Faculty Abstracts A-C

• ASTRONOMY
• BIOLOGY
• CHEMISTRY
• COMPUTER SCIENCE

ASTRONOMY

Kan Balam Meets Halley's Comet

Samuel Y. Edgerton, Jr., Professor of Art, Jay M. Pasachoff 1993, Professor of Astronomy

Archaeology, November/December 1993, p24, by Angela M. H. Schuster and in Corriere de la Sera; submitted to Ancient Mesoamerica

Astronomy Education in the United States: U.S. National Report on Astronomy Education

Jay M. Pasachoff, Professor of Astronomy

I.A.U. Commission 46 Newsletter

Coronal Images from the 1984 Solar Eclipse

Jay M. Pasachoff, Professor of Astronomy, Kevin P. Reardon '92, and John MacKenty 1993

Solar Physics 146, 405-408

We present digitized photographs of the white-light solar corona taken during the total solar eclipse of 22-23 November, 1984, on both calibrated black-and-white film and on color film. Conditions on site in Hula, Papua New Guinea, were exceptionally clear. The color image was used to produce an isophotal map of the inner corona, from which a flattening coefficient of 0.23 was measured. The black-and-white image was enhanced through a digital radial filter. Our images are the best processed images available from the 1984 eclipse and so provide important data for synoptic observations.

"Dazzled by Solar Eclipses"

Deborah Pasachoff with Jay M. Pasachoff 1993, Professor of Astronomy

Odyssey 2, #6 (July-August), 14-19

"Diffuse X-Ray Emission From the Dumbbell Nebula?"

Y.-H. Chu, K.B. Kwitter, Professor of Astronomy, and J.B. Kaler

Astronomical Journal, 106, 650, 1994

We have analyzed ROSAT Position Sensitive Proportional Counter pointed observations of the Dumbbell Nebula and find that the previously reported "extended" x-ray emission is an instrumental electronic ghost image at the softest energy band. At slightly higher energy bands, the image of the Dumbbell is not very different from that of the white dwarf HZ43. We conclude that the x-ray emission of the Dumbbell Nebula comes from its central star. A blackbody model is fitted to the spectrum and the best-fit temperature of ¾136,000 ±10,000 K is in excellent agreement with the Zanstra temperatures.

BIOLOGY

Sphingolipid Long-Chain Base Synthesis in Plants

Daniel V. Lynch, Assistant Professor of Biology and Scott R. Fairfield

Plant Physiology, 103: 1421-1429 (1993)

The activity of serine palmitoyltransferase (palmitoyl-CoA:L-serine-C-palmitoyltransferase (decarboxylating), EC2.3.1.50), the enzyme catalyzing the first step in the synthesis of long-chain base required for sphingolipid assembly, has been characterized in a plant system. Enzyme activity in a microsomal membrane fraction from summer squash fruit (Cucurbita pepo L. cv Early Prolific Straightneck) was assayed by monitoring the incorporation of L-(3H)serine into the chloroform- soluble product, 3-ketosphinganine. Addition of NADPH to the assay system resulted in the conversion of 3-ketosphinganine to sphinganine. The apparent Km for serine was approximately 1.8 mM. The enzyme exhibited a strong preference for palmitoyl-CoA, with optimal activity at a substrate concentration of 200 mM. Pyridoxal 5'-phosphate was required as a coenzyme. The pH optimum was 7.6 and the temperature optimum was 36 to 40 C. Enzyme activity was greatest in the microsomal fraction obtained by differential centrifugation, and, using marker enzymes, was localized to the endoplasmic reticulum. Two known mechanism-based inhibitors of the mammalian enzyme, L- cycloserine and b-chloro-l-alanine, were effective inhibitors of enzyme activity in squash microsomes. Changes in enzyme activity with size (age) of squash fruit were observed. The results from this study suggest that the properties and catalytic mechanism of serine palmitoyltransferase from squash are similar to those of the animal, fungal and bacterial enzyme in most respects. However, the specific activity of the enzyme from squash microsomes was two-fold to 20-fold higher than values previously reported for preparations from animal tissues.

