Current affiliation: Graduate Institute of Audiology and Speech Therapy, National Kaohsiung Normal University, Taiwan
For children with specific language impairment (SLI), comprehension of reversible sentences is often a source of particular difficulty (
Most research investigating this phenomenon has focussed on reversible active and passive constructions, testing children’s ability to comprehend sentences such as “the boy chases the dog”, “the dog chases the boy” or “the dog is chased by the boy”. Young children may make errors when asked to act out such sentences because they adopt simple strategies such as “pick up the first-named object and do something with it”—a strategy that is usually effective for active sentences but not for passive ones (
Poor comprehension of grammatically complex sentences has been recognised as a problem for many language-impaired children and a handful of experimental interventions have been developed to address these (
Like other open-class words, spatial prepositions are very common in the language. In a rank-ordered list of words in the British National Corpus, the prepositions “before”, “after”, “above” and “below” occur at ranks 185, 111, 786 and 1656 respectively. These frequencies are similar to common concrete nouns such as “house” (rank 191), “cup” (rank 831) or “cat” (rank 1758). Children appear to master the meanings of spatial prepositions, however, relatively late (
This pattern of performance suggested two alternative explanations for children’s comprehension difficulties. One possibility is that the errors arise because of problems maintaining a sentence in short-term memory while computing its meaning. Verbal short-term memory is usually impaired in SLI, and it has been shown to play a role in comprehension problems (
Another possibility is that there is difficulty learning concepts that express relationship between other items. For instance, the prepositions “above” and “below” express a spatial relationship between two arguments,
The current study was designed to explore problems that children with SLI have in understanding reversible sentences. It used a training task modelled on that used by
Do children with SLI have trouble learning to respond accurately to reversible sentences compared with their age-matched controls, even when a very simple syntactic frame is used, and training focuses on just one meaning contrast?
Does variability of nouns used in training sentences help or hinder comprehension of children with SLI? This question was addressed by comparing performance on a set of unique sentences, each of which used different nouns, with a set of sentences using the same nouns, presented repeatedly throughout training. The performance of children with SLI was compared with that of younger typically-developing children matched on overall comprehension level.
Does comprehension of reversible sentences depend on capacity of verbal short-term memory?
Approval for this study was given by the University of Oxford Medical Sciences Division Research Ethics Committee, approval reference MSD/IDREC/2009/28. Parents of all participants gave written informed consent, and the children gave assent after the study was explained in age-appropriate language.
A total of 96 children took part in the study, subdivided into four groups: (a) 6 to 11 year-old children with SLI who received language training (SLI-T,
The children with SLI were recruited from special schools for children with language impairment or support units in mainstream schools. Children were included if they met all of the following criteria:
Performed at least 1 SD below the normative mean on at least two out of the following six standardized tests: the British Picture Vocabulary Scales II (BPVS II) (
Had nonverbal ability within the normal range (no more than 1 SD below average), as measured with Raven’s Coloured Progressive Matrices (
Were able to hear a pure tone of 20 dB or less in the better ear, at 500, 1,000, 2,000 and 4,000 Hz.
Had English as their native language.
Did not have a diagnosis of other neurodevelopmental disorders such as autism or Down Syndrome.
Children meeting the inclusion criteria for SLI were randomly assigned to either SLI-T (i.e., trained group) or SLI-U (i.e., untrained group). These two SLI groups did not differ in age, nonverbal IQ, or any of the standardized language tests.
