What follows is (i) my summary of some of the salient points in Hacker's 1981 paper; (ii) some excerpts from that paper on W's move away from the picture theory; and (iii) a couple of comments of my own. I continue to use the courier font for my own stuff and the georgia font when I'm either quoting or nearly quoting (i.e., paraphrasing) . I also continue to use "W" for "Wittgenstein". The basic premises utilized by W to end up with a picture theory of representation are: 1. "Ordinary language is in order as it is" though this might not always be readily apparent and might take analysis to show it. I.e., it is consistent with the laws of logic that require bivalency and excluded middle. Since so many ordinary language statements are vague, this must be a function of vacuous proper names. These must be handled by the Russellian Theory of Description and the Fregean requirement of determinacy of sense. [Why such changes do not amount to altering ordinary language rather than leaving it as it is, seems odd to me.] 2. The sense of a prop does not generally determine its truth-value, so knowing what something means will not generally require us to know whether it is true or false. That a prop is false does not make it meaningless. So, what makes a prop true can't be its sense alone. 3. Languages have generative powers, which means that props must be composite. Only if props consist of elements can there be rule-governed ways to generate an infinite number of new props out of existing elements. 4. We do in fact understand things—representatio n is not only possible, but actual. It is isomorphism that is necessary to make 1-4 true. If we make a model of a state of affairs, the model will represent in virtue of being isomorphic with what it models. Elements of the model must stand for the elements of what is represented. This, for W, is "the pictorial relation." Models represent states of affairs, with the structure of the latter consisting in the way its constitutive elements are connected with each other. For a model to represent some state of affairs, the elements of the model must be arranged isomorphically with the elements of the represented state of affairs, given the appropriate method of projection. A model is true if things are as the model represents them as being; otherwise it's false, and to know whether some model is true or not, it must be compared with reality. There must be an internal relation between a model and what it represents whether it represents truly or falsely. That is, the "logical form" (or the multiplicity and combinatorial possibilities) of the model and what it represents must be identical. No model can represent its own (internal) relation to what it is a model of—it can only display it. Propositions are a particular type of model and so, too, must have a logical form matching what they represent, whether they are true or false. Their logical form is what is possible for them to say given the rules of logical syntax. Hacker then says this regarding W's development after 1929: In recent years there has been a justifiable reaction to the initial conception of the relationship between W's two masterpieces. To be sure there is profound change in his philosophy, but there is also profound continuity. But exactly what changes and what continues is no easy matter to discern. This is not surprising, for if what W has done is rotate the axis of reference of his investigation 180 degrees [see PI, sec. 108] then the difference of the sameness, as it were, will be difficult to perceive…. W continued to think that psychological features of thought processes are logically irrelevant. In the PI, he insists repeatedly that mental representations and accompanying experiences are irrelevant to sense and understanding. The doctrine of avowals underlines the principle that it must always be possible to distinguish being true from being believed to be true, and the private language argument emphasizes the necessity of the distinction between being right and believing oneself to be right. So here we find an anti-psychologism, and affinity with Realism, which, because of the criteria link neither involves the Realist disregard for the conditions of possible knowledge as determining the bounds of sense, nor slips into the typical reductionism of Anti-realism. Despite this affinity, however, een the anti-psychologism is transformed. In the first place, it is no longer wedded to Realist dogmas—in particular the transcendence (as opposed to the independence) of truth. In the second, the boundary between philosophy and psychology has shifted dramatically. The Tractatus was tacitly or explicitly committed to a host of psychological hypotheses about arcane mental processes whose relation to reality was mediated by language. Thought, understanding and belief, although they had a logical structure similar to the proposition, and contained unknown psychic constituents, were of no philosophical consequence (except in so far as sentences like "A believes p" threaten the thesis of extensionality) . The assignment of meaning to indefinables, the forging of links between language and reality, applying the method of projection are all mental processes. How they are done is a matter for psychology; all that concerns logic is that they are done. In the later work this is repudiated. The subjects of meaning, understanding and thinking are essential to a proper grasp of the nature of language. For the relations between meaning that p, understanding `p' and the ense of `p' are internal. Therefore no psychological explanation or hypothesis can replace a philosophical account of these relations…. This…leads to the total repudiation of…the Realist dogma…of the irrelevance to logic of the grounds of judgment. The grounds of judgment, being what justify assertion, constitute at least in certain cases, the sense of a proposition. . The grounds are grammatically related to the proposition and tell us what proposition it is (Zettel, sec. 437). … In short, the later philosophy replaces the Realist methodological principles by diametrically opposed principles. The bounds of sense and the limits of possible knowledge must coincide. We can squeeze no more sense out of a proposition than we can put into one. We can assign sense to a prop only in so far as we can stipulate the conditions which would justify its employment. Consequently the crucial strategic principle that sense is given by truth-conditions independently of means of recognition of truth, which dominates the Tractatus semantics, is not rejected….[By the time of PI] the grounds for an assertion are part of its grammar and tell us what proposition it is. To specify the grounds for an assertion is to explain its sense….The contrast with the picture theory of meaning here runs deep…. Hacker's interesting paper concludes as follows: To be sure, [PI's] answer to the great problem of the harmony between language and reality seems, by comparison with the picture theory of meaning and its exciting logico-metaphysical atomism, trivial, even uninteresting. Madness is more interesting than sanity. But it is much better to be sane than to be mad. I just want to add my own sense that the continuing fascination of the Tractatus among analytic philosophers likely stems from an unwillingness (which I share) to drop Tractarian doctrines regarding the independence of logic from epistemology and psychology (as described by Hacker above) just because one wants to drop the atomism and the picture theory of representation. I think the "ordinary language philosopher' s" use of "paradigm case" arguments and the regular rejection of the distinction between ratio essendi and ratio cognoscendi by W and his Oxford followers has been enough to cause many analytic philosophers to look back longingly on the Tractatus with its clear distinctions between what is being said and why one is saying it. Hacker is right that is important not to yield to madness, but one must also be careful that the treatment not be as harmful as the disease. Walto