Landscape ecology has generated much excitement in the past two decades. One reason was that it brought spatial analysis and modeling to the forefront of ecological research. However, high expectations for landscape analysis to improve our understanding and prediction of ecological processes have largely been unfulfilled. We identified three kinds of critical issues: conceptual flaws in landscape pattern analysis, inherent limitations of landscape indices, and improper use of pattern indices. For example, many landscape analyses treat quantitative description of spatial pattern as an end itself and fail to explore relationships between pattern and process. Landscape indices and map data are sometimes used without testing their ecological relevance, which may not only confound interpretation of results, but also lead to meaningless results. In addition, correlation analysis with indices is impeded by the lack of data because of difficulties in large-scale experimentation and by complicated behavior of indices because of their varying responses to changes in scale and spatial pattern. These problems represent significant challenges to landscape pattern analysis, especially in terms of relating pattern to process. In this perspective paper, we examine the underlying problems of these challenges and offer some solutions.