The same test battery was used to confirm language status for children in the two control groups. Both these groups met the same criteria for nonverbal IQ, hearing and native language, and did not have a history of speech, language, social or psychological impairments. At least five of six standardized language test scores were within normal limits (no more than 1 SD below average). Descriptive information on the participants is given in
Means with different superscripts differ significantly from one another on post hoc Sidak test,
SLI-Trained; |
SLI-Untrained; |
Grammar-matched; |
Age-matched |
|
---|---|---|---|---|
Age (yr) | 8.6 (1.32) | 9.1 (1.32) | 5.8 (0.86) | 8.9 (0.77) |
RCPM SS | 102.9 (13.25) | 100.6 (10.38) | 105.2 (8.47) | 105.8 (11.35) |
TROG-E raw blocks | 8.3 (4.00)a | 9.7 (3.34)a | 9.8 (3.11)a | 14.8 (2.38)b |
TROG-E SS | 73.3 (14.29)a | 75.7 (10.64)a | 102.3 (13.96)b | 97.8 (10.31)b |
BPVSII raw | 69.5 (17.77)a | 75.4 (14.68)a | 67.1 (12.95)a | 92.6 (9.16)b |
BPVSII SS | 87.8 (13.18)a | 87.7 (8.89)a | 108.3 (9.51)b | 102.6 (7.58)b |
NEPSY nonwords raw | 22.3 (8.90)a | 23.9 (8.39)a | 27.0 (7.60) | 32.2 (8.48)b |
NEPSY nonwords SS | 84.1 (16.78)a | 85.5 (14.13)a | 104.3 (15.01)b | 101.8 (15.75)b |
ERRNI Comprehension raw | 8.3 (3.25)a | 9.7 (3.96)a | 8.4 (3.12)a | 12.9 (2.56)b |
ERRNI Comprehension SS | 82.5 (15.17)a | 87.9 (16.29)a | 103.0 (14.6)b | 101.9 (12.58)b |
ACE Naming raw | 10.2 (4.32)a | 12.3 (2.94)a | 10.0 (2.84)a | 16.3 (3.01)b |
ACE Naming SS | 81.6 (13.2)a | 83.8 (9.16)a | 101.1 (7.98)b | 98.8 (10.99)b |
ACE Syntax raw | 15.0 (5.16)a | 17.2 (6.64)a | 17.8 (5.93)a | 24.1 (5.13)b |
ACE Syntax SS | 80.7 (9.2)a | 83.3 (13.7)a | 105.5 (14.16)b | 98.5 (15.05)b |
3.3 (1.38)a | 3.2 (1.15)a | 0.3 (0.48)b | 0.4 (0.5)b | |
Word Span | 4.0 (1.07)a | 4.1 (0.91)a | 4.1 (0.88)a | 4.9 (0.79)b |
The children in the grammar-matched group were aged between 4 and 7 years and were matched as closely as possible with the children in the SLI-T group on receptive grammar using TROG-E, a standardized test that contains four four-choice test items for each of twenty different syntactic contrasts (e.g., negation, subject relative clause, post-modified subject, etc.), including one block that tests comprehension of reversible above/below sentences. A block is scored as passed if all four items in that block are correct. Each child in the grammar-matched group had a TROG-E raw score within 3 blocks of one of the children in the SLI-T group.
The computerized training program was based on that used by
In each session, children were presented with 24 sentences containing either
Children were randomly assigned to be trained with reversible sentences containing either
The child hears a sentence such as “the hedgehog is above the boot”, and must click first on the hedgehog and then on the boot to move them into correct positions in the rocket. The question mark denotes the next position to be filled. The child can press “talk” to have the sentence repeated, and can press “help” to have visual cues added to denote which items must be moved. The format for the preposition pair “before/after” was identical except that a train was shown, and the task was to select items to fill trucks of the train in the correct horizontal order.
A game-like format using an errorless learning procedure was adopted. If the child made a correct response, there was a visible reward, with a cartoon character heading a football over a bar, to add to the child’s collection. If a wrong response was made, the child had an opportunity to try again. The tester would say “Do you want to try again? Or you can click on the Help button here”. The Help button provided a visual cue to indicate which items needed to be selected. To minimize errors due to misperception or forgetting, a Talk button was included: the tester explained to the child that when this button was pressed, the test sentence would be spoken again.
