We conducted a series of eye movement experiments in order to empirically examine our MCU hypothesis. We tested whether frequently occurring two-character phrases, three-character idioms with different structures (e.g., “2+1” idioms, 2-character modifier and 1-character noun structure; “1+2” idioms, 1-character modifier and 2-character noun structure; “1+2” idioms, 1-character verb and 2-character noun structure), four-character idioms and frequently occurring four-character phrases, famous people’s names, place names, and popular internet phrases, are likely to be represented and processed as MCUs. All these experiments have shown statistically robust results providing solid support for the MCU Hypothesis that we described in our proposal. The project reshaped our understanding of eye movement control during reading, and opened up a series of interesting research questions in both alphabetic and non-alphabetic languages – what determines a sequence of adjacent words to be processed as a single unit and how visual and lexical processing is operationalized over a flexible unit of text rather than the constant word during reading.