A nonlinear dynamic model for the quotes issued by nasdaq dealers is considered. The model focusses on the top two electronic communication networks (ecns), island and instinet, and the three most active market makers for a sample of twenty stocks. The model extends the standard linear vector error correction model for price discovery in three different ways. First, quote adjustments are set relative to the inside quote, i.e. The best bid and ask in the market. Second, dealers react to the inside spread. Third, adjustments differ according to which dealer is currently at the inside. Adjustments are different if an ecn is currently at the inside compared to an individual dealer. This difference is attributed to the asymmetric information among dealers. Price discovery dynamics are studied using generalized impulse response functions.