One point was given for each trial when a correct answer was provided on the first attempt by the child, even if the child had to click on the Talk button to listen to the target sentence again. No points were given if the target picture was selected after more than one attempt by the child or if the Help button was used. The program automatically recorded timing of responses in milliseconds. Although times recorded on a laptop computer are not very accurate, and likely to vary from one machine to another (
One to three days after training session 4, children were seen for another session (session 5) in which the same computerized sentence comprehension task was conducted, but using the set of prepositions (i.e.,
Raw score on the NEPSY repetition of nonsense words (
A computerized word span task was used, in which the child’s task was to select pictures to be placed in a horizontal line of numbered fishing nets in the correct order. Children listened to lists of familiar words and saw pictures of the words appear on the computer screen immediately after the auditory presentation of the words. They then clicked the pictures in the same order as they had heard the corresponding words. Once the first picture was clicked, it moved automatically to the first fishing net, and so on for the rest of the pictures until all the pictures were clicked. Items started with a list length of three, and increased by one item each time the list was recalled correctly in the right order. When an incorrect response was given, a second attempt at the same list length was provided. If both trials were failed at list length three, a list length of two was presented. The program stopped automatically when two successive trials were failed at a given list level. The dependent measure was word span, i.e., the longest list length correctly reproduced at least once. The word lists were composed of monosyllabic nouns that were found in the lexicon of two-year-old typically-developing children (
All children, except those assigned to the SLI-U group, were seen for two weeks during which they completed two screening sessions (assessing language, hearing, nonverbal IQ), followed by four sessions of language training and a post-test session (see
Session | Groups | Activity |
---|---|---|
1 | All | TROG-E, BPVS-II, Raven’s Matrices, hearing screen |
2 | All | Other language tests (see |
3 | All except SLI-U | Training session 1 |
4 | All except SLI-U | Training session 2 |
5 | All except SLI-U | Training session 3 |
6 | All except SLI-U | Training session 4 |
7 | All except SLI-U | Transfer of training, session 5 (untrained preposition) |
7 | All | Posttest: BPVS-II and TROG-E (parallel form) |
It would not be possible to demonstrate improvement with training for children who were highly accurate in session 1. We therefore restricted analysis of training effects to those who performed below ceiling (i.e., less than 90% correct) with reversible sentences on session 1. We had anticipated that age-matched controls would find this task very easy, and this proved to be the case, with only one child scoring below ceiling. Accordingly the age-matched group was excluded from further consideration. Of the 28 children in the SLI group, 15 (53%) scored below 90% in the first training session, as did 16 of 28 (57%) children in the grammar-matched control group. It is also of interest to consider how many children score at chance—indicating no understanding of the contrast. Since virtually all errors consisted of selecting the correct items in the wrong order, this can effectively be treated as a two-choice task, so a score of between 8–15 out of 24 correct can be regarded as chance level performance (binomial theorem,
Analysis proceeded in four stages: first, exploratory comparisons were made between the children who did and did not score close to ceiling in session 1, to identify characteristics of those with adequate comprehension. Second, for those who scored below ceiling, we compared rates of learning for unique and repeated items in the SLI and grammar-matched groups. Third, we examined scores on the post-test to check for generalization of training effects. Finally, we conducted further analyses to examine memory predictors of learning.
