Committee in charge:

Professor Paul M. Churchland, Chair
Professor Patricia Smith Churchland
Professor Garrison Cottrell
Professor Adrian Cussins
Professor Robert Hecht-Nielsen
Professor Vilayanur Ramachandran

The dissertation of Rick Grush is approved, and it is acceptable in quality and form for publication on microfilm:

_____________________________________________________

_____________________________________________________

_____________________________________________________

_____________________________________________________

_____________________________________________________

_____________________________________________________
Chair

University of California, San Diego

1995

Signature Page iii
Table of Contents iv
List of Figures vi
Acknowledgments viii
Vita, Publications and Fields of Study x
Abstract xi

Chapter One: Preliminaries 1
1.1 Introduction 1
1.2 Brief history of this project 3
1.3 Survey of the current project 5

Chapter Two: Emulation and Control 11
2.1 A control thought-experiment 11
2.2 Inverse vs. forward mappings 13
2.3 What is an emulator 15
2.4 Benefits of emulation 19
2.5 Emulation and neural networks 23

Chapter Three: Perception, Imagery, and the Sensorimotor Loop 29
3.1 Motor control 29
3.2 Motor imagery 36
3.3 Visual imagery 39
3.4 Perception and closed-loop imagery 46
3.5 Conclusion 49

Chapter Four: Emulation, Representation and Learning 50
4.1 Dreaming, pretense, and altered control loops 50
4.2 Representational redescription 51
4.3 Ways of emulator making 56
4.4 What can ETM say about development? 62
4.5 Parameters, distributed representations and semi-locality 66
4.6 Concluding remarks 68

Chapter Five: The Mind-Body Solution 70
5.1 How the folk get their psychology 71
5.2 Josef Perner: Modeling models 72
5.3 Further thoughts on Leslie, Wellman and the simulation theory 78
5.4 Emulating emulators 81
5.5 Conclusion 82

Chapter Six: The Grammar of Thought 84
6.1 Innateness and syntactic autonomy 85
6.2 Anaphora and extraction 88
6.3 Outline of the GB account 93
6.4 Cognitive grammar 106
6.5 Binding 112
6.6 Temporal characteristics of language processing 119
6.7 What all of this buys us 122
6.8 Conclusion 132

Chapter Seven: Semantics 134
7.1 Neo-Fregean theories of meaning 135
7.2 X-role semantics 141
7.3 Towards a positive theory 146
7.4 Interpretational semantics 148
7.5 Problems with interpretational semantics 153
7.6 The general interpretational semantic theory (GIST) 154
7.7 The prospects for communication 171
7.8 The role of the interpreter 172
7.9 Conclusion 174

References 176

Figure 2.1: Robot arm and control console. 11
Figure 2.2: Schematic of control loop. 13
Figure 2.3: Forward and inverse mappings. 15
Figure 2.4: Controlling the plant and emulator in parallel. 16
Figure 2.5: Enhanced control console. 17
Figure 2.6: Bioreactor. 24
Figure 2.7: Neural controller for the bioreactor. 25
Figure 2.8: Training the emulator. 26
Figure 2.9: Training the neural controller with an emulator. 27
Figure 3.1: Pseudo-closed-loop control. 31
Figure 3.2: Schematic of musculoskeletal emulator. 33
Figure 3.3: Left and right eye views of a tetrahedron. 40
Figure 3.4: 'Contextron' processing unit. 41
Figure 3.5: Effects of movement on retinal projection. 42
Figure 3.6 Real and imagined rotation. 43
Figure 3.7: Sensorimotor feedback apparatus. 44
Figure 4.1: Plot of some dynamic variables during a typical arm motion. 59
Figure 4.2: Schematic mapping between physical parameters and analog model. 60
Figure 4.3: Blocks from the block balancing task. 62
Figure 5.1: Schematic of brain emulating the external world. 81
Figure 5.2: Schematic of brain emulating an agent as emulating the world. 82
Figure 6.1: 'Jerry understands himself.' 94
Figure 6.2: 'Jerry thinks everyone understands himself.' 95
Figure 6.3: 'The woman with the hat ate salmon.' 99
Figure 6.4: 'What did the woman with eat salmon?' 100
Figure 6.5: 'What did the woman with the hat eat?' 101
Figure 6.6: 'I drink two rum and cokes with a twist of lemon every day.' 102
Figure 6.7: 'I drink every day two rum and cokes with a twist of lemon.' 103
Figure 6.8: Schematic of 'on.' 109
Figure 6.9: 'Cat on mat.' 109
Figure 6.10: 'Go,' 'gone' and 'away.' 110
Figure 6.11: Hierarchical scene segmentation. 123
Figure 7.1: Cummins' Tower Bridge. 151
Figure 7.2: Generalized tower bridge. 156
Figure 7.3: Tower bridge with non-real target domain. 158
Figure 7.4: Two interpretive projects. 160
Figure 7.5: Interpretation of Twin Earth example. 162
Figure 7.6: Imperfect interpretation and opacity. 170

In addition to the great intellectual debts I owe to many researchers in a number of fields, there are a number of people who have aided me, in one way or another, in the completion of this project. Robert Hecht-Nielsen first presented the ideas to me that began this project. I learned more than just neurocomputing from his course and from the number of conversations we have had. As CEO of HNC, in addition to teaching graduate courses in electrical and computer engineering at UCSD, he has huge demands on his time, and yet he has graciously made time to have a number of long conversations with me about a variety of issues relating to this project, on topics from theories of brain function to issues in engineering generally.

