ArticlePDF Available

Measuring Software Timing Errors in the Presentation of Visual Stimuli in Cognitive Neuroscience Experiments

Article

Measuring Software Timing Errors in the Presentation of Visual Stimuli in Cognitive Neuroscience Experiments

Abstract and Figures

Because of the features provided by an abundance of specialized experimental software packages, personal computers have become prominent and powerful tools in cognitive research. Most of these programs have mechanisms to control the precision and accuracy with which visual stimuli are presented as well as the response times. However, external factors, often related to the technology used to display the visual information, can have a noticeable impact on the actual performance and may be easily overlooked by researchers. The aim of this study is to measure the precision and accuracy of the timing mechanisms of some of the most popular software packages used in a typical laboratory scenario in order to assess whether presentation times configured by researchers do not differ from measured times more than what is expected due to the hardware limitations. Despite the apparent precision and accuracy of the results, important issues related to timing setups in the presentation of visual stimuli were found, and they should be taken into account by researchers in their experiments.
Content may be subject to copyright.
Measuring Software Timing Errors in the Presentation of
Visual Stimuli in Cognitive Neuroscience Experiments
Pablo Garaizar
1
*, Miguel A. Vadillo
2
, Diego Lo
´pez-de-Ipin
˜a
1
, Helena Matute
3
1Deusto Institute of Technology, DeustoTech, Universidad de Deusto, Bilbao, Spain, 2Cognitive, Perceptual and Brain Sciences, University College London, London,
United Kingdom, 3Faculty of Psychology and Education, Universidad de Deusto, Bilbao, Spain
Abstract
Because of the features provided by an abundance of specialized experimental software packages, personal computers
have become prominent and powerful tools in cognitive research. Most of these programs have mechanisms to control the
precision and accuracy with which visual stimuli are presented as well as the response times. However, external factors,
often related to the technology used to display the visual information, can have a noticeable impact on the actual
performance and may be easily overlooked by researchers. The aim of this study is to measure the precision and accuracy of
the timing mechanisms of some of the most popular software packages used in a typical laboratory scenario in order to
assess whether presentation times configured by researchers do not differ from measured times more than what is
expected due to the hardware limitations. Despite the apparent precision and accuracy of the results, important issues
related to timing setups in the presentation of visual stimuli were found, and they should be taken into account by
researchers in their experiments.
Citation: Garaizar P, Vadillo MA, Lo
´pez-de-Ipin
˜a D, Matute H (2014) Measuring Software Timing Errors in the Presentation of Visual Stimuli in Cognitive
Neuroscience Experiments. PLoS ONE 9(1): e85108. doi:10.1371/journal.pone.0085108
Editor: Suliann Ben Hamed, Centre de Neuroscience Cognitive, France
Received September 19, 2013; Accepted December 1, 2013; Published January 7, 2014
Copyright: ß2014 Garaizar et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Support for this research
was provided by Grant PSI2011-26965 from Direccio
´n General de Investigacio
´n of the Spanish Government and Grant IT363-10 from the Basque Government.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: garaizar@deusto.es
Introduction
Since the days of Wilhelm Wundt, experimental psychologists
have studied the temporal dynamics of the cognitive processes
involved in perception and attention. Experimental apparatus that
allow researchers to present visual stimuli for very short periods of
time, such as tachistoscopes, quickly became popular in most
experimental psychology laboratories. Since the 1980s, these
complex and delicate devices have been replaced by personal
computers and standard monitors, which are cheaper and easier to
use. Moreover, experimental psychologists and cognitive neuro-
scientists now have several software alternatives that accurately
present visual stimuli [1]. However, most of these tools run on
standard personal computers (PCs), and this type of hardware is
not optimized for the accurate presentation of visual stimuli. There
are many limitations related to stimuli presentation durations
caused by the underlying technologies of CRT and LCD displays
and the timing mechanisms provided by non-real-time operating
systems (e.g., Microsoft Windows on a PC). Both factors can have
a significant impact on the accuracy and precision of the visual
stimuli presentation, particularly in experimental paradigms that
have very short Stimulus Onset Asynchrony (SOA) [2] [3] [4].
The goal of this research was to assess the potential
discrepancies between the timing conditions defined by the
researchers when using a subset of specialized software for the
presentation of visual stimuli on common-use hardware and
software (i.e., low-refresh-rate displays and non-real-time operat-
ing systems) and the actual onset and offset times of the visual
stimuli detected using external photodetectors. Assessing the size of
these potential timing errors is a first and necessary step in order to
minimize and correct them. Whenever possible, we will also
describe the way in which we were able to compensate these
errors.
CRT and LCD Displays
The widespread success of displays based on the cathode ray
tube (CRT) established this type of monitor as the standard in
computer-based experimental laboratories. A CRT monitor has
an electron gun at the back that points to a glass screen covered
with phosphors located at the front of the monitor. To display an
image, the electron gun points to the top left corner of the screen
and shoots a beam at the phosphors, briefly lighting them up.
Once the image displaying process has started, the gun moves
rapidly, and the beam energizes the entire screen, line by line
down to the lower right corner. Once there, the electron gun turns
off, points again to the top left corner, and the painting process
starts again. The last step is controlled by the VSYNC (vertical
synchronization) signal, and the refresh rate of the CRT monitor
depends on how fast this signal is triggered (see Figure 1). The
period between two VSYNC signals is known as a ‘‘tick,’’ and its
duration depends on the refresh rate (e.g., 16.667 ms at 60 Hz and
10 ms at 100 Hz).
It is important to notice that the luminance of each dot of the
image displayed by a CRT monitor is not constant but dependent
on two factors: the refresh rate (i.e., the time the electron gun
needs to energize it again) and the decay time, which depends on
the phosphor type (1.5 to 6 ms for the frequently used P-22
phosphor). Therefore, researchers conducting experiments with
CRT monitors should not assume a rectangular-shaped signal of
PLOS ONE | www.plosone.org 1 January 2014 | Volume 9 | Issue 1 | e85108
the configured number of ticks of duration in their visual stimuli
presentations, but a set of discrete pulses separated by the duration
of a tick and the decay time.
Despite the continuous advances in liquid crystal display (LCD)
technology, CRT monitors have been considered more suitable for
the precise and accurate presentation of visual stimuli [5]. Their
higher frame rates (typically around 85–100 Hz, but some models
can achieve 240 Hz), greater number of frames per second (FPS),
and shorter rise and decay times enable abrupt changes from one
frame to another.
