Ataxia is the inability to generate a smooth, accurately scaled voluntary movement — the trajectory overshoots, undershoots, or breaks into a series of corrective jerks — in the absence of weakness or extraneous involuntary movements. The deficit is one of coordination, not power. Mechanistically, the cerebellum acts as a feed-forward predictor that pre-computes the motor command needed to hit a target; when it fails, the patient must rely on slower visual and proprioceptive feedback loops, which produces the characteristic terminal dysmetria and intention tremor.

Three systems can each produce ataxia, and distinguishing them is the heart of localization:

• Cerebellar — the prediction machinery itself is broken. Signs persist with the eyes open and are not improved by vision. Hemispheric lesions cause ipsilateral appendicular dysmetria and dysdiadochokinesia (the cerebellum controls the same side of the body); vermal/midline lesions cause truncal and gait ataxia with relatively spared limbs.
• Sensory (proprioceptive) — the cerebellum is intact but starved of position information from large-fiber peripheral nerves or the dorsal columns. Because vision can substitute for the missing proprioceptive input, these patients are dramatically worse with the eyes closed (positive Romberg) and lack nystagmus or scanning speech.
• Vestibular — loss of the gravitational/head-motion reference frame, producing imbalance and oscillopsia. This becomes diagnostically central in CANVAS, where vestibular areflexia joins cerebellar and sensory failure.

The single most useful first question is not where but how fast. The temporal profile — hyperacute, acute, episodic/recurrent, subacute, or chronic-progressive — reorders the entire differential and sets the urgency. An ataxia evolving over hours forces an immediate hunt for stroke, intoxication, or Wernicke encephalopathy; one evolving over years in a young person points toward an inherited degeneration. Building the differential around tempo first, and localization second, prevents the common error of ordering a genetic panel before excluding a treatable or emergent cause.

Chronic progressive ataxia spans genetic, metabolic, structural, and acquired causes, and the differential is too large to attack by brute force. The way to make it tractable is to let a handful of phenotypic axes partition it: age of onset, mode of inheritance, the company the ataxia keeps (neuropathy, pyramidal signs, ophthalmoplegia, movement disorder), and systemic clues (cardiomyopathy, diabetes, telangiectasia). Each axis collapses dozens of possibilities into a short list.

Inheritance and onset age are the strongest discriminators. As a clinical rule of thumb, ataxia beginning before about age 25 is far more likely to be autosomal recessive (Friedreich ataxia, ataxia-telangiectasia, AVED, the oculomotor-apraxia ataxias), while adult onset with a vertical family history favors a dominant SCA. The recessive disorders tend toward a spinocerebellar/sensory phenotype with neuropathy and areflexia, reflecting their frequent involvement of mitochondrial or DNA-repair machinery; the dominant SCAs are predominantly cerebellar-plus syndromes.

Two reasoning traps deserve emphasis. First, a negative family history does not exclude a genetic cause: recessive disease produces unaffected carrier parents, de novo dominant expansions occur, penetrance is age-dependent, and biallelic RFC1/CANVAS — now recognized as one of the commonest late-onset recessive ataxias — typically presents sporadically. Second, treatable mimics must be excluded before a degenerative label is accepted. Vitamin E deficiency (AVED), vitamin B12 and thiamine deficiency, hypothyroidism, Wilson disease, celiac/gluten ataxia, and paraneoplastic cerebellar degeneration (anti-Yo, anti-Hu, classically in adults over 40) are all reversible or arrestable if caught early — and several are silently progressive if missed. The associated-features table below operationalizes this pattern-matching across the acute, episodic, and chronic presentations.

The workup is best thought of as a tiered funnel that spends cheap, high-yield, treatment-changing tests first and reserves expensive genetic testing for after the treatable and acquired causes have been excluded. The sequence is deliberate: imaging and neurophysiology characterize the lesion and narrow the genetic differential, the metabolic screen rescues the treatable cases, and only then does genetic testing confirm a degenerative diagnosis.

Neuroimaging does more than exclude a mass. The pattern of atrophy is itself diagnostic data: isolated spinal cord (especially cervical) atrophy with relatively preserved cerebellum points to Friedreich ataxia early in its course; pan-cerebellar atrophy fits the SCAs; and dentate or brainstem T2 signal, or an MRS lactate peak, redirects toward mitochondrial or metabolic disease. Nerve conduction studies are pivotal because a large-fiber sensory axonal neuropathy is a fingerprint shared by Friedreich ataxia, AVED, and CANVAS — finding it immediately reshapes which genes to test.

