Professional Documents
Culture Documents
1103haebig Saffran Ellis Weismer 2017
1103haebig Saffran Ellis Weismer 2017
12734
Background: Word learning is an important component of language development that influences child outcomes
across multiple domains. Despite the importance of word knowledge, word-learning mechanisms are poorly
understood in children with specific language impairment (SLI) and children with autism spectrum disorder (ASD).
This study examined underlying mechanisms of word learning, specifically, statistical learning and fast-mapping, in
school-aged children with typical and atypical development. Methods: Statistical learning was assessed through a
word segmentation task and fast-mapping was examined in an object-label association task. We also examined
children’s ability to map meaning onto newly segmented words in a third task that combined exposure to an artificial
language and a fast-mapping task. Results: Children with SLI had poorer performance on the word segmentation
and fast-mapping tasks relative to the typically developing and ASD groups, who did not differ from one another.
However, when children with SLI were exposed to an artificial language with phonemes used in the subsequent fast-
mapping task, they successfully learned more words than in the isolated fast-mapping task. There was some
evidence that word segmentation abilities are associated with word learning in school-aged children with typical
development and ASD, but not SLI. Follow-up analyses also examined performance in children with ASD who did and
did not have a language impairment. Children with ASD with language impairment evidenced intact statistical
learning abilities, but subtle weaknesses in fast-mapping abilities. Conclusions: As the Procedural Deficit
Hypothesis (PDH) predicts, children with SLI have impairments in statistical learning. However, children with SLI
also have impairments in fast-mapping. Nonetheless, they are able to take advantage of additional phonological
exposure to boost subsequent word-learning performance. In contrast to the PDH, children with ASD appear to have
intact statistical learning, regardless of language status; however, fast-mapping abilities differ according to broader
language skills. Keywords: Specific language impairment; autism spectrum disorder.
learning in word learning, Graf Estes and colleagues chronological age-matched peers (Kan & Windsor,
examined how infants establish connections 2010). Children with SLI have deficits in encoding
between sounds and meanings (Graf Estes, Evans, phonological and semantic information about words
Alibali, & Saffran, 2007). Specifically, they asked (Alt & Plante, 2006). In addition, they have been
whether newly segmented sound sequences – dis- found to demonstrate deficits in word-learning
covered via statistical learning processes – were mechanisms like shape bias (Collisson, Grela,
more easily mapped onto a referent than sound Spaulding, Rueckl, & Magnuson, 2015). As such,
sequences that had not been previously segmented. lexical-semantic weaknesses exist in both breadth
They found that infants were able to map meanings and depth of word knowledge (Sheng & McGregor,
onto high-TP words (TP = 1.0), but not the nonwords 2010). Beyond language, children with SLI have
(TP = 0.0) or part-words (TP = 0.33), indicating that deficits in working memory (WM) and processing
the output of statistical learning can support word speed (Gathercole & Baddeley, 1990; Leonard et al.,
learning in infants. There has been a call for research 2007).
to explain how the output of statistical learning Evans et al. (2009) also identified deficits in
contributes to the process of linking knowledge of statistical learning, with children with SLI requiring
sound patterns to meaning in children with ASD and additional exposure to an artificial language before
SLI (Arunachalam & Luyster, 2016; Nation, 2014). demonstrating above-chance learning of word
Given the high comorbidity of language impair- boundaries. Under the PDH, statistical learning is
ments in ASD, it is particularly important to examine crucial for learning rule-based features of language
the link between statistical learning and fast-map- such as grammar and phonology (Ullman & Pier-
ping in children with ASD and SLI. The Procedural pont, 2005; but see Hsu & Bishop, 2014 for a
Deficit Hypothesis (PDH) has been suggested to nuanced interpretation of findings). Indeed, Hede-
account for language impairments in both groups nius et al. (2011) demonstrated that procedural
(Ullman & Pierpont, 2005; Walenski, Tager-Flus- learning abilities are associated with grammatical
berg, & Ullman, 2006); however, others suggest that deficits in children with SLI. However, Evans et al.
language deficits in children with ASD and SLI stem (2009) also identified an association between seg-
from distinct or partially nonoverlapping mecha- mentation skills and vocabulary knowledge.
nisms (Boucher, 2012; Williams, Botting, & Bou- Autism spectrum disorder (ASD) is characterized
cher, 2008). Importantly, procedural learning and by core deficits in social communication and by
statistical learning both fit under the umbrella of restricted interests and repetitive behaviors (Ameri-
implicit learning. Implicit learning is the learning of can Psychiatric Association, 2013). Pragmatic defi-
information in an incidental form, without aware- cits are a defining feature of ASD; however, some
ness of the newly formed information. Although the children also have structural language deficits,
declarative/procedural model distinctly links declar- including grammatical and lexical deficits (McGregor
ative abilities with vocabulary and procedural abil- et al., 2012). Specifically, some children have
ities with grammar within the language domain, it reduced breadth of word knowledge (Kjelgaard &
has been suggested that aspects of word learning Tager-Flusberg, 2001) and many children with ASD
also may be supported by the implicit system (Evans, have limited depth of word knowledge (McGregor
Saffran, & Robe-Torres, 2009). A better understand- et al., 2012). Although children with ASD demon-
ing of the nature of procedural and declarative strate use of learning mechanisms such as cross-
learning in children with ASD and SLI is necessary. situational learning (McGregor, Rost, Arenas, Farris-
As will be outlined below, language deficits are Trimble, & Stiles, 2013) and mutual exclusivity (de
apparent in children with ASD and SLI (Durkin et al., Marchena, Eigsti, Worek, Ono, & Snedeker, 2011),
2012; Kjelgaard & Tager-Flusberg, 2001). Addition- they have been found to not use others, like shape
ally, both groups share risk factors spanning demo- bias (Tek, Jaffery, Fein, & Naigles, 2008). Further-
graphic, behavioral, and neural (Tager-Flusberg, more, studies have indicated that when children
2016). It is thus important to determine whether with ASD learn new words, they encode new phono-
similar learning mechanisms account for language logical forms of words, but less-readily integrate
deficits within the two populations (Obeid, Brooks, phonological information (Henderson, Powell, Gar-
Powers, Gillespie-Lynch, & Lum, 2016; Rice, 2016). eth Gaskell, & Norbury, 2014) and/or less-readily
consolidate new word knowledge (Norbury, Griffiths,
& Nation, 2010) over extended periods of time.
