known. Bands recovered in Texas were further designated 
by either North or South Texas. In addition, all recoveries 
of birds banded outside of the CMU during the specified 
years but recovered either in the CMU or in Latin America 
were also used in the analysis. Locations of banded birds 
recovered in Mexico were further identified by Mexican 
State of recovery. 
The banding and recovery data were reformatted to 
facilitate their analysis. Major steps included (1) changing 
locations of banding and recovery from the waterfowl fly- 
way orientation of the original files to mourning dove 
management unit orientation, (2) realigning States within 
the mourning dove management units and nations south 
of the United States from an alphabetical order to a north 
to south orientation, and (3) establishing major and minor 
time periods of band recovery. These procedures generally 
followed those outlined by Reeves and Fiehrer (unpublished 
report). 
The final edited and reformatted records were then sub- 
jected to programs in the EDP system noted above. Output 
was on computer printouts, which provided data directly 
useful for analysis or source data for additional analysis. 
To illustrate the latter, banding and recovery matrices from 
the Laurel tabulations were analyzed by the Texas Parks 
and Wildlife Department in Austin, using various new 
statistical methods developed for estimating survival rates. 
A few other special tabulations not in the above EDP sys- 
tem were produced to show the number of doves banded 
in non-CMU States during the study period, and to show 
the band recovery distribution in Mexico by Mexican State. 
The 50-odd computer-produced tabulations comprising the 
foundation for this study are filed in the Office of Migra- 
tory Bird Management, Laurel, Maryland, for future ref- 
erence. 
Data Limitations 
At least four weaknesses are inherent in mourning dove 
banding data: 
(1) Few mourning doves were banded in Canadian areas 
north of the CMU during the study period. We elected to 
exclude these Canadian records from the analysis, There- 
fore, not all segments of the migrating mourning dove popu- 
lation in the CMU were represented by banded samples. 
In addition, Mexico’s breeding population of mourning 
doves was not sampled. Thus, all discussions in this analy- 
sis of distribution or derivation of doves in Mexico and 
Central America stem from birds banded in the United 
States, 
(2) Species with low recovery rates and relatively low 
survival rates (such as mourning doves) do not readily lend 
themselves to banding studies because of the large sample 
sizes necessary for analysis. Although more advanced than 
earlier methods (Brownie et al. 1978:17; Anderson and 
Burnham 1976:12), the stochastic models used in our 
13 
analysis were limited in their ability to provide mourning 
dove survival and recovery rate estimates with acceptable 
precision because of the stringent requirements of the 
models. The banded samples by age and sex for some States 
were marginal or inadequate for computation of survival 
and recovery rates. This situation necessitated pooling of 
data for the management unit as a whole and tended to 
mask some of the individual variation by State. 
(3) Not all hunters report all banded doves they encoun- 
ter, Furthermore, this reporting rate is not equal in all areas 
where dove bands are recovered (Tomlinson 1968). Al- 
though a reward band study was undertaken to determine 
reporting rates during the CMU banding program, the reli- 
ability of the derived band reporting estimates is question- 
able (see discussion in Reeves 1979). An average reporting 
rate of 45% was arbitrarily used in this analysis unless 
otherwise stated, 
(4) The absence of a nationwide hunter and harvest sur- 
vey restricts the application of data derived from the band- 
ing analysis. In this analysis, we summarized State-obtained 
harvest data when practicable and estimated harvest when 
that figure was not available from the State (e.g., Arkansas). 
As a result, projected preseason population estimates are 
subject to considerable error. 
Representativeness of Doves Banded 
One of the fundamental assumptions underlying all 
analyses of banding data is that the marked animals are 
representative of the population from which they come. 
This assumption may be violated in the present analysis for 
reasons discussed below. 
The preseason banding period of June through August 
was selected to ensure that trapped doves were either nest- 
ing adults or immatures hatched in the sampled area. If 
there was substantial movement into or away from specific 
areas, representativeness of the banded sample would be 
compromised. Movement into and out of areas most logi- 
cally would occur among early-hatched immatures, if it 
occurred at all. 
Although the banding period was 3 months long, it was 
assumed that recovery patterns of birds banded early in the 
banding season were similar to those banded late. The 
validity of this assumption was determined by testing the 
hypothesis that proportions of in-State to out-of-State direct 
recoveries were similar throughout the banding period. 
Data within each hunting State were examined separately 
by using the G-test (Sokal and Rohlf 1969:599-600). Data 
for immature (Table A-8) and adult (Table A-9) doves were 
tested separately. 
No chronological differences in proportions of in-State 
to out-of-State direct recoveries (P > 0.05) were detected 
among 10-day periods for immature doves banded in Colo- 
rado, Kansas, North Texas, and South Texas. Differences 
(P < 0.05) were detected for South Dakota, Missouri, New 
Mexico, Oklahoma, and Arkansas where immatures banded 
late had a higher proportion of in-State recoveries (Table 
