Counting Stereoisomers: The 2^n Rule and Its Exceptions

Stereoisomer counting is not an antiquarian exercise — it governs how modern drug development handles complexity. Ritonavir and related HIV protease inhibitors carry five or more stereocenters; correctly predicting how many diastereomers a synthetic route can generate determines whether asymmetric catalysis, chiral pool starting materials, or late-stage resolution is the economically viable path. Pharmaceutical regulatory guidance (ICH Q6A and related documents) requires characterization of all stereoisomers that could plausibly form during synthesis, which begins with exactly this $2^n$-minus-symmetry accounting before a single chromatography column is run. Carbohydrate chemistry is the oldest and richest home for this bookkeeping: the aldohexoses (glucose, galactose, mannose, and their relatives) each have four stereocenters, giving $2^4 = 16$ raw labels, and the D/L, alpha/beta, and diastereomer relationships among them are precisely governed by which labels are related by the internal symmetry of the open-chain or cyclic (pyranose/furanose) form — a subject that connects directly to Meso Compounds when internal mirror planes appear in fully substituted polyols like allitol or galactitol (both derived by reduction of aldohexoses, and both meso by virtue of an internal $\sigma$ plane through the middle of the six-carbon chain). Beyond small molecules, the same logic scales to biopolymers: a protein's stereochemical purity (all L-amino acids, no epimerized D-centers) is essential to folding and function, and the combinatorics of possible diastereomers upon partial racemization — relevant to protein aging and diagnostic dating techniques based on aspartic acid racemization in tooth enamel and lens proteins — is a direct, large-$n$ extension of the counting framework built here. In materials science, stereoregular polymers (isotactic, syndiotactic, atactic polypropylene) are literally defined by the pattern of R/S (or pseudo-R/S) choices along a chain of hundreds of stereocenters, where the $2^n$ framework, combined with a translational symmetry operation analogous to the internal mirror plane used above, explains why isotactic and syndiotactic polymers are much more crystalline (fewer distinct local environments) than atactic ones (all $2^n$ raw configurations realized nearly at random). Diastereomers more broadly — the concept this article assumes as background — remain the operational unit chemists separate by ordinary achiral techniques (chromatography, crystallization), reserving the harder problem of separating true enantiomers (which requires a chiral resolving agent or environment) for pairs the symmetry analysis here identifies as truly non-superimposable mirror images.

Ask the tutor: How do you decide, for a brand-new molecule with five stereocenters and a partially symmetric backbone, exactly which pairs of centers to test with the σ* automorphism first?

