Which Of The Following Are True About Outgroups

You're staring at a phylogenetic tree on your screen, and something feels off. The branching pattern makes sense, the bootstrap values look decent, but the root — the root is weird. Maybe it's stuck in the middle of your ingroup. Maybe the whole topology flips when you swap one taxon for another. Nine times out of ten, the problem isn't your alignment or your model. It's your outgroup.

What Is an Outgroup

An outgroup is a taxon — or a set of taxa — that you deliberately include in a phylogenetic analysis because it sits outside* the group you actually care about. That group you care about? Day to day, root the tree. The outgroup's job is simple on paper: give you a reference point so you can figure out which character states are ancestral and which are derived. That's your ingroup. Polarize the changes.

In practice, it's one of the most consequential decisions you'll make in a systematic study. Get it right, and your tree tells a coherent evolutionary story. Get it wrong, and you might spend months chasing artifacts that don't exist.

The logic behind it

Think of it like this. In real terms, you're trying to understand the family relationships among a group of cousins at a reunion. You could just look at who looks like whom. But if you bring in a great-aunt — someone you know* is a generation older and not part of the cousin cluster — suddenly you have a baseline. Still, traits shared by the cousins but not the great-aunt? Probably new. Traits shared by everyone including the great-aunt? Because of that, probably old. That great-aunt is your outgroup.

In phylogenetics, we formalize this. Which means the outgroup branches off before* the most recent common ancestor of the ingroup. That position — external to the ingroup clade — is what lets it serve as an evolutionary reference.

Monophyly matters

Here's where people trip up. On top of that, your ingroup should be monophyletic — a complete clade, ancestor plus all descendants. Your outgroup doesn't have* to be monophyletic (though it often is), but it must* fall outside that ingroup clade. Practically speaking, if your outgroup nests inside* the ingroup, you've failed. You've just added another ingroup taxon and lost your root.

Why Outgroups Matter in Phylogenetics

You can run a phylogenetic analysis without an outgroup. Consider this: you can't map trait evolution. But an unrooted tree is just a network of relationships — it tells you who's related to whom*, not which direction evolution flowed*. Most software will happily give you an unrooted tree. You don't know which node represents the common ancestor. You can't say "this trait evolved once" versus "this trait evolved three times independently.

Rooting changes everything.

Character polarization

This is the big one. When you have a rooted tree, you can ask: for any given character — a nucleotide position, a morphological trait, a gene presence/absence — what was the ancestral state? Because of that, the outgroup provides the answer. Day to day, if the outgroup has state A and some ingroup taxa have state B, the transition A→B happened within* the ingroup. Which means state B is derived. State A is ancestral.

Without that reference, you're guessing. And in systematics, guessing is expensive.

Detecting long-branch attraction

Long-branch attraction — LBA — is the phantom menace of phylogenetics. A well-chosen outgroup can break up long branches. Fast-evolving lineages artifactually cluster together because they've both accumulated lots of changes, not because they're genuinely close relatives. A poorly chosen one can create* them.

If your outgroup is very distant — say, using a bacterium to root a mammal tree — the branch leading to that outgroup is enormous. This isn't theoretical. It can pull fast-evolving ingroup taxa toward the root, distorting the whole topology. It's a documented, recurring headache in molecular systematics.

Model testing and clock calibration

If you're doing molecular dating or testing clock models, the outgroup isn't just a rooting tool. Consider this: it's part of your calibration strategy. Fossil calibrations often sit on the stem lineage between* outgroup and ingroup. The outgroup's branch length, its substitution rate, its saturation level — all of these feed into your divergence time estimates. A bad outgroup choice propagates error through the entire timescale.

It looks simple on paper, but it's easy to get wrong.

How to Choose an Outgroup

This is where the art lives. There's no universal formula, but there are principles that separate solid choices from regrettable ones.

Phylogenetic distance: the Goldilocks zone

You want an outgroup that's close enough to align reliably, but distant enough to be unambiguously outside the ingroup. Too close, and you risk the outgroup actually nesting within* the ingroup due to incomplete lineage sorting, hybridization, or just insufficient signal. Too distant, and you get alignment ambiguity, saturation, and LBA.

The sweet spot? That's why if you're studying birds, use crocodilians. Usually the sister group to your ingroup — the clade that shares the most recent common ancestor with your ingroup exclusive* of everything else. If you're studying rodents, use lagomorphs (rabbits, pikas). Sister groups minimize branch length while maximizing topological certainty.

When the sister group is unavailable

Sometimes the sister group is extinct. Sometimes it's unsampled. Sometimes you're working on a group where relationships at that depth are the question* — you can't use the answer to root the analysis that's trying to find the answer.

In those cases, you go one node deeper. So keep stepping back until you hit a taxon you can confidently place and reliably sequence. Use the sister group of the (ingroup + sister group) clade. Just know that each step back increases branch length and alignment difficulty.

