How Devanagari works: a script where vowels ride on consonants

In an alphabet such as English, consonants and vowels are separate letters. In an abugida, the basic letter is a consonant that includes a vowel. The letter क is not k but ka. To change the vowel, a sign called a matra is attached to the consonant: का is kaa, कि is ki, की is kii, कु is ku, कू is kuu, के is ke, कै is kai, को is ko and कौ is kau. A vowel that begins a word or follows another vowel is written with its own full letter, such as अ, आ, इ or उ.

One of the matras is written to the left of the consonant, though it is pronounced after it. In कि, the vowel sign for i is drawn before the क, and learners soon get used to reading it in the right order.

To write a consonant with no vowel at all, a small mark called the halant, or virama, is placed under it: क् is just k. When a vowel-less consonant is followed immediately by another, the two are often joined into a single shape called a conjunct. The conjuncts क्ष (kṣa), त्र (tra), ज्ञ (gya) and श्र (shra) are among the most common. Some are easy to see as combinations, and others have changed so much that they are learned as separate characters.

The traditional chart of consonants is not alphabetical in the English sense. It is organised by the place in the mouth where each sound is made, from the back of the throat to the lips, in five rows of five: क ख ग घ ङ, then च छ ज झ ञ, then ट ठ ड ढ ण, then त थ द ध न, then प फ ब भ म. Within a row the order is the same each time: unvoiced, unvoiced aspirated, voiced, voiced aspirated, and then the nasal. After these come the semivowels य र ल व, the sibilants श ष स, and ह, which brings the total to 33 consonants. This system was refined by ancient Sanskrit grammarians, and is one of the oldest scientific analyses of speech sounds.

Most Devanagari letters hang from a horizontal line, the shirorekha, which runs along the top of a word and joins its letters together. A dot above a letter, the anusvara, and a dot with a crescent, the chandrabindu, mark nasal sounds. A dot placed underneath a letter, the nukta, is used for sounds that came from Persian, Arabic and English, such as the z in ज़ or the f in फ़.

Devanagari descends from Brahmi, the script of ancient India, through later forms used in the north, and had taken roughly its modern form by about the eleventh century. Different languages adapt it a little to their own sounds. It is written from left to right, and it is used for Hindi, an official language of India, and for Nepali, the official language of Nepal, as well as for scripture, newspapers and film titles.

In Unicode, Devanagari occupies its own block of code points, and a visible syllable often consists of several of them. The syllable क्षि, for instance, is made of four: क, the halant, ष and the vowel sign ि. The software that draws the text has to join, reorder and position them, so the number of characters in a string is not the number of shapes on the screen. That is why a Hindi headline can behave differently from an English one when a program measures its length or breaks it across lines.

Hindi news also mixes scripts. Company names, abbreviations and many technical terms are often written in Latin letters inside Hindi sentences. For a program that needs to tell whether a text is in Hindi, the sensible approach is to look at what share of its letters are Devanagari, not to expect every one of them to be.

• Where to start with Indian literature: twelve writers and a first book for each

  A starter guide to Indian writing, from Tagore and Premchand to Booker winners and writers in translation, with one place to begin for each.

• Visual hierarchy: how designers direct your eye

  Size, contrast, spacing and alignment decide what you notice first on a page or screen. A plain guide to visual hierarchy, with a news card as the example.