FAB

A dummy’s program for self-paced forward and backward reading.

FAB is designed to run psychological or linguistic tasks that allow participants to go through sequential stimuli at their own paces, with reaction time recorded. The primary goal of FAB is to provide a free and handy instrument that simulates the eye-tracking process to explore the mechanism of human cognition. At the same time, the flexibility of FAB also enables it to carry out other psychological experiments (e.g., self-paced memorization, self-paced study, Stroop, IAT, mental rotation, perceptual simulation, and stimuli display for developmental experiments).

Based on website web languages (JavaScript, HTML, and CSS), FAB does not require you to install any special software on your own computer or on computers used to run experiments. All you need is a common web browser (e.g., Chrome, Firefox, etc). Both task generation and data collection run locally by default, so the Internet connection is not required. Reach out for any question: gongtiw@gmail.com

FABdesign (set up a task)

For designing your self-paced reading task, please open FABdesign.html with the browser.

Part 1: Create Your Parameter File

Step1: The Stimuli Setting

Please choose the type of stimuli.

Choose picture if you only use pictures in the experiments, otherwise choose the text. The difference between the two is that text will trigger two extra divisions to adjust the text layout and the segmentation pattern.

Please choose background color for your screen.

The default color is RGB 192,192,192.

Please set a layout for your text stimuli (only for text stimuli option).

Adjust the color, font size, font styles for the text. The stimulus layout will be identical to the example layout here. The font size depends on the screen resolution ratio and whether you zoom with the browser, so please make sure not to zoom the browser now and at least try on the lab computer once before collecting data .

Please choose how the stimuli should display. Differences between two display patterns are shown in the picture. Usually we choose Moving window option for linguistic experiments.

Please choose your segmentation rule (only for text stimuli option).

There are three ways to segment the stimuli sequences. You should choose here and then reform your stimulus list correspondingly later.

• by space: for languages that use space to separate words or picture stimuli. If you write sentence nice to meet you., it will display as nice, to, meet, you.
• by character: for character-based languages. If you write sentence 很高兴认识你。, it will display 很, 高, 兴, 认, 识, 你。 The punctuations 。, ， , 、, ？, ！, ：, ； will be displayed together with the last character rather than independently.
• by '|', customized: for other segmentation rule.

Step 2: The Keyboard Setting

Do you allow participants to go back during the learning process?

If you choose yes, participants will be able to go back and forth during the task, like an eye-tracking process. Data would be richer but also more complicated. If you choose no, participants can never look at the previous words/pictures once they press the forward key.

Please press and save the response keys.

After considering the forward/backward response key, you can press it, check its information in the key cube(at lower right), and then click save it as forward key or save it as backward key to save it. If you want to make a change, you can press another key and click the button again. The keyboard information is compatible since different browsers use the same key code, that is, you are confident to tell participants to press say "J " key, whenever they are using IOS or Windows, Safari or Chrome, etc. However, the physical positions might differ among keyboards.

Step 3: Instruction

Do you want a special layout for instruction?

If you choose yes, you can set up a special layout for the instruction part. The instruction content will be designed in the CSV file together with your stimuli.

Step 4: Comprehension Questions and Feedback

Do you have any comprehension question for participants?

Choose yes if you have comprehension questions for some/all stimuli. The questions will be listed in the CSV file but here you can set up how participants answer those questions.

• Yes, and they are multichoice questions: for like yes/no questions. Four answer levels at most could be set up. The keyboard setting procedure is identical to forward/backward keys. Forward/backward keys chosen above are allowed to be also set up as answer responses. You can indicate the labels for different answers in the dataset. If you want to record accuracy or give feedback to participants, the answer labels set here should be identical to the labels in fab_key column in CSV (see details later).
• Yes, and they are text-input questions: for all more complex questions. Participants will see the question followed by a textarea input. Once finishing the answer, they can click the submit button to continue.
• No: if you choose this option, this step is finished. You do not need to indicate anything about the comprehension question in CSV as well.

Do you want a special layout for comprehension questions?

You can set the special layout for comprehension questions, or the layout would be identical to the stimuli's layout.

The questions will appear __ ms later after the final windows disappear.

Usually the comprehension question should appear immediately after the end of the stimuli. However, the function here not only facilitates complex linguistic experimental setting, but also enables many other psychological experiments. If you want to give different delays for different trials, you can set up an extra column fab_qwait in your CSV (see details later).

Do you want to give the feedback to participants' answer?

You can give feedback to participants. After receiving the feedback, participants need to press the forward key to continue (which should be indicated in the instruction). FAB will record the answer as correct only when it exactly matches the correct answer. Therefore, feedback for text-input answers is allowed but not recommended. The font and size of feedback will be the same as the questions, and an italic style is added. The feedback color will be blue (RGB 0,0,255) for positive and red (RGB 255,0,0) for negative.

If you want to give feedback only for some special trials (for practical trials), you can choose yes here and then set up an extra column fab_feedback in CSV file (see details later).

