Elsevier

Cortex

Volume 134, January 2021, Pages 145-161
Cortex

Research Report
Understanding differing outcomes from semantic and phonological interventions with children with word-finding difficulties: A group and case series study

https://doi.org/10.1016/j.cortex.2020.09.030Get rights and content

Abstract

Developmental Language Disorder occurs in up to 10% of children and many of these children have difficulty retrieving words in their receptive vocabulary. Such word-finding difficulties (WFD) can impact social development and educational outcomes. This research aims to develop the evidence-base for supporting children with WFD and inform the design and analysis of intervention studies.

We included 20 children (age 6 to 8) with WFD each of whom participated in two interventions one targeting semantic attributes and the other phonological attributes of target words. The interventions, employing word-webs, were carefully constructed to facilitate direct comparison of outcome which was analysed at both group and case-series level.

The study used a robust crossover design with pre-intervention baseline, between–intervention wash-out and post-intervention follow-up testing. We incorporated: matching of item sets on individual performance at baseline, independent randomisation of order of intervention and items to condition, blinding of assessor, evaluation of fidelity and control items. The interventions were clinically feasible, with weekly sessions over six weeks.

Intervention improved children's word-finding abilities with statistically significant change only during treatment phases of the study and not over baseline, wash-out or follow-up phases. For the group the semantic intervention resulted in a gain of almost twice as many items as the phonological intervention, a significant difference.

However, children differed in their response to intervention. Importantly, case-series analysis revealed outcomes predictable on the basis of children's theoretically driven language profiles. Taking account of individual profiles in determining choice of intervention would enable more children to benefit.

The study provides new evidence to inform and refine clinical practice with this population. Future studies should be designed such that results can be analysed at both group and case series levels to extend theoretical understanding and optimise use of appropriate interventions.

Introduction

The aim of this paper is to report findings from a study in which children with WFD participated in two different interventions, and also to explore the value of presenting and analysing data from a case series as well as at the group level. We first highlight the need and rationale for the intervention study (1.1) and then consider the research design (1.2). Finally, we provide a theoretical model in the domain of the study, word production, within which findings are interpreted and clinical implications determined (1.3) and outline the resulting research questions (1.4).

Low language (≥1.25 SD below the mean) is found in up to a fifth of 7 year olds and is associated with concurrent literacy and school difficulties (McKean et al., 2017). If a child has a language difficulty on school entry it is likely to persist (Stothard, Snowling, Bishop, Chipchase, & Kaplan, 1998). An international consensus study agreed on the term ‘developmental language disorder’ (DLD) for children “who are likely to have language problems enduring into middle childhood and beyond, with a significant impact on everyday social interactions or educational progress” (p.1070, Bishop, Snowling, Thompson, & Greenhalgh, 2016). Children with DLD have heterogeneous profiles, yet within this variety sub-groups with particular areas of language strength and need can be identified. One such group is children with word-finding difficulties (WFD), which occur when a child is unable to produce words despite having an understanding of their meaning. WFDs occur in up to a quarter of children attending language support services (Dockrell, Messer, George, & Wilson, 1998) and have implications for education and social development (e.g., Messer, Dockrell, & Murphy, 2004). Thus, language difficulties are prevalent in childhood and word-finding difficulties (WFD) occur in many children with DLD. Given the impact of WFD on development it is crucial to investigate interventions for children with this difficulty.

There is a growing evidence base demonstrating that interventions for children with WFD can be effective in improving retrieval for treated words. Research studies tend to provide intervention focusing either on semantics, exploring the meaning of treated items (e.g., Ebbels et al., 2012), or on phonology/orthography, exploring an item's spoken or written form (e.g., McGregor, 1994). A range of activities have been used, including categorisation tasks, odd one out tasks and word-webs where features of meaning or form are requested or provided (e.g., semantic category, first sound, etc.).

Due to heterogeneity in the nature of WFD (see section 1.3), it is important to determine which approach works best for which child/ren. There are very few studies which compare interventions directly, but Bragard and colleagues' 2012 case series study with four participants with WFD is an exception. The results showed long-lasting effects of intervention with a pattern of response that could be related to the nature of the children's difficulties. Clinical practice is likely to be informed by such studies. In the meantime, in our experience (which combines to over half a century of working in the UK with children with DLD and training students to do the same) clinical practice is eclectic, with many practitioners using approaches including word-webs (the word-webs used in this study are described in detail in section 2.4) which entail information about both meaning and form of targeted items. In sum, whilst the research evidence is encouraging there is very little that contrasts different approaches directly or which relates outcome to the nature of children's language profiles.

Research designs in the field of intervention for DLD vary. They include experimentally controlled single case and case-series studies, and randomised controlled trials (RCTs). In each case there are debates over the optimum way of exercising experimental control and analysing the research findings (Gold et al., 2017; Tate et al., 2015). Given the lack of consensus, it is important to compare findings from different designs and to explore the value of the different approaches.

