Outline the paper

[CrumpLab]

We should outline the paper before we write it. This will help us think about how we order what we want to say.

Here is a first crack at the outline:

1. Big issues: how people learn about the structure of their environment
   • need to have a way to measure the structure in the environment (information theory)
   • need to have a theory/model about how structure in the enviroment is learned (instance theory
   • need to have laboratory/real-world environment where behavior in interaction with environment can be measured (typing)
2. Information theory and reaction time
   • explain what entropy means, how to calculate it
   • brief history of importance in reaction time research
   • end with where we at, and how it can be useful
3. Current aims and broader implications of what we are doing beyond typing domain
   • What we are trying to do here, using typing (which measures learning about a structured natural language environment) as a laboratory tool to develop explanatory models of how people learn from the structure of the environment
   • brief foreshadowing of how this can be achieved (e.g., by focusing on phenomena such as first-letter and mid-word slowing, and thinking about whether this phenomena emerges naturally from learning about the structure of the typing environment, or requires special process explanations)
4. Typing phenomena (first-letter and mid-word slowing)
   • introduce each (Ostry, 1983)
5. Existing explanations of typing phenomena
   • brief review of existing explanations. These are likely to be special process explanations (e.g., a word-buffereing planning process)
6. Propose a more general learning and memory process explanation
   • Establish plausibility by describing prior evidence supporting the idea that general learning and memory processes are involved in typing (e.g., prior evidence that typists are sensitive to frequency effects).
   • Describe how the first-letter and mid-word slowing phenomena could emerge from a process sensitive to letter uncertainty
   • Describe what would be required for this to work...e.g., letter uncertainty would have to roughly follow the pattern of first-letter and mid-word slowing
   • Describe that a model would be needed to explain how a general learning and memory process could learn from experience, and produce performance that was constrained by letter uncertainty (ie., the structure in the natural language).
7. Roadmap what will we do in the paper
   • Part 1: Show that we can replicate the phenomena (Experiment 1)
   • Part 2: Show that we can measure letter uncertainty, and that it does explain some of the variance in the behavioral data
   • Part 3: Show that we can model the relationship between mean_iksi and letter uncertainty with an instance theory
   • Part 4: Discuss

Experiment 1 (present the data)

• introductory paragraphs explaining how the data was obtained
• Methods
  • participants
  • stimuli apparatus
  • design and procedure
  • data-analysis
  • Results & Discussion
  • report mean IKSIs by position
  • mention that we replicated first-letter and mid-word slowing

Letter Uncertainty analysis

• explain how we got the letter frequencies
• describe how turned them into H's
• show plot of H's
• report our analysis of shared variance between iksis and H

Modeling Section

• Brief review of instance theory and general principles for how it works
• Explain our model and how we made it, reference code on github
• Show the results of the model
• discuss

General Discussion

• brief synopsis of the highlights of what we did
• talk about what it all means
• talk about the bigger and broader implications
• pat ourselves on the back

References

Figures and tables

[nbrosowsky]

How about the big data / naturally occurring data sets angle?

I'm thinking of this paper we read in class: https://pdfs.semanticscholar.org/f398/11213179b9f8eb1c67f249a1a4b9c6c5bb3b.pdf

[CrumpLab]

It's a good fit. I've got a nearly finished first draft going. How about you read it over and see where you would talk about this. Also, would be good to know if you are able to compile the paper on your end using papaja.

[nbrosowsky]

Hm. I tried to make a few small edits and recompile it, but it ran into errors. Something about not finding pdflatex. I'll have to look into it some more.

[CrumpLab]

Check out the installation requirements for papaja

https://github.com/crsh/papaja

You probably need the latex installation

[nbrosowsky]

ok, after a bit of fussing I got it to work. I pushed a small edit to the paper to test it out.

[nbrosowsky]

One thing I noticed:

The git comparison is done per line (as in where you put line breaks). In your .rmd file you've only put line breaks after each paragraph, which makes sense in terms of writing text. I just tested what happens when you add line breaks after each sentence. It looks like the pdf compiles correctly, but now I assume the git diff comparison will show you which sentences have changed (rather than which paragraph).