Handbook of Psychology, Volume 4: Experimental Psychology

108 Foundations of Visual Perception

(LTM). To determine the LTM, Barraza and Colombo (2001)
showed the observers two gratings in succession. One was
drifting to the right, and the other was drifting to the left. The
observer had to report whether the first or the second interval
contained the leftward-drifting grating. Such tasks are called
forced-choice tasks.More specifically, this is an instance of
a temporal two-alternative forced-choice task (2AFC; to
learn more about forced-choice designs, see Macmillan &
Creelman, 1991, chap. 5, and Hartmann, 1998, chap. 24).
To simulate the effect of glare, Barraza and Colombo
(2001) used an incandescent lamp located 10° away from the
observer’s line of sight. On each trial, they first turned on the
glare stimulus, and then after a predetermined interval of
time, they showed the drifting grating. Because neither the
glare stimulus nor the grating had an abrupt onset, they de-
fined the effective onsetof each as the moment at which the
stimulus reached a certain proportion of its maximum effec-
tiveness (as shown in Figure 4.14). The time interval between
the onset of two stimuli is called stimulus-onset asynchrony
(SOA). In this experiment the SOA between the glare stimu-
lus and the drifting grating took on one of five values: 50,
150, 250, 350, or 450 ms.
Barraza and Colombo (2001) were particularly interested
in determining whether the moments just after the glare stim-
ulus was turned on were the ones at which the glare was the
most detrimental to the detection of motion (i.e., it caused
the LTM to rise). To measure the LTM for each condition,
they used the method of constant stimuli:They presented the
gratings repeatedly at a given drift velocity so that they could

estimate the probability that the observer could discriminate

between left- and right-drifting gratings.

To calculate the LTM, they plotted the proportion of cor-

rect responses for a given SOA as a function of the rate at

which the grating drifted (Figure 4.15, top panel). They then

fitted a Weibull function to these data and determined the

LTM by finding the grating velocity that corresponded to

80% correct responses (dashed lines). Although there is no

substitute for publishing the best-fitting normal, logistic, or

Weibull distribution function to such data (using logistic re-

gressionfor a logistic distribution or a probit modelfor the

normal; Agresti, 1996), the easiest way to look at such data is

to transform the percentage of correct data into log odds. Let

us denote motion frequency by f and the corresponding

proportion of correct responses by (f). We plot the log-odds

of being right (using the natural logarithm, denoted by ln) as

a function of f. In other words, we fit a linear function,

ln 1 ^ ( f)(f)=+f, to the data obtained. Figure 4.15, bot-

tom panel, shows the results. Fitting the linear regression

does not require specialized software, and the results are

usually close to estimates obtained with more complex fitting

routines.

Adaptive Methods

Adaptive methods combine the best features of the method

of limits and forced-choice procedures. Instead of exploring

the response to many levels of the independent variable, as in

the method of constant stimuli, adaptive methods quickly

Stimulus contrast modulation

Contrast

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Effective glare onset

Glare

intensity

Time (s)

SOA = 0.250 s

Effective

stimulus

duration

Figure 4.14 Scheme of presentation of glare and test stimulus in a trial for a 250-ms value of SOA. After

Barraza and Colombo (2001, Figure 1).