Linear Coverage View
The linear coverage view is the position, advanced by Schmitt, Jiang, and Grabe (2011), that L2 reading comprehension scales as a roughly linear function of lexical coverage across the upper range, with no discontinuous jump at 95% or 98%. The result frames the much-cited 98% and 95% thresholds as conventional reference points on a continuous gradient rather than as empirically distinct cliffs in the comprehension curve.
The Schmitt, Jiang, and Grabe study
The study sampled 661 readers of two academic texts across eight countries. For each text-reader pairing, the authors computed the reader's actual coverage (the share of word families in the text that the reader knew, measured directly with a yes/no vocabulary test built from the texts themselves) and the reader's comprehension score on a written comprehension test of the same text.
Plotting comprehension against coverage produced a smooth, near-linear ascent from the lowest coverage levels through the upper 90s. The authors found no inflection point at 95% and no inflection point at 98%; comprehension simply continued to improve as coverage rose, with the steepest gains accumulating between roughly 90% and 100% coverage. The variance in comprehension at any given coverage level was substantial, meaning that no single coverage figure cleanly separated successful from unsuccessful readers.
Implications
Three implications followed. First, the framing of "the threshold" as a discrete pedagogical target became harder to sustain on the data; coverage targets are conventions chosen for design tractability, not empirical singularities. Second, even at very high coverage many readers fell below adequate comprehension, indicating that lexical knowledge is a necessary but far from sufficient condition. Third, in the authors' reading of the magnitude of effects, 98% coverage remained the more reasonable target for academic reading because of how steeply comprehension fell off below it, even without a sharp threshold.
Status
The linear-coverage finding is widely cited as a corrective to threshold rhetoric, including by Nation in subsequent editions of Learning Vocabulary in Another Language. In practice, materials designers continue to use 95% and 98% as targets because the graded reader and frequency-band machinery requires concrete numbers to grade against, but the theoretical justification for these numbers now rests on the steepness of the gradient near the top of the curve rather than on a clean discontinuity.
Subsequent work in listening (van Zeeland and Schmitt, 2013) and viewing comprehension extended the linear pattern beyond reading, broadly confirming that comprehension scales smoothly with coverage across modalities, though the coverage-comprehension function differs across input types.
References
- Nation, I. S. P. (2013). Learning Vocabulary in Another Language (2nd ed.). Cambridge University Press.
- Schmitt, N., Jiang, X., & Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal, 95(1), 26–43. https://doi.org/10.1111/j.1540-4781.2011.01146.x
- van Zeeland, H., & Schmitt, N. (2013). Lexical coverage in L1 and L2 listening comprehension: The same or different from reading comprehension? Applied Linguistics, 34(4), 457–479. https://doi.org/10.1093/applin/ams074