Linguistic Experiment

Origin. Lev Shcherba, Soviet linguistics, 1930s. Introduced in his famous example "Glokaya kuzdra shteko budlanula bokra i kurdyachit bokryonka" — a grammatically well-formed Russian sentence using nonsense words. The method tests grammatical competence by eliminating semantic content.

Mechanism. Natural language conflates grammar and meaning: we understand sentences because we know both the words and the structure. To isolate grammatical knowledge, construct sentences where the words are unknown but the grammatical markers are intact. Native speakers can still parse the structure, assign grammatical roles, and make judgments about well-formedness. This reveals what they know about grammar independent of vocabulary. The method is introspective — it uses native speaker intuitions as data — but controlled by eliminating the confounding variable of semantic content.

Procedure. Test grammatical structure through constructed nonsense: (1) Identify the grammatical phenomenon — what aspect of grammar are you investigating? Case marking, agreement, word order, derivational morphology? (2) Construct nonsense stems — create novel roots that follow the phonotactics of the language but have no meaning. They should be pronounceable and feel like possible words. (3) Apply grammatical morphology — attach the real grammatical markers of the language to the nonsense stems. Case endings, agreement markers, tense markers, derivational affixes. The grammatical structure should be fully specified. (4) Embed in sentence context — place the nonsense words in sentence frames that provide grammatical context. Use real function words (articles, prepositions, conjunctions) to signal structure. (5) Elicit judgments — present the sentences to native speakers. Ask: What is the subject? What is the object? What is happening? Is this sentence well-formed? Could you add X to this sentence? (6) Analyze responses — consistent responses across speakers indicate shared grammatical knowledge. Variation indicates areas of grammatical uncertainty or optionality. (7) Vary systematically — change one grammatical feature at a time to test its contribution. Compare minimal pairs of nonsense sentences.

Applies to. Grammatical analysis. Language description. Testing linguistic hypotheses. Language teaching (demonstrating grammatical patterns). Psycholinguistic research on syntactic processing.

Limitations. Works only for languages with overt grammatical morphology; isolating languages provide fewer markers to test. Depends on native speaker intuitions, which can be influenced by prescriptive norms or metalinguistic awareness. The nonsense words must be constructed carefully; if they accidentally resemble real words, meaning contaminates the test. The method reveals grammatical competence but not processing — it shows what speakers know, not how they use that knowledge in real time.