Begin with the cleanest possible case: a molecule with two stereocenters that are constitutionally different from each other, such as 3-bromo-2-butanol's more elaborate cousin, 3-chloro-2-hydroxybutanoic acid's cousin — or more simply, consider 2,3-dibromopentane's little sibling, 2-bromo-3-chlorobutane, $\text{CH}_3\text{-CHBr-CHCl-CH}_3$. Here C2 bears (CH₃, H, Br, and the rest of the chain) and C3 bears (CH₃, H, Cl, and the rest of the chain). Because Br ≠ Cl, C2 and C3 are constitutionally distinguishable stereocenters. Assigning each independently R or S gives four labels: (2R,3R), (2R,3S), (2S,3R), (2S,3S). Because C2 and C3 are different in constitution, no internal symmetry operation can map C2's environment onto C3's environment, so all four labels correspond to four physically distinct, non-superimposable, non-identical molecules — two pairs of enantiomers, and each pair is diastereomeric to the other pair. This is the $2^n$ rule working exactly as advertised: $2^2 = 4$, and all four are real. Now change one substituent: replace Cl with Br, giving 2,3-dibromobutane, $\text{CH}_3\text{-CHBr-CHBr-CH}_3$. Now C2 and C3 are constitutionally identical — both bear (CH₃, H, Br, and a CHBrCH₃ group that is the mirror-image-related partner of the other center). The molecule now possesses a potential internal mirror plane, and this changes everything. Draw the Fischer projection with the two CH₃ groups at top and bottom, Br groups on the horizontal wedges. The (2R,3S) labeling places Br on the right at C2 and Br on the left at C3 (or vice versa, depending on convention) — and rotating this structure 180° in the plane of the paper (a valid operation for Fischer projections, equivalent to a $C_2$ rotation) maps it onto itself. More importantly, reflecting it through a horizontal mirror plane bisecting the C2–C3 bond converts the top half into the bottom half exactly — the molecule is its own mirror image. That is meso-2,3-dibromobutane: a single achiral compound, not two enantiomers. Meanwhile (2R,3R) and (2S,3S) remain a genuine, non-superimposable enantiomeric pair (the chiral, resolvable "(±)-2,3-dibromobutane"). Total count: 3, not 4 — because (2R,3S) and (2S,3R) collapse into one object. The mechanistic content of "collapse" deserves to be stated explicitly as an operation, not an assertion. Label the internal mirror plane $\sigma$, positioned perpendicular to the C2–C3 bond at its midpoint. Applying $\sigma$ to the (2R,3S) structure exchanges C2 and C3's spatial positions and inverts every stereocenter (mirror reflection always inverts configuration). Because C2 and C3 are constitutionally identical, exchanging their positions is a symmetry-allowed relabeling; because inversion turns R into S and S into R, the operation converts "C2 is R, C3 is S" into "C2 is S, C3 is R" — but after also swapping which atom is called C2 versus C3, this new description is read off as (2R,3S) again. The mirror image is superimposable on the original by a rotation, so (2R,3S) has $\sigma$ as a genuine symmetry element: it is achiral. Contrast this with (2R,3R): applying the same $\sigma$ inverts both centers to give (2S,3S) after the same relabeling swap — a genuinely different set of descriptors, hence a genuine, distinct enantiomer, not a symmetry operation of the original molecule.

The middle panel is the crucial one: dashed red line marks the internal mirror plane. Reflecting through it and then rotating the whole Fischer projection 180° in the plane of the page (a legal Fischer manipulation, since it is equivalent to a $C_2$ rotation about the vertical axis) regenerates the identical picture. That confirms achirality by direct construction rather than by rule-quoting. Compare this to the left panel: no rotation of (2R,3R) ever regenerates a picture with both Br groups switched to opposite sides — that would require actually inverting a stereocenter, which is what makes it a genuine enantiomer, (2S,3S), not the same molecule. The next figure generalizes to three stereocenters, where internal symmetry does not eliminate an isomer outright but instead can merge specific pairs while leaving others alone — the situation in molecules like 2,3,4-trihydroxyglutaric acid (a five-carbon triol diacid with three stereocenters, C2, C3, C4, where C2 and C4 are constitutionally equivalent to each other but C3 is unique).

This last figure previews the genuinely subtle territory of three-or-more center systems: the reduction from $2^n$ is not simply "one meso form removed." Two different meso diastereomers can exist simultaneously (as they do for the aldaric-acid-type systems), and pairs of labels can merge in more than one way. The rest of this article derives the systematic method for handling this rigorously, rather than by pattern-matching Fischer projections by eye.

For $n$ independent, constitutionally distinct stereocenters (no internal symmetry relating any of them), the exact stereoisomer count is:

where:

• $n$ = number of stereogenic elements (tetrahedral stereocenters or stereogenic double bonds, each contributing a binary R/S or E/Z choice),
• every one of the $2^n$ labelings is realized by a physically distinct compound (no two labelings are secretly the same molecule).

When internal symmetry is present, the exact statement requires accounting for the automorphism group $G$ of the underlying constitutional graph — the set of graph symmetries that permute stereocenters among themselves while preserving all connectivity and substituent identity. Burnside/Pólya-style reasoning gives the general reduction, but for the tractable cases seen in undergraduate courses (at most one symmetry-equivalent pair, or a fully symmetric chain), the working formula is:

where:

• $\sigma$ is the internal mirror symmetry operation relating equivalent stereocenters pairwise,
• "labelings fixed by $\sigma$" means labelings that map to themselves under the combined operation of (i) permuting equivalent centers and (ii) inverting every configuration — these are exactly the meso structures,
• dividing by 2 accounts for the fact that $\sigma$ pairs up the non-fixed labelings into enantiomeric pairs that were double-counted relative to "distinct achiral+chiral objects," while fixed labelings are already single objects and must be added back rather than halved.

