We use laboratory experiments to compare allocation rules in uniform price divisible good auctions. 'Standard' and 'uniform' allocation rules admit different types of low-price equilibria, which are eliminated by a 'hybrid' rule. We observe little evidence of revenue differences among the allocation rules. Under all three allocation rules, prices are competitive when bidders cannot explicitly communicate, and collusive when bidders can explicitly communicate. With explicit communication we find that collusive agreements are more fragile when the gain from a unilateral deviation is larger. We also find that the standard allocation rule is particularly vulnerable to collusive equilibria.