The decision that most often goes wrong is the genetic strategy, and the reason is a structural blind spot in modern sequencing. The commonest hereditary ataxias — FXN, the polyglutamine SCAs, and RFC1 — are all caused by expanded short tandem repeats, and standard short-read exome/genome sequencing cannot reliably size them: 150-bp reads cannot span hundreds to thousands of repeat units, and the repetitive sequence defeats alignment. A short-read exome reported as normal therefore does not exclude the most likely diagnoses. Practically, this means choosing the test by phenotype: lead with FXN GAA repeat-primed PCR when the picture is recessive/Friedreich-like, a targeted SCA repeat panel when the pedigree is dominant, and reserve a comprehensive panel or exome (explicitly paired with dedicated repeat analysis or long-read sequencing) for the remainder. Verifying that the ordered test actually includes repeat sizing is the step that most directly improves diagnostic yield.

Friedreich ataxia (FRDA) is the textbook example of a recessive, intronic, loss-of-function repeat expansion — a mechanism that looks nothing like the dominant polyglutamine SCAs despite both being triplet-repeat diseases. The defect is biallelic expansion of a GAA trinucleotide repeat in intron 1 of FXN, which encodes frataxin, a small mitochondrial matrix protein required for the assembly of iron-sulfur (Fe-S) clusters — the redox cofactors of respiratory-chain complexes I-III and aconitase. The landmark identification of this intronic GAA expansion as the cause of an autosomal recessive disease was made by Campuzano et al. 1996, who showed that the overwhelming majority of patients are homozygous for the expansion.

The key conceptual point is why an intronic repeat causes disease at all. Because the GAA sits in an intron, it does not alter the protein's amino-acid sequence; instead the expanded repeat adopts non-B DNA structures (triplex/sticky-DNA) and nucleates heterochromatin formation across the locus, silencing transcription of an otherwise normal gene. FRDA is thus a disorder of frataxin quantity, not quality — a partial loss of expression — which is exactly why longer repeats (more silencing, less residual frataxin) track with earlier onset and more severe disease, and why the rare patients with one point mutation can be more severely affected. Loss of frataxin starves Fe-S cluster biogenesis, mitochondrial iron accumulates, and the resulting oxidative stress drives degeneration of the most metabolically demanding neurons — dorsal root ganglia, dorsal columns, spinocerebellar tracts, and the dentate nucleus — explaining the signature combination of sensory ataxia, areflexia, and pyramidal signs. This silencing model also frames the therapeutics: the GAA repeat itself is the upstream lesion, but approved treatment (omaveloxolone) acts downstream on the oxidative-stress consequence rather than restoring frataxin. FRDA is the most common hereditary ataxia worldwide, with a prevalence of approximately 1/50,000.

The autosomal dominant spinocerebellar ataxias (SCAs) are a heterogeneous family of >40 named disorders, but the most common and most instructive subset — SCA1, 2, 3, 6, 7, 17 and DRPLA — share a single mechanism: an expanded CAG repeat translated into an elongated polyglutamine (polyQ) tract within an otherwise unrelated protein. This is the mechanistic mirror image of Friedreich ataxia. There, an intronic repeat causes recessive loss of a normal protein; here, a coding repeat creates a dominant gain of toxic function. The expanded polyQ protein misfolds, aggregates, and sequesters transcription factors and chaperones — which is why a single mutant allele is sufficient (dominant) and why these are progressive neurodegenerations rather than simple deficiency states.

The polyQ mechanism explains the clinical signatures that let you separate the SCAs at the bedside. Anticipation — earlier onset and greater severity down the generations — is a direct consequence of repeat instability during meiosis: the CAG tract tends to lengthen, most dramatically in paternal transmission (spermatogenesis), so a child of an affected father can present decades earlier than the parent. Repeat length inversely correlates with onset age, the same dose-response logic seen in FRDA. And because each SCA expresses its toxic protein in a partly distinct neuronal population, each carries discriminating features: pyramidal signs in SCA1, strikingly slow saccades and neuropathy in SCA2, ophthalmoplegia and a comparatively pure motor picture in SCA3 (Machado-Joseph disease, the most prevalent SCA worldwide), a nearly pure late-onset cerebellar syndrome from small expansions in SCA6, and pathognomonic progressive macular degeneration in SCA7.

Two allelic relationships are worth internalizing because they recur on exams. SCA6 and episodic ataxia type 2 are different mutations in the same gene, CACNA1A — small CAG expansions yield the chronic degeneration, while loss-of-function/missense variants yield the paroxysmal channelopathy. And the contrast that closes the loop on this module: not every repeat-expansion ataxia is dominant. CANVAS — cerebellar ataxia with sensory neuropathy and bilateral vestibular areflexia — is caused by a biallelic, recessive intronic AAGGG pentanucleotide expansion in RFC1, identified by Cortese et al. 2019 as a frequent cause of late-onset, apparently sporadic ataxia. Like FXN and the SCAs, RFC1 is invisible to standard short-read sequencing and must be sought with dedicated repeat testing.