Binocular vision: two possible central interactions between signals from two eyes.

Both foveae of light-adapted subjects were stimulated at the same time with monocularly presented lights of increasing or decreasing luminance. Combinations judged just detectable violated predictions of the energy summation and the probability summation hypotheses of binocular interaction. Rather, the results can be explained by independent central neural mechanisms that signal the sum or the difference of stimuli to two eyes.

Receiver operating characteristic analysis. Application to the study of quantum fluctuation effects in optic nerve of Rana pipiens.

Receiver operating characteristic (ROC) analysis of nerve messages is described. The hypothesis that quantum fluctuations provide the only limit to the ability of frog ganglion cells to signal luminance change information is examined using ROC analysis. In the context of ROC analysis, the quantum fluctuation hypothesis predicts (a) the detectability of a luminance change signal should rise proportionally to the size of the change, (b) detectability should decrease as the square root of background, an implication of which is the deVries-Rose law, and (c) ROC curves should exhibit a shape particular to underlying Poisson distributions. Each of these predictions is confirmed for the responses of dimming ganglion cells to brief luminance decrements at scotopic levels, but none could have been tested using classical nerve message analysis procedures.

Performance of cat retinal ganglion cells at low light levels.

Responses of brisk-sustained cat retinal ganglion cells were examined using receiver operating characteristic (ROC) analysis. Stimuli were brief luminance changes superimposed upon a weak steady pedestal ranging from 27 to 47,000 quanta (507 nm) per second at the cornea. Overall quantum efficiencies of cells ranged up to approximately 13% and were compatible with previous estimates at absolute threshold. The main work was done on on-center cells, but a small sample of off-center units behaved similarly. Experimental ROC curves verified a set of qualitative predictions based on a theoretical treatment of performance, assuming that response variability resulted solely from quantum fluctuations. However, quantitative predictions were not fulfilled. The discrepancy could be resolved by postulating a source of added internal variance, R, the value of which could then be deduced from the experimental measurements. A ganglion cell model limited by a fixed amount of added variance from physiological sources and having access to a fixed fraction of incident quanta can account quantitatively for (a) slopes of ROC curves, (b) variation of detectability with magnitude of both increments and decrements, and (c) performance over a range of pedestal intensities. Estimates of the proportion of incident quanta used ranged up to 29% under some conditions, a figure approximately matching estimates of the fraction of corneal quanta that isomerize rhodopsin in the cat.

Relative efficiency for the detection of apparent motion.

We measured the relative efficiency for motion and position discriminations of brief, localized spot stimuli with a technique that makes no assumptions about sites of noise or information loss in the visual system. In one task, the observer had to discriminate whether an increment was located at one (left) or another (right) closely spaced spots. In the other task, the observer had to discriminate two successive brief increments of the left spot from a left spot increment followed by a right spot increment. Ideal observer theory predicts identical performance on the two tasks. Observers' thresholds, however, were significantly lower in the motion task at all intervals between flashes (ISIs) less than 60 msec in one observer and all ISIs less than 150 msec in two other observers (P < 0.01, t-test). We conclude that this apparent motion stimulus is seen more efficiently than a non-moving stimulus, and that the higher efficiency may be due to use of a motion sensitive channel in addition to independent position sensitive channels.

Postural stability and stereo-ambiguity in man-designed visual environments.

Our modern rectilinear visual environment contains visual stimuli for which evolution has not had time to optimally shape visual processing. One such stimulus, periodic stripes, is known to lead to visual depth ambiguity. In this paper we show that postural instability, as measured by the variance of fore and aft sway, is increased by viewing such stimuli. This instability may be the precursor of falls. Designers must evaluate the visual impressions conveyed by their systems in order to avoid postural instability due to visual ambiguity.

Visual depth illusion and falls in the elderly.

This article investigates the relation of vision and the effects of age on the maintenance of posture. This relationship in the elderly is explored within the context of visual depth illusions induced by repeating patterns that occur on escalator treads and elsewhere in the environment. Age does not appear to reduce the susceptibility of the elderly to visual depth illusions. However, if age is coupled with declines in motor control and strength, the elderly are probably more susceptible to falls.

Are rods and cones of dyslexics anomalous?

Grosser and Spafford (in this journal, 1989) have advanced an hypothesis and presented measurements which they believe support the idea of an excess of cones in the peripheral retinae of dyslexics. This note points out that their hypothesis is based on the erroneous assumption that normals have no peripheral cones. Further, their data can be explained by at least two alternative, though uninteresting, methodological hypotheses, that uncontrolled eye movements or experimenter suggestion (or both) could have produced their results. Finally, the requisite methods for assessing color vision, and the cones, were not met in the study.

The universe of activity that is bioengineering research.