Nevertheless, because of their small size and low weight, LCDs
are becoming more popular. Moreover, they offer some charac-
teristics that increase visibility, such as the absence of flickering
when displaying static images, and the poor temporal character-
istics typically associated with LCD screens are based on studies of
older LCD technology. Using current LCD displays, Lagroix,
Yanko, and Spalek [6] found photometric estimates of the rise
time that are far shorter (1–6 ms) than earlier estimates (20–
150 ms) and approach those of CRTs (,1 ms). In addition, Wang
and Nikolic [7] tested an inexpensive 120 Hz LCD display that
was shown to have timing and stability characteristics at least as
good as a CRT monitor. These findings suggest that LCD displays
are suitable for studies that require high accuracy in the timing of
visual presentations, but the truth is that both display technologies
are affected by different limitations that make them inappropriate
for brief stimuli presentations [8] [9].
The luminance of each dot of an LCD display is defined by the
orientation of the liquid crystal molecules placed between two
polarized glasses and two electrodes. To display an image, the
array of electrodes modifies the amount of visible light emitted
from the back of the display. Although this backlight is often
assumed to be constant, its luminance level is pulse-width
modulated. Instead of varying its intensity to adjust the luminance,
it is switched off for brief periods of time (the higher the amplitude
of the modulation, the lower the backlight luminance). The
backlight modulation frequencies are far from the critical flicker
frequency [10] and usually neglected, but might introduce
undesirable effects in the visualization of stimuli (e.g., steady-state
evoked visual potentials). The time needed to switch a dot of an
LCD display from one luminance level to another is known as
response time (RT). Instead of calculating the RT as the time
needed to switch from full black to white as suggested by ISO
9241–305, most manufacturers provide the RT as the time needed
to perform a grey-to-grey transition which is significantly shorter
due to Response Time Compensation (RTC) mechanisms. RTC,
also known as overdrive, is based on applying an over-voltage to
accelerate the orientation change of the liquid crystals. This
relation between luminance levels and response time in LCD
displays hinders synchronous presentations of stimuli formed by
different luminance levels [8].
Although there are significant differences between the under-
lying technologies of CRTs and LCDs, some mechanisms are
similar [11] and they are operated in the same way from the
operating system due to compatibility issues. Even in LCD
monitors, the refresh rate configurations affect the monitor’s
performance. In order to minimize flicker, tearing, and other
artifacts, most programs use a technique called double-buffer (or
multiple-buffer for cases with more than 2 buffers). While the
graphics card is rastering one buffer on the screen, the software is
preparing the next frame in the other buffer. At VSYNC signal,
both buffers are flipped, and the process starts again. Therefore,
different images cannot be shown on the screen at a rate faster
than the refresh rate.
Because CRT monitors are decreasing in popularity with
consumers, fewer are being produced, and it is becoming difficult
for experimental laboratories to find CRT monitors through
conventional supply channels. Some researchers fight this trend by
buying and storing CRT monitors for future use or purchasing
them through nonconventional supply channels (e.g., eBay) or
even repairing them when they malfunction. However, these
strategies do not seem viable in the long term. Therefore, we use a
computer equipped with an LCD display to perform tests of
software specialized for the presentation of visual stimuli.
General Purpose and Real-time Operating Systems
To evaluate the influence of the real-time nature of an operating
system in these experiments, some of the tests conducted in our
study were run under a real-time operating system (Linux 2.6.33–
29-realtime). It is important to recall how a real-time operating
system differs from a general-purpose operating system. Contrary
to popular belief, real-time operating systems are not necessarily
faster, but they are predictable. Their schedulers work like traffic
lights on a road. The schedulers’ main goal is not to maximize the
overall throughput but, rather, to meet the predictability and
fairness requirements. By contrast, the goal of general-purpose
operating systems is usually the opposite. Throughput is
maximized, even if background tasks are starved for long periods.
As a result of their scheduling policies, the length of these periods is
unpredictable. These non-real-time scheduling policies manage
multitasking concurrency similar to a yield sign on a road (i.e., no
task will wait unnecessarily if it can be attended to instantly, but
the maximum running delay for a task when the system is
overloaded is unpredictable). In general, a real-time and low-
latency operating system is recommended for experiments that
have high accuracy requirements. Unfortunately, nearly every
specialized software package used in experimental psychology and
neuroscience research runs on a general-purpose operating system
(e.g., Microsoft Windows) and might be affected by the
unpredictability of the task scheduling process. The good news is
that specialized software eliminates eventual delays by implement-
ing several optimizations (e.g., priority boosting and frame
precomputation). However, optimizing twice (i.e., using a low-
latency operating system and specialized software with built-in
performance boosts) may lead to suboptimal resource allocation,
which will result in poorer overall performance. Therefore, we
tested whether the combination of a real-time operating system
and the specialized software is optimal compared with the same
software running on a general-purpose operating system. As will
be explained further in the Results section, not all tests run using a
real-time operating system were better than tests run using a
general-purpose operating system in all conditions.
Figure 1. Representation of the path followed by the electron
gun for a CRT monitor. At the end of each frame, the electron gun
returns to the top left corner and starts again (VSYNC signal).
doi:10.1371/journal.pone.0085108.g001
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 2 January 2014 | Volume 9 | Issue 1 | e85108
Materials and Methods
We used Black Box Toolkit (BBTK) [12] to test the accuracy
and precision of the timing mechanisms used to present visual
stimuli by several specialized software packages on PCs: E-Prime
2.0.8.90 [13], DMDX 4.0.4.8 [14] and PsychoPy 1.64.00 [15]. All
were tested on Microsoft Windows 7 32-bit Professional Edition.
Because PsychoPy 1.64.00 can run natively on both operating
systems, it was also tested on Ubuntu Linux 10.04 with the Linux
2.6.33-29-realtime kernel.
This subset of software packages is not aimed to be exhaustive
but to serve as an example of the type of errors that can be found
in their underlying timing mechanisms and the type of corrections
that should be applied for time-sensitive experiments’ configura-
tions. The selection of the investigated software packages has been
done with the following criteria: select a popular commercial
software package, a popular free software package, and a
multiplatform software package. Regarding popularity, E-Prime
is one of the most popular experimental software packages
worldwide. Its combination of a graphical authoring tool and
custom programming language to configure the experiments, its
licensing model, and its underlying mechanisms to schedule and
present stimuli make it a representative instance of this class
(without detriment or other popular software packages such as
DirectRT, NBS Presentation, etc.). DMDX is representative of a
long tradition of experimental software, starting at 1975 with
DMASTR and ending with its recent port to Microsoft DirectX
libraries. In fact, these libraries are used by other software
packages (e.g., NBS Presentation), therefore their limitations might
affect all of them in a similar way. PsychoPy is a multiplatform
software package that can run natively in Microsoft Windows,
GNU/Linux and Apple Mac OS X. It is based on Python, like
many other alternatives available (e.g., Experiment Builder,
PyEPL, OpenSesame, Vision Egg), and provides a graphical
authoring tool and a set of Python libraries to set up the
experiments. All software tests were conducted using the same
hardware: an Apple Macbook Pro with an Intel Core 2 Duo
T7600 processor and an ATI Radeon Mobility X1600 graphics
card. The native resolution of the display was 14406900 pixels at
60 Hz. The suitability of the display was confirmed using the 6-
hour RefreshClockTest of E-Prime (available at: http://www.pstnet.
com/support/kb.asp?TopicID = 3003) and an analysis of the
results provided by the timeByFrames test of PsychoPy.