Specific language impairment and autism spectrum
Given previous suggestions that the PDH may also
disorders
explain deficits in children with ASD (Ruffman,
Children with SLI have a primary deficit in language Taumoepeau, & Perkins, 2012; Walenski et al.,
without hearing impairment, intellectual disability, 2006), segmentation tasks also have been used to
or neurological impairments. Although grammatical assess statistical learning in individuals with ASD.
skills are particularly poor in children with SLI, other Scott-Van Zeeland et al. (2010) found that children
abilities often are impaired. Notably, children with with ASD failed to demonstrate neural evidence of
SLI have poorer word-learning abilities relative to statistical learning. However, Mayo and Eigsti (2012)
found that children with ASD evidenced intact sta- statistical learning abilities relative to the SLI group
tistical learning abilities, but that they were not due to surface-level attention biases. We also predict
strongly associated with language abilities. Attention that both language impaired subgroups would have
to phonological information and failure to integrate deficits in fast-mapping. Lastly, we expect that
semantic information to support language develop- difficulties mapping meaning to newly segmented
ment aligns with theories that suggest that children words would be most apparent in the SLI group,
with ASD may have a featural/surface-biased infor- relative to the subgroup with ALI. See Table 1 for an
mation-processing style that may enhance process- outline of the hypotheses.
ing of a single stimulus cue instead of focusing on
multiple cues (J€ arvinen-Pasley, Wallace, Ramus,
Happ" e, & Heaton, 2008). Methods
Given the deficits in consolidating linguistic infor- Seventy-four school-aged children participated in the current
mation observed in children with ASD and weak study (TD n = 26, ASD n = 25, SLI n = 23). Given the importance
of vocabulary knowledge in academic and social success in the
phonological representations or phonological WM school-aged years, we included children between the ages of 8
observed in children with SLI, it is unclear whether and 12 years. This age range also was similar to previous studies
these children are able to map meaning onto newly examining statistical learning abilities in children with ASD and
segmented words. Furthermore, the breakdown in SLI, which facilitates study comparisons. Participants lived in
mapping meaning to sound patterns may stem from the greater Madison metropolitan area (USA) and were recruited
from a larger two-year longitudinal study examining the rela-
different causes. Thus, a systematic comparison tionship between language and executive functions conducted at
between children with ASD and SLI is needed to the University of Wisconsin-Madison.
improve our understanding of the mechanisms that Standardized assessments were administered to measure
drive language impairment and to determine language and cognitive skills in the first year of the larger
whether overlapping deficits contribute to the lan- study. Nonverbal cognitive abilities were assessed using the
Perceptual Reasoning Index of the Wechsler Intelligence Scale
guage deficits seen in both populations. Although for Children-IV (WISC-IV; Wechsler, 2003). The Peabody Pic-
researchers have begun to assess statistical learning ture Vocabulary Test-4 (PPVT-4; Dunn & Dunn, 2007) mea-
in children with ASD and SLI separately, no study sured receptive vocabulary. The Clinical Evaluation of
has evaluated the output of statistical learning – Language Fundamentals-4 (CELF-4; Semel, Wiig, & Secord,
combining word segmentation and word learning – in 2003) was used to assess receptive and expressive lexical and
grammatical skills. Children in all groups had WISC-IV stan-
either of these populations. Therefore, in the current dard scores above 85 (with the exception of two children with
study, we asked: Do children with ASD or SLI have ASD who had scores of 79), and were monolingual American
deficits in (a) statistical learning, (b) fast-mapping, English speakers. Children with typical development (TD)
and/or (c) mapping of meaning onto newly seg- obtained standard scores that were <1 standard deviation
mented words, compared to typically developing below the mean on the CELF-4 and had no history of special
education services. Children with SLI scored at least 1.25
children? In addition, we asked whether language standard deviations below the mean on one or more of the
abilities more strongly influence statistical learning composite CELF-4 measures or demonstrated at least a 14-
and fast-mapping than diagnostic classification. point gap between one of the CELF-4 composite measures and
If atypical language development in ASD and SLI is nonverbal cognition, and had a history of or were currently
derived from underlying impairments in statistical receiving language therapy. Children in the ASD group had a
documented community diagnosis of an autism spectrum
learning as proposed by the PDH, children in both disorder. Diagnoses were confirmed by an experienced psy-
diagnostic groups should demonstrate poor statisti- chologist; children with ASD had a score of 25 or higher on the
cal learning. However, if children with ASD present Childhood Autism Rating Scale-2 (CARS-2; Schopler, Van
with a surface-biased information-processing style, Bourgondien, Wellman, & Love, 2010). To rule out ASD in
we would expect them to demonstrate intact statis- the other groups, children with TD and SLI were required to
score below the core autism cutoff score for on the Social
tical learning abilities. In addition, given word-
learning delays and difficulties with word meanings
(McGregor et al., 2012), we hypothesized that chil- Table 1 Outline of hypotheses
dren with ASD who also had a language impairment
would demonstrate poor fast-mapping abilities. We Task Hypotheses
also hypothesized that children with SLI would have Word a
TD & ASD > SLI TD = ALN;
poor fast-mapping abilities, given their known segmentation ALI > SLI; ALN = ALIa
encoding deficits. Additionally, in reference to our task
third research question, if children were able to Fast- TD & ASD > SLI TD = ALN;
mapping ALI = SLI; ALN > ALIb
segment high-TP words, we hypothesize that high-TP task
words would be more readily mapped onto referents Combination TD & ASD > SLI; TD = ALN;
than low-TP words. When comparing performance task Word > Nonword ALI > SLI; ALN = ALI
based on language impairment rather than diagnos-
tic classification, we expect children with language TD, typical development; ASD, autism spectrum disorder; ALN,
ASD with normal language; ALI, ASD with language impair-
impairment to have mixed profiles of statistical ment; SLI, specific language impairment.
