Data of sequential nature arise in many application domains in forms of, e.g.\ntextual data, DNA sequences, and software execution traces. Different research\ndisciplines have developed methods to learn sequence models from such datasets:\n(i) in the machine learning field methods such as (hidden) Markov models and\nrecurrent neural networks have been developed and successfully applied to a\nwide-range of tasks, (ii) in process mining process discovery techniques aim to\ngenerate human-interpretable descriptive models, and (iii) in the grammar\ninference field the focus is on finding descriptive models in the form of\nformal grammars. Despite their different focuses, these fields share a common\ngoal - learning a model that accurately describes the behavior in the\nunderlying data. Those sequence models are generative, i.e, they can predict\nwhat elements are likely to occur after a given unfinished sequence. So far,\nthese fields have developed mainly in isolation from each other and no\ncomparison exists. This paper presents an interdisciplinary experimental\nevaluation that compares sequence modeling techniques on the task of\nnext-element prediction on four real-life sequence datasets. The results\nindicate that machine learning techniques that generally have no aim at\ninterpretability in terms of accuracy outperform techniques from the process\nmining and grammar inference fields that aim to yield interpretable models.\n