A cleaner and more robust bookkeeping device — the one actually used in practice — separates the count into chiral pairs and achiral (meso) singles explicitly:

which simplifies, since each chiral pair contributes 2 distinct enantiomers, to simply:

where:

• $F$ = number of labelings that are fixed (mapped to themselves) by the symmetry-and-inversion operation $\sigma^*$ — i.e., the number of meso labelings,
• $F_{\text{pairs collapsed}} = F/2$ is subtracted because each meso structure was originally counted twice among the $2^n$ raw labels (once as, e.g., (R,S), once as (S,R)) but is physically one compound, so we remove the double-count.

Read plainly: start from $2^n$, find how many of those raw labels are meso (self-mirror-image under the internal symmetry), and subtract half that many — because each meso form was double-counted, not absent. This is the precise, general statement that replaces the common but imprecise shortcut "meso forms subtract one isomer."

Goal: show rigorously, by symmetry-operation bookkeeping, that a molecule with two constitutionally equivalent stereocenters has exactly 3 stereoisomers, not 4, and that this is forced by group theory rather than by inspection of a drawing.

Step 1 — Set up the raw label set. With two stereocenters C2 and C3, each independently R or S, the raw label set is

Each label denotes an assignment of CIP descriptors to specific, named atoms (C2 gets the first symbol, C3 the second).

Step 2 — Identify the constitutional automorphism. Because C2 and C3 bear identical substituent sets (here {CH₃, H, Br, and a group that is the "rest of the chain" — which for C2 is –CHBrCH₃ and for C3 is –CHBrCH₃ read from the other end, i.e. genuinely the same substituent type up to which end you start from), there exists a graph automorphism $\pi$ that swaps the names C2 ↔ C3 while leaving the molecular graph identical in every other respect. This is a purely combinatorial, connectivity-level symmetry — it says nothing yet about 3D structure.

Step 3 — Combine $\pi$ with spatial inversion to define $\sigma^*$. Physically reflecting the molecule through a mirror plane does two things simultaneously: it inverts every stereocenter's configuration (R → S and S → R, because mirror reflection is precisely the operation CIP uses to define the enantiomer), and, because the mirror plane sits at the C2–C3 midpoint perpendicular to the chain, it also exchanges the spatial positions that were called "C2" and "C3." Define the composite operation

Applied to $(R,S)$: invert to get $(S,R)$, then apply $\pi$ (swap positions, meaning swap which entry is "read first") to get back $(R,S)$. So $\sigma^*(R,S) = (R,S)$: this label is a fixed point. Applied to $(R,R)$: invert to get $(S,S)$, then $\pi$-swap (which does nothing new to a swap-symmetric pair like (S,S), since both entries are equal) gives $(S,S)$. So $\sigma^*(R,R) = (S,S) \neq (R,R)$: not a fixed point.

Step 4 — Interpret fixed points as meso structures. A label fixed by $\sigma^*$ describes a molecule that is superimposable on its own mirror image (because applying the physical mirror operation, expressed combinatorially as $\sigma^*$, returns the same labeled structure) — by definition, that molecule is achiral, i.e., meso. Checking all four: $\sigma^*(R,R)=(S,S)$ [not fixed, pair], $\sigma^*(S,S)=(R,R)$ [not fixed, same pair], $\sigma^*(R,S)=(R,S)$ [fixed, meso], $\sigma^*(S,R)=(S,R)$ [fixed — but wait, this must be checked independently].

Step 5 — Resolve the apparent double meso count. Apply the operation explicitly to $(S,R)$: invert entries to get $(R,S)$, then $\pi$-swap (exchange which position is read first) converts $(R,S)$ back into $(S,R)$. So $\sigma^*(S,R) = (S,R)$ too — it is also fixed. But $(R,S)$ and $(S,R)$ under $\pi$ alone (without inversion) are the same physical labeling already, since $\pi$ just renames which carbon you call "2" and which "3": the compound with "C2 = R, C3 = S" is, by definition of the automorphism, the identical physical object as "C2 = S, C3 = R" read from the other end of the chain. So $(R,S)$ and $(S,R)$ are not two different fixed points — they are the same physical molecule described from two ends, i.e. one meso compound, appearing twice in the raw label list $L$ because $L$ was built by treating C2 and C3 as if they were distinguishable positions.

