Ling 425/709

Explanation

Older Studies

   1. General Issues

      a. Are children using semantic or syntactic rules?

            Agent-Action vs. N-V

      b. Are children’s rules specific or general?

            “people who play with toys” vs. Agent

            sentence-initial noun vs. Subject

      c. Are children using linguistic or distributional features to acquire syntax?

            linguistic features include semantic and syntactic knowledge, e.g., Agent, Subject

            distributional features include sentence position, relative order

      d. Do children’s syntactic rules have limited or extensive productivity?

            Brown & Fraser (1964) described two grammars:

      e. How observationally adequate is the grammar?

            Lean–only accounts for a subset of the child’s data

            Rich–account for all of the child’s utterances

      f. How complex are the rules?

      g. How similar is the child’s grammar to the adult’s?

            Continuous–constrained by UG; uses adult rules

            Discontinuous–not constrained by UG; uses rules not found in adult grammars

   2. Pivot Grammar (Braine 1963)

      a. Child selects a small group of words as pivots (Table 7.11, p. 265)

            i. pivots make up a small class of words

            ii. pivots are frequent

            iii. pivots have a restricted position (utterance initial or utterance final)

            iv. pivots don’t occur together

            v. pivots don’t occur alone

      b. Child combines pivots with open class words to produce multiword utterances

      c. Andrew’s pivot grammar

            S –> P1 + O

            S –> O + P2

            P1 –> all, I, no, see, more, hi, other

            P2 –> off, by, come, there

      d. Criticisms

i. the grammar is inconsistent with the data, e.g., “I see”, “off + O” & “O + off” (Brown 1973)

            ii. the grammar is too lean; it does not apply to other children (from Late Stage I?)

            iii. the grammar is inadequate; it overlooks semantic distinctions,

                  e.g., Kathryn’s “mommy sock” (Bloom 1970; from Late Stage I, 226 syntactic types)

      e. Assessment

Pivot grammar offers an initial learning heuristic for building word classes and rules from distributional evidence (positional consistency). It could be revised to include semantic information, c.f., Braine (1976), Maratsos & Chalkley (1980).

   3. Transformational grammar (Bloom 1970)

Attempted to use transformational grammar to describe children’s early sentences. Bloom changed features of the adult grammar to describe the children’s grammar.

      a. Grammars for Gia I (1;7.14, MLU 1.12, 55 syntactic types)

Grammar 2 is not strictly a pivot grammar. Bloom uses the N + N sequence for ‘locative’ relations and ‘subject-object’ or ‘genitive’ relations, e.g., ‘Gia eyes’ (reaching for doll ‘Blueyes’). It is very similar to Andrew’s grammar.

      b. Grammar for Kathryn I (MLU 1.32, 226 syntactic types, p. 275)

            1. S1 –> Nom (Ng) {NP/VP}

            2. S2 –> Pivot + N

            3. VP –> VB {NP/Part}

            4. NP –> ([6]) (ADJ) N

            5. Nom –> {N/Dem}

            6. Pivot –> ‘hi’, ‘oh’, ‘OK’, ‘thank you’

            7. Ng –> ‘no’

      Formal differences from the adult grammar

            1. Nom–restricted to the subject position

                  Problem

                        i. What motivates the child to form this category?

                        ii. How does the child unlearn the non-adult categories?

            2. N N sequences produced by S1, e.g., ‘mommy sock’ and ‘this book’

                  Bloom argued that these sequences derive from different underlying structures, p. 277

      c. Bloom’s tests for syntactic relations

            i. Sentence patterning–a construction occurs in the same place as its head

            ii. Linear order–a construction shows the adult word order (ambiguous)

iii. Replacement sequences–shorter constructions have the same relation as longer ones, e.g., ‘mommy milk’ and ‘mommy’s milk’

            iv. Replacement and deletion–substitutions indicate the syntactic relation,

                  e.g., ‘Baby milk’ and ‘touch milk’

            v. Non-linguistic context–the child’s actions and objects in the environment

      d. overgeneration

            i. Bloom’s grammar for Kathryn I generate three and four-word utterances

            ii. Bloom proposed using a reduction transformation to account for this constraint

            iii. criticisms of reduction transformations (Wall 1972, Brown 1973)

                  a. they violate the ‘recoverability’ constraint

                  b. they are unique to the child grammar

            iv. reduction transformations may describe performance factors rather than child grammar

      e. assessment

Bloom’s grammars demonstrate significant constraints typical of children’s grammars. Bloom also provides criteria for establishing constituents in children’s language samples.

            Bloom showed that two word utterances may be structurally ambiguous.

   4. Semantically oriented approaches (Bowerman 1973)

      a. arguments against establishing a VP constituent in children’s grammars (Table 7.15)

      b. arguments against establishing the subject relation in children’s grammars

            i. Children do not use passive constructions that demonstrate the syntactic relation

            ii. Children do not show Verb Agreement, Subject Pronominalization or Case Marking

   5. Brown (1973)

      Brown’s eleven semantic relations for Stage I grammars (1973: 187-98; Table 7.16)

a. Brown suggested that all children acquire the same set of semantic relations as part of their cognitive development. ‘Representation starts with just those meanings that are most available to it, propositions about action schemes involving agents and objects, assertions of nonexistence, recurrence, location, and so on’ (p. 200).

      b. criticism (Howe 1976)

            i. Semantic approaches yield inconsistent descriptions

            ii. Researchers imposed adult semantic relations on the children’s utterances

      c. Assessment

i. Brown provides a description of semantic relations children use across a number of different languages.

ii. His semantic relations produce a discontinuity with the adult grammar in that they allow children to express propositions without an Action.

iii. Brown also has to explain why children do not use more than two semantic relations, e.g., Agent + Action + Object.

   6. Braine (1976)

a. Subjects: 11 children in early Stage I acquiring English, Samoan, Finnish, Hebrew and Swedish

      b. Method: proposed three criteria for identifying children’s productive syntactic rules

            i. Positional consistency

                  Number of utterances with

            ii. Productivity–evidence of a novel creation, e.g., Andrew’s ‘no wet’

            iii. Semantic consistency–utterance conveys a consistent semantic relation

            iv. Classification of children’s utterances

                  a. positional productive patterns: patterns that meet all three criteria

b. positional associative patterns: patterns that meet positional and semantic consistency

                  c. groping patterns: patterns that meet semantic consistency, and which are

                        1. small in number

                        2. produced with some uncertainty or hesitation

d. free order patterns: patterns that meet productivity and semantic consistency, and are not groping patterns

                  e. ‘other’ patterns: ~25% of the data; possibly imitations of adult models

c. Results: ‘Children differ considerably in the kinds of contents expressed by their productive patterns and in the order in which they acquire them’ (p. 57).

            Children begin with restricted patterns of expression–formulae of limited scope

            Argument against using a more general phrase structure grammar for Kendall

                  Assume Kendall starts with rules for Actor/Action, e.g.,

                        1. S –> NP + VP

                        2. VP –> (V) + (NP)

                  Adding locatives would change the second rule to

                        3. VP –> (V) + {(NP)/Locative}

Prediction: The relative order of subject and locative is fixed by rule one, so Kendall should use the locatives with the appropriate order. Kendall initially used locatives with a groping pattern.

      d. Assessment

            i. Braine contributes important methodological criteria for identifying productive rules

            ii. He identified considerable variation in the semantic relations children start with

iii. His patterns do not solve the discontinuity problem–how do children attain the adult grammar?