The letters are arranged by where in the mouth they are made, which no European alphabet does, and that arrangement is the fastest way to learn them. Five rows of five, each row a place of articulation, each column a manner: unvoiced, unvoiced with a puff of air, voiced, voiced with a puff, and the nasal that belongs to that place. Learn the grid rather than the list and the forty six letters become five patterns.
What actually costs the fortnight is not the letters. It is the conjuncts, where two consonants meeting inside a word are written as a single fused shape, and there are several hundred of those.
Six facts and the script becomes learnable
- each consonant carries a built in a. The letter for k is really ka, and writing k alone requires a mark to cancel the vowel. This is the single idea that makes Devanagari click.
- matras change the vowel, attaching above, below, before or after the consonant. The one that goes before is written first and read second, which is the one genuine trap.
- the line across the top joins the word, so word boundaries are visible at a glance in a way they are not in Arabic script.
- conjuncts join two consonants when no vowel comes between them, and they form a new shape. There are a few hundred and the common ones are learned by exposure rather than by table.
- ka, kha, ga, gha are four sounds, not two. Each place of articulation has an unaspirated and an aspirated form, voiced and unvoiced, and the script gives each one its own letter.
- there are no capitals, and the script is written left to right, which makes it the easier of the two Indian systems for an English reader to start with.
An alphabet where each consonant carries a vowel
Every consonant sign comes with a short a already inside it, so a bare letter is a syllable rather than a sound. Other vowels are written as marks attached to the consonant, above it, below it or to either side, and a vowel with nothing before it has its own standalone form. That is an economical system and it means a Devanagari word has fewer symbols than a transliteration of it does.
The conjuncts follow from the same design. If two consonants come together with no vowel between them, the built-in vowel of the first has to be suppressed, and the usual way to do that is to fuse the two signs. Some fusions are transparent, with one letter losing its vertical stroke and sitting beside the other. Some are not, and a handful of the commonest ones look like nothing either parent letter looks like.
The headline running along the top is the last piece and it is worth knowing how it is produced. It is drawn after the letters rather than before them, and letters hang from it, which is why Devanagari handwriting is taught with the line last and why a page written the other way round looks wrong in a way that is hard to name.
Learning forty six letters and stopping
Here is the stall. The grid goes in over a few days, the vowel marks take a few more, and a learner reads simple words and declares the script done. Then real text arrives, full of conjuncts, and every second word contains a shape that was not on the chart.
What follows is usually a period of guessing, because a conjunct is decodable if you know it is a conjunct and unrecognisable if you expect a letter. The exit is a short and unglamorous piece of work: take the thirty or forty commonest fusions, learn them as shapes, and treat the rest as decodable on sight once the principle is clear. That is a weekend, and skipping it is what produces a reader who is still slow at Devanagari a year later.
Two letters, one shape
How a conjunct is made, frame by frame
Each frame is the same two consonants. Nothing is being abbreviated: the fusion is what suppressing the built-in vowel looks like.
- 01Two consonants, side by sideEach carrying its own built-in short a
- 02The first vowel has to goBecause the two consonants are meant to meet with nothing between them
- 03The first letter loses its strokeThe vertical stroke that the headline hangs it from
- 04The two fuse into one shapeTransparent in most cases, and unrecognisable in a handful
The third frame is the pattern that covers most of the several hundred fusions, which is why a learner who knows the principle can decode a conjunct they have never seen. The exceptions are a short list and are worth memorising as shapes.
Two things for the fortnight
A conjunct chart is the object to work through, and it is worth finding one that groups the fusions by their pattern rather than listing them alphabetically. It replaced meeting each conjunct individually in text and treating it as a new letter, which is what happens by default and which turns a system into several hundred facts. Free versions exist and the useful ones show the two parent letters beside the fused form.
The other is a pen. Write the letters with the headline drawn last, the way the script is actually produced, because the order teaches the shapes: the vertical stroke that most letters share is what the line hangs them from, and the letters that lose that stroke in a conjunct are exactly the ones that fuse. It replaced typing, which produces correct Devanagari and shows you none of this, since the software assembles the conjuncts for you.
The mistake: pronouncing the vowel that is written
The built-in vowel is dropped in predictable places and most reliably at the end of a word, so the letters spell one thing and everybody says another. The rule is not marked anywhere, it is learnable in an evening, and it is usually taught late or not at all.
A learner who has not met it reads correctly and sounds like a recitation, adding a syllable to the end of a great many words and to the middle of some. Since they are understood, nothing corrects it. Learn the rule with the alphabet rather than after it, because it is not an advanced refinement: it is the difference between reading the script and pronouncing the language, and the same distinction runs through retroflex and aspirated pairs that English does not distinguish.
What the script opens and what it does not
Devanagari decides what you can read rather than what you can say, because the everyday spoken language is shared with Urdu and the two diverge above the level anybody speaks, in the formal vocabulary and in the script. That division is the subject of the Sanskrit half, the Persian half, and English on top. The words being read carry gender on every noun and change shape before a postposition, in gender on every noun, and postpositions where English puts prepositions. And reading fast enough for any of it to be automatic is the ordinary problem of volume in material just past your level, and why that is the sweet spot.