leff@smu.UUCP (Laurence Leff) (12/15/88)
Subject: AI-Related Dissertations from Sigart No. 102, part 2 of 3 The following is a list of dissertation titles and abstracts related to Artificial Intelligence taken taken from the Dissertation Abstracts International (DAI) database. The list is assembled by Susanne Humphrey and myself and is published in the SIGART Newsletter (that list doesn't include the abstracts). The dissertation titles and abstracts contained here are published with the permission of University Microfilms International, publishers of the DAI database. University Microfilms has granted permission for this list to be redistributed electronically and for extracts and hardcopies to be made of it, provided that this notice is included and provided that the list is not sold. Copies of the dissertations may be obtained by addressing your request to: University Microfilms International Dissertation Copies Post Office Box 1764 Ann Arbor, Michigan 48106 or by telephoning (toll-free) 1-800-521-3042 (except for Michigan, Hawaii, and Alaska). In Canada: 1-800-268-6090. From SIGART Newsletter No. 102 part 2 of 3 Economics to Linguistics ------------------------------------------------------------------------ AN University Microfilms Order Number ADG87-09801. AU HALL, HOMER KEITH, JR. IN Purdue University Ph.D. 1986, 196 pages. TI THE PROCESS OF DECISION MAKING UNDER UNCERTAINTY WITH AN APPLICATION TO THE THEORY OF EXPERT SYSTEMS. SO DAI V48(01), SecA, pp194. DE Economics, Theory. AB The broad objective of this dissertation is to examine the process of decision making under uncertainty. First, within the framework of the economics of information, a theoretical model is developed that describes rational choice as a process of gathering information in a sequence of r > 1 actions, determining an appropriate stopping time for the information collection, and then making a final decision. The formulation of an a priori strategy is described where each information gathering action, as well as the final decision, is contingent upon information signals received from previous actions. As information is collected during the execution of such a strategy uncertainty is reduced. Therefore, at any step in the sequence of information gathering actions the remaining portion of the strategy, called a substrategy, is seen to solve a subproblem of the original problem. It is shown that if a strategy is optimal for a decision problem (maximizes the expected net payoff), then any of its substrategies will be optimal for its associated subproblem. Second, the possible use of this, or some other decision-theoretical model, as a framework for computer based decision making systems is discussed. Since very little work has been done on the decision-theoretic basis of expert systems as a subject of artificial intelligence, this approach is first compared to the approach taken in most current expert systems. The major advantage is seen to be the focus on the efficiency of the decision process and therefore necessarily on the trade-off between the "correctness" and the cost of making a decision. Then, some of the computational aspects of using our theoretical model as a basis for an expert system are discussed. A method for acquiring sufficient information from a decision maker to calculate a gross optimal joint information and decision strategy (maximizing the expected gross payoff of the decision process) is discussed that does not depend upon fully specifying an outcome function. AN University Microfilms Order Number ADG87-10235. AU WRITT, PATRICK JAMES. IN Columbia University Ph.D. 1987, 149 pages. TI MATHEMATICAL PROBLEM SOLVING: AN EXPLORATION OF THE RELATIONSHIP BETWEEN STRATEGIES AND HEURISTICS. SO DAI V48(01), SecA, pp72. DE Education, Mathematics. AB The purpose of this study was to explore the effect strategy has on the problem solving process, its relationship to the heuristic process and, in particular, its effect on each of Polya's four phases of problem solving: Understanding the Problem, Devising a Plan, Carrying Out the Plan, and Looking Back. By strategy is meant a problem specific procedure which will solve the problem correctly and by heuristic is meant a non-problem specific procedure which does not necessarily solve it. To investigate the effect of strategy on the heuristic process a computer program, The Square Problem, was written which measured both the subject's strategy and heuristic process. The Square Problem could be solved using four strategies--three successful and one unsuccessful--and 12 heuristics, each assigned by a group of experts to one of Polya's four phases of problem solving. Seventy-five high school seniors and juniors attending seven different independent high schools in the New York City metropolitan area solved the problem four different times. The results of the study indicate that there is a relationship between the strategy used to solve the problem and the heuristic process, with observed differences in the heuristic process directly linked to the strategy used to solve the problem. The parts of the process that were similar tended to be related more to the problem itself, while the differences tended to be related to the strategy employed to solve the problem. Examined from the perspective of Polya's four-phase-model, most of the differences among the three successful strategy groups are associated with the phase Carrying Out the Plan. Unsuccessful problem solvers showed no difference on the phase Carrying Out the Plan but did spend more time than the successful problem solvers on the phases Understanding the Problem, and Devising a Plan. None of the strategy groups employed the phase Looking Back. Based on the results it is recommended that strategy--determined by a logical analysis of the problem--be taken into consideration when examining the heuristic process. AN University Microfilms Order Number ADG87-07794. AU IRZIK, GUROL. IN Indiana University Ph.D. 1986, 111 pages. TI CAUSAL MODELING IN THE SOCIAL SCIENCES: FOUNDATIONS AND APPLICATIONS TO PHILOSOPHY. SO DAI v47(12), SecA, pp4317. DE Education, Philosophy of. AB Although causal modeling employed in the social sciences is highly relevant to philosophical issues of statistical explanation and probabilistic causation, it has gone largely unnoticed by philosophers. The few exceptions to this were critical of the enterprise of causal modeling as a whole. However, the similarities between the sociological applications of causal modeling and recent philosophical theories of causality extend beyond the technical details, to cover the fundamental intuitions behind them: from the use of statistical machinery to the principle of common cause, and to the idea of process as the basis of causal relationships. Accordingly, the first part of this dissertation includes a discussion of different types of models (such as path models) and various techniques associated with them and defends causal modeling against charges of being methodologically defective, empiricist, and reductionist. The second part uses causal methodology to modify and extend Wesley C. Salmon's model of statistical explanation, and to establish the proper connection between causal and probabilistic claims. Finally, it is concluded that an intuitive and irreducible idea of causality based on causal processes provides a suitable framework both for models of scientific explanation and causal methodology in the sciences. AN University Microfilms Order Number ADG87-10163. AU TULLY, MARIANNE C. IN Columbia University Teachers College Ed.D. 1987, 195 pages. TI THE EVENT OF KNOWING: AN EDUCATOR'S PERSPECTIVE ON HEIDEGGER AND HERMENEUTICS. SO DAI V48(01), SecA, pp75. DE Education, Philosophy of. AB A phenomenological description of the event of knowing yields the essential characteristic of resonance. Resonance is stipulated to be a sense of already being familiar with the world to be known. Martin Heidegger's writings on knowledge and knowing support this description, especially as regards his notion of knowing as a founded mode of being-in-the-world, of knowing as ontological understanding, and of knowing as a hermeneutic experience. The Heideggerian images of the broken hammer and the clearing illustrate each successive level of the explanation of what knowing is according to Heidegger, and in turn offer a connection to the educating experience. Certain conditions, created by the teacher, may facilitate an educating experience in which the students resonate with the lesson. Heidegger's views on knowing provide a philosophical grounding for these conditions of learning. AN University Microfilms Order Number ADG87-09066. AU GAGE, BARBARA ANN. IN University of Maryland Ph.D. 1986, 184 pages. TI AN ANALYSIS OF PROBLEM SOLVING PROCESSES USED IN COLLEGE CHEMISTRY QUANTITATIVE EQUILIBRIUM PROBLEMS. SO DAI V48(01), SecA, pp94. DE Education, Sciences. AB This study investigates and compares the problem solving behavior of college chemistry faculty (experts) and undergraduate chemistry