SLI-T and grammar-matched groups subdivided according to whether above or below ceiling in Session 1. Means with different superscripts differ significantly from one another on post hoc Sidak test,
SLI-T, |
SLI-T, |
Grammar-matched, |
Grammar-matched, |
|
---|---|---|---|---|
Age (yr) | 8.4 (1.16) | 8.8 (1.48) | 5.5 (0.78) | 6.2 (0.80) |
RCPM SS | 100.9 (12.52) | 105.2 (14.18) | 103.4 (8.45) | 107.6 (8.24) |
TROG-E raw blocks | 7.7 (2.97) | 9.1 (4.96) | 8.9 (3.21) | 10.8 (2.73) |
TROG-E SS | 70.7 (11.26) | 76.2 (17.13) | 103.3 (13.59) | 101.1 (14.95) |
BPVSII raw | 70.4 (14.31) | 68.5 (21.66) | 64.1 (12.59) | 71.1 (12.85) |
BPVSII SS | 90.5 (10.48) | 84.5 (15.54) | 109.3 (9.97) | 106.8 (9.09) |
NEPSY nonwords raw | 19.9 (7.47) | 25.0 (9.92) | 25.0 (8.59) | 29.8 (5.19) |
NEPSY nonwords SS | 81.0 (15.26) | 87.7 (18.33) | 100.9 (16.35) | 108.8 (12.27) |
ERRNI Comprehension raw | 7.7 (3.54) | 8.9 (2.9) | 7.5 (3.18) | 9.6 (2.71) |
ERRNI Comprehension SS | 82.2 (15.29) | 82.9 (15.64) | 102.5 (13.64) | 103.7 (16.39) |
ACE Naming raw | 10.0 (4.88) | 10.4 (3.75) | 9.8 (2.67) | 10.4 (3.12) |
ACE Naming SS | 83.7 (14.08) | 79.2 (12.22) | 102.5 (7.07) | 99.2 (9.00) |
ACE Syntax raw | 14.0 (5.40) | 16.1 (4.84) | 15.9 (6.27) | 20.2 (4.65) |
ACE Syntax SS | 80.7 (9.61) | 80.8 (9.09) | 102.8 (13.90) | 109.2 (14.28) |
3.1 (1.46) | 3.4 (1.33) | 0.4 (0.50) | 0.3 (0.45) | |
Word Span | 3.7 (1.05) | 4.3 (1.03) | 4.0 (0.89) | 4.3 (0.87) |
At the suggestion of a reviewer, we also divided the SLI-T children according to whether or not they scored at chance on session 1. We found that the learning profiles of these two subgroups are very similar to that based on whole group data.
(A) and (B) show mean items correct, and (C) and (D) show response times. Grammar-matched children in (A) and (C) and SLI-T in (B) and (D). Error bars show standard errors.
Data were analysed using a mixed-design ANOVA, with session (1 to 4) and item type (unique vs. repeated) as repeated measures, and group as between-subjects factor. One potential problem with multiway analysis of variance is that the probability of a Type 1 error is increased because separate null hypotheses are tested (e.g., two main effects and one interaction in a 2-way design) (
Looking at main effects first, the overall effect of group was not significant:
We next considered interactions of group with the within-subjects factors. We were specifically interested in the interaction between item type and group, which was non-significant,
Finally, the difference between unique and repeated items was compared for each group at each session using a matched pairs
Data on response times are shown in
The overall effect of group was marginal, reflecting a trend for faster responses by the SLI group compared to grammar-matched controls,
Data gathered after the training allowed us to see whether there had been any generalization from the training. The first source of data was from session 5, in which children were given the same training game, but with the preposition pair they had not been trained on (above-below or before-after). Although the prepositions were new, half the items contained the noun sequences that had previously been used in repeated items. Data were missing for this post-test for one child in the SLI-T group. Means and standard errors for accuracy and response times are shown in
The expectation is that performance on a new preposition pair should be no better than was seen in session 1, unless the child had learned some general strategy for performing this kind of comprehension task. The mean scores for session 5 were higher than that of session 1 for both groups, the mean (both item types combined) increasing from 17.25 (SD = 3.49) to 19.13 (SD = 4.35) in the grammar-matched group and from 14.71 (SD = 3.67) to 18.50 (SD = 4.29) in the SLI-T group. This effect was examined using a mixed-design ANOVA comparing session 1 and session 5 scores for Unique and Repeated items in SLI and grammar-matched groups. The main effect of session was significant,
The marginal interaction between session and item type was unexpected, because all sentences were novel in session 5 (i.e., they used a different preposition). Because of its potential clinical significance, we explored this effect further, using single-sample
A parallel analysis was conducted on log response times, comparing session 1 and session 5. There was a marginal overall effect of group, reflecting faster response times by the SLI group,
Another way of assessing the impact of training is to consider changes in overall performance on TROG-E, where different parallel forms had been administered before and after the training.
