430 coraciiDjE. 



one each, Banff two, Elgin and Inverness one each, Sutherland 

 one or two, Caithness one, Orkney nine or ten, Shetland two 

 or three, St. Kilda one, Ayr one ; and, returning to England, 

 Curnheiiand, Lancashire, Flintshire and Somerset one each. 

 Of English counties not bordering on the sea we have one 

 for Hertfordshire and one for Cambridgeshire ; and of Scot- 

 tish, Lanarkshire one and Dumfriesshire two ; while six or 

 seven have occurred in Ireland. These records give an 

 average of more than one appearance in each year, and when 

 the paucity of observations during the early part of the 

 century is considered, that average will be at once seen to be 

 below the mark. Making also allowance for occurrences 

 that take place without being noticed in any Natural-History 

 journal, it is obvious that ornithologists need not account the 

 appearance of the Roller any extraordinary event, but simply 

 one that is to be expected annually. Though many of the 

 records, on which these results are founded, do not mention 

 the date of observation, this information is given in enough 

 instances to allow of its being said that the majority of 

 appearances is in the autumn-months — chiefly September 

 and October, next to which this species may be looked for in 

 May or early June. In the former case the birds have doubt- 

 less either nested or been bred during the preceding summer 

 in the northern parts of Europe, and in the latter they have 

 as doubtless been on the way to their breeding- quarters in 

 that direction. There are however a few exceptional cases 

 deserving remark. A bird is recorded by Mr. Stevenson as 

 having been shot in Norfolk in February 1824, and several 

 are said to have been observed in the height of summer. 

 The example killed so early in the year had possibly failed 

 to accomplish its normal migration southward, and had con- 

 trived to maintain its existence in our northern lands through- 

 out the winter, while those that have occurred about mid- 

 summer have probably failed in their search for a suitable 

 home, and have thus wandered to our shores.* There is no 



* Newman, in his edition of Montagu's ' Dictionary ', says that the Flintshire 

 bird, already mentioned, was killed Feb. 1st, 1858 ; but his authority (Zool. p. 

 5976) gave no date for the incident, and as the communication is dated Jan. 22d, 

 1858, the assertion must be wrong. 



