Domain-general processes: representation of gist
7.2 The present study
7.2.4 Surface form and gist representation
It will be important to examine the relation between surface form representation and gist representation (i.e. story events) in the normal group, and then to see whether there may be abnormalities in this relation in the groups with developmental disorders. It is expected that a strong dissociation between surface form and gist representation is present in the hyperlexic group, and to a lesser degree in the non-hyperlexic group. If the dyslexic and MLD groups’ poor retention of surface form constitutes a compensatory response, one would expect negative correlations between surface form and gist representation. On the other hand, if there is a true deficit in comprehension in both groups, then one would expect that the representation of gist may be more constrained by the amount of surface form retained (i.e. positive correlations between surface form and number of meaningfully retold story events).
Method
Materials as chapter 6
Procedure
After the participant had completed the story reading task and surface form recognition test for a story, he or she was asked to retell the story. In the ordered condition, the experimenter asked: “Can you tell me what happened in this story” and tape recorded the participant’s response. In the scrambled condition, the experimenter said: “These stories got mixed up. When you read these, try to
figure out what really happened. Tell me how you think the story would go, if it would be in the right order.” The response was then tape recorded.
Representation of story events (i.e. gist) - Scoring System
The most important story events were selected for each story and each received a score o f 1 point, if retold in the correct (i.e. meaningful) order. A story event was considerd to be correctly retold, if the gist of it was captured, details such as exact names and places or illustrative adjectives, etc. were not required. If essential information was missing or wrongly interpreted, the response scored 0. The setting and ending events (i.e. first and last sentences) were not included in the scoring, since the position of these was identical in both scrambled and ordered stories. In the ordered condition story events had to be retold in the order that subjects read them in the stories. On the other hand, in the scrambled condition the subjects had to re-order the story events so that they matched the original stories (i.e. in the order they were presented in the ordered stories).
The exact scoring system was developed after story construction. It was found that the number of story events was not identical and the maximum scores varied slightly. Story condition was counterbalanced across subjects, and the maximum scores per condition was the same (i.e. 40). Table 53 shows the total number of story events per story in the two story sets. Each set of stories was presented to half of the subjects in ordered format, and to the other half in the scrambled format.
Table 53: Total number of story events per story in the two story sets
1 11 2 10
3 9 4 9
5 10 6 10
7 10 8 11
In each condition a total number o f 40 story events were thus obtainable. The scores are presented as averages (i.e. divided by the number o f stories a participant read per condition). One story is shown below, both in the ordered and the scrambled condition, to illustrate the scoring system. The story events and the sequence in which they were presented is shown in table 54.
O rd ered story
Bob and Jane set out to go swimming one day. They went to a beach they had never been to before. They swam far out to sea.
Suddenly Bob spotted a big cave in the cliff. They swam into the cave.
Then they climbed onto a ledge to have a rest.
The swimming had made them very tired and they soon fell fast asleep. When they woke up the tide had risen.
There was only one way to get out. They had to swim underwater out of the cave. They were very lucky to get home safely.
Scrambled story
Bob and Jane set out to go swimming one day. When they woke up the tide had risen. Suddenly Bob spotted a big cave in the cliff. They had to swim underwater out of the cave. Then they climbed onto a ledge to have a rest. They went to a beach they had never been to before. There was only one way to get out.
They swam into the cave.
The swimming had made them very tired and they soon fell fast asleep. They swam far out to sea.
They were very lucky to get home safely.
Table 54: Story events and their sequence in the ordered and scrambled condition
Beginning set out to go swimming Beginning
1 went to a beach 6
2 swam far out 10
3 spotted a cave 3
4 swam into cave 8
5 climbed on a ledge 5 6 tired from swimming 9
7 fell asleep 9
8 woke up 2
9 tide has risen 2
10 only one way out 7
11 swam underwater 4
Ending get home safely Ending
7.3 Representation of gist in autism
Subjects
Fourteen subjects with pervasive developmental disorder could be matched on FIQ with either normal control subjects or subjects with learning difficulties. The means and standard deviations of chronological age and reading abilities (e.g.
Basic Reading and Reading Comprehension - raw and standard scores) are presented in table 55.
Table 55: WORD (BR and RC), FIQ and CA for autistic/ normal-MLD comparison: means, sds
Measure CA FIQ VIQ BR - Raw BR - Standard RC - Raw RC - Standard Autistic, n=14 Normal/MLD, n=14 13.0(3.0) 80.5 (24.7) 79.6(22.8) 39.6(7.7) 93.0(19.1) 18.5(7.6) 78.9(13.6) 12.0 (2.3) 80.0 (24.8) 81.7 (26.8) 35.6 (9.3) 86.9 (20.2) 20.4 (6.5) 84.7(21.0) R esults
The mean number of story events retold in the correct meaningful order per subject per condition was compared. A repeated measures ANOVA with one within-group factor with two levels (story condition: ordered and scrambled) and one between group factor with two levels (group: autistic and MLD/normal). There was a significant main effect of story condition (F(26,l)=54,2, p=.000); subjects retold significantly more story events in the ordered than the scrambled condition. There was no significant effect of group (F(26,l)=2.52, p=, 125); but there was a significant group by story interaction (F(26,l)=9.36, p=.005. A follow-up t-test showed that the MLD/normal group retold more story events in the ordered condition than the autistic group (t(26)=-1.95, p=.062). Table 56 presents the means and standard deviations of the scores in the ordered and the scrambled condition, and for the difference between these scores.
Table 56: Number of story events retold for autistic/normal-MLD comparison: means, sds
ordered condition 10 5.3 (1.7) 7.5 (2.9) scrambled condition 10 3.9(1.1) 3.9 (2.0) difference ordered minus scrambled 1.4 (1.7) 3.6 (1.7)
Discussion
The results show that subjects with autism, matched on FIQ with a mixed group of normal and MLD control subjects, retold the same number of correctly connected story events when presented with scrambled stories. On the other hand, when presented with ordered (i.e. coherent) stories, the story retelling abilities of the control group were significantly better than those of the autistic group. It thus seems that the autistic group is less able to make use of coherently presented information, when extraction of gist is required as opposed to surface form retention..
The relative inability to use meaning is reminiscent of the finding that autistic subjects provide the most frequent pronunciation of homographs rather than the pronunciation required by a sentence context (Happé, 1997). It is analogous to findings such as the unusual strength in memorising scrambled word strings (Hermelin & O ’Connor, 1967) as opposed to memory for ordered word strings (i.e. sentences), or the greater ability to memorise unrelated items (Tager-Flusberg, 1991) as opposed to memory for related items.
Enhanced surface form retention (see chapter 6) together with decreased representation of meaning or gist for ordered stories is a new finding supporting the account of weak central coherence. The fact that both groups recalled the same number of correctly connected story events in the scrambled condition is a strong indication that the scrambling manipulation has equalised the groups in terms of the ability to extract gist. The scrambled condition removed the benefits of story coherence for the control group, thus they performed significantly worse in this condition. However, the autistic group’s processing itself is weakly coherent, and thus this group was unable to improve significantly when information was presented in strongly coherent format (i.e. ordered condition).