Magical Realism and Computational Analysis2016

Magical realism is one of the most widely recognized modes in world literature, but scholars have struggled for decades to define what it actually is. The term has been debated since 1925, tangled up with fantasy and the Latin American Boom's marketing machinery. Most critics agree on one thing: magical realism is reproducible. Writers from entirely different backgrounds keep producing it, and readers keep recognizing it. But nobody has been able to point to concrete, measurable qualities that make a text magical realist. My thesis asked a simple question: what if those qualities exist but are invisible to close reading? I assembled a corpus of magical realist texts spanning multiple regions and historical periods, then ran unsupervised computational methods (topic modeling and text mining) to look for patterns no human reader had articulated. Unsupervised methods were key: rather than telling the algorithm what to look for, I let the texts speak for themselves. The analysis surfaced two unexpected markers. First, themes of death and violence appeared consistently across the corpus, regardless of the author's cultural background. Second, the texts showed a broader shift toward urbanized settings than the mode's pastoral reputation would suggest. Together, these findings support reading magical realism as a mode of resistance, not tied to any single tradition, but recurring wherever writers push back against dominant narratives. This was the first application of digital humanities methods to the magical realism archive. The hope is that it opens a door: if the empirical fingerprint of magical realism is consistent across regions and eras, then computational analysis at a much larger scale could finally give the mode the kind of formal definition that has eluded critics for a century.