of one or more of the parts of the spectrum 

 they transmit in common. Thus the green Ulva (in more 

 than one layer) transmits from red to green, the green being 

 somewhat more vivid than the red, but with no great 

 difference of intensity. The red Nitophyllum also reflects 

 red to green, but whereas the red is bright the green is exceed- 

 ingly dim. Laminaria also transmits from red to green, 

 but here the whole spectrum is very faint. 



The results show that brown weed backgrounds produce 

 the same effect on the coloration of young Crenilabrus 

 Mielops as does a black background. The fish may undergo 

 temporary flushing and pallor under the conditions of 

 examination, and there is a tendency for the dark bands to lose 

 their distinctness, but the result (PI. 23, fig. 1) is decisive. The 

 amount of red pigment is greater than in similar specimens 

 exposed to light reflected from red or green weed. The 

 reflected light is more dim and is diffused over the whole 

 spectrum in the case of black backgrounds than in that of the 

 brown weed, and it is probably this difference which explains 

 a tendency to greenness in some of the records. 



The results with green and i-ed weeds en masse are some- 



