Consider the following excerpts. Do you Agree or Disagree and Why?

- p. 32.  McCarthy writes "For a phonologist to try to analyze all of
  the alternations in the Latin third declension is a bit like a
  syntactician trying to analyze all of the six-word sentences in a
  language. The task isn’t meaningful."

- p. 32. "Which phenomena make good research topics? [...] The
  phenomenon should be reasonably self-contained. A pretty good test
  of this is whether the scope of the project can be described in a
  single sentence that uses only the terminology of the field in which
  the paper is being written." Agree or Disagree and why?

- p. 33. "In OT, nonuniformity is always a consequence of constraint
  interaction, as we saw with the Yawelmani example in §1.6."

- p. 33. "A <i>descriptive generalization</i> is the essential
  intermediate step between data and analysis. Good descriptive
  generalizations are accurate characterizations of the systematic
  patterns that can be observed in the data. They may not require any
  formalization or theoretical constructs beyond the usual linguistic
  terminology [...]  Proceeding straight from the data to the
  analysis, without taking time to formulate an accurate descriptive
  generalization, is never a good idea."


- p. 34.  "A good descriptive generalization will foreshadow the
  analysis. An excellent descriptive generalization will make the
  analysis seem almost inevitable."

- p.34.  "Ideally, descriptive generalizations [developed for an OT
  analysis] will contain statements about target output configurations
  and unfaithful mappings related to those configurations."

- p. 37. "When the data involve phonological alternations, start with
  the faithfulness constraints. They are the best place to start
  because faithfulness constraints are more limited in variety and
  better understood than markedness constraints."

- p. 38. "OT itself does not say much about constraints, but it says
  that there are only two types of them: constraints that evaluate
  individual output forms, and constraints that require individual
  output forms to be identical to their inputs."

- p. 39. "The danger with an insufficiently precise constraint
  definition is that the interpretation and hence the effect of the
  constraint can vary slightly as the analysis progresses."

- p. 42. "Because ranking arguments are so important, it’s unwise to
  claim rankings without first doing the arguments needed to back them
  up."

- p. 43. "Occasionally, someone will object: “Max >> Dep >> Onset
  isn’t exactly wrong. Since it just overspecifies the known rankings,
  it’s completely consistent with the data.” Regardless of whether
  it’s “wrong,” it’s unwise to assert rankings for which there is no
  support."

- p. 47 "The combination tableau is the ideal instrument for
  constructing and presenting ranking arguments, and we will be using
  it for that purpose throughout this book."

- p. 48 "We now have two main tableau formats [violation and
  comparative]. Brasoveanu and Prince (2005: 3–5) have a good way of
  explaining which tableau format to use depending on the
  situation. They distinguish between the ranking problem and the
  selection problem. The comparative or combination format is best for
  the ranking problem, whereas the violation format is best for the
  selection problem."

- p. 49. "Sometimes, however, the known information about ranking
  cannot be accurately represented in a tableau, even with the aid of
  the broken-line convention. [...] [On the other hand, ] The Hasse
  diagram (12) accurately represents the known rankings.

- p. 54 (Doing an analysis). "The first step in the analysis is to
  formulate OT-friendly descriptive generalizations ..."

- p. 55 (Doing an analysis). "We could interrupt the analysis at this
  point to research the topic of syllable-structure constraints, but a
  better move is to put the matter aside and simply formulate an ad
  hoc constraint that is violated by any syllable that is too big."

- p. 55 (Doing an analysis). "Now that we have a hypothesis about the
  constraints that are involved, we can move on to the ranking. It’s
  easiest to begin the ranking process with a faithfulness violation."

- p. 57 (Doing an analysis). "In general, whenever a constraint is
  introduced into an analysis, its potential effects on all existing
  claims about ranking need to be considered."

- p. 60 (Doing an analysis). "Since we have now completed an initial
  pass through the entire descriptive generalization, it’s appropriate
  to summarize the proven rankings in (25) and assess the overall
  analysis."

- p. 60 (Doing an analysis). "The next step is to check all of the
  pairs of constraints that aren’t ranked with respect to one
  another. Do these constraints ever conflict? If so, the analysis is
  incomplete until they have been ranked. Checking all of the pairs of
  unranked constraints is an excellent way of discovering interactions
  that we might have overlooked."

