Perceptual Coherence : Hearing and Seeing

242 Perceptual Coherence
estimate of the minimum number of quanta necessary to see. Cornsweet
(1970) beautifully summarized this work. Hecht et al. estimated that under
the optimum conditions, humans could see as few as 5–9 absorbed quanta.
Due to reflection off the cornea and absorption in the eye itself, roughly
100 quanta are necessary to yield the 5–9 quanta at the retina that excite retinal
cells. (These values are for light presented off-center in the eye where
the rod density is highest. The minimum number of quanta is roughly
five times greater for light presented to the cone fovea.) These quanta were
spread over a spatial summation region in the retina composed of about 300
rods, so that it seemed that one quantum is sufficient to trigger a receptor.
This assumption was confirmed by electrophysiological recordings showing
that neural responses could be recorded due to the absorption of a
single quanta of light (Baylor, Lamb, & Yau, 1979).
Absolute thresholds always are defined statistically—60% detection here.
Hecht et al. (1942) attempted to identify any source that could lead to variation
in the subject’s response. Was it due to variation in the light or in the subject?
From a decision theory perspective, we can assume that subjects report
seeing the dim light when the number of spikes in the trial interval exceeds
some fixed number. But, on any trial, the number of quanta actually emitted
by the light source can vary; the emitted quanta may not reach the retina; and
any quanta reaching the retina may not stimulate a cell. Whatever the criteria,
there are going to be instances when the number of spikes does not reach criterion
when the stimulus was presented, and conversely there are going to be
instances when the variability of the firing patterns within the visual system
generates the required number of spikes even when the dim light is not presented.
Hecht et al. concluded that any variation in the detection of the light
flash was due to the inherent variability of the light source itself and resulting
variability in quanta absorbed by the retinal receptors, rather than some internal
biological or cognitive randomness. The finding that spontaneous firings
of the retinal cells that could result in a false detection of light are amazingly
infrequent supports this conclusion. It has been estimated that the firing of
rods in the primate eye yielding a false alarm due to the spontaneous transformations
of rhodopsin is less than one every 100–160 s (Aho, Donner, Hyden,
Larsen, & Reuter, 1988; Lamb, 1987).
Nonetheless, a completely dark-adapted human observer will see flickering
specks and flashes that occur randomly in a dark field. Barlow (1956)
termed the lightness level resulting from the spontaneous firing dark noise
and conceptualized that it operated in the same way as any physical light.
Thus, the effective background for perceiving is the actual background
light (however dim) plus the dark noise, and together they set limits for the
detection of dim light flashes. In a decision theory coneption, observers
must set their criteria in terms of the variability of the signal and noise (i.e.,