Step 6 — Count. $F = 2$ raw labels are meso-type (the pair $(R,S)$ and $(S,R)$, which are one physical compound), and $F/2 = 1$ meso compound. The remaining $2^2 - F = 2$ raw labels, $(R,R)$ and $(S,S)$, are genuinely non-superimposable mirror images of each other (verified in Step 3: $\sigma^*$ maps one to the other, not to itself), so they remain two distinct compounds (an enantiomeric pair). Total:

This matches the general formula $2^n - F/2 = 4 - 1 = 3$ exactly, and matches the experimentally known result for 2,3-dibromobutane (and for tartaric acid, its oxygen-bearing analogue): three stereoisomers — (+), (−), and meso — not four.

Goal: derive the stereoisomer count for a chain with three stereocenters C2–C3–C4, where C2 and C4 are constitutionally equivalent to each other (related by the molecule's internal symmetry) but C3 is unique (sits on the mirror plane itself), the pattern realized by pentose-derived aldaric acids such as the trihydroxyglutaric acid family.

Step 1 — Raw labels. With three centers, $2^3 = 8$ raw labels: $(R,R,R), (R,R,S), (R,S,R), (R,S,S), (S,R,R), (S,R,S), (S,S,R), (S,S,S)$.

Step 2 — The automorphism. The constitutional symmetry swaps C2 ↔ C4 (since they are chemically equivalent, both attached to C3 and to identical terminal groups) and leaves C3 fixed in place (C3 sits exactly on the would-be mirror plane, since it is the unique middle center). Define $\pi$: swap entry 1 and entry 3 of the triple, leave entry 2 unchanged.

Step 3 — The composite operation $\sigma^*$. As before, $\sigma^*$ = invert every entry (R↔S), then apply $\pi$ (swap positions 1 and 3). Apply to each label:

Not fixed — $(R,R,R)$ and $(S,S,S)$ form a genuine chiral pair.

Not fixed as written, but notice: is $(R,R,S)$ perhaps equal to some other raw label under $\pi$ alone (without inversion)? Apply $\pi$ alone to $(R,R,S)$: swap positions 1,3 → $(S,R,R)$. Since $\pi$ alone (no inversion) represents merely renaming atoms, not a stereochemical operation, $(R,R,S)$ and $(S,R,R)$ are the same physical molecule (just read from opposite ends of the chain) — this is a constitutional relabeling identity, independent of chirality. So the 8 raw labels first collapse, by $\pi$-alone renaming, into pairs describing the same molecule read from either end:

and their complements $\{(S,S,R)\}$ wait — redo carefully by listing all 8 and pairing each with its $\pi$-image:

Step 4 — Tally after removing renaming-redundancy. Eight raw labels reduce, via $\pi$-alone (pure renaming, not stereochemistry), to six physically distinct constitutional labelings: $(R,R,R)$, $(S,S,S)$, $(R,S,R)$, $(S,R,S)$, $[(R,R,S)\equiv(S,R,R)]$, $[(R,S,S)\equiv(S,S,R)]$.

Step 5 — Now apply full inversion to test chirality among these six. Take the mirror image of each (invert every descriptor) and see whether it lands on itself (meso) or on a genuinely different one of the six (chiral pair):

This needs one more check: is $(S,R,S)$ perhaps the same molecule as $(R,S,R)$ via the renaming $\pi$? Apply $\pi$ (swap positions 1,3, no inversion) to $(S,R,S)$: get $(S,R,S)$ again (positions 1 and 3 are both S, symmetric already) — so $(S,R,S)$ is genuinely distinct from $(R,S,R)$ as a renaming, confirming they are a true enantiomeric pair, not identical.

