Zhong-Lin Lu


(213)-740-2282
zhonglin@almaak.usc.edu

Welcome to my home page. For those who want to know more about me, I have included my Vita, Publications, Invited Talks, and Conference Presentations. If you would like to get in touch with me, feel free to send email to: <zhonglin@almaak.usc.edu>.


Research Statement

For an overview of some of my recent work in vision that is intended for nonspecialist audiences, see two articles: Nature, "Attention-generated apparent motion", 1995, and a review that is in press in Current Directions in Psychological Science, "Three mechanisms of human visual motion detection." The technical source paper was published in Vision Research. "The functional architecture of human visual motion perception".

Although the articles mentioned above concern primarily visual psychophysics, I am also working on attention. Some of this work is described in the Nature article, some in an article published in Attention and Performance XIV, "Using repetition detection to define and localize the processes of selective attention", 1992. Both of these lines of research are being pursued with graduate students, and there are very promising new results. For example, Sperling, Wurst and Lu found that in repetition detection knowing the size of the target characters does not increase detection efficiency. The same phenomena was observed by Shih and Sperling in perceptual search experiments. However, in research with graduate student D. Mcbridge, we are finding that there are actually large positive size effects in both search and repetition detection. The difference between our new results and the previous ones reflects channel bandwidth of the attentional systems. A mathematical model of attentional bandwidth is currently being developed to account jointly for both sets of results. The Nature article described only the initial study of the attention-sensitive third-order motion system. A graduate student, E. Blazer, and I are working on several interesting ramifications: How is the ability to attend to particular stimuli learned? and What are some of the limitations on attention?

I obtained my Ph.D at NYU in 1992 in low-temperature physics under Professor Samuel Williamson. My doctoral thesis utilized advanced physical techniques to measure magnetic fields generated by the human cerebral cortex. The 14 channel SQUID system allowed us to localize psychological processes to an accuracy of about 3-5 mm. The topics studied were: neuronal sources of alpha rhythm and auditory cognition. We established that alpha rhythm is formed by a parade of spatially well-defined local excitations. Our discovery that the reduction of alpha activity is so spatially localized makes it possible to study localized alpha blocking caused by attention processes in various perceptual and cognitive tasks (see "Study of human occipital alpha rhythm: the alphon hypothesis and alpha suppression", 1995). The neuromagentic research in human auditory cortex, for the first time, successfully localized, separated, and recorded signals from two simultaneously active brain regions: auditory primary and association cortex (see "Human auditory primary and association cortex have differing lifetimes for activation traces", 1992). Through study of the habituation properties of these brain regions and related psychophysical experiments, we discovered a remarkable brain correlate of auditory iconic memory in human auditory primary cortex. We actually found that the lifetime for decay of the neuronal activation trace in primary auditory cortex predicts the psychophysically determined duration of memory for the loudness of a tone. This work was published in Science: "Behavioral lifetime of human auditory sensory memory predicted by physiological measures", 1992. Indeed, this study of sensory memory is what led me to come to study and work with Professor George Sperling.

It is not practical at the moment to continue work in audition, but I am planning to resume work in neuromagnetism in relation to my work on visual motion perception with Professor George Sperling. We have identified three systems (which we call first-, second-, and third-order) that process luminance modulation motion, texture-contrast motion, and feature-salience motion, respectively. The first two systems are primarily monocular, so there are five sites that perform the initial processing of a normal moving object. We believe we have isolated each one and measured its temporal frequency characteristic. Neurophysiologists have been unable to locate the second-order system. And finding the third-order system, which also involves a general mechanism for attentional control, would be a great coup. No one knows how to look for it yet. These problems seem to be ideal for a neuromagnetism approach, especially if it can be done jointly with single cell recording. It would be especially interesting to localize and learn more about the attentional control processes, especially the attentional "salience map" described in the Nature paper.