The group RCT is held to be the ‘gold standard’ in medically related research, and increasingly in education research. This has a profound influence on the likelihood of obtaining research funding. Group studies with randomisation to different treatment and control conditions have many strengths, including generalisation of findings to others meeting the inclusion criteria and enabling combination of effect sizes across studies in systematic reviews.

Systematic reviews and ‘n-of-1 trials’ share the ‘top-spot’ in the Oxford Centre for Evidence Based Medicine's hierarchy of evidence in order to answer the question ‘does this intervention help?’ ‘N of 1 trials’ refer to a variety of designs which investigate the effect of treatment with one individual (Shamseer et al., 2015). A key point highlighted by these authors is that findings for a group may not be true for all the individuals within the group. In order for us to determine progress over time compared with progress with an intervention at an individual level, more than one (and preferably several) pre-intervention baselines are required (Howard, Best, & Nickels, 2015).

There are differences in what can be concluded from these contrasting research designs. This is important because discrepancies in the interpretation of findings from group and single case studies may result in different theoretical and practical implications. Investigating outcome at these different levels within the same intervention study is novel and likely to have important implications for the design of future studies.

Within the vast literature on language production there is agreement that there are at least three levels of processing necessary for single word production: semantic, lexical and phonological (e.g., Dell, Schwartz, Martin, Saffran, & Gagnon, 1997). This theoretical framework is discussed in detail in relation to WFD in children in Best (2005) and (Friedmann, Biran, & Dotan, 2013). During the widely used task of picture naming, for example, there will be activation of semantic representations. This will, in turn, activate representations in a phonological output lexicon. There is then a post-lexical stage of processing during which phonological forms are assembled for production. In the present research, we employ the Restricted Interaction Account (RIA, Goldrick & Rapp, 2002) in which activation is largely feed-forward but there is also feedback from phonological representations to the output lexicon. In summary, within a child's developing language system there are several levels of processing necessary for word-retrieval and production, each of which may have developed typically or atypically.

The results of the two different interventions employed in the present study are interpreted in relation to three potential sub-types of difficulties in word-retrieval, with the recognition that individual children may experience difficulty with multiple levels of processing; here we aim to identify the principle level of difficulty for each child. It is important at this point to note that children were not recruited to the current study on the basis of having different profiles; instead these were identified and used in analyses after the intervention study in order to look for potential explanations of the different outcomes. We consider three sub-types: (i) classic WFD, (ii) semantic difficulties and (iii) phonological difficulties (in the context of strong semantic processing).

Classic WFD (sub-type i) occur when children have identified difficulties in word-finding despite relatively good semantic processing and good phonological processing. In this case the difficulty is in accessing the phonological form for production (see, e.g., Best, 2005).

WFDs may reflect less elaborate semantic representations (sub-type ii). For example, McGregor and colleagues investigated the drawings and definitions provided by children with DLD and found less detail was provided for items for which children made semantic errors in naming (McGregor & Appel, 2002; McGregor, Newman, Reilly, & Capone, 2002). In line with this, some children with WFDs show relative strengths in phonological awareness and decoding whilst demonstrating difficulty with reading comprehension and semantic fluency tasks (Messer & Dockrell, 2013; Messer et al., 2004).

Finally, for other children, WFDs appear to stem from limitations with phonological processing, i.e., at a later stage within the language production process (sub-type iii). For example, Constable, Stackhouse, and Wells (1997) used a psycholinguistic framework approach and present a careful single case study of a seven-year-old who had no observable semantic difficulties but pervasive difficulties with tasks requiring phonological processing. German and Newman (2004) found the naming errors of children with WFDs were influenced by the post-semantic psycholinguistic variables of word frequency and phonological neighbourhood density, providing further support for a later-stage account.

While there is an increasing evidence base for interventions for WFDs and this spans semantic (e.g., Ebbels et al., 2012) and phonological/orthographic approaches (e.g., Best, 2005; McGregor, 1994), we remain unclear about which approach is appropriate for an individual child. This is particularly important in the context of limited resources and the impact WFD can have on everyday communication and education. Group studies have produced conflicting results (Wing, 1990; Wright, Gorrie, Haynes, & Shipman, 1993). Bragard, Schelstraete, Snyers, and James (2012) were able to provide some evidence that children may respond differently according to their background language profile. They reported that, of the four children in their study, the two with more semantically based difficulties responded better to phonological intervention, whereas one participant with apparently more phonologically based WFDs responded more favourably to semantic intervention. However, this study involved only four children, one of whom did not show a positive response to intervention of either type. We consequently have little direct evidence on the optimal intervention approach for a child given their language profile.

