LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Top Lang Disorders

Vol. 33, No. 4, pp. 282–297
Copyright c 2013 Wolters Kluwer Health | Lippincott Williams & Wilkins

Long-Term Memory
A Review and Meta-Analysis of
Studies of Declarative and
Procedural Memory in Specific
Language Impairment
Jarrad A. G. Lum and Gina Conti-Ramsden
This review examined the status of long-term memory systems in specific language impairment
(SLI)—declarative memory and aspects of procedural memory in particular. Studies included in
the review were identified following a systematic search of the literature and findings combined
using meta-analysis. This review showed that individuals with SLI are poorer than age-matched
controls in the learning and retrieval of verbal information from the declarative memory. However,
there is evidence to suggest that the problems with declarative learning and memory for verbal
information in SLI might be due to difficulties with verbal working memory and language. The
learning and retrieval of nonverbal information from declarative memory appears relatively intact.
In relation to procedural learning and memory, evidence indicates poor implicit learning of verbal
information. Findings pertaining to nonverbal information have been mixed. This review of the
literature indicates that there are now substantial grounds for suspecting that multiple memory
systems may be implicated in the impairment. Key words: declarative memory, procedural
memory, specific language impairment, working memory

problems characterizing specific language

T HE IDEA THAT memory impairments
may cause or contribute to the language
impairment (SLI) has been the focus of much
research and debate in the field (e.g., Gather-
cole & Alloway, 2005; Gathercole & Baddeley,
1990; Joanisse & Seidenberg, 1998; Van der
Lely, 2005). An important development in this
Author Affiliations: School of Psychology, Deakin area has been the identification of working
University, Melbourne, Australia (Dr Lum); and memory impairments in SLI (for reviews, see
School of Psychological Sciences, The University of
Manchester, Manchester, United Kingdom Alloway & Gathercole, 2012; Montgomery,
(Dr Conti-Ramsden). Magimairaj, & Finney, 2010). From this litera-
The authors acknowledge the support of the Wellcome ture, it appears that individuals with SLI have
Trust (Grant #079305). problems with the short-term storage and
Supplemental digital content is available for this manipulation of verbal information but not
article. Direct URL citations appear in the printed necessarily with visual (and not easily verbally
text and are provided in the HTML and PDF ver- coded) information (Archibald & Gathercole,
sions of this article on the journal’s Web site (www.
[Link]). 2006a, 2006b; Lum, Conti-Ramsden, Page,
& Ullman, 2012a). In relatively more recent
The authors have indicated that they have no financial
and no nonfinancial relationships to disclose. times, there has been growing interest
in the memory systems that support the
Corresponding Author: Jarrad A. G. Lum, PhD,
BBsc (Hons), School of Psychology, Deakin Univer- learning and long-term storage of informa-
sity, 221 Burwood Highway, Melbourne, Australia (jar- tion (Evans, Saffran, & Robe-Torres, 2009;
[Link]@[Link]). Hsu & Bishop, 2011; Ullman & Pierpont,
DOI: 10.1097/[Link].0000437939.01237.6a 2005). This review summarizes research

282

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 283

investigating two long-term memory systems An overview of the declarative

in SLI—the declarative and procedural memory system
memory systems. The declarative memory system principally
supports learning, storage, and retrieval of
knowledge or memory pertaining to personal
MULTIPLE MEMORY SYSTEMS IN THE events, referred to as episodic memory, and
BRAIN: THE DECLARATIVE AND general information, referred to as semantic
PROCEDURAL SYSTEMS memory (Squire, 1992; Squire et al., 2004;
Tulving & Markowitsch, 1998). Examples of
It is widely accepted that humans’ abil- episodic memory include knowing particular
ity to learn, store, and retrieve information events that occurred at a dinner party. An ex-
is supported by separate memory systems ample of semantic memory includes knowing
(Squire, 2004). In this review, the term “mem- that the word “dog” corresponds to the object
ory system” refers collectively to the learn- or concept of a dog. Learning via the declara-
ing, storage, and retrieval of information. tive memory system is achieved through bind-
The declarative and procedural systems differ ing arbitrarily related pieces of information to-
with respect to supporting brain structures gether (Mayes, Montaldi, & Migo, 2007).
and cognitive processes that underlie learn- Learning via the declarative memory sys-
ing and memory (Eichenbaum, 2000; Packard tem can be fast. In some cases, only a sin-
& Knowlton, 2002; Squire, Stark, & Clark, gle exposure to the information is required
2004). However, although distinct systems in order for information to be learned. How-
have been identified, it is important to recog- ever, repeated exposures to the information
nize that different memory systems cooperate will increase the propensity that learning will
and compete to support learning and mem- take place and also the speed at which it
ory (Packard & Knowlton, 2002; Poldrack can be retrieved (Alvarez & Squire, 1994).
et al., 2001). Subsequently, the knowledge Retrieval of information from the declara-
and skills that we learn, including language, tive memory system occurs through the pro-
can be thought of as the product of interacting cesses of recall and recognition (Knowlton &
memory systems rather than a set of systems Squire, 1995). In many contexts, but not all,
working independently. the learning and retrieval of information re-
A widely accepted taxonomy of memory quires awareness. That is, cognitive effort is
systems differentiates between declarative required to learn (e.g., studying for an exam)
and nondeclarative memory (Squire, 1992). and retrieve the information (e.g., respond-
Procedural memory is one of a number of ing to a question on an exam). The learn-
systems that comprise the nondeclarative sys- ing and retrieval processes of the declarative
tem. Both declarative and procedural systems memory systems are supported by the medial
have the capacity to store information from temporal lobes, in particular the hippocam-
minutes to years. This can be contrasted with pus (Eichenbaum, 2000, 2004; Squire et al.,
working memory, which holds information 2004).
in an active state for use in cognition for a
short period of time (Baddeley, 2003). Col-
lectively, the declarative and procedural sys- An overview of the procedural
tems allow us to learn and use information memory system
and skills across the life span. It is known that The procedural memory system under-
disruption to the declarative or procedural lies the implicit learning, storage, and re-
system significantly reduces the ability to trieval of skills and knowledge (Gabrieli, 1998;
learn, even if working memory is intact (e.g., Squire & Zola, 1996). The main neural struc-
Corkin, 2002; Heindel, Butters, & Salmon, tures that support this memory system are a
1988). collection of subcortical structures referred to