The BBTK photodetectors were used to measure the timing
under all conditions. The BBTK was designed for this type of
measurement [12] [16] and is able to transfer detected changes in
luminance from the photodetector to the parallel port in less than
100 ns.
A well-known procedure was used to test the precision and
accuracy of the visual stimuli presentations [17] [18]. For each of
the experimental software packages examined, the photodetector,
placed in the middle of the screen, was programmed to detect
black to white and white to black screen transitions in the
presentation display. In five independent series of 60 seconds, a
continuous alternation of white and black screens was scheduled to
be repeated with interval durations of 16.667, 50, 100, 200, 500,
and 1000 ms (i.e., 1, 3, 6, 12, 30 and 60 ticks at 60 Hz,
respectively). To compute the amount of missed frames, the whole
set of measurements for each testing condition was used (i.e.,
approximately 300, 600, 1500, 3000, 6000, and 18000 for 1000,
500, 200, 100, 50, and 16.667 ms, respectively).
As in previous studies using this measurement procedure [17]
[18], the presentation and measurement equipment used for all
timing conditions are independent in order to avoid undesired
interferences between the timing mechanisms used to generate the
black to white and white to black transitions of the specialized
software and the real-time application used to gather the data
provided by the photodetector. The Apple MacBook Pro runs the
experimental software independently (as it would be run in a real
experiment with human participants), and the BBTK detects all
changes in luminance from the Macbook’s display and sends them,
via a parallel-port connection, to an auxiliary computer (AMD
Sempron 2200 running the BBTK’s capture data software under
Microsoft Windows XP 32-bit Professional edition).
Results
After considering all the previously mentioned details regarding
displays, we expected that the difference between the interval
configured by the experimenter for a visual stimulus presentation
and the actual interval displayed by the software would be a
multiple of the duration of a tick (see Figure 2a). However, we
observed that all measured timing errors (MTEs) were concen-
trated around two peaks: one slightly before the VSYNC signal,
the other slightly later. Figure 2b shows an ideal representation,
and Figure 2c shows an example of the measurements collected by
our photodetectors during the analysis of a 1000-ms interval
animation performed using E-Prime.
The discrepancy between our expectations and the actual
measurements can be explained by considering the rise and decay
times of the LCD display. Because the response time is greater
than 0 ms, these times are slightly longer for black-to-white
transitions and slightly shorter for white-to-black transitions.
Further technical details about rise and decay times of LCD
displays and their implications in the duration of the stimuli are
explained by Elze [8] [9] and Elze and Tanner [10]. Moreover,
the BBTK photosensors do not provide a continuous analog value
but a discrete digital one based on an adjustable threshold. This
MTE is attributed to the technology used to show and measure the
visual stimuli and not to the experimental software. Therefore, we
decided to convert the MTE to full missed frames using the
formula:
Missed frames ~
MTE{tick
2
,MTE v0
MTEztick
2
,MTE §0
8
>
>
>
<
>
>
>
:
ð1Þ
where jdenotes the floor function.
Even with this conversion, we found discrepancies between
stipulated and measured intervals in most of the setups we
configured in DMDX, E-Prime and PsychoPy. These discrepan-
cies were so consistent across all tests (each test was repeated 5
times per interval and software) that we decided to identify ways to
compensate for them, as described below. We then repeated the
measurements to assess the optimal precision and accuracy of the
presentation of visual stimuli using the tested software. The results
shown in Table 1 are the final measurements, once all adjustments
and corrections had been made (the complete dataset can be found
in https://openscienceframework.org/project/F2gBN.).
In our initial DMDX configurations, we could not attain
transition intervals shorter than three ticks (i.e., 50 ms at 60 Hz),
even if the delay between trials was set to 0 ticks. In a personal
communication, K. I. Forster confirmed that fast sequences of
different items should not be presented as separate items but as a
single item with multiple frames. This advice should be heeded by
DMDX users when very short transitions between different trials
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 3 January 2014 | Volume 9 | Issue 1 | e85108
are needed. Researchers should adjust related aspects of the
experiment (e.g., non-visual stimuli presentations, reaction time
measurement, or external apparatus synchronization) to compen-
sate for this change.
We faced a different problem with E-Prime. Because durations
in E-Prime are configured in milliseconds and not ticks, the
duration of each interval was defined as a multiple of the number
of milliseconds in one tick (i.e., 1000 ms for 60 ticks, 500 ms for 30
ticks, 200 ms for 12 ticks, etc.), and we requested VSYNC
synchronization. After analyzing our initial measurements, we
found that E-Prime was consistently missing a frame in each
interval. Fortunately, information about these timing errors was
provided in E-Prime’s log files, and they are related to the
preparation time of the following stimulus (configured via the
PreRelease value, which allows the current stimulus to release a
portion of its execution time to a following stimulus in order to
allow the following stimulus to perform setup activities) and E-
Prime’s timing modes (i.e., Cumulative Mode and Event Mode).
After testing several configurations, we concluded that the best
way to optimize the precision and accuracy of visual stimuli
presentations using E-Prime for this simple animation is to use E-
Prime’s Event Mode timing (i.e., delays in the onset of an event
will not affect the specified duration of the event), to subtract the
duration of one tick from the interval duration in each transition
(e.g., 1000 ms–16.667 ms = 983.333 ms) and to force the synchro-
nization with VSYNC signal at the onset and offset of the stimulus.
This adjustment is coherent with the ‘‘rule of thumb’’ provided by
vendors in the user manual, according to which the stimulus
Figure 2. Measured timing errors distribution. Measured timing errors distribution: (a) Expected distribution of measured timing errors. (b)
Actual distribution of measured timing errors. (c) Distribution of measured timing errors (in ms) of the PsychoPy 1.64.00 software displaying an
animation from black to white every 1000 ms onLinux 2.6.33-29-realtime.
doi:10.1371/journal.pone.0085108.g002
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 4 January 2014 | Volume 9 | Issue 1 | e85108
duration should be set to 10ms below the expected total duration
of the stimulus (‘‘A good rule of thumb is to set the stimulus
duration to 10 ms below the expected total duration of all refresh
cycles desired for the stimulus. Since the visual duration is always
rounded up to the next refresh, the display duration must be
specified as some portion of one refresh below the targeted refresh
duration plus the expected measurement error for when to look for
the vertical blanking signal indicating the time of a refresh.’’ [13]).
