We describe the use of partially overlapping galaxies to provide direct measurements of the effective absorption in galaxy disks , independent of assumptions about internal disk structure .
The non-overlapping parts of the galaxies and symmetry considerations are used to reconstruct , via differential photometry , how much background galaxy light is lost in passing through the foreground disks .
Extensive catalog searches yield \sim 15 - 25 nearby galaxy pairs suitable for varying degrees of our analysis ; ten of the best such examples are presented here .
From these pairs , we find that interarm extinction is modest , declining from A _ { B } \sim 1 magnitude at 0.3 R _ { 25 } ^ { B } to essentially zero by R _ { 25 } ^ { B } ; the interarm dust has a scale length consistent with that of the disk starlight .
In contrast , dust in spiral arms and resonance rings may be optically thick ( A _ { B } > 2 ) at virtually any radius .
Some disks have flatter extinction curves than the Galaxy , with A _ { B } / A _ { I } \approx 1.6 ; this is probably the signature of clumpy dust distributions .
Even though typical spirals are not optically thick throughout their disks , where they are optically thick is correlated with where they are most luminous : in spiral arms and inner disks .
This correlation between absorption and emission regions may account for their apparent surface brightness being only mildly dependent on inclination , erroneously indicating that spirals are generally optically thick .
Taken as an ensemble , the opacities of spiral galaxies may be just great enough to significantly affect QSO counts , though not enough to cause their high redshift cutoff .