We present observations of the early stages of formation of a filament channel when a compact activity complex emerged in a previously quiet, near-equatorial area. In a few hours, and while flux was rising rapidly in one bipolar component in the complex, Hα fine structure overlying a polarity inversion zone organized into a conspicuous pattern of parallel fibrils enclosing the trailing end of the new activity complex. Yet it took another 4 days for a stable filament to form inside that pattern. It did so at a place where migrating positive polarity flux from the new activity complex contacted the negative polarity flux in a plage of an adjacent decaying bipolar active region. In contrast, no filament formed along an existing channel inside the adjacent decaying region; the opposite-polarity fluxes on the borders of the existing channel showed no signs of convergence. We attribute the fibril-aligning forces in the new channel to the horizontal component of an extended nonpotential magnetic field caused by currents in the multipolar activity complex. The channel is, in this view, an elementary part of the magnetic topology of an activity complex. We propose that the later formation of the filament in the new channel requires an additional and separate process. A plausible candidate for this second step is the development of a current sheet at the site of converging magnetic flux.