Modern far-infrared and submillimeter instruments require large-format arrays. We consider the relative performance of filled-array (bare pixel) and feedhorn-coupled architectures for bolometer focal planes. Based on typical array parameters, we quantify the relative observing speeds and comment on the merits of the different architectures. Filled arrays can provide higher mapping speed (by a factor of as much as 3.5) and simpler observing modes at the expense of reduced sensitivity for pointed observations, increased detector numbers, and greater vulnerability to stray light and electromagnetic interference. Taking advantage of the filled-array architecture requires strongly background-limited detectors. At millimeter wavelengths, filled arrays must be surrounded by a sufficiently cold enclosure to minimize the background power from the instrument itself.