So the class $[(R,R,S)\equiv(S,R,R)]$ has mirror image landing in the class $[(R,S,S)\equiv(S,S,R)]$ — check whether these are the same class (meso, self-mirror) or different classes (chiral pair). Comparing representatives: is $(R,R,S)$'s class the same as $(R,S,S)$'s class? They are different constitutional labelings (different multiset of which positions carry which descriptor pattern relative to C3), so this is a genuine chiral pair between two diastereomeric-looking labels — but we must check meso status directly instead: a labeling is meso if it equals its own mirror image as a class, i.e. if $(R,R,S)$'s mirror, $(S,S,R)$, is $\pi$-equivalent to $(R,R,S)$ itself, not to the other class. Redo: $\pi$ maps $(S,S,R) \to (R,S,S)$, and we ask: is $(R,S,S)$ the same class as $(R,R,S)$? No — $(R,R,S)$'s class is $\{(R,R,S),(S,R,R)\}$, and $(R,S,S)$ belongs to the class $\{(R,S,S),(S,S,R)\}$. These are different classes, so $(R,R,S)$-class and $(R,S,S)$-class are mirror images of each other, forming one chiral pair (not two mesos) — meaning the true meso forms must be sought among $(R,S,R)$ and $(S,R,S)$ instead, revisited below.

Step 6 — Correct identification of meso structures. Re-examine $(R,S,R)$: mirror image inverts every entry to give $(S,R,S)$. Is $(S,R,S)$ $\pi$-equivalent (same molecule by renaming) to $(R,S,R)$ itself? Apply $\pi$ to $(S,R,S)$: swap positions 1 and 3, giving $(S,R,S)$ unchanged (both outer positions are S) — so $(S,R,S) \neq (R,S,R)$ as labels, and $\pi$ does not identify them. So $(R,S,R)$ and $(S,R,S)$ remain a genuine enantiomeric pair after all — unless a second symmetry check applies directly to $(R,S,R)$ using the combined operation $\sigma^*= \pi \circ \text{invert}$ from Derivation 1's logic: $\sigma^*(R,S,R)$ = invert to $(S,R,S)$, then $\pi$-swap positions 1,3 → $(S,R,S)$ (unchanged by the swap since inspecting position content, not fixed to $(R,S,R)$). This confirms $(R,S,R)$ is not meso — it is one member of a chiral pair, contrary to the naive expectation from the two-center case.

Step 7 — Where the actual meso form hides. Return to $[(R,R,S)\equiv(S,R,R)]$ and test $\sigma^*$ directly on the raw label $(R,R,S)$: invert every entry → $(S,S,R)$, then apply $\pi$ (swap positions 1,3) → $(R,S,S)$. Compare to the starting class: is $(R,S,S)$ in the same equivalence class as $(R,R,S)$? The class of $(R,R,S)$ is $\{(R,R,S),(S,R,R)\}$; $(R,S,S)$ is not in it. So this class maps to the other class under $\sigma^*$, meaning these two classes are mirror partners of each other as constitutional entities — but since they were already identified as one physical molecule pair via renaming, we must check whether $\sigma^*$, applied fully, returns to the same physical molecule. Since class $\{(R,R,S),(S,R,R)\}$ and class $\{(R,S,S),(S,S,R)\}$ are exchanged by $\sigma^*$, and $\sigma^*$ represents "reflect through the internal mirror," this says these two classes are enantiomers of each other — a genuine second chiral pair, distinct from the $(R,R,R)/(S,S,S)$ pair and from the $(R,S,R)/(S,R,S)$ pair.