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

284 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

as the basal ganglia and cerebellum as well as aspects of language. In the case of language
the prefrontal cortex (Packard & Knowlton, development in nonimpaired populations,
2002; Parent & Hazrati, 1995). The learning the proposal is that procedural memory sup-
and retrieval of information from the proce- ports the use and learning of grammar, and
dural memory system is said to be implicit the declarative memory system supports the
because conscious awareness is not required. use of lexical knowledge (for exceptions, see
Unlike declarative memory, procedural Hartshorne & Ullman, 2005). In SLI, Ullman
memory is better suited to learning sequen- and Pierpont (2005) argued that the grammat-
tially or learning probabilistically structured ical impairments in this group are primarily
information (Packard & Knowlton, 2002). caused by neurological problems that under-
This includes learning new motor and per- lie the procedural memory system. It is also
ceptual skills (Nissen & Bullemer, 1987) and proposed that the learning and memory func-
forming associations between pieces of ver- tions of the declarative memory system are
bal or visual information that are probabilis- relatively intact. Subsequently, lexical knowl-
tically or statistically structured (Conway & edge and other aspects of language hypothe-
Pisoni, 2008; Karuza et al., 2013; Knowlton, sized to be dependent on this memory system
Squire, & Gluck, 1994). The neurological ar- should not be key areas of impairment in SLI.
chitecture of the procedural memory sys- A number of studies have been conducted
tem permits implicit learning, storage, and re- that have investigated the learning and mem-
trieval processes to be carried out on visual, ory functions of the declarative and proce-
verbal, cognitive, and also linguistic informa- dural memory systems in SLI. In this litera-
tion (Alexander & Crutcher, 1990; Alexander, ture, findings concerning the status of these
DeLong, & Strick, 1986; Ullman, 2006). two memory systems in SLI have been mixed.
Learning via the procedural memory system Specifically, contrasting data have been pre-
is often slow, with repetition or repeated ex- sented showing that declarative and proce-
posures to the information required in order dural memory systems are intact (Gabriel,
for a skill or knowledge to be learned (Packard Maillart, Guillaume, Stefaniak, & Meulemans,
& Knowlton, 2002). However, once informa- 2011; Mayor-Dubois, Zesiger, van der Linden,
tion has been acquired, new knowledge and & Roulet-Perez, 2012) and impaired (Evans
skills may be used without awareness and et al., 2009; Kemény & Lukács, 2010; Lum,
sometimes only following the presence of a Gelgec, & Conti-Ramsden, 2010). Thus, con-
preceding stimulus. An often-cited example sensus is yet to be reached whether declar-
is our ability to learn and then automatically ative and procedural memory systems are
execute new motor skills without awareness. impaired in SLI.
The act of repeating a motor skill (e.g., reach- In this report, research investigating these
ing for an object) will come to be performed long-term memory systems in SLI is reviewed.
rapidly and without awareness. To provide a concise and objective review
of the literature, studies were identified us-
ing a systematic search of databases and
DECLARATIVE AND PROCEDURAL their results synthesized using meta-analytic
MEMORY FUNCTIONING IN SLI techniques.

At present, there are empirical and the- METHODS

oretical grounds to suspect that language
learning is supported by both declarative and
procedural memory systems. Ullman (2001, Database searching
2004) has proposed that the declarative and Articles that were published or were
procedural memory systems support different “in press” up to December 2012 were

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 285

identified following a systematic search of Study inclusion criteria

CINHAL, ERIC, MEDLINE, and PsycINFO Studies were included in the meta-analysis
electronic databases. Databases were ac- if they met two inclusionary criteria. First, the
cessed via EBSCOHost. Two sets of searches study was required to be published or be “in
were undertaken. The first identified studies press” in a peer-review journal (written in any
that investigated declarative memory for ver- language) and to report on an original study.
bal or visual information in SLI using standard- Data reported in literature reviews, disserta-
ized tests (for a review of the performance of tions, or book chapters were not included in
individuals with SLI on experimental declar- this meta-analysis. Second, the study needed
ative memory tasks, see Ullman & Pierpont, to present a task that assessed the declarative
2005). A description of the search strategy or procedural memory system using verbal
and key words used to identify these studies or visual stimuli to one group of individuals
is presented in Supplemental Digital Content with SLI and a control group comprising indi-
Table A, available at [Link] viduals of comparable age without language
TLD/A24. impairments. The search that aimed to iden-
The second search identified studies that tify studies investigating declarative memory
had investigated procedural learning and in SLI initially returned 1724 records. After
memory in SLI. In keeping in line with the duplicates were removed and inclusionary cri-
declarative memory studies, the search strat- teria applied, a total of 11 studies were iden-
egy was designed to identify studies that tified. The search that aimed to identify stud-
have investigated procedural memory for ver- ies investigating procedural memory initially
bal or visual information in this group. It identified 250 records. A total of four studies
should be noted that there are very few stan- were identified after applying the inclusionary
dardized tests that assess procedural mem- criteria.
ory (Baron, 2004). However, a number of
validated experimental tasks have been de- Meta-analytic methods
veloped that are known to tap the proce- Results from each individual study were
dural memory system (e.g., Aslin, Saffran, & summarized using the effect size Cohen’s d
Newport, 1998; Karuza et al., 2013; Nissen along with 95% confidence intervals and sig-
& Bullemer, 1987; Poldrack et al., 2001). nificance test. Cohen’s d describes differences
The search strategy aimed to identify stud- between groups in standard deviation units
ies that had presented such tasks to individ- (for an introduction to effect sizes, see Durlak,
uals with SLI. A description of the search 2009). For all comparisons, Cohen’s d was cal-
strategy and key words used in this search culated so that positive values indicated that
is presented in Supplemental Digital Content the control group performed better than the
Table B, available at [Link] SLI group on the memory task. The 95% con-
TLD/A25. fidence interval accompanying each Cohen’s
We note that research also has been un- d value provides an estimate of the precision
dertaken investigating procedural learning of of the study’s effect size. Larger confidence
visuomotor information (e.g., Gabriel et al., intervals denote poorer precision (for
2011; Lum et al., 2012a; Tomblin, Mainela- an introduction to interpreting confidence
Arnold, & Zhang, 2007). For systematic intervals, see Cumming & Finch, 2005). In
reviews and meta-analyses of this literature some cases, it was possible to combine the re-
with respect to neurodevelopmental disor- sults of several studies investigating the same
ders such as SLI, see the work by Lum aspect of memory and compute an average ef-
et al. (Lum, Conti-Rasmden, Morgan & Ull- fect size for a group of studies. Averaging of
man, submitted for publication; Lum, Ullman, effect sizes was undertaken using a random
& Conti-Ramsden, 2013). effects model (Hedges & Vevea, 1998).

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

286 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

RESULTS AND DISCUSSION Kramer, Kaplan, & Ober, 1994). In both scor-
ing methods, the assumption is that as more
Results from the meta-analysis investigating words are encoded into declarative mem-
learning and memory functions of the declar- ory, the number of words correctly recalled
ative memory system in SLI are reviewed increases.
first. Figure 1 summarizes results from studies
investigating verbal learning in SLI using list
Verbal learning via the declarative learning/retrieval tasks in a forest plot. Forest
memory system in SLI plots show individual study effect sizes and
Learning verbal information by the declar- their confidence intervals as well as the over-
ative memory system in SLI has been investi- all average effect size. As is standard prac-
gated using list learning/retrieval tasks. These tice, data points showing results from indi-
tasks assess both learning and retrieval of vidual studies are presented as circles and the
verbal information from declarative memory weighted average as a diamond (for an intro-
(Baron, 2004; Lezak, 2012). The learning com- duction to forest plots, see Lewis & Clarke,
ponent of the task assesses how well an ex- 2001). All studies summarized in Figure 1 re-
aminee can encode arbitrary pieces of ver- ported that children with SLI recalled signifi-
bal information given multiple exposures. The cantly fewer words during the learning trials
pieces of information in the task are individ- than age-matched control groups. The effect
ual words or, pairs of words (e.g., nurse–fire) sizes observed in each study are consistent;
that are presented in a list. During the learn- the smallest effect size was found to be 0.769
ing component, the complete list of words (Riccio, Cash, & Cohen, 2007) and the largest
is repeatedly presented. Depending on the 1.150 (Dewey & Wall, 1997). The average ef-
test, the list may be repeated three or four fect size computed using all studies summa-
times. After each presentation, the examinee rized in Figure 1 was found to be 0.899 and
is asked to recall the entire list. Declarative statistically significant. Thus, on average, the
learning for verbal information on these tasks literature shows that verbal learning is signif-
is quantified by summing the total number icantly poorer in children with SLI than in
of words correctly recalled over the learning typically developing children of the same age.
trials (e.g., Cohen, 1997; Sheslow & Adams, The studies summarized in Figure 1 sug-
1990) or the number of words correctly re- gest that children with SLI learn fewer pieces
called on the final learning trial (e.g., Delis, of verbal information than nonlanguage

