[Vor. 8 
344 ANNALS OF THE MISSOURI BOTANICAL GARDEN 
in the mosaic disease of tobacco is a filterable virus, the ‘‘con- 
tagium vivum fluidum" of Beijerinck (798). Confirmation of 
Beijerinck's porous filter experiments is not lacking. On the 
other hand, the agency in this disease has been found to be held 
back, or non-filterable, when certain filters are employed. Ex- 
periments establishing the last-mentioned fact have been con- 
tributed by the work of Allard (’16) and also for the cucumber. 
mosaic by Doolittle (20). All too frequently, it would seem, 
our knowledge respecting the particles or individuals of the 
so-called filterable organisms has been chiefly the fact of the 
passage of infective particles through some bacteriological 
filler, more partieularly the Chamberland or the Berkefeld, 
with no particular effort to effect a more precise standardization 
of both the filters permitting the passage of such particles and 
of those filters restraining them, so as to permit a more definite 
measurement of the particles concerned. 
In this work some of the methods of ultrafiltration have been 
employed. In general, the method or technique of the experi- 
mentation may be divided into 3 phases: (1) filtration (or diffus- 
ion) of diseased juice through various ultrafilters, (2) inoculation 
of healthy plants with the filtrates obtained, and (3) the standardi- 
zation of the filters by a determination of their capacity to permit 
or prevent the passage of colloidal particles of known, or ap- 
proximately known, sizes. 
PRELIMINARY EXPERIMENTS 
Preceding a discussion of the later work under 3 headings 
corresponding to the 3 phases, or aspects, above noted, it seems 
well to report certain preliminary data, secured during the 
previous year, which led to the more definite formulation of the 
chief experimental work reported in this paper. The preliminary 
experiments consisted of: (1) a filtration test of infected juice 
through a Livingston spherical atmometer cup, (2) filtration 
through layers of 1.5 and 3.0 per cent agar, (3) diffusion through 
Schleicher and Shiill parchment diffusion shells. 
In these preliminary experiments with the atmometer the 
filter cup was partially filled with the juice, and suction was 
