Homework 8

Analyzing the Speech Perception Lab Results

Having run yourself as a subject, your task now is to analyze your data from the experiments. This is best done using a spreadsheet program such as Microsoft Excel. Most spreadsheet programs nowadays also allow you to plot graphs. Since the Psyscope data files are plain text files, they can be read on any kind of computer. Note: You may have to rename them to have a .txt file extension on them. Example: myfile is renamed to myfile.txt.

Objectives

Identification Tasks: graph judgments and reaction times for both the English voicing contrast and the Hindi dental/retroflex contrast.

Discrimination Tasks: graph percentage accuracy and reaction times for both the English and the Hindi contrasts.

Discussion: write-up a description and analysis of your results. What differences, if any, do you observe between your performance with the English and the Hindi contrasts? Do you observe any patterns in your reaction times - do shorter reaction times always correspond to more accurate judgments? For the patterns that you observe in the data, try to suggest explanations. The experiments by Werker (1995) should be helpful in this regard.

In displaying the results, there is no specific expectation for how this should be presented -- you should figure out for yourself how your findings can be presented most clearly. But, we expect that line graphs are most likely to yield easy-to-read results.

Note a: If this was a real experiment, then obviously we would want to provide statistical justification for any claims about the data, and we would also aim to use data from a number of subjects. For the purposes of this lab, though, this is impractical.

Note b: However, if you wish to work with other students in the class and combine your data for analysis, then this would be an excellent thing to do. Nevertheless, we recommend that you analyze your own individual data, too, because this will give you an idea of how much can be learned from the individual responses you gathered from yourself.

Analyzing the Data

You are strongly advised to keep a back-up copy of your original data files, in case you accidentally lose data at some point in your analysis, and need to go back to your original files.

Psyscope Data Files look like the following. After some initial header information about what experiment the data is from, when the experiment was run and some subject information, the data is listed trial-by-trial. In the data file below the information listed for each trial consists of the trial number, the stimulus code (see below - for this discrimination experiment, each stimulus code stands for a pair of stimuli), the reaction time (in milliseconds, timed from the beginning of the second stimulus of the pair of sounds, the mouse state (irrelevant in this case), and the response button that was pressed.

First open your spreadsheet program, such as MS Excel, and then open the data file from within the spreadsheet program, i.e. use File > Open, rather than simply double-clicking on the file's icon. When you attempt to open a plain text file in Excel, the program asks you to specify how the file is formatted - you will see a window like the one below.

If you click on Finish in the Text Import Window, the file will open into a regular Excel window, like the following:

The first few lines of the file are not needed for your analysis. Line 18 of the file above is the first one that you'll need ('Trial ... Condition ... Time' etc.). You can delete the unneeded lines by highlighting the rows in the left-hand margin, and then selecting Delete (Edit menu; Mac shortcut - command+k; Windows shortcut - alt+e, d).

If you try to open your data file and it does not appear in the Open dialog box, you should check that your Open dialog box is set to display all files, and not just files in the special spreadsheet format. If your spreadsheet program asks you how the data in the file is organized, you should say that it is tab-delimited.

Three things that you are likely to want to do with your data in the spreadsheet program:

• Search and replace: for example, if you want to change the 'f' and 'j' responses to '0 and '1', then you should select the relevant column of data and run Edit > Replace. Keep in mind that for the ID tasks, 'f' was the response for dental d (which is the English one) and 'j' was the t or the Hindi retroflex D, depending on which experiment you were running. For the discrimination tasks, the 'f' key was the response for "same", and the 'j' key was the response for "different".
  Changing the letters to numbers is useful for computing average accuracy scores.
• Sort: the data starts out ordered by trial number, but if you want to look at all of the trials of the same kind together (e.g., all of the response to a 10ms [dæ]), then you should sort your data such that the data is grouped by stimulus name. To do this, select the entire range of data that you need to sort (all the columns of data, not just the one you want to group by), and then run the Data > Sort command. If you sort by both the stimuli and responses, then you will find that it is easier to find your overall response patterns at a glance.
  (1) You'll start with the file looking like this:
  
  (2) Highlight all of the cells:
  
  (3) Select the columns you want to sort by, using Data > Sort
  
  (4) After pressing ok, you get a reorganized file, in which response to the same sound appear on adjacent lines.
  
