Linger Documentation

Version 2.88


Linger is a Tcl/Tk application that performs a variety of language experiments, particularly self-paced reading. Because it is Tcl/Tk-based, it is easily modified and runs on all major platforms: Unix, Windows, and Mac.

Linger can be used with just about any left-to-right horizontal language for which a font is available. It handles true/false, multiple choice questions, and difficulty ratings, as well as audio playing and recording.

It currently includes the following experiment types:

These different experiment types are explained in their own sections below. Many variations on these experiments are possible by overriding default parameters in an experiment preferences file. Additional experiment types can be created by adding new code to the preferences file. For example, Linger can also be used to generate HTML questionnaires.

Getting Started

To run Linger, you must have a working version of Tcl/Tk installed. To obtain this, read the instructions for your machine type.

Linux

Tcl/Tk comes pre-installed on most Linux machines. Just check that you have an executable named wish. If not, Tcl/Tk can be obtained from your favorite RPM site, or from the Tcl/Tk website.

To run Linger, simply execute the linger.tcl program, which should be marked executable.

Windows

To run Linger on Windows, you will need to install Tcl/Tk. To do that, go to the Tcl/Tk website and follow the links for downloading the latest ActiveTcl distribution. Once Tcl/Tk and Linger are installed you should be able to run Linger by simply double clicking on the linger.tcl file, which should have an icon that looks like a feather.

On Windows you may find that the Tcl/Tk window doesn't completely fill the screen and that the mouse remains visible. I can't find any way around these problems. I recommend moving the mouse to the edge of the screen once you start an experiment.

To figure out which fonts are available on your machine, run Wish (this comes with Tcl/Tk) and type "font families".

Macintosh OS X

First install the latest Tcl/Tk Aqua. Once that is installed, you should be able to run linger.tcl from the Terminal command line in the same way you would for Linux.

Or you can go to the Linger directory and double click the Linger.sit file. This will create an AppleScript file called Linger. Double click that file to run Linger. Note that the Linger window may not pop to the front when it starts.

Macintosh OS 9

First install Tcl/Tk, as explained on the Tcl/Tk website. The OS 9 version of Tcl/Tk should come with a program called "Drag & Drop Tclets". Simply drag the linger.tcl file onto this icon and it will generate an executable. I suggest you name it Linger. The first time you run "Drag & Drop Tclets" it will ask you for a stub to clone. Select the Wish application that is in the same directory as "Drag & Drop Tclets".

On the Mac, you will find that the toolbar menu at the top and the mouse remain visible. I don't know any way around this. I recommend moving the mouse to the edge of the screen once you start an experiment.

To figure out which fonts are available on your machine, run wish (this comes with Tcl/Tk) and type "font families".

Sound and Images

Linger is able to play and record sounds and work with a variety of image types. However, you may need to use some Tcl/Tk extension packages to get the full functionality. In order to display JPEG images, you will need the Img package. To include it, place this at the top of your preferences file:

package require Img
The Snack package is recommended for playing sounds and is required for recording them. It will be included automatically if it is available. To check its availability, run wish and type:
package require snack
The Img and Snack packages are included in recent distributions for Windows and Macintosh OS X. If you are running under Unix, you may need to download and install the packages from these web sites:

Running Linger

It is easiest to run Linger from within the main Linger directory, which contains linger.tcl. If you want to run it from elsewhere under Linux, you must set the LINGER_HOME environment variable to point to the main Linger directory. If the Linger directory is located in your home directory, you can do this as follows:

export LINGER_HOME=${HOME}/Linger
You may want to put this command in your .bash_profile file.

To run Linger on a Unix machine, simply run the linger.tcl script, which should be executable. On a Windows machine, you can double-click the linger.tcl file. On Mac OS 9, you can double-click the executable program you created with "Drag & Drop Tclets". Under Max OS X, you can run linger.tcl from the command line, or you can double click on the Linger script, which is produced when you unpack the Linger.sit file.

When you start Linger, it will look in the Experiments subdirectory of the main Linger directory to find all available experiments. Choose the one you want using the pull-down menu.

The subject number will be set to one more than the highest number for which data has been saved in the chosen experiment. You can type a different subject number if you like. The subject number must be an integer, because it determines which set of items will be selected.

If the subject number is blank or you select Test, the experiment will be run but no data will be collected. If you choose a subject number for which data already exists, Linger will ask before overwriting the old data file.

