Visual attention has been extensively studied in psychology, but some fundamental questions
remain controversial. We focus on two questions in this study. First, we investigate how a
neuron in visual cortex responds to multiple stimuli inside the receptive eld, described by
either a response-averaging or a probability-mixing model. Second, we discuss how stimuli
are processed during visual search, explained by either a serial or a parallel mechanism.
Here we present novel mathematical methods to answer the psychology questions from a
neural perspective, combining the formulation of neural explanations for the visual attention
theories and spiking neuron models for single spike trains. Statistical inference and model
selection are performed and various numerical methods are explored. The designed methods
also give a framework for neural coding under visual attention theories. We conduct both
analysis on real data and theoretical study with simulations.
Our ndings are shown in separate projects. First, the probability-mixing model is favored
over the response-averaging model, shown by analysis on experimental data from monkeys.
Second, both parallel and serial processing exist, with a tendency of being parallel in the
beginning and a tendency of being serial later on, shown by another set of experimental data
from monkeys. Third, we show that the probability-mixing and response-averaging model
can be separated and parameters can be successfully estimated for either model in a more
realistic biophysical system, supported by simulation study. Finally, we present the decoding
of multiple temporal stimuli under these visual attention theories, also in a realistic biophysical
situation with simulations.