A Comparison of the Impacts of Leaf and Sap-Feeding Insects on Growth and Allocation of Goldenrod

G. A. Meyer, Assistant Professor of Biology

Ecology, 74: 1101-1116 (1993)

Many plants support both leaf and saf-feeding insects, yet comparative studies of plant responses to different types of damage are rare. I compared the impacts of a leaf-chewing beetle (Trirhabda sp.), a xylem-sap feeding spittlebug (Philaenus spumarius) and a phloem-sap feeding aphid (Uroleucon caligatum) on growth of goldenrod (Solidago altissima). My goals were (1) to measure insect impacts on plant size as a function of herbivore loads, and (2) to determine the mechanisms by which insect feeding affects plant growth, using the techniques of plant growth analysis. Plants were grown outside in large pots and insects were placed on them at a range of densities. A series of four whole-plant harvests was performed, with the final harvest occurring after 3 weeks of insect feeding. Insect mass gain per plant was used as the basis for the comparison, since the insects differed greatly in size.

Spittlebugs were the most damaging insect, beetles were intermediate, and no effects of aphid feeding were detected. After 3 weeks of feeding, both spittlebugs and beetles reduced total leaf mass, total leaf area, and root mass, and the impact of the spittlebug was 5 to 6 times greater than that of the beetle. Spittlebug feeding also reduced the mass of the apical buds, stem mass, and the number of lateral stems, while beetle and aphid damage had no effect on these variables. Dry mass allocation to leaves, stems and roots was generally not affected by insect feeding, resulting in smaller but proportionally similar plants by the end of the feeding period.

Plant relative growth rates (dry mass increase per unit dry mass per day) were reduced by both spittlebugs and beetles but not aphids, and the impact of the spittlebug was more severe than that of the beetle. None of the insects affected net assimilation rates (dry mass increase per unit leaf area per day), indicating that the insects did not reduce plant growth by altering plant physiology. Instead, reductions in relative growth rates were explained by changes in plant morphology; feeding by both spittlebugs and beetles strongly reduced specific leaf area (leaf area/leaf mass). This reduction of leaf area relative to leaf mass was the mechanism by which both insects decreased plant relative growth rates. This result suggests an explanation for why simulated damage frequently fails to mimic damage by actual herbivores. Then leaves are artificially removed by clipping there will be an immediate change in dry mass allocation patterns, but specific leaf area will not necessarily be affected.

Effects of Herbivorous Insects and Soil Fertility on Reproduction of Goldenrod

G.A. Meyer, Assistant Professor of Biology, and R.B. Root

Ecology, 74: 1117-1128 (1993)

Plant responses to herbivore damage vary enormously, but the causes of this variation are not well understood. We examined how both insect feeding style and soil fertility influence the severity of herbivore impacts, by measuring the effects of feeding by a leaf-chewing beetle (Trirhabda sp), a xylem-sap feeding spittlebug (Philaenus spuraius) and a phloem-sap feeding aphid (Uroleucon caligatum) on growth and reproduction of goldenrod (Solidago altissima). Our goals were to determine (1) whether these three insects differentially affect plant fitness, (23) whether soil fertility affects a plant's ability to compensate for damage, and (3) whether soil fertility and insect feeding style interact in determining herbivore impacts. Plants were grown outside in large pots at two levels of soil fertility, and insects were placed on them at a range of densities. Insects fed on the plants for approximately 3 weeks in spring, and then the plants were grown insect-free for the rest of the season. We used insect mass gain per plant as the basis for comparing the effects of three insects, since they differed greatly in size.

Spittlebug feeding reduced plant growth rates 3 times more than beetle feeding; aphid feeding had no detectable effect. The plants grew faster at higher soil fertility, but there was no interaction between insect impact and fertilizer level. Stem heights at the end of the season were increased by fertilizer and not affected by herbivory. Spittlebug feeding, and to a lesser extent aphid feeding, decreased the number of lateral stems, and again there was no interaction with soil fertility.

