The big question is: what are they?
The LRT nests multituberculates with rodents, but currently that’s a minority view of one.
Kielan-Jaworowska Z and Hurum 2001 wrote:
“Traditionally palaeontologists believed that multituberculates might have originated from cynodonts independently from all other mammals, or diverged from other mammals at a very early stage of mammalian evolution.” Unfortunately these authors do not say which taxa attract multis to the the pre-eutherian grades and clades.
Simpson (1945, p. 168) stated:
“The multituberculate structure was so radically distinctive throughout their history that it seems hardly possible that they are related to other mammals except by a common origin at, or even before, the class as such”
Hahn et al. (1989) and Miao (1993)
reported that multituberculates might be a sister taxon of all other mammals. On the other hand, Kielan-Jaworowska et al. 1986; Miao 1988; Wible 1991; Rougier et al. 1992; Wible and Hopson 1993, 1995; Hurum 1994, 1998a, b) demonstrated the homogeneity of the internal structure of the skull and vascular system of all mammals, including multituberculates.
Hurum et al. 1996; Rougier et al. 1996a report
Multituberculate ear ossicles display the same pattern as those of all other mammals.
that multituberculates might form a sister taxon of all other mammals is related to the idea that they are close relatives to the Haramiyidae, a family represented until recently only by isolated teeth, with numerous cusps arranged in longitudinal rows, known from the Late Triassic and Early Jurassic mostly in Europe. Key to these thoughts are the idea that most fossil material comes from teeth.
Jenkins et al. (1997)
described from the Upper Triassic of Greenland Haramiyavia clemmenseni, assigned to the Haramiyidae, represented by dentaries and partial maxillae with teeth and fragments of the postcranial skeleton. Haramiyavia has been interpreted as having orthal jaw movement (standard up-down rotation on a glenoid axis). On this basis Jenkins et al. excluded the Haramiyida from the Allotheria, which have propalinal (fore-and-aft) movement of the dentary and backward (palinal) power stroke. In turn Butler (2000) revised all known allotherians and argued that dental resemblance supports the hypothesis that the Multituberculata originated from the Haramiyida.
Kielan-Jaworowska Z and Hurum 2001 wrote:
“Finally, the most recent analyses of mammalian relationships, including analysis of the skeleton of a symmetrodont Zhangheotherium (Hu et al. 1997; here recovered as a pangolin ancestor), and the skeleton of the eutriconodont Jeholodens (Ji et al. 1999; here recovered as a tritylodontid), did not support multituberculate-therian sister-group relationship. In both of these papers the Multituberculata were placed between Monotremata (Ornithorhynchus) and Symmetrodonta (Zhangheotherium), being a sister taxon of all the Holotheria” (last common ancestor of Kuehneotherium and Theria). In other words, close to monotremes.
Kielan-Jaworowska Z and Hurum 2001 wrote about the multi brain:
“The multituberculate brain, designated cryptomesencephalic (characterised by an expanded vermis, no cerebellar hemispheres, and lack of the dorsal midbrain exposure) is very different from that in Theria, which originally had eumesencephalic brains (characterised by a wide cerebellum with extensive cerebellar hemispheres and large dorsal midbrain exposure).”
This appears to assume only one direction for brain development, with no evolutionary backsliding. Unfortunately Kielan-Jaworowska and Hurum employed a hypothetical ancestor for their multituberculate cladogram.
We’ve already seen teeth in whales reverse from the typical W and Y molar cusp patterns, to linear molar cusps to simple pegs.
Kielan-Jaworowska Z and Hurum 2001 wrote about the multituberculate foot:
“Another character neglected until recently in phylogenetic analyses of early mammals involves the foot structure. In the multituberculate foot the middle metatarsal (M3) is abducted from the longitudinal axis of the tuber calcanei, while the calcaneus contacts distally the 5th metatarsal (Kielan-Jaworowska and Gambaryan 1994). This type of foot appeared at that time to be unique among mammals, but Ji et al. (1999) described a similar type of foot in the eutriconodont Jeholodens. It follows that there are two groups of characters related to brain and foot structure, which ally multituberculates with eutriconodonts.”
The trouble is
a large gamut analysis of mammalian relationships does not find a better nesting for those highly-derived, but primitive-brained, rodent-like mutituberculates than with rodents. They have similar teeth, similar extremities, similar skulls. And that twisted heel-bone (calcaneum) is a derived trait. So why are multis supposed to nest with mammals earlier than placentals?
If anyone can produce a pre-therian that attracts multis, please bring it to my attention. So far, I have failed to find out, and so multis continue to nest with rodents and plesiadapids.
Kielan-Jaworowska Z and Hurum JH 2001. Phylogeny and Systematics of multituberculate mammals. Paleontology 44, 389–429.