APHASIOLOGY, 2009, 23 (11), 1338–1352

PAPH0268-70381464-5041APHASIOLOGY,APHASIOLOGY Vol.A 0, No. 0, June 2009:meta-analysis pp. 0–0 of word-finding treatments for

Meta-AnalysisWisenburn and of Mahoney Word-Finding Bruce Wisenburn and Kate Mahoney State University of New York at Fredonia, NY, USA

Background : The research literature on treatment methods for word-finding deficits in aphasia is extensive. A meta-analysis of studies for word-finding therapy was conducted in order to objectively synthesise this information to answer large-scale questions of treatment efficacy. Aims : The purpose of this study was to examine the efficacy of various treatment approaches for word-finding deficits for individuals with aphasia. This analysis also exam- ined gains made to trained and untrained words, the level of maintenance after therapy, and the effect of the time post-onset of aphasia on the recovery of function. Methods & Procedures : Various search methods were used to gather anomia treatment studies for this analysis. From 44 studies, 107 effect sizes were calculated for the final analysis. These data were sorted according to the following moderator variables: treat- ment category (semantic, phonological, or mixed), word set (trained, exposed-related, exposed-unrelated, unexposed-unrelated, and unexposed-related), follow-up measures, and median number of months post-onset. Outcomes & Results : All therapy approaches showed evidence of efficacy, although the variance between studies was large. Strong gains were seen for trained and exposed words, but only minor gains for unexposed words. Large effects were seen for up to 2 months post-therapy, with lingering effects at 3 months post-therapy. Treatment appeared efficacious even for individuals that were years post-onset. Conclusions : An objective synthesis of the literature shows that intervention for word-finding deficits is efficacious. However, the level of gains varied widely across studies and therapy approaches. As expected, little generalisation was found for untrained-unexposed words. Keywords: Anomia; Treatment; Meta-analysis. Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Word-finding deficits are one of the most common characteristics of aphasia (Goodglass & Wingfield, 1997, Raymer, 2005), and may be the only obvious symptom in cases of (Brookshire, 2007). As a result, this disorder

Address correspondence to: Bruce Wisenburn, 2300 Adams Avenue, Department of Communication Sciences and Disorders, Marywood University, Scranton, PA 19509-1598, USA. E-mail: wisenburn@ marywood.edu The authors would like to thank Kevin Kearns for reviewing this manuscript and offering helpful recommendations. We also wish to thank the following students for searching and reviewing the literature in preparation for this manuscript: Jessica Brocki, Justin Carr, Katie Catania, Jennifer Dechert, Elizabeth Derx, Shannon DeSantis, Erin Fenar, Rachel Garguilo, Amy Gibson, Cody Goodwin, Sarah Hunt, Keri Jakubowski, Helen Johnson, Lindsay Johnson, Melissa Kron, Melanie Lescynski, Meaghan Linehan, Delaney Lowery, Elizabeth Murphy, Shannon O’Leary, Laurie Pollinger, Pam Prentice, Amy Pugh, Jennifer Rosney, Lindsey Taylor, Kim Uminski, Elizabeth Wilger, Sarah Williams, Lindsay Zoldos, Alycia Zwahlen, and Christie Zyhowski. © 2009 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business http://www.psypress.com/aphasiology DOI: 10.1080/02687030902732745 META-ANALYSIS OF WORD-FINDING 1339

has been heavily researched, and numerous treatment methods have been proposed and studied. The vast literature may prove overwhelming to both researchers and clinicians who seek to compare the evidence for a variety of treatment approaches. Other variables in research studies also add to the confu- sion, such as participant differences in severity, type of aphasia, and time post- onset. In this paper we have conducted a search of the literature in order to gather and synthesise, using meta-analysis, efficacy studies related to treatment for word-finding deficits. Numerous sources provide a detailed summary of the literature (e.g., Li, 1996; Raymer, 2005; Wilshire & Coslett, 2000). Nickels and Best (1996a) and Nickels (2002a) provided a more detailed interpretation of the research literature to draw clinical conclusions. Nickels and Best (1996a) conducted a thorough review of both facilitation and therapy experiments related to word-finding deficits. Facilitation studies examined the effects on confrontational naming of a single therapy application, while therapy studies looked at the long-term effects of repeated applications of one or more techniques (Nickels & Best, 1996a; Howard, Patterson, Franklin, Orchard-Lisle, & Morton, 1985). The review of anomia therapy by Nickels and Best (1996a) covered 19 group and single-case designs throughout the 1980s and 1990s. Nickels and Best further analysed the treatment approach (semantic or phonological) in comparison with the participant charac- teristics. Their conclusion was that both semantic and phonological therapies were efficacious, but that semantic therapy appeared to show more generalisa- tion to untreated items. Yet the relationship of participant characteristics to optimum therapy type was still unclear, as was the pattern of generalisation. Nickels and Best (1996a) also noted other unanswered questions, such as the effect of combined semantic and phonological therapy, and of errorless versus errorful therapy. Nickels (2002a) reviewed single-case studies published between 1980 and 2002. The experiments reviewed directly compared the participant characteristics, treat- ment methods, and outcome results for treated and untreated words. Nickels con- cluded that the effectiveness of treatment for word-finding deficits was thoroughly established. However, therapeutic success was largely for words treated in ther- apy. Generalisation of therapeutic effects to untreated items was far less frequent. Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Nickels stated that instances of generalisation might have been due to participant exposure to the untreated items during repeated probes throughout the experi- ments, a view also supported by Howard (2000), and Nickels (2002b). Nickels (2002a) could not find a clear relationship from the literature between participant characteristics and success with particular treatment tasks, and expressed pessimism that such an answer could be definitively found considering individual variability in numerous areas that may affect naming skills. Nickels did find evid- ence that individuals often benefited from a combined semantic and phonological approach, and that these benefits are typically long lasting for the treated items. Semantic therapy by itself appeared to be most beneficial for participants with semantically based word-finding deficits. Nickels also tentatively concluded that phonological therapy by itself can also be successful, with some participants showing generalisation to untrained words. Although these studies have effectively summarised the literature, these qualitative conclusions must be sup- ported by objective means of synthesising the research (Robey & Dalebout, 1998; Robey & Schultz, 1998). 1340 WISENBURN AND MAHONEY