students (novices) in solving three quantitative homogeneous gas phase equilibrium problems. Steps and sequence taken by experts (n = 5) and novices (n = 20) were compared to a standard general college chemistry textbook presentation for three problem types: (1) computing K(,c) from equilibrium concentrations of all species; (2) calculating new equilibrium concentrations of species when a product is added to a system at equilibrium; (3) calculating species equilibrium concentrations starting with amount of one reactant. Subjects interviewed during solution of the problems were asked to think-aloud as they progressed, explaining each step taken. Interviews were tape-recorded and transcribed. Resulting protocols were analyzed to: (1) identify procedural steps taken; (2) record sequence of steps taken; (3) compare expert and novice sequences to textbook model; (4) identify procedural and conceptual errors made by novices. Textbook solution presentations were found to represent the step sequence taken by experts. Novice approaches vary from textbook and expert approaches in sequence. Step sequence was generally not related to novice success. Experts consistently wrote chemical equations for each problem while textbook presentations and novices did not. Major errors were committed by novices, independent of their previous chemistry grades. Novices recognized problem types and applied learned algorithms rather than analyzing problem systems. When presented with a "disturbed equilibrium system" problem, novices had difficulty visualizing the system and quantitatively adjusting for new concentrations. Students confused amount for concentration but generally knew that concentrations are used in K(,c) expressions. Results of this study support previous findings that novices are algorithm or rule learners. Novices depend on problem type recognition and recall of algorithms rather than analysis of problem systems. No other general heuristics were found. Findings also confirm that students do not apply the implications of LeChatelier's Principle consistently but employ algorithms instead of analysis in dealing with quantitative shifts in equilibrium problems. Further work is recommended in equilibrium problem type recognition, problem system visualization, and effects of problem descriptions on novice performance. AN University Microfilms Order Number ADG87-11457. AU SKOLYAN, KENNETH STEPHEN. IN United States International University Ed.D. 1987, 199 pages. TI ASSESSING AND FORECASTING THE IMPLICATIONS OF ARTIFICIAL INTELLIGENCE SYSTEMS ON PEDAGOGY IN THE PUBLIC SECTOR. SO DAI V48(02), SecA, pp371. DE Education, Technology. AB The Problem. The purpose of this study was to asssess and forecast the implications Artificial Intelligence Systems would have on teaching methods, teacher training, curriculum and teacher-student roles in public education. Method. A three round Delphi Study was conducted. The Delphi Panel included 23 participants from the areas of A. I. research, computer using teachers from universities and high schools and writers in the field. Results. (I) Teacher Training and Reaction. (1) Resistance to the introduction of Artificial Intelligence Systems will occur. (2) Additional training for teachers in computers is likely. (3) The structure of teacher training will change from informational transmittal to learning how people learn and the structure of what they know. (II) Curriculum. (1) Schools will not be radically changed. The goals of education and the importance of reading and writing would remain. (2) Students would not be learning a large portion of their lesson at home. (3) Elitist separation by subject matter or social adjustment would not occur. (4) The classroom would change. Multi-leveled, multi-topical learning centers would be developed where students could learn at their own pace. Changes in testing would occur along with deeper student involvement with subject matter. (5) New subjects and increases in strategy development and problem solving would occur as a result of A. I. systems in the classroom. (6) The A. I. system would become an indispensable tool with frequent upgrade of use skills and development of new subjects occurring. (III) Teacher/Student Roles. (1) Teacher roles would not change from deliverance of knowledge and skills to parent, counselor or psychologist. (2) A. I. systems would not fill the role of parent, counselor or psychologist nor transmit values. (3) The role of teachers would change from learning director to co-problem solver. (4) A. I. systems would greatly assist teachers in improving learning experiences, diagnosing problems and assisting students with learning handicaps. AN University Microfilms Order Number ADG87-10410. AU FRESNEDA, PAULO SERGIO VILCHES. IN The George Washington University D.Sc. 1986, 281 pages. TI ASSESSING THE POTENTIAL OF MICROCOMPUTER-BASED EXPERT SYSTEMS IN THE PROCESS OF AGRICULTURAL TECHNOLOGY TRANSFER IN BRAZIL. SO DAI V48(01), SecB, pp193. DE Engineering, Agricultural. AB This study focused on technology transfer problems in the agricultural sector. The research hypotheses were to assess the potential use of microcomputer-based expert systems as (1) mechanisms for transferring technical information between agricultural research and rural extension programs, (2) training aids for extensionists' (county agents') programs, and (3) tools for gathering relevant information from farmers and extensionists for research and extension management. The study also addressed the integrative role that expert systems technology plays in the overall process of technology transfer in the agricultural area, as well as the self-improving feature the technology introduces to the Total System (Research + Extension + Farmers) of agricultural technology development. A prototype expert system was developed for diagnosing and recommending treatment for selected potato diseases. In an experiment carried out in an extension organization in Brazil, 56 extensionists used the prototype and filled out a questionnaire designed to test the research hypotheses. Forty-five agricultural researchers and university professors in the agricultural field were also interviewed. Categorical data analysis procedures and chi-square tests were used to test the research hypotheses and to check for relationships between the various variables. The findings of the study indicate that microcomputer-based expert system technology has the potential to accomplish the three objectives presented above. Research results have also indicated that microcomputer and expert system technology can not only integrate the information flow between research centers, extension programs, and farmers, but can also introduce a tool for self-improvement in the agricultural technology development system. AN University Microfilms Order Number ADG87-06268. AU KASSEM, MOSALLEM D. IN The Catholic University of America Ph.D. 1987, 220 pages. TI APPLICATION OF ARTIFICIAL INTELLIGENCE APPROACH TO DESIGN OF ONE-WAY FLEXURAL MEMBERS IN REINFORCED CONCRETE. SO DAI v47(12), SecB, pp4990. DE Engineering, Civil. AB This dissertation presents a methodology to use the principles of artificial intelligence in engineering design using the design of reinforced concrete one-way flexural members as a specific application. The end product is a LISP-BASED AI-aided computer program for the design of beams and slabs bending essentially in one direction. The computer program was designed to act somewhat like a human designer in that it simulated the learning or gaining experience part of the human ability. This was done by introducing into the American Concrete Institute (ACI) design procedures a "preliminary work" module. The key element of the module was the introduction of a feedback mechanism composed of three steps: acquisition of experience, application of experience, and database management. The final product was a program capable of applying pattern recognition to obtain an educated, experience-based, preliminary estimate of the cross-sectional dimensions, thus considerably increasing design efficiency. As the program is used, it gathers design experience in the form of a database which selectively stores the input and output data of a processed design problem. The benefits of this experience are fed back to enhance the processing of subsequent designs. It appears from this work that it is possible to incorporate the use of some AI techniques in the process of engineering design and thus gain considerable operational efficiency. AN University Microfilms Order Number ADG87-12388. AU MASOOD, MUHAMMAD TAHIR. IN Virginia Polytechnic Institute and State University Ph.D. 1987, 187 pages. TI FURTHER DEVELOPMENT AND APPLICATION OF COMPUTER-ASSISTED CREATIVITY TO RURAL ROAD RESOURCES MANAGEMENT PROJECTS. SO DAI V48(02), SecB, pp519. DE Engineering, Civil. Transportation. AB Artificial Intelligence (AI) is the part of computer science concerned with designing computer systems, that is, systems that exhibit the characteristics we associate