Text difficulty and the Arens-Altmann law

The study of text difficulty is considered to be an important issue for many branches of applied research. In the fields of journalism or education, for example, it is particularly important to know if (or to what degree) a given text is likely to cause difficulties for a recipient, or a group of recipients, i.e., if it is likely to be on an adequate (intended) level of difficulty or beyond. In order to achieve this goal, a specific line of text difficulty research has developed over the last decades, beginning in the 1920s, which attempts to combine linguistic analysis with informants’ ratings of text difficulty.1 Text difficulty thus is a double-faced kind of empirical research in two directions, either of which may be emphasized in individual studies: it is text-based, on the one hand, and informant-oriented, on the other. Due to this dual perspective, alternative terms such as ‘text readability’2 or ‘text comprehensibility’ have been used to refer to the related area(s) of research, the first term emphasizing the predominantly written (rather than oral) basis of communication, the second being broader in its understanding. Compared to these alternatives, ‘text difficulty’ as a term primarily refers to the analysis of linguistic structures, aiming at the identification and characterization of linguistic factors rendering a given text more or less easily comprehensible to a given person (or a group of persons), and at the (cor)relation of these structures to informants’ ratings about text difficulty. Such a definition is in line with research from the last decades: Klare (1963: 1), for example, understands this term as referring to “the ease of understanding or comprehension due to the style of writing”, and DuBay (2004: 3), more recently, has defined the overall aim of text difficulty research as the study of “what makes some texts easier to read than others”. Given this general orientation, research in this field, from its beginnings on, has continually tried to develop, modify and improve formulae to predict text difficulty and, by way of it, prognose comprehension ability. This is to say that attempts have been undertaken to develop measures of text difficulty, including formulae which combine quantitative (or quantified) linguistic characteristics in such a way that these characteristics serve as (possibly combined and