While in the experiment, you can hit End to jump to the next screen or Esc to kill the program. If you are in the middle of an experiment, you will have to confirm the exit. If Linger is killed, any data collected so far will still be saved. See the section on Aborting and Continuing Experiments.

The Notifier

If Linger is run on a computer connected to the internet, it can send out an AOL instant message to let you know when a subject is done with the practice session and is starting an experiment and when the subject is nearly done (with four items left). This may be useful if the subject is in another room. The file Notifier/experimenters contains a list of experimenter names and their AOL instant messenger ids, as in:

Doug	DougTheMan
John	johnny87
{Duane Wayne} ddog1023

When you start Linger, if you choose one of these names from the pulldown menu, notifications will be sent to that address.

Files and Directories

The Experiments directory should contain one subdirectory for each experiment. Each experiment directory contains a subdirectory called "Results". This will contain the data files for each subject. You don't need to create the Results directory. It will be created when the first subject is run.

The experiment directory should also have the following files:
items contains the experimental and filler sentences and questions
preferences contains commands that override the default preferences
introduction contains commands that run the intro and practice trials
break contains commands that run if a rest break occurs
conclusion contains commands that run at the end of the experiment

The Items File

The items file contains all of the experimental and filler sentences used in the experiment. Blank lines can occur anywhere and are ignored unless they fall between two lines of sentence text. Here is a sample entry in the sentence file:
# e2 16 c
There_were_two guitarists practicing at a_night_club.
One made a living as a musician. The other only played for fun.
A pianist played with the guitarist who@W played@verb for fun in a concert, and they got a record deal.
? Did the guitarist who played for fun get a record deal? Y

Each item begins with a header. The format of the header is:

# experiment_name item_name condition_name

Item Headers

The header must begin with a # symbol, unless you choose a different HeaderMark symbol in the preferences. The experiment_name is used to distinguish items drawn from multiple experiments used in the same subject run. If you only have one experiment, set it to something like E1 for the experimental items. item_name is the name of the item, which will probably just be a number. condition_name is the name of the condition, which might be a number or letter or any combination. The three *_name fields cannot contain colons (:) or whitespace.

A filler item must be given the experiment_name "filler". The condition_name will be ignored, but must be specified. It is best to set it to "-". Practice items can also be included in the items file. They must be given the experiment_name "practice". Practice items will not be shown during the experiment, but you can choose to present them during the introduction using the presentItem command.

Sentences

Following the header is (are?) one or more lines containing the sentence(s). The sentence lines can't begin with a ?, !, or #. Each new line in the sentence file will start a new line in the display. If you don't wrap the lines yourself, any line that is too long for the display will be wrapped for you.

Using the default font on a 1024x768 display, you should be able to fit 100 characters on a line. If your items are longer than one line, be sure to consider where the line breaks might fall and how this could bias your results. It is probably best if the line breaks occur between sentences or during a non-critical region. If lines are too long to fit on the screen, Linger will wrap them for you. You may want to avoid this. You can check if any lines are too long by running Linger with the -c command-line option.

The blank spaces (other than those at the beginning or end of a line) will be retained in the actual experiment. So you should be sure that double spaces only appear where they should (after periods).

If you want several consecutive words to be revealed (in masked reading) as a block, they should be connected with underscores. In the first line of the example, "There were two" appears as a block and "a night club" appears as a block. On the other hand, if you want to split a word into two separate regions with no space between them, use a pipe symbol ("|") to split the words in the items file. This is often necessary in languages like Chinese which does not use spaces between the words.

Each word (or word block, really) has a tag, which will appear in the data file. The default tag is -. If you append @foo to a word, the word's tag will be set to "foo". Tags can be used to identify the words of interest in your experiment. You might tag the main verb with @mv and the noun following it with @obj. Then you can easily extract reading times of interest from the data file by selecting lines whose tag column matches a certain tag.

You could also use tags to label the words in regions of interest. If you have regions A, B, and C, you might label the words: A1 A2 A3 B1 B2 C1 C2, etc. Then you can easily compute region-by-region averages by extracting the lines in the results file with matching tags.

Important note: If your regions of interest do not always start and end on the same word number (or word block number, really) for every item in a given condition, you must use tags to label the regions. Otherwise, once the experiment is done it will be extremely difficult to analyze the data because the regions won't line up.

If a word is given an empty tag, meaning that nothing follows the @, it will be a static word. Static words are not masked during self-paced reading and are visible throughout the sentence. No data will be recorded for them in the results file. Static words might be used to identify the speakers in a dialogue.