95% CI SLI group Control group

Study Test Name Cohen's d p
Lower Upper performs better performs better
Baird et al. (2010) WRAML (Verbal Learning subtest) 0.945 0.449 1.440 <.001**
Dewey & Wall (1997) WRAML (Verbal Learning subtest) 1.150 0.428 1.871 .002*
Duinmeijer et al. (2012) CVLT-C (List Recall subtest) 0.760 0.281 1.239 .002*
Lum et al. (2010) CMS (Word Pairs subtest) 0.769 0.047 1.492 .037*
Lum et al. (2012a) CMS (Word Pairs subtest) 0.997 0.586 1.409 <.001**
Nichols et al. (2004) CVLT-C (Learning Efficiency subtest) 0.775 0.236 1.313 .005*
Records et al. (1995) RAVLT 0.818 0.273 1.363 .003*
Riccio et al. (2007) CMS (Word Pairs and Word List subtests) 0.797 0.268 1.325 .003*
Shear et al. (1992) CVLT-C (Learning Efficiency subtest) 0.947 0.103 1.791 .028*
Average effect size 0.879 0.696 1.061 <.001**
-1.0 0.0 1.0 2.0 3.0

Figure 1. Effect sizes of studies examining declarative learning for verbal information in SLI using list
learning/retrieval tasks. *p < .05; **p < .001. CI = confidence interval; CMS = Children’s Memory Scale;
CVLT-C = California Verbal Learning Test for Children; RAVLT = Rey Auditory Verbal Learning Test; SLI =
specific language impairment; WRAML = Wide Range Assessment of Memory and Learning.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 287

impaired children of the same age. These initial differences on the first trial only, which
findings might indicate impairments related in turn suggest verbal short-term memory
to the learning mechanisms of the declara- limitations.
tive memory system in SLI. Based on simi- Other evidence showing that poor working
lar observations, Lum et al. (2010) suggested memory might underlie poor verbal learning
that declarative memory for verbal informa- in SLI was presented by Lum et al. (2012a). In
tion might be impaired in SLI. This suggestion that study, the performance of children with
was forwarded after finding that children with SLI and age-matched peers on verbal learn-
SLI performed significantly more poorly on a ing was compared and differences in verbal
composite measure of learning and recall on working memory statistically controlled. Be-
list learning/retrieval tasks than age-matched fore controlling for verbal working memory,
controls. the observed value for Cohen’s d on the mea-
Subsequent work (Lum & Bleses, 2012; Lum sure of verbal learning was found to be 0.997
et al., 2012a) coupled with closer inspection (see Figure 1). However, after controlling for
of past research appears to question whether verbal working memory, the effect size de-
declarative learning is impaired in SLI (e.g., creased to 0.263, and the difference between
Records, Tomblin, & Buckwalter, 1995). It has the groups was no longer found to be statisti-
been found that the rate at which new words cally significant. This study showed that differ-
are learned from one trial to the next in list ences in verbal learning between SLI and con-
learning/retrieval tasks is similar between par- trol groups can be reduced once the influence
ticipants with SLI and their age-matched coun- of verbal working memory is removed. Third,
terparts. For example, using a nonstandard- there is also the influence of language level
ized list learning task, Lum and Bleses (2012) to consider. Having a higher level of language
did not find a difference in the learning rate functioning may help children “see” connec-
of new words between children with SLI and tions between arbitrary words while this is
age-matched controls. This result was previ- likely to be more difficult for children with
ously observed by Nichols et al. (2004) and poor language. Better language can, there-
Records et al. (1995). These studies found fore, help children to do better even from the
that the number of new words learned be- first trial in recalling lists. Thus, differences in
tween learning trials was not significantly dif- the severity of the language problems in SLI
ferent between children with SLI and age- may at least partly explain the lack of replica-
matched controls. These findings suggest that tion of findings across the studies described
differences between individuals with SLI and previously.
control groups on list learning/retrieval tasks
might be due to difficulties with other mem- Nonverbal learning in the declarative
ory, for example, difficulties in verbal working memory system in SLI
memory. There is some evidence to support The ability of children with SLI to learn
this claim. Children with SLI have been found information via the declarative memory sys-
to recall significantly fewer words after the tem has also been investigated using non-
first trial on list learning tasks (e.g., Lum & verbal tasks. The results from these studies
Bleses, 2012; Nichols et al., 2004; Records potentially shed light on the functioning of
et al., 1995; Shear, Tallal, & Delis, 1992). the declarative learning mechanisms in SLI.
In the neuropsychological literature, perfor- This is because nonverbal tasks place mini-
mance on this part of the list-learning/retrieval mal demands on verbal working memory ca-
tasks is considered to be an indicator of short- pacity. Subsequently, the difference between
term memory (Lezak, 2012). Thus, observed individuals with and without SLI on nonverbal
differences between children with and with- learning tasks should be smaller; assuming
out SLI on verbal learning appear to reflect that declarative learning is relatively intact.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