with intelligence in human behavior--understanding language, learning, reasoning, solving problems, and so on. Many believe that insights into the nature of the mind can be gained by studying the operation of such programs. The AI concept has formed the basis for developing the computer-assisted creativity techniques called The Computer Consultant (TCC), and The Idea Machine (TIM). TIM has, so far, been applied to topics in the engineering and "hard sciences" fields. In this study these techniques are presented/reviewed in detail and the research concentrated on the expansion/development of a methodology for computer-assisted creativity. This research will help in further evolution of TIM into a richer process for idea generation and general problem solving, and in enhancing the application capabilities. This is done by: (1) expanding the conceptual and ideas data bases from which analogies can be drawn; (2) conducting comprehensive trials with TIM to establish its strengths and limitations; and (3) doing research on techniques for the screening and packaging of ideas techniques. Rural road projects are an important part of rural development programs in the Third World countries. For some years the construction of such road projects, funded in part by international donor agencies, has been a subject of some controversy. Most policy makers in the developing or underdeveloped countries support the practice of expanding the rural dirt (unpaved) roads rather than spending limited resources on maintenance. Some donor agencies are now inclined to only support maintenance-biased road projects. A similar situation arose in Pakistan where the U.S. Agency for International Development (USAID) proposed to fund a road resources development project in the Sind Province. This real life situation is selected as a basis for developing a road resources management model, and generating ideas using TIM. These ideas are screened and packaged to be used in revising the model for further trials. The application of TIM to this problem from the civil engineering field results in some useful outputs. This study provides a good basis for further enhancing TIM capabilities. AN University Microfilms Order Number ADG87-09476. AU WHARTON, STEPHEN WAYNE. IN University of Maryland Ph.D. 1986, 169 pages. TI A SPECTRAL TARGET RECOGNITION EXPERT FOR URBAN LAND COVER DISCRIMINATION IN HIGH RESOLUTION REMOTELY SENSED DATA. SO DAI V48(01), SecB, pp214. DE Engineering, Civil. AB Parametric methods for spectral classification use statistics or other thresholds to identify spectral categories in multispectral remotely sensed data. A limitation of statistical classification methods is that the analyst cannot utilize nonparametric spectral knowledge in the classification process. It is necessary to treat each set of observations as a separate case in which the relationship between the observed spectral patterns and the set of categories must be empirically defined. The results are not generally applicable to other areas or dates because of spectral variations induced by differing atmospheric or illumination conditions. A prototype expert system was developed to demonstrate the feasibility of classifying multispectral remotely sensed data on the basis of knowledge of the spectral relationships within and between land cover categories. The spectral expert was developed and tested with Thematic Mapper Simulator (TMS) data having eight spectral bands and a spatial resolution of 5m. A knowledge base was developed that describes the target categories in terms of three types of spectral features: band to band relations that describe the shape of the reflectance curve; category to background relations that describe local contrast; and category to category relations that describe contrast with other designated categories. The knowledge base is used to direct the accumulation of spectral evidence for each target category. The system makes classification decisions on the basis of convergent evidence, i.e., two or more features that support the same category. The spectral expert achieved an accuracy of 80 percent correct or higher in recognizing eleven spectral categories in TMS data for the Washington D.C. area. AN University Microfilms Order Number ADG87-09669. AU PARRA-LOERA, RAMON. IN New Mexico State University Ph.D. 1986, 94 pages. TI AN AUTOMATIC OBJECT RECOGNITION SYSTEM. SO DAI V48(01), SecB, pp223. DE Engineering, Electronics and Electrical. AB The problem of devising a fast object recognition system is addressed here. An implementation of the system is presented and discussed. The concept Normalized Interval/Vertex Descriptors (NI/VD) is presented and used to represent contours. Two systems are presented: Global Shape Recognition System and Partial Shape Recognition System (for occluded objects). Polygonal approximation methods for contour representation are discussed and some of them are implemented. Presorting techniques to increase classification speed are developed. Data base management system concepts and structural pattern recognition concepts are used in the development of the presorting algorithms. Experimental results that demonstrate the performance of the present system are given. Comparison of the presented approach with different existing approaches is also included. The system is implemented in "C" language and runs under Unix Berkeley 4.2 in a VAX 11/750. AN University Microfilms Order Number ADG87-05726. AU TKACIK, THOMAS E. IN University of Virginia Ph.D. 1986, 170 pages. TI MACHINE VISION FOR MOBILE ROBOTS. SO DAI V48(01), SecB, pp225. DE Engineering, Electronics and Electrical. AB A vision system using a pair of Linear Image Arrays is proposed for use with mobile robots in a factory environment. This vision system was developed to perform the function of guidance, navigation, and obstacle detection required by mobile robots. Using a standard 16-bit microprocessor for all of the computing tasks, low cost, small size, and low power consumption can be obtained. Real-time operation is achieved by using line images, which provide less data then 2-dimensional images. However, this results in the need for image processing techniques that are tailored for use with line images. Both guidance and navigation are aided by a new technique called the difference of medians, a robust method of detecting and locating known features in line images. Obstacle detection is performed using a technique that locates feature points in thresholded images, and template matching for determining the distances to these feature points. A distance map is then generated, from which obstacles can be located. The algorithms used are evaluated and chosen for their efficiency, with the result that a pair of images can be analyzed in less then 0.5 sec. Finally, an analysis is made of measurement errors as a function of camera misalignment. AN University Microfilms Order Number ADG87-08554. AU VALAVANIS, KIMON P. IN Rensselaer Polytechnic Institute Ph.D. 1986, 308 pages. TI A MATHEMATICAL FORMULATION FOR THE ANALYTICAL DESIGN OF INTELLIGENT MACHINES. SO DAI v47(12), SecB, pp5007. DE Engineering, Electronics and Electrical. AB A mathematical formulation for the analytical design of Intelligent Machines operating in structured or unstructured uncertain environments with minimum supervision and interaction with a human operator has been developed. The structure of the Intelligent Machine is defined to be the structure of a Hierarchically Intelligent Control System, composed of three levels of interactive controls ordered according to the principle of Increasing Intelligence with Decreasing Precision, namely: the organization level performing information processing tasks in association with a long-term memory, the coordination level dealing with specific information processing tasks with a short-term memory only, and the execution level which performs the execution of various tasks through hardware using feedback control methods. A three level probabilistic model is derived for such a system and an Entropy function is proposed as a common measure of the performance of the three levels. Special simple architectures specifically designed to implement the mathematical model are derived. A Generalized Partition Law of Information Rates which considers the internal control procedures, interaction between the three levels and memory exchange procedures within the Intelligent Machines is also derived to indicate the flow of knowledge (information) in such machines. The intelligent control problem is cast as the mathematical programming solution that minimizes the total Entropy of Intelligent Machines. AN University Microfilms Order Number ADG87-06135. AU WRIGHT, JAMES AUSTIN. IN The University of Texas at Austin Ph.D. 1986, 128 pages. TI QUANTITATIVE ANALYSIS OF HUMAN PERCEPTION AND JUDGMENT. so DAI v47(12), SecB, pp5008. DE Engineering, Electronics and Electrical. Psychology, Psychometrics. AB Understanding human perception is becoming increasingly important in the field of