I have done my best to optimize the reaction time scripts that are offered here (IAT script and the more general RT script) to make them run smoothly and with as little error as possible. However, Javascript has a couple fundamental limitations with how it runs timed events. First, its internal clock can be off by about +/-15ms in some browsers (info). This can lead to a higher error rate for some browsers. Second, the processing speed of computers may cause events to take longer than they should (i.e. a 20ms prime may take 25ms instead). This is inescapable, as Javascript events must “wait in line” alongside other events for their turn to be processed by computer.

Please remember that these RT scripts are not suitable for all contexts. In some cases, these issues may be inconsequential, but it may matter drastically for other studies. It is convenient to be able to gather RT data from an online sample, but by using these scripts you cede a fair degree of control over the equipment and environment involved. Please use your best judgment and consider adopting some of the solutions below.

Methods to Compensate for Javascript Limitations

1. Forward and/or backward mask primes. Internet Explorer versions 8 and below are especially poor at running timed events in the time they are supposed to run (info). Simple primes meant to be subliminal may thus end up not actually being subliminal. Masking primes can help to further disguise primes. To set up a forward/backward mask, simply add an extra "prime" to a given word. Set a delay of 0 between the mask and the prime.
2. Use the position and colour parameters for primes. Another option for keeping primes subliminal is to place them in the parafoveal field, outside the focus of attention. My script offers the ability to place primes (and words) above, below, to the left, or to the right of the centre of the screen. It also allows you to change the colours, so another option is to set primes to a grey colour.
3. Ask participants if they could identify primes. A couple questions can probe whether participants noticed primes, and if so, whether they could identify the words. This can be useful in pilot testing as well as full studies.
4. Control for browser and operating system. There may be numerous other factors that could influence the error rate of a RT task, but browser and operating system can have a large influence, and are easy to ask participants about. Common browsers: Internet Explorer, Firefox, Chrome, Safari. Common operating systems: Windows, Macintosh.
5. Replicate the study in a lab setting. For IATs or other RT tasks, it may be useful to replicate the study in a lab environment, where there is more control over the equipment and environment. If you are worried about possible error in the RT data of about +/-15ms, consider relying on online RT data for more exploratory purposes.