Lab 4: Syllable Restoration

In the phoneme restoration studies we talked about in class (Warren 1970, Samuel 1981), a phoneme was replaced with noise, but people were still able to recognize the word. Samuel (1981) found some evidence that people were more likely to notice that the phoneme was missing when it was at the beginning of the word versus the end of the word, giving some support to speech perception models which put a special emphasis on word onsets, like the COHORT model of Marslen-Wilson and colleagues. However, other work has questioned this emphasis on onsets (e.g. Nooteboom & van der Vlugt, 1988; our own Rochelle Newman, Sawusch, & Luce, 2005).

In Lab 4, we will explore the onset question using a similar paradigm, syllable restoration. Rather than simply blocking out phonemes with noise, we will block out whole syllables. Our experiment will manipulate two factors: 1) the position of the blocked out syllable in the word, and 2) whether an entire syllable or parts of two different syllables are blocked out. Participants will have to judge whether the stimuli is a word or a nonword, and our dependent measures will be reaction time and accuracy.

The first manipulation will let us add a new piece of evidence to the debate about the importance of onsets. The second manipulation will let us begin to explore questions about syllable representation. Some models of speech perception put a special emphasis on syllable-sized chunks as input. However, it is unclear whether these models would predict comprehension to be more difficult when two syllables are partially disrupted, or whether it would be more difficult when a single syllable is completed disrupted.

Everyone in the class will be involved with all levels of the experiment: material creation, testing, and analysis. Below we’ve listed the main components of the project. We will post specific instruction sheets for each main component separately. Tentative deadlines will be posted on the schedule but are subject to change! Since Lab 4 will involve more work than the other labs, and since we decided not to have you guys do a Lab 5 this semester, Lab 4 will count as two labs in your final grade.

Project components

- Materials

- Find words

- Find nonwords

- Have someone record all materials

- Splice in noise in Praat

- Collect all materials together

- Put into Linger software for experiment presentation

- Running experiment

- Everybody bring two friends into the lab

- Analysis

- Collect data into one set

- Analyze group data

Manipulations:

- position of noise in 3- or 4-syllable words (syllable 1, 2, or 3)

- whether noise is over a whole syllable or a separate coda and an onset (bachelor: ba__lor vs bach__or)

Measure:

Lexical decision ~ reaction time and accuracy

Items:

At least 90 word test items, 90 nonwords

Around 15 per condition:

-15 intact

-15 first syllable in noise

-15 second syllable in noise

-15 third syllable in noise

-15 first syllable rhyme and second syllable onset in noise

-15 second syllable rhyme and third syllable onset in noise

Design:

Latin square (subject sees one of four lists, in each list a given word will appear in a different condition)