In simple terms

The AX-CPT is a task where people see two letters in a row, first a cue and then a target (or also known as "probe"). The task was originally designed by Rosvold and his colleagues at Yale University.

In this task people must press a target button only when they see A followed by X. All other pairs (e.g., A–Y, B–X, B–Y) require a different or no response. It tests how well people use context (the first letter) to decide how to respond to the second.

Research history

The AX-CPT was designed in the 1950s and the first paper was by Rosvold and colleagues in 1956. In that time, they did have a special apparatus for letter presentation.

They called it simply the Continuous Performance Test (CPT). It was designed to detect brain damage.

Since then, the AX-CPT has been used a lot in different braches of cognitive research, including fMRI research, cognitive psychology, and neuropsychology, and neuroscience.

Polizetti and colleagues (2018) showed the AX-CPT performance across different ages (8 to 22 years old) using a large cross-sectional study. They have many interesting conclusions, such as: "Starting from adolescence, early developing resources appear to become increasingly integrated into optimized performance and it is only in early adulthood that a flexible and continually updated representation of task relevant information mature.".

Versions

This task has been used a lot and there are now many different variations. Apart from trivial things (such as the colors of stimuli), examples of important things that researchers varied:

  1. Whether there is a keyboard response to acknowledge the cue. This can help participants to engage/attend the cue better.

  2. The delay (time) between the cue and the probe, ranging from 0 to 6 seconds.

  3. The allowed response time window (how long do people get to respond before it is considered a timeout error).

  4. Type of stimuli. Instead of letters, researchers can use non-linguistic stimulus, such as shapes, faces, dot-patterns. This can be good for working with different species (e.g., animal research), ages (e.g., small children), cultural contexts (e.g., where the Latin alphabet is not common).

In the current version, we use letters, and we use a cue response as well.

Task design

In this task, there are multiple trials (typically more than 100) in which the event sequence looks like this (although there are variations).

  1. Show white fixation point on black screen

  2. Show a cue letter in cyan

  3. Optional: Wait for a response to cue (max time 1000 ms)

  4. Show a target letter in white

  5. Wait for a response to target (max time 1000 ms)

This type of trial comes in four different conditions. In the current implementation we use a response to the cue, but you can decide not to use that.

Trial type / condition The cue is The target is The response to the Cue The response to the target

A-X

The letter A

The letter X

M

C

B-X

Any letter other than A

The letter X

M

M

A-Y

The letter A

Any letter other than X

M

M

B-Y

Any letter other than A

Any letter other than X

M

M
A response to the cue is not necessary per se. It is always the same across conditions. The benefit is that it makes sure people are attending the task and it might increase people’s engagement.

About our implementation

In this task, we have a practise block of up to 50 trials. Participants need to do at least 10 trials in a row correctly to complete practise or up to 50 trials if they do not reach this. This to ensure people are used to the rules.

Because we use the letters M and C for responses, these are not used as stimulus letters. As described by Servan-Schreiber and colleagues, all other letters except the K (due to its similarity to X) were used.

Run the demo

Click here to run a demo

Data output file

You do not need this information, unless you want to understand the output data file. You can ignore this if you just want to find out your own score. This is only necessary if you want to carry out the experiment with multiple participants.
In PsyToolkit, the data output file is simply a textfile. The save line of the PsyToolkit experiment script determines what is being saved in the data output file. Typically, for each experimental trial, you would have exactly one line in your text file, and each number/word on that line gives you the information you need for your data analysis, such as the condition, response speed, and whether an error was made.

Meaning of the columns in the output datafile. You need this information for your data analysis.

Colum Meaning

1

The name of the block

2

Trial type (AX,AY,BX,BY)

3

The status of the trial (1=correct trial, 0=mistake was made)

4

RT to the the cue

5

status of the response to the cue (1=correct, 2=wrong key, 3=timeout)

6

RT to the target

7

status of the response to the target (1=correct, 2=wrong key, 3=timeout)

8

correct count in a row so far

How to analyze data from experiments

When you embed a AX-CPT in your survey, you need (as always), make sure you set the "analyze" parameters of your experiment. In this case, it is very simple, for the dependent variable you just use value 6. For correct, you choose column 5. For condition you use column 2

This way, you get the response time for each of the four conditions. You can do more sophisticated analyses. For this we recommend the R software.

Check out the source code

The PsyToolkit code

Download

The PsyToolkit code zip file, including stimulus file

If you have a PsyToolkit account, you can upload the zipfile directly to your PsyToolkit account.

If you want to upload the zipfile into your PsyToolkit account, make sure the file is not automatically uncompressed (some browsers, especially Mac Safari, by default uncompress zip files). Read here how to easily deal with this.

Further reading

  • Rosvold, H. E., Mirsky, A. F., Sarason, I., Bransome Jr, E. D., & Beck, L. H. (1956). A continuous performance test of brain damage. Journal of consulting psychology, 20(5), 343. Link to journal (behind paywall)

  • Polizzotto, N. R., Hill-Jarrett, T., Walker, C., & Cho, R. Y. (2018). Normal development of context processing using the AXCPT paradigm. Plos one, 13(5), e0197812. Open access link