night in the nest box for a time. Widmann (1898) be- 
lieved that martins tended to leave the nest boxes, even 
as roosting sites, as soon as possible because of the 
infestation of mites. He noted that the adult males 
deserted the box first, then the adult females, and 
finally the young, after they had been flying about a 
week. Martins, young and old, may return to the 
vicinity of the nest box almost daily for at least short 
periods, particularly mornings or evenings, often well 
into August, even though they make no apparent use 
of the box. In this period most martins actually spend 
the nights at some communal roost. 
The communal roosts are most often located in 
dense groves of trees. Dense willow groves 10-20 
feet high along water courses are particularly favored 
(Widmann 1898), but large roosting or preroosting 
ageregations have been seen on city buildings, power 
lines, and sand bars in rivers (Widmann 1922, ‘Tanner 
1933, Musselman 1932, White 1933, Wright 1939, and 
Lobik 1954). A large roost on the Des Plaines River 
was located in about 30 acres of hawthorn thicket 
(Craigmile 1939). Typically, the birds mass each eve- 
ning at the same “staging areas” (power lines, sand 
bars, etc.), then move, as it is getting dark, to the 
roosting trees. For excellent accounts see especially 
Widmann (1922), Anderson (1965), and Lobik 
(1954). Other species than martins, notably other 
swallows, starlings, robins, and blackbirds, may share 
a roost (Craigmile 1939, Bailey 1932, and ‘Taverner 
1906) . 
Loucks (unpublished notes) found a few martins 
roosting with tree swallows in the Peoria area on June 
4, 1893, but notable aggregations at roosts are not 
usually detected before July, with peak numbers oc- 
curring generally in August (Lobik 1954, and Ander- 
son 1965). The roosts vary in size from a few martins 
to several hundred to tens of thousands (Bailey 1932, 
Taverner 1906, Smith 1942, Widmann 1898, Anderson 
1965, and Hill 1922). Near Arsenal Island (St. Clair 
County) Widmann (1884) observed a roosting aggre- 
gation of 10,000 or more martins on August 24, and 
though the numbers remained about the same for 
another 2 weeks, after September 1, an increasing 
percentage of the birds at the roost were young. Large 
numbers were still present on September 8, but only 
about a hundred were left on September 12, and the 
last few were seen September 18. 
Anderson’s (1965) observations at a large roost on 
the Missouri River, near St. Charles, Missouri, show 
the seasonal build-up and decline of a roosting pop- 
ulation. On July 10, at his first visit to the roost, he 
estimated the martin population to be about 30,000 
birds. By August 3 there were 90,000 and on August 
13 there were 130,000. After this peak the numbers 
declined coincidentally with cold-front passage. By 
August 22 only 900 were left, and all were gone by 
September 1. The roost in this same area in 1968 
peaked at 30,000—40,000 (Hamilton 1968). At a roost 
in northeastern Illinois, the population peaked (more 
34 
than 10,000 birds) at the end of August, and decline 
after the first week in September. By September | 
all but a few were gone (Lobik 1954). : 
The precise roosting locales may change from yez 
to year, but there appears to be something of a trad 
tion in the general location of large roosts. ‘T'wo r 
gions with long histories of large roosts are the Chicag 
area, especially the lake front from North Evanston | 
Lincoln Park (Evans 1918, White 1933, Wright 193 
and Lobik 1954), and the St. Louis, Missouri area, 
sites on both sides of the Mississippi River (Widmai 
1922, and Anderson 1963). Widmann’s (1922) co 
jecture that the large roosts near St. Louis represent: 
the martin populations of half the states of Missou 
and Illinois needs corroboration with banding da 
The locations of only a few Illinois roosts have be: 
published. 
Fall Migration 
Both Widmann (1922) and Anderson (1965) ha 
observed what were apparently late evening (incli 
ing after dark) migration departures of martins fri 
roost areas, following the passage of cold fronts in / 
gust and September. ‘The indication is that fall 1 
eration is nocturnal at least in part. Barnes (192 
and Strode (1892) referred to (diurnal?) southwe 
flights in late July and late September in western I 
nois, and in the north Gault (unpublished no 
1894) saw a small southward flight at 8:00 A.M. 
August 17. 
Most observers report that the last martins of 
season are seen around nest boxes during August. J 
evacuation of roosts occurs in late August and Sept 
ber, as discussed earlier. In extreme southern Illin 
Nelson (1877) saw few purple martins in early A 
eust, but “immense numbers” in late August. Our 
counts for southern and central Illinois were relati’ 
high throughout August, but in the north peak n 
bers were seen in late August and early Septem 
(Fig. 29). This is contrary to expectations, as 
would expect to see migrant populations later in 
south than in the north. However, with the avail 
data we cannot differentiate annual variation in 
migration from regional variation, if any, and the 
atively late migration we observed in the north : 
have been a matter of annual variation. Such yea. 
year variation was noted by Mitchell (1930), 
found that in most years some martins at the colo 
in Hinsdale, Illinois, lingered until August 20 
whereas in some years they left in late July or € 
August. The causes of this variation have not | 
determined. In some years a few martins have | 
found in central and northern Illinois in Oct 
(Oberholser 1918, Jones 1933, and Robert Ru: 
fide Graber 1962), but we have no October rec 
for southern Illinois, a gap which probably reflect: 
paucity of observers in the south. | 
The ratios between our spring and fall coun) 
purple martins show no consistent pattern. In 
