A seismic line was shot such that the source-ghost effect from two different dynamite source patterns could be compared. Two 15-shot seismic datasets were created that were identical in all respects except that one used 4 kg of explosives in a single 18-m hole, while the other used 2 kg of explosives in each of two 9-m holes. After identical processing, the final stacked sections of the 18-m and 9-m datasets are dramatically different in character and temporal resolution. An f-x spectral analysis of the stacked sections reveals that the 18-m data shows a loss in power and phase coherence from 45 to 58 Hz, while the 9-m data shows a similar effect from 65 to 78 Hz. A spectral notch, centered near 55 Hz, due to a source ghost is suggested as the reason for the lower power in the 18-m dataset. The 9-m data is consistent with a spectral notch at a higher central frequency near 72 Hz. Above its spectral notch, 18-m data shows a reemergence of weak signal that persists to near 80 Hz; the 9 m dataset shows little signal above 65 Hz. Examination of raw shot records shows that these effects are very difficult to observe in field records. Without specialized deghosting, the 9-m dataset shows greater temporal resolution; however, the 18-m dataset has a broader signal bandwidth and is potentially superior.