At every jump from original IEEE 802.11-1997 DSSS -> 802.11b -> 802.11g -> 802.11n, there have always been small optimizations that the higher-speed devices could use to maximize their throughput when no legacy devices were present. Or to say it the other way around, there have always been small adjustments they'd have to make to not interfere too badly with legacy clients when legacy clients were present. These optimizations and adjustments are generally "minor but measurable", like a 10% speedup/slowdown.
It's a huge myth that the presence of legacy devices forces all clients on the network to completely abandon their higher-speed capabilities and use only legacy rates. That has never been true. So if you had some 802.11n 2x2 clients getting 144mbps on HT20 (20MHz wide channels), the presence of an 802.11b client would not knock them down to 11mbps.
Part of the Wi-FI certification testing (required if you want to use the Wi-Fi logo on your products) ensures that newer higher-rate products do not slow down to legacy rates in the presence of legacy clients, and also ensures that the newer higher-rate products don't destroy the legacy clients' ability to get reasonable rates.
article's from 2004, and only discusses 802.11b/g mixed-mode. need something more recent. – quack quixote – 2010-02-04T00:14:05.080