Whether it is in theoretical works or in automated processing, the most common approach in studies on the relationships between rhythm and meaning in French versified poetry consists in an alignment between the syntactical and metrical structures in order to identify enjambments. I will show that this approach brings poor results and that enjambment is a notion unsatisfying in many respects. Given that rhythm is a decisive characteristic of metrical poetry, I consider as more relevant a corpus-based analysis in which what is systematically and carefully observed are the metrical expressions themselves. The large corpus Malherbǝ from Anamètre project has been used to build-up sub-corpora according to the line's length, the period, the genre and the poet considered. The automated processing uses the punctuation marks as reflecting the syntactic and semantic structures of the metrical expressions (hemistiches, lines, stanza modules, stanzas). The observation focuses on the metrical expressions, their beginning, their end, and their internal consistency. This way of processing, metrical expression after metrical expression, is an attempt to integrate the mental processing time in the analysis itself, that is, the specific reader's expectations. This systematic scrutiny shows up some regularities which will be presented and discussed in this talk.