But instead of using an arbitrary correction value, we decided to
follow a more detailed suggestion provided in the E-Prime users’
manual and make this correction value dependent on the refresh
rate of the monitor and the time needed to prepare the next frame
of the animation (i.e., a non-zero but negligible value in our simple
animation).
Our tests using PsychoPy also required adjustments. We found
that a black frame lasting 1 tick was introduced in every frame
transition, which increased the duration of the black intervals by
two ticks (one before the beginning of the interval and another
after the end) and had no effect in the duration of white intervals.
To compensate for this, we subtracted two ticks from the duration
of the black intervals (e.g., 1000 ms–2616.667 ms = 966.667 ms).
For intervals lasting 1 tick, a continuous white to white transition 1
tick in length was configured because PsychoPy automatically
inserted one black frame 1 tick in duration between them.
As mentioned previously, the results shown in Table 1 represent
the missed frames (i.e., the number of frames away from the target)
for intervals lasting 1000, 500, 200, 100, 50, and 16.667 ms,
respectively once all adjustments were performed. Note that the
number of missed frames in each row do not necessarily add up to
the expected number of frames. This happens because exactly 5
minutes of black-to-white and white-to-black transitions were
recorded. If any of the frames happened to be presented later than
scheduled, then the 5 minute-interval might not suffice to present
the whole sequence of programmed transitions. Positive values
represent intervals longer than the configured interval, negative
values represent shorter intervals, and zero means the interval
lasted the stipulated duration. Quite interestingly, the number of
missed frames is noticeably higher when the shortest intervals were
tested with PsychoPy, both in Microsoft Windows and Linux.
Figure 3 depicts the mean number of missed frames when using
PsychoPy under both OS. To improve the comparability across
time intervals, we computed these means using only the first 50
Table 1. Missed frames for each testing condition.
Software
/OS
Interval
(ms)
Expectancy
(frames) Missed frames
2
2
2
10 1 2
.
2
DMDX 1000 300 0 0 300 0 0 0
/Windows 7 500 600 0 0 600 0 0 0
200 1500 0 0 1500 0 0 0
100 3000 0 0 2999 1 0 0
50 6000 0 0 6000 0 0 0
16.667 18000 0 0 17998 1 0 0
2
2
2
10 1 2
.
2
E-Prime 1000 300 0 0 299 1 0 0
/Windows 7 500 600 0 0 600 0 0 0
200 1500 0 0 300 0 0 0
100 3000 0 0 3000 0 0 0
50 6000 0 0 5997 1 1 0
16.667 18000 0 0 17857 28 20 12*
2
2
2
10 1 2
.
2
PsychoPy 1000 300 0 0 300 0 0 0
/Windows 7 500 600 0 0 600 0 0 0
200 1500 0 2 1496 2 0 0
100 3000 0 1179 634 1187 0 0
50 6000 0 0 2572 2573 0 0
16.667 18000 0 0 5981 5980 0 0
2
2
2
10 1 2
.
2
PsychoPy 1000 300 0 0 300 0 0 0
/Linux RT 500 600 0 9 582 9 0 0
200 1500 2 13 1447 13 0 0
100 3000 2 25 2889 32 2 0
50 6000 0 0 5113 575 0 0
16.667 18000 0 0 17908 14 3 10**
*E-Prime, 16.667 ms interval, .2 missed frames distribution: 3 missed frames: 8; 4 missed frames: 3; 11 missed frames: 1.
**PsychoPy under Linux RT, 16.667 ms interval, .2 missed frames distribution: 3 missed frames: 10.
doi:10.1371/journal.pone.0085108.t001
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 5 January 2014 | Volume 9 | Issue 1 | e85108
transitions in each round (250 transitions in each condition). As
can be seen, the performance of PsychoPy is worse under very
brief intervals, but this effect is somewhat ameliorated when
PsychoPy operates under realtime-Linux.
Discussion
In all of our tests, DMDX presented visual stimuli with high
precision and accuracy. Were it not for the difficulty in
programming experiments using the DMASTR syntax, it would
be a perfect tool. DMDX allows defining time intervals both in
ticks and in milliseconds. Although defining stimuli durations in
ticks can be tricky, it is closer to the actual implementation of the
experiment and makes the limitations of the hardware explicit to
researchers, which discourages them from setting intervals that are
impossible to meet (i.e., those that are not a multiple of the tick
duration). In fact, as Forster explains in the DMDX Online Help
page [19], millisecond-oriented keywords are only provided for
one of two situations: (1) when researchers do not care about
precise tachistoscopic presentations, or (2) when the item file is
needed to work on multiple machines regardless of their refresh
rates. Moreover, very short stimuli presentations have to be
aggregated as different frames of the same trial to eliminate inter-
trial delays. This change can interfere with other aspects of the
experimental design (e.g., non-visual stimuli presentations, reac-
tion time measurements, etc.) or be overlooked by experimenters.
We conclude that E-Prime is a highly precise software package
for the presentation of visual stimuli. However, setting the stimuli
durations in milliseconds (and not in ticks) might introduce a
source of error. On the one hand, it is much easier for researchers
to use. They are accustomed to thinking in terms of seconds or
milliseconds and may not be aware of the concept of a tick. On the
other hand, using presentation times that are not multiple of the
refresh rate may lead to timing errors. These errors can be smaller
than the duration of one tick, but will still increase the measured
error. According to our tests, even visual stimuli with a duration
that is a multiple of the refresh rate may have to wait to be
synchronized with the VSYNC signal of the display. This delay
results in larger presentation times for the previous stimulus. As
mentioned previously, correcting this deviation is straightforward.
However, if a researcher focuses on the reaction times in the E-
Prime logs and does not analyze the registered timing errors
during the presentation of stimuli, this error can be overlooked.
We recommend that researchers set the VSYNC signal synchro-
nization and subtract the duration of one tick when defining
stimuli intervals, and ensure that the pre-computation time needed
to prepare stimuli does not exceed on that subtracted tick plus the
amount of time configured in the PreRelease value of the stimulus.