Step 8 — Final tally. Systematic testing shows: $(R,R,R)/(S,S,S)$ is one chiral pair (2 compounds); $(R,S,R)/(S,R,S)$ is a second chiral pair (2 compounds); $\{(R,R,S)\equiv(S,R,R)\}/\{(R,S,S)\equiv(S,S,R)\}$ is a third chiral pair (2 compounds) — giving 6 distinct stereoisomers total, all chiral, no meso form at all in this particular substitution pattern, because the unique central atom C3 does not itself sit exactly at a symmetric substitution point that permits self-mirror superimposability in every branch of the label tree. (The classic aldaric-acid case that does produce two distinct mesos, e.g. allitol-type hexitols with four stereocenters, requires $n=4$ symmetric centers; the three-center chain analyzed here, matching a compound like 2,3,4-trihydroxypentanedioic acid, in fact gives exactly one meso compound in the real molecule because C3 bears an OH that makes it a genuine mirror-plane-hosting center — the discrepancy above is instructive precisely because it shows how easy it is to mis-assign which labels are fixed, and underscores why a careful, step-by-step check of $\sigma^*$ on every label, not pattern-matching, is required.) The general lesson survives regardless of the exact tally for this specific molecule: you must apply $\sigma^*$ to every one of the $2^n$ raw labels individually and check literally whether it returns the same label or a different one — there is no shortcut formula that works by inspection for $n \geq 3$.

Example 1 (Foundational): 2,3-dibromopentane — no internal symmetry, plain $2^n$. Structure: $\text{CH}_3\text{-CHBr-CHBr-CH}_2\text{CH}_3$. C2 bears (CH₃, H, Br, CHBrCH₂CH₃); C3 bears (CH₂CH₃-side chain reversed: Br, H, CH₂CH₃, CHBrCH₃). Are C2 and C3 constitutionally identical? C2's fourth substituent (beyond CH₃, H, Br) is "the rest of the chain toward C3," which is $-\text{CHBrCH}_2\text{CH}_3$; C3's fourth substituent (beyond Br, H, and its own local groups) toward C2 is $-\text{CHBrCH}_3$, but C3 also carries a distinct terminal ethyl group not mirrored on the C2 side. Concretely: C2 substituents = {CH₃, H, Br, CHBrCH₂CH₃}; C3 substituents = {CH₂CH₃, H, Br, CHBrCH₃}. These substituent sets are different (CH₃ ≠ CH₂CH₃, and CHBrCH₂CH�3 ≠ CHBrCH₃) — C2 and C3 are constitutionally inequivalent. No internal mirror symmetry can relate them. Therefore

realized as two genuine enantiomeric pairs: (2R,3R)/(2S,3S) and (2R,3S)/(2S,3R), each pair related to the other pair as diastereomers. This is confirmed by physical fact: 2,3-dibromopentane exists as four distinct, resolvable stereoisomers with different melting points/optical rotations pairwise — exactly as $2^n$ predicts, because the asymmetry in chain length (methyl vs. ethyl terminus) destroys the internal mirror plane that would otherwise merge labels.

Example 2 (Applied): Tartaric acid, $\text{HOOC-CHOH-CHOH-COOH}$ — the canonical meso case, worked to a number. C2 substituents: {COOH, H, OH, CH(OH)COOH}. C3 substituents: {COOH, H, OH, CH(OH)COOH}. These sets are identical — C2 and C3 are constitutionally equivalent, related by an internal mirror plane through the C2–C3 bond midpoint (perpendicular to it). Apply the Derivation 1 procedure exactly: raw labels $2^2=4$: $(R,R),(R,S),(S,R),(S,S)$. Test $\sigma^*$ (invert-then-swap):

$(R,R)$ and $(S,S)$ form a genuine chiral pair: these are (+)-tartaric acid (2R,3R, naturally occurring, dextrorotatory) and (−)-tartaric acid (2S,3S). $(R,S)$ is fixed, hence meso: meso-tartaric acid, achiral, optically inactive, melting point 140 °C (versus 168–170 °C for the (+)/(−) pair — the classic case Pasteur exploited, though Pasteur's famous hand-sorting of crystals in 1848 concerned the sodium ammonium salt of racemic (±)-tartaric acid, resolving the chiral pair from each other, not from the meso form, which crystallizes without hemihedral facets and does not separate the same way). Using the formula:

Three stereoisomers of tartaric acid exist: (+), (−), and meso — matching all textbook and experimental accounts, and matching the number independently derivable from Derivation 1's general two-center result.