Questions

Linger supports true/false and multiple choice questions, as well as collecting ratings. Each item can be followed by one or more questions. They can be of mixed type, but I don't recommend it.

Each type of question must be on a single line, starting with a special symbol. If you want to have a line that is part of an item's sentence, but which should begin with one of the marks that identify a question or item header (#?!*), you can precede the mark with a backslash (\), which will be stripped off but which will prevent the line from being treated as a question or header.

True/False
In the items file, true/false questions consist of a single line beginning with a question mark, followed by the question, and finally the answer. Answers should be Y or N (not yes or no). Here's an example:
? Did the guitarist who played for fun get a record deal? Y

Optionally, the first ? can be immediately followed by a tag, as in ?q1b. In this case, the question is given the tag "q1b". The tag is included in the results file. This can be useful in cases where several different types of questions are used and you would like to perform separate analyses on them.

Multiple Choice
Multiple choice questions are introduced by an exclamation point, rather than a question mark. Following the question is a list of possible answers, enclosed in curly braces, and then the correct answer. If the answers consist of multiple words, underscores should be used instead of spaces. These will be replaced with spaces in the display. Alternately, you may put double quotes around multi-word choices. There may be one or more correct answers. In that case, place the correct answers within curly braces, just like the list of choices. If there is a single correct answer and it has multiple words and you are using double quotes instead of underscores, you must still use the curly braces around the correct answer. That is why using underscores is recommended.

Here's an example with a single correct answer using underscores versus double quotes:

!foo The ________ ate the flowers. {cat dog pink_pig fat_baby} pink_pig
!foo The ________ ate the flowers. {cat dog "pink pig" "fat baby"} {"pink pig"}

In this case, the question has been given the tag "foo". Here is how it would look with two possible answers:

!foo The ________ ate the flowers. {cat dog pink_pig fat_baby} {pink_pig dog}
!foo The ________ ate the flowers. {cat dog "pink pig" "fat baby"} {"pink pig"
dog}
You can place double quotes around single-word answers if you want. It doesn't matter.

The following variables can be set in the preferences file to alter the way questions are asked:

# If true, multiple questions are randomly ordered:
set QuestRand    0
# This is the number of questions asked per item.  If it is the null string,
# all are asked:
set QuestNum     {}
# The probability of asking each question if one or more is available:
set QuestProb    1.0
If an item has four questions, all of them will be asked by default, in the order given. If you set QuestRand to 1, they will be randomly ordered. If you set QuestNum to 2, only the first two questions will be asked. If you set QuestRand to 1 and QuestNum to 2, then two of the four items will be asked, in random order. If you set QuestProb to 0.5, then each of the questions will be asked with 50% probability, which could mean that none are asked, or all of them. If you set QuestNum to 3 and QuestProb to 0.5, then three of the four items will be chosen initially, and then they will each be asked with 50% chance, so you will get from 0 to 3 questions.

If you need more customized question asking behavior, you can override the chooseQuestions procedure, which does the job of choosing the questions to ask for each item. For example, you may have two types of question (about the first clause versus the second clause), with three questions of each type, and you want to ask exactly one question of each type. In that case, you would ise the question tag to label its type and write your own chooseQuestions procedure to make use of this label.

Fill-in-the-Blank
Fill-in-the-Blank questions are preceded by an asterisk (*). The question should be all on one line in the items file, although it may be wrapped in the display. There are no answers following the question. In the results file, the QCRC column will always be 1, indicating correct. You will have to do your own correctness analysis.

The width of the entry box for typing the answer is controlled by the FillInBlankWidth variable. It defaults to 50. The entry box will scroll, so subjects may type more than this if they want.

Fillers and Combining Multiple Experiments

If you are running a single experiment, you will want at least as many fillers as you have items, or subjects may detect the pattern in your items. At least 60% fillers is advisable for most experiments. The more consistent and unusual your items, the more fillers you will need to disguise them.

It can be more efficient, however, to run items from multiple experiments in the same session. The items from different experiments will serve as fillers for each other. Of course, you would want those experiments to use different sentence types. One experiment using sentential complements would not make a good set of fillers for another experiment using sentential complements.

Normally, you will want the number of subjects who run an experiment to be a multiple of the number of conditions in the experiment to balance the number of subjects in each condition for each item. If you are combining several experiments with different numbers of conditions, the naive thing to do would be to run as many subjects as the least common multiple of the two condition numbers, or some multiple thereof. However, if you had three experiments with 4, 5, and 7 conditions, you would need 140 subjects to balance the conditions in this way.

