Page 164 - Contributed Paper Session (CPS) - Volume 4

CPS2165 Carla Susete Gonçalves Francisco et al.
1. Introduction
Reaction time (TR), or latency, is the interval between presenting a
stimulus and making a response to it. Saccade is the eye movement we
make to look at a target in our field of view. We make two or three saccades
every second of our lives. Latency is in the order of 200 ms, see Robinson
(1964). With modern equipment based on computational systems, it is
possible to obtain very large datasets of saccadic latency measurements
and to determine the form of their variability. The result is always a skewed
distribution, with a longer tail to the right. This distribution does not fit
particularly well in any of the most common standard mathematical
distributions (Gaussian, Poisson, Gamma, etc.). Observed variability in
reaction time might will be due to a variability in the rate of the underlying
process. Looking at the reciprocal of reaction time (1 / T) promptness. The
distribution of the reciprocal of reaction is not only symmetrical, but
actually looks as though it might be Gaussian. If it were, that would not
only make for easier mathematical analysis, but would also suggest that
we had reached a genuinely fundamental phenomenon. A graphical
procedure is designed to convert our histogram into a cumulative
histogram. For it, we are using a specially-distorted scale, this time on the
vertical, probability axis (a reciprobit plot). In this case, if the distribution is
indeed Gaussian, we should get a straight line. This approach provides
means of characterizing the behavior of the reaction time using
experimental data that is summarized through a very small number of
parameters, since it is sufficient to specify a median and the intercept of
the main distribution, Burle (2004).

2. Methodology
The idea of analyzing reciprocal latency, results from the treatment of
reaction time, due to a process whose rate varies randomly from one
experiment to another. Some type of decision signal, starting at an initial
level S0, rises to a constant rate r until it reaches a threshold value ST, at
which point a response is initiated. If r is randomly varied from one test to
2
another, such as a Gaussian with mean μ and variance σ , the asymmetry
of observed latency distributions is immediately explained. There are four
approaches:
• LATER (Linear Threshold Approximation with Ergodic Rate) Model.
• ELATER (Extended LATER) Model.
• Fieller distribution with parameters κ, λ1, λ2, ρ.
• DDM Model.
The DDM (Drift- Diffusion Model) is the most successful of the model
family to simulate growths up, to the threshold level. This model shares
many of its features with the LATER model. DDM family models are

153 | I S I W S C 2 0 1 9