Section 13, Translating Logical Connectives: Logical connectives are susceptible to radical translation by use of short occasion or standing sentences along with semantic criteria. For example, the idiom for negation will turn affirmation into dissent and vice-versa. Once a native construction fulfills the semantic criteria, we can consider it understood. This approach does not work if the natives have a pre-logical mentality, but we must assume that logical laws are preserved in translation. Dropping a logical law unhinges truth values in many sentences. Sentences that appear surprisingly false probably turn on hidden linguistic differences. We cannot do the same for categoricals, and the reason is fundamental. Categoricals depend on the objects for their truth, and objects cannot be uniquely determined by stimulus meanings. So, only the truth-functional part of logic can be translated via behavioral criteria. Mereological relations are more susceptible to radical semantic criteria, but the correspondence there is still poor.
Section 14, Synonymous and Analytic Sentences: "Synonymous" here carries the full generality of "same in meaning." The broad type of synonymy is when two sentences receive concomitant assent and dissent, and strictly due to word usage. Carnap's narrower "intensional isomorphism" can be reduced to this broader approach. This notion of synonymy works fairly well for occasion sentences, but not so for standing sentences, despite the fact that the latter remain important in theories. We can lengthen the modulus of stimulation, which improves stimulus meanings and stimulus synonymy only at the expense of the scrutability of the synonyms. The cause of this issue is the interconnection of sentences - there are many ways to accommodate experiences in the network of connections. Grice and Strawson's attempt to solve this using experiences that confirm and disconfirm proposed synonymous sentences fails to improve upon stimulus synonymy. This still-unclear notion of intrasubjective sentence synonymy is equivalent to that of an analytic sentence, which are sentences that we would affirm "come what may." In the end, socialized stimulus synonymy and stimulus analyticity are a behavioristic ersatz. In the face of denials of analytic sentences we tend to assume that there is a problem in understanding the language, and if a native does not agree on such things we tend to think that he cannot be depended on in general. This should not be used to justify an analytic / synthetic dichotomy.
Section 15, Analytical Hypotheses: Results so far: observation sentences and truth functions can be translated; stimulus-analytic sentences can be recognized; native occasion sentences cannot be translated. To get past this, the linguist creates analytical hypotheses where he segments utterances into short recurrent parts and maps them to English words and phrases. The translator must apply parsimony in trading complex analytical hypotheses for consistency of stimulus analyticity. Banal messages are the breath of life in translation. The analytical hypotheses do not imply strict equivalence of words. They are also used to explain syntax. This method accelerates translation by leveraging the destination language. We can re-analyze all this by adding the assumption that the translator becomes bilingual. The finished translation manual is an infinite semantic correlation of sentences, supported primarily only by analytical hypotheses. In part because of the ambiguity of stimulus meanings of terms, native sentences may be expected to be translated in a variety of incompatible ways. Stimuli vastly under-determine the analytical hypotheses. Thus rival systems of analytical hypotheses can fit the overall dispositions to speech behavior yet specify mutually incompatible translation manuals.
Section 16, On Failure to Perceive the Indeterminacy: There are at least seven reasons why one would fail to appreciate the indeterminacy. (1) Analytical hypotheses are confirmed in the field. (2) Confusion with the superficial claim that uniqueness of translation is not expected. (3) Confusion with the platitude that uniqueness of translation is absurd. (4) A feeling that a true bilingual is in a position to make a uniquely correct correlation. (5) Linguists adhere to implicit rules that constrain their analytic hypotheses. (6) A few early analytical hypotheses go far in translation. (7) In framing analytical hypotheses the linguist is subject to practical constraints. There is a parallel between truth of sentences in a theory and interlinguistic synonymy in that they are only meaningful in the context of a theory or set of analytical hypotheses. We should avoid thinking that there is a linguistically neutral meaning of a theoretical sentence; per Wittgenstein, "Understanding a sentence means understanding a language." The main lesson is the empirical slack in our beliefs: radical translation of sentences is underdetermined by dispositions to speech behavior, and our theories and beliefs are underdetermined by the possible sensory evidence.