PLID50留学生讲解、辅导Java，C++程序设计、讲解Python编程语言 解析SPSS|调试Matlab程序
- 首页 >> 其他 Due: 26 November 2019 PLID50 | Lab 4/5
Lab 4/5: Top-down effects on Perception
These final two laboratory assignments examine a very similar phenomenon: how other cues
(e.g., faces, words) affect your perception of speech sounds. Here, you will synthesize the skills
you have acquired in the course to this point – data summarization, plotting, outlier rejection,
reasoning about results. Assignments that are explanatory and well-written will receive higher
marks. The writing will be coherent and demonstrate thought about the purpose of the assignment
and the potential research question tackled by the lab. Each lab will be worth 50 points.
This will result in 100 points total. Please turn in one PDF file to Quercus by November 26.
Lab 4
Methods Lab 4 was a phoneme identification experiment. Your task was to determine if you
heard [sɑ] or [ʃɑ]. Sounds were paired with faces that appeared on the screen. A nine-step [s]-
[ʃ] continuum was constructed from a female voice who produces the sounds naturally. The continuum
was created by gradually cross-mixing the two sounds together such that at one point on
the continuum, the sound was 100% [s]/0% [ʃ] (s1) and at the other end of the continuum, the
sound was 0% [s]/100% [ʃ] (s9). The midpoint (s5) was 50% [s]/50% [ʃ]. Using the Change
Gender function in Praat (Boersma & Weenik 2001), we adjusted the median pitch of each token
to 140 Hz and adjusted the Formant Shift Ratio by a factor of 0.7. This was done to construct
tokens that were relatively ambiguous with respect to the gender of the talker. Each of the
nine tokens were paired with two faces: a male caucasian face and a female caucasian face.
This led to a 2 x 9 design: Condition (male vs female) x Continuum (nine steps). The ITI was
1000 ms. The face and sound file were played simultaneously. The face remained on the
screen for 1000 ms or until a response was made and the next trial began. Ten repetitions per
stimulus/face combination were played, resulting in 180 total trials (18 combinations (2x9) presented
10 times each). The experiment took approximately 5 minutes to complete. Eleven students
participated. Participants were instructed to press the ’s’ key if they heard [sɑ] or the ‘h’
key if they heard [ʃɑ]. Reaction times were calculated at the onset of the syllable presentation.
1. Methods: Provide a brief synopsis in your own words about the goal of the study (you might
have to write this after you analyze the data) and the methods employed. Minimally, state
the task, design, stimuli, number of participants and number of trials. This information is provided
above, but write it in your own words. Do not copy and paste what I have provided.
2. Outlier rejection: Provide a by-participant histogram for the reaction times for each participant
(you should have a total of 20 histograms). Identify a reasonable outlier rejection criterion.
State this. Apply this rejection method and re-plot the by-participant RT histograms. You
should present two plots (1 before and 1 after outlier rejection and the criterion that was
used).
3. Plot: Provide a geom_line() + geom_point() + geom_errorbar() ggpplot of the proportion [s]-
responses on the y-axis and continuum on the x-axis. You should be plotting your cleaned
data here. Note that you may have to create a new column that numerically codes for
whether the participant responded “s” or “sh”. Your plot should contain error bars of the
standard error of the mean. Moreover, the axes should be labeled coherently and not using
the default labels.
4. Interpret: Is a perceptual bias present in the results? That is, does the presentation of a face
in conjunction with a speech sound change our perception of the speech sound. Be specific.
Where is this bias present? How do we conclude, specifically, from the results that this bias
exists?
Due: 26 November 2019 PLID50 | Lab 4/5
5. Compare: Consult Strand & Johnson (1996) Experiment 3, where visually presented faces
were played in conjunction with the speech sounds. In particular, focus on Figure 2.3. Compare
the results in Q3 here to their Figure 2.3. What are the similarities and differences?
Note that I am not interested in the particulars of the plot (e.g., they used percentages, s1
was most [ʃ] like). Focus on the qualitative pattern of data.
6. Statistics: Finally. identify the ambiguous point in the plot and run a t-test (though note that
we are ignoring the assumptions of the test for the purposes of this assignment). State your
findings in paragraph form, include your null and alternative hypotheses, and conclude
whether or not we can accept or reject the null hypothesis, i.e., no effect of gender of face
on perception of speech sounds. Remember that p < 0.05 is our criterion for rejecting the
null hypothesis.