A smarter thing to do is to run at least enough subjects for each experiment individually, and then discard the data for the last few subjects for an experiment if the number of subjects is not a multiple of the number of conditions for that experiment. So, in this example, you might run 49 subjects. All of them will be used for the experiment with 7 conditions. The last 4 will be discarded for the experiment with 5 conditions, and the last 1 will be discarded for the experiment with 4 conditions.

Errors

The most common source of errors when using Linger is bad formatting in the items file, such as a missing or extra parenthesis or curly brace. If Linger dies with an error message after you have chosen the experiment, the error could be due to the items file. In this case, run Linger from the command-line using the "-v 1" option. This increases the verbosity level, causing it to print out each items file line as it is parsed. The last line printed before Linger aborts may be the culprit.

The Introduction, Break, and Conclusion Files

The introduction, break, and conclusion files define the start, rest periods, and end of the experiment, respectively. Each file is just a Tcl script, so you could actually do arbitrarily complex things with them. But there are some default commands that are normally used. They are as follows:

instruct {This is the first line of the text.
This is the second line.
This is the third.}

The instruct command prints some text in the small font and then waits for a key press to continue. The text will automatically be wrapped if the lines are two long, but you can put your own line breaks in as well. Lines will be left justified. This is useful for giving instructions in the introduction file.

command {This is a big font command.}

The command command is similar to instruct but it uses a big font and centers the text. It should be used for short messages, like "Here we go...".

createItem {# practice 1 -
This is a practice sentence.
? Did you read the sentence? Y}

The createItem command is used to create a temporary item. This is most often used to create practice items for immediate presentation. createItem takes one argument, which should be enclosed in curly braces and which should consist of an item description, just as it would appear in the "items" file. You should give practice items the experiment_name "practice" and the condition name "-". Practice items can also be defined in the items file. As long as they have the experiment_name "practice", they won't be presented during the actual experiment.

presentItem practice:1:-

The presentItem command runs a single item using the item-presentation function that is associated with the current experiment. If the experiment is selfPacedReading, then the command that is actually run will be selfPacedReadingItem. Normally this is used to immediately present a practice item that has just been created. The argument consists of the experiment_name, item_name, and condition_name of the item, separated by colons.

The break file is run periodically to give the subjects a break. Normally it will just contain something like:

command {Time for a break.}

The BreakInterval variable determines the number of sentences between breaks. If it is 0, no breaks will be offered.

The Preference File

The preference file is also just a Tcl script file. It will be run just after the experiment has been chosen. The preference file should contain overrides to the preferences set at the top of the linger.tcl file. You can override a parameter simply by setting it to a different value. For example, if you want to put masks over spaces between words, you should set the MaskSpaces parameter to 1 as follows:

set MaskSpaces 1

You can customize many different aspects of Linger in the preference file, including the experiment type, the display fonts, the response keys, the various timing delays, the names of the other files, the frequency of rest breaks, the frequency of questions, the feedback messages for right or wrong answers, and even the special symbols used in the items file.

To see all the variables that can be customized, with short descriptions, just read the top of the linger.tcl file.

If you want to do more profound customizations, such as creating a new experiment type, you should place the code in the preferences file. The preferences file will be read after most of the linger.tcl file, so you can also override any of the functions that are defined in linger.tcl by re-defining them.

Difficulty Ratings

In any of the standard experiment types, subjects can be asked to rate the difficulty or some other aspect of the item. This does not require any addition to the items file. In order to collect ratings, put this in the preferences file:
set CollectRatings 1
The rating will be presented after the sentence, but before any other questions. Subjects can respond my clicking with the mouse or using the number keys. These options affect the ratings:
# The maximum value of the rating scale:
set RatingChoices 7
# The message presented along with ratings:
set RatingMessage "How difficult was it to understand that sentence?"
# The label at the low end of the rating scale:
set RatingEasy "Very Easy"
# The label at the high end of the rating scale:
set RatingHard "Very Hard"
The rating and response time are stored in the Results file.

Aborting and Continuing Experiments

If the Esc key is pressed while Linger is running, or it is killed in some other way, it is possible to resume the experiment where it left off. This is useful if you have subjects who may not be able to sit through a full experiment.

