User guide for the Ocular Dominance Experiment 2.21


By Malcolm Handley, Michael Bevin, and Robert O’Shea.

Download the complete document [160 kB pdf file].

1    Introduction

1.1    The philosophy and the approach

1.1.1    What ODE does

The Ocular Dominance Experiment program (ODE) displays stimuli on two monitor screens and records the user’s responses via key presses. We designed ODE to run experiments in binocular rivalry; the two screens would normally be viewed through a stereoscope. Experiments comprise one or more trials.

We designed ODE to work closely with StatView and with BBEdit Lite. StatView is a statistics program that can be used to analyse results from ODE (see 1.5). BBEdit Lite is a text (ASCII) editor. All input files for, and output files from, ODE are text files. All of these files have the same first name, differing in their extensions.

1.1.2    How ODE does it

We specify all the experimental conditions that stay the same over an experiment in a parameter file (e.g., expname.param). We specify all the experimental conditions that vary within an experiment in a trials file (e.g., expname.trials).

ODE saves results into a results file (e.g., expname.results) that can be imported effortlessly into StatView for analysis. ODE generates a counter file (e.g., expname.counter) in which it keeps track of which trials the observer has completed. This means the observer can take a break in the middle of a block of trials, then come back and complete them, even if the computer has been shut down in the meantime. ODE can run two main types of trials: rivalry or pseudorivalry trials and probe trials.

1.1.3    Rivalry and pseudorivalry trials

In an example rivalry trial, ODE could display for one minute a red vertical grating and a blue horizontal grating on one screen and a green disk on the other. The user could press three response keys to indicate which stimuli were visible. Another form of rivalry trial is a pseudorivalry trial, during which, for example, one stimulus (say a vertical grating) appears for a time on both screens, and is then replaced by another (say a horizontal grating), which is in turn replaced by the first, and so on, according to a prearranged sequence (contained in a contrast-control file). The observer in this case could press two keys to report visibility of the stimuli. Rivalry and pseudorivalry trials can be intermixed in the one experiment. In both cases, ODE stores summary statistics about the user’s key presses (total duration, total number of presses, and average press duration). As well, the program can create files describing exactly which keys the user was pressing at any time during a trial (key-press record files); these files can then be used to make contrast-control files for pseudorivalry sequences.

1.1.4    Probe trials

In an example probe trial, ODE can be made to show rival stimuli, and then introduce one or more instances of a probe stimulus during a trial to which the observer responds with a key-press. In its simplest form, the probe experiment could wait until the observer presses a single key to signal that he or she is ready, then, after a random time, show a spot to which the observer releases the key. ODE records the time between each onset of the probe and the release of the key. Probe and rivalry trials cannot be mixed in the one experiment.

1.1.5    General experimental methods: Rivalry trials

For rivalry trials, we normally have a trial duration of 60 seconds, and an inter-trial interval (ITI) of at least 45 seconds (meaning that, once time spent getting ready for the next trial is added in, the elapsed time is more like 60 seconds). In some of our experiments with complicated designs, one block might contain more than 50 trials, so to do 50 trials in one session would take 2 x 50 = 100 minutes. We recommend, however, that observers limit each session to 60 minutes, and to have about 24 hours’ rest between sessions. So two blocks of such an experiment (2 x 100 minutes) would take about four sessions to complete, one each for four days.

We normally construct our trials files so they contain one run through the design, with one trial per cell. This means we cannot statistically test our design competely until we have run at least two blocks of trials (i.e., e1JW1.trials, and e1JW2.trials). Usually we will do at least four replications of all crucial trials in an experiment. (Remember that each different trial file will require its own parameter file: i.e., e1JW1.params, and e1JW2.params.)

1.1.6    General experimental methods: Probe trials

For probe trials, the trial duration is determined by how long the observer takes to respond to all the probes. We normally present 25 probes in each trial. We use a hazard function for the random foreperiod, with the earliest probes appearing 500 ms after the observer’s key press, and the latest probes appearing 7000 ms after the observer’s key press. We normally use a key release for the observer’s response. We normally set the minimum delay between a response and the ready signal for a new probe at 1000 ms.
Back to Robert O'Shea's laboratory facilities.

This page was created on 20 Mar 2003.
This page was last updated on 10 Oct 2005.