Groups subdivided according to initial level of performance on Session 1 (Age-matched, grammar-matched and SLI-T groups) or TROG-E (SLI-U group).
Group | Initial performance |
Trained? |
|
Mean (SD) pre | Mean (SD) post |
|
|
---|---|---|---|---|---|---|---|
Age-matched control | At ceiling | Yes | 19 | 14.79 (2.44) | 15.32 (2.36) | 0.87 | .399 |
Grammar-matched control | Below ceiling | Yes | 16 | 8.94 (3.21) | 9.94 (4.02) | 1.71 | .108 |
Grammar-matched control | At ceiling | Yes | 12 | 10.83 (2.73) | 11.42 (3.68) | 0.71 | .492 |
SLI-T | Below ceiling | Yes | 15 | 7.67 (2.97) | 7.13 (4.45) | −0.71 | .488 |
SLI-T | At ceiling | Yes | 13 | 9.08 (4.96) | 9.85 (5.14) | 0.70 | .495 |
SLI-U | Low TROG-E | No | 8 | 6.54 (2.77) | 7.25 (3.62) | 0.58 | .576 |
SLI-U | High TROG-E | No | 12 | 11.83 (1.40) | 12.25 (2.95) | 0.57 | .581 |
To further assess impact of training on TROG-E performance, for children trained on the “above-below” contrast, performance on the four reversible above/below items of TROG-E was compared before and after training. For the eight grammar-matched children who were trained on “above” and “below”, the number of correct responses on TROG-E before and after training was 2.75 and 3.25 out of 4 test items, a non-significant improvement (
Two multiple regression analyses were run using the full sample of grammar-matched controls and SLI children. These examined the relationship between comprehension and memory measures. First, age and raw scores on the two memory measures, nonword repetition and word span, were used as predictors of comprehension performance on session 1. Results are shown in
Using the full sample has the benefit of enhancing statistical power, but the disadvantage that results are somewhat skewed by inclusion of cases who scored close to ceiling levels. The analysis was therefore repeated just using the 31 children scoring below 90% on session 1 (
(A) All children from grammar-matched and SLI groups (
Zero-order |
|||||
---|---|---|---|---|---|
Variable | Age (yr) | Nonword repetition | Word span | Session 1 score |
|
|
|||||
Age (yr) | −.20 | .16 | .13 | .16 | |
Nonword rep. | .32 |
.34 |
.29 |
||
Word span | .35 |
.23 | |||
Mean | 7.19 | 24.6 | 4.03 | 19.1 | |
SD | 1.78 | 8.54 | 0.97 | 4.62 | |
|
|||||
Age (yr) | −.29 | .04 | −.13 | −.11 | |
Nonword rep. | .27 | .27 | .15 | ||
Word span | .35 | .32 | |||
Mean | 6.88 | 22.6 | 3.83 | 15.9 | |
SD | 1.74 | 8.34 | 0.97 | 3.84 |
Next, number of items correct on the last day of training (session 4) was taken as the dependent measure, with number correct in session 1, age, nonword repetition and word span as predictors. This analysis was run both with all 56 children (including those who performed near ceiling), and with the smaller sample of 31 children who performed below ceiling on session 1.