Meta-analysis is a method proposed by Glass (1976) to integrate and summarise findings from a body of research. Meta-analysis is the statistical analysis of a collec- tion of individual studies. As Glass (1976, p. 3) states: Meta-analysis refers to the analysis of analyses. I use it to refer to the statistical analysis of a large collection of results from individual studies for the purpose of integrating the findings. It connotes a rigorous alternative to the casual, narrative discussions of the research studies which typify our attempts to make sense of the rapidly expanding research literature. Glass’s approach advocated that researchers include as many studies as possible. As Glass, McGaw, and Smith (1981, p. 64) put it: The goal of the meta-analyst should be to provide an accurate, impartial, quantitative description of the findings in a population of studies on a particular topic . . . No survey would be considered valid if a sizable subset (or stratum) of the population was not represented in the cumulative results. Neither should a meta-analysis be considered complete if a subset of its population is omitted. Meta-analytic techniques may be particularly useful for the field of aphasia, as the literature consists of numerous studies with a small number of participants. Signifi- cant findings in these studies are often difficult to achieve due to the low power level, and may not be calculated for single-participant designs. A meta-analysis, however, evaluates the effect size, rather than the significance of studies. While tests of signifi- cance may suggest whether there is a difference between pre- and post-test results, the effect size is a calculation of the size of the difference between the pre- and post-test results. Thus, effect size may indicate more accurately how efficacious a therapy is, not just whether it is efficacious. A few meta-analyses that investigated actual effect sizes for aphasia therapy have been published (Greenhouse et al., 1990; Robey, 1994, 1998; Robey, Schultz, Crawford, & Sinner, 1999; Whurr, Lorch, & Nye, 1992). These studies have concluded that aphasia therapy has shown an overall positive effect. However, these meta-analyses have focused on therapy for a wide range of communi- cation disorders due to aphasia, such as , comprehension deficits, and simple verbal production deficits. Although these studies have shown strong evidence of aphasia therapy in general, meta-analytic techniques may now be needed to show

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 the efficacy of specific treatments for a specific characteristic of aphasia. The present meta-analysis is focused specifically on therapy for word-finding deficits, as measured by confrontational naming tasks. Moderator variables related to the gains made in therapy are analysed to gain an understanding of how treatment affects recovery. This research sought to answer the following five research questions about word- finding therapy, as measured by confrontational naming tasks: (1) What therapies work best for word-finding? (2) Are the gains only for trained words? (3) Do the gains persist after therapy? (4) How is the time post-onset related to the level of gains? (5) What ther- apies generalise best to unexposed words? Meta-analysis of word-finding treatment studies was chosen as the best method of inquiry to answer the research questions.

METHOD Selecting the studies In an effort to cast a wide net around all naming studies we conducted our library search focusing on all aphasia therapy studies. Thus, we searched ERIC, MEDLINE, META-ANALYSIS OF WORD-FINDING 1341

PsychArticles, PsychInfo, PubMed, and EBSCOHost for aphasia therapy studies with no restriction on publication date. The following search terms were used in various combinations: anomia , word-finding , word retrieval , word recall , treatment , therapy , and efficacy . In addition, the table of contents for each issue of journals related to the field were reviewed for the previous 10 years. The search also included reference lists for recent textbooks, literature reviews, and articles related to aphasia therapy. Approximately 500 studies were identified and reviewed. These studies were then categorised according to the specific characteristic focused upon in therapy, such as agrammatism, paraphasic productions, comprehension deficits, and word- finding deficits. From these, 106 studies were found to be related to treatment for word-finding deficits. This analysis applied Glass’s original advice (1976) to cast the widest net possible in selecting studies. Therefore strict selection criteria were not applied to the studies, so that the maximum number of studies could be included in the analysis. Studies were included in the present meta-analysis according to the following selection crite- ria: They (1) involved at least two aphasic patients; (2) used a picture confrontational naming task as the dependent measure; (3) included statistical details needed to perform the meta-analysis; (4) measured one treatment at a time and provided a description of the treatment; and (5) used word sets with baseline percent correct > 0. We focused specifically on studies with n > 1, as this has been the precedent for most previous meta-analyses (e.g., Robey, 1998), and including studies where n = 1 was not within the scope of this study. However, it is possible to calculate an effect size for single-participant studies with n = 1, as explained by Beeson and Robey (2006). Future meta-analyses on word-finding deficits may focus specifically on studies using a single-participant design with n = 1. For this study we did not eliminate studies based on methodological adequacy. In consideration of methodological adequacy about a priori selection of studies, we fol- lowed Glass’s original advice to cast the widest net in beginning a meta-analysis (Glass, 1976; Glass et al., 1981). Although picture confrontational naming tasks have shown questionable validity in relation to functional word finding in connected speech (Mayer & Murray, 2003), these tasks were found to be strongly correlated to the severity of aphasia. Confrontational naming tasks were selected for this analysis because of their near-universal use in aphasia treatment research. In addition, the Downloaded By: [University of Oregon] At: 22:07 27 March 2010 analysis of studies with the same outcome measure (i.e., confrontational naming tasks) may allow for a higher level of confidence when comparing effect sizes. In our search of the literature we found only a single word-finding efficacy study that did not use confrontational naming as a measurement tool. Based on previously stated criteria, the number of articles used in this analysis was 47 (later reduced to 44 after outliers were omitted).