Soil fertility had strong effects on plant sexual and asexual reproduction. Plants at the high level of soil fertility showed a 3-fold increase in total seed production, a 5% increase in mean achene mass, almost a 2-fold increase in rhizome mass, and increased percent nitrogen in seeds and rhizomes. Total seed production was reduced by herbivory. The pattern was the same as for plant growth rates; the spittlebug caused the strongest decrease, beetle feeding was second, and aphid feeding caused slight reductions. There was a strong interaction between insect impact and soil fertility. Total seed production was reduced only at the high level of soil fertility for all three insects. Spittlebug and beetle damage also delayed flowering, but only at the high level of soil fertility. Rhizome mass, mean achene mass, and percent nitrogen in seeds and rhizomes were not affected by herbivory.

The common assumption that plants are better able to compensate for herbivore damage at higher levels of soil fertility must be reconsidered. This study also suggests that sexual reproduction may be more sensitive to damage than asexual reproduction in perennial plants.

Vaccinia Virus Ribonucleotide Reductase Expression and Isolation of the Recombinant Large Subunit

Mary B. Slabaugh, Ralph E. Davis, Nancy A. Roseman, Assistant Professor of Biology, and Christopher K. Mathews

Journal of Biological Chemistry , 268(24): 17803-17810 (1993)

The vaccinia virus gene encoding the 87-kDa protein that comprises the large subunit of ribonucleotide reductase (vvR1) was cloned into a bacterial expression vector under the control of an inducible promoter. Culture of Escherichia coli cells harboring the recombinant plasmid under standard induction conditions (0.4 mM isopropyl ß-D-thiogalactopyranoside, 37oC) resulted in synthesis of a completely insoluble product. Production of soluble vvR1 was achieved by growing bacteria at low temperature (15oC) during the induction period, initiating induction at low cell density, and using a low concentration (0.05 mM) of the inducer isopropyl ß-D-thiogalactopyranoside. Hydroxyurea, an inhibitor of ribonucleotide reductase, increased production of soluble vvR1 in a dose-dependent manner. Recombinant vvR1 was purified from a high salt extract of the E. coli lysine in four steps, the last utilizing an affinity column consisting of the carboxyl-terminal seven amino acids of the protein linked to an insoluble resin. Using purified recombinant vvR2 to reconstitute active enzyme, we determined that maximizing the rate of CDP reduction required pH 8.0-8/8, 50 mM dithiothreitol, and 2 nM ATP. Specific activity of purified vvR1 was 122 nmol/min/mg. Limited proteolysis of the vvR1 protein revealed protease-resistant fragments approximately 30 and 59 kDa in size. To our knowledge, this study represents the first expression, solubilization, and isolation of a recombinant "eukaryotic" form of ribonucleotide reductase large subunit.

Temporal Patterning of Song Production: Participation of Nucleus Uvaeformis of the Thalamus

Heather Williams, Assistant Professor of Biology, and David S. Vicario

Journal of Neurobiology, ,24: 903-912 (1993)

Birdsong is a learned vocal behavior used in intraspecific communication. The motor pathway serving learned vocalizations includes the forebrain nuclei NIf, HVC, and RA; RA projects to midbrain and brainstem areas that control the temporal and acoustic features of song. Nucleus Uvaeformis of the thalamus (Uva) sends input to two of these forebrain nuclei (NIf and HVC) but has not been thought to be important for song production. We used three experimental approaches to re-examine Uva's function in adult male zebra finches. 1) Electrical stimulation applied to Uva activated HVC and the vocal motor pathway, including tracheosyringealmotor neurons that innervate the bird's vocal organ. 2) Bilateral lesions of Uva including the dorso-medial portion of the nucleus affected the normal temporal organization of song. 3) Chronic multi-unit recordings from Uva during normal song and calls show burst of pre-motor activity that lead the onset of some song components, and also larger bursts that mark the end of complete song motifs. These results implicate Uva in the production of learned vocalizations, and further suggest that Uva contributes more to the temporal structure than to the acoustic characteristics of song.