Multiple outgroups beat single outgroups

One outgroup is a gamble. Practically speaking, two or three — especially if they form a small clade themselves — give you internal validation. On the flip side, if your multiple outgroups form a monophyletic group that sits cleanly outside the ingroup, you've got confidence. Still, if they don't — if one outgroup jumps into the ingroup, or they scatter across the tree — something's wrong. But could be contamination. Could be misidentification. Could be genuine phylogenetic conflict. But you'd never know with a single outgroup.

If you found this helpful, you might also enjoy how many days in 3 years or how many hours is 4 days.

Practical constraints

Let's be honest. Sometimes you choose an outgroup because it's the only one with a genome assembly. And or the only one you could get tissue for. Plus, or the only one your lab has primers for. Real-world systematics is messy. And the perfect outgroup doesn't exist for every study. But you should know* when you're compromising, and you should test the sensitivity of your results to that compromise.

Common Mistakes with Outgroups

I've seen these derail publications. You've probably seen them too.

Using a paralog by accident

This one stings. This leads to you're doing a gene tree. You pick an outgroup sequence from GenBank. It aligns beautifully.

actually a paralog — a gene duplicate that diverged before* the speciation event you're trying to root. On top of that, your tree now roots on a duplication node, not a speciation node. The ingroup topology might look fine, but the root is wrong, and every downstream inference — ancestral state reconstruction, divergence dating, trait evolution — inherits that error.

Always verify orthology. Reciprocal best BLAST hits help. Day to day, synteny helps. Plus, gene tree–species tree reconciliation helps. But the only real safeguard is knowing your gene family history before* you pick the outgroup sequence.

Rooting on a long branch because it's convenient

That basal lineage with the fast evolutionary rate? Consider this: it's tempting. And you think you've resolved the basal polytomy. But if it's the only thing connecting your ingroup to the rest of the tree, you've built a LBA magnet. Fast-evolving outgroups attract other long branches — often the deepest splits within your ingroup — pulling them artifactually toward the root. Consider this: the one that's "easy to amplify" or "already in the dataset"? You've actually manufactured one.

Check branch lengths. But run site-stripping analyses. Use models that account for heterotachy. If your root placement collapses when you remove the fastest-evolving sites or the longest-branched outgroup, you never had a root — you had an artifact.

Treating the outgroup as an afterthought

You spent months designing ingroup sampling. In real terms, you ran analyses for weeks. And you sequenced 200 taxa. The outgroup? You grabbed three species from a 2015 paper, downloaded their COI sequences, and called it a day.

The outgroup is the analysis. Even so, " It anchors every polarity decision. That said, it determines where the tree starts. Day to day, it defines "basal. If your outgroup sampling is thinner than your ingroup sampling, if you haven't tested alternative outgroups, if you haven't checked for rogue taxa among them — you haven't finished the study.

Ignoring outgroup monophyly

You included four outgroup taxa. One nests inside the ingroup. Now, they don't form a clade. In real terms, two form a clade with each other but exclude the third. The fourth floats somewhere near the root with 42% bootstrap.

You have two problems. So first, at least one of those taxa is misidentified, contaminated, or misplaced. Second, you don't have a root — you have a hypothesis about rooting that your own data reject. That's why don't force the tree to be rooted. Figure out why the outgroups disagree.

Testing Your Root

You've picked your outgroup(s). The tree is rooted. Now what?

Outgroup jackknifing. Re-run the analysis dropping one outgroup at a time. Does the ingroup topology hold? Does the root position shift? If removing a single taxon changes the basal split of your ingroup, your root is unstable.

Alternative outgroup testing. Pick a different outgroup clade — one node deeper, or a different sister lineage if phylogeny permits. Re-root. Compare. Congruence across outgroup choices is the strongest evidence you've got the root right.

Unrooted network approaches. Neighbor-nets, split graphs, consensus networks — these show conflict without* forcing a root. If the unrooted signal shows a clear "center" that matches your rooted tree, good. If the network is a starburst or shows competing splits near the base, your root is speculative.

Model-based rooting. Non-reversible substitution models (e.g., GTR+FO+R in IQ-TREE, or the non-reversible models in PhyloBayes) can estimate the root without* an outgroup, using strand asymmetry or compositional heterogeneity. They're computationally heavy and assumption-laden, but they provide an independent check.

The Root Is a Hypothesis

Not a fact. So not a default. A hypothesis — tested by taxon sampling, by model choice, by alternative outgroups, by the congruence of multiple lines of evidence.

Every rooted tree in the literature represents a bet: this* is where the history of this group begins. The root of eukaryotes. Which means the ctenophore-sister vs. The position of acoels in Metazoa. Sometimes a new genome, a new fossil, a new method flips the root and rewrites the narrative. sponge-sister debate. Sometimes the bet pays off. These aren't settled by better outgroups alone — they're settled by critical* outgroup use.

So choose deliberately. Report honestly. Think about it: test aggressively. And never, ever treat the outgroup as the easy part.