Results for the full sample are shown in
(A) All children from grammar-matched and SLI groups (
Zero-order |
|||||
---|---|---|---|---|---|
Variable | Nonword repetition | Word span | Session 1 score | Session 4 score |
|
|
|||||
Age (yr) | −.20 | .16 | .14 | .19 | .08 |
Nonword rep. | .32 |
.33 |
.16 | −.06 | |
Word span | .35 |
.45 |
.33 |
||
Score day 1 | .46 |
.35 |
|||
|
|||||
Age (yr) | −.29 | .04 | −.14 | .21 | .24 |
Nonword rep. | .27 | .27 | 0 | −.14 | |
Word span | .35 | .43 |
.34 | ||
Score day 1 | .42 |
.38 |
|||
Our study found that for many children with SLI, language comprehension ability measured with TROG-E lags well behind their peer group. In addition, just over half of these children exhibited difficulty with even short and syntactically simple sentences that their age-matched peer group found extremely easy. Overall, the comprehension level of children with SLI aged 8 to 9 years was comparable to typically-developing children who were around three years younger.
Children with SLI who showed poor comprehension of spatial prepositions on the initial training session had significantly higher scores on the repeated vs. the unique sentences in the second and third sessions, though they had found these sentence types equally difficult in session 1. In session 4 their performance on unique items rose to be as good as for repeated items. Together, these findings suggest a dynamic pattern of learning of syntactic constructions in these children with SLI: they first rely on repeated items in order to gain basic understanding of the meaning of specific sentence exemplars. Once they have gained some understanding of the meaning of individual sentences using these prepositions, then a more general meaning can be reliably extracted and generalised to new contexts. In contrast, no difference between repeated and unique training sentences was seen for younger grammar-matched controls until the final training session, when, unexpectedly, performance on unique items deteriorated.
Note that the pattern of performance that was observed in the children with SLI was the opposite to what might have been predicted from artificial grammar learning studies, where it has been found that variability of items facilitates learning of rule-based sequences (
In our study, however, high variability (i.e., unique items) did not benefit typically-developing children either. Both groups responded faster to repeated items. To explain this, we need to consider the task demands of learning an artificial grammar compared with those of our training task. In a classic artificial grammar learning task, all that is required is for a person to demonstrate awareness of an underlying sequential structure. In our tasks, children had to assign meaning to a sequential structure. For this, we suggest, repetition may facilitate performance. Comprehension of a simple reversible sentence involves building a representation of its syntactic and semantic structure that distinguishes between the subject of the sentence and the object of the preposition, and this involves holding verbal material in a memory buffer as the structural representation is generated. There are three ways in which this process may be facilitated when items are repeated. First, it is possible that children do not analyse the sentence structure, but simply rote learn the meaning of the whole sentence: e.g., the child may learn to associate the sentence: “the apple is above the horse”, with a particular spatial configuration, but not form an abstract representation of the meaning of “above”. This explanation corresponds to the “rote learning” account of SLI proposed by
Children who did poorly in comprehending short reversible sentences were similar to other children in many regards but they tended to have low scores on a nonword repetition test of phonological short-term memory and on a syntax formulation task. Their comprehension scores at the end of training were significantly predicted by a measure of memory span for words, even after taking into account their initial scores in session 1. Thus ability to remember a series of unconnected words predicted how far children benefit from repeated exposure to short reversible sentences containing a spatial preposition.