In the course of doing my thesis at NYU, I was responsible for all the technical aspects of the neuromagnetism laboratory, and I feel confident that I could operate a Neuromagnetism laboratory. However, because of the huge expense involved, neuromagnetism research must initially be collaborative. The planned extensions to single-unit recording would be in collaboration with neurophysiologists who are already set up for this kind of work.

Since 1992, I have been a post-doctoral fellow/assistant researcher with Professor George Sperling. In addition to motion perception, we have worked on a variety of problems. Many deal with second-order phenomena. Second order means features replace photons, same processes that govern first-order (luminance) perception operate on the output of feature extraction. We have demonstrated second-order Mach bands, Chevreul and Craik-O'Brien-Cornsweet illusions and established that the processes that control these illusions involve fullwave (vs halfwave) rectification (see "Second-order illusions: Mach bands, Chevreul and Craik-O'Brien-Cornsweet", 1996). We also demonstrated the existence of second-order reversed phi, an early proof of existence of second- and higher-order motion computations (ARVO'96 abstract).

We are currently finishing a project on the Lincoln picture problem (ARVO'96 abstract), and on the question of whether there is a specialized detector that is specific for detecting the motion of flicking fields. With a graduate student, S. Richman, we showed that indeed there is. This result was so surprising that we spent almost a year doing various control experiments to rule out alternatives (ARVO'96).

A paper ("Contrast saturation in first- and second-order motion perception") on contrast gain control in motion systems was submitted to Journal of Optical Society of America in October. This is the first report of gain control in second-order motion. The paper presents a review and a coherent model of the different gain control mechanisms that precede the first- and second-order motion computations.

For the past several months, my main effort has been on a neural network problem that grew out of experiments on odd and even textures. How can a neural network, without guidance, acquire knowledge of textures in the environment, and more generally, of any environmental regularities. This transforms Gibson's search for environmental invariants from intuitive approach into an actual computation. Charles Chubb (a mathematical psychophysicist from Rugers University, currently visiting UC Irvine) and I are exploring a fascinating hypothesis: To detect structure, all a neural network needs to know is that the input is not random. We have developed some powerful tools for determining whether an input really is random and, when it is not, for tuning a network to the latent structure in the input. One satisfying aspect of this work is that I have been able to apply methods from mathematical physics to derive mathematical properties of such networks. To actually see how such networks operate in "real" environments, simulation is necessary. So far, we have applied our network only to a set of 10,000 "natural" texture images selected from photographs of natural scenes. The network learned to successfully extract structures from natural images with a success rate of 95%. As a result, the neural network develops simple cell-like receptive fields. It is probable that receptive fields are innate, not learned, but the same regularities in nature that can tune neural networks are likely also to tune genes. Gibson was unable, and indeed any human would have been unable, to say what the optimally efficient description of a natural environment might be. But a neural network can be designed to automatically compute such a description. And evolution ensures that when it is sufficiently universal the genes discover it. This preliminary work is described briefly in an abstract (ARVO'96); a manuscript is in preparation.

Psychophysical experiments, physiological investigation, and mathematical plus computational modeling are all essential ingredients in understanding how the mind and brain operate. In each of these domains, modern techniques are quite complex, and there is too much for one person to do. Collaboration is necessary. I believe I have acquired the skills to be able to organize collaborative research with students and with faculty colleagues.


Selected Publications:

Zhong-Lin Lu and Samuel J. Williamson (1991) Spatial extent of coherent sensory-evoked cortical activity. Experimental Brain Research, 84: 411-416.

Samuel J. Williamson, Zhong-Lin Lu, Daniel Karron and Lloyd Kaufman (1991) Advantages and limitations of magnetic source imaging. Special Issue: Functional localization with EEG and MEG: Comparative aspects. Brain Topography, 4: 169-180.

Zhong-Lin Lu, Samuel J. Williamson and Lloyd Kaufman (1992) Human Auditory Primary and Association Cortex Have Differing Lifetimes for Activation Traces. Brain Research, 572: 236-241.