We can use models of language production to support the development of theoretically motivated predictions as to the outcome of intervention given the strengths and difficulties of an individual child. The current study, termed ‘WORD’ (WOrd Retrieval and Development), is the first to include a large number of children with WFD, each of whom participated in two types of intervention. It therefore provides us with a unique opportunity to explore how the outcome of different interventions relates to language profiles.

If a child receives multiple interventions, there is the possibility that the order in which the interventions are deployed will have an effect. There is limited evidence regarding whether the order in which semantic and phonological interventions occur can influence outcome. Zens, Gillon, and Moran (2009) reported that for new word-learning, children with language impairment who participated in a phonological awareness intervention followed by semantic intervention showed improved word production of the new forms while the reverse order did not result in improved performance at post-test, and neither order impacted on the comprehension of new words (Zens et al., 2009). Since models of word production describe the process as involving the use of semantic information to access phonological forms, it would seem that there is sense in predicting the opposite result i.e., that semantic intervention preceding phonological intervention may be more effective than the reverse. In conclusion, while order effects may be present, there is no strong evidence on the direction these may take. If effects are present there will be clinical implications for the order of treatment approaches.

The overarching research question addressed by the study is whether for children with WFDs, naming skills will improve with WORD intervention. The intervention is described in detail in 2.4. A second key question is whether, when the results are considered at different levels (group and case series), the conclusions differ.

Hypotheses are divided into sections: hypotheses addressed by data from the group (1.4.1) and case series (1.4.2), hypotheses addressed by considering the children in theoretically informed sub-groups (1.4.3) and a hypothesis pertaining to the design and level of analysis (1.4.4).

  • i)

    children's naming will improve as a result of the WORD intervention, specifically there will be significant improvement over the intervention phases of the study and not over the non-intervention phases

  • ii)

    there will be a difference between the effect of the two interventions (semantic and phonological)

  • iii)

    the order in which the semantic and phonological interventions occur will influence the outcome

  • iv)

    the children will differ from one another in their response to intervention

  • i)

    individual children will show greater change during therapy phases for treated items (n = 50)

  • ii)

    individual children will show greater change during the semantic therapy phase of the study for the sub-sets of items given semantic therapy (n = 25).

  • iii)

    individual children will show greater change during the phonological therapy phase of the study for the subsets of items given phonological therapy (n = 25).

Predictions for the outcomes for children with different language profiles will be outlined in turn and are summarised in Fig. 1.

  • (i)

    classic WFD

Children with this profile are predicted to benefit from the semantic or phonological intervention or both. By definition their difficulty is in accessing phonology from meaning and providing and processing either additional semantic or phonological information may aid their word retrieval.

  • (ii)

    semantic difficulties

Children with this profile are predicted to benefit from semantic intervention but not to benefit from phonological intervention. This is because they may not have adequate information about an item's meaning to drive retrieval, therefore, the provision and processing of semantic information is likely to aid retrieval of the word form. In contrast, information about phonology will not help if this later stage of processing is not activated for hard to retrieve items.

  • (iii)

    phonological difficulties (in the context of strong semantic processing)

Children whose profiles show only phonological difficulties are predicted to benefit from phonological intervention but not to benefit from semantic intervention. This is because, by definition, they have relatively good knowledge of word-meanings, in their case benefit is likely to derive from providing and processing additional phonological information.

  • (i)

    The results from the group analysis will provide an understanding of outcomes on average which may be generalised to other children with WFD meeting the inclusion criteria.

  • (ii)

    Consideration of heterogeneity in individual children's profiles and outcomes may enable more specific theoretically and clinically informative conclusions.

Section snippets

Materials and methods

We report all measures in the study. how we determined our sample size, all inclusion/exclusion criteria (which were established prior to data analysis) and all manipulations.

Group analysis

A mixed ANOVA was carried out with one within-subjects variable, TIME, with 7 levels: assessment from A1 to A7, and one between-subjects variable, GROUP, with 2 levels according to which of the two interventions happened first. Significant main effects were followed up with pairwise comparisons. A homogeneity test was employed to determine whether the individual children responded in the same way to the interventions.

Case series analysis

Analysis of the case series data employed established item-by-item statistics

Fidelity

A fidelity checklist included number of sessions, length of intervention period for each type of therapy (semantic or phonological), session length and key aspects of the protocol. Twenty-four intervention sessions were selected and examined according to a pre-determined template. Nineteen children of the twenty included (95%) received exactly six intervention sessions for each type of intervention, occurring over an approximate six-week period (mean length of intervention period for each

Overview of findings in relation to predictions

The findings from 20 children with WFD, each of whom participated in two types of intervention, provide us with new insights into outcomes at group and case-series level. The intervention improved children's word-finding abilities as measured by picture naming with statistically significant change occurring only during treatment phases of the study. At the group level, significantly more benefit resulted from the semantic than from the phonological word-webs. This is in line with the success of

Conclusions

The study provides a theoretical framework and evidence-based recommendations that aim to optimise the time spent on key lexical features to help children find the words with which they are struggling. Generalisation of word learning and retrieval beyond trained items continues to be a key challenge for researchers and clinicians. Analysis of case series data can reveal patterns beyond those that emerge from analysis at group level and can produce elevated rates of successful outcomes.