Coding the studies Once studies were identified, selected characteristics were coded for the following four variables: treatment approach (semantic, phonological, or mixed), word set (trained, related-exposed, unrelated-exposed, unrelated-unexposed, and related- unexposed), follow-up measures, and median number of months post-onset of aphasia. Because therapy types are often described differently, special caution was taken to code therapy type according to its linguistic focus: semantic, phonological, or mixed. Examples of semantic therapy techniques included the following: sorting words into 1342 WISENBURN AND MAHONEY

categories, defining words, describing words, selecting words with semantic foils, answering semantic questions about words, using words in a sentence, and the use of gestures. Therapies that focused on syntactic properties, such as argument structure verb retrieval treatment (Schneider & Thompson, 2003), were also included in the semantic therapy category. Phonological techniques included repeating words, read- ing and writing words in isolation, phonemic cues, syllable cues, guessing sounds or number of syllables, and orthographic cues. Mixed techniques combined any semantic and phonological techniques listed, and also included functional therapies, such as role- playing or using target words in conversation. Confrontational picture naming was used as a therapy procedure in some studies (such as Miceli, Amitrano, Capasso, & Caramazza, 1996) or as a frequent probe, as for most studies that were of single- participant design. Confrontational naming involves both the phonological and semantic system, and so may be considered as a mixed therapy task (Nickels, 2002a). However, confrontational naming tasks, even if used as a therapy procedure, were omitted as a consideration for treatment categorisation. This was due to their preva- lent use throughout all of the studies, either for treatment or as a frequent probe. Consideration of confrontational naming for treatment categorisation would have rendered most studies as “mixed”, making it difficult to separate studies that were primarily semantic versus those that were phonological. Therefore, studies that utilised semantic tasks in combination with confrontational naming tasks were cate- gorised as semantic; studies with phonological tasks combined with naming tasks were categorised as phonological therapy. The word set used for assessment was divided into five categories: trained, related- exposed, unrelated-exposed, unrelated-unexposed, and related-unexposed. The trained set was directly used in therapy with feedback provided to the participant. A word set was considered to be exposed if it was probed numerous times between baseline and the final measurement at the conclusion of therapy. However, unlike the trained set, no feedback was provided to the participant as to the correctness of the response to the exposed word sets. Exposed word sets that were related to the trained word set were most frequently semantically related, but were also phonologically related (i.e., rhymed words) in one study (i.e., Raymer, Thompson, Jacobs, & Le Grand, 1993). Unexposed sets consisted of both standardised tests (such as the Boston Naming Test and the Philadelphia Naming Test) and non-standardised sets Downloaded By: [University of Oregon] At: 22:07 27 March 2010 that were administered in only a few instances, usually only at the start and comple- tion of therapy. In only one study (Pring, Hamilton, Harwood, & Macbride, 1993) was the unexposed word set related to the trained set.

Calculating effect size A computer program called Meta-Stat (Rudner, Glass, Evartt, & Emery, 2002) was used to perform this meta-analysis. This software is a comprehensive package designed to help in the meta-analysis of research studies in the social and behavioural sciences. Meta-Stat transforms study findings into a common metric, called an effect size, so that studies can be compared. Three sources of data for calculating effect size were used in this study: the standardisation mean difference formula (substituting the pre-test as the control group and the post-test as the experimental group), the F Statistic, and the Paired T Statistic. To evaluate the baseline to post-therapy gains, all effect sizes were calculated using within-group comparisons. Whenever the information was available, we calculated the mean scores on confrontational naming META-ANALYSIS OF WORD-FINDING 1343

tasks immediately before and after therapy. We also calculated the standard devi- ation of the pre-therapy scores between participants. The formula used for stand- ard effect size was d = ( Mpost tx – Mpre tx) / SD pre tx and the formula used to calculate an unbiased effect size was unbiased d = (1 – (3 / ((4 * n) – 9))) * d. When the pre- and post-therapy scores for each participant were unavailable, the F Sta- tistic or Paired T statistic was entered into Meta-Stat in order to calculate the unbiased effect size. The magnitude of an effect size indicates the difference in units of the standard deviation of the post-test when compared to the pre-test, with an adjustment by the unbiased effect size to lower the magnitude as the number of participants is decreased. A positive effect size indicates that the group did better at the post-test than at the pre-test, while a negative effect size indicates that the group fared better before the treatment than after. The unbi- ased effect size was used for all analyses, and will henceforth be referred to sim- ply as the effect size . We calculated 155 effect sizes from the 47 studies. Some studies produced an effect size many times larger than the median of 0.98. Chauvenet’s criterion (Taylor, 1997) was applied to limit the influence of outliers. Through the use of Chauvenet’s crite- rion, three outliers from three studies were eliminated. All effect sizes generated from these three studies (for a total of 16) were also eliminated, based on examples set by other meta-analyses (e.g., Camilli, Vargas, & Yurecko, 2003; Rolstad, Mahoney, & Glass, 2005). To balance the number of effect sizes per study for the analysis, studies containing multiple effect sizes within the same moderator variables were collapsed into one by calculating the mean. Therefore some studies still provided more than one effect size, as long as each effect size was differentiated by at least one moderator variable. For example, Schneider and Thompson (2003) produced 32 distinguishable effect sizes, often separated by the type of verb named by the participants. These 32 effect sizes were collapsed into four effect sizes that were differentiated by the word set and follow-up moderator variables. The two different therapy techniques studied by Schneider and Thompson (semantic verb retrieval treatment and argument structure verb retrieval treatment), both fit under the same treatment category (involving both semantic and syntactic therapies). By this method, 126 of the 155 effect sizes were collapsed into 48, leaving a total of 107 effect sizes from 44 studies that were included Downloaded By: [University of Oregon] At: 22:07 27 March 2010 in the final analysis. For this analysis, we applied the same formula to calculate the effect size for group, single-participant, and case studies. For all studies, measurements were made both before and after therapy using the same outcome measure. Therefore, for single-participant and case studies with n > 1, we calculated pre and post data for the effect size formula by combining all participants. We combined the data for the par- ticipants in these single-participant and case studies when no differences were seen in the treatment. These studies appeared to have a high level of control over potential confounding variables, similar to the group studies. Although single-participant and single-case designs showed differences between participants, we observed differences in aphasia type, severity, and time post-onset for group studies as well.