Lab 5
Methods Lab 5 was also a phoneme identification experiment. In particular, you heard a CV(C)
(C) syllable and were asked to respond whether the initial consonant was [k] or [g]. A seven step
continuum was constructed by manipulating the duration of VOT: 11 ms, 18 ms, 25 ms, 33 ms,
40 ms, 47 ms, and 54 ms in duration. Recall that VOT is primary acoustic cue for voiced versus
voiceless stop/plosive consonants in English. These seven tokens were embedded inside of
three _V(C)(C) frames: [_ɪ], [_ɪft] and [_ɪs]. The ITI was 1000 ms. The next trial began after a
response via button press: ‘k’ if they heard the [k] sound and ‘g’ if they heard the [g] sound . The
experiment design was 3 x 7 (Frame (3 frames) x Continuum (7 steps)). This led to 21 total
combinations. Each stimulus was played 12 times, resulting in 252 total trials per participant. A
total of 14 students participated in the experiment. Each experimental session lasted approximately
15 minutes.
7. Methods: Provide a brief synopsis in your own words about the goal of the study (you might
have to write this after you analyze the data) and the methods employed. Minimally, state
the task, design, stimuli, number of participants and number of trials. This information is provided
above, but write it in your own words. Do not copy and paste what I have provided.
8. Outlier rejection: Provide a by-participant histogram for the reaction times for each participant
(you should have a total of 14 histograms). Identify a reasonable outlier rejection criterion
(e.g., > 5 sec). Apply this rejection and re-plot the by-participant RT histograms. You
should present two plots (1 before and 1 after outlier rejection and the criterion that was
used).
9. Plot: Provide a plot of the proportion [g]-responses on the y-axis and continuum on the xaxis.
This should be one geom_line + geom_point() ggplot with three distinct lines, one for
each Condition/Word Frame. These should be colour coded.
10. Interpret: Does the presentation of a _V(C)(C) frame in conjunction with a speech sound
change our perception of the speech sound? Be specific. Where is this bias present? How
do we conclude, specifically, from the results that this bias exists?
11. Extend: What does this perceptual bias mean? It is important to consider the relationship
between the interpretation of the speech sound and its phonetic frame. What is the information
that is causing the observed perceptual bias? Is there a control in the experiment? If so,
what do you consider to be the experimental control?
12. Compare: Consult Ganong (1980) Experiment 1, where visually presented faces were
played in conjunction with the speech sounds. In particular, focus on Figure 3. Compare the
results in Q9 here to his Figure 3. What are the similarities and differences? Note that I am
not interested in the particulars of the plot. Focus on the qualitative pattern of data.
Lab 4/5: Top-down effects on Perception
These final two laboratory assignments examine a very similar phenomenon: how other cues
(e.g., faces, words) affect your perception of speech sounds. Here, you will synthesize the skills
you have acquired in the course to this point – data summarization, plotting, outlier rejection,
reasoning about results. Assignments that are explanatory and well-written will receive higher
marks. The writing will be coherent and demonstrate thought about the purpose of the assignment
and the potential research question tackled by the lab. Each lab will be worth 50 points.
This will result in 100 points total. Please turn in one PDF file to Quercus by November 26.
Lab 4
Methods Lab 4 was a phoneme identification experiment. Your task was to determine if you
heard [sɑ] or [ʃɑ]. Sounds were paired with faces that appeared on the screen. A nine-step [s]-
[ʃ] continuum was constructed from a female voice who produces the sounds naturally. The continuum
was created by gradually cross-mixing the two sounds together such that at one point on
the continuum, the sound was 100% [s]/0% [ʃ] (s1) and at the other end of the continuum, the
sound was 0% [s]/100% [ʃ] (s9). The midpoint (s5) was 50% [s]/50% [ʃ]. Using the Change
Gender function in Praat (Boersma & Weenik 2001), we adjusted the median pitch of each token
to 140 Hz and adjusted the Formant Shift Ratio by a factor of 0.7. This was done to construct
tokens that were relatively ambiguous with respect to the gender of the talker. Each of the
nine tokens were paired with two faces: a male caucasian face and a female caucasian face.
This led to a 2 x 9 design: Condition (male vs female) x Continuum (nine steps). The ITI was
1000 ms. The face and sound file were played simultaneously. The face remained on the
screen for 1000 ms or until a response was made and the next trial began. Ten repetitions per
stimulus/face combination were played, resulting in 180 total trials (18 combinations (2x9) presented
10 times each). The experiment took approximately 5 minutes to complete. Eleven students
participated. Participants were instructed to press the ’s’ key if they heard [sɑ] or the ‘h’
key if they heard [ʃɑ]. Reaction times were calculated at the onset of the syllable presentation.