learning abilities. We predict that children with a
Following work by J€ arvinen-Pasley et al. (2008).
b
ASD and language impairment (ALI) will have better Following work by McGregor et al. (2012).
Communication Questionnaire (SCQ; Rutter, Bailey, & Lord, the 2AFC test, the first stimulus played while the number 1
2003). Each child was required to complete the three exper- appeared on the left side of the computer screen and as the
imental tasks and meet inclusionary criteria listed above to be second stimulus played the number 2 appeared on the right
included in the current study. Children were matched on side of the screen. After both items had been presented, a
chronological age (p = .73) and WISC-IV standard scores question mark appeared in the center of the screen to prompt
(p = .72); see Table 2 for participant characteristics. the child to select the stimulus that sounded like the ‘Martian
language’ by pressing the button-box key with either the
number 1 or 2 above it.
Ethical considerations
Parents provided informed written consent and children pro- Fast-mapping. A fast-mapping task assessed word-learn-
vided verbal assent. The study was approved by the univer- ing abilities. Four novel words (/timo/,/bole/,/deno/,/pɑdu/)
sity’s Institutional Review Board. were paired to two-dimensional novel objects with a solid color.
During teaching, each novel object was displayed and its
corresponding auditory label was presented within a 2 s
Experimental procedures teaching trial. Each object-label pair was presented individu-
ally three times in a nonsequential pseudorandomized order.
Participants completed three experimental tasks over two In the test phase, two objects appeared at opposite sides of the
visits. Task order was counterbalanced across participants. bottom of the screen. An auditory cue directed the child’s
Half of the participants completed the segmentation and fast- attention to one of the two images (e.g. ‘Find the __.’ ‘Where’s
mapping tasks during visit 1 and the combination task during the __?’). Each object-label association was tested in four test
visit 2, and the other half of the participants completed the trials, yielding a total of 16 test trials. The task was presented
combination task during visit 1 and the segmentation and fast- using Matlab and lasted 4.5 min. A video camera recorded the
mapping tasks during visit 2. child’s face. To derive the eye-gaze data and examine learning
from the eye-gaze test phase, trained coders analyzed the
Statistical learning. A word segmentation task was used videos offline using Looking-While-Listening (LWL) coding
to assess statistical learning. Children were assigned to one of procedures (Fernald, Zangl, & Marchman, 2008). After the
two artificial languages (Graf Estes et al., 2007; Experiment 1). eye-gaze test phase completed, a second test phase ensued.
Each artificial language consisted of four disyllabic novel The examiner showed the child a piece of paper with one of the
words (Language A:/time/,/mɑno/,/dobu/,/pigɑ/; Language four object images in each corner and asked the child to point
B:/nome/,/mɑti/,/gɑbu/,/pido/) that were repeated in ran- to the object that she named (e.g. ‘Find the ___.’ ‘Where’s the
dom order without pauses or other acoustic cues to word ___?’). The examiner recorded in writing the child’s response for
boundaries. A female American English speaker recorded each each of the four labels.
