A group of linguists tested how listeners perceived accents after hearing biased comments about a person’s speech. The participants in the experiment were English speakers living in Montreal, where both English and French are spoken, and participants rated recordings of local French speakers on their L2 English. The listeners were tested with three types of comments about the recordings: one group heard no comments (baseline), one group heard comments that were positive about local French speakers’ English, and one group heard negative comments about their English. The researchers hypothesized that hearing positive comments would lead to higher ratings for the French speakers’ English, and that hearing negative comments would lead to lower ratings.

Dataset: attitudes-fall2022.txt.

Columns: participant is a number identifying the participant. rating is the mean rating a participant gave, bias_group is the group they were in, and age is their age bracket.

Part A (SA5): Load the dataset attitudes-fall2023.txt into the variable sa5 and then compute the following.

1. For each level of bias_group, compute the mean and standard deviation of the rating for that group. Store these as sa5.negative.mean, sa5.baseline.mean, sa5.positive.mean, sa5.negative.sd, and so on.

2. Now for each combination of bias_group and age, compute the mean and standard deviation of the rating. Name the variables as sa5.older.negative.mean, sa5.younger.negative.mean, sa5.younger.baseline.sd, and so on (e.g. the format is sa5.AGE.BIAS_GROUP.STATISTIC).

3. Compute an ANOVA to test if there is a main effect of bias_group on rating. (In other words, does the rating depend on the group participants were in?) Assume that the observations are independent. Store the result of your computation in sa5.anova1.

4. Conduct a post-hoc test using TukeyHSD() to show which levels of bias_group (if any) exhibit the effect. Store the results in sa5.anova1.posthoc.

5. Compute an ANOVA to test if there is a main effect of bias_group and age on rating, and an interaction between bias_group and age. Assume that the observations are independent. Store the result of your computation in sa5.anova2.

6. Conduct a post-hoc test using TukeyHSD() on the results of your ANOVA. Store the results in sa5.anova2.posthoc.

7. Compute an ANOVA to test if there is a main effect of bias_group and age on rating, and an interaction between bias_group and age. The observations are not independent, so you’ll need to use a repeated measures ANOVA this time, with participant as the subject variable. Store the result of your computation in sa5.anova3.

8. Finally, let’s try to compute a mixed ANOVA. In a mixed ANOVA, at least one independent variable is presented within-subjects, and at least one independent variable is presented between-subjects. In this experiment, age is a between-subjects variable, while bias_group is a within-subjects variable. Store your result in sa2.anova3. 

 

Part B:

Using your computations from part A (the previous section), answer the following questions.

1. Produce a boxplot to visualize the relationships between the different levels of bias_group. Remember to adequately label your plot.

2. Report the result of your one-way classic ANOVA (stored as sa5.anova1).

3. Produce an interaction plot to visualize any interactions between the effects of group and age on ratings. Use bias_group as your trace factor, age as your x factor, and rating as your response. Remember to adequately label your plot.

4. Produce a table to present the means of all the groups in the data. Along the top should be the names of the bias groups, and along the left should be the name of the age groups. The following is a starting point:

                    Baseline                           Positive                           Negative

Old

Young

 

5. Report the result of your two-way classic ANOVA (stored as sa5.anova2).

6. Report the result of your two-way mixed ANOVA (stored as sa5.anova3). 

Using PsychoPy Builder, create a semantic decision task where participants are asked to identify if a word refers to an animal or not. The experiment should have four items (2 animal words and 2 non-animal words) which are presented in a random order. Participants should be given 4 seconds to respond per trial. Give feedback if their response was incorrect only.

RStudio (data is below):

Setup: A psycholinguist was interested in how three different types of sentences were processed in real time. She created stimuli, and presented these sentences to participants in a speeded grammaticality judgement task. In this type of task, participants give an acceptability judgement for the sentence on an ordinal scale, and the time the participant takes to give the judgement is measured. The file la1.txt is partial results from this experiment: Participant is the participant number, Group is the type of sentence tested with that participant (1, 2, or 3), and RT is the mean reaction time for that participant. (This data doesn’t contain any information about the rating itself.)