The items that the subject would have seen are stored in a file in the Results directory whose name is the subject number, with the extension .itm. If you run Linger again and choose the same experiment and subject number and this file is found, you will be asked if you want to continue where you left off.

If you do choose to continue, the items will be loaded from the .itm file. Linger will then read the .dat file to figure out where the previous run left off. The new results will be appended to the .dat file.

When you continue an experiment, you may not want to make the subject go through the entire introduction again. In your introduction file, you can check the "Continuing" variable to determine if this is a continuation. Here is how you might do that:

if {$Continuing} {
  put code here that is run when continuing
} else {
  put code here that is run for a new subject
}

The Results Files

Unless it is in "test" mode, Linger will write a data file to the Results subdirectory of the experiment directory. The file name will be the subject number followed by ".dat". The data file contains both the reading times and the question answering correctness and response time.

Each line in the data file contains a single observation and has 8 columns. The columns are separated by whitespace. The columns for a reading time response will be:

subject_num experiment_name item_name condition_name word_num word word_tag time

The word is the actual word or region, from which you can get the word length.
The word_tag is its tag.
The time is in milliseconds.

The columns for a T/F question will be:

subject_num experiment_name item_name condition_name ?tag answer correct time

The answer is the actual response given (1 for yes, 0 for no).
The correct field is 1 if the response was correct, 0 if wrong.
The tag is the tag that followed the ? in the items file.

The columns for a multiple-choice question will be:

subject_num experiment_name item_name condition_name !tag answer correct time

The answer is the actual response given by the subject.

The columns for a difficulty rating will be:

subject_num experiment_name item_name condition_name %tag answer {} time

The answer is the response given by the subject. There is no correct answer, so the 7th column is the null string.

Command-Line Arguments

Normally Linger is run without any command-line arguments. However, you can use arguments to avoid the experiment-choosing window or to use special modes. The following arguments are recognized:
-e experimentThis sets the experiment name, causing it to skip the experiment-choosing window.
-s subjectSets the subject number. If -e is used and this is not, the next available subject number will be used.
-tRuns in testing mode, so no data will be recorded.
-bRuns in batch mode, which is non-interactive. This can be used to have Linger perform special functions, like spewing out information for an HTML form.
-cThis causes Linger to check the length of the lines in every item and report those that will wrap.
-v #Sets the verbosity level. 1 prints out the lines in the items file as they are parsed, which is useful for finding the source of errors. 2 prints out lots of info as the experiment runs.

To check the length of all of the items in experiment Exp1, you could do this:

linger.tcl -e Exp1 -c

Masked Self-Paced Reading

Masked self-paced reading requires the subject to press a key (the spacebar by default) to reveal each word or block of words. A block of words is simply a group of words that appear together. They are joined by underscores in the items file. Hidden words are masked by dashes. By default these don't cover spaces and have small gaps between characters.

This is the default experiment type, so the preferences file could be empty if you just want the standard behavior. Here are some options that you may want to override:

# The font used for the sentences:
set TextFont     "-family {courier new} -size -14 -weight bold"

# Should spaces within blocks be masked:
set MaskJoins    0
# Should spaces between blocks be masked:
set MaskSpaces   0
# The gap between mask dashes (in pixels):
set MaskGap      1
# The number of blocks to show at once with a moving window:
set WindowSize   1

# The key to see the next word:
set GoKey        space
# The key to respond no:
set NoKey        f
# The key to respond yes:
set YesKey       j

# The delay (in msec) before presenting each sentence:
set PreDelay     1000

# The reward message, leave blank for no feedback:
set RightAnswer  ""
# The error message, leave blank for no feedback:
set WrongAnswer  "Oops.  Wrong answer."
# The time the feedback message stays on the screen:
set FeedbackDelay 1500

If the RightAnswer or WrongAnswer message is left blank, no feedback will be given in that case.

If a word is given an empty tag (by appending an @ symbol to the word), it is a "fixed" word, which means it remains on the screen permanently and no timing data is recorded for it. If such a word appeared in the middle of a sentence, the moving window would jump over it.

The moving window can be made larger than a single block by changing the WindowSize parameter. In this case, it will really appear to be more like a window moving over the sentence. Fixed words do not participate in the moving window. The reading time is recorded as if it applies to the first word in the window.