syllable with all possible coarticulatory contexts. The syllables
were later concatenated to create the artificial language Mapping meaning to sounds. A combination task (Graf
containing the appropriate coarticulatory versions for each Estes et al., 2007) was designed to investigate children’s ability
syllable at a rate of approximately 200 syllables/minute. The to map meaning onto newly segmented words. The task
only reliable cue for segmentation of words was the statistical included an artificial language presentation phase, word-
structure of the artificial language (within-word TP = 1.0, learning phase, and test phase. A new artificial language –
across-word TP = 0.33). While listening, children watched a different from the one used in the segmentation task – was
silent nature slide-show for 4.75 min, presented using E-Prime created with the same statistical structure used in the
software (Psychology Software Tools, Inc., 2012). Children segmentation task, using a new set of syllables (Language
were informed that they were going to listen to a ‘Martian 2A:/pɑtu/,/midɑ/,/gine/,/bodu/; Language 2B:/tune/,/
language’ and their job was to sit and listen. After the exposure pɑgi/,/mipo/,/dɑdu/). Children were randomly assigned to
phase, children completed a practice phase with commonly one of the two artificial languages. The same methods were
known disyllabic real words and disyllabic novel words used to create the artificial language for the combination task
following American English phonotactics. Then, a 32-item as the segmentation task. The labels for the word-learning
two-alternative forced-choice (2AFC; word vs. nonword) test phase of the combination task came from artificial language
phase was administered. During the 2AFC, children were that was presented in the artificial language exposure phase of
presented with two auditory stimuli – a word and nonword (a the combination task. Two of the four words in the artificial
pair of syllables that never occurred together in the speech language, with TPs of 1.0, served as the labels in the subse-
stream; TP = 0.0) from the artificial language. Items that were quent fast-mapping phase. The other labels consisted of two
words for participants who heard Language A were nonwords nonwords that were comprised of nonsequential syllables from
for participants who heard Language B, and vice versa. During the artificial language, yielding a TP of 0. Items that were words
Chronological Age 10.38 1.28 10.13 1.36 10.28 1.18 TD = ASD = SLI
WISC-4a 104.50 9.06 103.65 16.63 102.35 10.88 TD = ASD = SLI
CELF-CLb 103.69 12.37 88.35 20.21 81.91 14.23 TD > ASD* = SLIc
PPVT-4d 110.96 17.02 106.27 18.90 94.22 13.77 TD = ASD > SLI*
TD, typical development; ASD, autism spectrum disorder; ALN, ASD with normal language; ALI, ASD with language impairment;
SLI, specific language impairment.
a
Perceptual reasoning index standard score.
b
Core language standard score.
c
ASD versus SLI p = .101; *p < .05
d
Receptive vocabulary standard score.
The random slope GCA model yielded a significant nonword trials. However, there was a three-way
effect of intercept (Estimate = 0.63; SE = .12; interaction of cubic time, condition, and SLI group
t = 5.45) and linear time (Estimate = 2.06; SE = .52; (Estimate = !1.41; SE = .64; t = !2.22), indicating
t = 3.96), indicating that children increased their that children with SLI had a different trajectory of
looks to the target during the test window and spent looking in the word condition, with looks to the target
more time looking at the target than the distractor. object, then away, and then back to the target. See
Additionally, TD and ASD groups looked to the target Appendix S2 for the full model results. As with the
image significantly more than the SLI group (Esti- eye-gaze data, the pointing data revealed that there
mate = !0.41; SE = .17; t = !2.43, Esti- were no significant group differences in pointing
mate = !0.37; SE = .17; t = !2.16, respectively). accuracy for the combination task, ps > .10, and
None of the interactions between group and time that all groups had above-chance pointing accuracy.
terms were significant. Performance did not differ Mean pointing scores on the combination task were
according to visit order. Figure 2 depicts the average 3.2 (SD = 1.4) for the TD group, 2.9 (SD = 1.5) for the
group looking behavior. ASD group, and 2.6 (SD = 1.4) for the SLI group.
Pointing data from the fast-mapping test phase When testing for visit order effects, we found that
yielded accuracy scores of 2.8 (SD = 1.4) for the TD children had higher proportion of looks to the target
group, 2.8 (SD = 1.4) for the ASD group, and 2.0 image and higher pointing accuracy when complet-
(SD = 1.0) for the SLI group, out of a possible total of ing the combination task during the second visit;
four points. The pointing data were evaluated using however, after controlling for visit order, we contin-
linear regression (which is relatively robust to viola- ued to see similar performance across the groups
tions of statistical assumptions) and nonparametric (i.e. a lack of a group effect). Figure 3 depicts group
analyses. Since results were equivalent only regres- performance.
sion findings are presented. Analysis of the pointing Our fourth research question asked whether sta-
data from the fast-mapping task revealed similar tistical learning and fast-mapping abilities were
findings to the eye-gaze data; although all groups more strongly related to child language characteris-
performed significantly above chance, the TD and tics than diagnostic group classification. As an initial
ASD groups had significantly higher pointing accu- step, we examined correlations within each group
racy relative to the SLI group (Estimate = !1.30; between performance on the experimental tasks and
SE = .57; t = !2.28, Estimate = !1.38; SE = .59; standardized language assessments (see Table 3).
t = !2.36, respectively), but the TD and ASD groups Only the ASD group exhibited significant correla-
did not differ (Estimate = 0.08; SE = .57; t = 0.89). tions between performance on the experimental
The third research question asked whether chil- measures and language assessments. Given the
dren are able to map meaning onto newly segmented heterogeneity in language abilities in children with
words. In the model of the eye-gaze data from the ASD, we also examined correlations when the TD
combination task, the intercept and linear time were and SLI groups were combined into a non-ASD
significant (Estimate = 1.68; SE = .14; t = 4.81, Esti- group, to have similar ranges in language assess-
mate = 1.86; SE = .65; t = 2.87, respectively), indi- ment scores. Despite the increased range of perfor-
cating that children increased their looks to the mance with this non-ASD group, the only significant
target image and spent more time looking at the correlation was between fast-mapping performance
target than the distractor across the test window. and receptive vocabulary on the PPVT-4 (r = .405).