1. Methods: Provide a brief synopsis in your own words about the goal of the study (you might
have to write this after you analyze the data) and the methods employed. Minimally, state
the task, design, stimuli, number of participants and number of trials. This information is provided
above, but write it in your own words. Do not copy and paste what I have provided.
2. Outlier rejection: Provide a by-participant histogram for the reaction times for each participant
(you should have a total of 20 histograms). Identify a reasonable outlier rejection criterion.
State this. Apply this rejection method and re-plot the by-participant RT histograms. You
should present two plots (1 before and 1 after outlier rejection and the criterion that was
used).
3. Plot: Provide a geom_line() + geom_point() + geom_errorbar() ggpplot of the proportion [s]-
responses on the y-axis and continuum on the x-axis. You should be plotting your cleaned
data here. Note that you may have to create a new column that numerically codes for
whether the participant responded “s” or “sh”. Your plot should contain error bars of the
standard error of the mean. Moreover, the axes should be labeled coherently and not using
the default labels.
4. Interpret: Is a perceptual bias present in the results? That is, does the presentation of a face
in conjunction with a speech sound change our perception of the speech sound. Be specific.
Where is this bias present? How do we conclude, specifically, from the results that this bias
exists?
Due: 26 November 2019 PLID50 | Lab 4/5
5. Compare: Consult Strand & Johnson (1996) Experiment 3, where visually presented faces
were played in conjunction with the speech sounds. In particular, focus on Figure 2.3. Compare
the results in Q3 here to their Figure 2.3. What are the similarities and differences?
Note that I am not interested in the particulars of the plot (e.g., they used percentages, s1
was most [ʃ] like). Focus on the qualitative pattern of data.
6. Statistics: Finally. identify the ambiguous point in the plot and run a t-test (though note that
we are ignoring the assumptions of the test for the purposes of this assignment). State your
findings in paragraph form, include your null and alternative hypotheses, and conclude
whether or not we can accept or reject the null hypothesis, i.e., no effect of gender of face
on perception of speech sounds. Remember that p < 0.05 is our criterion for rejecting the
null hypothesis.
Lab 5
Methods Lab 5 was also a phoneme identification experiment. In particular, you heard a CV(C)
(C) syllable and were asked to respond whether the initial consonant was [k] or [g]. A seven step
continuum was constructed by manipulating the duration of VOT: 11 ms, 18 ms, 25 ms, 33 ms,
40 ms, 47 ms, and 54 ms in duration. Recall that VOT is primary acoustic cue for voiced versus
voiceless stop/plosive consonants in English. These seven tokens were embedded inside of
three _V(C)(C) frames: [_ɪ], [_ɪft] and [_ɪs]. The ITI was 1000 ms. The next trial began after a
response via button press: ‘k’ if they heard the [k] sound and ‘g’ if they heard the [g] sound . The
experiment design was 3 x 7 (Frame (3 frames) x Continuum (7 steps)). This led to 21 total
combinations. Each stimulus was played 12 times, resulting in 252 total trials per participant. A
total of 14 students participated in the experiment. Each experimental session lasted approximately
15 minutes.
7. Methods: Provide a brief synopsis in your own words about the goal of the study (you might
have to write this after you analyze the data) and the methods employed. Minimally, state
the task, design, stimuli, number of participants and number of trials. This information is provided
above, but write it in your own words. Do not copy and paste what I have provided.
8. Outlier rejection: Provide a by-participant histogram for the reaction times for each participant
(you should have a total of 14 histograms). Identify a reasonable outlier rejection criterion
(e.g., > 5 sec). Apply this rejection and re-plot the by-participant RT histograms. You
should present two plots (1 before and 1 after outlier rejection and the criterion that was
used).
9. Plot: Provide a plot of the proportion [g]-responses on the y-axis and continuum on the xaxis.
This should be one geom_line + geom_point() ggplot with three distinct lines, one for
each Condition/Word Frame. These should be colour coded.
10. Interpret: Does the presentation of a _V(C)(C) frame in conjunction with a speech sound
change our perception of the speech sound? Be specific. Where is this bias present? How
do we conclude, specifically, from the results that this bias exists?
11. Extend: What does this perceptual bias mean? It is important to consider the relationship
between the interpretation of the speech sound and its phonetic frame. What is the information
that is causing the observed perceptual bias? Is there a control in the experiment? If so,
what do you consider to be the experimental control?
12. Compare: Consult Ganong (1980) Experiment 1, where visually presented faces were
played in conjunction with the speech sounds. In particular, focus on Figure 3. Compare the
results in Q9 here to his Figure 3. What are the similarities and differences? Note that I am
not interested in the particulars of the plot. Focus on the qualitative pattern of data.