Self-Paced Listening

Word-by-word listening can be used instead of or in conjunction with reading in any of the self-paced or auto-paced experiments. In order to play the words, you will want to use the Snack package. It is possible to play words with an external application, but the timing will not be as accurate. In order to play word-by-word audio, put this in the preferences file:
set PlayWords 1
If you don't want to display the words at the same time, use this option as well:
set ShowWords 0
A variety of sound formats can be used, although WAV files are standard. By default, Linger will expect the first word of experiment E1, item 4, condition b in the following location:
Audio/E1-4-b/E1-4-b-1.wav
and the second word in file:
Audio/E1-4-b/E1-4-b-2.wav
The Audio directory, in this case, is a subdirectory of your experiment directory. You can change the name of the audio directory as well as the mark used to separate the elements of the item name and the extension of the sound file by changing these options in the preference file:
# The subdirectory in which audio is stored.  For self-paced listening, this
# needs a subdirectory for each item:
set AudioDir     "Audio"
# This separates an item's exper., item, condit., and word in audio filenames:
set AudioJoinMark "-"
# The extension on audio files:
set AudioExtension "wav"
Note that each item and condition gets its own subdirectory within the main audio directory. If a sound file is missing, a beep will be placed in place of the word and an error message will be printed to standard error.

Centered Self-Paced Reading

In centered self-paced reading, the words or phrases appear roughly in the center of the screen, with no masking. In the preferences file, you will need to set the experiment type:

set Experiment centerSelfPacedReading

By default, the words are not actually centered, but left justified a certain distance to the left of the center. This is generally better because it allows the reader to always saccade back to the same point to read the next word. You can adjust whether the words are actually centered, and the distance from the center on which they are aligned:

# When doing center*PacedReading, if the words are left justified relative to 
# one another:
set CenterLeftAligned  1
# When doing center*PacedReading, the amount the words are shifted to the left
# from the middle of the screen:
set CenterLeftShift  50
Before the first word appears, there will be a crosshair on which subjects can fixate, to indicate the impending start of a sentence. If the words are left-aligned, it is best if the crosshair appears one or two characters to the right of the alignment point. The location and size of the crosshair can be adjusted with these parameters:
# The size of a fixation crosshair:
set CrossSize    10
# The amount the crosshair is shifted up from center:
set CrossUpShift 0
# The amount the crosshair is shifted left from center:
set CrossLeftShift 0
The delay during which the crosshair is visible is set by this:
# The delay (in msec) before presenting each item:
set PreDelay     1000
And the duration of blank screen between the crosshair and the first word is:
# The delay before the first word in auto-paced reading (and centerSelfPaced):
set AutoPacedInitDelay 1000

Masked Auto-Paced Reading

Auto-paced reading is like self-paced reading, except that it isn't self paced. The word blocks are revealed at a fixed rate. To use auto-paced reading, add the following to the preferences file:

set Experiment autoPacedReading
Most of the options useful for self-paced reading are also useful here. Two addition parameters are:
# The delay between words in auto-paced reading:
set AutoPacedDelay 400
# The delay before the first word in auto-paced reading:
set AutoPacedInitDelay 1000

Centered Auto-Paced Reading

Centered auto-paced reading is like centered self-paced reading, except that it isn't self paced. The word blocks are revealed at a fixed rate. To use this, add the following to the preferences file:
set Experiment centerAutoPacedReading
Most of the options useful for centered self-paced reading are also useful here. The one additional parameters is:
# The delay between words in auto-paced reading:
set AutoPacedDelay 400

Block Reading

This is a simple paragraph reading experiment. The sentence or sentences are shown all at once and the subject presses a key when he or she is done reading. The reading time and the question answering times are recorded in the Results file.

The following variables are useful:

TextFont
TextColor
PreDelay

Speak and Listen

This is an experiment for two subjects running simultaneously on two computers. One subject, the speaker, will read sentences out loud and will be recorded. The other subject, the listener, will listen to the speaker and then answer questions and/or report difficulty ratings. In order to record the speaker, you must have the Snack package installed.

When starting Linger, the two subjects must be given the same subject number, so the item ordering can be synchronized. In order to ensure that the subjects see the same item ordering, the random number generator will be re-seeded using the subject number before randomizing the item order.

Instead of setting the Experiment variable to speakAndListen, you must run this command in the preferences file:

speakAndListenInit
This will present a dialog box that allows you to indicate whether the subject is the speaker or listener. This will set the SALRole global variable to either "Speaker" or "Listener". Then, in the introduction file, you can check this variable in order to present different instructions to the speaker and listener as follows:
if {$SALRole == "Speaker"} {
  put the speaker's introduction code here
} else {
  put the listener's introduction code here
}
This variable sets the message shown to the speaker at the start of each item:
set SpeakerWaitMessage "Wait until the listener is ready."
If the subject number is 5, the speaker's data will be stored in the file Results/5S.dat and the listener's in Results/5L.dat. For the speaker it records the reading time prior to speaking and the speaking time. For the listener it records the listening time and the question responses.

