RTbox: A device for highly accurate response time measurements

ZHONG-LIN LU 0 0 University of Southern California , Los Angeles, California AND Although computer keyboards and mice are frequently used in measuring response times (RTs), the accuracy of these measurements is quite low. Specialized RT collection devices must be used to obtain more accurate measurements. However, all the existing devices have some shortcomings. We have developed and implemented a new, commercially available device, the RTbox, for highly accurate RT measurements. The RTbox has its own microprocessor and high-resolution clock. It can record the identities and timing of button events with high accuracy, unaffected by potential timing uncertainty or biases during data transmission and processing in the host computer. It stores button events until the host computer chooses to retrieve them. The asynchronous storage greatly simplifies the design of user programs. The RTbox can also receive and record external signals as triggers and can measure RTs with respect to external events. The internal clock of the RTbox can be synchronized with the computer clock, so the device can be used without external triggers. A simple USB connection is sufficient to integrate the RTbox with any standard computer and operating system. - In behavioral studies, response times (RTs) provide valuable measures of human performance (Jastrow, 1890; Luce, 1991). Defined as the lapse of time between stimulus or task onset and a subjects response, RTs have been measured in a variety of tasks, ranging from simple visual and auditory detection (Arieh & Marks, 2008), choice reaction (Brown, Marley, Donkin, & Heathcote, 2008), and object recognition (Lu, Morrison, Hummel, & Holyoak, 2006) tasks to more complex lexical decision tasks (Yap, Balota, Cortese, & Watson, 2006). RT is also a key component in speedaccuracy trade-off paradigms (Dosher, 1976; Ratcliff & Smith, 2004). In this article, we describe the design, implementation, and test results of a new device, the RTbox, for accurate RT measurements in behavioral experiments. We focus on the most commonly used response mode, button responses. Accurate measurements of RTs require recording of two time stamps, the onset time of a stimulus or task and the time of a subjects response, with millisecond accuracy. It would be ideal that both time stamps are represented in the same time basethat is, according to the same clock. They should also be reliable, free of systematic bias and excessive measurement noise, and not affected by any timing jitter caused by standard computer hardware and operating systems. For an RT measurement device to be widely applicable, it must be (1) compatible with the standard, widely used computer hardware, such as Intel PC-compatible computers and Apple Macintosh computers; (2) compatible with the commonly used operating systems, including Microsoft Windows, Apple Mac OS X, and GNU/Linux; (3) easy to use and control from stimulus presentation software; and (4) easy to set up (only a minimal amount of work is required to integrate the device into common experimental setups). To reduce the impact of other component devices in experimental setups on RT measurements, it is also essential that the device can collect a subjects responses asynchronously: It should be capable of storing the timing of all the response events until it is convenient for the user software to retrieve them. The ability to detect and time stamp signals from external devicesfor example, transistortransistor logic trigger signals from a stimulus or data acquisition deviceis also a very useful feature. Before describing our RTbox, we review a few commonly used RT collection methods and devices. Computer keyboards and mice are widely used for RT measurements. These devices are cheap, do not require any special hardware setup or programming, and naturally work with any computer hardware, operating system, and standard software toolkit. However, the use of a keyboard and mouse often leads to huge variations and biases in RT measurements. Figure 1 shows a typical sequence of events involved in obtaining a buttonpress response with a regular computer keyboard. There are several sources of timing error in keyboard responses. 1. Mechanical lag. Typical keyboard and mouse buttons are implemented as electrical contacts that close and open an electronic circuit when the subject presses and releases a key or button. A keyboard encoder chip is used to register each closing of the electric circuit as a key-/buttonpress. Depending on the mechanics of the keys/buttons, there is some delay between key/button press/release and circuit switch on/off. This kind of delay cannot be eliminated totally, but some keyboards and mice may introduce unknown long delays. 2. Debouncing. Due to mechanical imperfections, button bouncingthat is, an electrical contact opening and closing the circuit multiple times in quick succession for a single key-/buttonpressoften happens. To eliminate the errors caused by button bouncing, the encoder chip performs debouncingtreating switch-on and switch-off events within a certain time window as a single event. Although the length of the time window and the exact implementation of the debouncing algorithm vary across keyboard models and manufacturers, a 20-msec delay due to debouncing is not uncommon. This introduces a systematic delay but does not affect relative RT measurements. Mechanical lag and debouncing delay Computer system gets event from encoder User code reads event from computer system Scanning bias: 520 msec Operating system, CPU load, and user programdependent variable delay 3. Scanning. To save manufacturing costs, there is not a dedicated electric connection from each key on a keyboard to an input of the encoder chip. Instead, the electrical contacts are arranged in a matrix form, with the switches placed at the intersections of the rows and columns in the matrix. A keyboard with up to 128 keys typically uses 24 connections on the chip and an 8 16 matrix. Pressing a key closes the electric circuit in one row and one column of the matrix. The keyboard encoder sequentially and periodically scans the connections in all the columns and rows and uses the state of the matrix to identify the key/button pressed/released. Keys connected to different rows and columns are checked at different times in the cycle. Depending on the keyboard encoder, the scan cycle typically takes 520 msec and may introduce a 5- to 20-msec timing bias among keys (Shimizu, 2002) if multiple keys at different rows/columns of the matrix are used in an experiment. A mouse typically has only one to three buttons, so it is not vulnerable to scanning bias. However, some mice may have up to a 70-msec delay and variations due to other design issues (Plant, Hammond, & Whitehouse, 2003). 4. Polling. Mice and keyboards are usually connected to their host computers via the USB using the USBHID protocolan inexpensive, flexible, and standardized communicati (...truncated)