When we read, our eyes don’t move smoothly across the page; instead, they make series of quick jumps called saccades, followed by brief pauses known as fixations. During fixations – typically lasting 200-300 milliseconds – the brain processes the information in focus².

Interestingly, we don’t fixate on every word. Skilled readers tend to skip shorter or highly predictable words and adjust their reading pace depending on the text difficulty. When the meaning becomes unclear or the syntax is challenging, we may backtrack – this is known as a regression².

Research has shown that fluent reading follows some predictable eye movement patterns: fixations cluster around content words like nouns and verbs, while function words such as “the” or “and” are often skipped²; saccades usually span 7-9 characters²; regressions occur in response to ambiguity³; and initial passes through the text are frequently followed by re-reading key sections⁴. These patterns can be disrupted by factors like poor text structure, low screen contrast, or increased cognitive load^5,6.

Unlike traditional reading research methods that rely on self-reports or reaction times, eye tracking offers a precise, real-time window into the reading process. Modern eye tracking systems capture exactly where someone is looking, how long they dwell on each word, and whether they skip or revisit sections of text. This data allows researchers to analyze the cognitive mechanisms underlying reading in detail.

Commonly used metrics include fixation duration (how long the eyes pause on a word), saccades (quick jumps between fixations), and regressions (backward movements indicating rereading or confusion). Researchers also examine scanpaths, which reveal the overall pattern of eye movements during reading. Additional metrics such as first-pass reading time, total reading time, and skipping rate help uncover how readers process language, identify word predictability, and manage cognitive load⁷. Captured at millisecond level precision, these measures provide valuable insights into language comprehension, attention, and working memory – enabling the development of robust models that reflect the complexity of reading behavior.