The audio recordings are stored in a subdirectory of the Results directory whose name is the subject number. The file name will be composed of the experiment name, item number, and condition, joined by the AudioJoinMark, which defaults to "-".

You can change the rate and format of the sound recordings by setting the RecordOptions variable:

# Flags for the SNACK recorder:
set RecordOptions {-rate 44100}
# -encoding Lin16 | Lin8offset | Lin8 | Lin24 | Lin32 | Float | Alaw | Mulaw
# -rate 8000 | 11025 | 16000 | 22050 | 32000 | 44100 | 48000
# see the Snack manual for other options

Listen and Answer

This is a primitive form of the Speak and Listen experiment type. It has the subject listen to someone else read a sentence and then answer a question about the sentence. No sentences are actually presented by Linger, only the questions. The sentences will probably be read by the speaker from a hard copy.

The following line should appear in the preferences file:

set Experiment listenAndAnswer
This option sets the message displayed while the subject listens to the sentence:
# The message used in listenAndAnswer trials:
set ListenMessage "Listen to the sentence"
In this type of experiment, you probably don't want to randomize the order of the items. Therefore, you should add:
set RandomOrder  0

Auditory Prime Lexical Decision

In this task, a sound file is played while the subject looks at a fixation cross and then the cross is replaced by a word, on which the subject must perform a lexical decision.

This will only work on a system that is capable of playing a sound file from the command-line. The command for playing a file is stored in the PlayAudio variable. The default is appropriate for most Linux systems:

set PlayAudio /usr/bin/play
Other relevant variables are:
# The size of a fixation crosshair:
set CrossSize    10
The test word is presented using the BigFont:
set BigFont      "-family {helvetica} -size 24 -weight bold"
In writing the items file, the name of the sound file should be used in place of the sentence and the question should just contain the lexical decision word:
# APLD 1 a
/home/foo/Primes/1a.wav
? cantaloupe Y

Using Non-English Languages

Linger can be extended to non-English languages, including Chinese, Japanese, and Russian. You will need to encode the items, introduction, break, and conclusion files using a language-specific extended-ASCII encoding, if they are to include any non-English text. This can be done with Word and possibly with Wordpad or Notepad under Windows, or SimpleText on a Mac.

When you save the file, choose plain text (not rich text format) and then select the appropriate Unicode encoding for your language. Try to save the files with no .txt extension. If you can't save without an extension, try removing it once the file is saved. You can also leave the .txt extension on the files, but you'll need to make a change to preferences, as explained below.

The preferences file itself must be in plain English. In the preferences file, you need to tell Linger what encoding was used for the other files. You do this by setting the LangEncoding variable. For Chinese, the preferences file might include the following:

set LangEncoding big5
set TextFont "-family {song ti} -size -24 -weight bold"

This says that the encoding used on the experiment files is big5. "Song ti" is a Chinese font available on most Linux machines. Other fonts may be downloadable and others are probably available under Mac and Windows.

In order to see which encodings are available in Linger, run the program wish and, at the command line, run "encoding names". The names won't be the same as those available in Word, but you will probably see the correspondence. Make sure you choose the same encoding when you save the files as you do in the preferences file.

I recommend big5 or euc-cn for Chinese, enc-jp or jis0208 for Japanese, macCyrillic for Russian (on a Mac).

In order to see which fonts are available, run wish and type "font families". Look for a name that is similar to a non-English font that you are used to using in Word or some other program.

If you were forced to save the introduction, items, and other files with a .txt extension, you will need to tell Linger about this by putting the following lines (whichever apply) in the preferences file:

set ItemFile     "items.txt"
set IntroFile    "introduction.txt"
set BreakFile    "break.txt"
set ConclFile    "conclusion.txt"

If you need to include encoded text in the preferences file (for example, if you are changing the WrongAnswer message), this may present a small problem. The desired encoding is not normally known until you set the LangEncoding inside the preferences file, and by that time it's too late to encode the file properly. Most changes that would require non-English text can be done at the top of the introduction file instead of in the preferences file. Otherwise, you should put the encoded portions of the preferences in another file, like preferences2. Then you can put this line in the original preferences file, after you set the LangEncoding:

encSource preferences2

How Items are Selected

Linger automatically decides which items to give to each subject in a round-robin sort of fashion based on the subject's number. Subject 1 will get the first condition of item 1, the second condition of item 2, the third of item 3, and so on. Subject 2 will get the second condition of item 1, the third of item 2, and so on. Each subject will thus have 1 condition for each item and an as-even-as-possible number of sentences from each condition.

