I think he's got something of a point, but has to be careful not to overplay it. In several places, including my actual work blog, I've written how there are tipping points in the density of development for each environmental value where marginal increases become generally beneficial, and below that point are generally detrimental. See the link for more details, but because the tipping point is a gradual transition in each case, and occurs at a different level of density for each particular environmental value (walkability tipping point is at higher density, farming and open space at lower levels), I think it's better to talk about a tipping point range.
At a residential density of 1 house per 10 acres, I can think of no environmental reason to support a marginal increase in that density. That would reduce the natural habitat value, make farming more difficult, and put more SUVs on the roads that have to drive long distances to get anywhere useful, all for a tiny increase in housing stock. This is where land use restrictions make enormous sense from an environmental viewpoint (and where conservatives put all their efforts to eliminate restrictions). At a density of 10 houses per acre, the reverse is true - the environmental advantages of low density for farming and open space no longer exist and so can't be harmed further, while walkability and public transit use are feasible, and increases in density mean large increases in housing stock.
So I agree with Matt on the high density end but not at the low end, leaving the small matter of the two orders of magnitude in the middle unresolved. I think for most environmental values, though, the tipping point range can be narrowed to fall between 2 residences per acre and 5 residences per acre. Half acre lots have some, modest, value for open space and wildlife. Just as I can think of no environmental reason to slightly increase densities at very low levels, there are few reasons to do so at the half acre lot size or lower. Conversely, increases from 5 residences per acre to something higher can at least add significant housing amounts and get closer to urban densities that reduce driving.
Of course, the useless zone of 2 to 5 residences per acre is what most of suburban construction has created in the last 60 years.
UPDATE: I should add that the policy relevance is primarily regarding rezoning areas that haven't been fully developed, and redevelopment of urban areas. Incremental changes like whether to permit "granny units" on parcels also apply. And as per the comments, all the above is a generalization subject to exceptions. Clustering development can maximize open space, and dense development can be a stupid idea in the wrong place (like a local proposal to put new development in San Francisco Bay).