Origin and Function of Spiral Fibers Projecting to the Goldfish Mauthner Cell

John W. Scott, Steven J. Zottoli, Professor of Biology, Nicholas P. Beatty and Henri Korn

Journal of Comparative Neurology, 339: 76-90 (1994)

Two neuron types contact the Mauthner cell (M cell) in the zcon cap, a specialized region of high electrical resistance surrounding the initial segment of the M cell axon. One type produces a mixed electrical and chemical inhibition of the M cell. The second sends axons into the central core of the axon cap, where they spiral around the initial segment making both conventional synapses and gap junction contacts. The origin and synaptic effects of these spiral fibers have not been studies previously.

When goldfish M cells were filled with Lucifer yellow, pre synaptic spiral fibers were seen in the axon cap. These fibers could be traced back through the medial longitudinal fasciculus to their somata, near the contralateral fifth nerve motor nucleus. The same somata were labeled by horseradish peroxidase injected extracellularly into the axon cap.

Recordings were made in the axon cap and the M cell after stimulation of hindbrain areas near the spiral fiber somata and axons. Extracellularly, a negative potential was observed close to the termination of the spiral fibers and termed the spiral fiber potential (SFP). Intracellularly, a graded, short latency depolarization of the M cell corresponded to the SFP and could cause the M cell to spike. This depolarization did not shunt the membrane, indicating that it may be produced through gap junctions. Intracellular responses to hindbrain stimulation also had a chloride-dependent, second component that shunted the membrane during paired-pulse testing. This inhibitory second component was probably evoked by cells other than the spiral fiber cells themselves.

Julia B. Platt (1857-1935): Pioneer Comparative Embryologist and Neuroscientist

Steven J. Zottoli, Professor of Biology, and Ernst-August Seyfarth

Brain, Behavior and Evolution, 43: 92-106 (1994)

Julia Barlow Platt was a comparative embryologist and neurobiologist who was primarily interested in segmentation of the head in vertebrates. She was born on September 14, 1857 in San Francisco, California. Platt grew up in Burlington, Vermont, attended the University of Vermont and began graduate studies at Harvard University. Her nine years as a graduate student were spent on two continents with some of the most influential comparative zoologists of the time. Platt's remarkable scientific accomplishments over a ten year period include a description of axial segmentation currently used in the staging of chick embryos and the first description of a separate anterior head segment in Squalus embryos. Her most controversial study identified ectodermal cells in Necturus embryos that gave rise to head cartilage and dentine, a discovery which was the impetus for the reassessment and modification of the germ layer concept. She was one of the first women to 'matriculate' at a German university and receive a Ph.D. degree. Platt played a pioneer role in opening opportunities for other women who followed her.

Platt was one of the first women neuroscientists. Among her contributions. she distinguished dorsolateral placodes, epibranchial placodes, and the first stages of lateral line organs in Necturus, and she described nerve fibers originating in the spinal cord and extending to the notochord in Branchiostoma ( = Amphioxus).

After receiving a Ph.D. degree in Freiburg, Germany in 1898, Platt was unable to secure a suitable teaching position and, as a result, her scientific career came to an end. She retired to Pacific Grove, California, where she pursued civic duty with the same vigor and energy she had dedicated to scientific research. We provide a sketch of her remarkable life and work as a comparative embryologist, neuroscientist and civic leader.

Comparative Studies of the Mauthner Cell in Teleosts

Steven J. Zottoli, Professor of Biology, Graeme W. Davis and Susan C. Northern

Neurology Today, 53-63 (1992)

Alberto Stefanelli has been an advocate of the comparative morpho-ecological approach to the study of Mauthner cells in fish and amphibians. As a result of his extensive studies, he has proposed that the M-cell may play a broad role in "starting" a number of behaviors, in addition to startle reactions. Thus, the M-cell might be involved in more than one behavior depending on the behavioral context in which the fish finds itself.

