The compound leaves of our Berberis species disarticulate at points of leaflet attachment along the rachis, by separation of leaflet or rachis segment or both - above the basal attachment of the leaf petiole. The base of the petiole does not form an abscission zone, and the section below the basal node may persist for many years. Which sections will separate is apparently determined by forces external to the plant, such as wind, rain, contact with other objects. The abscission joints are clearly marked by swelling as well as (sub)annular ridges and grooves.
 |  |
|---|---|
| B. swaseyi short shoot with remnants of 7 leaves. (Nodes are indexed #1 to #n counting from the base.) | Petioles persist for many years. |
The cooccurrence of two features, (1) the disarticulation of individual rachis segments and (2) morphologically similar and strongly developed rachis and leaflet abscission zones, may well constitute a unique property of Berberidaceae as a whole (cf. below for Nandina).
Given the relatively stiff nature of even green leaflets, it is not uncommon for mechanically disturbed leaves to lose leaflets at any time.
B. trifoliolata has equally well defined abscission joints, although the trifoliate terminal rachis node has a subannular ridge with the abscission zone divided into three equal triangular sections rising toward the center. Contrast the B. swaseyi terminal nodes with those of B. trifoliolata in the following set of images.
- 
Other Berberis species (e.g. B. repens, B. bealei) exhibit the same swollen, annular articulation of leaflet and rachis.
A. D. Bell (Plant Form 1991) uses B. bealei (= M. japonica) to illustrate a general presentation of leaf articulation (p. 49). I have not found any discussion of this feature in any treatments of Berberis/Mahonia.
It may well be that this is a common feature of the Berberidaceae as a family, since I have noted it with Nandina domestica as well - especially if Berberidaceae is seen as consisting of two subfamilies, Berberidoideae and Nandinoideae. Although I have been able to observe the manner of abscission for only a few species, disarticulation of the tripinnately compound leaves of N. domestica follows the same pattern as Berberis.
Finally, by way of contrast, I note that Berberis leaves, especially the segmentation of the rachis, seem to be constructed quite differently from the more typical compound leaves, which are sometimes said to separate as a whole (at the base of the leaf/petiole). Cf. e.g., web resources which inform without qualification that:
Examination of compound leaves from a number of species of Fabaceae (pea family) and Anacardiaceae (sumac family) indicate (1) that although the rachis/petiole falls as unit, individual leaflets often detach independently and (2) that rachis segments and leaflets do not share attachment morphology.