RESULTS There was a wide range of variability in the studies (see Appendix). The unbiased effect size, as measured by confrontational naming tasks, ranged from –0.42 to 19.11. 1344 WISENBURN AND MAHONEY

From the final 44 studies, 17 evaluated semantic therapy, 15 phonological therapy, and 16 mixed therapy. Four studies evaluated more than one type of treatment. The word set used in the confrontational naming task was trained in therapy for 46 effect sizes from 30 studies, with 9 effect sizes from 5 studies for related-exposed, 14 effect sizes from 10 studies for unrelated-exposed, 36 effect sizes from 27 studies for the unrelated-unexposed, and 2 effect sizes from 1 study for the related-unexposed group. Although most of the effect sizes were made for measurements immediately upon the conclusion of therapy, 28 effect sizes from 14 studies included a follow-up measure between 1 to 3 months after therapy. The median amount of time post onset for participants was under 6 months post onset for 7 effect sizes from 4 studies, 6 months to 1 year for 22 effect sizes from 9 studies, and over 1 year for 64 effect sizes from 25 studies. For 14 effect sizes from 7 studies, the time post onset for individual participants was not listed and so the median could not be determined. The mean effect size across all therapies was 1.66, SD = 2.45. This figure was based on 107 effect sizes from 44 studies. This overall mean effect size should be interpreted with caution, as it combines data across all moderator variables. These moderator variables (in addition to other variables not analysed, such as participant severity, aphasia type, age, etc.) may act as a confounding variable in the interpretation of this overall score. It is functionally impossible to eliminate all potential confounding vari- ables from an analysis of aphasia therapy. Despite these inherent limitations, the over- all score does give a general indication of the efficacy of word-finding therapy. Table 1 shows all moderator variables. For the analysis of word sets, all effect sizes were evaluated depending on whether they were trained, exposed without training, or unexposed except for pre- and post-therapy administrations, and also

TABLE 1 Effect size for various moderator variables

Moderator variable N of d N of studies M of d SD of d

Word set Trained 45 29 2.66 3.23

Related-exposed 9 5 1.73 1.06 Unrelated-exposed 14 10 1.78 2.02 Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Unrelated-unexposed 37 28 0.44 0.52 Related-unexposed 2 1 0.44 0.13 Treatment Semantic 40 17 1.41 1.31 Phonological 30 15 2.43 4.12 Mixed 37 16 1.32 1.23 Immediate and follow-up measurements Immediate post treatment 79 44 1.61 2.52 1 month post treatment 18 8 2.39 2.69 2 months post treatment 6 3 1.04 0.64 3 months post treatment 4 3 0.48 0.33 Median months post-onset < 6 7 4 0.53 0.40 7–12 22 9 1.55 2.04 13–18 9 4 1.78 1.33 19–24 12 6 1.71 1.62 25–36 15 4 1.57 1.32 37–48 10 4 0.73 0.65 > 49 18 7 1.66 1.44 META-ANALYSIS OF WORD-FINDING 1345

whether the word set for the effect size was related (semantically or phonologically) or not to the words trained in therapy. As would be expected, the gains were higher for trained sets versus unexposed sets. However, exposure to words without any training in therapy also appeared to have efficacious effects, regardless of whether the words were related or not to the trained set. We also calculated the gains made in therapy for unrelated-unexposed words for individuals past 12 months post-onset (typically identified as the final point for most spontaneous recovery), and found a mean effect size of 0.42, SD = 0.54. Table 1 also shows the magnitude of the mean effect size based on therapy cate- gory, along with the standard deviation and sample size. All types of therapy proved efficacious, with a higher mean effect size for phonological therapies. The results of immediate post-therapy and follow-up evaluations of 1, 2, and 3 months can be seen in Table 1. Substantial gains persisted for up to 3 months. However, the level of gains sharply decreased with each subsequent month, with a mean effect size of 0.48 at 3 months post onset. An analysis of all mean effect sizes showed an unexpected increase (from 1.61 to 2.39) at 1 month post treatment. How- ever, if reliability is increased by looking only at the eight studies that performed a follow-up measure at 1 month post-treatment with the matching measurement imme- diately post-treatment, then this increase disappears. The immediate post-treatment measures (that match the 1-month follow-up effect sizes) showed a mean of 2.92 (SD = 4.23), which was higher than the 1-month follow-up measure of 2.39. The effect size depending on the median number of months post-onset for the par- ticipants is also shown in Table 1. As the table shows, substantial gains were made in therapy even past 4 years post-onset. The discrepancy between word sets was relatively large (as shown in Table 1), so we calculated the mean effect size of therapies for different word sets. These results are shown in Table 2. As can be seen from Table 2, the gains for trained word sets were larger than the unexposed words. Semantic therapy appeared to generalise

TABLE 2 Effect size of various therapies for different word sets

Grouping N of d N of studies M of d SD of d Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Trained words Semantic 17 11 1.90 1.42 Phonological 14 11 4.06 5.30 Mixed 15 10 2.17 1.31 Related-exposed words Semantic 5 3 1.99 1.43 Phonological 2 1 1.26 0.13 Mixed 2 1 1.53 0.05 Unrelated-exposed words Semantic 6 4 1.42 1.25 Phonological 2 2 4.59 4.84 Mixed 6 4 1.29 1.21 Unrelated-unexposed words Semantic 13 10 0.57 0.73 Phonological 12 10 0.37 0.44 Mixed 12 9 0.39 0.29 Related-unexposed words Mixed 2 1 0.44 0.13 1346 WISENBURN AND MAHONEY

TABLE 3 Effect size of immediately post-therapy and follow-up measurements for various therapies for trained words

N of d N of studies M of d SD of d

Immediate post-therapy Semantic 11 11 1.98 1.40 Phonological 14 11 3.59 5.20 Mixed 10 10 2.12 1.36 1–3-month follow-up Semantic 5 5 1.71 1.61 Phonological 2 1 6.87 7.04 Mixed 5 5 2.26 1.35

slightly more than either phonological or mixed therapies to the unexposed words, although these gains were still relatively small compared to the gains made for trained or exposed sets. Table 3 focuses on the effect size for immediate post-onset and follow-up meas- urements for various therapies for trained words. The low number of follow-up measurements for phonological therapy makes it difficult to compare the level of maintenance of trained words to between therapies. However, the results show that the effect size for trained words did not diminish substantially after therapy for semantic, phonological, or mixed therapy.