Step 5: The "End of Experiment" Message

Here you can input the message you want to show when participants finish the task. All other instructions or breaks could be set later in the CSV file (see details later).

Step 6: Download the Parameter File

Now you have finished the first part and can get the first “card” for FAB.

• Click saving the experimental parameter file and an exp.js file will be saved in the default download folder of your browser.
• Move exp.js to jsfile folder under FABrunning.

IMPORTANT: The file name should be exact exp.js, not exp(1).js, exp(2).js or anything else. If so, please rename it as exp.js.

Part 2: Create Your Stimulus File

Step 1: Create and Check Your CSV File

(you can get more examples of stimuli lists at the templetes/ALLSTIMULI folder)

Since almost all psycholinguistic researchers write their stimuli in spreadsheet software (e.g., Excel) firstly before conducting the experiment, FAB will just use sheets to make stimulus files. Here is one example of a valid stimulus list:

• fab_stimulus necessary: Add / at the beginning and the end of the stimuli (You can do it instantly, check here). Add * to separate stimuli into different regions, then FAB will support region analysis. If you want to add a fixation + at the beginning of sentences, but don't want participants to go back to the fixation later (see the demo above), you can rewrite the sentence as /+/The old lady lived further away from the city center./. Namely, / can help separate stimuli into different loops, where participants can only go back-and-forth within the loop.

• fab_question: You can add one question for each stimulus, which could be displayed as texts or a picture. If you want to present multiple questions for one stimulus, add blank rows below that stimulus and only type in your second, third... questions.

• fab_key: The correct answer to comprehension questions. This column is necessary if you want the accuracy data or feedback presentation.

• fab_feedback: A column for customizing which trials to present to feedback.

• fab_instruction: You can add picture (recommended) or text instruction before each stimulus. Participants will click a next button to continue. Scrolling is allowed so the content could be long.

• fab_qwait: A column for customizing the delay between stimuli and comprehension questions.

• list, block, condition: You can add other columns. The information will not be used in FAB but will appear in the dataset, which may benefit data analysis.

Notes for Your CSV File

Only fab_stimuli is necessary for successfully running the experimetal program.

All keywords starting with fab_ are case-sensitive so make sure you give the right names.

There is another important thing to check before exporting the CSV. Since CSV interpret the English comma as a default delimiter, we cannot include any English comma in the cells, or the delimiter process would go wrong. This sounds ridiculous for linguistic tasks but there is really an easy way to solve it. Just find all , and replace them with &#44, a code for the English comma in HTML. Participants will see commas as usual in the experiment.

Step 2: Compile Your CSV File

• Export the sheet as CSV in your spreadsheet software.
• Upload the CSV by click Choose File.
• After uploading (choosing) CSV, you can get the sti.js file by click save the experimental stimulus file.
• Put sti.js on jsfile folder under FABrunning(just as exp.js).

Step 3: Prepare Other Picture Files

If you indicate any pictures (jpg, bmp, png) in instruction or stimuli, please put them into the media folder under under FABrunning.

FABrunning (run the task)

For running your self-paced reading task, open FABrunning.html with the browser.

• enter the subject ID
• FAB will record the time participants spend on each window.
• After the experiment, the data file data_subjectID.csv will be saved in the default download folder of your browser.

Tips:

• If you want to make it fullscreen, open the fullscreen mode according to your browser.
• Terminate and save the incomplete data: Shift+q

FABanalysis (analyze data)

For cleaning data and generating eye-tracking like indices, open FABanalysis.html with the browser.

• Choose data files of all participants by clicking Choose Files at the top of the website. The column names should be identical for all files.
• Click the indices you want to generate.
• Input the reading-time threshold to remove outliers.
• Click Save the data processing file to save the dataset.

The definition of indices are similar to eye-tracking tasks. Here we use an example to describe them.

Window-based Indices

• Gazes is defined as the single dwelling time in any given window. In this demo, we set the outlier as gazes below 80 ms or beyond 5000 ms. We firstly took a look at the indices in the window level.
• First pass time refers to the first gaze on the window (e.g., 344 ms for “kids”, 1361 ms for “felt”).
• Rereading time refers to the sum of the duration of second and later gazes on the window (e.g., 485+207 = 692 ms for “kids”, 1361+863 = 2224 ms for “felt”).
• Total reading time refers to the sum of the duration of all gazes on the window (e.g., 344 + 485 + 207 = 1036 ms for “kids” and 1361+863 = 2224 ms for “felt”). Total reading time equals the addition of first-pass reading time and rereading time.