288 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

A number of standardized memory tests in- Range Assessment of Memory and Learn-
clude subtests designed to probe the learning ing (WRAML; Sheslow & Adams, 1990). The
mechanism of the declarative memory system performance of children with SLI on these
for nonverbal information. These tasks are de- subtests compared with age-matched peers
signed to be nonverbal analogues of the list has been reported in several studies (Baird,
learning/retrieval tasks described previously. Dworzynski, Slonims, & Simonoff, 2010;
In these tasks, an examinee is repeatedly pre- Bavin, Wilson, Maruff, & Sleeman, 2005; Lum
sented with a picture that cannot be ver- et al., 2010). Results from studies investigating
balized easily. Following each presentation, declarative learning of nonverbal information
the participant is prompted to reproduce the are summarized in Figure 2. This figure shows
picture or recognize elements of the picture that overall, there are no differences be-
from a set of distracters. Learning on the task tween SLI and age-matched control groups on
is quantified by measuring whether there is tasks assessing declarative learning for visual
an increase in nonverbal information recalled information.
during the learning trials. Figure 2 shows that no study identified in
Lum et al. (2012a) and Riccio et al. (2007) our literature search reported significant dif-
presented the Dot Locations subtests from the ferences between children with and without
Children’s Memory Scale (CMS; Cohen, 1997) SLI on the tests assessing learning of nonverbal
to children with and without SLI. On that task, information over multiple trials. In the studies
children are presented with a picture com- by Riccio et al. (2007), Lum et al. (2012a), and
prising randomly placed dots. The picture is Lum et al. (2010), the observed effect sizes
shown multiple times and after each presen- were close to zero. That is, the SLI and con-
tation, children aim to recreate the picture trol groups were statistically indistinguishable
using small tokens. A measure of learning is on the task. The average effect size computed
obtained by summing how many tokens were from all the studies summarized in Figure 2
placed in a position that matched their loca- was found to be 0.228, which although signif-
tion in the stimulus picture. A variant of this icant at p = .048, reflects a small effect size
task is for children to learn the spatial loca- (Cohen, 1988). Thus, the literature on declar-
tion of an abstract object over a number of ative learning in SLI indicates a small differ-
trials. This ability is assessed by the Paired ence (just more than 0.2 SD) between chil-
Associates Learning Subtest from the Cam- dren with SLI and their typically developing
bridge Neuropsychological Test Automated peers. Overall, there is little evidence to sug-
Battery (Cambridge Cognition, 1996) and gest that learning of nonverbal information is
Visual Learning subtest from the Wide impaired in SLI.

95% CI SLI group Control group

Study Test Name Cohen's d p
Lower Upper performs better performs better
Baired et al. (2010) WRAML (Visual Learning subtest) 0.410 –0.066 0.887 .092
Bavin et al. (2005) CANTAB (Paired Associates subtest) 0.398 –0.212 1.009 .201
Lum et al. (2010) CANTAB (Paired Associates subtest) 0.000 –0.716 0.716 .999
Lum et al. (2012a) CMS (Dot Locations subtest) 0.239 –0.151 0.628 .229
Riccio et al. (2007) CMS (Dot Locations subtest) 0.000 –0.506 0.506 .999
Average effect size 0.228 0.002 0.454 .048*

–1.0 0.0 1.0 2.0

Figure 2. Effect sizes of studies examining declarative learning for nonverbal information in SLI using
learning/retrieval tasks. *p < .05. CANTAB = Cambridge Neuropsychological Test Automated Battery;
CI = confidence interval; CMS = Children’s Memory Scale; SLI = specific language impairment;
WRAML = Wide Range Assessment of Memory and Learning.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 289

Retrieval of information from the The results from studies investigating re-
declarative memory system in SLI trieval of verbal information are summarized
The other component of the declarative in Figure 3. To assist with interpretation, the
memory system studied in SLI is the retrieval computed effect sizes are averaged across re-
of information. There are two broad meth- call and recognition conditions and across
ods used to study the retrieval of verbal infor- short and delayed recall retrieval conditions.
mation from the declarative memory system. In several studies, the processes of recall
One method assesses how well an examinee and recognition in immediate and delayed
can recall or recognize arbitrarily related in- retrieval conditions have been found to be
formation that has been learned over multiple highly correlated (Baird et al., 2010; Lum
trials. This method of investigating retrieval is et al., 2012a; Riccio et al., 2007). Thus, for
typically assessed in an additional component the purposes of this review, averaging over
of the list learning/retrieval tasks described these conditions is not masking any process-
previously (e.g., Records et al., 1995; Shear specific effects.
et al., 1992). In the retrieval component of Figure 3 shows that most of the studies have
the task, after the learning phase has been reported that participants with SLI retrieve
completed, participants are asked to recall or significantly less verbal information than chil-
recognize previously studied list of words. A dren of the same age who are not impaired.
different method assesses how well a verbally There are some exceptions to this trend.
presented episodic event can be retrieved fol- Studies by Baird et al. (2010) and Records
lowing a single exposure (e.g., Cohen, 1997; et al. (1995) observed small effect sizes and
Sheslow & Adams, 1990). This type of re- nonsignificant differences between partici-
trieval is assessed in story recall/recognition pants with SLI and age-matched peers. Shear
tasks. In these tasks, the examinee is pre- et al. (1992) also observed a nonsignificant
sented with a short story. At the end of the difference between children with and with-
story, the examinee is asked to either recall out SLI. In that study, the effect size was
the entire story or recognize different similar to those studies that did find a sig-
elements. nificant difference (e.g., Lum et al., 2012a;

95% CI SLI group Control group

Task Type/Study Cohen's d p
Lower Upper performs better performs better
List recall/recognition
Baird et al. (2010) 0.100 –0.372 0.572 .678
Lum et al. (2012a)a 0.781 0.377 1.184 <.001**
Nichols et al. (2004)a 0.690 0.152 1.228 .012*
Records et al. (1995) 0.196 –0.329 0.721 .464
Riccio et al. (2007)a 0.607 0.088 1.125 .022*
Shear et al. (1992)b 0.607 –0.215 1.428 .148
Average effect size 0.497 0.253 0.741 <.001**

Story recall/recognition
Baird et al. (2010)a 0.224 –0.252 0.701 .356
Dewey & Wall (1997) 1.799 1.007 2.591 <.001**
Lum et al. (2012a)a 1.003 0.591 1.415 <.001**
Riccio et al. (2007)a 0.582 0.064 1.100 .028*
Average effect size 0.845 0.292 1.398 .003**
–1.0 0.0 1.0 2.0

Figure 3. Effect sizes of studies examining retrieval of verbal information from declarative memory in SLI.
a
Average effect size for recall and recognition in immediate and delayed conditions. b Average effect size
for recall in short and delayed conditions. *p < .05; **p < .001. CI = confidence interval; SLI = specific
language impairment.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