At first glance, our results may seem to contradict findings by
Overall, our results are in line with the previous findings that poor short-term memory plays a role in comprehension problems in children with SLI (
Results from this study offer both negative and positive messages to those concerned with intervention for comprehension problems. On the negative side, even when training was restricted to a single pair of antonyms, performance still fell below the level of age-matched controls, who made very few errors on these items. Also, neither children with SLI nor their grammar-matched controls showed improvement from pre- to post-test on TROG-E. A lack of transfer would be consistent with the idea that children tend to learn sentence meanings by rote: such knowledge would not generalize to new meanings. Nevertheless, this latter result needs to be treated cautiously, given that there were only four relevant items in TROG-E; it is possible that a longer and more sensitive test of preposition comprehension may have revealed some improvement.It is noted that successful transfer to TROG items has been reported by
Results from this study also offer positive messages to those concerned with intervention for comprehension problems. First of all, children with SLI benefited from multiple exposures to the same sentence exemplars during the initial phase of intervention. This might compensate for their limitations in memory and provide the children with an opportunity to build more solid representations of the first few sentences they encountered before the general meaning of the related sentences could be detected and learned. The ultimate goal is for children to learn the general meanings of constructions such that their comprehension is not restricted to sentences they have heard before; the current results suggest that when a certain level of comprehension accuracy of repeated sentences is reached, different sentence exemplars of the same constructions could be emphasized in order to facilitate learning of general meanings. Furthermore, we found an effect of generalization from the trained prepositions to a new pair of untrained prepositions, and this effect was not limited to the younger grammar-matched children but was also observed in the children with SLI. This finding reveals positive effects of receptive grammar training that could generalize to comprehension of sentences using similar syntactic constructions. It is unclear whether or not the generalization depends on a particular training format, but our results suggest that children with SLI may do best if given repeated exposure to specific sentences with a given sentence frame, perhaps because this facilitates rapid interpretation of the nouns in the sentence, freeing up mental capacity for sentence interpretation. Accordingly, generalisation to other sentences using the same syntactic frame may be more effective if preceded by rote learning of specific items. In addition, when training children’s grammatical skills, it might be beneficial to restrict consideration to a small, highly-familiar vocabulary, and not to attempt generalization to new vocabulary until performance on a restricted set of items is highly reliable. It would be of interest to conduct future studies examining how different presentations of training stimuli affect learning of a given sentence frame and subsequent generalization to a similar but different sentence type.
In conclusion, a high proportion of children with SLI had difficulty understanding reversible sentences even when these are short and syntactically simple. It is not plausible that these difficulties were due to impaired auditory perception, because the words used in the sentences were simple, discriminable and illustrated with pictures. Although some children performed at chance, others scored above chance, indicating they had some grasp of the meanings of prepositions, but they failed to perform consistently over repeated trials. The pattern of results suggested that their comprehension failures may be related to weaknesses in verbal short-term memory. Our study suggested that performance could be improved by computerised training of sentence comprehension, incorporating rote learning of specific items in a sentence frame. Nevertheless, the training regime used in this study was not sufficient to bring performance up to the level of an age-matched peer group.
Items marked Unique contain new nouns, whereas those marked Repeated have exactly the same words on each presentation. The four training sessions use the same preposition pair, whereas at post-test, the other preposition pair is presented.
Sample sentences for Session 1:
Repeated A: “The apple is above the horse”; Repeated B: “The chair is above the rabbit”; Repeated C: “The cow is below the soap”; Repeated D: “The ball is below the turtle”; Unique 1: “The baby is below the owl”; Unique 2: “The hedgehog is below the crown”; Unique 3: “The gate is below the butterfly”; Unique 4: “The orange is above the table”; Unique 5: “The book is above the lamp”; Unique 6: “The flag is below the bear”; Unique 7: “The zebra is below the cake”; Unique 8: “The bus is below the eye”; Unique 9: “The bike is above the fox”; Unique 10: “The bell is above the kite”; Unique 11: “The cushion is above the wall”; Unique 12: “The cup is below the door”.
We thank Annie Brookman, Nikki Gratton, Mervyn Hardiman, Anneka Holden, Georgina Holt and Eleanor Paine and for their invaluable assistance in data collection.
Dorothy V.M. Bishop is the author of three of the standardized assessments used to assess children in this study (CCC-2, TROG-E and ERRNI). These are published by Pearson and the royalties are paid direct to a charity.
The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers):
University of Oxford Medical Sciences Division Interdivisional Research Ethics Committee: Reference number MSD/IDREC/2009/28.
The following information was supplied regarding the deposition of related data:
We plan to deposit the raw data in a database on the Open Science Framework once we have published one further study on a different set of variables from the same participants.