Abstract

Noise is ubiquitous in nature. For this reason, it makes up a widely investigated topic. Taking into account experimental physics, whatever the measurement process under investigation, noise, defined as a random fluctuation of the measured signal, is usually considered to be detrimental. Thus, many techniques have been developed to reduce its impact: for example, the most common post–processing procedure to improve a noisy measurement consists in averaging the measured signal over repeated sessions. Rather surprisingly, the opposite, non–detrimental effect has been also observed: the response of a nonlinear system to a weak input signal is, under suitable conditions, optimized by the presence of a particular, non–vanishing noise level. In the last decades, the wide spectrum of such phenomena has been referred to as stochastic resonance. With regard to neuroscience, noise can be considered ubiquitous also in the brain processes. Theoretical considerations, as well as a robust experimental evidence, demonstrate its role in many phenomena occuring at different levels in neural system: for example, firing rate of neurons is not predictable by reason of their intrinsic variability [1, 5, 6]; measurements of brain activity with imaging techniques, like EEG, MEG or fMRI, are always affected by inner, noisy mechanisms. Recently, neuroscientists put forward the idea that stochastic fluctuations of neural activity have a functional role in brain dynamics. However, this functional role has been thus far observed in very few experimental situations. Regarding the human behaviour, decision making (representing our main field of investigation) is also influenced by several noisy processes. Also in this case, random fluctuations of brain processes typically have a detrimental role, limiting, for example, the possibility to exactly predict the human behaviour not only in everyday life, but also under controlled conditions, such as psychophysical experiments. Aim of this work is to gain insight into the problem of how noise influences the decisional mechanisms. In this framework, two different kinds of noise were investigated, namely endogenous and exogenous. The endogenous noise refers to stochastic fluctuations that are present within the neural system whereas with exogenous some form of enviromental noise, external to the perceptual system, are meant. These two distinct types of noise define two indipendent research fields, that were both investigated by means of behavioural experiments. In other words, we were interested in investigating how noise acts on human behavior, in particular in case of discrimination processes. To this purpose, psychophysical experiments were carried out, addressing both the acoustic and the visual modality. To investigate how a proper amount of exogenous noise can act positively on human perception, improving performance in detection experiments, – an effect that, as mentioned above, is interpreted as an occurence of stochastic resonance – we carried out experiments in the acoustic modality. This is the topic of the first part of this work. In particular, we used a detection paradigm where pure tone stimuli were superimposed with different levels of noise and subjects were requested to signalize the presence of the tone. Usually a sufficient noise level masks the signal. However,what we observed was a tiny, yet statistically significant improvement of stimulus detection ability in correspondence to a specific noise level. The used experimental approach – “Yes/No” experiments – is usually interpreted in terms of Signal Detection Theory (SDT). The two most important SDT parameters are the sensitivity d′ and the decisional criterion. Since improvement of detection ability driven by noise is, if any, a tiny effect, all the ingredients combined to formulate a decision must lay under the experimenter’s control. In particular, in addition to the stimulus detectability, also the knowledge of the decisional strategy at any time is crucial so as to achieve reliable data. In other words, the demand of criterion stability, and more in general the problem of its dynamics, turned out to have a critical role and urged us to focus our attention to the specific topic of criterion dynamics. The scientific literature on the possibility to condition the subject’s criterion and reconstruct its dynamics with the highest possible time resolution (the single trial) is extremely scant: the only two works on this subject do not provide robust methods to tackle the issue. The second part of the dissertation is completely focused on our theory of criterion setting dynamics and the related experimental evidence. An ad-hoc experiment involving the visual perceptual modality allowed to test a model for trial–by–trial criterion dynamics based on the theory of feedback. Feedback loop were implemented by informing the subject, after each trial, relatively to his/her performance. When requested to maximize the rate of correct response in an orientation discrimination experiment, subjects showed the ability to continuosly change their internal criterion. More in detail, the optimal criterion position oscillates at a certain frequency, set a-priori by the experimenter; we observed that subjects were able to modify their decisional criterion in order “to follow” the optimal position. Two main assumptions of our model are that the subject stores information coming from previous trials, and is willing of in improving his/her performance. One of the most important assumption of SDT is that the adopted strategy, i.e. the criterion positioning, by an observer performing a task is completely indipendent from his/her discrimination ability. We implemented this consideration in a model for criterion dynamics so that this parameter turns out to be completely indipendent from d′. The possibility to disentangle sensitivity and criterion allows the experimenter to force the subject’s inclination to be more liberal or conservative, indipendently from his/her ability in performing the task, and monitor at each trial the result of this conditioning. The problem of how the human neural system set and maintain a decision criterion over time is still an open question. This problem recently received particular attention, within the more general context of the neural mechanisms underlying the decision process. Our approach, based on behavioural experiments, provides an novel investigation tool to tackle the issue.