Previous studies suggest that, within particular groups of plant species, biological attributes can be used to predict the potential invasiveness of species that are intentionally introduced for horticultural or agricultural purposes. We examined the broad question of whether commonly available biological information can predict the invasiveness of a wide range of intentionally and accidentally introduced species. We collected information from published floras on 165 pairs of plant species. In each pair, one species originated in Europe and successfully invaded New Brunswick, Canada, and the other was a congeneric species that has not invaded North America. Only three biological characters—lifeform, stem height, and flowering period—and European geographic range were known for all species. We conducted multiple logistic regression analyses using two‐thirds (110) of the species pairs and tested the predictive ability of resulting models using the remaining 55 pairs. Although a significant logistic regression model was obtained using the biological attributes, the model could not predict invasiveness of the test species pairs. In contrast, a model using only European range successfully predicted invasiveness in 70% of the test species. The importance of geographic range suggests that prediction of invasiveness on a species‐by‐species basis is not likely to help stem the flow of accidentally introduced invasive species. Species that are inadvertently picked up and moved to a new location due to their wide distribution are the same species that are likely to succeed in a new environment due to their wide environmental tolerances.

No abstract

roviding adequate support environments for the development of software has been a recognized problem for several years. Unfortunately, support environments have emerged only recently that seriously aid the development of large software systems. Perhaps the most well-known support of these efforts is the Ada programming-support environment. However, it focuses on the life cycle's implementation stage, with facilities t o improve program reliability and t o promote the development of portable software and software-development tools.Several support environments have been developed to aid the requirements-analysis and design stages.',* Recently, commercial CASE environments have been introduced that apply these concepts to the entire life cycle. In addition, some support systems, such as the User Software-Engineering methodology,' extend system analysis and design support with prototyping facilities.McAllister worked on this research while at the University of Saskatchewan. 0740-74S9/88/0300/0030/$01 .OO 01988 IEEE BrunswickThe effort to develop such support environments is considerable, as are enhancements or changes t o a s u p p o r t environment to address an application's special requirements. To significantly reduce the implementation effort of producing a support environment, some researchers have proposed m e t a~y s t e m s .~ A metasystem's primary purpose is to generate automatically the major parts of a particular software-development environment. Metasystems are systems used to develop system-development environments just as compiler-writing systems are systems used to develop compilers. Figure 1 shows a metasystem's architectural overview. Figure l a shows the components of an environment that could be used to specify some aspects of a system description. Many specification-environment support facilities (commonly called tool support) are included, such as graphics interfaces, graphics layout, and query languages. Figure l b shows the generalized metasystem approach to developing an environment-support facility that includes an environment-model IEEE Software

This paper introduces an automated grader for Java programs called GUI_Grader that allows students a degree of flexibility in graphical user interface (GUI) design. GUI_Grader allows students to build multiwindow Java applications, choose among alternative GUI components, and decide how to order, position and label components. This enables students to practice some aspects of designing their own GUI applications while still providing automated grading `based on a single test plan. The data-driven approach helps to maintain consistency between test plans and program specifications. Testing GUI_Grader on Java assignments from both a first-year and an upper-year course confirms the feasibility of the approach.

SUMMARYOne of the most significant challenges in the use of entity relationship data models is in deciding whether to use a single relationship between several entities or a set of simpler relationships to represent a complex association. A practical eight-step approach is presented for analysis of n-ary relationships and decomposition into simpler relationships where appropriate. Relational database design concepts form the basis for this approach. An extended specification of cardinality constraints is used to support the decomposition approach and to ensure applicability to a variety of modeling styles. This approach defines a fundamental analysis skill for data modeling practitioners.