21 



ahead of others, notably 1920. A glance at the temperature 

 chart will show that there exists a strong correlation between 

 the negative deviations and the late flowering. The season of 

 191 7 is marked by a preponderance of negative deviations of 

 comparatively large magnitude. The few positive deviations 

 almost coincide with the periods of flowering in which the season 

 was not completely behind all other seasons. Certainly the cool 

 temperature seems to have been responsible for the late flowering. 



igi8 and igiQ — ^These two years can well be discussed to- 

 gether since on the chart their curves alternate for the second 

 place to the earliest. Up through No. 22, 1919 (o) leads 1918 

 (I) with a very few exceptions. At that point the flowering 

 in April has passed. A study of the temperature deviations 

 shows that the greater number of positives are on the side of 

 191 9, especially during the months preceding flowering. 



And with the coming of large positive deviations in the 19 18 

 column for the month of May, if the temperature control is 

 strong, the J curve should lead the o curv^e through that month. 

 With but four exceptions (nos. 27, 29, 30 and 40) it does lead 

 through No. 72 where the flowering passes over into June, and 

 indeed well on through the season with a few exceptions toward 

 the end. 



This last fact is true in spite of the winning of the temperature 

 balance by 1919 from June ist on. Either this is evidence 

 against the importance of the temperature control, or, what 

 seems more probable in the face of the other evidence, the start 

 given through May to the 191 8 plants was sufficient to keep 

 most species well ahead for the rest of the season, when it had 

 ceased to be a question of getting minimum warmth to grow 

 and flower. 



So the study of these two seasons seems to give evidence in 

 favor of temperature control, particularly b}^ that of the pre- 

 flowering months, possibly by optimum conditions in mid- 

 season that even carry over another month. 



ig20 — ^Thls season (recorded by a f) is of little interest except 

 as it partially confirms the results of 1917. With temperatures 

 generalty below normal from January on, it is not surprising 

 to find that the cur\-e generally falls behind all but 191 7 (the 

 coldest season), though there are glaring exceptions to that con- 

 dition. The general lateness of the dates makes the writer feel 

 that there is a distinct posIti\-e correlation. 