Both PsychoPy and E-Prime provide the experimenter with a
user-friendly interface for setting up experiments and the ability to
define arbitrary times for the duration of the stimuli. Therefore,
similar recommendations apply. In addition, PsychoPy automat-
ically includes a black frame lasting 1 tick in every stimuli
transition. This inclusion should be taken into account in
experimental paradigms with high precision and accuracy
requirements, and the frame duration should be corrected by
subtracting the black frames when possible. Moreover, PsychoPy is
able to eliminate the limitations of the Graphical User Interface
(GUI) by providing an Application Programming Interface (API)
from which experiments can be easily developed in Python.
Combining both high-level (i.e., GUI) and low-level (i.e., Python
API) approaches, experimenters are able to generate experimental
setups that are as accurate as the underlying platform (e.g., using
frame refresh periods to control presentation timing accurately).
Therefore, PsychoPy is an attractive alternative for designing and
running cognitive experiments with some very interesting charac-
teristics. For example, it is published under a Free Software
license. In addition, it runs under the most popular general-
purpose operating systems (i.e., Microsoft Windows, GNU/Linux
and Mac OS X) and is even able to take advantage of the
reliability provided by real-time operating systems (e.g., Linux
2.6.33–29-realtime).
The significant differences between a multi-platform software
package such as PsychoPy and single-platform alternatives like E-
Prime or DMDX are related to the software abstraction layers
involved in each case. DMDX uses a multimedia library
specifically designed for the development of real-time applications
(i.e., DirectX) and requires pre-configuration of display equipment
Figure 3. Absolute number of missed frames in PsychoPy tests. Absolute number of missed frames for each testing condition of PsychoPy
running on Microsoft Windows 7 Professional 32-bit edition and Ubuntu 10.04 LTS with Linux 2.6.33–29-realtime.
doi:10.1371/journal.pone.0085108.g003
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 6 January 2014 | Volume 9 | Issue 1 | e85108
to achieve the maximum accuracy and precision. E-Prime is not as
close to the hardware as DMDX, but leverages best timing APIs in
Microsoft Windows (i.e., QueryPerformanceCounters and Multi-
media Timers) to maximize the accuracy and precision of stimulus
presentations. PsychoPy, by contrast, relies on a high-level and
multi-platform interpreted language (i.e., Python). The benefits of
this approach are clear: any PsychoPy-based experiment can run
smoothly on all major operating systems. However, the transition
from PsychoPy experiment setup files (i.e., XML-based psyexp
files) to the control of the display hardware might be performed
sub-optimally if not tuned manually. The good news is that
PsychoPy gives advanced users the ability to maximize the
accuracy and precision of stimulus presentation when used as a
Python library through the use of a non-slip (global) clock timing
mechanism. This functionality is also available from the Exper-
iment Builder of PsychoPy since version 1.74.00. Moreover,
Python scripts generated by PsychoPy could improve their
performance if compiled to native executable code or used an
optimized Python interpreter.
There may be several causes that explain the differences
between the results gathered using PsychoPy on Microsoft
Windows and on Linux real-time. The best-effort policy of a
non-real-time operating system like Microsoft Windows is able to
provide a better throughput than a real-time operating system in
non demanding cases (i.e., tests with larger intervals), while
performs worse in demanding situations (i.e., tests with shorter
intervals). Moreover, the software architecture of each operating
system is significantly different regarding the graphical layer (i.e., it
runs as a user process in Linux, whereas it is integrated in the
kernel in Microsoft Windows), and might be the cause of a slightly
poorer performance in Linux in tests where the stress of the system
is not an issue.
Further research is needed to extend these conclusions to a
wider range of experimental software. Providing sub-millisecond
accuracy is a typical claim of vision-related experimental software,
but the validity of these claims should be tested by third-party
researchers. Nevertheless, although some parameters (e.g., the
refresh rate of the LCD display) were not optimal, the overall
performance of the tested operating systems and experimental
software combinations were sufficient to fulfill the requirements of
most of the experimental tasks performed by researchers. This is
good news for researchers who need to present stimuli with high
accuracy. Provided that the recommendations explained in the
present work are taken into account by researchers, we consider
DMDX, E-Prime and PsychoPy suitable for most of the
experimental paradigms used in cognitive research that require
high levels of precision and accuracy.
Author Contributions
Conceived and designed the experiments: PG MAV DLI HM. Performed
the experiments: PG. Analyzed the data: PG MAV. Contributed reagents/
materials/analysis tools: HM. Wrote the paper: PG MAV DLI HM.
References
1. Mathoˆ t S, Schreij D, Theeuwes J (2011) OpenSesame: An open-source,
graphical experiment builder for the social sciences. Beh Res Methods, 44, 314–
324.
2. Neely JH (1977) Semantic priming and retrieval from lexical memory: Roles of
inhibitionless spreading activation and limited-capacity attention. J Exp Psychol-
Gen, 106, 225–254.
3. Ratcliff R, McKoon G (1978) Priming in item recognition: Evidence for the
prepositional structure of sentences. J Verbal Learn Verb Beh, 17, 403–417.
4. Fazio RH, Sanbonmatsu DM, Powell MC, Kardes FR (1986) On the automatic
activation of attitudes. J Pers Soc Psychol, 50, 229–238.
5. Cowan WB (1995) Displays for vision research. In M. Bass (Ed.), Handbook of
optics, Vol. 1: Fundamentals, techniques, and design (pp. 27.21–27.44). New
York: McGraw-Hill.
6. Lagroix H, Yanko M, Spalek T (2012) LCDs are better: Psychophysical and
photometric estimates of the temporal characteristics of CRT and LCD
monitors. Atten Percep Psychophys., 74, 1033–41.
7. Wang P, Nikolic D (2011) An LCD monitor with sufficiently precise timing for
research in vision. Front Hum Neurosci, 5, 85.
8. Elze T (2010) Achieving precise display timing in visual neuroscience
experiments. J Neurosci Methods, 191, 171–179.
9. Elze T (2010) Misspecifications of Stimulus Presentation Durations in
Experimental Psychology: A Systematic Review of the Psychophysics Literature.
PLoS ONE, 5, e12792.
10. Elze T, Tanner T (2009) Liquid crystal display response times estimation for
medical applications. Med Phys, 36, 4984–4990.
11. Castellano JA (1992) Handbook of display technology. New York: Academic
Press.
12. Plant RR, Hammond N, Turner G (2004) Self-validating presentation and
response timing in cognitive paradigms: How and why? Beh Res Meth Instr C,
36, 291–303.
13. Schneider W, Eschman A, Zuccolotto A (2002) E-Prime user’s guide. Pittsburgh,
PA: Psychology Software Tools.
14. Forster KI, Forster JC (2003) DMDX: A Windows display program with
millisecond accuracy. Beh Res Methods, 35, 116–124.