DISCUSSION The synthesis of 44 studies clearly shows that word-finding therapy is efficacious. The large overall effect size for therapy (1.66) corresponds with the conclusions of the literature reviews by Nickels and Best (1996a) and Nickels (2002a). To interpret the magnitude of the effect size for individual studies, the results may be divided into quartiles (as shown by Robey et al., 1999). The first, second, and third quartiles are 0.32, 0.98, and 1.91. This may give an overall idea of the magni-

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 tude of the effect size when comparing various treatments, with 0.40 to be considered small, 1.05 would be medium, and 2.02 would be large. However, the effect sizes showed substantial differences within the treatment set moderator variable. There- fore, it may be more beneficial to divide the effect sizes into quartiles within each type of treatment set. The results of this calculation are shown in Table 4. No definite conclusions could be made as to whether semantic, phonologic, or mixed therapy is most efficacious. The overall mean effect sizes for each therapy were

TABLE 4 Magnitude of the effect size for different word sets

Word set Small Medium Large

Trained words 1.07 1.88 3.16 Related-exposed words 1.26 1.54 2.04 Unrelated-exposed words 0.43 1.02 2.63 Unrelated-unexposed words 0.16 0.31 0.62 META-ANALYSIS OF WORD-FINDING 1347

high (over 1.0). Considering the extremely high standard deviation for each therapy, the success of any one type of therapy may vary substantially depending on the individual. The effect size for semantic treatment for unrelated-unexposed words (0.57) was, although relatively small, still slightly larger than for other therapies. This supports the conclusion of Nickels that semantic therapy may provide more general- isation than other therapies. The literature also provided encouraging results regarding the persistence of the gains in therapy. Although the gains from therapy decreased after each month, the effect size at 3 months (0.48) was still substantial. What is more positive is the high level of maintenance for trained words, regardless of the type of therapy (as shown in Table 3). The results for the word sets used in measurement show that most of the gains in therapy are for exposed words. The surprisingly high level of gains for exposed words suggests that frequent attempts to name pictures, even without feedback or training, will lead to gains for those words. This idea was suggested by Howard (2000) and Nickels (2002a), and has been shown by Nickels (2002b) in a single- participant study. Exposed words that were semantically or phonologically related to the trained words showed a slightly higher effect size than unrelated words. More research may be needed in this area to verify these findings. Studies may focus partic- ularly on the gains made for related words that were not exposed in therapy. The level of gains for unrelated-unexposed words was fairly low, with a mean effect size of 0.43, SD = 0.52. The mean effect size for participants past 12 months post-onset for unexposed words was 0.42, SD = 0.54. Robey (1998) found in his meta-analysis of aphasia treatment that the mean effect size of gains made for untreated aphasia was 0.63 when the pre-test measure was taken in the acute phase of aphasia, and 0.05 when the pre-test measure was taken on or after the twelfth month post-onset. Robey stated that these figures showed that, as expected, the peak period of natural recov- ery (without any aphasia treatment) was within 1 year post-onset. Therefore, the mean effect size for unrelated-unexposed words for participants past 12 months post- onset (0.43) was substantially larger than what would be expected by spontaneous recovery without treatment for this population (0.05). This suggests that the effect size for unrelated-unexposed words represents mild generalisation gains in therapy, rather than simply gains due to spontaneous recovery. Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Another important finding was the substantial gains made in therapy for participants long past their neurological injury. In most cases, the effect size for par- ticipants after 1 year post-onset was larger than for the more acute cases. This is further evidence contrary to the well-accepted idea that little gains can be made for patients with aphasia who are more than one year post-onset. Studies with a median participant post-onset time of greater than 4 years still showed a large overall effect size (1.52, SD = 1.35).

CONCLUSION The findings of this meta-analysis support the conclusions that therapy for anomia is efficacious for aphasic individuals. This is consistent with the previous literature reviews of Nickels and Best (1996a) and Nickels (2002a). Semantic, phonological, and mixed therapies all showed benefits, although semantic therapy appeared to have more generalisation to unexposed words. The large standard deviations seen throughout this study are an indication that therapeutic gains are highly variable. 1348 WISENBURN AND MAHONEY

As Nickels stated, we still have little knowledge as to what type of therapy works best for individual patient characteristics. Clinicians must take care to find the therapeutic method that is the most efficacious for each individual client.

Manuscript received 1 March 2008 Manuscript accepted 30 December 2008 First published online 16 September 2009