290 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

Nichols et al., 2004; Riccio et al., 2007). and retrieval of verbal information from
Consequently, insufficient statistical power declarative memory in this group. Some evi-
most likely accounts for their nonsignificant dence already supports this suggestion. Lum
results. Overall, significant differences have et al. (2012a) first observed significant differ-
been found irrespective of whether the infor- ences between children with and without SLI
mation to be retrieved is a list of unrelated on declarative memory retrieval tasks (see
words, presented over multiple trials, or a Figure 3). After controlling for verbal work-
short story presented once. This suggests that ing memory, the average effect size across
individuals with SLI retrieve less verbal infor- different retrieval tasks was found to be 0.23
mation from declarative memory than typical and nonsignificant. Interestingly, a further
children of the same age. reduction in the effect size was observed
However, as with verbal learning, the ap- when both working memory and language
parent difficulty children with SLI have with abilities were simultaneously controlled.
retrieval might be due to other factors and not Thus, the difference in the retrieval of verbal
poor declarative memory. Initial evidence to information in SLI may be secondary to both
support this claim comes from closer inspec- working memory and language problems. In
tion of the study by Baird et al. (2010), who sum, individuals with SLI struggle to retrieve
observed a small effect size. Memory func- verbal information. However, the difficulty
tioning in that study was assessed using the observed in this area may be due to verbal
WRAML (Sheslow & Adams, 1990). The mea- working memory and language problems and
sure of retrieval in the WRAML is calculated not declarative memory deficits.
by finding the difference between the total Data also exist concerning retrieval of non-
number of words recalled in the learning and verbal and visual information from declarative
retrieval conditions. In contrast, the instru- memory in SLI. This aspect of memory in SLI
ments used in the other studies that found has been assessed with three different tasks.
significant differences between SLI and con- One task assesses retrieval of nonverbal or vi-
trols (e.g.,Lum et al., 2012a; Nichols et al., suospatial information following multiple tri-
2004; Riccio et al., 2007) did not directly con- als. This task is designed to be a visual ana-
trol for performance in the learning condition. logue of the list learning/retrieval tasks. For
Rather, the standardization of subtest scores example, on the Dot Locations task from the
from learning conditions aims to correct for CMS (Cohen, 1997), the examinee is asked
performance on learning trials. It is also in- to recreate a previously shown picture de-
teresting to note that Records et al. (1995) picting a random array of dots, using tokens,
adjusted participants’ retrieval scores to ac- after a delay. Another commonly used non-
count for learning and immediate retrieval verbal task assesses recognition of faces (e.g.,
conditions. After this calculation, they found Cohen, 1997). During the assessment, the ex-
a nonsignificant trend in which the partici- aminee is shown a number of faces that he
pants with SLI retrieved more words than chil- or she is asked to commit to memory. Re-
dren who were not impaired. Thus, control- trieval is examined by testing how accurately
ling for differences in verbal learning reduces the faces previously studied can be identi-
differences between SLI and non-SLI groups fied from “distracter” faces that were not pre-
on declarative memory retrieval tasks. viously shown. Finally, several standardized
It was noted earlier that verbal learning tests (Cohen, 1997; Sheslow & Adams, 1990)
problems in SLI might be due to poor verbal assess retrieval of information from pictures
working memory. An interesting proposition that show common events (e.g., eating lunch
that will need to be explored in future at a picnic). In these “Picture Tasks,” the ex-
research is whether poor verbal working aminee is shown the picture once for a short
memory problems in SLI underlie learning period of time (e.g., 10 s) and then asked to

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 291

retrieve specific information about the pic- 2012b), the difference between the SLI and
ture. The task is designed to be a visual ana- control groups was found to be statistically
logue of the story recall/recognition task. significant. In the remaining studies (Baird
A summary of study findings from tests as- et al., 2010; Riccio et al., 2007), the differ-
sessing retrieval of nonverbal or visual infor- ences between groups was not significant but
mation in SLI is presented in Figure 4. On effect sizes were medium in magnitude. The
tasks assessing retrieval of nonverbal/visual average effect size for all studies using the Pic-
spatial information including faces, most stud- ture Task was found to be 0.526, which was
ies have observed small effect sizes and non- statistically significant.
significant differences between SLI and con- The difficulty children with SLI have with
trol groups. The average effect size observed retrieving information from pictures show-
on tasks assessing retrieval of abstract non- ing everyday events may be related to ver-
verbal information and faces is 0.198 and bal memory. In the adult neuropsychological
0.157, respectively. These findings indicate literature, concerns have been raised about
that children with SLI are able to retrieve some the validity of Picture Tasks as a test of vi-
types of nonverbal information from declara- sual memory. Studies have found that perfor-
tive memory with equal proficiency as non- mance on this task in adults with neurological
language impaired children of the same age. impairment can be best predicted by learning
It is certainly not the case that children and memory for verbal information (Chapin,
with SLI are proficient at retrieving all types Busch, Naugle, & Najm, 2009; Dulay et al.,
of nonverbal/visual information from declara- 2002). One explanation for this association is
tive memory. Figure 4 shows that in the Pic- that the events shown in Picture Tasks can be
ture Tasks, children with SLI typically retrieve verbally encoded. Lum et al. (2012b) further
significantly less information than typical chil- observed that, in children with and without
dren of the same age. In two studies (Dewey SLI, performance on the Picture Task could
& Wall, 1997; Lum, Conti-Ramsden, & Ullman, be predicted by verbal working memory. This

95% CI SLI group Control group

Information Type/Study Cohen's d p
Lower Upper performs better performs better
Dots and abstract pictures
Baird et al. (2010) 0.166 –0.307 0.640 .490
Dewey & Wall (1997) 0.743 0.051 1.435 .035*
Lum et al. (2012a)a 0.241 –0.149 0.630 .227
Riccio et al. (2007)a –0.172 –0.679 0.335 .506
Average effect size 0.198 –0.106 0.501 .203

Faces
Lum et al. (2012a)a 0.090 –0.298 0.479 .648
Riccio et al. (2007)a 0.271 –0.237 0.780 .296
Average effect size 0.157 –0.152 0.466 .319

Pictures showing everyday events

Baird et al. (2010) 0.439 –0.039 0.916 .072
Dewey & Wall (1997) 0.746 0.054 1.438 .035*
Lum et al. (2012b)a 0.592 0.218 0.966 .002*
Riccio et al. (2007)a 0.383 –0.127 0.894 .141
Average effect size 0.526 0.287 0.765 <.001**
-1.0 -0.5 0.0 0.5 1.0

Figure 4. Effect sizes of studies examining retrieval of nonverbal information from declarative memory
in SLI. a Average effect size for short and delayed recall conditions. *p < .05; **p < .001. CI = confidence
interval; SLI = specific language impairment.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

292 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

association was found even after controlling (Knowlton et al., 1994), auditory/verbal
for the effects of language and declarative (Saffran, 2003; Saffran, Aslin, & Newport,
memory skills. In light of these findings, there 1996), and visuospatial-motor domains
is evidence to suggest that verbal working (Nissen & Bullemer, 1987).
memory may underlie the difficulty children
with SLI have in retrieving information about Procedural learning and memory
pictures. for auditory/verbal information

Is the declarative memory system Implicit statistical learning of auditory in-

impaired in SLI? formation in SLI has been investigated by
Evans et al. (2009) and Mayor-Dubois et al.
Overall, the evidence on the declarative (2012). In both studies, children with SLI and
memory system in SLI mirrors the data on age-matched peers listened to a continuous
verbal working memory. Children with SLI stream of nonwords or tones. The stimuli are
perform poorly, relative to age-matched con- constructed so that the adjacent probabili-
trols, on tasks that assess the learning and ties between specific phonemes/tones within
retrieval of verbal information from declar- nonwords/phases are higher than those be-
ative memory. The analogous functions for tween nonwords/phases (Saffran, Newport,
visual/visuospatial information appear to be Aslin, Tunick, & Barrueco, 1997). During test-
intact in SLI. An outstanding question to ing, the stimuli are presented while the chil-
be addressed is whether declarative mem- dren are engaged in another activity (e.g.,
ory for verbal information is impaired in SLI drawing a picture). This ensures that children
or whether the problems are more closely are not actively attending to the verbal infor-
tied to verbal working memory and/or lan- mation and any learning that takes places is
guage problems. Experimental work will be implicit. After the exposure period, knowl-
required to clarify this issue. edge of the phonemes or tones presented ear-
lier is assessed using a recognition task. Re-
The procedural memory system in SLI cent evidence indicates that learning on this
Unlike declarative memory, there are no type of task draws upon the same neural net-
standardized assessment tools that evaluate works that support other forms of procedural
the learning and memory mechanisms of learning (Karuza et al., 2013). Thus, the im-
the procedural memory system. However, plicit learning of verbal information can be
a number of experimental paradigms have considered to engage the procedural learning
been developed to investigate the implicit system. Figure 5 presents a summary of indi-
learning and retrieval of information in visual vidual study findings.