In this paper I review some of our recent work on two morphologically and ecologically diverse species, the winter flounder and sea robin and compare these to the extensive investigations on the goldfish, the typical species for such studies. Although the Mauthner cells and their axons are smaller in sea robins and winter flounder when compared to those of other fish such as the goldfish, the cells appear to be similar in their general morphology and physiology. All these teleosts can display a startle response to an abrupt sound stimulus. As a result, one is tempted to speculate that the M- cell may function as part of startle behavior in all fish. However, many fish in which the M-cell cannot be found can give startle responses, and the M-cell is not necessarily needed for all such responses.

Central Nervous System Lesion Triggers Inappropriate Pathway Choice in Adult Vertebrate System

Adrienne P. Bentley, Steven J. Zottoli, Professor of Biology

Brain Research, 630: 333-336 (1993)

Damaged neurons within the CNS of the goldfish are able to regrow to appropriate target areas with resultant recovery of swimming behavior. However, after a whole spino-medullary level crush, many adult goldfish do not recover all behavior. Brain neurons regenerating past a crush wound at this level have a choice between the spinal cord and the first ventral root. Many CNS neurons faced with this decision do not make the same pathway choice as they made during development but rather project axons into the first ventral root, away from their normal target areas in the spinal cord. In fact, more regenerating fibers, including those of reticulospinal and vestibulospinal neurons, choose the peripheral nervous system (PNS) over the CNS, which may limit behavioral recovery. The goldfish PNS may present a more permissive environment to regenerating fibers than the CNS, as is the case in mammals. We suggest that the goldfish is a better model for mammalian regeneration than previously thought.

CHEMISTRY

Forensic Science: Crime in the Chemistry Curriculum

Lawrence J. Kaplan, Professor of Chemistry

Journal of Chemical Education, 70, 574 (1993).

Much has been written about and many symposia have focused on the current state of science education in the undergraduate curriculum and on the status of the laboratory experience as an integral component of science courses for non-majors as well as majors. One consensus has emerged: a series of unrelated experiments (apparently) illustrating unrelated concepts is not effective in stimulating students' interest, capturing their attention and motivating them toward further study in the sciences. At Williams College, we addressed these issues with the development of a course for nonscience majors called Chemistry and Crime: From Sherlock Holmes to Modern Forensic Science. The institution of the creative, exciting, and real life experiments which serve to illustrate the concepts presented in the lecture make the material relevant to the student's experiences and background. Forensic science is the perfect vehicle for this because it is exciting and it cuts across many scientific disciplines.

Chemistry and Crime: From Sherlock Holmes to Modern Forensic Science - A Science Course for Non-Science Majors

Lawrence J. Kaplan, Professor of Chemistry

Crime Laboratory Digest, 19(4), 107-132 (October 1992; actually published in February 1993)

The structure of a course in forensic science, formally called Chemistry and Crime: From Sherlock Holmes to Modern Forensic Science, is discussed with illustrations of how the scientific, technological and legal aspects of crime detection are woven into this unique course for nonscience majors. A course in forensic science is an ideal vehicle to introduce students who are not majoring in the natural sciences to the way in which scientists actually work; the way they define a problem, collect data, analyze the results and then present those results in meaningful ways. The course includes a full scale laboratory program which allows the students to have a practical hands-on appreciation for many of the procedures discussed in the class and for the operation of a "crime lab." A unique feature of the laboratory experience is the assignment of the students to an investigative unit which is called to the scene of a crime. There the students process the crime scene collecting the physical evidence which will be studied in the crime lab. Crime scenarios such as a hit-and-run accident, a drug bust, or a burglary provide ideal opportunities for the students to catch the excitement of forensic science.