engineering as we develop computing and robotic systems which attempt to model the psychophysics and psychology of the human being. A large amount of effort has been expended in modeling human decision processes but less is known about exactly what information is used by people in making decisions. This research is concerned with identifying, evaluating and expanding a method for finding out what knowledge is used by people in making decisions and quantifying its significance to the decision making process. The specific application covered in the dissertation is analyzing print quality surveys to determine what physical characteristics of printed matter people use to judge between samples and how the physical characteristics affect judgments of preference. We demonstrate the survey design and analysis technique on two print quality surveys, one of which shows how the techniques can be used to judge samples relative to a reference sample. These same techniques can be applied in many other areas to ascertain exactly what attributes of a particular object or situation are perceivable to people and what affect they have on decisions related to them--for example, one could study the characteristics of strip-chart recordings from well logging equipment to identify the attributes used by experts to judge such recordings. This could then be incorporated in an expert system to identify potentially promising wells by computer. We also present a new multidimensional scaling technique based on maximum entropy and compare it to the previously preferred method for this type of survey analysis which was based on maximum likelihood. AN University Microfilms Order Number ADG87-07650. AU CURRAN, ALLEN RICHARD. IN Stanford University Ph.D. 1987, 120 pages. TI AN INTELLIGENT CONTROL SYSTEM DESIGN AID. SO DAI v47(12), SecB, pp5025. DE Engineering, Mechanical. AB The objective of this thesis has been to develop the framework for an "intelligent" Computer Aided Control System Design (CACSD) Aid. Such systems are intended to allow engineers who are not control system specialists to effectively solve control system problems. To achieve this objective, a model was developed to explain the behavior of a human designer when solving routine controls problems. The model reflects that the designer solves control system problems by breaking the overall problem into successively smaller parts, each of which fall into one of three classes: (1) Analyzing the problem and/or deciding on a design approach to take, i.e., diagnostic or deductive tasks which are heuristic in nature; (2) Following a "recipe" for design until either the solution is reached or "something goes wrong"; (3) Calculating a result, number crunching, etc. The Design Aid has the following "intelligent" features: (1) Introspection. The Design Aid is able to explain decisions made en route to a problem solution. (2) Modifiable Knowledge Base. The user can create, modify, and delete rules according to his own preference and experience. (3) Flexible Division of Labor. The user is able to work on any part of the design at any time and have the Design Aid "fill in the rest." (4) Expertise. The problem solving performance of the prototype system is roughly comparable to that of a student who has taken an introductory controls course. The Design Aid is structured as a hierarchy of specialists that solve parts of the design problem and/or pass parts of the problem to sub-specialists. The design process proceeds forward from the initial state to the final goal state, with each specialist satisfying an intermediate goal. Three kinds of specialists are utilized by the Design Aid, each of which corresponds to one type of human design activity enumerated above. The thesis addresses a number of issues relevant to design expert systems including representation of design plans, failure handling, passing advice to design procedures, and dealing with interacting design goals. AN University Microfilms Order Number ADG87-08097. AU LI, PEIGEN. IN The University of Wisconsin - Madison Ph.D. 1987, 202 pages. TI DEVELOPMENT OF AN EXPERT SYSTEM FOR MULTIFACET DRILLS. SO DAI V48(02), SecB, pp540. DE Engineering, Mechanical. AB Multifacet drills (MFDs) have shown great success in real drilling operations. However, their applications have been limited due to the complicated drill geometry design and the difficulty in drill point grinding. An expert system called DRILLEX with high-level expertise about MFDs and with learning algorithms is proposed and developed for use in automated manufacturing environments. The system is possessed of capabilities for (1) design of MFD geometry and grinding parameters; (2) selection of drilling conditions (speed, feedrate, drill material, coolant); (3) trouble diagnosis; (4) recognition of drill wear states; and (5) selection of optimal speed and feedrate by use stochastic automation. The major elements of DRILLEX, i.e. the knowledge base and the inference engine are designed and analyzed. This thesis investigates the details needed for the development of knowledge base by use of VAX-11 Datatrieve. The fuzzy concept is adopted in this study to deal with some phenomena which are vague in nature. The strategy of Fuzzy Logic Control is used to select suitable MFD types during the geometry design of Multifacet Drills. The expert system DRILLEX is capable of recognizing the drill wear states through the learning process known as Fuzzy C-Means algorithm. The thrust force and torque are considered as features for clustering and classification. The feasibility of this algorithm is shown by experiments and through simulations. The approach of this fuzzy pattern recognition method, indicates good suitability for varying environments and better classification accuracy than the conventional pattern recognition techniques. The learning process for optimizing cutting speed and feedrate is introduced based on stochastic automaton theory. Analysis of the algorithm, formulated as a learning controller of cutting parameters, is performed. Simulation studies have been carried out to verify the effectiveness of the proposed method. AN This item is not available from University Microfilms International ADG03-74262. AU SCHNEITER, JOHN LEWIS. IN Massachusetts Institute of Technology SC.D. 1986. TI AUTOMATED TACTILE SENSING FOR OBJECT RECOGNITION AND LOCALIZATION. SO ADD X1986. DE Engineering, Mechanical. (no abstract provided in the DAI database]) AN University Microfilms Order Number ADG87-11622. AU ZAREFAR, HORMOZD. IN The University of Texas at Arlington Ph.D. 1986, 190 pages. TI AN APPROACH TO MECHANICAL DESIGN USING A NETWORK OF INTERACTIVE HYBRID EXPERT SYSTEM. SO DAI V48(02), SecB, pp544. DE Engineering, Mechanical. AB Artificial intelligence (AI) is the branch of computer science that attempts to create "intelligent" behavior in computing machines. Expert systems is an applied field in the realm of AI which deals with emulating human expertise and problem solving capabilities in a particular domain. Successful systems have been developed in a number of disciplines such as medicine, geology, biochemistry and computer configuration. Mechanical design is a process of combining informal and common sense knowledge about a design domain with the more formal and well developed analysis and synthesis methods. Those aspects of the design which deal with the intuitive human thought process are not well defined for computer programming at any level. However, the more routine redesign class can be formalized for expert systems application. The objective of this research is to develop a taxonomy for a class of redesign process within the framework of available expert systems technology and mechanical design practice. The approach follows the practical design procedure in an industrial environment. It involves partitioning of the design space into subspaces and developing interactive expert subsystems to address the design process. A coordinator at each level of the design activity is suggested to emulate the lead engineer in a design team and direct the flow of the design process. An important aspect of this research is integration of data processing routines with the rule-based expert systems. To illustrate the proposed approach, a prototype parallel axis spur gear-drive system was developed. The program encompasses a network of interacting rule-based and procedural modules to design the components of a prototype speed reducer. Experience gained from this research suggests that the proposed approach could be employed for a variety of decomposable design tasks. Employing similar systems in the real industrial environment should result in more creativity and productivity from the design engineers. AN University Microfilms Order Number ADG87-11623. AU ATHEY, SUSAN. IN The University of Arizona Ph.D. 1987, 167 pages. TI A MENTOR SYSTEM INCORPORATING EXPERTISE TO GUIDE AND TEACH STATISTICAL DECISION-MAKING. SO DAI V48(02), SecA, pp238. DE Information Science. Business Administration, Management. Education, Business. AB