Adrian Cussins has had much more influence on my thought than he suspects, and more than is evident in this project. Earlier drafts of this dissertation included a chapter on metaphysics that manifested that influence, but for purposes of continuity I elected to omit that chapter from the final project. He has made me a better philosopher than I would have been.

Vilayanur Ramachandran was kind enough to allow me to participate in his neuropsychology laboratory for three years even though the amount of time I was able to devote to the lab in hands-on projects was much less than is usual for a lab participant. I had the opportunity to learn from a brilliant experimentalist that amazing revelations are sometimes sitting right under your nose, waiting patiently for you to look at them. His insight, knowledge, and knack for transcending the conventional has been a great inspiration to me.

Ron Langacker has painstakingly developed what I consider to be the most plausible, powerful and elegant theory of human linguistic competence extant. Not only has his work convinced me that language is intimately linked to, and provides a window on, many other facets of cognition, but he has patiently read and provided crucially helpful feedback on drafts of Chapter Six.
A number of others have helped me in one way or another: by providing valuable feedback on earlier drafts, by forcing me to be less unclear, by offering encouragement, ideas, or inspiration, or by providing opportunities that I would not otherwise have had. These include, but are not limited to Andy Clark, Gary Cottrell, Gilles Fauconnier, Bob Gordon, Peter and Vicki Grush, Valerie Hardcastle, Susan Hibbs, Bill Hirstein, Jordan Hughes, George Lakoff, Megan Lauppe, Phillip Kitcher, Nili Mandelblit, Mary Powers, Joe Ramsey, Georg Schwartz, Oron Shagrir, Micheal Wedin, and Dorene and Dan Wetzel.

Finally, but certainly not least, my debts to Paul and Pat Churchland are countless. Paul's Scientific Realism and the Plasticity of Mind, and Pat's Neurophilosophy, both of which I read as an undergraduate, were my guiding philosophical inspirations. So much so that the only graduate program I applied to was UCSD's. The lofty standards they set as intellectual inspirations when I was an undergraduate, they matched as mentors during my graduate career. They have provided an environment in which interdisciplinary work is encouraged and where I could follow my interests. They helped in a number of ways (often going far beyond the call of duty) with funding and other administrative issues. When I needed space and freedom, it was there, and when I needed feedback and encouragement, it was there as well. If philosophy really is an honest search for enlightenment, a free investigation into basic questions of existence and human nature, if it is a love of intellectual curiosity unshackled by methodological pretenses and posturings, then no one deserves the title of Philosopher more than Paul and Pat Churchland.

February 2, 1965 Born, Klamath Falls, Oregon
1990 B.A. (Philosophy) University of California, Davis
1995 Ph.D. (Cognitive Science and Philosophy)
University of California, San Diego

PUBLICATIONS

Grush, R. (1994a) 'Motor models as steps to higher cognition' Behavioral and Brain Sciences 17:2:209-210, commentary on Jeannerod, M. (1994) 'The representing brain - Neural correlates of motor intention and imagery' Behavioral and Brain Sciences 17:2:187-202

Grush, R. (1994b) 'Beyond connectionist vs. classical AI: A control theoretic perspective on development and cognitive science' Behavioral and Brain Sciences 17:4:720 commentary on Karmiloff-Smith, A. (1994) Precis of Beyond Modularity: A developmental perspective on cognitive science Behavioral and Brain Sciences 17(4):693-706

Grush, R. and Churchland, P.S. (1995) 'Gaps in Penrose's Toilings' Journal of Consciousness Studies

FIELDS OF STUDY
Major Field: Philosophy

Studies in Philosophy of Mind
Professors Paul Churchland and Patricia Churchland

Studies in Philosophy of Language
Professor Adrian Cussins

Studies in Cognitive Linguistics
Professors Ronald Langacker and Gilles Fauconnier

Studies in Neurocomputing
Adjunct Professor Robert Hecht-Nielsen

Studies in Neuropsychology
Professor Vilayanur Ramachandran

ABSTRACT OF THE DISSERTATION

Emulation and Cognition

by

Rick Grush

Doctor of Philosophy in Cognitive Science and Philosophy

University of California, San Diego, 1995

Professor Paul M. Churchland, Chair

I explain a strategy, called model-based control, which has proven useful in control theory, and argue that many aspects of brain function can be understood as applications of this strategy. I first demonstrate that in the domain of motor control, there is good evidence that the brain constructs models, or emulators, of musculoskeletal dynamics. I then argue that imagery, motor, visual and otherwise, can be supported by these emulatory mechanisms. I argue that the same apparatus to understanding aspects of psychological development, including the development of theory of mind. I then show how features of linguistic competence can be addressed with the same mechanisms. Finally, I develop a semantic theory applicable to these emulators.