In preparing for a fifth attempt to bring the Colorado under control, they determined to modify the plan of operations previously followed by substituting rock for the materials that had before been used in the construction of dams. Practical experience had shown that piling, brush, sandbags and earth could not be made to support the pressure of the river in full flood, while a series of rock-fill barrier dams, of sufficient width and height, might be strong enough to stand even a flood discharge of 115,000 cubic feet of water per second. In making this change of plan, Mr. Randolph acted on his own judgment and in direct opposition to the views and advice of experts who were acquainted with the situation. Almost all of the engineers who had visited the break, including many of national and international reputation, regarded a rock-fill barrier dam as wholly unworthy of consideration, for at least two reasons. First, the rock would probably sink into the soft silt bottom, and keep on going down indefinitely. It might perhaps be supported by a strong brush-mattress foundation, but even then, the mattress would be likely to break under the weight of the load and thus fail to answer its purpose. Second, the water going over a rock-fill dam, while it was in course of construction, would almost certainly wash away some one rock at the top. This, by increasing the overflow at that point, would dislodge more rocks, and finally create a breach that could not be closed. President Randolph who had used brush-mattresses and rock-fill dams on the Tombigbee River in Alabama many years before, fully considered these objections but did not find them convincing and steadfastly adhered to his own plan.