Study Sanskrit without inventing grammar

The safest first record is the form exactly as printed. Preserve the script, transliteration, verse reference and source revision. Do not silently turn a joined form into two words or replace a visible ending with the dictionary form you expect. The copied surface is evidence; the proposed analysis is a hypothesis.

This distinction prevents a common loop: an assumed split produces an assumed lemma, the assumed lemma supplies a convenient meaning, and that meaning is then cited as proof of the original split. Keeping the source form beside every proposal breaks the loop and gives a teacher something precise to correct.

When a form is unfamiliar, list more than one plausible route. Could it be a finite verb, a nominal ending, a compound member or the result of sandhi? Use the sentence to rule candidates out. Agreement, government and nearby particles often constrain the field before a dictionary definition becomes useful.

A tool can accelerate this step by displaying paradigms or applying a known rule. Its output should be treated as a candidate generator. Sanskrit forms can be ambiguous, and a small rule engine has a deliberately narrow scope. Write down which rule was applied and which remaining question the tool did not answer.

Use a paradigm or grammar rule to check form, a lexicon to check attested range, and the sentence to check fit. These are different tests. A lexicon can show that a sense exists without proving it belongs here. A paradigm can show that a form is possible without identifying which possible analysis the verse uses.

Then compare an identified edition. Agreement is supporting evidence, not permission to erase your own reasoning. Disagreement is especially valuable: mark the exact junction where the analyses diverge. That junction becomes a focused question for a teacher rather than a vague feeling that Sanskrit is difficult.

Use status words such as observed, proposed, checked and unresolved. They make a notebook more accurate, not less learned. A learner can confidently say, “I observe this ending; I propose this case; I have not yet resolved the compound relation.” That statement carries more knowledge than an unlabelled gloss copied from elsewhere.

Return after a delay and reproduce the analysis without looking. Reopen the evidence, correct the weakest link, and date the revision. Language learning becomes cumulative when earlier guesses remain inspectable. The goal is not to avoid error; it is to make error easy to locate and repair.