15. Peirce JW (2007) PsychoPy: Psychophysics software in Python. J Neurosci
Methods, 162, 8–13.
16. Plant RR, Turner G (2009) Millisecond precision psychological research in a
world of commodity computers: New hardware, new problems? Beh Res
Methods, 41, 598–614.
17. Schmidt WC (2001) Presentation accuracy of Web animation methods. Beh Res
Meth Instr C, 33, 187–200.
18. Stewart N (2006) Millisecond accuracy video display using OpenGL under
Linux. Beh Res Methods, 8, 142–145.
19. Foster K (2012) DMDX Display Software. Retrieved 18 November 2013, from
http://www.u.arizona.edu/,kforster/dmdx/dmdx.htm.
Experimental Software Timing Errors
PLOS ONE | www.plosone.org 7 January 2014 | Volume 9 | Issue 1 | e85108
... Visual stimuli were presented on a 22-inches computer screen (resolution 1680 × 1050, refresh rate 60.0 Hz; 16.67 ms ± 12.37 ms) located at 1.5 m away from the participants. Refresh rate was assessed via a photosensor and corresponded to normative values 60 . ...
Article
Full-text available
Visual processing of emotional stimuli has been shown to engage complex cortical and subcortical networks, but it is still unclear how it affects sensorimotor integration processes. To fill this gap, here, we used a TMS protocol named short-latency afferent inhibition (SAI), capturing sensorimotor interactions, while healthy participants were observing emotional body language (EBL) and International Affective Picture System (IAPS) stimuli. Participants were presented with emotional (fear- and happiness-related) or non-emotional (neutral) EBL and IAPS stimuli while SAI was tested at 120 ms and 300 ms after pictures presentation. At the earlier time point (120 ms), we found that fear-related EBL and IAPS stimuli selectively enhanced SAI as indexed by the greater inhibitory effect of somatosensory afferents on motor excitability. Larger early SAI enhancement was associated with lower scores at the Behavioural Inhibition Scale (BIS). At the later time point (300 ms), we found a generalized SAI decrease for all kind of stimuli (fear, happiness or neutral). Because the SAI index reflects integrative activity of cholinergic sensorimotor circuits, our findings suggest greater sensitivity of such circuits during early (120 ms) processing of threat-related information. Moreover, the correlation with BIS score may suggest increased attention and sensory vigilance in participants with greater anxiety-related dispositions. In conclusion, the results of this study show that sensorimotor inhibition is rapidly enhanced while processing threatening stimuli and that SAI protocol might be a valuable option in evaluating emotional-motor interactions in physiological and pathological conditions.
... The prime duration in most masked priming experiments is 50 ms, which is three ticks when the refresh rate is 60 HZ. More details about software timing in cognitive experiments can be found in Garaizar et al. (2014). The sequence of each trial was: a pattern mask (####) for 500 ms, followed by a 50 ms prime, and then the target appeared for either 3,000 ms or until the participant's response. ...
Article
Using two-character Chinese word targets in a masked priming lexical-decision task, Gu and colleagues (2015) demonstrated a significant transposed character (TC) priming effect. More importantly, the priming effect was the same size for single-morpheme words and multiple-morpheme words, suggesting that TC priming effects are not influenced by morphemic structure. In Chinese, there are, however, two types of single-morpheme words, single-morpheme simple words (e.g., [similar to practice in English]) and single-morpheme complex words (e.g., [similar to carpet in English in that both components are words themselves and, hence, when presented in transposed order, may activate morphological information reflecting the individual components rather than the word itself]), a contrast that Gu et al. did not examine. In Experiment 1, we replicated Gu et al.'s finding of equal TC priming effects for their single- and multiple-morpheme words, although our priming effects were noticeably smaller than theirs. In Experiment 2, we split the single-morpheme condition in order to examine the TC priming effects for single-morpheme simple words, single-morpheme complex words and multiple-morpheme words. The results showed that the single-morpheme complex words produced the smallest priming effect, indicating that transposed morphemes can influence masked priming in Chinese; however, apparently only in an inhibitory fashion. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... If the projector technique is to become a useful test in glaucoma research, it should be directly compared with other modern stimulation techniques [38]. Generally, stimulators cannot simply be exchanged for another without considering that spatiotemporal differences in display technologies and stimuli may affect various facts in experiments [42][43][44]. Prior to clinical application, the establishment of guidelines for the use of a projector is recommended. Further research and investigation on these more recent technologies may enable us to use BYVEP with projector stimulation in the diagnosis and management of glaucoma. ...
Article
Full-text available
Background and aim In the past, increased latencies of the blue-on-yellow pattern visually evoked potentials (BY-VEP), which predominantly originate in the koniocellular pathway, have proven to be a sensitive biomarker for early glaucoma. However, a complex experimental setup based on an optical bench was necessary to obtain these measurements because computer screens lack sufficient temporal, spatial, spectral, and luminance resolution. Here, we evaluated the diagnostic value of a novel setup based on a commercially available video projector. Methods BY-VEPs were recorded in 126 participants (42 healthy control participants, 12 patients with ocular hypertension, 17 with “preperimetric” glaucoma, and 55 with perimetric glaucoma). Stimuli were created with a video projector (DLP technology) by rear projection of a blue checkerboard pattern (460 nm) for 200 ms (onset) superimposed on a bright yellow background (574 nm), followed by an offset interval where only the background was active. Thus, predominantly S-cones were stimulated while L- and M-cone responses were suppressed by light adaptation. Times of stimulus onset to VEP onset-trough (N-peak time) and offset-peak (P-peak time) were analyzed after age-correction based on linear regression in the normal participants. Results The resulting BY-VEPs were quite similar to those obtained in the past with the optical bench: pattern-onset generated a negative deflection of the VEP, whereas the offset-response was dominated by a positive component. N-peak times were significantly increased in glaucoma patients (preperimetric 136.1 ± 10 ms, p < 0.05; perimetric 153.1 ± 17.8 ms, p < 0.001) compared with normal participants (123.6 ± 7.7 ms). Furthermore, they were significantly correlated with disease severity as determined by visual field losses retinal nerve fiber thinning (Spearman R = –0.7, p < 0.001). Conclusions Video projectors can be used to create optical stimuli with high temporal and spatial resolution, thus potentially enabling sophisticated electrophysiological measurements in clinical practice. BY-VEPs based on such a projector had a high diagnostic value for detection of early glaucoma. Registration of study Registration site: www.clinicaltrials.gov Trial registration number: NCT00494923.
