There seems to be a distinction between 'descriptive' and 'evaluative' meanings of sentences. Descriptive meanings are cognitive (have truth values) and express propositions. It might be suggested that evaluative meanings are non-cognitive, expressing attitudes instead. Some statements might be purely descriptive, such as "Pandas are animals." Other utterances seem purely evaluative, e.g. "Hooray for philosophy!" Consider the phrase: "These are good scissors," which most clearly expresses a positive evaluation of the scissors. The phrase also seems to involve a descriptive component, i.e. that the scissors are sharp and able to cut well, or something along those lines. ("Good" is rather generic - the descriptive meaning more obvious in the case of more specific evaluative terms such as "intelligent", "courageous", etc.)
What is the relationship between descriptive and evaluative meaning? Non-cognitivists will want to say that the two are entirely distinct, mirroring the fact-value distinction. Moral realists will suggest that there is only one type of meaning (descriptive meaning), though of course language can be employed for a variety of purposes. As a reductionist, I will argue that evaluative meanings are contained within the broader category of descriptive meanings.
I think a desire-based analysis of value (where we take "good" to mean "is such to fulfill the relevant desires") offers a neat explanation of the link between descriptive and evaluative meanings. "These scissors are good" (value) just means "these scissors will fulfill the relevant desires" (fact). To fill in the details: we all recognise that people usually want scissors for cutting stuff. We know that to best fulfill these desires, the scissors should be sharp. So, in the appropriate contexts, we can infer "these scissors are sharp" from "these scissors are good". The descriptive meaning follows from the evaluative meaning, for they are really just one and the same.
It's not clear to me that the non-cognitivist can so readily explain this connection. They hold facts and values to be entirely separate spheres, such that no descriptive facts can entail an evaluative conclusion. So you should, in principle, be able to couple a positive evaluation with any descriptive claim whatsoever. This disallows the sort of reasoning (from one 'type' of meaning to another) demonstrated in the previous paragraph. For example, the non-cognitivist might agree that the scissors are sharp and so would best fulfill our 'cutting' desires, yet disagree that the scissors are in any way "good". This strikes me as inconsistent.
Now, it's true that someone might admit the scissors to be good for our purposes, yet still hold a negative attitude towards them himself. This suggests to me that we need to distinguish between evaluations and attitudes. Evaluations are made objectively, relative to some set of standards. Attitudes are purely subjective - expressions of personal whim rather than real value. So my understanding of 'evaluative meaning' is actually quite different from that initially described in this post. Utterances like "philosophy rules!" express positive attitudes, but have no real evaluative meaning, since it isn't made clear just what the value of philosophy is. (Compare this to "philosophy develops the mind", which makes a descriptive claim about what good it is that philosophy does, and thus could be said to have genuine evaluative meaning.)
So, all meaning is descriptive meaning. Evaluative meaning is just a particular kind of descriptive meaning, namely, the kind concerned with facts about desire fulfillment (or whatever the natural 'reduction basis' for value turns out to be). Value claims are about desires, and thus are cognitive. Expressions of attitude are something else entirely, distinct from the value claims involved in typical evaluative terms like 'good', 'intelligent', etc. The non-cognitivist is mistaken to conflate these.
But we might now instead consider the "meaning/attitude distinction". This looks similar to how the non-cognitivist originally conceived the descriptive/evaluative distinction, except that value is now located in the left box. We can safely agree with the non-cognitivist that this is a strict distinction, for it is also a fairly unimportant one. It no longer allows the non-cognitivist to insist that descriptive premises cannot yield evaluative conclusions, for instance. The analogous claim that meanings cannot yield attitudes seems far less threatening.