Alan Garnham Language and Thought

1
LANGUAGE AND THOUGHT

• COMPREHENSION 2

2
BRANSFORDS THREE CLAIMS

• The mental representation of a sentence does not
  correspond to any of its linguistic
  representations
• Comprehension is an integrative process
• pieces of information from different parts of the
  text have to be put together properly
• Comprehension is a constructive process
• information from the text is understood in the
  context of relevant background information that
  is not explicitly mentioned in the text

3
INFERENCE IN COMPREHENSION

• Integrative and constructive processes often go
  hand in hand.
• Background information has to be used to link
  pieces of information in different parts of a
  text.

4
HAVILAND AND CLARK (1974)

• We checked the picnic supplies
• The beer was warm
• We took the beer out of the trunk
• The beer was warm

5
NECESSARY AND ELABORATIVE INFERENCES

• NECESSARY INFERENCES
• The inference required in the Haviland and Clark
  passage (that the beer was part of the picnic
  supplies) is NECESSARY for a coherent
  interpretation of the text.
• This does not mean it HAS to be made, only that,
  if it is not, a coherent interpretation will be
  lacking.

6
NECESSARY AND ELABORATIVE INFERENCES

• ELABORATIVE INFERENCES
• Other inferences are MERELY ELABORATIVE (e.g.
  that the delicate glass pitcher broke when it was
  dropped)
• Are such MERELY ELABORATIVE inferences made
  during reading?

7
ELABORATIVE INFERENCES (cont.)

• Bransford claimed that many of them were made.
• This idea raises two questions
• Which ones?
• Why waste cognitive effort?

8
ELABORATIVE INFERENCES (cont.)

• Subsequent researchers (e.g. Thorndyke, Corbett
  Dosher, Singer, McKoon Ratcliff) have claimed
  that they are not.
• Bransford's own data came from memory tests, and
  it was claimed that his subjects made inferences
  when they were asked memory questions,not while
  they were reading passages or listening to them.

9
SOME RECENT IDEAS ABOUT ELABORATIVE INFERENCES

• They aren't made in "normal reading".
• They are made only when they depend of
  information that is "readily available" in memory
  (McKoon Ratcliff - Minimalist Hypothesis)
• Some general types of inference are made and
  others are not (e.g. superordinate goals, but not
  causal consequences, Graesser et al.)
• They are only made if context is highly
  supporting (O'Brien, Myers et al.)

10
IF MENTAL REPRESENTATIONS OF TEXT AREN'T
LINGUISTIC REPRESENTATIONS, WHAT ARE THEY?

• Representations of events (actions, states,
  processes), and their participants (people,
  animals, objects) and the relations between them
  (spatial, temporal, logical, causal, intentional,
  moral)
• called mental models, situation models,
  discourse models

11
GLENBERG, MEYER, LINDEM (1987)Example Materials

• John was preparing for a marathon in August.
• After doing a few warm-up exercises, he put on
  (ASSOCIATED VERSION) OR took off (DISSOCIATED
  VERSION)his sweatshirt and went jogging)(1st)
• He jogged halfway around the lake without too
  much difficulty.(2nd)
• Further along his route, however, John's muscles
  began to ache.(3rd)
• PROBE Sweatshirt - at one of the "" points
• ASSOCIATED lt DISSOCIATED at 2nd and 3rd test
  points

12
SOME OTHER IDEAS ABOUT TEXT COMPREHENSION

• Story Grammars (e.g. Rumelhart, 1975) - stories
  have a grammatical structure similar to that of
  sentences.
• What about other types of text?
• Construction-Integration (Kintsch, 1988) - Of the
  possible meanings that parts of a text might
  have, those that are related support each other,
  while those that are not are eliminated (e.g.
  "wrong" meanings of ambiguous words)