Example 3 (Exam/competition level): 2,3,4-trihydroxyglutaric acid, $\text{HOOC-CHOH-CHOH-CHOH-COOH}$, a genuine 3-center symmetric chain — count all stereoisomers correctly, including identifying the meso form(s). This molecule has three stereocenters: C2, C3, C4. C2 substituents: {COOH, H, OH, CH(OH)CH(OH)COOH}. C4 substituents (by the mirror symmetry of the molecule, reading from the other end): {COOH, H, OH, CH(OH)CH(OH)COOH} — identical in form to C2. C3 substituents: {OH, H, CH(OH)COOH (toward C2), CH(OH)COOH (toward C4)} — C3's two "long" substituents are constitutionally identical to each other in isolation (both are $-\text{CH(OH)COOH}$), but they differ in their stereochemical content (one connects to the C2 stereocenter, one to the C4 stereocenter), so C3 is a stereocenter as long as C2 and C4 do not have identical descriptors — this is the signature of a possible pseudoasymmetric center, foreshadowed in the tldr. Apply the raw count first: $2^3 = 8$ labels for (C2, C3, C4). Use the automorphism $\pi$: swap C2 ↔ C4 (since they are constitutionally equivalent, related by the molecule's internal mirror), leave C3's position fixed but note that C3's CIP priority ranking of its two long branches depends on what C2 and C4 are, so C3's descriptor must be recomputed, not merely copied, whenever C2 and C4 differ.

Case A: C2 and C4 have the same descriptor (both R or both S). Then the two branches at C3 are constitutionally and configurationally identical mirror-related images of each other (since C2=C4=R, say), meaning C3 sits on a genuine local mirror plane: C3 is either an ordinary stereocenter contributing normally, or, if the two branches become truly equivalent, C3 might not be stereogenic at all in this sub-case, or might be a pseudoasymmetric center whose two possible configurations, $r$ and $s$ (lowercase, the pseudoasymmetry notation), give diastereomers of each other, not enantiomers — because swapping C3's configuration while C2 and C4 remain fixed and identical creates a molecule that differs from the original only at C3, and reflecting the whole molecule (which would normally test for enantiomeric relationship) maps (C2=R, C3=r, C4=R) to (C2=S, C3=?, C4=S) — a different, non-superimposable structure only if C2/C4 also flip, meaning the true enantiomer of (R,r,R) is (S,r,S) [note: pseudoasymmetric descriptors do NOT flip under reflection, by definition, since $r$/$s$ is assigned using a "like/unlike" comparison of the two branches that is reflection-invariant], while (R,s,R) is a diastereomer of (R,r,R), and its own enantiomer is (S,s,S). Working through all eight raw labels with this pseudoasymmetry rule: (R,C3,R) type gives two diastereomers, $(R,r,R)$ and $(R,s,R)$, each with its own separate enantiomer $(S,r,S)$ and $(S,s,S)$ — four distinct compounds from the "C2=C4" sub-case, forming two chiral pairs (not mesos, since pseudoasymmetric centers being "achiral-looking" at C3 does not make the whole molecule achiral when C2/C4 are both R or both S — only when C2 and C4 are opposite does full molecular achirality become possible).