Author contributions

WB led the WORD (Word Retrieval and Development) project. Funding was secured by PI WB, and Co-PIs JM and MT. The intervention strand reported in this paper was conceptualised by WB and LH, with involvement of the wider advisory group. All co-authors contributed to an aspect of study development. Research design was partially inspired by related work with adults with aphasia carried out by DH. Data collection: intervention LH, KS and WB, assessment AK, LH, WB, JM. WB supervised the work of LH

Funding sources

This work was supported by the ESRC [grant number RES-062-23-2721, UKRI grant ref. ES/H046925/1].

Open practices

The study in this article earned an Open Data – Protected Access badge for transparent practices. Raw data on core assessment tasks from typically developing participants and children with WFD are available via UK Dataservice: https://beta.ukdataservice.ac.uk/datacatalogue/studies/study?id=851771.

Acknowledgements

We are grateful to the children involved in the project and to their parents and teachers. Several speech and language therapists and researchers contributed to data collection during the wider study: Vivien Gibson, Liory Fern-Pollak, Jo Piper, Silvia Roncoli and Donna Lynn Shepherd. We are especially grateful to our expert advisory panel Kathleen Cavin and Susan Ebbels and our wider advisory group. Colleagues who have contributed to the wider WORD project of which this forms a part include

References (46)

  • S.M. Gold et al.

    Control conditions for randomised trials of behavioural interventions in psychiatry: A decision framework

    The Lancet Psychiatry

    (2017)
  • M. Goldrick et al.

    Lexical and post-lexical phonological representations in spoken production

    Cognition

    (2007)
  • W. Best

    Investigation of a new intervention for children with word-finding problems

    International Journal of Language and Communication Disorders

    (2005)
  • W. Best et al.

    Intervening to alleviate word-finding difficulties in children: Case series data and a computational modelling foundation

    Cognitive Neuropsychology

    (2015)
  • D.V.M. Bishop

    Test for the reception of grammar

    (1996)
  • W. Best et al.

    Intervention for children with word-finding difficulties: A parallel group randomised control trial

    International Journal of Speech-Language Pathology

    (2017)
  • D.V.M. Bishop et al.

    When words fail us: Insights into language processing from developmental and acquired disorders

    Philosophical Transactions of the Royal Society B: Biological Sciences

    (2014)
  • D.V.M. Bishop et al.

    CATALISE: A multinational and multidisciplinary delphi consensus study of problems with language development. Phase 2. terminology

    PeerJPreprints

    (2016)
  • A. Bragard et al.

    Word-finding intervention for children with specific language impairment: A multiple single-case study

    Language Speech and Hearing Services in Schools

    (2012)
  • L. Campbell et al.

    The effectiveness of semantic intervention for word-finding difficulties in college-aged students (16–19 years) with persistent Language Disorder

    Autism & Developmental Language Impairments

    (2019)
  • A. Constable et al.

    Developmental word-finding difficulties and phonological processing: The case of the missing handcuffs

    Applied Psycholinguistics

    (1997)
  • G.S. Dell et al.

    Lexical access in aphasic and nonaphasic speakers

    Psychological Review

    (1997)
  • J.E. Dockrell et al.

    Notes and discussion: Children with word-finding difficulties prevalence, presentation and naming problems

    International Journal of Language and Communication Disorders

    (1998)
  • J. Druks et al.

    Object and action naming Battery

    (2000)
  • S.H. Ebbels et al.

    Effectiveness of semantic therapy for word-finding difficulties in pupils with persistent language impairments: A randomized control trial

    International Journal of Language and Communication Disorders

    (2012)
  • C.D. Elliot et al.

    British ability scales

    (1996)
  • F. Faul et al.

    G∗Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences

    Behavior Research Methods

    (2007)
  • K.I. Forster et al.

    DMDX: A windows display program with millisecond accuracy

    Behavior Research Methods, Instruments, & Computers

    (2003)
  • N. Friedmann et al.

    Lexical retrieval and its breakdown in aphasia and developmental language impairment

  • E. Funnell et al.

    Age of acquisition for naming and knowing: A new hypothesis

    Quarterly Journal of Experimental Psychology

    (2006)
  • S. Gathercole et al.

    Children's test of Nonword repetition

    (1996)
  • D.J. German

    Word-finding intervention for children and adolescents

    Topics in Language Disorders

    (1992)
  • D. German

    Test of word finding (TWF-2)

    (2000)
  • View full text