95% C.I. SLI group Control group

Task Type/Study Cohen's d p
Lower Upper performs better performs better
Auditory statistical learning
Evans et al. (2009)a 0.832 0.165 1.499 .015*
Mayor-Dubois et al. (2012) 0.836 0.298 1.373 .002*
Average effect size 0.834 0.416 1.253 <.001**

Visual statistical learning

Kemény & Lukács (2010) 1.059 0.319 1.799 .005
Mayor-Dubois et al. (2012) 0.224 -0.299 0.747 .401
Average effect size 0.502 0.075 0.930 .021

-1.0 0.0 1.0 2.0

Figure 5. Effect sizes of studies investigating implicit learning and memory of verbal and nonverbal
information. a Average effect size for verbal and tonal stimuli in experiment 1 and 2. CI = confidence
interval; SLI = specific language impairment. *p < .05; **p < .001.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 293

The results from both Evans et al. (2009) 75% of the time and another set of cues 25% of
and Mayor-Dubois et al. (2012) showed that the time. At the start of the task, participants’
children with SLI performed significantly accuracy is around chance level. However,
more poorly than non–language impaired chil- following more trials, accuracy increases. The
dren of the same age in tasks that involved learning on the task is implicit because partici-
recognizing implicitly learned auditory infor- pants are unable to articulate the relationship
mation. Figure 5 shows that the observed ef- between cues and outcomes. However, evi-
fect sizes for both studies were around 0.8. dence that participants have learned associa-
The average effect size for these two stud- tions is observed as accuracy increases over
ies is also statistically significant. Interestingly, trials.
Evans et al. (2009) also found that the magni- Two studies have investigated probabilistic
tude of the difference between children in classification in SLI. Results from these two
the SLI and comparison groups could be re- studies are also summarized in Figure 5. The
duced by increasing the exposure period to available evidence on probabilistic classifica-
the auditory stimuli. In one experiment, chil- tion in SLI has produced conflicting results. In
dren with and without SLI were presented the study by Kemény and Lukács (2010), a sig-
with the auditory stimuli for 24 minutes and nificant difference between children with and
in another, 48 minutes. The difference be- without SLI was observed on the task. Specifi-
tween the SLI and control groups was signifi- cally, after 150 training trials the children with
cant in the 24-minute exposure period but not SLI were not able to predict outcomes above
in the 48-minute exposure period. This pat- chance level whereas children in the control
tern of results does not support the view that group could. Also, as seen in Figure 5, the ob-
SLI is associated with an absence of implicit served effect size in that study was large. In
learning. Rather, it seems that implicit learn- contrast, Mayor-Dubois et al. (2012) found no
ing occurs in SLI, but these affected individ- significant differences in accuracy between
uals require increased exposure to the target children with and without SLI after 200 trials.
information. In that study, both groups of children were
able to learn the association between the cues
Procedural learning and memory and outcomes.
for visual information In accounting for the discrepant findings,
Implicit learning in SLI in the visual do- one suggestion forwarded by Mayor-Dubois
main has been investigated using probabilis- et al. (2012) is that differences in study
tic classification tasks. A considerable volume findings on the probabilistic classification
of research has shown that these tasks are task might reflect differences in declarative
supported by parts of the brain that under- memory. They point out that the relationship
lie the procedural memory system (Poldrack between cues and outcomes used in the study
et al., 2001; Poldrack & Packard, 2003). In by Kemény and Lukács (2010) was simple,
these tasks, participants are seated in front of a and participants may have obtained explicit
computer display and different combinations awareness of the association. If this were
of four cues appear on the screen. Participants the case, learning on the task by the SLI and
are asked to work out which combinations of control group participants might have been
cues are associated with one of two outcomes. supported by the declarative memory system.
During testing, different cue combinations ap- Given previous research showing intact
pear and the participant presses one of two declarative memory for visual information in
buttons on a keyboard to indicate a particular SLI, it would be reasonable to expect compa-
outcome. Feedback is provided on each trial. rable performance between the two groups.
The association between the cues and out- Another possibility is that children with SLI
comes is probabilistic. For example, one set require more trials in order to implicitly
of cues will be associated with one outcome learn associations with equal proficiency to

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

294 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

non–language-impaired children. As noted and retrieval in SLI may be secondary to work-

previously, in the study by Kemény and ing memory problems and/or language diffi-
Lukács, participants received 150 trials, culties and do not reflect declarative memory
whereas in the study by Mayor-Dubois deficits. Indeed, it appears that declarative
et al. (2012), participants received 200 trials. memory functioning is similar in children with
To address these issues, additional research and without SLI. With respect to the procedu-
will be required to test these alternative ral memory system, implicit learning of verbal
explanations. information appears to be affected SLI. Thus,
in terms of future research directions, to fur-
Is the procedural memory system ther our understanding about the relationship
impaired in SLI? between memory and language problems in
At present, the available data indicate that SLI, it may be beneficial to broaden our focus
auditory/verbal aspects of procedural mem- beyond working memory to other memory
ory are likely to be impaired in SLI. The two systems.
studies involving auditory statistical learning There is a history of identifying underlying
both report poor implicit learning in SLI. How- deficits in children with developmental disor-
ever, when visually based probabilistic clas- ders such as SLI. This has informed and will
sification tasks are used, findings have been continue to inform clinical practice in speech
inconsistent. The difficulty in making general and language pathology. In this review, we
statements about these aspects of procedural identify potential difficulties in long-term
memory is that the evidence is very sparse. memory, such as implicit learning of verbal
Nonetheless, one trend that is emerging from material via the procedural memory system
this literature is that children with SLI ap- that clinically adds to our existing knowl-
pear to require more exposures to informa- edge about short-term and working memory
tion, irrespective of whether the information difficulties in SLI. This review, in addition,
is verbal or visual. Thus, at this stage it does identifies declarative memory as a system
not seem appropriate to characterize SLI as that appears to function relatively well in
a developmental disorder whereby informa- SLI. Declarative memory is a powerful and
tion cannot be implicitly learned. Rather, it flexible long-term memory system that can
seems that individuals with SLI require more have a compensatory role in language learning
exposures to the information. Consequently, (Ullman & Pierpont, 2005). This information
detecting differences between children with provides an evidence-based rationale for exist-
and without SLI on measures assessing the ing clinical practice, that is, the use of explicit
procedural memory system may require tasks learning strategies in therapeutic interven-
that assess implicit learning rather than tion. For example, when talking about yes-
retrieval. terday, put “ed” on the verb, such as played,
and if talking about today, it is “ing,” such
CONCLUDING REMARKS AND as playing. The evidence reviewed here also
CLINICAL IMPLICATIONS implies that clinical interventions that harness
the declarative memory system and also incor-
This review examined research investigat- porate approaches that are effective in learn-
ing the declarative memory system and as- ing through that system, for example, spaced
pects of the procedural memory system in SLI. presentations, imageable items, deep encod-
From this literature, it appears that in children ing, should result in more effective language
with SLI, learning is affected by functioning outcomes in children with SLI (Ullman &
in long-term memory systems in addition to Pullman, submitted for publication).
known deficits in working memory. In the A trend emerging from investigations
case of declarative memory, there is some evi- into the procedural memory system in SLI
dence to suggest that problems with learning highlights a potential benefit in increasing

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 295

the amount of language to which chil- systems is still very much in its infancy,
dren with SLI are exposed, presumably more experimental work and then ran-
above that encountered by typically develop- domized controlled trials of interventions
ing children. However, given that research are required before translating findings into
into declarative and procedural memory treatments.