Purification of an SOS Repressor from Bacillus subtilis

Charles M. Lovett, Jr., Associate Professor of Chemistry, Kerry C. Cho '92, and Thomas M. O'Gara

Journal of Bacteriology, 175, 6842-6849 (1993)

We have identified in Bacillus subtilis a DNA binding protein that is functionally analogous to the Escherichia coli LexA protein. We show that the 23-kDa B. subtilis protein binds specifically to the consensus sequence 5'-GAACN4GTTC-3' located within the putative promoter regions of four distinct B. subtilis DNA damage-inducible genes: dinA, dinB, dinC, and recA. In RecA+ strains the protein's specific DNA binding activity was abolished following treatment with mitomycin C; the decrease in DNA binding activity after DNA damage had a half-life of about 5 minutes and was followed by an increase in SOS gene expression. There was no detectable decrease in DNA binding activity in B. subtilis strains deficient in RecA (recA1, recA4) or otherwise deficient in SOS induction (recM13) following mitomycin C treatment. The addition of purified B. subtilis RecA protein, activated by single-stranded DNA and dATP, abolished the specific DNA binding activity in crude extracts of RecA+ strains and strains deficient in SOS induction. We purified the B. subtilis DNA binding protein more than 4000-fold using an affinity resin in which a 199 bp-DNA fragment containing the dinC promoter region was coupled to cellulose. We show that B. subtilis RecA inactivates the DNA binding activity of the purified B. subtilis protein in a reaction that requires single-stranded DNA and nucleoside triphosphate. By analogy with E. coli, our results indicate that the DNA binding protein is the repressor of the B. subtilis SOS DNA repair system.

Analysis of the SOS Inducing Signal in Bacillus subtilis using Escherichia coli LexA as a Probe

Charles M. Lovett, Jr., Associate Professor of Chemistry, Thomas M. O'Gara and James N. Woodruff

Journal of Bacteriology, 176, in press (1994)

We analyzed the Bacillus subtilis SOS response using Escherichia coli LexA protein as a probe to measure the kinetics of SOS activation and DNA repair in wild-type and DNA repair-deficient strains. By examining the effects of DNA damaging agents that produce the SOS inducing signal in E. coli by three distinct pathways, we present evidence that the nature of the SOS inducing signal has been conserved in B. subtilis. In particular, we used the B. subtilis DNA polymerase III inhibitor, 6-(p- hydroxyphenylazo)-uracil, to show that DNA replication is required to generate the SOS inducing signal following UV irradiation. We also present evidence that single-stranded gaps, generated by excision repair, serve as part of the UV inducing signal. By assaying the SOS response in B. subtilis dinA, dinB, and dinC mutants we identified distinct deficiencies in SOS activation and DNA repair that suggest roles for the corresponding gene products in the SOS response.

15N-NMR Spectra of Carbolines

J. Hodge Markgraf, Professor of Chemistry, David P. Richardson, Associate Professor of Chemistry, and Robert C. Livingston '92

Magnetic Resonance in Chemistry, 31, 694-695 (1993).

The 15N nuclear magnetic resonance spectra of a-, b-, g-, and d-carboline and a series of methyl, dihydro, and tetrahydro derivatives were determined. The nitrogen chemical shift assignments provided a new and convenient basis for structural assignments of this important class of alkaloids.

Synthesis of Five- and Six- Coordinate Alkylidene Complexes of the Type Mo(CHR)(NAr)[OCMe(CF3)2]2Sx and Their Use as Living ROMP Initiators or Wittig Reagents

Lee Y. Park, Assistant Professor of Chemistry

Organometallics , 12, 759 (1992)

A variety of Mo(VI) hexafluoro-tert-butoxide alkylidene complexes can be prepared if they are stabilized by internal or external coordination of a base. These compounds can be used as living ROMP initiators and they also react smoothly with a variety of aldehydes in Wittig-like reactions. The techniques described should expand the opportunities for making polymers via living ROMP. In particular, one can now introduce conjugated sequences of known, fixed length into a variety of polymers and to attach functional groups to both ends of the polymers.

A Model for Temperature-Dependent Collisional Quenching of NO A2S+

P.H. Paul, J.A. Gray, J.L. Durant, Jr., and J.W. Thoman, Jr., Assistant Professor of Chemistry

Applied Physics B 57, 249 (1993).

