Barriers to Adoption of Dictionary-Based Text-Entry Methods: A Field Study

MacKenzie et al. (2001) published user-study and simulation results which suggest that text-entry speed on cell phone keypads backed by predictive dictionary-based software is highly dependent on the fraction of words the user needs which are present in the dictionary. They found that when the fraction of non-dictionary words is greater than approximately 15% percent, dictionary-based methods are slower than multi-tap. To assess the relevance of these laboratory findings to user experience in practice, we conducted a field study of the use of these dictionary-based methods. The field study comprised 230 interviews with people who had cell phones and who had some experience with dictionary-based text-entry methods. Participants were asked questions concerning their use of messaging and of a dictionary-based method, if any. Non-users most frequently cite difficulty of use or dictionary deficiencies as their reason for not using a dictionary-based method, and dictionary deficiencies are also often cited by users. We supply evidence that dictionary deficiencies contribute to making the systems hard to learn as well as hard to use. A simulation was performed to account for this latter result. In this survey, somewhat under half of those exposed to predictive text use it. This is to be compared to the results of Doring (2002) who reports a predictive-text usage rate of 30%.