What is bioengineering? A concise and meaningful answer to this question is important for pedagogy. This paper demonstrates that research activity within 'bioengineering' resides in a multidimensional space. A specific study can be characterized by an area into which it falls, a problem that it attacks, a specific level of analysis, a species of focus, pertinent disciplines, relevant tools that are employed, and temporal features. Each represents a dimension of the universe and each dimension may be shown to be largely orthogonal to every other dimension. The universe thus defined is sparsely populated. Accordingly, bioengineering is a field of rich opportunity within which frontiers lie almost everywhere.

Detection of 1-of-M orthogonal signals: asymptotic equivalence of likelihood ratio and multiband models.

A closed-form solution for the optimal detection of 1-of-M orthogonal signals is presented for low signal energy. The solution is obtained for the Poisson case and applies as well in the Gaussian equal-variance case. Detectability, d', is given by S/(MB)((1/2)), where S is the signal energy (in photons) and B is the background. This result is obtained by demonstrating that the decision rule of the likelihood ratio (optimal) detector is identical to that of the multiband detector, which sums photon counts for all M signal loci.

Saccadic suppression and stimulus uncertainty.

Saccadic suppression is a decline in detectability of a weak flash presented during a saccadic eye movement. We examined the hypothesis of Matin [Psychol. Bull. 81, 899 (1974)] that saccadic suppression may be due to increased stimulus uncertainty during the saccade. Uncertainty could arise from variability and inhomogeneities in the visual frame of reference translation that must accompany a saccade. We measured an average 0.6-log-unit suppression for a brief foveal 1 degree flash in a light-adapted detection task. Receiver-operating-characteristic (ROC) slopes for flash detection during saccades, compared with those when fixating, were reduced, indicating the presence of increased uncertainty. The magnitude of this uncertainty change was estimated and found to be consistent with that required to account for the measured detectability decline. When a flashed pedestal was employed to reduce the effect of uncertainty, there was no saccadic suppression and no ROC slope change. Also, spatially separate flashed markers, intended to reduce uncertainty, led to a significant reduction in saccadic suppression for one of two subjects. Our results are consistent with the hypothesis that a saccade leaves the observer with increased uncertainty as to which subjective visual direction to attend for a stimulus of fixed retinal locus. The magnitude of this uncertainty change can account fully for the saccadic suppression measured.

Effect of chromatic uncertainty on detectability of a visual stimulus.

The detectability of a brief color shift towards red or towards green of a foveally viewed yellow target is less by 25% if the observer is uncertain as to the direction of the color shift. This result matches a prediction of the theory of signal detectability: When signal parameters become uncertain, detectability declines.

Effect of large spatial uncertainty on foveal luminance increment detectability.

In principle, the ability to detect a luminance increment is lowered when there is uncertainty for its spatial location. Frequency-of-seeing curves were generated for small foveal targets. When fixed in space the target's detectability was more than 10 times higher than when it could occur at one of 140 locations.

Flash masking and facilitation by nearby luminance perturbations.

Light-adapted foveal luminance increment thresholds were measured for white photopic targets of 1.5-arc min diameter and 220-ms duration. We aimed to learn about the properties of mechanisms that subserve the detection of these targets. To study this subject we developed a noise probe technique that inserts noise close to the site of the stimulus. Threshold is more than doubled when zero-mean luminance noise is placed at a pair of flanking spots in the horizontal meridian centered on the test spot and 1.5 arc min distant. The detection mechanisms thus has a broad field, since noise effects persist at 5-arc min separation. The masking effect increases when the noise is in antiphase at the two flanking spots. Neither even- nor odd-symmetric mechanisms are able to explain these findings, regardless of whether linear or nonlinear processing is employed. The target detection may be mediated in part by a motion-sensitive mechanism.

Site of the accelerating nonlinearity underlying luminance-change detection.

The visual psychometric function for luminance-increment detectability (d') is known to be an accelerating function of stimulus energy. Two different models have been suggested to explain this fact and also why a luminance-increment pedestal linearizes the psychometric function. In the present experiment it is shown that a dichoptically presented increment linearizes the psychometric function and facilitates detection of weak signals. Since the dichoptic pedestal combines with the signal centrally, the nonlinearity must then originate more centrally. This result is compatible with the uncertainty model of the nonlinear psychometric function but not with a model that requires a nonlinear transducer.

Detection and recognition of visual targets.

To aid the development of detection-based interpretations of visual resolution, we evaluated theorems that (1) relate observers' performance in detection of a single target to that in 1-of-m signal detection and (2) predict recognition performance from measured performance at 1-of-m detection. These theorems require that the sensory effects of the stimuli be continuous and that the m signals be equally detectable and mutually orthogonal. To evaluate the theorems, we tested observers in simple detection, 1-of-m detection, and recognition of Landolt C targets and compared predicted and observed performance. Predictions of 1-of-m detection performance from that for simple detection and predictions of recognition from the 1-of-m receiver operating characteristic (ROC) were both accurate to within 0.03 in P(A). In addition, predictions of recognition based on the 1-of-m ROC predicted from simple detection were generally accurate. Thus, under restricted experimental conditions, recognition is determined completely by that for simple detection.