The statistical mentor system incorporates a knowledge base into an educational tool for novices in statistical decision making to use in choosing a statistical technique. The novices are students in a business school curriculum who are expected to learn the basic statistical processes in business applications. The purpose of the system is to stimulate learning of the data analysis process on the part of the novice, usually a difficult task. The system acts as a consultant to the novice and approaches the task using a top-down problem solving strategy rather than the traditional bottom-up strategy used by novices. The heart of the system is the rule base for differentiating between statistics. These rules were built by gathering expertise from two experts in statistical analysis. The rules are based on five questions which the data can answer, as well as the type of data, the number of variables, and any dependent/independent relationships which exist between the variables. The knowledge base consists of five rule sets and can be represented either by condition/conclusion rules or by a set of multi-dimensional tables. Twenty-nine statistics and the rules for choosing them are in the rules sets. The knowledge base was used to define the logic incorporated in the consultant system in order to aid the user in selecting a correct technique. A dialogue mode is employed in the consultant to determine which conditions are true for the problem and data set. The rule sets are then checked to find the conclusion satisfying the conditions. The computer mentor was tested against the usual textbook mentor method (search through a textbook until one finds a statistic that looks promising) with two different groups of subjects, 25 undergraduates and 19 doctoral students. The results were that the computer-assisted students in both samples correctly solved a larger proportion of problems and had a higher average number of problems correct than did the textbook assisted groups. AN University Microfilms Order Number ADG87-02901. AU MULLIN, THERESA M. IN Carnegie-Mellon University Ph.D. 1986, 196 pages. TI UNDERSTANDING AND SUPPORTING THE PROCESS OF PROBABILISTIC ESTIMATION. SO DAI v47(12), SecA, pp4217. DE Information Science. AB There is an increasing interest in and recognition of the need to explicitly incorporate factors of uncertainty in public policy analysis, often in the form of subjective probability distributions. Many experimental studies involving the subjective judgments of nonexperts have shown there to be widespread and systematic biases towards overconfidence. These biases appear to result from subjects' failure to use available data in a theoretically appropriate way, and through the faulty use of a variety of judgmental heuristics, such as anchoring and adjustment. The implications of these laboratory studies for the quality of expert judgments are not clear, however. Many decision analysts consider these potential shortcomings in judgments can be ameliorated through structuring the judgment problem and through conditioning the assessor to be more aware of and account for various sources of uncertainty. A study conducted to investigate the potential benefits of problem decomposition in the probabilistic estimation of almanac quantities. It was found that assessments based on decomposition models were no more accurate nor better calibrated than estimates made directly, and use of the simple multiplicative decomposition models significantly altered the direction of the bias in subjects' judgments, from systematic underestimation to overestimation of the unknown quantities. A second study involved the probabilistic estimation of almanac quantities in a test of three suggested debiasing strategies thought to encourage assessors to account for more uncertainties in their estimates: these strategies involved telling the assessor (1) to think of contradictory reasons; (2) to describe alternative scenarios; and (3) about problems with anchoring and adjustment. Only warnings about the use of anchoring and adjustment reduced subjects' overconfidence significantly. To study the underlying processes involved in probabilistic estimation and determine whether there were any appreciable differences between the estimation processes of experts and nonexperts, estimation protocols for these two types of assessors were examined. Results of this study indicate that there may be significant differences in the approaches to probabilistic estimation used by experts versus nonexperts. These include differences in attitude, in the way the assessor works through the estimation problem, in the use of the anchoring and adjustment heuristic and in responses to debiasing questions. (Abstract shortened with permission of author.). AN University Microfilms Order Number ADG87-06126. AU WIDMEYER, GEORGE ROBERT, III. IN The University of Texas at Austin Ph.D. 1986, 150 pages. TI PREFERENCE DIRECTED REASONING IN DECISION SUPPORT SYSTEMS. SO DAI v47(12), SecA, pp4218. DE Information Science. AB This research compares two rule-based reasoning approaches and a cardinal dominance method with an assumed multi-attribute value function. The goal of these approaches is to reduce the cognitive demands on the decisionmaker in terms of preference elicitation from that required by conjoint analysis. This benefit is offset by the error that can result when compared with a multiattribute value function. A measure of effort is developed by deriving a value function based on the number of pairwise comparisons along attributes and the number of tradeoff comparisons across attributes. The accuracy of each approach is measured by the number of alternatives eliminated and the probability that the optimal alternative is not eliminated. This is done using a Monte-Carlo simulation for 10, 20 and 30 alternatives versus 2 through 10 attributes. The results of the research are that rule-based methods are not significantly less accurate than a normative model and this is more than offset by the reduced effort. Unlike conjoint analysis, the rule-based method studied is valid when only partial information is available and since it uses symbolic logic its results can be explained more directly. The significance of this research is that it presents a theory for decision support systems that integrates symbolic reasoning and preference technologies. AN University Microfilms Order Number ADG87-05945. AU ALAM, YUKIKO SASAKI. IN The University of Texas at Austin Ph.D. 1986, 283 pages. TI A THEORETICAL APPROACH TO THE JAPANESE VERBAL SYSTEM WITH COMPUTATIONAL IMPLICATIONS. SO DAI v47(12), SecA, pp4373. DE Language, Linguistics. AB This study presents a set of phrase structure rules for Japanese and an analysis of the verbal component, both of which are intended to serve a unified treatment of case assignment. Analyses presented here are expressed in X-bar schema as well as in the formalism of Lexical Functional Grammar. The set of phrase structure rules is formulated on the following hypotheses: (a) Japanese phrase strucure is strictly binary-branching and head-final, (b) the surface linear order of constituents is base-generated, (c) grammatical relations in Japanese are not defined configurationally, but mainly through cases, (d) Japanese differentiates between logical and constituent sentences. Logical sentences presuppose the presence of a subject, whereas constituent sentences do not. Finally, (e) the minimal elements necessary to compose a Japanese constituent sentence are a verb, a tense marker and an element indicating the use of the sentence, whereas the minimal elements for a logical sentence are a verb and a subject. This study also postulates a new syntactic category U(se), which stands for not only 'function words' but inflections. The postulation of this category permits the claim that Japanese phrase structure consists of the complement and the head at all levels, thus making it possible for the set of phrase structure rules to be compact. Comparison is made between the present model of grammar with other hypotheses: the flat structure hypothesis, the move-alpha hypothesis and case-linking grammar. It is also shown that the present model facilitates an attempt at an overall treatment of the causative construction which has been problematic to other analyses. In addition, a case theory is presented which is an extension of the work by localist case theorists. Unlike those of case theorists, however, the aspect system supporting the proposed case theory is an extension of the work by Vendler. Based on this theory, lexical entries for verbs are postulated and designed so that they, together with the proposed phrase structure rules and lexical entries for elements in the category U, make possible a unified treatment of case assignment. AN University Microfilms Order Number ADG87-07094. AU BEAUVAIS, PAUL JUDE. IN Michigan State University Ph.D. 1986, 157 pages. TI A SPEECH ACT THEORY OF METADISCOURSE. SO DAI v47(12), SecA, pp4374. DE