... They vary both in terms of specialization of specific functionality and research area. However, all of the software is still dependent on computer hardware, drivers, and operating systems (OS), resulting in variability of configuration settings and installations (Plant et al., 2004;Garaizar et al., 2014). ...
Article
In auditory behavioral and EEG experiments, the variability of stimulation solutions, for both software and hardware, adds unnecessary technical constraints. Currently, there is no easy to use, inexpensive, and shareable solution that could improve collaborations and data comparisons across different sites and contexts. This article outlines a system composed by a Raspberry Pi coupled with Python programming and associated with a HifiBerry sound card. We compare its sound performances with those of a wide variety of materials and configurations. This solution achieves the high timing accuracy and sound quality important in auditory cognition experiments, while being simple to use and open source. The present system shows high performances and results along with excellent feedback from users. It is inexpensive, easy to build, share, and improve on. Working with such low-cost, powerful, and collaborative hardware and software tools allows people to create their own specific, adapted, and shareable system that can be standardized across different collaborative sites, while being extremely simple and robust in use.
... Nevertheless, it is still a code generator's responsibility to provide the best timing performance possible with the target experiment platform. One of the improvements for achieving higher precision with visual stimuli, due to the limitation of the display's refresh rate, would be an option to specify durations in ticks instead of milliseconds [47]. ...
Article
Full-text available
The majority of studies in psychology are nowadays performed using computers. In the past, access to good quality software was limited, but in the last two decades things have changed and today we have an array of good and easily accessible open-source software to choose from. However, experiment builders are either GUI-centric or based on general-purpose programming languages which require programming skills. In this paper, we investigate an approach based on domain-specific languages which enables a text-based experiment development using domain-specific concepts, enabling practitioners with limited or no programming skills to develop psychology tests. To investigate our approach, we created PyFlies, a domain-specific language for designing experiments in psychology, which we present in this paper. The language is tailored for the domain of psychological studies. The aim is to capture the essence of the experiment design in a concise and highly readable textual form. The editor for the language is built as an extension for Visual Studio Code, one of the most popular programming editors today. From the experiment description, various targets can be automatically produced. In this version, we provide a code generator for the PsychoPy library while generators for other target platforms are planned. We discuss the language, its concepts, syntax, some current limitations, and development directions. We investigate the language using a case study of the implementation of the Eriksen flanker task.
... Computer-based experiments were carried out using the Window program DMDX [29] widely used in the community of experimental and cognitive psychologists [16,18,19,24,[29][30][31]. Using this program, stimuli were presented to the participants and RTs recorded. ...
Article
Full-text available
An individual’s reaction time data to visual stimuli have usually been represented in Experimental Psychology by means of an ex-Gaussian function. In most previous works, researchers have mainly aimed at finding a meaning for the parameters of the ex-Gaussian function which are known to correlate with cognitive disorders. Based on the recent evidence of correlations between the reaction time series to visual stimuli produced by different individuals within a group, we go beyond and propose a Physics-inspired model to represent the reaction time data of a coetaneous group of individuals. In doing so, a Maxwell–Boltzmann-like distribution appeared, the same distribution as for the velocities of the molecules in an Ideal Gas model. We describe step by step the methodology we use to go from the individual reaction times to the distribution of the individuals response within the coetaneous group. In practical terms, by means of this model we also provide a simple entropy-based methodology for the classification of the individuals within the collective they belong to with no need for an external reference which can be applicable in diverse areas of social sciences.
... computer screen (resolution: 1,680 × 1,050, refresh rate: 60.0 Hz; 16.67 ± 12.37 ms) located at 80 cm away from the subjects. Refresh rate was assessed via a photosensor connected to the response box and corresponded to normative values (Garaizar et al., 2014). Stimuli were presented using the E-Prime 3.0 software (Psychology Software Tools, Pittsburgh, PA, United States). ...
Article
Full-text available
The investigation of how humans perceive and respond to emotional signals conveyed by the human body has been for a long time secondary compared with the investigation of facial expressions and emotional scenes recognition. The aims of this behavioral study were to assess the ability to process emotional body postures and to test whether motor response is mainly driven by the emotional content of the picture or if it is influenced by motor resonance. Emotional body postures and scenes (IAPS) divided into three clusters (fear, happiness, and neutral) were shown to 25 healthy subjects (13 males, mean age ± SD: 22.3 ± 1.8 years) in a three-alternative forced choice task. Subjects were asked to recognize the emotional content of the pictures by pressing one of three keys as fast as possible in order to estimate response times (RTs). The rating of valence and arousal was also performed. We found shorter RTs for fearful body postures as compared with happy and neutral postures. In contrast, no differences across emotional categories were found for the IAPS stimuli. Analysis on valence and arousal and the subsequent item analysis showed an excellent reliability of the two sets of images used in the experiment. Our results show that fearful body postures are rapidly recognized and processed, probably thanks to the automatic activation of a series of central nervous system structures orchestrating the defensive threat reactions, strengthening and supporting previous neurophysiological and behavioral findings in body language processing.
Article
Available as free and open-source software, lab.js Builder is easy to use to create experiments and surveys for both online and laboratory data collection. Its visual interface makes it possible to design stimuli and integrate them into experiments and surveys without programming, although it can also be customized with HTML, CSS, and JavaScript code. The software may be beneficial for students and researchers in allowing them to conduct research under home-schooling and remote-working conditions. In this tutorial article, after introducing lab.js Builder’s easy-to-use functions, we also demonstrate its application to conducting practical experiments within university classes.
Article
Full-text available
Web-based data collection is increasingly popular in both experimental and survey-based research because it is flexible, efficient, and location-independent. While dedicated software for laboratory-based experimentation and online surveys is commonplace, researchers looking to implement experiments in the browser have, heretofore, often had to manually construct their studies’ content and logic using code. We introduce , a free, open-source experiment builder that makes it easy to build studies for both online and in-laboratory data collection. Through its visual interface, stimuli can be designed and combined into a study without programming, though studies’ appearance and behavior can be fully customized using html , css , and JavaScript code if required. Presentation and response times are kept and measured with high accuracy and precision heretofore unmatched in browser-based studies. Experiments constructed with can be run directly on a local computer and published online with ease, with direct deployment to cloud hosting, export to web servers, and integration with popular data collection platforms. Studies can also be shared in an editable format, archived, re-used and adapted, enabling effortless, transparent replications, and thus facilitating open, cumulative science. The software is provided free of charge under an open-source license; further information, code, and extensive documentation are available from https://lab.js.org/ .