Path-based Indices

• Number of regression out refers to how many times the comprehender go back to the previous window from the present window (e.g., 1 for “kids”, 1 for “felt”). In the second line of gaze sequences, the gazes of “showed”, “all”, “the” were outliers. Therefore, the regression path should be regarded as from “dances” to “kids” and from “kids” to “older”. Accordingly, the number of regression out was zero for “show”, “all”, “the”.
• Number of regression in refers to how many times the comprehender go back in the present window from the later window (e.g., 1 for “kids”, 0 for “felt”). Similarly, the number of regression in “showed”, “all”, “the” was zero.
• Probability of regression out refers to the quotient of the number of regression out and the total gaze numbers of the present window (e.g., 0.33 for “kids”, 0.5 for “felt”).
• Probability of regression in refers to the quotient of the number of regression in and the total gaze numbers of the present window (e.g., 0.33 for “kids”, 0 for “felt”). Since comprehenders cannot regress out from the first window, or regress into the final window, the number/probability of regression out for the first window and the number/probability of regression in for the final window is always zero.
• Regression path duration is also called “go-past” time, referring to the time from first fixating the window to first moving past the window to the right, including time spent on rereading earlier windows. For “felt” in the example sentence, the regression path duration was 1361+499+485+712+207+287+440+663+704+863 = 6221. Since the comprehender of this example moved directly forward after the first gaze for other windows, the regression path duration for other windows was equal to the first pass time.
• Selected regression path duration was regression path duration detached from time spent in rereading earlier windows (1361+863= 2224 for “felt”, and first pass time for all other windows). Since the regression path does not exist for the first window, the regression path duration and selected regression path duration will always equal the first pass time for the first window.

Exploratory Measures

The previous measures were adopted from global eye-tracking measures. Besides, FAB also provides two exploratory measures.

• Number of gazes refers to how many valid gazes there are for the present window (e.g., 3 for “kids”, 2 for “felt”).
• Mean gaze duration refers to the average duration of all gazes in the present window, equal to the total reading time divided by the number of gazes (e.g., 1036/3=345 ms for “kids”, 2224/2= 1112 ms for “felt”).

Area-level Indices

In order to conduct analysis on the area level, FAB would remove the gaze outliers and then merge all gazes by areas as shown in the figure. After that, FAB treats areas as windows and calculates all indices following the same ways mentioned above.

FAQ

Why my Fabrunning.html crushes?

check the following things for debugging 🐛

• The separating notion for fab_stimuli is / not \\.
• Remember to remove English commas in CSV stimulus list (see here).
• Remember to use HTML code to replace other symbols on your language materials/questions/instructions that could have conflicts with FAB's inner setting
  • all ,should be replaced with &#44 (because , would be used as CSV delimiter/separator)
  • all * should be replaced with &#42 (because * would be used to mark interest areas)
  • all / should be replaced with &#47 (because / would be used to mark reading areas)

What browser should I use to run FAB

FAB is ideally supported by browsers that provide web developer tools, like Chrome, Firefox, Safari, Edge. You could recommend participants open FABrunning.html with one of these browsers for online experiments. This is a very common requirement in online experiments with customized websites (Gureckis et al., 2016).

A good thing you can do is to always test your programs on different browsers before releasing them.

How can I link the picture stimuli/instructions/questions to the stimulus list?

• For instructions, you can input the picture name (e.g., ins1.jpg) in the cell directly.
• For stimuli, you can write in this way: /ram.jpg goat.jpg beetle.jpg/ or /There is a picture1.jpg in my room./ (and choose by space as segmentation in FABdesign.html)
• For questions, you can input the question name (e.g., question1.jpg) in the cell directly. If you want to combine words and pictures in your questions, you need to add HTML code, but that is easy, just write in this way: Did you see this object in the previous sentence?<br><img src='media/picture1.jpg'> (<br> is for line breaking).

Could FAB randomize my trial order?

FAB does not have the function to randomize the presentation of your trial, for you might have constraints that the automatic procedure cannot satisfy (e.g., three trials from the same condition cannot be present continuously; there should not be continuous four "yes" answers for the comprehension questions). Therefore, the best way for you is to make different randomized stimulus lists yourself, and add one list column to indicate that.

Could FAB display multiple pages of instructions/questions?

Yes. For multiple pages of instructions, just add one blank row, input things on fab_instruction cell, and leave other cells empty. See Examples in templetes/ALLSTIMULI/causalRelatedness. Similarly, if you want to ask multiple questions for the same stimulus, you can add several blank rows below that stimulus and input questions for the fab_question cells while leaving other cells empty.

How can I use FAB for online experiments?

So far FAB does not have its own web server. We recommend researchers to use any form tool (e.g., Qualtrics, Google Forms) + any online storage tool (e.g., Google Drive, One Drive; it should be chosen according to your privacy concerns) for online experiments at the current stage. Taking Google Drive and Google Forms for example:

1. Compress the procedure folder and upload it to Google Drive , then you can get the link to the procedure and share it with participants afterward.
2. Create a Google Forms, add content, demographic questions, etc. Paste the google drive link of your procedure for participants to download. And finally add one file-upload question for participants to send their data back.

How precision is the timing in FAB?

The timing precision of FAB should be close to a range of online experiment tools, espcially jsPsych. Here is a great paper that compared the precision of different online and laboratory tools. For response recording, both FAB and jsPsych use the addEventListener("keydown") function. For stimulus updating, jsPsych rewrites HTML while FAB just toggles the visibility of different divisions, which is regarded as faster than rewriting HTML.