REFERENCES References marked with an asterisk indicate studies included in the meta-analysis. *Bastiaanse, R., Hurkmans, J., & Links, P. (2006). The training of verb production in Broca’s aphasia: A multiple-baseline across-behaviours study. Aphasiology , 20 , 298–311. *Beeson, P., & Egnor, H. (2006). Combining treatment for written and spoken naming. Journal of International Neuropsychological Society , 12 , 816–827. Beeson, P. M., & Robey, R. R. (2006). Evaluating single-subject treatment research: Lessons learned from the aphasia literature. Neuropsychology Review , 16 , 161–169. *Best, W., Hickin, J., Herbert, R., Howard, D., & Osborne, F. (2000). Phonological facilitation of aphasic naming and predicting the outcome of treatment for anomia. Brain and Language , 74 , 435–438. Brookshire, R. (2007). Introduction to neurogenic communication disorders , 7th ed. St. Louis, MO: Mosby. Camilli, G., Vargas, S., & Yurecko, M. (2003). Teaching children to read: The fragile link between science and federal education policy. Education Policy Analysis Archives , 11 . Retrieved 18 June 2008, from http://epaa.asu.edu/epaa/v11n15 *Cornelissen, K., Laine, M., Tarkiainen, A., Järvensivu, T., Martin, N., & Salmelin, R. (2003). Adult brain plasticity elicited by anomia treatment. Journal of Cognitive Neuroscience , 15 , 444–461. *Deloche, G., Dordain, M., & Kremin, H. (1993). Rehabilitation of confrontation naming in aphasia: Relations between oral and written modalities. Aphasiology , 7, 201–216. *Deloche, G., Ferrand, I., Metz-Lutz, M., Dordain, M., Kremin, H., Hannequin, D., et al. (1992). Con- frontation naming rehabilitation in aphasics: A computerised written technique. Neuropsychological Rehabilitation , 2, 117–124. *Doesborgh, S. J. C., Van de Sandt-Koudstaal, M. W. M. E., Dippel, D. W. J., van Harskamp, F., Koudstaal, P. J., & Visch-Brink, E. G. (2004). Cues on request: The efficacy of Multicue, a computer program for word-finding therapy. Aphasiology , 18 , 213–222. *Drew, R. L., & Thompson, C. K. (1999). Model-based semantic treatment for naming deficits in aphasia. Journal of Speech, Language, and Hearing Research , 42 , 972–989. *Edmonds, L. A., & Kiran, S. (2006). Effect of semantic naming treatment on crosslinguistic generalisa- tion in bilingual aphasia. Journal of Speech, Language, and Hearing Research , 49 , 729–748. *Edwards, S., & Tucker, K. (2006) Verb retrieval in fluent aphasia: A clinical study. Aphasiology , 20 , Downloaded By: [University of Oregon] At: 22:07 27 March 2010 644–657. *Fillingham, J. K., Sage, K., & Lambon Ralph, M. A. (2005). Further explorations and an overview of errorless and errorful therapy for aphasic word-finding difficulties: The number of naming attempts during therapy affects outcome. Aphasiology , 19 , 597–614. *Fink, R. B., Schwartz, M. F., Sobel, P.R., & Myers, J. (1997). Effects of multilevel training on verb retrieval: Is more always better? Brain and Language , 60 , 41–44. *Fridriksson, J., Morrow-Odom, L., Moser, D., Fridriksson, A., & Baylis, G. (2006). Neural recruitment associated with anomia treatment in aphasia. NeuroImage , 32 , 1403–1412. *Fridriksson, J., Moser, D., Bonilha, L., Morrow-Odom, K. L., Shaw, H., Fridriksson, A., et al. (2007). Neural correlates of phonological and semantic-based anomia treatment in aphasia. Neuropsychologia , 45 , 1812–1822. Glass, G. (1976). Primary, secondary, and meta-analysis of research. The Educational Researcher , 10 , 3–8. Glass, G., McGaw, B., & Smith, M.L. (1981). Meta-analysis in social research . Beverly Hills, CA: Sage. Goodglass, H., & Wingfield, A. (1997). Word-finding deficits in aphasia: Brain–behavior relations and clinical symptomatology. In H. Goodglass & A. Wingfield (Eds.), Anomia: Neuroanatomical and cognitive correlates . San Diego, CA: Academic Press. Greenhouse, J. B., Fromm, D., Iyengar, S., Dew, M. A., Holland, A., & Kass, R. (1990). The making of a meta-analysis: A quantitative review of the aphasia treatment literature. In M. Straf & K. Wachter (Eds.), The future of meta-analysis . Beverly Hills, CA: Sage. META-ANALYSIS OF WORD-FINDING 1349

*Greenwald, M. L., Raymer, A. M., Richardson, M. E., & Rothi, L. J. G. (1995). Contrasting treatments for severe impairments of picture naming. Neuropsychological Rehabilitation , 5, 17–49. *Helm-Estabrooks, N., Emery, P., & Albert, M. L. (1987). Treatment of aphasic perseveration (TAP) program: A new approach to aphasia therapy. Archives of Neurology , 44 , 1253–1255. *Herbert, R., Best, W., Hickin, J., Howard, D., & Osborne, F. (2003). Combining lexical and interactional approaches to therapy for word-finding deficits in aphasia. Aphasiology , 17 , 1163–1186. *Hickin, J., Best, W., Herbert, R., Howard, D., & Osborne, F. (2002). Phonological therapy for word- finding difficulties: A re-evaluation. Aphasiology , 16 , 981–999. *Hickin, J., Greenwood, A., Grassly, J., Herbert, R., Howard, D., & Best, W. (2005) Therapy for word- finding difficulties using phonological and orthographic cues: A clinical application in progress . Paper presented at the 2005 Clinical Aphasiology Conference. Retrieved 6 February 2008, from http:// aphasiology.pitt.edu/archive/00001562/01/e8d78f6604edcfca58a04a07fb48.pdf. *Hillis, A. E. (1989). Efficacy and generalisation of treatment for aphasic naming errors. Archives of Physical Medicine and Rehabilitation , 70 , 632–636. *Hinckley, J. J., & Carr, T. H. (2005). Comparing the outcomes of intensive and non-intensive context- based aphasia treatment. Aphasiology , 19 , 965–974. *Hinckley, J. J., & Craig, H. K. (1998). Influence of rate of treatment on the naming abilities of adults with chronic aphasia. Aphasiology , 12 , 989–1006. Howard, D. (2000). Cognitive neuropsychology and aphasia therapy: The case of word retrieval. In I. Papathanasiou (Ed.), Acquired neurogenic communication disorders: A clinical perspective . London: Whurr. Howard, D., Patterson, K. E., Franklin, S., Orchard-Lisle, V., & Morton, J. (1985). The facilitation of picture naming in aphasia. Cognitive Neuropsychology , 2, 49–80. *Kiran, S., & Thompson, C. K. (2003). The role of semantic complexity in treatment of naming deficits: Training semantic categories in fluent aphasia by controlling exemplar typicality. Journal of Speech, Language, & Hearing Research , 46 , 773–787. *Laganaro, M., Di Pietro, M., & Schnider, A. (2003). Computerised treatment of anomia in chronic and acute aphasia: An exploratory study. Aphasiology , 17 , 709–721. *Laganaro, M., Di Pietro, M., & Schnider, A. (2006). Computerised treatment of anomia in acute aphasia: Treatment intensity and training size. Neuropsychological Rehabilitation , 16 , 630–640. *Law, S., Wong, W., Sung, F., & Hon, J. (2006). A study of semantic treatment of three Chinese anomic patients. Neuropsychological Rehabilitation , 16 , 601–629. Li, E. C. (1996). Treatment of naming impairment. In G. L. Wallace (Ed), Adult aphasia rehabilitation . Boston: Butterworth-Heinemann. *Linebaugh, C. W., Baron, C. R., & Corcoran, K. J. (1998). Assessing treatment efficacy in acute aphasia: Paradoxes, presumptions, problems and principles. Aphasiology , 12 , 519–536. *Lowell, S., Beeson, P. M., & Holland, A. L. (1995). The efficacy of a semantic cueing procedure on nam- ing performance of adults with aphasia. American Journal of Speech-Language Pathology , 4, 109–114. *Mackenzie, C. (1991). An aphasia group intensive efficacy study. British Journal of Disorders of Commu-

