Denis G. Pelli

The VideoToolbox software for visual psychophysics: transforming numbers into movies.

The VideoToolbox is a free collection of two hundred C subroutines for Macintosh computers that calibrates and controls the computer-display interface to create accurately specified visual stimuli. High-level platform-independent languages like MATLAB are best for creating the numbers that describe the desired images. Low-level, computer-specific VideoToolbox routines control the hardware that transforms those numbers into a movie. Transcending the particular computer and language, we discuss the nature of the computer-display interface, and how to calibrate and control it.

Quest: A Bayesian adaptive psychometric method

An adaptive psychometric procedure that places each trial at the current most probable Bayesian estimate of threshold is described. The procedure takes advantage of the common finding that the human psychometric function is invariant in form when expressed as a function of log intensity. The procedure is simple, fast, and efficient, and may be easily implemented on any computer.

Accurate control of contrast on microcomputer displays

Off-the-shelf microcomputers can now display arbitrary 8-bit images, but accurate control of these images requires dealing with several undesirable properties of real digital to analog converters (DACs) and analog video monitors. The limitations of DACs and video monitors are presented in the form of a model for their calibration and use in vision experiments. Low contrasts can be accurately rendered by summing a small accurate a.c. signal and a large less-accurate d.c. signal (Watson et al., 1986; Behavior Research, Method, Instrument and Computers, 18, 587-594). Exploiting that idea, this note presents an easy-to-build passive resistor network, a video attenuator, that combines the outputs of three 8-bit DACs to render low contrasts with 12-bit accuracy at the display. Measurements confirm the 12-bit accuracy.

Psychophysics of reading-I. Normal vision

This paper is about the visual requirements for reading with normal vision. It is the first in a series devoted to the psychophysics of reading with normal and low vision. We have measured reading rates for text scanned across the face of a TV monitor while varying parameters that are important in current theories of pattern vision. Our results provide estimates of the stimulus parameters required for optimal reading of scanned text. We have found that maximum reading rates are achieved for characters subtending 0.3 degree to 2 degrees. Contrast polarity (black-on-white vs white-on-black text) has no effect. Reading rate increases with field size, but only up to 4 characters, independent of character size. When text is low-pass spatial-frequency filtered, reading rate increases with bandwidth, but only up to two cycles/character, independent of character size. When text is matrix sampled, reading rate increases with sample density, but only up to a critical sample density which depends on character size. The critical sample density increases from about 4 X 4 samples/character for 0.1 degree characters to more than 20 X 20 samples/character for 24 degrees characters. We suggest that one spatial-frequency channel suffices for reading.

Psychophysics of reading--II. Low vision.

Very little is known about the effects of visual impairment on reading. We used psychophysical methods to study reading by 16 low-vision observers. Reading rates were measured for text scanned across the face of a TV monitor while varying parameters that are likely to be important in low vision: angular character size, number of characters in the field, number of dots composing each character, contrast polarity (white-on-black vs black-on-white text), and character spacing. Despite diverse pathologies and degrees of vision loss in our sample, several major generalizations emerged. There is a wide variation in peak reading rates among low-vision observers, but 64% of the variance can be accounted for by two major distinctions: intact central fields vs central-field loss and cloudy vs clear ocular media. Peak reading rates for observers with central-field loss were very low (median 25 words/minute), while peak reading rates for observers with intact central fields were at least 90 words/minute (median 130 words/minute). Most low-vision readers require magnification to obtain characters of optimal size. Sloan M acuity was a better predictor of optimal character size than Snellen acuity, accounting for 72% of the variance. Low-vision reading is similar to normal reading in several respects. For example, both show the same dependence on the number of characters in the field. Our results provide estimates of the best reading performance to be expected from low-vision observers with characteristic forms of vision loss, and the stimulus parameters necessary for optimal performance. These results will be useful in the development of clinical tests of low vision, and in the design of low-vision reading aids.

Pixel independence: measuring spatial interactions on a CRT display.

The standard working assumption of careful CRT imaging is that each pixel is imaged independently, through a point nonlinearity (the monitors gamma function, relating screen luminance to input voltage), and then blurred by the point-spread function of the beam spot on the phosphor. Unfortunately most monitors have inadequate video bandwidth, DC restoration, and high-voltage regulation to live up to this ideal model. Two tests are recommended for assessing a CRTs deviation from the pixel-independence model.