Article
Behavioral research in cognitive and human systems neuroscience has been largely carried out in-person in laboratory settings. Underpowering and lack of reproducibility due to small sample sizes have weakened conclusions of these investigations. In other disciplines, such as neuroeconomics and social sciences, crowdsourcing has been extensively utilized as a data collection tool, and a means to increase sample sizes. Recent methodological advances allow scientists, for the first time, to test online more complex cognitive, perceptual, and motor tasks. Here we review the nascent literature on the use of online crowdsourcing in cognitive and human systems neuroscience. These investigations take advantage of the ability to reliably track the activity of a participant’s computer keyboard, mouse, and eye gaze in the context of large-scale studies online that involve diverse research participant pools. Crowdsourcing allows for testing the generalizability of behavioral hypotheses in real-life environments that are less accessible to lab-designed investigations. Crowdsourcing is further useful when in-laboratory studies are limited, for example during the current COVID-19 pandemic. We also discuss current limitations of crowdsourcing research, and suggest pathways to address them. We conclude that online crowdsourcing is likely to widen the scope and strengthen conclusions of cognitive and human systems neuroscience investigations.
Article
Full-text available
We hypothesized that attitudes characterized by a strong association between the attitude object and an evaluation of that object are capable of being activated from memory automatically upon mere presentation of the attitude object. We used a priming procedure to examine the extent to which the mere presentation of an attitude object would facilitate the latency with which subjects could indicate whether a subsequently presented target adjective had a positive or a negative connotation. Across three experiments, facilitation was observed on trials involving evaluatively congruent primes (attitude objects) and targets, provided that the attitude object possessed a strong evaluative association. In Experiments 1 and 2, preexperimentally strong and weak associations were identified via a measurement procedure. In Experiment 3, the strength of the object-evaluation association was manipulated. The results indicated that attitudes can be automatically activated and that the strength of the objectevaluation association determines the likelihood of such automatic activation. The implications of these findings for a variety of issues regarding attitudes—including their functional value, stability, effects on later behavior, and measurement—are discussed.
Article
Full-text available
Prior to each target letter string presented visually to 120 university students in a speeded word–nonword classification task, either {bird, body, building,} or {xxx} appeared as a priming event. Five types of word-prime/word-target trials were used: bird-robin, bird-arm, body-door, body-sparrow, and body-heart. The stimulus onset asynchrony (SOA) between prime and target letter string varied between 250 and 2,000 msec. At 2,000-msec SOA, reaction times (RTs) on bird-robin type trials were faster than on xxx-prime trials (facilitation), whereas RTs on bird-arm type trials were slower than on xxx-prime (inhibition). As SOA decreased, the facilitation effect on bird-robin trials remained constant, but the inhibition effect on bird-arm decreased until, at 250-msec SOA, there was no inhibition. For Shift conditions at 2,000-msec SOA, facilitation was obtained on body-door type trials and inhibition was obtained on body-sparrow type. These effects decreased as SOA decreased until there was no facilitation or inhibition. On body-heart type trials, there was an inhibition effect at 2,000 msec SOA, which decreased as SOA decreased until, at 250-msec SOA, it became a facilitation effect. Results support the theory of M. I. Posner and S. R. Snyder (1975) that postulated 2 distinct components of attention: a fast automatic inhibitionless spreading-activation process and a slow limited-capacity conscious-attention mechanism. (27 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Full-text available
Several Web animation methods were independently assessed on fast and slow systems running two popular Web browsers under MacOS and Windows. The methods assessed included those requiring programming (Authorware, Java, Javascript/Jscript), browser extensions (Flash and Authorware), or neither (animated GIF). The number of raster scans that an image in an animation was presented for was counted. This was used as an estimate of the minimum presentation time for the image when the software was set to update the animation as quickly as possible. In a second condition, the image was set to be displayed for 100 msec, and differences between observed and expected presentations were used to assess accuracy. In general, all the methods except Java deteriorated as a function of the speed of the computer system, with the poorest temporal resolutions and greatest variability occurring on slower systems. For some animation methods, poor performance was dependent on browser, operating system, system speed, or combinations of these.
Article
Full-text available
Many cognitive and perceptual phenomena, such as iconic memory and temporal integration, require brief displays. A critical requirement is that the image not remain visible after its offset. It is commonly believed that liquid crystal displays (LCD) are unsuitable because of their poor temporal response characteristics relative to cathode-ray-tube (CRT) screens. Remarkably, no psychophysical estimates of visible persistence are available to verify this belief. A series of experiments in which white stimuli on a black background produced discernible persistence on CRT but not on LCD screens, during both dark- and light-adapted viewing, falsified this belief. Similar estimates using black stimuli on a white background produced no visible persistence on either screen. That said, photometric measurements are available that seem to confirm the poor temporal characteristics of LCD screens, but they were obtained before recent advances in LCD technology. Using current LCD screens, we obtained photometric estimates of rise time far shorter (1-6 ms) than earlier estimates (20-150 ms), and approaching those of CRTs (<1 ms). We conclude that LCDs are preferable to CRTs when visible persistence is a concern, except when black-on-white displays are used.
Article
Full-text available
In the present article, we introduce OpenSesame, a graphical experiment builder for the social sciences. OpenSesame is free, open-source, and cross-platform. It features a comprehensive and intuitive graphical user interface and supports Python scripting for complex tasks. Additional functionality, such as support for eyetrackers, input devices, and video playback, is available through plug-ins. OpenSesame can be used in combination with existing software for creating experiments.
Article
Full-text available
Until now, liquid crystal display (LCD) monitors have not been used widely for research in vision. Despite their main advantages of continuous illumination and low electromagnetic emission, these monitors had problems with timing and reliability. Here we report that there is at least one new inexpensive 120 Hz model, whose timing and stability is on a par with a benchmark cathode-ray tube monitor, or even better. The onset time was stable across repetitions, 95% confidence interval (the error) of which was <0.01 ms. Brightness was also delivered reliably across repeated presentations (<0.04% error) and across blocks with different durations (<3% error). The LCD monitor seems suitable for many applications in vision research, including the studies that require combined accuracy of timing and intensity of visual stimulation.
Article
A priming technique using item recognition was employed to investigate the structure of the memory representation of simple sentences. The experimental procedure involved presenting sentences to the subject for study and then testing single words for recognition (the subject had to decide whether the test word was in one of the study sentences). A large priming effect was obtained. Response time to a word from one of the study sentences preceded by another word from the same study sentence was over 100 milliseconds faster than response time to a word preceded by a word from a different study sentence. It was hypothesized that, if sentences are stored propositionally, then priming should be greater between words within a proposition than between words in different propositions in the same sentence. This result was obtained and provides strong evidence for the propositional structure of the sentences used in this study. In the discussion, parallels are drawn between priming and the processes used in sentence comprehension, and the process of activation is examined in some detail.