176 ' ANNUAL REPORT SMITHSONIAN INSTITUTION, 1920. 



bles on a beach. But this tells us nothing about the sizes of the 

 minute disks. Have I gone too far perhaps in saying there is no 

 disk visible in any star ? Nebulae do show disks, and though they are 

 nebulae and not stars they may become stars. The new secrets 

 wrested from the stars have chiefly come, not from the increase in 

 size of telescopes, but from the new appliances attached to them, such 

 as the photographic plate, the spectroscope, and by this time many 

 others. The lines in the spectra of stars tell us what the stars are 

 made of, how they may be classified accordingly, how fast they are 

 moving, how bright they really are (this is an amazing recent dis- 

 covery), and by inference how far away, and may yet have other 

 surprises in store. For the moment we are chiefly concerned with 

 the classification. The Harvard system gives us a number of classes 

 denoted by the capital letters OBAFGKMRN. The fact that 

 the order is not quite the same as that of the alphabet represents a 

 revision of early ideas, chiefly due to the gradual accumulation of 

 intermediate types, which make a nearly continuous series. 



Now a series of stars in order is probably a representation of 

 growth; just as the growth of trees may be illustrated by selecting 

 various stages from the same wood, an illustration originally given 

 by Sir W. Herschel. But we have seen a tree grow, and we know 

 independently that it grows up from the acorn through the sapling 

 to the giant oak ; while we have not had time to see a star grow and 

 were thus in ignorance whether the changes are from B toward M 

 or from M toward B, though by this time we have an immense 

 number classified. The classification has been largely the work of an 

 American lady, Miss Cannon. I am told that there is a man who 

 can deftly straighten rifle barrels — he gives a glance along the barrel, 

 a tap with a hammer, and lo ! it is straight. His value is recognized 

 at some £15 a week. Miss Cannon has the same deftness with 

 spectra — but I fear that (to judge from the report of the Board of 

 Visitors of Harvard Observatory) her great skill is not so appro- 

 priately rewarded. 



Now, it is obviously important to find out, if we can, which is 

 the direction of a star's growth, and we seemed to have an important 

 clue when the spectral classification was connected with the tempera- 

 ture of a star, or rather its surface temperature, which is all we can 

 get at. The outside is the coolest, just as the edges of a plate of 

 porridge are the coolest, as most of us have learned by early and 

 rather painful experience. And yet the outside of a star is hot 

 enough. The temperature is again estimated from the spectrum, 

 though this time not from the lines but from the relative intensities 

 of its parts, and the O B A end is undoubtedly hotter than the other. 

 We may give as illustrations 15,000° for B, 5,000° for G, and 