The remarkable inefficiency of word recognition

Do we recognize common objects by parts, or as wholes? Holistic recognition would be efficient, yet people detect a grating of light and dark stripes by parts. Thus efficiency falls as the number of stripes increases, in inverse proportion, as explained by probability summation among independent feature detectors. It is inefficient to detect correlated components independently. But gratings are uncommon artificial stimuli that may fail to tap the full power of visual object recognition. Familiar objects become special as people become expert at judging them, possibly because the processing becomes more holistic. Letters and words were designed to be easily recognized, and, through a lifetime of reading, our visual system presumably has adapted to do this as well as it possibly can. Here we show that in identifying familiar English words, even the five most common three-letter words, observers have the handicap predicted by recognition by parts: a word is unreadable unless its letters are separately identifiable. Efficiency is inversely proportional to word length, independent of how many possible words (5, 26 or thousands) the test word is drawn from. Human performance never exceeds that attainable by strictly letter- or feature-based models. Thus, everything seen is a pattern of features. Despite our virtuosity at recognizing patterns and our expertise from reading a billion letters, we never learn to see a word as a feature; our efficiency is limited by the bottleneck of having to rigorously and independently detect simple features.

The role of spatial frequency channels in letter identification

How we see is today explained by physical optics and retinal transduction, followed by feature detection, in the cortex, by a bank of parallel independent spatial-frequency-selective channels. It is assumed that the observer uses whichever channels are best for the task at hand. Our current results demand a revision of this framework: Observers are not free to choose which channels they use. We used critical-band masking to characterize the channels mediating identification of broadband signals: letters in a wide range of fonts (Sloan, Bookman, Künstler, Yung), alphabets (Roman and Chinese), and sizes (0.1-55 degrees ). We also tested sinewave and squarewave gratings. Masking always revealed a single channel, 1.6+/-0.7 octaves wide, with a center frequency that depends on letter size and alphabet. We define an alphabets stroke frequency as the average number of lines crossed by a slice through a letter, divided by the letter width. For sharp-edged (i.e. broadband) signals, we find that stroke frequency completely determines channel frequency, independent of alphabet, font, and size. Moreover, even though observers have multiple channels, they always use the same channel for the same signals, even after hundreds of trials, regardless of whether the noise is low-pass, high-pass, or all-pass. This shows that observers identify letters through a single channel that is selected bottom-up, by the signal, not top-down by the observer. We thought shape would be processed similarly at all sizes. Bandlimited signals conform more to this expectation than do broadband signals. Here, we characterize processing by channel frequency. For sinewave gratings, as expected, channel frequency equals sinewave frequency f(channel)=f. For bandpass-filtered letters, channel frequency is proportional to center frequency f(channel) proportional, variantf(center) (log-log slope 1) when size is varied and the band (c/letter) is fixed, but channel frequency is less than proportional to center frequency f(channel) proportional, variantf(center)(2/3) (log-log slope 2/3) when the band is varied and size is fixed. Finally, our main result, for sharp-edged (i.e. broadband) letters and squarewaves, channel frequency depends solely on stroke frequency, f(channel)/10c/deg=(2/3), with a log-log slope of 2/3. Thus, large letters (and coarse squarewaves) are identified by their edges; small letters (and fine squarewaves) are identified by their gross strokes.

Are faces processed like words? A diagnostic test for recognition by parts

Do we identify an object as a whole or by its parts? This simple question has been surprisingly hard to answer. It has been suggested that faces are recognized as wholes and words are recognized by parts. Here we answer the question by applying a test for crowding. In crowding, a target is harder to identify in the presence of nearby flankers. Previous work has described crowding between objects. We show that crowding also occurs between the parts of an object. Such internal crowding severely impairs perception, identification, and fMRI face-area activation. We apply a diagnostic test for crowding to a word and a face, and we find that the critical spacing of the parts required for recognition is proportional to distance from fixation and independent of size and kind. The critical spacing defines an isolation field around the target. Some objects can be recognized only when each part is isolated from the rest of the object by the critical spacing. In that case, recognition is by parts. Recognition is holistic if the observer can recognize the object even when the whole object fits within a critical spacing. Such an object has only one part. Multiple parts within an isolation field will crowd each other and spoil recognition. To assess the robustness of the crowding test, we manipulated familiarity through inversion and the face- and word-superiority effects. We find that threshold contrast for word and face identification is the product of two factors: familiarity and crowding. Familiarity increases sensitivity by a factor of x1.5, independent of eccentricity, while crowding attenuates sensitivity more and more as eccentricity increases. Our findings show that observers process words and faces in much the same way: The effects of familiarity and crowding do not distinguish between them. Words and faces are both recognized by parts, and their parts -- letters and facial features -- are recognized holistically. We propose that internal crowding be taken as the signature of recognition by parts.

The Information Capacity of Visual Attention

Is visual attention mediated by a general-purpose processor with a small data capacity? Such an attentive processor could perform a wide range of transformations upon a small amount of image data. We suggest that this limited capacity corresponds to a fixed amount of information, measured in bits. We measure how much information an observers attention can handle by measuring how much we can restrict display information without impairing the observers performance. The attentive visual tasks we study are the detection of a stationary dot in a field of moving dots, and the detection of a static square in a field of flashing squares. Performance of these tasks is perfect up to a critical number of elements (the span of attention) and then falls as the number of elements increases beyond this critical number. The display information required for unimpaired performance in each of these tasks is low; the results indicate that visual attention processes only 30 to 60 bits of display information.