There were no significant main effects of group or Nevertheless, the somewhat larger amount of vari-
condition indicating that, overall, the groups spent a ability in the ASD group may have influenced this
similar amount of time looking to the target and that pattern of correlations. It is noteworthy that nonver-
looking to the target was similar for word and bal cognition scores were not significantly correlated
Typical development
Segmentation Task .207 .418* !.072 !.172
Fast-Mapping Task !.235 .360† .016
Combination Task .172 .219
ASD
Segmentation Task .261 .445* .469* .466*
Fast-Mapping Task .444* .566* .507*
Combination Task .541* .581*
SLI
Segmentation Task .108 .051 .170 .165
Fast-Mapping Task .249 .229 .159
Combination Task .365† .113
ASD, autism spectrum disorder; SLI, specific language impairment; Clinical Evaluation of Language Fundamentals; PPVT-4,
Peabody Picture Vocabulary Test-4.
a
Percent accuracy.
b
Mean proportion of looks to the target.
c
Receptive vocabulary standard score.
d
Core language standard score.
†
p < .1, *p < .05.
with performance on the experimental tasks for any strongly influenced statistical learning and word
group (ps > .05). learning, relative to diagnostic group classification,
To further examine our fourth research question, we compared the groups within each of the language
we classified our participants based on language classifications on the experimental measures. In
abilities. Children with ASD were subcategorized addition, we compared performance between the
according to the CELF-4 standard scores. Thirteen ALN and ALI groups, who by the nature of their
children with ASD were classified as having a language groupings differed on PPVT and CELF-4
language impairment (ALI). These children had a standard scores, ps < .001.
CELF standard score that was at least 1.25 standard First, we examined statistical learning abilities in
deviations below the mean, with the exception of two children with LI. The mean scores on the segmenta-
children who had >1 standard deviation below the tion task in the LI groups were 61.1% (SD = 9.51%) for
mean and were receiving special services for language the ALI group and 53.7% (SD = 11.6%) for the SLI
or reading. The children in the language impairment group. A mixed-effect logistic regression revealed that
(LI) classification consisted of children in the SLI the ALI group performed significantly better on the
and ALI groups; they were matched on PPVT-4 stan- segmentation task than the SLI group (Esti-
dard scores (p = .84) and CELF-4 Core Language mate = !0.31; SE = .15; z = !2.11). As before, each
standard scores (p = .54). The children classified as group was separately tested against chance perfor-
having normal language abilities consisted of children mance (50%). The ALI group performed significantly
in the TD group and the ASD group with normal above chance (t[12] = 4.19, p = .001), but the SLI
language (ALN); they also were matched on PPVT-4 group did not (t[22] = 1.52, p = .14). Next, we exam-
(p = .41) and CELF-4 (p = .82) scores. In order to ined statistical learning abilities in children with
examine whether child language characteristics more normal language abilities. The mean scores were
66.9% (SD = 15.1%) for the ALN group and 62.7% the LI comparison model, there were significant
(SD = 14.3%) for the TD group. The mixed-effect effects of intercept (Estimate = 0.40; SE = .09;
logistic regression model revealed that performance t = 4.47), linear time (Estimate = 1.76; SE = .49;
on the segmentation task did not differ between the t = 3.62), and quadratic time (Estimate = !1.22;
ALN and TD groups (Estimate = !0.19; SE = .24; SE = .39; t = !3.10), indicating that the children
z = 0.41). Both groups performed significantly above with LI increased their looks to the target image,
chance (50%, ALN t[11] = 3.88, p = .003; TD t[25] plateaued in looks to the target image, and spent
= 4.55, p < .001). Lastly, a mixed-effect logistic more overall time looking at the target than the
regression model revealed that performance on the distractor across the test window. There were no
segmentation task did not differ between the ALN and significant main effects of group or condition indi-
ALI groups (Estimate = 0.28; SE = .23; z = 1.25). cating that, overall, the LI groups spent a similar
To determine whether fast-mapping differences amount of time looking to the target and that looking
were observed in our language groups, we analyzed to the target was similar for word and nonword trials.
eye-gaze data from the fast-mapping task. As before, However, there again was a three-way interaction of
main effects included linear, quadratic, and cubic cubic time, condition, and SLI group (Esti-
orthogonal time terms, group, and group by orthogo- mate = !2.80; SE = .73; t = !3.85), indicating that
nal time term interactions. Random effects allowed children with SLI had a different trajectory of looking
children to vary by the orthogonal time terms. The in the word condition, with looks to the target object,
random slope GCA model comparing fast-mapping then away, and then back to the target. In addition,
performance between the ALI and SLI groups yielded a pointing accuracy did not differ between the ALI and
significant effect of intercept (Estimate = 0.31; SLI groups, p = .72.