• Average: calculating average reaction times is quite easy in a program like Excel. After you have sorted your file such that related trials are on adjacent rows, you just need to do the following.
  (i) In some blank cell in the same row as the first data point that you want to average, type =average(
  (ii) Use the arrow keys to move to the first cell in the range that you want to average. You should notice that the cell that you just typed into shows the label of the currently selected cell.
  (iii) Hold down the shift key, and then while still holding down the shift key use the arrow keys to move to the other end of the range of the cells that you want to average. You should now see that the cell containing your average command specifies the range that you have selected.
  At this point, your window should look something like this:
  
  (iv) Type a right bracket [')'] and then press return. The cell in which you started should now display the average of the range you selected.
  (v) If you are averaging a number of ranges of cells of the same size, select the average cell that you just created and copy it (i.e. command+c on Mac; control+c on Windows). If you paste this into cells in the same relative position to the new range of cells that you want to average, the average will automatically be generated. In this way it is quite easy to compute many averages fairly rapidly.

Stimuli

In the data files for the identification experiments, the sounds are listed by name; in the data files for the discrimination experiments, the sounds are listed by the name Item 1, Item 2 ... etc. The table below shows what the item numbers stand for.

The English stimuli are taken from a computer synthesized continuum, created using the Klatt synthesizer for Macintosh. The only property of the sounds that varies is the voicing onset time (VOT) -- i.e. the time lag between the release of the stop closure and the onset of voicing. The continuum spans times from 0msec to 70msec, in steps of 5 or 10msec. The Hindi stimuli are based on natural recordings from a native speaker of Hindi. Stimuli 1-4 are dental consonants, and stimuli 5-8 are retroflex consonants. The 4 stimuli within each group are acoustically different from one another, but they do not form part of a dental-retroflex acoustic continuum, unlike the English stimuli.

	English ID	Hindi ID	English Discrim	Hindi Discrim
Item 1	00dt	1hindi	00dt -- 00dt	1hindi -- 1hindi
Item 2	05dt	2hindi	10dt -- 10dt	2hindi -- 2hindi
Item 3	10dt	3hindi	20dt -- 20dt	3hindi -- 3hindi
Item 4	15dt	4hindi	30dt -- 30dt	4hindi -- 4hindi
Item 5	20dt	5hindi	40dt -- 40dt	5hindi -- 5hindi
Item 6	25dt	6hindi	50d t -- 50dt	6hindi -- 6hindi
Item 7	30dt	7hindi	60dt -- 60dt	7hindi -- 7hindi
Item 8	35dt	8hindi	70dt -- 70dt < /td>	8hindi -- 8hindi
Item 9	40dt		00dt -- 20dt	1hindi -- 3h indi
Item 10	45dt	< p>	10dt -- 30dt	2hindi -- 4hindi
Item 11	50dt		20dt -- 40dt	3hindi -- 5hindi
Item 12	55dt		30dt -- 50dt	4hindi -- 6hindi
Item 13	60dt		40dt -- 60dt	5hindi -- 7hindi
Item 14			< center> 50dt -- 70dt	6hindi -- 8hindi
Item 15			00dt -- 40dt	1hindi -- 5hindi
Item 16			10dt -- 50dt	2hindi -- 6hindi
Item 17			20dt -- 60dt	3hindi -- 7hindi
Item 18			30dt -- 70dt	4hindi -- 8hindi

Note on Graphing Accuracy Data

In the identification experiment, you will probably want to plot sound category (x-axis), against responses (e.g. % [dæ] responses). Remember that in this study you simply made judgments about individual sounds. There was no right or wrong answer about which category each sound belonged to, but your judgments will probably have been mostly quite consistent.

In the discrimination experiment, you are making a same/different judgment. On this task, there is a fact of the matter about whether the sounds in each pair are the same or not. For some pairs of sounds the answer "same" is correct, for other pairs of sounds the answer "different" is correct. Therefore, if you want to graph your accuracy across all pairs of stimuli, you do not want to just plot your percentage of "same" or "different" judgments. You'll need to convert the scores from percentage same/different to percentage correct/incorrect. This conversion should be pretty straightforward once you think about it - you'll need to use the table above for this.

For tips on how to create graphs in Excel, see the on-line Help that is included with the program.

Questions: email llsp AT wam DOT umd DOT edu