- p. 63 (Doing an analysis). "The best way to [check that all of the
  original data have been correctly analyzed] is with summary tableaux
  that challenge the full constraint hierarchy with all of the
  relevant inputs and candidates."

- p. 71. "We have seen that there are three circumstances where direct
  ranking arguments are impossible: undominated constraints,
  constraints in a stringency relationship, and constraints that only
  break ties. This doesn’t exhaust the possibilities"

- p. 72. "Figuring out which losing candidates need to be considered
  is probably the hardest thing about doing analysis in OT. There are
  two aspects of this problem, one arising in the early stages of
  analysis and the other arising later on. In the early stages, losing
  candidates are needed for the ranking arguments that emerge from our
  descriptive generalizations. In the later stages, the analysis needs
  to be challenged with losing candidates that might reveal its
  inadequacies."

- p. 73 "Given a suspected constraint ranking Const1 >> Const2 and a
  winner [w], we want to find a loser *[l] to prove the
  ranking. Starting from [w], eliminate one or more Const2 violations
  and add one or more Const1 violations with- out adding violations of
  any other constraints, except for those that are already known to be
  ranked below C ONST 1."

- p. 75 "[...] a constraint introduced later in the process of
  analysis can cast a shadow on a ranking argument that had previously
  been OK. That is why it’s so important to do summary tableaux like
  (31)–(34). Summary tableaux provide a final check on the soundness
  of the analysis."

- p. 76 "As Karttunen says, “The specter of an unexpected competitor
  suddenly emerging to eliminate the desired winner is the bane of OT
  analyses.” [...] Karttunen’s solution is to take the problem out of
  the hands of the analyst by implementing a candidate-generation
  algorithm (like Gen, but more limited) and using the grammar to
  check all of the resulting candidates against the intended
  winner. [...] The weakness of this analytic technique is that it’s
  only as good as the algorithm that generates the candidates."

- p. 82 "Harmonic bounding is important for three reasons. First,
  harmonically bounded candidates need to be recognized as a potential
  distraction from the task of ranking constraints.... Second,
  discovering that the intended winner is harmonically bounded by some
  loser is a serious problem. Third, harmonic bounding is important in
  studying language typology."

- p. 88 "Faithful input–output mappings [...] are relevant to the
  analysis because of the assumption that all constraints are
  universally present in the grammars of all languages (§1.3). It’s
  important to discover which markedness constraints a language
  violates, since those constraints must be ranked below
  faithfulness. The only way to discover these constraints is to look
  at faithful input–output mappings.

- p. 88 "One of OT’s basic hypotheses is that constraint ranking is
  the only systematic difference between languages (§1.7). If this is
  true, then languages cannot differ systematically in their
  lexicons."

- p. 89 "There are two main arguments in support of richness of the base.
  One is parsimony: since ranking can differ from language to
  language, the strongest hypothesis is that ranking is the only
  possible difference between languages... The other argument goes
  back to the [observation that] restrictions on the lexicon often had
  the same effect as the phonological rules... This kind of conspiracy
  was referred to as the Duplication Problem ...  Some researchers
  proposed to solve the Duplication Problem by eliminating
  restrictions on the lexicon and using just rules or output
  constraints to account for all generalizations. Richness of the base
  is OT’s instantiation of this idea."

- p. 90 "In general, the grammar of every language has to map every
  possible input to some well-formed output... In general, it isn’t
  enough that the analysis works when the inputs are well-behaved; the
  analysis has to work over all possible inputs."

- p. 92 "This analysis treats the distribution of nasalized vowels in
  Yoruba as a fact about surface forms ... In a traditional analysis,
  neutralization of a contrast is dealt with by a restriction on the
  inputs to the grammar."

- p. 122 "In short, RCD has no way of identifying the crucial rankings
  that make up an analysis."

- p. 132 "Linguists aren’t always accustomed to the idea of studying
  the formal foundations of a theory. As we’ve seen, however, research
  on the formal foundations of OT has yielded practical tools for
  analysis as well as insights of a more abstract character."