SE = .10; t = 3.15) and linear time (Estimate = 1.24; In the comparison model including the ALN and
SE = .45; t = 2.73), indicating that children with LI TD groups, there were significant effects of intercept
increased their looks to the target during the test (Estimate = 0.73; SE = .10; t = 7.15), linear time
window and spent more time looking at the target than (Estimate = 2.36; SE = .38; t = 6.22), and quadratic
the distractor. There were no significant group differ- time (Estimate = !0.65; SE = .26; t = !2.50), indi-
ences and none of the interactions between group and cating that the children with normal language abil-
time terms were significant. Comparison of the point- ities increased their looks to the target image,
ing data, however, revealed that the ALI group had plateaued in looks to the target image, and spent
significantly higher pointing accuracy than the SLI more overall time looking at the target than the
group, p < .05. The random slope GCA model compar- distractor across the test window. There were no
ing fast-mapping performance between the ALN and significant main effects of group, condition, or inter-
TD groups revealed a significant effect of intercept actions across group, condition, and time, indicating
(Estimate = 0.70; SE = .10; t = 6.93) and linear time that the ALN and TD groups spent a similar amount
(Estimate = 2.31; SE = .46; t = 5.01), indicating that of time looking to the target and that looking to the
children increased their looks to the target during the target was similar for word and nonword trials. There
test window and spent more time looking at the target also were no differences in pointing accuracy
than the distractor. The groups did not significantly between the ALN and TD groups, p = .65. Lastly,
differ and there were no interactions between group comparisons of the ALN and ALI groups on the eye-
and time terms. Furthermore, there were no signifi- gaze test phase of the combination task revealed
cant differences in pointing accuracy between the ALN significant effects of intercept (Estimate = 0.78;
and TD groups, p = .96. Lastly, the random slope GCA SE = .17; t = 4.63) and linear time (Estimate = 2.85;
model comparing fast-mapping performance between SE = .68; t = 4.17), indicating that the children with
the ALN and ALI groups revealed a significant effect of ASD increased their looks to the target image and
intercept (Estimate = 0.78; SE = .14; t = 5.57) and spent more overall time looking at the target than the
linear time (Estimate = 2.55; SE = .68; t = 3.78). distractor across the test window. Additionally, there
Although the ALN group spent slightly greater amount was a significant effect of group, with the ALN group
of time looking to the target image relative to the ALI looking more to the target image during the test
group, the groups did not significantly differ (Esti- window than the ALI group (Estimate = !0.54;
mate = !0.37; SE = .19; t = !1.92). There was a sig- SE = .23; t = !2.29). There were no significant main
nificant interaction of cubic time and group effects of condition, or interactions across condition
(Estimate = 1.09; SE = .55; t = 1.97), indicating that and time. However, there was a three-way interaction
the ALI group had a different trajectory of looking, with across cubic time, condition, and group (Esti-
looks to the target image, then away, and then back to mate = !1.73; SE = .86; t = !2.01), indicating that
the target. The ALN and ALI groups had similar the ALI group had a different trajectory of looking
pointing accuracy on the fast-mapping task, p = .98. behavior with looks to the target image in the word
Finally, we compared performance on the combi- condition, to looks away, and then looks back to the
nation task to examine whether the ability to map target toward the end of the trial. The pointing data
meaning onto newly segmented words differed revealed that the ALN group had marginally higher
among children with similar language abilities. In accuracy relative to the ALI group, p = .08.
representations of new words during the teaching easily than nonwords in the combination task. How-
phase. It is also possible that children with SLI have ever, as indicated by the interaction across the SLI
deficits in declarative memory, contrary to the PDH group, condition, and cubic time, children with SLI
claims. Lum, Ullman, and Conti-Ramsden (2015) shifted from the target to the distractor and then back
recently suggested that declarative memory is only to the target in the word condition. Additionally,
impaired in children with SLI if children have con- despite the lack of a main effect of condition, the
comitant WM deficits. Phonological WM and nonver- correlation analyses revealed a significant association
bal WM skills were not measured in the current study; between performance on the segmentation task and
therefore, this claim cannot be directly addressed. combination task in the TD and ASD groups. The
However, future studies should examine cognitive current study was the first to examine the link between
predictors of fast-mapping in children with SLI. the output of statistical learning (i.e. identifying
Unlike our findings from the word segmentation boundaries to word forms) and subsequent mapping
task, language status may have influenced fast- of word forms to referents in children with atypical
mapping abilities. Children in the ALI subgroup and development. Contrary to our hypotheses, the lack of a
SLI group performed similarly on the fast-mapping condition effect in the combination task and the lack of
task when measured using eye gaze, but children a significant correlation between segmentation perfor-
with ALI had significantly higher pointing accuracy mance and performance on the combination task in the
than the SLI group. It has been suggested that SLI group indicate that there is no support for the idea
children with ASD have relative strengths in object- that exposure to word forms with high transitional
label associative learning (Preissler, 2008). Addition- probability in a word segmentation task facilitates
ally, children with ASD may be particularly good at subsequent word learning in school-aged children.
learning the sound structure of language, such as the The most surprising finding in the current study
phonological structure of words (J€ arvinen-Pasley was that, unlike the fast-mapping task, children with
et al., 2008; Norbury et al., 2010). The current SLI did not perform significantly worse than the other
findings support this claim for children with ASD groups in the combination task. It is possible that the
who have normal language abilities. However, the statistical structure of the words taught during the
picture is more complex with our ALI group. It is teaching phase of the combination task slightly
possible that performance on the eye-gaze test phase enhanced learning in the SLI group. Alternatively,
of the fast-mapping task may have been driven by the SLI group’s overall increase in looking to the target
attention mechanisms (Nation, Marshall, & Altmann, in word and nonword trials may be explained by the
2003; Norbury, 2005) and that simple object-label artificial language exposure phase wherein children
associative learning in a fast-mapping study may be received additional exposure to the phonemes that
mostly preserved even in children with comorbid ASD subsequently appeared in the words and nonwords
and language impairment. However, children with during the object-label associative learning phase.