Language, Linguistics. Education, Teacher Training. AB Metadiscourse commonly is defined as "discourse about discoursing." In its brief history, the term has appeared in several models of text structure; however, theorists disagree concerning the range of metadiscursive structures and the role of metadiscourse in a larger theory of text linguistics. This dissertation provides a detailed history and a critical analysis of the existing metadiscourse theories, and it offers an alternative theory that defines metadiscourse as a component of speech act theory. The first chapter surveys the history of metadiscourse from Zellig S. Harris' early use of the term to recent studies by Joseph M. Williams, Avon Crismore, and William J. Vande Kopple. The second chapter introduces four criteria for evaluating the utility of theoretical models. The existing metadiscourse models are analyzed in light of these criteria and are found to contain imprecise definitions of key terms. The models also are found to be collections of disparate structures instead of principled systems. The third chapter provides an overview of important works on speech act theory by J. L. Austin and John R. Searle. Particular attention is devoted to the distinction between illocutionary acts and propositions, the differences between explicit performative structures and implicit expressions of illocutionary intent, and the types of illocutionary acts that are possible. In the fourth chapter, metadiscourse is defined as those illocutionary force indicators that identify expositive illocutionary acts. A taxonomy of metadiscourse types is provided, and canonical forms using performative or near-performative structures are identified for each type. Partially explicit forms of metadiscourse that do not provide an attributive subject also are identified. The dissertation concludes with suggestions for experimental studies using the proposed metadiscourse model. AN University Microfilms Order Number ADG87-08276. AU BLACK, EZRA WILLIAM. IN City University of New York Ph.D. 1987, 181 pages. TI TOWARDS COMPUTATIONAL DISCRIMINATION OF ENGLISH WORD SENSES. SO DAI v47(12), SecA, pp4374. DE Language, Linguistics. AB An experiment is conducted which compares three different methods of deciding which of three or four senses characterizes each occurrence of a word for which a Key Word In Context concordance has been constructed. The three methods consist of a dictionary-based approach (DG) where categories intended to classify the words and expressions occurring in each concordance line are simply the subject codes of a major dictionary; an approach (DS1) in which categories are obtained via a frequency analysis of words occurring in the immediate neighborhood of the "node word"--the word in focus--of the concordance, and of "content" words occurring anywhere in a given line; and an approach (DS2) chiefly based on the content-analytic categories obtained by closely reading the concordances of a 100-type sampling of words occurring in the 20-25-million-token English text source, consisting of the official proceedings of the Canadian House of Commons. Results are that DG performs extremely poorly--in fact, near-randomly; DS1 and DS2 yield better and substantially similar performances. The conclusion is that domain-general, syntax-based approaches to automatic word sense discrimination and domain-specific, content-analytic approaches need and complement each other. AN University Microfilms Order Number ADG87-07013. AU CAIN, EILEEN. IN University of Hawaii Ph.D. 1986, 319 pages. TI LEXICAL RETRIEVAL DISTURBANCES IN A CONDUCTION APHASIC. SO DAI v47(12), SecA, pp4374. DE Language, Linguistics. Health Sciences, Speech Pathology. AB This dissertation is a case study of a conduction aphasic, C.G. Chapter 1 presents the clinical features of conduction aphasia. This is followed by a discussion of many of the linguistic impairments manifested in this syndrome: an impairment affecting the use of various word classes (including noun facilitation), anomia, and lexical substitution errors, which include verbal paraphasias (replacement of a target word by a word similar to it in meaning and/or form in speech) and paralexias (lexical substitutions in reading). Lexical substitution errors can affect both the open and the closed class vocabularies. It will be my position in this dissertation that all of these errors involve lexical retrieval disturbances. Chapter 2 discusses accounts of conduction aphasia in the literature: the classical disconnection model of Wernicke (1874/1969), Goldstein's (1948) central aphasia, the encoding deficit approach, studies attributing the repetition deficit in conduction aphasia to a defect of audio-verbal short-term memory, and several phonological approaches. These accounts are described and evaluated. Chapter 3 describes the subject in this study and the methodology used for data-gathering. The corpus includes samples of spontaneous speech, repetition, naming, reading, and writing, as well as metalinguistic tasks where the subject was asked to judge sentences as to their grammaticality and to correct them if possible. The fourth chapter presents the results of the Boston Diagnostic Aphasia Examination (Goodglass and Kaplan 1972) and the Token Test (De Renzi and Vignolo 1962). Chapter 5 discusses the results of the tests which I designed in order to provide a more complete picture of C.G.'s language abilities and impairments than can be obtained from the Boston Diagnostic Aphasia Examination or the Token Test. Results from the Wug Test (Berko 1958) are also included in Chapter 5. The data includes instances of lexical omissions and substitutions in repetition, reading, and spontaneous speech. Chapter 6 presents accounts of the linguistic impairments described in the previous chapters in terms of lexical selection errors, relying on models by Merrill Garrett (1980, 1981, 1982) and Katz and Fodor (1963), as well as network models of the lexicon. AN University Microfilms Order Number ADG87-10433. AU CHAO, WYNN. IN University of Massachusetts Ph.D. 1987, 228 pages. TI ON ELLIPSIS. SO DAI V48(02), SecA, pp380. DE Language, Linguistics. AB This work is intended as an investigation into elliptical phenomena in natural language. It is argued that at least two major classes of elliptical constructions must be distinguished, on the basis of the presence or absence of their major phrasal heads. The recovery of the missing material in the class in which the relevant heads are missing (H('(TURN)) class) is of a syntactic nature, and this along with other 'characteristic properties' of H('(TURN)) constructions follow as a direct consequence of the omission of their syntactic heads. In constrast, constructions in the H+ class are pronominal in nature, and their characteristic properties follow from the fact that pronominals may be interpreted either in the syntax (e.g., as bound variables) or in the discourse representation. An account of these constructions is proposed within the Government-Binding framework, and consists of four main components: (i) a 'defective' X-bar schema, which allows for the base-generation of H('(TURN)) constructions, (ii) a licensing principle on D-structure representations, which constrains the output of the defective rule schema, and (iii) a process of elliptical reconstruction at LF, which applies to both H('(TURN)) and H+ constructions, and (iv) the reintroduction of the more general notion of 'recovery of content' to subsume the more narrowly defined notion of 'syntactic identification.' It is argued that this proposal has several desirable consequences. In the first place, it provides a strictly syntactic basis for the H+/H('(TURN)) classification, and derives the distributional and interpretive properties of these constructions from the interaction of this syntactic fact with existing principles in the theory. It accounts for interesting similarities between null arguments (pro) and H+ phenomena. It sheds light on various aspects of the interpretation of both elliptical and overt pronominal elements. And finally, it makes predictions about the range of variation in the manifestations of H('(TURN)) and H+ constructions that one may expect to find in natural language. The linguistic data in this work is drawn primarily from English, but constructions from French, Brazilian Portuguese and Chinese are considered as well. AN University Microfilms Order Number ADG87-07343. AU FINN, KATHLEEN ELLEN. IN Georgetown University Ph.D. 1986, 261 pages. TI AN INVESTIGATION OF VISIBLE LIP INFORMATION TO BE USED IN AUTOMATED SPEECH RECOGNITION. SO DAI v47(12), SecA, pp4376. DE Language, Linguistics. AB Although acoustically-based automatic speech recognition (ASR) systems have witnessed enormous improvements over the past ten years, they still experience difficulties in several areas, including operation in noisy environments. Research with human subjects has shown that visible information about the talker provides extensive benefit to speech recognition in difficult