The number of different choices (different subject lists) will be the least common multiple of the number of conditions in each item. If some items have 4 conditions and some have 3, there will be 12 distinct lists, effectively.

If you drop a subject from the experiment and want to replace his data, it is a good idea to reuse their subject number with the next subject to fill in the hole. Otherwise, the number of subjects per condition may not be balanced. Of course, you will probably want to copy the original .dat file to a backup location before overwriting it.

Once the experimental sentences have been chosen for a particular subject, they are ordered. If you want to present the items in the order specified in the items file, set the RandomOrder variable to 0 in the preferences file.

When ordering the items, items within each experiment are randomized and filler items and items from different experiments are interlaced. If possible, Linger will start with a filler item and no two items from the same experiment will be adjacent. If the items are all from the same experiment, you should have at least as many fillers as items. If you are running items from multiple experiments in one session, items from one experiment can act as fillers for the others, and fewer actual filler items may be necessary.

Customizing Linger

Many aspects of Linger's behavior can be customized by overriding global variables in the experiment's preferences file. Other relatively minor modifications to existing experiment types can be achieved by copying one or more procedures from the linger.tcl file to the preferences file and modifying them. All changes should be made in the preferences file, rather than in linger.tcl itself, so they don't need to be migrated to each new version of linger.tcl.

More advanced users may wish to create their own experiment types. The best way to learn to customize Linger is to begin by becoming proficient with the Tcl/Tk language. A simple tutorial is available here. The Tcl/Tk manual is also very helpful for reference.

The next thing to do is to read the code in linger.tcl. The most useful parts are the global variables at the top and the helper functions which follow them. The helper functions are simple procedures that can be used to facilitate commonly performed operations. They handle tasks such as opening files, printing warning and error messages, randomly sorting lists, measuring elapsed time, creating time delays, displaying various types of messages, and working with sounds.

The procedures for a new experiment can be written from scratch or based on copies of the procedures of an existing experiment. Most experiments will have two primary procedures, although defining some additional helper procedures may be useful. For example, the following procedures define the Auditory Prime Lexical Decision experiment type:

proc auditoryPrimeLexicalDecisionItem item {
  global QuestProb Question Sentence PreDelay
  
  loadSound $Sentence($item) snd
  crossHairOn
  after $PreDelay
  playSound snd
  crossHairOff
  askQuestions $item BigFont center
  deleteSound snd
}

proc auditoryPrimeLexicalDecision list {
  standardExperiment auditoryPrimeLexicalDecisionItem $list
}
The auditoryPrimeLexicalDecision procedure is the main one that runs the entire experiment, other than the introduction and conclusion. The name of this main procedure is the name of the experiment and it is what you must set the Experiment variable to.

The other necessary procedure is the item-presentation procedure. Its name should be the same as the experiment, with "Item" appended to the end. It takes as an argument the code for a single item and controls the presentation of the item and any questions that follow it.

The main experiment procedure takes a list of item codes as its argument. Most of these procedures simply call the standardExperiment procedure. This takes two arguments, the name of the experiment's item-presentation procedure and the list of items. The standardExperiment runs through the list of items, calling the item presentation procedure on each one. It also pauses for rest breaks and sends notifications at the appropriate time.

The auditoryPrimeLexicalDecisionItem procedure begins by declaring some global variables. Any variables that exist outside of the procedure must be declared global before they are used. It then calls the loadSound procedure to load the sound file. The name of the sound file in this experiment type is given in the items file in place of the sentence. So the global variable Sentence($item), which normally contains a list of the sentence lines, actually contains a filename in this case.

The crossHairOn procedure displays a crosshair in the center of the screen. The next line waits for a fixed delay. Then the sound is played in blocking mode, which won't return until it is done. Next, the crosshair is removed and the questions are asked. The question, in this case, should consist of a word on which a lexical decision will be performed. Finally, the sound is deleted to free up the memory it uses.

Other Tools

Linger includes a few other useful packages.


Written by Doug Rohde
Copyright 2001-2003