Aside from the Historical Introduction, Simpson's paper is somewhat tedious. The gist of it is the three-step definition of the Baldwin process, p112, second column. It is perhaps best to simply read this as a series of three observations. Baldwin à la Simpson:

1. Individual organisms interact with the environment in such a way as systematically to produce in them behavioral, physiological, or structural modifications that are not hereditary as such but that are advantageous for survival, i.e., are adaptive for the individuals having them.
2. There occur in the population genetic factors producing hereditary characteristics similar to the individual modifications referred to in (1), or having the same sorts of adaptive advantages.
3. The genetic factors of (2) are favored by natural selection and tend to spread in the population over the course of generations. The net result is that adaptation originally individual and nonhereditary becomes hereditary.

End Quote

The key is whether there is a connection between (1) and (2). In most of his discussion, Simpson seems not to assume a connection. Yet, absent a connection, there is no “effect” in the Baldwin effect. On p115 (second column, third sentence), Simpson finally points out that either there is or there is not a connection between (1) and (2).

(i)
If there is no connection, we are just dealing with statistical coincidence. In the simplest case: (1) Organisms adapt (in a non-hereditary fashion, i.e., by plasticity) to a change in the environment, (2) a mutation occurs that happens to have the same effect as the adaptation, (3) it gets selected. Simpson notes that in such a case step (1) is irrelevant.

(i.i)
It is actually not quite correct to say that step (1) is irrelevant in the absence of a connection. In this case, the utility of the adaptive move consists in permitting the organisms to remain in the changed environment, therefore inducing the selection pressure needed for the mutation with the appropriate effect to sweep the population. (Careful: this is different from making the appropriate mutational effect more likely; see below.) This point was apparently made by Huxley.

Example: An animal finds better food at high altitude. It moves to higher altitude and adapts plastically to lower oxygen tension (using allosteric effectors acting on hemoglobin). This permits the animal to stay there, thus causing selection pressure that rewards mutations producing hemoglobin appropriate for this altitude (i.e., without the need of allosteric effectors to function under default conditions). Suppose an animal has that plastic range, but chooses not to reside in a new environment (there is no better food at high altitude, say). In that case the above mutational effect on the phenotype would still occur with the same frequency, but obviously it would not sweep the population, since there is no selection pressure for it.

(ii)
Simpson then briefly ponders on the possibility that there is a causal connection between (1) and (2). But here he misses the point. He thinks that the only possible connection is Lamarckian: The adaptation somehow causes an appropriate mutation within the same individual during its lifetime (that’s, as far as I’m concerned, the meaning of “Lamarckian”). Naturally, Simpson dismisses this possibility.

(iii)
Waddington is a genius, but his writings are not always the clearest. He notes Simpson’s mistake in (ii). He writes that a causal connection between (1) and (2) need not be Lamarckian. It could be of a form that “the initial non-hereditary response […] enables natural selection to set the stage in such a way that the useful genetic effect is likely to occur.”This needs a bit of work.

In essence, Waddington realizes that selection does not have to be on a trait, it could also be on the likelihood of a trait. He states that genotypes with “the ability to produce an adaptive phenotype would […] encourage the appearance of genetically controlled variants [i.e. mutations] mimicking the adaptive type”. As far as I’m concerned, this is a forerunner of Kirschner's and Gerhart's concept of "facilitated variation". For Waddington a genotype was a range of phenotypic possibility in a particular environment. That range becomes manifest through plasticity, i.e., the capacity to adapt in a non-hereditary fashion through learning or physical interaction with the environment (during and after development). A genotype is selected on the basis of that range. The effect of a mutation is determined by how it modifies that range, which is unlikely to be arbitrary since that range is the outcome of prior selection, and thus presumably implemented by specific mechanisms (p386, towards the end of 2nd paragraph). Any mechanism, however, surely biases the consequences of its own modification in a specific way. (In this sense two implementations of the same behavior are not equivalent.) This is really all that Waddington could say at the time, given the knowledge available to him.

Kirschner and Gerhart rediscover this thought in the light of modern molecular developmental biology. I have come to it from a different angle in a much narrower setting based on a biophysical model of RNA folding. (Not knowing of facilitated variation, Lauren Ancel and I called it plasto-genetic congruence – a term sufficiently ugly that only Gary Odell ever picked it up in a paper.)

I believe Kirschner and Gerhart went farther by:

1. Generalizing Waddington’s thought to multiple levels of biological organization within an organism (i.e., a cell does not only experience aspects of the environment that is external to the organism, but also an environment generated within the organism by other cells),
2. Suggesting specific molecular and architectural design principles (e.g., weak linkage, exploratory behavior, compartmentalization) that provide a mechanistic foundation to Waddington,
3. Suggesting that the mechanisms in (2) interact with the evolutionary process in a way that leads to “innovation”, not only to “canalization” (as Waddington saw it),
4. Hypothesizing that these design principles are maintained, once discovered by evolution, because of their role in facilitating variation.

I don’t think it is necessary to go back much farther than Simpson/Waddington, as many agree that these gentlemen defined (or rather interpreted) the Baldwin effect better than Baldwin himself. (Some authors even refer to it as the Simpson-Baldwin effect, which is rather ironic given that Simpson did not think it was important.)

To sum up:
The Baldwin effect as described in (i.i) may lead to a speed-up in the fixation of a mutation within a population. However, it does not assert a speed-up in the production of a mutation with the desired effect. I disagree with Simpson when he claims that this process does not occur frequently, but I tend to agree with Simpson that it is not of fundamental importance (I guess it depends on what one means by “fundamental”). The Waddington-Kirschner-Gerhart (WKG) effect is described in (iii). It is deeper than the Baldwin effect (i.i) and, to quote a humble Waddington, “unduly neglected by neo-Darwinism”. It asserts the possibility that organizational architectures underlying plasticity make it more likely for mutations to unfold phenotypic effects similar to those attained within the plastic range of possibility.

wf