Universal grammar includes parameters whose values must be set through limited input exposure, such as the null-subject parameter (whether subjects must be phonetically realized). Acquisition involves rapid parameter-setting triggered by minimal positive evidence, explaining the relative ease of childhood language learning.
Examine cross-linguistic parameter values and study how children acquiring different languages set parameters with limited input; test cases where parameters interact (clustering of properties).
Parameter-setting is not learning rules or statistics; it is setting binary (or few-valued) switches in a universal grammar, which explains rapid acquisition despite sparse input.
From your study of Universal Grammar, you know the theoretical framework: all human languages share an innate grammatical endowment — UG — that constrains the space of possible grammars, making language acquisition feasible even for young children who receive impoverished, noisy input. From your study of parameter-setting in acquisition, you know that the variation between languages is not random but structured: it results from different parameter values — binary or small-valued switches within UG that get set by exposure to the target language. Parameter learning is the developmental question: *how* do children set these parameters correctly given limited input?
The canonical example is the null-subject parameter (also called the pro-drop parameter). Languages like Spanish and Italian allow sentences without an overt subject — *Habla español* ("Speaks Spanish" — meaning "He/she speaks Spanish") is grammatical because the subject is implied by verb morphology. English requires an overt subject: *He speaks Spanish* is well-formed but *Speaks Spanish* is not (except in special constructions). A child acquiring Spanish must set the null-subject parameter to [+pro-drop]; a child acquiring English must set it to [−pro-drop]. The poverty of the stimulus argument applies here: no one explicitly teaches children that English requires overt subjects. Yet English-acquiring children reliably produce subjects very early, and Spanish-acquiring children reliably omit them — without explicit instruction, and despite input that doesn't directly contrast the two possibilities.
The mechanism proposed for parameter-setting is triggering: specific structures in the input — positive evidence — activate parameter settings. A child hearing sentences with rich verbal morphology (which correlates with null-subject languages) might trigger the [+pro-drop] setting. The critical property is that triggering is fast and requires minimal exposure — not thousands of examples, but a few clear instances of the relevant construction. This contrasts sharply with statistical learning models, which require large amounts of data and gradual accumulation of evidence. Parameter-setting posits a qualitatively different cognitive process: a switch flips, and the child's grammar is restructured accordingly.
An important and still-debated aspect of parameter learning is parameter clustering: some parameters appear to come packaged with correlated properties. Pro-drop languages, for instance, tend to allow postverbal subjects (*Habló María* — "Mary spoke"), have richer verbal morphology, and permit null expletives. If these properties are linked to a single parameter, setting it [+pro-drop] should automatically produce all the correlated properties — a prediction that can be tested in acquisition data. When children begin allowing null subjects, do they simultaneously acquire the correlated properties? The evidence is mixed, but parameter clustering remains a powerful argument for the view that children are not learning surface patterns but setting deep grammatical switches whose effects cascade through the grammar. This is what distinguishes parameter-setting theory from purely empiricist accounts of language acquisition: the claim is not that children learn from input, but that input *triggers* innate structures that then generate grammatical knowledge that extends far beyond what the input explicitly contained.