MLP: a MATLAB toolbox for rapid and reliable auditory threshold estimation.

TLDR: MLP, a MATLAB toolbox enabling auditory thresholds estimation via the adaptive maximum likelihood procedure proposed by David Green (1990, 1993), is presented.

Abstract: In this article, we present MLP, a MATLAB toolbox enabling auditory thresholds estimation via the adaptive maximum likelihood procedure proposed by David Green (1990, 1993). This adaptive procedure is particularly appealing for those psychologists who need to estimate thresholds with a good degree of accuracy and in a short time. Together with a description of the toolbox, the present text provides an introduction to the threshold estimation theory and a theoretical explanation of the maximum likelihood adaptive procedure. MLP comes with a graphical interface, and it is provided with several built-in, classic psychoacoustics experiments ready to use at a mouse click.

Figures

TABLE 1 ABOUT HERE ----------------------------

Full text

MLP: a MATLAB toolbox for rapid and reliable auditory
threshold estimation
GRASSI, Massimo and SORANZO, Alessandro <http://orcid.org/0000-0002-
4445-1968>
Available from Sheffield Hallam University Research Archive (SHURA) at:
http://shura.shu.ac.uk/6129/
This document is the author deposited version. You are advised to consult the
publisher's version if you wish to cite from it.
Published version
GRASSI, Massimo and SORANZO, Alessandro (2009). MLP: a MATLAB toolbox for
rapid and reliable auditory threshold estimation. Behavior Research Methods, 41 (1),
20-28.
Copyright and re-use policy
See http://shura.shu.ac.uk/information.html
Sheffield Hallam University Research Archive
http://shura.shu.ac.uk

B411
1
Running head: MAXIMUM LIKELIHOOD PROCEDURE
MLP: a MATLAB toolbox for rapid and reliable auditory threshold estimation
Massimo Grassi* and Alessandro Soranzo#
*Dipartimento di Psicologia Generale - Università di Padova Via Venezia 8
35131 – Padova
Italy
#School of Social Science and Law - University of Teesside
Middlesbrough - UK
Email: massimo.grassi@unipd.it
Phone: +39 049 8277494
Fax: +39 049 8276600
* Corresponding author

B411
2

B411 - MAXIMUM LIKELIHOOD PROCEDURE
3
Abstract
In this paper, we present MLP, a MATLAB toolbox enabling auditory
thresholds estimation via the adaptive Maximum Likelihood procedure proposed
by David Green (1990, 1993). This adaptive procedure is particularly appealing for
those psychologists that need to estimate thresholds with a good degree of accuracy
and in a short time. Together with a description of the toolbox, the current text
provides an introduction to the threshold estimation theory and a theoretical
explanation of the maximum likelihood adaptive procedure. MLP comes with a
graphical interface and it is provided with several built-in, classic psychoacoustics
experiments ready to use at a mouse click.

B411 - MAXIMUM LIKELIHOOD PROCEDURE
4
MLP: a MATLAB toolbox for rapid and reliable auditory threshold estimation
In this paper, we present MLP, a MATLAB toolbox enabling auditory
thresholds estimation via the adaptive Maximum Likelihood procedure proposed
by David Green (1990, 1993). This procedure (hereafter referred to as ML) is
particularly suitable to estimate thresholds with an optimal compromise between
accuracy and rapidity. For this reason, the ML procedure has been used
successfully in clinical contexts (e.g., Florentine, Buus, & Geng, 2000), in studies
with children (e.g., Wright et al., 1997) as well as in studies with a large number of
subjects (e.g., Amitay, Irwin, & Moore 2006). For the same reason, it is suitable for
those studies where subjects perform various tasks, therefore, when each task has
to consume only a portion of the subject’s time. The ML procedure is largely
known, used and appreciated by the auditory community, it has collected more than
one hundred and twenty citations and the majority of these citations come from
journals specialized in the auditory research [footnote 1]. Thus, the user of this
procedure can benefit of a large background literature to optimise his/hers own
threshold estimation. As far as we know, MLP is the first software implementing
an adaptive psychophysical procedure with a graphical interface in a freely
downloadable version and it is provided with several built-in, classic
psychoacoustics experiments ready to use at a mouse click.
In the next section, we give a short introduction to the threshold estimation
theory. The reader familiar with these concepts may wish to skip this section. The
ML procedure and the MLP toolbox will be illustrated after this section.

Frequently asked questions

Q1. What is the significance of the ML procedure?

----------------------------The major benefit of the ML procedure is that it makes maximal use of theavailable data: the data of all trials are used to estimate the subject’s threshold. 

Q2. What is the parameter of the ML procedure?

After the selection of the parameters for the ML procedure, the parameter offor the stimulus selection policy (i.e., p-target) has to be chosen. 

Q3. What is the last thing the experimenter can do to control the goodness of the threshold estimates?

After the ML procedure has terminated there is one last thing theexperimenter can do to further control the goodness of the threshold estimates, that is controlling for attentional lapses. 

Q4. What is the threshold for a particular task?

The choice of the specific task depends on two factors: the desired experiment duration and the desired robustness of the threshold estimation. 

Q5. What types of psychometric functions can be adopted to fit experimental data?

Different types of psychometric functions can be adopted to fit experimental data, for example, the logistic, the Weibull and the cumulative Gaussian. 

Q6. How do the authors calculate the subject’s threshold?

By means of equation (6) the authors calculate the subject’s threshold: the authors take H1 and calculate the stimulus level that corresponds to 80.9% of correct responses. 

Q7. How many trials can be used to track the false alarm rate?

In other words, if the authors expect a false alarm rate ranging from 0% to 40%, the authors will track 63.1%, i.e., the average between the sweetpoint for 0% false alarm rate (i.e., 50%) and the sweetpoint for 40% false alarm rate (i.e., 76.2%). 

Q8. How can a subject estimate the detection threshold?

The detection threshold can be estimated either via yes/no tasks or viamultiple Alternative Forced Choice tasks (in brief nAFC, with n being the number of alternatives). 

Q9. How many trials should be required to calculate the threshold?

Recent studies suggest that an optimal threshold estimate should require about 24 (Leek et al., 2000) or about 30 trials (Amitay et al., 2006), that still remains a small figure [footnote 8].