ASD with comorbid language delays have been found Building on phonological explanations, exposure to
to have deficits in fast-mapping relative to mental-age the artificial language may have helped the children
expectations (McDuffie, Kover, Hagerman, & Abbe- with SLI to develop more robust phonetic representa-
duto, 2013). McDuffie et al. also found associations tions (Edwards & Lahey, 1996). The more stable
between fast-mapping and concurrent vocabulary phonetic representations may have allowed children
knowledge in children with ASD and language impair- with SLI to develop stronger representations of labels
ments. As previous studies have suggested, child during the teaching phase. Furthermore, the more
characteristics in children with ASD may influence the stable phonetic representations may have been easier
strength of associative learning in children with ASD to map to visual referents and to be differentiated from
and we will explore this in future studies. distractor images during the test phase. Dismissing
Our third research question asked whether the alternative candidates in lexical processing tasks can
output of statistical learning facilitates word learning. be difficult for children with SLI (McMurray et al.,
The results were somewhat mixed. Children with SLI 2014). Alternatively, attention during the teaching
demonstrated greater learning in the combination phase may have been primed because of the pre-
task than the fast-mapping task. As depicted in exposure of the phonemes in the artificial language.
Figures 2 and 3, the proportion of looking to the Previous language-learning studies in children with
target was higher in the combination task relative to SLI have utilized therapy strategies that provide tar-
the fast mapping task. In addition, the children in the geted input to facilitate language development. Such
ALI subgroup did not differ from the SLI group. therapeutic strategies include auditory bombardment
However, although pointing accuracy did not differ, (Alt, Meyers, & Ancharski, 2012), recast intervention
the children with ALI spent less time looking to the (Hassink & Leonard, 2010), and focused stimulation
target image relative to the ALN group in the combina- (Ellis Weismer, Venker, & Robertson, 2016). Additional
tion task, which may be explained by attention differ- work is needed to understand the types of exposure
ences in the two subgroups during the eye-gaze test that facilitate word learning. For example, is exposure
phase. The analyses also revealed that, in contrast to to the sounds/syllables of words that will be taught as
our predictions, children did not learn words more effective in enhancing word learning as exposure to the
novel words in isolation before object-label associa- artificial language also may have facilitated word
tions are presented? The results from this study learning in children with SLI. Finally, this study
indicate that children with SLI have deficits in statis- underscores the importance of examining language
tical learning and fast-mapping; however, the link abilities in children with ASD and taking this child
between these deficits and their problems in word characteristic into account when setting appropriate
learning is unclear. Our data suggest that children therapeutic goals and approaches.
with SLI may require additional phonetic input to
enhance learning. In contrast, children with ASD, as a
group, did not display significant deficits in these Supporting information
learning mechanisms. Although the current study did Additional Supporting Information may be found in the
not address slow mapping or long-term word learning, online version of this article:
practitioners may need to address this component of Appendix S1. Novel objects.
the learning and consolidation process with children Appendix S2. Model summary for research question 3.
with ASD who have language deficits.
Acknowledgements
Conclusion This research was supported by a research grant from
Statistical learning and fast-mapping are impaired in the National Institutes of Health (NIH R01 DC011750),
training grants (NIH F31 DC013485, NIH T32
children with SLI, relative to their age- and cognition-
DC005359, NIH T32 DC00030), and a core grant to
matched peers with ASD and TD. In addition to the Waisman Center (NIH P30 HD03352). The current
previously suggested deficits in procedural learning project also was supported by the Emma Allen Foun-
in the Procedural Deficit Hypothesis, declarative dation Grant and the Vicki Lord Larson and James R.
learning, measured through the fast-mapping task, Larson Research Grant. The views expressed in the
also seems to be impaired in children with SLI. In current article are those of the authors and not neces-
contrast, children with ASD demonstrated intact sta- sarily those of the National Institutes of Health. E.H.
tistical learning, despite heterogeneity in language has no conflicts of interest; S.E.W. consults with the
abilities. Fast-mapping abilities varied slightly NIH on the NIDCD Advisory Council; J.R.S. consults
according to language abilities in children with ASD. with the NIH as a grant reviewer. The authors are
This finding counters suggestions that language grateful to the Language Processes Lab members, the
Language Acquisition and Bilingualism Lab members,
impairments in children with ASD stem from deficits
and the Infant Learning Lab members. We wish to
in procedural learning.
specifically thank Rita Kaushanskaya, Jan Edwards,
The current study provides an initial step in exam- Audra Sterling, Heidi Sindberg, Anna Bina, Taryn
ining whether the output of statistical learning relates Stricker, Andrea Solochek, Kaitlin Meyer, and Lizzy
to subsequent word-learning opportunities. Children Elkin. Lastly, we wish to thank the families and
with SLI demonstrated increased looking to the target children who participated in this study.
and higher pointing accuracy in the combination task
and no longer differed from the TD and ASD groups,
suggesting that the SLI group benefitted from the Correspondence
artificial language exposure before word learning in Eileen Haebig, Department of Communication Sciences
the combination task. However, improvements in & Disorders, Louisiana State University, 68 Hatcher
performance seemed to only partially be attributed Hall, Field House Drive, Baton Rouge, LA 70803, USA;
to the condition (words vs. nonwords), suggesting that Email: ekhaebig@gmail.com
the additional exposure to the phonemes in the
SLI processing load
/
Key points
statistical learning fast-mapping ASD
• This study examined word-learning mechanisms – statistical learning and fast-mapping – in school-aged
children with SLI andASD
ASD. We also examined whether children map meaning onto newly segmentedASD words to
clarify the relationship between these two mechanisms.
