After seventeen days' captivity the other species (_N. Eleanora_) began to cover a small surface of the gauze with a semi-transparent substance (which resembled varnish rather than silk), secreted from its spinners, and four days later it began to weave a cell; this cell took twelve days to complete, and finally assumed the shape of a rudely-formed figure of 8, with a circular aperture at either end, each of which was kept open during the construction of the cell, and then closed. The gauze itself, covered with silk, formed the ceiling of the cell, while the floor was made of silk attached to the earth, and the sides of strong and rather opaque silk. This cell bore no resemblance to any portion of any trap-door nest that I have ever seen, and it is difficult to conceive how the idea of such a structure presented itself to the spider. Its outline indeed had some likeness to the general outline of the spider herself, one loop of the figure 8 being rather smaller than the other. The distance between the floor and the ceiling of this impromptu cell was a little over half an inch, its width varying from one inch in the broadest to eight lines in the narrowest part, while its length was an inch and a quarter. It would appear that the object which the spider had in view was to construct a warm and secure retreat for the winter, and accordingly after having completed this chamber, she no longer made excursions to catch the flies with which I supplied her, but remained self-immured in her cell.[106] [Footnote 106: My observations on the captive spider were still in progress at the time of going to print, so that the above notes must be considered as incomplete.] It would be interesting to discover whether any of the spiders of this group (but which do not construct trap-door nests) pass the winter in similar structures. H. _On the Structure of Cork Doors._ In order to test my theory to the effect that the trap-door nests are enlarged from time to time, and that the numbers of layers of silk in an undisturbed cork door should represent the number of enlargements which the nest has undergone, I examined the doors of twenty-eight nests of the cork type (all I believe of _N. cæmentaria_), in order to prove whether as a rule the larger cork doors do contain more layers of silk than the small ones, as they should on this hypothesis. This is, I think, fairly established by the following table:-- _Comparative Table._