listening situations. The primary purpose of this study was to demonstrate the feasibility of performing automatic speech recognition based only on optical information. Other purposes included characterizing the nature of the optical recognition process, comparing it to human visual speech perception, and estimating its potential value to acoustic-based ASR. An optically-based algorithm was developed to recognize a set of 23 English consonant phonemes. Distance measurements were derived from a set of 12 dots placed on or near the speaker's mouth. The dots offered a computationally simple means of tracking lip movements. After data reduction and selective weighting of the variables, five distance variables were found capable of identifying 74% of the phonemes, with no acoustic information whatsoever. The same variables correctly identified 87% of the phonemes by viseme groups. The machine's viseme set was very similar to the human set. The preponderance of the variables measured vertical distances, suggesting that vertical opening, as opposed to horizontal movement or area of mouth opening, is a critical cue to optical speech recognition. The optical recognition process was subjected to the effects of random visual noise at various levels, and was found to be fairly robust. The results of the optical recognition algorithm were compared against the results of an acoustic algorithm operating over the same speech tokens. The acoustic algorithm's performance, subjected to signal-to-noise ratios ranging from +25 dB to +65 dB, measured 64% correct phoneme recognition. It was estimated that an effective combination of the optical and acoustic recognition systems could result in 95% recognition. AN University Microfilms Order Number ADG87-11796. AU FUKADA, ATSUSHI. IN University of Illinois at Urbana-Champaign Ph.D. 1987, 149 pages. TI PRAGMATICS AND GRAMMATICAL DESCRIPTIONS (JAPANESE). SO DAI V48(02), SecA, pp380. DE Language, Linguistics. AB The goal of the dissertation is to argue for the recent position that strictly distinguishes between grammar and pragmatics and discuss its consequences. By now it is clear that "raw" linguistic data contain many pragmatic elements, whether they are speech act properties, implicatures, beliefs and intentions of the speech participants, or what not. In analyzing such data linguists, in my view, are constantly faced with two problems; one is how to distinguish pragmatic matters from purely grammatical aspects of the data, and the other is what to do with such pragmatic elements. The second problem has to do with a proper conceptualization of the relationship between pragmatics and grammar. In particular, linguists must have clear conception as to what the proper domain of each field is, and what the exact nature of the mode of their interaction is. This is, in my opinion, one of the outstandingly important empirical issues in current theoretical linguistics. The first problem concerns ways of determining, in a given situation, what is pragmatic and what is grammatical. If one decides to take the position that denies the heterogeneous nature of raw linguistic data, this problem will not arise at all. I will argue, however, that such a position cannot be seriously maintained. These are the two major issues this study addresses. The arguments in the body of the thesis will take the form of analyses or reanalyses of some problematic phenomena in Japanese and English where one's position on the above issues will have a serious effect on resulting grammatical descriptions of the phenomena. Two highly controversial areas of Japanese grammar, i.e. passives and causatives, issues concerning honorifics and politeness in general, and an analysis of the English complement-taking verb have are some of the major descriptive issues taken up in this study. In each case, it will be shown that the position being argued for can provide solutions to the controversies and/or lead to what seems to be the optimal over-all descriptions. AN University Microfilms Order Number ADG87-08406. AU GIOTTA, FRANCISC ANTONIO. IN University of California, Davis Ph.D. 1986, 887 pages. TI THE ARTICLE SYSTEM OF FRENCH AND FUZZY SEMANTIC MODELS. SO DAI v47(12), SecA, pp4376. DE Language, Linguistics. AB The study offers a comprehensive semantic representation of definiteness, indefiniteness and partitivity in French article uses. The crucial assumptions are: (1) natural languages use imprecise concepts and (2) the article marks the precision/imprecision level under which a nominal is used and becomes recoverable in the 'universe of discourse'. This suggests the application of a possibilistic model in the true/false semantic 'continuum' to account for subjective 'degrees of truth'. Explicit definition of fuzzy-semantic notions and a considerable amount of linguistic data supporting the theory makes the present work particularly useful in the area of French linguistics, universal grammar and artificial intelligence. The semantic representation of definite and indefinite reference in classical logic, pragmatics and discourse dynamics are described and discussed in terms of adequacy and explanatory power to handle linguistic evidence found in French article uses. A tentative categorial representation of fuzzy nominalizations with respect to verbal subcategorization yields three major types of assertive content: 'features', 'kinds' and 'abstracts'. This model is tested in special uses of the article: articled proper nouns, definites in topicalization and specific and non-specific indefinites 'a certain', 'such a', 'any'. Fuzzy set-theoretical categories of functors, 'monomorphism', 'epimorphism', 'representative' and 'isomorphism' define, respectively, indefinite, definite, demonstrative and articleless nominals. The consideration of plurality demands a drastic reformulation of the notions of set and combinatorics. The 'partitive' is analysed as a pairing between nameability (presence in the universe of discourse) and assertability (argument-value for the current predication) of a nominal theme. The last three chapters are extensions and applications of the fuzzy semantic model to anaphora, genericness and pragmatic presuppositions. Anaphora is defined as a behavior map in a discourse linear-ordered dynamic system (dynamics is described by standard information retrieval notions: reacheability, observability, minimal realization, free trajectory, memory and forgetful functors). The incrementation law of discourse is held as a generalized entailment structure: a 'crisp' conclusion is inferable from fuzzy premises under some nicety conditions. The pragmatic component makes use of fuzzy sentential connectors. The ordinary 'truth-tables' are replaced by offer/challenge risk-values distributed among speakers in a dialogue interpretation. AN University Microfilms Order Number ADG87-06620. AU LOBECK, ANNE C. IN University of Washington Ph.D. 1986, 300 pages. TI SYNTACTIC CONSTRAINTS ON VP ELLIPSIS. SO DAI v47(12), SecA, pp4378. DE Language, Linguistics. AB The theory of government proposed by Chomsky (1981) is argued to include a set of principles which restrict the distribution of empty categories, in particular those derived by movement. This study investigates two types of empty categories which are not derived by movement, and which have not been systematically analyzed in government-binding theory. They include those base-generated empty categories which arise from ellipsis in NP, S and S', and the set of non-phrase nodes which Emonds (1985) argues are generated empty in the base, and filled at a post-transformational level. The aim of this study is to show that these two types of empty categories are subject to a uniform set of restrictions, and interact with each other in a way best expressed in terms of proper government. To incorporate the restrictions on ellipted categories, a theory of government is developed where both X('0) heads and specifiers are potential governors, and where phrasal categories, but not heads, are subject to the ECP. As a result, intermediate projections, which are shown to be those which are systematically ellipted in English, must be properly governed. Such projections are licensed only by 'specifier' government, but not by lexical or antecedent government. Ellipsis in S is discussed in detail, where the constraints on the distribution of empty phrasal projections of V are shown to follow from principles of specifier government and their interaction with conditions on empty heads, the Empty Head Conditions and the Generalized Head Movement Constraint. INFL is analyzed as SP(V), which, aside from accounting for several properties INFL shares with specifiers, allows ellipsis across categories to be uniformly expressed as licensed by specifier government. The proposed analysis obviates analyzing INFL as a lexical head when tensed but not untensed, in contrast to previous proposals. AN This item is not available from University Microfilms International ADG05-60077. AU ROBERGE, YVES. IN The