•PPTV-4
Children with SLI displayedTDpoorer statistical
ASD SLI
learning than children with ASD and TD.
• Children with SLI and, to some extent, children with ASD with language impairment evidenced poorer fast-
mapping than TD and ASD peers with normal language.
• There was some evidence that word segmentation abilities are SLIassociated
ASD with word learning in school-aged
children with TD and ASD, but not SLI.
SLI&ASD
• Clinical implications include the need to directly address language goals in some children with ASD and to
provide additional phonetic input to children with SLI.
update, and eight principles of autistic perception. Journal Schopler, E., Van Bourgondien, M., Wellman, G., & Love, S.
of Autism and Developmental Disorders, 36, 27–43. (2010). Childhood autism rating scale (2nd edn). Los Angeles:
Nation, K. (2014). Lexical learning and lexical processing in Western Psychological Services.
children with developmental language impairments. Philo- Scott-Van Zeeland, A.A., McNealy, K., Wang, A.T., Sigman, M.,
sophical Transactions of the Royal Society of London. Series Bookheimer, S.Y., & Dapretto, M. (2010). No neural evidence
B, Biological Sciences, 369, 1–10. of statistical learning during autism spectrum disorders.
Nation, K., Marshall, C.M., & Altmann, G.T.M. (2003). Investi- Biological Psychiatry, 68, 345–351.
gating individual differences in children’s real-time sentence Semel, E., Wiig, E.H., & Secord, W.A. (2003). Clinical evalua-
comprehension using language-mediated eye movements. tion of language fundamentals (4th edn). San Antonio, TX:
Journal of Experimental Child Psychology, 86, 314–329. Harcourt Assessment.
Norbury, C.F. (2005). Barking up the wrong tree? Lexical Sheng, L., & McGregor, K.K. (2010). Lexical-semantic organi-
ambiguity resolution in children with language impairments zation in children with specific language impairment. Jour-
and autistic spectrum disorders. Journal of Experimental nal of Speech, Language, and Hearing Research, 53, 146–
Child Psychology, 90, 142–171. 159.
Norbury, C.F., Griffiths, H., & Nation, K. (2010). Sound before Smith, L., & Yu, C. (2008). Infants rapidly learn word-referent
meaning: Word learning in autistic disorders. Neuropsy- mappings via cross-situational statistics. Cognition, 106,
chologia, 48, 4012–4019. 1558–1568.
Obeid, R., Brooks, P.J., Powers, K.L., Gillespie-Lynch, K., & Tager-Flusberg, H. (2016). Risk factors associated with lan-
Lum, J.A.G. (2016). Statistical learning in specific language guage in autism spectrum disorder: Clues to underlying
impairment and autism spectrum disorder: A meta-analysis. mechanisms. Journal of Speech, Language, and Hearing
Frontiers in Psychology, 7, 1–18. Research, 59, 143–154.
Preissler, M.A. (2008). Associative learning of pictures and Tek, S., Jaffery, G., Fein, D., & Naigles, L.R. (2008). Do
words by low-functioning children with autism. Autism: the children with autism spectrum disorders show a shape bias
International Journal of Research and Practice, 12, 231–248. in word learning? Autism Research, 1, 208–222.
Psychology Software Tools, Inc. [E-Prime 2.0]. (2016). Tomblin, J.B., Mainela-Arnold, E., & Zhang, X. (2007). Proce-
Retrieved from http://www.pstnet.com [Last accessed, 01 dural learning in adolescents with and without specific
January 2012]. language impairment. Language Learning and Development,
Rice, M.L. (2016). Specific language impairment, nonverbal IQ, 3, 269–293.
attention-deficit/hyperactivity disorder, autism spectrum Ullman, M.T., & Pierpont, E.I. (2005). Specific language
disorder, cochlear implants, bilingualism, and dialectal impairment is not specific to language: The procedural
variants: Defining the boundaries, clarifying clinical condi- deficit hypothesis. Cortex, 41, 399–433.
tions, and sorting out cases. Journal of Speech, Language, Walenski, M., Tager-Flusberg, H., & Ullman, M.T. (2006).
and Hearing Research, 59, 122–132. Language in autism. In S.O. Moldin & J. L. R. Rubenstein
Rice, M.L., & Hoffman, L. (2015). Predicting vocabulary growth (Eds.), Understanding autism: From basic neuroscience to
in children with and without specific Llanguage impairment: treatment (pp. 175–203). London, UK: CRC/Taylor &
A longitudinal study from 2;6 to 21 years of age. Journal of Frances Books.
Speech, Language, and Hearing Research, 58, 345–359. Wechsler, D. (2003). Wechsler intelligence scale for children
Ruffman, T., Taumoepeau, M., & Perkins, C. (2012). Statistical (4th edn). San Antonio, TX: The Psychological Corporation.
learning as a basis for social understanding in children. Williams, D., Botting, N., & Boucher, J. (2008). Language in
British Journal of Developmental Psychology, 30, 87–104. autism and specific language impairment: Where are the
Rutter, M., Bailey, A., & Lord, C. (2003). Social communication links? Psychological Bulletin, 134, 944–963.
questionnaire. Los Angeles: Western Psychological Services.
Saffran, J.R., Aslin, R.N., & Newport, E.L. (1996). Statistical Accepted for publication: 21 March 2017
learning by 8-month-old infants. Science, 274, 1926–1928.