% Comments on the Lezgian assignment


# General Comments on Writing

1. The purpose of these squibs is for you to get practice writing at a
   professional, academic level. Please endeavor to write at that level!

2. Use section headings such as Introduction, Conclusion and so
   on. Make sure you have a conclusion!

3. The paper is supposed to be self-contained. Therefore, the data
   needs to be introduced in the paper. Don't show all the data once
   as a single block. Instead introduce the relevant data as needed in
   order to make your points. When introducing a set of data points,
   it is useful to use a table instead of listing them all in the
   text.

4. Check spelling and grammar. Every error distracts the reader from
   your points.

5. Follow conventions for referring to phonological representations.
  - URs are given inside //. If there are multiple items, it is more
    readable to present them like this /a, i, u, e, o/ instead of like
    this /a/, /i/, /u/, /e/, /o/.
  - SRs are given inside [].
  - Every time a form is presented in the text, it should be
    accompanied by a gloss in single quotes.
  - Features are presented in []. A bundle of features is also shown
    within [] with the features separated by new lines or commas. For
    example, mid vowels are given by [-high, -low, +syllabic]. Feature
    bundles are not given by [-high][-low][+syllabic].

6. If you invent features, you need to let the reader know. Features
   like [laryngeal], [ejective] and [aspirated] are not actual
   features as far as I know. If you are unsure, present a mini
   feature table showing the feature system you are using, and the
   values you are presuming for each segment.

7. When arguing against a competing analysis, you need to present the
   best version of the competitor, not the worst.

# General comments on Analysis

1. Be careful when making appeals to simplicity. Such appeals require
   a precise measure of simplicity, but this is hardly ever provided.

2. Correctness is an important criteria for a successful analysis in
   the sense it should derive the forms present in the data sample
   ("account for the data"). If your analysis does not derive them,
   then it is not correct.

3. Correctness is not all or nothing. Exceptions to otherwise robust
   generalizations may exist. Perhaps these are memorized or accounted
   for in some other way. A relatively small number of exceptions does
   not necessarily doom an analysis, and they should always be
   acknowledged. On the other hand, presenting a faulty rule or
   ranking argument may entail that a whole section of forms are not
   derived correctly. This can be a more serious form of
   incorrectness.

4. Explanatory power is another important criteria. Suppose your
   account is simply that speakers memorize the forms in the
   data. Then you would be able to derive them correctly (by looking
   up the memorized form), but you have not provided any
   explanation. Explanations make predictions. Good analyses make
   predictions. It is generally a good idea to comment on the
   predictions your analysis makes.

5. These two criteria do not always align and a tension can exist
   between them. An analysis which mostly accounts for the data may be
   better than an analysis which completely accounts for the
   data. Suppose you have a simple theory that predicts 99% of the
   data and makes predictions beyond the data. That is arguably better
   than the analysis that says speakers memorize all the data. This is
   because this latter analysis makes no predictions to novel forms.