University of British Columbia (Canada) Ph.D. 1986. TI THE SYNTACTIC RECOVERABILITY OF NULL ARGUMENTS. SO DAI V48(01), SecA, pp119. DE Language, Linguistics. AB In most natural languages, a sentence may include a variety of missing elements the recoverability of which is made possible by different processes. This thesis investigates the type of syntactic recoverability found in null argument languages. It is supposed that the mechanisms responsible for this type of recoverability are deeply embedded in Universal Grammar and that this suggests that there is no need for a parameter designed to allow empty arguments per se. The main goal pursued here is to present a systematic account of the similarities between recoverability through verbal agreement and recoverability through clitics. This results in the proposal that languages with subject clitics and/or object clitics are the same as languages with rich subject agreement and/or object agreement as far as the licensing of the empty pronominal pro is concerned. We then examine the relationship between clitics and overt NPs in the so-called clitic doubling constructions. The hypothesis defended here is that subject clitics and object clitics are surface realizations of the same abstract element and that this can account for the symmetry existing between various types of clitic regarding the licensing of pro, the possibilities for doubling, and extractions out of doubling constructions at S-structure and at LF. AN University Microfilms Order Number ADG87-10499. AU ROBERTS, CRAIGE. IN University of Massachusetts Ph.D. 1987, 364 pages. TI MODAL SUBORDINATION, ANAPHORA, AND DISTRIBUTIVITY. SO DAI V48(02), SecA, pp381. DE Language, Linguistics. AB The analysis of pronominal anaphora provides us with tools to explore linguistic structures involving the scope of operators. In this dissertation, I develop a theory of anaphora, modifying and extending existing proposals in the literature, and then use it to explore distributivity and related phenomena. I assume that pronouns are interpreted as variables, and base a theory of anaphora on the claim that there are two kinds of constraints on how these variables may be bound. One type of constraint involves the relative positions of antecedents and anaphors in the hierarchical structure of discourse. I propose an extension of Discourse Representation Theory wherein a relation of subordination between propositions is induced by their mood. Mood is analyzed in terms of modality, and establishes the position of a proposition in the Discourse Representation. The structure which results constrains both inference and the potential for anaphora. The other type of constraint on anaphoric binding is based on the configurational notion of c-command in the Government and Binding Theory. Recognizing that the Binding Theory and the theory of discourse anaphora are both necessary in a comprehensive theory of anaphora permits a clarification and simplification of each. It is argued that the Binding Principles hold at S-Structure, and that coindexation is only a guide to interpretation in discourse, and not necessarily an indication of coreference. This comprehensive theory of anaphora serves as a tool for the exploration of the phenomenon of distributivity, including the group/distributive ambiguity in examples such as four men lifted a piano. It is argued that distributivity arises in predication when either the determiner in the subject is quantificational or there is an implicit or explicit adverbial distributivity operator. Anaphoric phenomena associated with distributivity are shown to be a consequence of the scope of operators. This theory of distributivity, implemented in the mapping from S-Structures onto Discourse Representations, then provides further arguments that coindexation is not to be interpreted as coreference, and also illuminates the contribution of the number of a pronoun to its interpretation. AN University Microfilms Order Number ADG87-07465. AU ROSS, GARRY. IN Texas A&M University Ph.D. 1986, 174 pages. TI COHERENCE THEORY: AN INTERDISCIPLINARY STUDY. SO DAI v47(12), SecA, pp4378. DE Language, Linguistics. AB What writers of composition texts have written about coherence is not useful to teachers or students. The valuative nature of the language they use to describe coherence does not further instruction. They confuse three aspects of language use--correct grammar, correct usage, and cohesion--with coherence. Their failure to recognize the separateness of these three has resulted in them writing about coherence in a way that is uninstructive and confusing. Those relevant aspects of cohesion which textbooks discuss are reference, connection, and lexical cohesion. Reference is of two kinds, exophoric and endophoric. Exophoric reference points outside the text. Endophoric reference, which can be broken down into anaphoric and cataphoric reference, functions within a text. Connection is of four types: additive, adversative, causal, and temporal. It can point out internal or external relationships. Lexical cohesion depends on the lexicosemantic relationship between two words. The second chapter of Huckleberry Finn shows that standard English is not what gives a text coherence. Twain's use of dialect and non-standard English does not distract from the global structure of the episode. Just as standard English is not a clear requirement for coherence, cohesion is not a clear requirement for coherence. Cohesion does not deal with global restraints. Coherence does. Philosophy offers much to the discussion of coherence. Kant's Categories and logical form offer a base from which to discern form in discourse. The coherence theory of truth and the philosophy of language-in-use suggest that there are realities which can function as bases for determining the wholeness of a work. Gestalt psychology and cognitive psychology show that structure is important to perception. Cognitive structures function in the process of knowing. These structures are discernible and can be used as bases to identify textual coherence. Linguistics offers the global structures of text grammars as clues to the generation of coherent texts. Linguists working from the base of language-in-use philosophy have identified structures that make reference to context of situation to establish the boundaries of a text. Both linguistic endeavors offer grounds from which texts can be judged coherent. AN This item is not available from University Microfilms International ADG03-74350. AU SCHEIN, BARRY JAY. IN Massachusetts Institute of Technology Ph.D. 1986. TI EVENT LOGIC AND THE INTERPRETATION OF PLURALS. SO ADD X1986. DE Language, Linguistics. (no abstract provided in the DAI database]) AN This item is not available from University Microfilms International ADG03-74644. AU SPROAT, RICHARD WILLIAM. IN Massachusetts Institute of Technology Ph.D. 1985. TI ON DERIVING THE LEXICON. SO ADD X1986. DE Language, Linguistics. (no abstract provided in DAI database]) AN University Microfilms Order Number ADG87-06125. AU WEIR, CARL EDWARD, JR. IN The University of Texas at Austin Ph.D. 1986, 117 pages. TI ENGLISH GERUNDIVE CONSTRUCTIONS. SO DAI v47(12), SecA, pp4379. DE Language, Linguistics. AB English gerundives are nominalizations containing a verb phrase constituent whose initial, non-adverbial element has the 'ing' morpheme suffixed to it. Often gerundives also contain a possessive NP constituent, and consequently they are sometimes referred to as POSS-ing constructions. Evidence is presented in this dissertation which demonstrates that gerundives contribute to the formation of characteristic and episodic readings in sentences in much the same way that bare plurals do, and that they too may serve as complements of kind-predicates. Additional evidence is presented to show that on one reading gerundives denote events. Two fundamental points of view on how to formally represent events are discussed, and an analysis of gerundives proposed by Gennaro Chierchia is shown to have serious flaws. An analysis of gerundives is formulated in which stages in the sense of Greg Carlson's work are abandoned--this is made possible through the use of Davidsonian predicates, since the event argument in such a predicate can be accessed to compute the relevant spatio-temporal slice of an individual under consideration. Also abandoned is Carlson's view that bare plurals unambiguously serve as proper names for kinds of things. Instead, they are taken to denote kinds of things or indefinite manifestations of kinds of things. By asserting that bare plurals have an indefinite reading, it is possible to explain their ability to arise in the object position of existential 'there' constructions, a position in which definite NPs have traditionally been observed to sound anomalous. Gerundives are claimed to be similar to bare plurals, except that instead of denoting indefinite manifestations of kinds on one of their readings, they denote a definite event or state. end of part 2 of 3 *************************