Visual pathways

In this dissertat ion, the early visual system is used to explore the role of efficiency in
the general organization of the nervous system. Efficient representation theory predicts
that neurons dynamically change their responses to changes in the environment in order to
maintain their efficiency. To directly test the predication of this theory, a computational
model and a neurophysiological experiment are used. Using a computational model, we investigate the sparseness of the response of filters at
each stage of the model of the visual pathway. We find that the temporal bandpass filter
and the rectification in each stage improves the efficiency of the response representation.
Moreover, we find that ON/nonlagged responses carry more information than OFF/ lagged
responses in signals with low signal-to-noise ratios. In the neurophysiological experiment, the response of LGN cells is measured and compared
to their input from the retina in awake cats during free-viewing of natural time-varying
images using quasi-intracellular recording technique. We find that the neural responses in
the retina and the LGN are efficient. However, the LGN response is more efficient, sparser and less correlated than the retina's response, and it carries less information about eye
movements than the retina's. As a result the LGN represents the visual world with fewer
spikes. The LGN response changes with the variation of visual input. The temporal correlation
of the visual input changes with saccade timing. Accordingly, the temporal receptive field of
the LGN also changes in order to maintain the decorrelation of the LGN response regardless
of the saccade. The retina-thalamic transmission changes during and after a saccade in order to transmit
useful information to the visual cortex and decreases during a saccade in order to eliminate
the variation of the visual input during a saccade. However, the transmission increases after
a saccade to facilitate the transmission of new information due to the new gaze direction in
the visual environment. The temporal receptive field of the LGN, derived from the efficacy of the thalamic
transmission, is causal and bimodal. Such a receptive field decorrelates the visual input
and improves the sparseness of the LGN response representation.

In the present study, it was examined whether the spatiotemporal dynamics of
transitions towards target dominance in motion-induced blindness (MIB) were wave-like,
similar to those in binocular rivalry. The spatiotemporal dynamics of transitions towards
dominance in MIB were further compared with those in binocular rivalry to reveal a
potential neural locus of MIB. Across a series of experiments, the relationship between
target length, stimulus structure, presentation location and the latency for circular arc
segment-shaped targets to reappear was examined, respectively. It was found that target
reappearance durations increase with target length, as if they reappear in a gradual, wavelike
fashion. Target reappearance durations were decreased for collinear compared to
radial targets, but they were not influenced by the location of target presentation. The
results suggest MIB target reappearances are associated with traveling waves of
dominance, and early visual cortex is a likely neural substrate in which these wave-like
transitions occur.

Prior research has explored the counterchange model of motion detection in terms of counterchanging information that originates in the stimulus foreground (or objects). These experiments explore counterchange apparent motion with regard to a new apparent motion stimulus where the necessary counterchanging information required for apparent motion is provided by altering the luminance of the background. It was found that apparent motion produced by background-counterchange requires longer frame durations and lower levels of average stimulus contrast compared to foreground-counterchange. Furthermore, inter-object distance does not influence apparent motion produced by background-counterchange to the degree it influences apparent motion produced by foreground-counterchange.

An array of four motion quartets (stimuli for which either horizontal or vertical motion is perceived depending on quartet aspect ratio) is arranged in a diamond configuration such that two global motion patterns are formed: (1) Rotation---alternating counterclockwise and clockwise motion is perceived, and (2) Parallel path motion---the perceived motions of all the elements are simultaneously horizontal or simultaneously vertical. The perception of rotation resulted in global feedback that biases the motion perceived for an individual component motion quartet to be more consistent with rotation than aspect ratio. Stronger rotation produced greater bias. Under certain conditions, the feedback-induced bias occurred even though global rotation was not perceived. The results were interpreted in the context of neurophysiological evidence regarding neurons in Areas MT and MSTd, and a dynamical theory of motion pattern formation (Hock, Schoner & Giese, 2003; Nichols, Hock & Schoner, 2006).

The interplexiform cells(IP cells) are the most recently discovered neurons in the retina and their function is to provide centrifugal feedback in retina. The anatomical structure of the IP cells has been well studied, but the function of these neurons is largely unknown. I systematically studied the excitatory and inhibitory inputs from IP cells in salamander retina. I found that L-EPSCs in IP cells are mediated by AMPA and NMDA receptors; in addition, L-IPSCs are mediated by glycine receptors and GABAC receptors. In response to light, IP cells reaction potentials transiently at the onset and onset of light stimulation. The major neural transmitter of IP cells in salamander retina is glycine. We also studied the distribution and function of glycine transporters. Our result indicates that GlyT1- and GlyT2-like transporters were present in Muller cells and neurons. The glycine feedback at outer plexiform layer (OPL) has effects on both the bipolar cell dendrites and rod photoreceptor terminals. At bipolar cell dendrites, glycine selectively depolarizes rod-dominant On-bipolar cells, and hyperpolarizes Off- bipolar cells. At rod photoreceptor terminals, 10 M glycine activates voltage-gated Ca2+ channels. These effects facilitated glutamate vesicle release in photoreceptors. It increases the sEPSC in OFF bipolar cells. The combined effect of glycine at rod terminals and bipolar cell dendrites leads to enhanced dim light signal transduction in the rod photoreceptor to ganglion cell pathway. This study provides a model that displays the function of centrifugal feedback through IP cells in the retina.

Although it is well known that the pupil responds dynamically to changes in ambient light levels, the results from this dissertation show for the first time that the pupil also responds dynamically to changes in spatially distributed attention. Using a variety of orientating tasks, subjects alternated between focusing attention on a central stimulus and spreading attention over a larger area. Fourier analysis of the fluctuating pupil diameter indicated that: 1) pupil diameter changed at the rate of attention variation, dilating with broadly spread attention and contracting with narrowly focused attention, and 2) pupillary differences required changes in attentional spread; there were no differences in pupil diameter between sustained broad and sustained spread attention. Given that broadly spread attention increases the relative activation of large receptive fields and narrowly focused attention increases the relative activation of small receptive fields (Balz & Hock, 1997), the results of this study indicate that these attentional effects on receptive field activation can be mediated by changes in pupil diameter. That is, under broad attention, the corresponding pupillary dilation observed would increase spherical aberration, blurring the image thereby reducing high spatial frequency information and decreasing the activation of relatively small cortical receptive fields compared to relatively large receptive fields. This increased perception of low spatial frequencies would be beneficial in cases where attention is spread over a large area. Alternatively, under narrow attention the resulting pupillary constriction reduces spherical aberration sharpening the image and preserving high spatial frequency information resulting in a relatively increased response of small receptive fields.