Zhong-Lin Lu, Samuel J. Williamson and Lloyd Kaufman (1992) Physiological Measures Predict Behavioral Lifetime of Human Auditory Sensory Memory. Science, 258: 1668-1670.

Zhong-Lin Lu (1993) Magnetic Source Imaging of the Human Brain. Proceedings of physiological imaging, spectroscopy, and early-detection diagnostic methods, edited by Randall L. Barbour, Mark J. Carvlin and Abraham Katzir. SPIE proceeding series, vol 1887, page 2-15. Bellimgham, Washington.

George Sperling, Stephens A. Wurst and Zhong-Lin Lu (1993) Using Repetition Detection to Define and Localize the Processes of Selective Attention. Attention and Performance XIV, edited by David E. Meyer and Sylvan Kornblum. Cambridge, Massachusetts: The MIT Press, Page 265-298.

George Sperling, Charles Chubb, Joshua Solomon and Zhong-Lin Lu (1994) Visual preprocessing: first and second order processes in the perception of motion and texture. In: Computational Intelligence: Imitating Life, edited by Zurada, J. M., Marks II, R. J. & Robinson, C. J., New York: IEEE Press, Page 223-236.

George Sperling, Charles Chubb, Joshua Solomon and Zhong-Lin Lu (1994) Fullwave and halfwave processes in 2nd-order motion and texture. In: Higher-order processing in the visual system. Wiley, Chichester (Ciba Found Symp 184), page: 287-308.

Zhong-Lin Lu, Jia-Zhu Wang, Daniel Karron and Samuel J. Williamson (1995) Support for the alphon hypothesis as a description of human alpha rhythm. In: Biomagnetism: Fundamental Research and Clinical Applications, Proceedings of the 9th International Conference on Biomagnetism, Studies in Applied Electromagnetics and Mechanics, Vol. 7, Baumgartner, C., Deecke, L., Stroink, G. & Williamson, S.J., Eds. (Elsevier and IOS Press, Amsterdam), page: 295-298.

Samuel J. Williamson and Zhong-Lin Lu (1995) Habituation and sensory Memory. In: Quantitative and Topological EEG and MEG Analysis, edited by Eiselt, M., Zwiener, U. & Witte, H., Universitoatsverlag Druckhaus-Mayer GmbH, Jena, page: 19-26.

Samuel J. Williamson, Lloyd Kaufman, Zhong-Lin Lu, Jia-Zhu Wang and Daniel Karron (1995) Study of human occipital alpha rhythm: the alphon hypothesis and alpha suppression. In: Alpha Activity: Cognitive and Sensory Behavior, Basar, E., Ed. Birkhauser, Boston, MA

Zhong-Lin Lu and George Sperling (1995) The Functional Architecture of Human Visual Motion Perception. Vision Research, 35: 2697-2722.

Zhong-Lin Lu and George Sperling (1995) Attention-Generated Motion Perception. Nature, 377: 237-239.

Zhong-Lin Lu and George Sperling (1995) Three mechanisms of human visual motion detection. Current Directions of Psychological Science, in press.

Zhong-Lin Lu and George Sperling (1995) Second-order illusions: Mach bands, Chevreal and Craik-O'Brien-Cornsweet. Vision Research, 35, in press.

Samuel J. Williamson, Lloyd Kaufman, Sally Curtis, Zhong-Lin Lu, Christoph Michel and Jia-Zhu Wang (1995) Neural substrates of working memory are revealed magnetically by the local suppr ession of alpha rhythm. EEG Journal, in press.

Zhong-Lin Lu, and George Sperling (1995) Contrast saturation in first- and second-order motion perception. Technical report MBS 95-25 , Institute for Mathematical Behavioral Sciences, University of California, Irvine. Journal of Optical Society of America A , submitted.

Zhong-Lin Lu, and George Sperling (1995) The existence of second-order reversed phi: A prediction from a theory of motion perception. Perception and Psychophysics, submitted.


<zhonglin@mach2.hipl.uci.edu>
December 28, 1995 at 9:01 AM