Questions: Answer each of the following questions using the R functions and techniques we’ve learned in class. Use la1 as the variable where you initially store the data. Only a couple questions require you to store calculations in specific variables, but make sure you use the correct variable names. Use comments and whitespace when appropriate to make your code more readable to others.

1. Compute mean, median, and standard deviation for each of the three groups. Store your calculations as group1.mean, group1.median, group1.sd, group2.mean, and so on.

2. Construct a boxplot to summarize reaction times for each level of the independent variable. Include the plot as your answer to this question. See the notes for boxplot() on Quercus for information on how to change group names in a boxplot.

3. Based on the boxplot, does it look like Groups 1 and 2 might have central tendencies that are statistically different from each other? Explain your reasoning.

4. Compute 95% confidence intervals of the mean reaction time for Groups 1 and

2. Do the confidence intervals support your hunch, or do they provide conflicting information? Explain your reasoning.

5. Conduct a relevant statistical test to compare the central tendencies of Groups 1 and 2 and store the result of the test in the variable test1. 1 Report the results, and also report the results of any statistical tests you ran to check assumptions.

6. Does the boxplot mischaracterize the data for Group 3 in any way? Use other methods we’ve learned about in class to examine Group 3. Briefly explain what you did, why you did it, and what you found out. 

For instance, if you had computed a one-sample t-test, you might do test1 <- t.test(a, mu=400).

7. Group 3 looks like it is bimodally distributed (e.g. it’s sampled from a distribution with two peaks). Use subset() to subset Group 3 into two subgroups: one group should be all the values of the dependent variable that are strictly less than the median reaction time, and the other group should be all the values greater than or equal to the median reaction time.

8. Compute mean, median, and standard deviation for these subgroups. Store the computations as group3a.mean, group3a.median, group3a.sd, group3b.mean, and so on.

9. Draw a boxplot for the two subgroups. See the R function list on Quercus for pointers on how to use the names parameter to make the group names more informative. Include the plot as your answer to this question.

10. From the plot, do you expect the Group 3 subgroups to have measures of central tendency that are significantly different from each other? Explain your reasoning.

11. From the plot, do you expect that the subgroups have similar dispersions? Explain your reasoning.

12. Conduct a relevant statistical test to compare the central tendencies of the subgroups you created from Group 3. Store the results in test2. Report the results, and also report the results of any additional statistical tests you ran to check assumptions.

Encourage you to use comments and whitespace (blank lines, spaces, indentation) to make your code more readable.

you might want/need to use some functions with vectors instead of formulas. You can use boxplot() with two vectors instead of a formula by using boxplot(a,b), where a and b are vectors. This works for t.test() and wilcox.test() as well.

DATA TXT:

RT Participant Group
466.76 1 1
489.6 2 1
487.72 3 1
402.51 4 1
504.46 5 1
421.09 6 1
527.49 7 1
420.32 8 1
360.27 9 1
390.98 10 1
527.74 11 1
456.07 12 1
523.72 13 1
519.42 14 1
487.03 15 1
431.2 16 1
644.31 17 1
564.61 18 1
619.6 19 1
405.39 20 1
428.09 21 1
520.42 22 1
357.35 23 1
424.67 24 1
419.1 25 1
488.97 26 1
470.67 27 1
533.38 28 1
527.72 29 1
552.7 30 1
315.92 31 2
456.53 32 2
442.27 33 2
342.5 34 2
493.79 35 2
426.27 36 2
377.36 37 2
515.34 38 2
492.56 39 2
476.48 40 2
486.05 41 2
479.56 42 2
425.08 43 2
488.37 44 2
315.88 45 2
417.15 46 2
434.63 47 2
371.77 48 2
375.57 49 2
375.58 50 2
504.51 51 2
350.41 52 2
582.64 53 2
452.7 54 2
458.69 55 2
314.68 56 2
397.13 57 2
410.09 58 2
394.01 59 2
572.44 60 2
529.42 61 3
495.63 62 3
323.5 63 3
292.77 64 3
516.84 65 3
317.3 66 3
534.84 67 3
350.02 68 3
523.41 69 3
744.39 70 3
307.04 71 3
657.69 72 3
619.51 73 3
588.96 74 3
315.94 75 3
497.8 76 3
443.79 77 3
347.26 78 3
342.11 79 3
321.96 80 3
329.19 81 3
526.74 82 3
594.58 83 3
279.41 84 3
285.41 85 3
520.06 86 3
620.46 87 3
647.73 88 3
292.59 89 3
346.79 90 3