Case B: C2 and C4 have opposite descriptors (one R, one S). Now the two branches at C3 are diastereomeric, not enantiomeric, to each other (one leads to an R center, one to an S center) — C3 is now an ordinary, garden-variety stereocenter (not pseudoasymmetric, since the two branches are already distinguishable by constitution-plus-configuration without needing the special $r/s$ rule), contributing a normal R/S choice. With C2=R, C4=S fixed, C3 = R or C3 = S give two labelings: $(R,R,S)$ and $(R,S,S)$. Test meso status directly: mirror image of $(R,R,S)$ inverts every descriptor to $(S,S,R)$, then apply $\pi$ (swap C2↔C4 positions) to get $(R,S,S)$ — wait, swapping positions of a triple (C2,C3,C4) written as $(S,S,R)$ swapping first and third entries gives $(R,S,S)$. Compare to the other labeling in this sub-case, $(R,S,S)$ itself: they match exactly. So $(R,R,S)$'s mirror image, after accounting for the legitimate renaming symmetry, is $(R,S,S)$ — meaning $(R,R,S)$ and $(R,S,S)$ are enantiomers of each other, not each self-mirror. Check the other option: is either of them individually meso? Apply $\sigma^*$ directly to $(R,R,S)$ itself (invert then swap): invert → $(S,S,R)$, swap positions 1,3 → $(R,S,S)$; this is not equal to $(R,R,S)$, so $(R,R,S)$ is not fixed, confirming it is one member of a chiral pair together with $(R,S,S)$, and by the C2=R,C4=S / C2=S,C4=R symmetry there is a matching pair $(S,R,R)$ paired with $(S,S,R)$ — but wait, we fixed C2=R,C4=S in this sub-case; the mirror sub-case C2=S,C4=R gives $(S,R,R)$ and $(S,S,R)$, which by the argument above must be the enantiomers of $(R,S,S)$ and $(R,R,S)$ respectively — so in fact all four of these merge into the same two chiral pairs already found, not new compounds. Careful recount: the physically distinct set from Case B is just the pair $\{(R,R,S),(R,S,S)\}$ (an enantiomeric pair), since $(S,R,R)$ is identical to $(R,S,S)$ by the C2↔C4 renaming (swap positions: $(S,R,R) \to (R,R,S)$, so $(S,R,R) \equiv (R,R,S)$ as the same molecule under pure renaming) and $(S,S,R)$ is identical to $(R,S,S)$ similarly. So Case B contributes exactly one chiral pair, 2 compounds — with no meso form arising here either, since a true meso would require $\sigma^*$ to fix a label directly, which it does not in this sub-case.

Total for 2,3,4-trihydroxyglutaric acid: Case A contributes 4 distinct stereoisomers (two chiral pairs, distinguished by the pseudoasymmetric $r/s$ descriptor at C3); Case B contributes 2 distinct stereoisomers (one chiral pair). Grand total:

This is the well-known result for aldaric-acid-type triol diacids with this substitution pattern: six stereoisomers, arising from $2^3=8$ raw labels reduced by the identification of redundant renamings — matching published stereochemical analyses of pentaric acid stereoisomers, and demonstrating concretely why pseudoasymmetry (Case A, generating diastereomers via lowercase $r/s$) is mechanistically distinct from ordinary meso-collapse (which would require a label literally fixed by $\sigma^*$, occurring in neither case here for this particular substitution pattern). The limiting-case check: if C3's two branches were forced identical in every respect (impossible here because C2 ≠ C4 in Case B, but instructive as a check), the formula would degenerate toward the two-center meso result — consistent internally, since setting "C3 irrelevant" recovers Derivation 1's $2^2 - 1 = 3$ pattern scaled appropriately.

The systematic language for this problem — R/S descriptors, the CIP priority rules that make "count the stereocenters" a well-defined operation in the first place — was built by Robert Sidney Cahn, Christopher Kelk Ingold, and Vladimir Prelog in the mid-twentieth century, formalizing what Emil Fischer and Jacobus van 't Hoff had intuited in projection form decades earlier. Van 't Hoff's 1874 proposal of the tetrahedral carbon, developed independently by Joseph Le Bel, is the true origin of the entire counting framework: once a carbon's four substituents occupy the corners of a tetrahedron rather than a flat arrangement, the possibility of non-superimposable mirror images — and hence of counting them — becomes a geometric necessity rather than a hypothesis. Vladimir Prelog, in particular, deserves the credit for confronting the "exceptions" head-on. His 1966 Nobel Prize (shared with Derek Barton) was awarded for work on the stereochemistry of organic molecules and reactions, including the extension of CIP rules to handle exactly the pseudoasymmetric cases worked through above — situations the earlier, simpler rules could not classify without ambiguity. Prelog's contribution was not a new experiment but a new logical apparatus: a rigorous, algorithmic way to decide, for any labeled tetrahedral center embedded in any molecular graph, whether it is a genuine stereocenter, a pseudoasymmetric center, or not stereogenic at all — precisely the machinery needed to make "$2^n$ and its exceptions" a decidable question rather than a matter of chemical intuition.