February 5, 2001

The readings for this class were:

• review Hand Signals by MacNeilage, Studdert-Kennedy, Lindblom (on reserve)
• Whitney, Chapter 2

The outline for the class is:

1. Announcements
2. Review of MacNeilage, et. al.
3. The physics of sound
   • handout of complex wave
   • handout of speech spectrogram
   • Spectrogram Reading online
   • categorization of speech sounds and VOTs
4. Phrase Structure rules and embedding
   • basic example
   • embedding examples
   • "time flies" example
5. Chomsky's deep/surface structures
   • Basic generative model
   • TG diagram from Figure 2.8 of your text
   • ambiguity example
6. Case Grammar
   • sentence description
   • proposition description
   • The case frame for "open" would be ( _____, O, (I), (A) ) (example)
   • reduced version of previous example

Digressions/Miscellaneous

• Ladle Rat Rotten Hut
• Spectrogram v4.1.2 is a dual channel audio spectrum analyzer for Windows 95 which can provide either a scrolling time-frequency display or a spectrum analyzer scope display in real time for any sound source connected to your sound card. Spectrogram requires Windows 95, at least 16 mbytes of RAM, and a Windows compatible sound card. Spectrogram is freeware.
  • Gram.exe, 258K
  • Gram.hlp, 299K
  • Gram.gid, 9K
  • GramReadme.txt, 1K

Overhead Notes

1. Voice Onset Time (VOT)(onset time is time from release from closure to voicing
   bat /b/ voiced stop
   pat /p/ voiceless stop
   CV syllables (consonant-vowel)
   VOTs < 30 milliseconds perceived as voiced
   VOTs > 30 milliseconds perceived as voiceless
   labials ahve 0-30 msec VOTs, velars (/g/ /k/) have 30-150 msec VOTs
2. drawing of overlapping normal curves illustrating possible Von and Voff detectors, where continuous stimulation by (for example) /b/ will fatigue the Von detector and shift the boundary so Voff is relatively stronger
3. experiments by Warren. "The state governors met with their respective legislatures convening in the capital city." Replace first /s/ in legislatures and replace it with a cough yields phonemic restoration
   "The #eel was on the {table, car}." If # is cough (or other noise), then phonemic restoration is dependent upon the terminal word. Perception isn't phoneme by phoneme, then.
4. Problems with PS (finite state grammars):
   
   1. missing elements: Stop! v. Otto stop it!
   2. ambiguity (structure, phonetic, lexical, or PS)
   3. identical form yet different (Noam is easy to please v. Noam is eager to please; Noam is object, subject respectively)
   4. different form by same meaning (Everyone in the room knows 2 languages v. Two languages are known by everyone in the room)
   5. discontiguous elements (The girl that Bill met lives here. middle 3 words separate elements)
5. Equi-NP deletion transformation rule example
   

   structural 
   description: NP X [ {for} NP Y]
                1  2     3   4  5
        change: 1  2     -   -  5
     condition: 1=4
                with verb such as "want," 1 must be subject of verb
   

   Woodrow wanted for Woodrow to dance -> Woodrow wanted to dance

Parting Thoughts

• Read Whitney, Chapters 3 and 4
• Read The Function of Language Classifications in Behavior by Carroll and Casagrande (on reserve).
• Assignment #3 is due in class on February 12, 2001.

Self-test Questions

• Is it possible to point to any particular word and identify it as the longest word in the English language? Explain.
• Draw a phrase structure tree for "George stayed after school with the teacher." Now extend the tree one level further by showing the morphemes that compose each word (where there is more than one in a single word). Should this tree diagram be considered as one kind of knowledge or two (syntactic and morphemic)? Explain.
• Test your understanding of relative clause formation by determining the main and subordinate sentences in the Ambiguity Example sentence 1b.
• Describe the structural ambiguity of the sentence, "The princess thought the shooting of the prince was terrible." If possible, illustrate your answer with tree diagrams.