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 nication , 26 , 275–291. Mayer, J. F., & Murray, L. L. (2003). Functional measures of naming in aphasia: Word retrieval in confrontation naming versus connected speech. Aphasiology , 17 , 481–497. *Miceli, G., Amitrano, A., Capasso, T., & Caramazza, A. (1996). The treatment of anomia resulting from output lexical damage: Analysis of two cases. Brain and Language , 52 , 150–174. *Nadeau, S. & Kendall, D. (2006). Significance and possible mechanisms underlying generalisation in aphasia therapy: Semantic treatment of anomia. Brain and Language , 99 , 12–13. *Nettleton, J., & Lesser, R. (1991). Therapy for naming difficulties in aphasia: Application of a cognitive neuropsychological model. Journal of , 6, 139–157. Nickels, L. (2002a). Therapy for naming disorders: Revisiting, revising and reviewing. Aphasiology , 16 , 935–979. Nickels, L. (2002b). Improving word-finding: Practice makes (closer to) perfect? Aphasiology , 16 , 1047–1060. Nickels, L., & Best, W. (1996a). Therapy for naming deficits (part I): Principles, puzzles and progress. Aphasiology , 10 , 21–47. *Nickels, L., & Best, W. (1996b). Therapy for naming deficits (part II): Specifics, surprises and suggestions. Aphasiology , 10 , 109–136. *Pedersen, P. M., Vinter, K., & Olsen, T. S. (2001). Improvement of oral naming by unsupervised computerized rehabilitation. Aphasiology , 15 , 151–169. *Pring, T., Hamilton, A., Harwood, A., & Macbride, L. (1993). Generalisation of naming after picture/ word matching tasks: Only items appearing in therapy benefit. Aphasiology , 7, 383–394. 1350 WISENBURN AND MAHONEY

Raymer, A. (2005). Naming and word-retrieval problems. In L. L. LaPointe (Ed.), Aphasia and related language disorders , 3rd ed. New York: Thieme. *Raymer, A., & Kohen, F. (2006). Word retrieval treatment in aphasia: Effects of sentence context. Journal of Rehabilitation Research & Development , 43 , 367–378. *Raymer, A. M., Kohen, F., & Saffell, D. (2006). Computerised training for impairments of word comprehension and retrieval in aphasia. Aphasiology , 20 , 257–268. *Raymer, A. M., Thompson, C. K., Jacobs, B., & Le Grand, H. R. (1993). Phonological treatment of naming deficits in aphasia: Model based generalisation analysis. Aphasiology , 7, 27–53. Robey, R. (1998). A meta-analysis of clinical outcomes in the treatment of aphasia. Journal of Speech, Language, and Hearing Research , 41 , 172–187. Robey, R., & Dalebout, S. (1998). A tutorial on conducting meta-analyses of clinical outcome research. Journal of Speech, Language, and Hearing Research , 41 , 1227–1241. Robey, R. R. (1994). The efficacy of treatment for aphasic persons: A meta-analysis. Brain and Language , 47 , 585–608. Robey, R. R., & Schultz, M. C. (1998). A model for conducting clinical-outcome research: An adaptation of the standard protocol for use in aphasiology. Aphasiology , 12 , 787–810. Robey, R. R., Schultz, M. C., Crawford, A. B., & Sinner, C. A. (1999). Single-subject clinical-outcome research: Designs, data, effects sizes, and analyses. Aphasiology , 13 , 445–473. *Robson, J., Marshall, J., Pring, T., Montagu, A., & Chiat, S. (2004). Processing proper nouns in aphasia: Evidence from assessment and therapy. Aphasiology , 18 , 917–935. *Rodriguez, A. D., Raymer, A. M., & Rothi, L. J. G. (2006). Effects of gesture+verbal and semantic- phonologic treatments for verb retrieval in aphasia. Aphasiology , 20 , 286–297. Rolstad, K., Mahoney, K., & Glass, G. V. (2005). The big picture: A meta-analysis of program effective- ness research on English language learners. Educational Policy , 19 , 572—594. Rudner, L, Glass, G., Evartt, D., & Emery, P. (2002). Meta-Stat (version 1.5) [computer software]. College Park, MD: ERIC Clearinghouse on Assessment & Evaluation. *Schneider, S. L., & Thompson, C. K. (2003). Verb production in agrammatic aphasia: The influence of semantic class and argument structure properties on generalisation. Aphasiology , 17 , 213–241. Taylor, J. (1997). An introduction to error analysis: The study of uncertainties in physical measurements (2nd ed.). Sausalito, CA: University Science Books. *Van Mourik, M., & Van De Sandt-Koenderman, W. M. E. (1992). Multicue. Aphasiology , 6, 179–183. *Wambaugh, J., Cameron, R., Kalinyak-Fliszar, M., Nessler, C., & Wright, S. (2004). Retrieval of action names in aphasia: Effects of two cueing treatments. Aphasiology , 18 , 979–1004. *Wambaugh, J. L., Doyle, P. J., Martinez, A. L., & Kalinyak-Fliszar, M. (2002). Effects of two lexical retrieval cueing treatments on action naming in aphasia. Journal of Rehabilitation Research and Devel- opment , 39 , 455–466. Whurr, R., Lorch, M. P., & Nye, C. (1992). A meta-analysis of studies carried out between 1946 and 1988 concerned with the efficacy of speech and language therapy treatment for aphasic patients. European Journal of Disorders of Communication , 27 , 1–17.