REFERENCES

Alexander, G. E., & Crutcher, M. D. (1990). Functional Cambridge, United Kingdom: Cambridge Cognition
architecture of basal ganglia circuits: neural substrates Limited.
of parallel processing. Trends in Neurosciences, 13, Chapin, J. S., Busch, R. M., Naugle, R. I., & Najm, I. M.
266–271. (2009). The Family Pictures subtest of the WMS-III:
Alexander, G. E., DeLong, M. R., & Strick, P. L. (1986). Par- Relationship to verbal and visual memory following
allel organization of functionally segregated circuits temporal lobectomy for intractable epilepsy. Journal
linking basal ganglia and cortex. Annual Review of of Clinical and Experimental Neuropsychology, 31,
Neuroscience, 9, 357–381. 498–504.
Alloway, T. P., & Gathercole, S. E. (2012). Working Cohen, J. (1988). Statistical power analysis for the
memory and neurodevelopmental disorders. Hove: behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum.
Psychology Press. Cohen, M. J. (1997). Children’s memory scale. San
Alvarez, P., & Squire, L. R. (1994). Memory consolida- Antonio, TX: The Psychological Corporation.
tion and the medial temporal lobe: A simple network Conway, C. M., & Pisoni, D. B. (2008). Neurocognitive
model. Proceedings of the National Academy of Sci- basis of implicit learning of sequential structure and
ences, 91, 7041–7045. its relation to language processing. Annals of the
Archibald, L. M., & Gathercole, S. E. (2006a). Visuospatial New York Academy of Sciences, 1145, 113–131.
immediate memory in specific language impairment. Corkin, S. (2002). What’s new with the amnesic patient
Journal of Speech, Language, and Hearing Research, HM? Nature Reviews Neuroscience, 3, 153–160.
49, 265–277. Cumming, G., & Finch, S. (2005). Inference by eye:
Archibald, L. M., & Gathercole, S. E. (2006b). Short-term Confidence intervals and how to read pictures of
and working memory in specific language impair- data. The American Psychologist, 60, 170–180.
ment. International Journal of Language & Com- Delis, D. C., Kramer, J. H., Kaplan, E., & Ober, B. A.
munication Disorders, 41, 675–693. (1994). The California verbal learning test: Chil-
Aslin, R. N., Saffran, J. R., & Newport, E. L. (1998). Compu- dren’s version (CVLT-C). San Antonio, TX: The
tation of conditional probability statistics by 8-month- Psychological Corporation.
old infants. Psychological Science, 9, 321–324. *Dewey, D., & Wall, K. (1997). Praxis and memory
Baddeley, A. (2003). Working memory: Looking back and deficits in language-impaired children. Developmen-
looking forward. Nature Reviews Neuroscience, 4, tal Neuropsychology, 13, 507–512.
829–839. *Duinmeijer, I., de Jong, J., & Scheper, A. (2012).
*Baird, G., Dworzynski, K., Slonims, V., & Simonoff, E. Narrative abilities, memory and attention in children
(2010). Memory impairment in children with language with a specific language impairment. International
impairment. Developmental Medicine & Child Neu- Journal of Language & Communication Disorders,
rology, 52, 535–540. 47, 542–555.
Baron, I. S. (2004). Neuropsychological evaluation of Dulay, M. F., Schefft, B. K., Testa, S. M., Fargo, J. D.,
the child. Oxford: Oxford University Press. Privitera, M., & Yeh, H. (2002). What does the Family
*Bavin, E. L., Wilson, P. H., Maruff, P., & Sleeman, F. Pictures subtest of the Wechsler Memory Scale-III
(2005). Spatio-visual memory of children with spe- measure? Insight gained from patients evaluated for
cific language impairment: Evidence for generalized epilepsy surgery. The Clinical Neuropsychologist,
processing problems. Research report. International 16, 452–462.
Journal of Language & Communication Disorders, Durlak, J. A. (2009). How to select, calculate, and inter-
40, 319–332. pret effect sizes. Journal of Pediatric Psychology,
Cambridge Cognition. (1996). Cambridge neuropsy- 34, 917–928.
chological test automated battery (CANTAB). Eichenbaum, H. (2000). A cortical–hippocampal sys-
tem for declarative memory. Nature Reviews,
Neuroscience, 1, 41–50.
Eichenbaum, H. (2004). Hippocampus: cognitive pro-
*References marked with an asterisk indicate studies in- cesses and neural representations that underlie
cluded in the meta-analysis. declarative memory. Neuron, 44, 109–120.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