Analyze and write about the results of a visual and an auditory lexical decision task. Analyze a large data set that includes students’ responses for both the visual and auditory modality. Lab is a write-up of these findings.

1. One large data file that includes all experimental trial responses for both the Visual and Auditory modality tasks. Note that the sample size for the visual experiment is 16 and for the auditory experiment is 15.

Presentedz with visual words and pseudo-words in one experiment and auditory words and pseudo-words in a second experiment. Respond whether you saw/heard a real word of English or not. In the visual task, remained fixated on a cross (i.e., ‘+’) presented in the centre of the screen, while the stimulus was visually presented either to the left or right of the fixation point. In the auditory experiment, you heard words and pseudo-words presented either to your left or right ear.

  Words Pseudo-words Number 60 60 Pronounceable N/A Yes Hemi-field/Ear Presentation 30 left/30 right 30 left/30 right String Length

Range: 5-7

Mean: 6.01 (SD = 0.71)

Range: 5-7

Mean: 5.85 (SD = 0.78)

Phoneme Length

Range: 5-7

Mean: 5.45 (SD = 0.59)

Range: 3-6

Mean: 4.58 (SD = 0.87)

Number of Morphemes 1 N/A

In each experiment, there were 60 words and 60 pseudo-words. It was the same 120 items presented in each experiment. Only their modality of presentation changed. Thirty words and 30 pseudo-words were presented to the left (side of the screen in the visual experiment, ear in the auditory experiment), while the other 30 words and other 30 pseudo-words are presented to the right. Side of presentation was balanced across participants using two lists. Each list contained the same set of words and pseudo-words; only presentation side was counter-balanced across lists.

Independent samples t-tests revealed no difference in orthographic length between the words and pseudo-words (t(117) = -1.59, p = 0.11); there is, however, a reliable phoneme length difference between the words and pseudo-words (t(104.28) = -6.37, p < 0.001). We also recorded whether the participant was a native speaker of English or not and whether they are right- or left-hand dominant. Feel free to use these factors in your analysis.

1. Given what you know about the organization of the visual and auditory pathways and lateralization of language processing, state, in very clear terms, what your predictions are. Provide specific predictions for your observation (dependent) variable(s), and select and name your predictor (independent) variable(s). 
2.  
3. Provide a detailed list of the processing steps in R that is performed to arrive at your results. For example, do you want to include incorrect trials trials in your analysis of both dependent variables? Do not provide the R code, but instead write out these steps in plain English.  
4. Provide a series of plots that provide an illustration of your results. You should minimally include a plot of the accuracy and reaction time results and this should be done for each modality. You may also included histograms of your reaction time distributions, but this is not necessary. Be sure your axes are appropriately labeled. 
5. Provide a table that presents your numerical findings in a clear manner. Remember to include both the sample-level averages and standard deviations per condition. 
6. Discuss your results in light of the predictions in Q1 and your findings in Q3 and Q4. Make specific reference to your predictions and the extent to which they are borne out. 
7. Get creative: Incorporate one other factor (e.g., native language, handedness). State your predictions and provide a breakdown of the data in terms of these factors. 
8. Design your own research question: Consider how you might use the same experimental design to understand another linguistic phenomenon and population (e.g. stop VOT, vowel length, lexical tone, Mandarin speakers, French speakers etc.). State your research question, and provide and motivate specific predictions for your dependent variable(s) as a function of your predictor variable(s). Note: Ideally, the new research question will still involve language lateralization. Consult the Gazzaniga readings online for ideas based on the particular specializations of each hemisphere.