This study used classical psychometric, latent trait, and nonparametric approaches to analyze 13- and 14-item scales of English language proficiency. Tests of English listening comprehension (dictation) and reading ("copytest") were constructed by modifying the standard dictation testing procedure to create items of text segments which varied widely in both length and difficulty. Both the dictation and copytest were found to be homogeneous, cumulative scales of language proficiency with high reliability and validity. Log ability scores provided by Rasch analyses were found to correlate better with other measures of language proficiency than did the dictation and copytest raw scores. These findings indicate that the two language testing techniques investigated provide a useful innovative approach to measuring general aspects of language proficiency. The theoretical and practical advantages of this approach over other language proficiency measurement techniques are discussed as well as implications for measuring language proficiency and other cognitive variables. (Author)