296 TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2013

Evans, J. L., Saffran, J. R., & Robe-Torres, K. (2009). Lum, J. A., & Bleses, D. (2012). Declarative and procedu-
Statistical learning in children with specific language ral memory in Danish speaking children with specific
impairment. Journal of Speech, Language, and language impairment. Journal of Communication
Hearing Research, 52, 321–335. Disorders, 45, 46–58.
Gabriel, A., Maillart, C., Guillaume, M., Stefaniak, N., Lum, J. A., Conti-Ramsden, G., Morgan, A., & Ullman, M.
& Meulemans, T. (2011). Exploration of serial T. Procedural learning deficits in specific language
structure procedural learning in children with impairment: A meta-analysis of serial reaction
language impairment. Journal of the International time task performance. Manuscript submitted for
Neuropsychological Society, 17, 336–343. publication.
Gabrieli, J. D. (1998). Cognitive neuroscience of human *Lum, J. A., Conti-Ramsden, G., Page, D., & Ullman,
memory. Annual Review of Psychology, 49, 87–115. M. T. (2012a). Working, declarative and procedural
Gathercole, S. E., & Alloway, T. P. (2005). Practitioner memory in specific language impairment. Cortex, 48,
review: Short-term and working memory impair- 1138–1154.
ments in neurodevelopmental disorders: Diagnosis *Lum, J. A., Conti-Ramsden, G., & Ullman, M. T.
and remedial support. Journal of Child Psychology (2012b). The role of verbal and nonverbal mem-
and Psychiatry, 47, 4–15. ory in the Family Pictures Subtest: Data from
Gathercole, S. E., & Baddeley, A. D. (1990). Phonological children with specific language impairment.
memory deficits in language disordered children: Is Child Neuropsychology. Retrieved from http:
there a causal connection? Journal of Memory and //[Link]/10.1080/09297049.09292012.09734294
Language, 29, 336–360. *Lum, J. A., Gelgec, C., & Conti-Ramsden, G. (2010). Pro-
Hartshorne, J. K., & Ullman, M. T. (2005). Why girls say cedural and declarative memory in children with and
‘holded’ more than boys. Developmental Science, 9, without specific language impairment. International
21–32. Journal of Language & Communication Disorders,
Hedges, L. V., & Vevea, J. L. (1998). Fixed-and random- 45, 96–107.
effects models in meta-analysis. Psychological Lum, J. A., Ullman, M. T., & Conti-Ramsden, G. (2013).
Methods, 3, 486. Procedural learning is impaired in dyslexia: Evidence
Heindel, W. C., Butters, N., & Salmon, D. P. (1988). from a meta-analysis of serial reaction time stud-
Impaired learning of a motor skill in patients with ies. Research in Developmental Disabilities, 34,
Huntington’s disease. Behavioral Neuroscience, 102, 3460–3476.
141–147. Mayes, A., Montaldi, D., & Migo, E. (2007). Associative
Hsu, H. J., & Bishop, D. V. M. (2011). Grammatical diffi- memory and the medial temporal lobes. Trends in
culties in children with specific language impairment: Cognitive Sciences, 11, 126–135.
Is learning deficient? Human Development, 53, 264. *Mayor-Dubois, C., Zesiger, P., van der Linden,
Joanisse, M. F., & Seidenberg, M. S. (1998). Specific M., & Roulet-Perez, E. (2012). Nondeclarative
language impairment: A deficit in grammar or pro- learning in children with Specific Language
cessing? Trends in Cognitive Sciences, 2, 240–247. Impairment: Predicting regularities in the visuo-
Karuza, E. A., Newport, E. L., Aslin, R. N., Star- motor, phonological, and cognitive domains.
ling, S. J., Tivarus, M. E., & Bavelier, D. Child Neuropsychology. Retrieved from http:
(2013). The neural correlates of statistical learn- //[Link]/10.1080/09297049.09292012.09734293
ing in a word segmentation task: An fMRI Montgomery, J. W., Magimairaj, B. M., & Finney, M. C.
study. Brain and Language. Retrieved from (2010). Working memory and specific language im-
[Link] pairment: An update on the relation and perspectives
*Kemény, F., & Lukács, Á. (2010). Impaired procedural on assessment and treatment. American Journal of
learning in language impairment: Results from Speech-Language Pathology, 19, 78–94.
probabilistic categorization. Journal of Clinical and *Nichols, S., Jones, W., Roman, M. J., Wulfeck, B., Delis,
Experimental Neuropsychology, 32, 249–258. D. C., Reilly, J., et al. (2004). Mechanisms of verbal
Knowlton, B. J., & Squire, L. R. (1995). Remembering and memory impairment in four neurodevelopmental
knowing: Two different expressions of declarative disorders. Brain and Language, 88, 180–189.
memory. Journal of Experimental Psychology: Nissen, M. J., & Bullemer, P. (1987). Attentional re-
Learning, Memory, and Cognition, 21, 699–710. quirements of learning: Evidence from performance
Knowlton, B. J., Squire, L. R., & Gluck, M. A. (1994). measures. Cognitive psychology, 19, 1–32.
Probabilistic classification learning in amnesia. Packard, M. G., & Knowlton, B. J. (2002). Learning
Learning & Memory, 1, 106–120. and memory functions of the basal ganglia. Annual
Lewis, S., & Clarke, M. (2001). Forest plots: Trying to Review of Neuroscience, 25, 563–593.
see the wood and the trees. British Medical Journal, Parent, A., & Hazrati, L. N. (1995). Functional anatomy of
322, 1479. the basal ganglia. I. The cortico-basal ganglia-thalamo-
Lezak, M. D. (2012). Neuropsychological assessment cortical loop. Brain Research Reviews, 20, 91–
(5th ed.). Oxford: Oxford University Press. 127.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
LWW/TLD TLD-D-13-00006 October 31, 2013 19:26

Long-Term Memory 297

Poldrack, R. A., Clark, J., Paré-Blagoev, E., Shohamy, D., Squire, L. R. (2004). Memory systems of the brain: A
Moyano, J. C., Myers, C., et al. (2001). Interactive brief history and current perspective. Neurobiology
memory systems in the human brain. Nature, 414, of Learning and Memory, 82, 171–177.
546–550. Squire, L. R., Stark, C. E. L., & Clark, R. E. (2004). The me-
Poldrack, R. A., & Packard, M. G. (2003). Competi- dial temporal lobe. Annual Review of Neuroscience,
tion among multiple memory systems: Converging 27, 279–306.
evidence from animal and human brain studies. Squire, L. R., & Zola, S. M. (1996). Structure and function
Neuropsychologia, 41, 245–251. of declarative and nondeclarative memory systems.
*Records, N. L., Tomblin, J. B., & Buckwalter, P. R. Proceedings of the National Academy of Sciences,
(1995). Auditory verbal learning and memory in 93, 13515–13522.
young adults with specific language impairment. The Tomblin, J. B., Mainela-Arnold, E., & Zhang, X. (2007).
Clinical Neuropsychologist, 9, 187–193. Procedural learning in adolescents with and without
*Riccio, C. A., Cash, D. L., & Cohen, M. J. (2007). specific language impairment. Language Learning
Learning and memory performance of children and Development, 3, 269–293.
with specific language impairment (SLI). Applied Tulving, E., & Markowitsch, H. J. (1998). Episodic
Neuropsychology, 14, 255–261. and declarative memory: role of the hippocampus.
Saffran, J. R. (2003). Statistical language learning Hippocampus, 8, 198–204.
mechanisms and constraints. Current Directions in Ullman, M. T. (2001). A neurocognitive perspective
Psychological Science, 12, 110–114. on language: The declarative/procedural model. 2,
Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996). 717–726.
Statistical learning by 8-month-old infants. Science, Ullman, M. T. (2004). Contributions of memory circuits
274, 1926–1928. to language: The declarative/procedural model.
Saffran, J. R., Newport, E. L., Aslin, R. N., Tunick, R. A., Cognition, 92, 231–270.
& Barrueco, S. (1997). Incidental language learning: Ullman, M. T. (2006). Is Broca’s area part of a
Listening (and learning) out of the corner of your ear. basal ganglia thalamocortical circuit? Cortex, 42,
Psychological Science, 8, 101–105. 480–485.
*Shear, P. K., Tallal, P., & Delis, D. C. (1992). Verbal Ullman, M. T., & Pierpont, E. I. (2005). Specific lan-
learning and memory in language impaired children. guage impairment is not specific to language: The
Neuropsychologia, 30, 451–458. procedural deficit hypothesis. Cortex, 41, 399–
Sheslow, D., & Adams, W. (1990). Wide range assess- 433.
ment of memory and learning administration Ullman, M. T., & Pullman, M. Y. A compensatory role
manual.. Wilmington, DE: Jastak Associates. for declarative memory in neurodevelopmental
Squire, L. R. (1992). Declarative and nondeclarative disorders. Manuscript submitted for publication.
memory: Multiple brain systems supporting learning Van der Lely, H. K. (2005). Domain-specific cognitive
and memory. Journal of Cognitive Neuroscience, 4, systems: Insight from Grammatical-SLI. Trends in
232–243. Cognitive Sciences, 9, 53–59.

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.