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Wilshire, C. E., & Coslett, H. B. (2000). Disorders of word retrieval in aphasia: Theories and potential applications. In S. E. Nadeau, L. J. G. Rothi, & B. Crosson (Eds.), Aphasia and language: Theory to practice . New York: Guilford Press. META-ANALYSIS OF WORD-FINDING 1351

APPENDIX

Unbiased effect size for trained, exposed-related, exposed-unrelated, and unexposed word sets (Follow-up measures in parentheses)

Study Trained Exp-Rel Exp-Unrel Unexposed Med MPO

Semantic therapy Bastiaanse, Hurkmans, & 0.65 (0.97 c) 0–6 Links (2006) Edmonds & Kiran (2006) 2.43 (1.92 a) 1.54 (1.54 a) 0.78 7–12 Edwards & Tucker (2006) 0.50 (0.44 c) –0.42 (0.31 c) 7–12 Fink, Schwartz, Sobel, & 4.22 N/A Myers (1997) Fridriksson et al. (2007) 0.53 19–24 Greenwald, Raymer, 2.02 (1.21 a) 7–12 Richardson, & Rothi (1995) Kiran & Thompson (2003) 2.72 13–18 Law, Wong, Sung, & Hon 3.28 (4.39 a) 4.11 (2.52 a) 1.98 (3.25 a) 49+ (2006) Lowell, Beeson, & Holland 1.32 0.27 2.23 7–12 (1995) Mackenzie (1991) 0.24 19–24 Nadeau & Kendall (2006) 0.56 (0.23 c) 37–48 Nettleton & Lesser (1991) 0.28 19–24 Nickels & Best (1996b) 1.23 0.43 37–48 Raymer & Kohen (2006) 0.28 0.13 0.10 49+ Schneider & Thompson (2003) 0.90 (0.57 a) 0.43 (0.50 a) N/A Wambaugh, Cameron, 4.22 25–36 Kalinyak-Fliszar, Nessler, & Wright (2004) Wambaugh, Doyle, Martinez, 1.41 49+ & Kalinyak-Fliszar (2002) Phonological therapy Beeson & Egnor (2006) 0.81 49+ Best, Hickin, Herbert, 1.11 0.14 N/A Howard, & Osborne (2000) Deloche et al. (1992) 6.64 0.02 7–12 a a

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Fillingham, Sage, & Lambon 19.11 (11.85 ) 1.08 (1.20 ) N/A Ralph (2005) Fridricksson, Morrow-Odom, 0.02 49+ Moser, Fridricksson, & Baylis (2006) Fridriksson et al. (2007) 0.17 19–24 Hickin, Best, Herbert, 1.05 0.17 37–48 Howard, & Osborne (2002) Hickin et al. (2005) 1.45 1.16 0.59 19–24 Hillis (1989) 8.01 7–12 Laganaro, Di Pietro, & −0.07 0.23 0–6 Schnider (2003) Laganaro, Di Pietro, & 1.08 0 7–12 Schnider (2003) Laganaro, Di Pietro, & 0.98 0–6 Schnider (2006)

(Continued ) 1352 WISENBURN AND MAHONEY

APPENDIX (Continued)

Study Trained Exp-Rel Exp-Unrel Unexposed Med MPO

Miceli, Amitrano, Capasso, & 2.36 0.00 13–18 Caramazza (1996) Raymer, Thompson, Jacobs, 2.35 (1.90 b) 1.35 (1.17 b) 13–18 & Le Grand (1993) Wambaugh, Cameron, 4.04 25–36 Kalinyak-Fliszar, Nessler, & Wright (2004) Wambaugh, Doyle, Martinez, 3.03 49+ & Kalinyak-Fliszar (2002) Mixed therapy Cornellisen et al. (2003) 0.45 N/A Deloche, Dordain, & Kremin 0.62 49+ (1993) Doesborgh et al. (2004) 0.33 7–12 Drew & Thompson (1999) 1.59 (1.02 b) 0.87 (0.41 b) 49+ Greenwald, Raymer, 1.41 7–12 Richardson, & Rothi (1995) Helm-Estabrooks, Emery, & 0.54 N/A Albert (1987) Herbert, Best, Hickin, 1.88 (1.51 b) 0 (0.23 b) 37–48 Howard, & Osborne (2003) Hinckley & Carr (2005) 0.40 25–36 Hinckley & Craig (1998) 1.09 N/A Linebaugh, Baron, & 0.69 0.28 0–6 Corcoran (1998) Pedersen, Vinter, & Olsen 1.27 19–24 (2001) Pring, Hamilton, Harwood, & 2.43 (1.99 a) 1.56 (1.49 a) 0.65 (0.48 a) 25–36 Macbride (1993) Pring, Hamilton, Harwood, & 0.54 25–36 Macbride (1993) (0.35 a)* Raymer, Kohen, & Saffell 4.66 (4.52 a) 3.03 (2.63 a) 19–24 (2006) Robson, Marshall, Pring, 2.50 (2.28 a) 0.31 (0.30 a) 25–36

Downloaded By: [University of Oregon] At: 22:07 27 March 2010 Montagu, & Chiat (2004) Rodriguez, Raymer, & Rothi 4.16 0 13–18 (2006) van Mourik & Van De 0.41 7–12 Sandt-Koenderman (1992)

Exp-Rel = Exposed-Related. Exp-Unrel = Exposed-Unrelated. Med MPO = median months post- onset. aFollow-up of approximately 1 month after therapy. b2-month follow-up. c3-month follow-up.