Simon Grondin - Psychology of Perception

Gustav Fechner

One could say that psychophysics started in 1860 with the publication of the book Elements of psychophysics by the German researcher Gustav Theodor Fechner (18011887). Philosopher and physicist, the founder of psychophysics wanted to study the links between the inner world and the outer world. Also known under the pseudonym of Dr. Mise, Fechner, who worked in Leipzig, had quite a special mind. We owe him various experimental methods still used in psychophysics, but he was also interested in, for example, the properties of the electric current, experimental aesthetics, and even life after death. Note that there is an annual meeting of psychophysics, usually held in October, called Fechner Day (October 22). This meeting is held in different locations around the world under the supervision of the International Society for Psychophysics ( http://www.ispsychophysics.org/ ), founded in 1985 in southern France.

For measuring an absolute threshold with the method of constant stimuli, also called the constant method, one must first determine the threshold roughly by locating a region for which a stimulus is almost never perceived and for which a stimulus is almost always perceived. Then, we generally select from five to nine stimuli located between these regions. After this selection, the selected stimuli are presented repeatedly in random order. The method requires an observer to make at least a hundred judgments, but of course, increasing the number of trials for estimating a threshold decreases the risk that the estimated value is far from what the real threshold is.

At each presentation, an observer has to indicate whether or not the stimulus is perceived. It becomes then possible to obtain a discrete (not continuous) frequency distribution, each point representing the number of times a stimulus was detected. These frequencies have to be transformed into probabilities. It is on the basis of these probabilities that the threshold value will be estimated. The probability calculated for each stimulus can be reported on a figure. As shown in Fig. , the percentage of times the stimulus is detected is placed on the y -axis and is plotted as a function of the magnitude of the stimuli, placed on the x -axis, in ascending order. The function that relates the probability of detecting to the magnitude of a physical continuum is called a psychometric function. Such a function generally has the shape of an ogivea kind of Sand the threshold is operationally defined as the point corresponding to an ability to perceive the stimulus 50% of the time. This value, 50%, represents the point for which an observer is able to detect the stimulus at a level higher than what would provide responses made randomly in a procedure involving two responses, yes or not.

For drawing a function on the basis of a series of points, it is necessary to posit some assumptions. First, the phenomenon under investigation is assumed to be a continuous random variable. Thus, we shall believe that the discrete distribution obtained (series of points) is an approximation of a continuous function. Also, it is necessary to make an assumption about the shape of this function. Mathematics offers several possibilities, but a function often used in psychology is the normal distribution. The reader is probably already familiar with the concept of normal distribution (normal or Gaussian curve or bell-shaped curve). The function used to draw a psychometric function is derived from the bell-shaped function (probability density function) and is called cumulative normal function. It is after drawing this function that it becomes possible to estimate the threshold value accurately. Besides the cumulative Gaussian function, Weibull and logistics functions are probably the most likely ones to be used (Macmillan & Creelman, ).

Despite the rigor used to estimate the ability to detect a stimulus with the constant stimuli method, a major problem may arise. The estimated capacity may depend not only on the sensitivity of an observer but also on the way in which this observer makes decisions. An observer might as well wait to be sure before making a decision, before declaring that a stimulus is perceived, whereas another observer, in spite of doubt, would tend to say yes, I perceive (Macmillan & Creelman, ).

There is a method, developed in the 1940s, to determine the sensitivity of the observer to detect a stimulus while correcting the problem associated with the involvement of decision making. Thus, the signal detection theory (SDT), also known as sensory decision theory, uses two parameters to describe the performance: one describing the sensitivity level and the other describing the way an observer makes a decision (Macmillan & Creelman, ).