1. IntroductionThe Generations and Gender Survey (GGS), which is part of the Generations and Gender Programme (GGP), has been one of the most ambitious projects in Population Studies in the last decade. Its stated goal is to improve understanding of demographic and social development and of the factors that influence these developments (UNECE 2005). This requires elaborate survey design and implementation, resulting in a long and complex questionnaire. Previous research has shown that with increasing length and complexity, the probability of distortions in a survey rises. In this paper we focus on the German GGS, in which distortions appear in parts of the questionnaire which are complex, i.e., the retrospective parts on fertility and partnerships. There are various types of distortion: compared to German vital statistics the total number of children in the GGS is too low for birth cohorts 1930-1954 and too high for the cohorts born thereafter. When looking at partnership history we have too many women who were never married in our data in the older cohorts and too many married ones in the younger cohorts (Kreyenfeld et al. 2010; 2011; Kreyenfeld, Hornung, and Kubisch 2013; Sauer, Ruckdeschel, and Naderi 2012; Vergauwen et al. 2015). So far, approaches explaining these distortions have not provided satisfactory results. In this paper we try another approach by investigating if and under which conditions the complexity of the questionnaire could hinder the GGS improving the knowledge of the factors that influence demographic and social development. We investigate if the distortions in the German GGS can be explained by questionnaire design factors, interview situations, or other aspects of the implementation of the survey. In the following we especially assume that length, complexity, and structure of the GGS questionnaire are factors that affect respondents' as well as interviewers' behaviour and which offer advantageous conditions for shortening interviews by recording or giving incomplete or wrong answers.The paper is structured as follows: first, it presents a summary of the current state of research concerning possible causes of these distortions. Next, we describe the GGS questionnaire and the distortions in the German GGS in greater detail. Based on this, hypotheses are formulated concerning the nature and causes of the distortions in the GGS. Data and methods used as well as the results are presented in the subsequent sections. Finally, the conclusion suggests lessons for future surveys and provides advice for handling data from the German GGS in empirical analyses.2. Possible distortions due to survey implementationAfter discovering and identifying the distortions in the German GGS, different explanations have been posited, but with limited success. Kreyenfeld and her colleagues came to the conclusion that wording, question placement and technical problems do not seem to be the major sources of error leading to the bias in the fertility (Kreyenfeld, Hornung, and Kubisch 2013: 20). A paper by Sauer, Ruckdeschel, and Naderi (2012) instead concluded that the complexity of the retrospective histories could be a main source of distortion. In this paper we will elaborate upon this aspect.The literature on sources of distortion in surveys covers a wide range of topics, such as the characteristics of the event itself (saliency of event), respondent characteristics, and survey design (Blasius and Thiessen 2012). The design of the questionnaire is a decisive factor in the accuracy of answers, as is the interview situation. In this context most approaches explain interviewers' and interviewees' response behaviour and social desirability by rational choice theory (e.g., Stocke 2004, 2007). This also applies to the description and explanation of possible distortions in the German GGS, where rational choice theory offers a useful approach (e.g., Esser 1985, 1990; Tourangeau, Rips, and Rasinski 2000). …