` Fitts Law Dataset


Fitts Law Dataset

The Fitts' Law is one of the well studied models for computer-human interaction in the HCI community. It determines the human motor performance using the Shanon Information Entropy. In the 50s Paul Fitts studied the expected time required for human motor system to reach an object from a starting point. He realized that the time to reach the target is a function of the distance to the object and the width of the object (in the 1D model). Scott MacKenzie later proposed a model that explained this behavior based on the information theory. Today the classical Fitts' model is known by the following equation, in which A is the distance to the target and W is the width. Parameters a and b depend on the environment and the pointing device and can be determined experimentally.


Our Research

We have been studying how people interact with computers and we are interested to see how human motor system improves over time. Our dataset is freely available for research use when properly acknowledged.

This dataset contains data for two sets of experiments. An in-lab experiment and a web-based experiment. Each one of the experiments consists of data for two conditions: Homogeneous Cursor Motion (Fixed Rectangles) in which the participant repeatedly clicks back and forth between two vertical rectangles of fixed width and amplitude. After 11 repetitions, the width and amplitude are changed. The second condition is called Heterogeneous Cursor Motion (Variable Circles) in which the participant clicks on target circles of varying diameter and location.

The table below shows the values of A and W for each condition

Controlled User Study Dataset

For this controlled experiment, we collected a total of 22,540 timing measurements (11,040 for homogenous targets and 11,500 for heterogenous targets). We cleaned this raw dataset by keeping only timing measurements for cases where the subject successfully clicks on all presented targets within a "reasonable" time period (within 3 std dev of the global mean time). Our goal is to remove most cases where subjects were distracted or decided not to complete the experiment. After cleaning, the dataset contains 16,170 valid timing measurements (8,250 for homogenous targets and 7,920 for heterogenous targets).

Uncontrolled User Study Dataset

To conduct the uncontrolled study, we made the same applet available online and advertised by emails and postings on user groups. The experimental applet and datasets are available online at http://www.tele-actor.net/fitts/. To comply with our Human Subjects approval, each online participant is asked to click an online consent box before starting the applet. An entry is created in the server database each time the consent box is clicked. We do not record IP addresses and cannot determine if the same person runs the experiment multiple times so we do not know the number of unique participants. We ask online visitors to indicate the type of mouse device they use (trackpad, mouse, trackball, etc), but cannot verify these responses.

The online applet presents visitors with 24 homogenous targets and 25 heterogenous targets and thus collects up to 49 timing measurements. Unlike the controlled experiment, online visitors were not asked to repeat each experiment 10 times. (The online applet includes a third experiment with variable-sized rectangular targets; we discovered a timing error in that experiment so we do not use data from that experiment.)

We cannot determine how many visits are repeats (by the same person). We collected timing data from 2,689 visits to the homogeneous target experiment and 2,811 visits to the heterogenous target experiment. As we did in the controlled study, we cleaned the raw dataset by keeping only timing measurements for cases where the subject successfully clicks on all presented targets within a "reasonable" time period (within 3 std dev of the global mean time). Our goal is to remove most cases where subjects were distracted or decided not to complete the experiment. After cleaning, the online study dataset includes 78,410 valid timing measurements (39,360 for the homogeneous targets and 39,050 for the heterogenous targets).

Below is a screenshot of the Excel spreadsheet for the web-based experiment. There are two separate sheets for the heterogeneous and homogeneous conditions. The first two rows present the width and the distance to the target. After the twelfth row, each row presents a participant with their pointing device (if known) in the first column. Different columns present the time (T) in miliseconds for a participant to finish each trial.

As a courtesy, if you use the data, we would appreciate knowing your name, what research group you are in, and the publications that may result.


         Fitts Law Dataset


Dedicated to Professor E.R.F.W."Ted" Crossman, 1925-2001

For technical questions please contact Siamak Faridani <faridani at berkeley dot edu>

For further information please contact:

Ken Goldberg
goldberg at berkeley dot edu
Prof of IEOR and EECS
4135 Etcheverry Hall
University of California
Berkeley, CA 94720-1777
(510) 643-9565 (phone)
(510) 642-1403 (fax)