A model for the temperature-dependent electronic quenching of NO A2S+ is presented. The model is appropriate for collision partners with stable negative ions, which are predicted to have large quenching cross-sections. Colliders with relatively large negative electron affinities are predicted to have near-zero cross-sections at room temperature which can increase dramatically at very high temperatures. A negligibly small electronic quenching cross-section is predicted for collision partners having negative ions which are unstable on a collisional time scale. Results of the model are compared to experimentally measured cross-sections for a number of species of interest in combustion and aerothermodynamic applications

COMPUTER SCIENCE

Safe and Decidable Type Checking in an Object-Oriented Language

Kim B. Bruce, Professor of Computer Science, Jon Crabtree'93, Thomas P. Murtagh, Associate Professor of Computer Science, Robert van Gent '93, Williams College and Allyn Dimock, Robert Muller, Harvard University

Proceedings of the ACM Symposium on Object-Oriented Programming, Systems, Languages, and Applications, 29-46 (1993)

Over the last several years, much interesting work has been done in modelling object-oriented programming languages in terms of extensions of the bounded second-order lambda calculus, F¾. Unfortunately, it has recently been shown by Pierce that type checking F¾ is undecidable. Moreover, he showed that the undecidability arises in the seemingly simpler problem of determining whether one type is a subtype of another.

In two earlier papers, the first author introduced a statically-typed, functional, object-oriented programming language, TOOPL, which supports classes, objects, methods, instance variables, subtypes, and inheritance. The semantics of TOOPL is based on F¾, so the question arises whether type checking in this language is decidable.

In this paper we show that type checking for TOOPLE, a minor variant of TOOPL (Typed Object- Oriented Programming Language), is decidable. The proof proceeds by showing that subtyping is decidable, that all terms of TOOPLE have minimum types (which are in fact computable), and then using these two results to show that type checking is decidable. Our algorithm fails to be polynomial in the size of the term because the size of its type can be exponential in the size of the term. Nevertheless, it performs well in practice.

Attracting (& Keeping) the Best and the Brightest: An Entry-Level Course for Experienced Introductory Students

Kim Bruce, Professor of Computer Science

Proceedings of the 1994 ACM SIGCSE Conference, 243 - 247 (1994)

This course addresses the design of an exciting introductory course for entering college students who have some programming background. The key innovation of the course is the introduction of material pertaining to experimental graph reduction and dataflow architectures for highly parallel computers. The success of the course is attributed to interesting, well-integrated material and the use of visually oriented interactive simulators for the architectures.

A Modest Model of Records, Inheritance, and Bounded Quantification

Kim B. Bruce, Professor of Computer Science and Giuseppe Longo, Ecole Normale Superiéure

Originally published in Information and Computation, 87, 196-240 (1990) Reprinted in Theoretical Aspects of Object-Oriented Programming, MIT Press, 151 - 195 (1994)

This paper provides the first denotational semantics for the language, Bounded Fun, a statically typed object-oriented computer language which supports both parametric and subtype polymorphism. Parametric polymorphism is similar to the "generics" in Ada, and provides the ability to support parameterized type functions (such as List(T), Tree(T), Stack(T), etc.) as well as operations which work on the types generated from these functions. Subtype polymorphism is based on subtyping, an important concept in object-oriented programming languages, and might also be called structural polymorphism. The semantics provided here uses partial equivalence relations to model types and subtypes in the language.

An Art Gallery Theorem for Line Segments in the Plane

George F. Jennings '90 and William J. Lenhart, Associate Professor of Computer Science

Pattern Recognition Letters, 14, 727-732 (1993)

Given a set L of n non-intersecting line segments in the plane, we show that it is possible to choose a set S of at most Î n/2û segments such that for each segment l of L there exists a point pl on one of the segments in S which sees every point of l. That is, for any point p on segment l the segment does not intersect the interior of any line segment other than those containing p and pl. This bound is also shown to be tight. Thus, by imagining that each segment of S contains an edge guard, we conclude that Î n/2û edge guards are sometimes necessary and always sufficient to guard any set of n segments in the plane.