Contextual priming in semantic anomia: A case study

Brain and Language 95 (2005) 327–341 www.elsevier.com/locate/b&l

Contextual priming in semantic anomia: A case study Kati Renvall a,*, Matti Laine b, Nadine Martin c a

Department of Psychology, University of Turku, Turku, Finland Department of Psychology, A˚bo Akademi University, A˚bo, Finland Department of Communication Sciences, Temple University, Philadelphia, USA b

c

Accepted 19 February 2005 Available online 25 March 2005

Abstract The present case continues the series of anomia treatment studies with contextual priming (CP), being the second in-depth treatment study conducted for an individual suffering from semantically based anomia. Our aim was to acquire further evidence of the facilitation and interference effects of the CP treatment on semantic anomia. Based on the results of the study of Martin, Fink, and Laine (2004a), our hypothesis before the treatment was that our participant would show short-term interference and at most modest and short-term benefit from treatment. To acquire such evidence would not only be important for the choice of anomia treatment methods in individual patients, but would also prompt further development of the CP method. The CP technique used for our participant included cycles of repeating and naming items in three contextual conditions (semantic, phonological, and unrelated). As predicted, the overall improvement of naming was modest and short-term. Interestingly, the contextual condition that corresponded with the nature of our patientÕs underlying naming deficit (semantic) elicited immediate interference in the form of contextual naming errors, as well as short-term improvement of naming. Based on this and a recent study by Martin et al. (2004a), it appears that despite short-term positive effects, in its current form the CP treatment is not sufficient for those aphasics who have a semantic deficit underlying their anomia. The possible mechanism and directions for future research are discussed.
2005 Elsevier Inc. All rights reserved. Keywords: Semantic anomia; Anomia rehabilitation; Contextual priming

1. Introduction Contextual priming (CP) is a theoretically motivated and relatively simple naming treatment technique that was originally developed to test predictions of different word-production models (Laine & Martin, 1996). It has since proved to have potential as a technique to be used in anomia treatment (see Martin, Fink, & Laine, 2004a; Martin, Fink, Laine, & Ayala, 2004b; Martin & Laine, 2000; Renvall, Laine, Laakso, & Martin, 2003). The CP technique includes cycles of repeating and naming items in different contextual conditions, including semantic condition in which items are only semantically *

Corresponding author: Fax: +358 2 3335060. E-mail address: kati.renvall@utu.fi (K. Renvall).

0093-934X/$ - see front matter
2005 Elsevier Inc. All rights reserved. doi:10.1016/j.bandl.2005.02.003

related, phonological condition in which items are only phonologically related, and an unrelated (baseline) condition in which items are neither semantically nor phonologically related to each other. The first two relatedness conditions are aiming at the main stages of word production. One of the aims of the present and the other recent CP studies is to determine whether it is better to treat a deficit directly or to use preserved abilities to improve lexical access. This issue has gained much interest in the field of anomia treatment (see e.g., Nickels, 1997, 2002; Nickels & Best, 1996) and has sparked a growing body of carefully designed treatment studies targeting this question (e.g., Bastiaanse, Bosje, & Franssen, 1996; Drew & Thompson, 1999; Hickin, Herbert, Best, Howard, & Osborne, 2002; Hillis & Caramazza, 1994;

328

K. Renvall et al. / Brain and Language 95 (2005) 327–341

Marshall, Robson, Pring, & Chiat, 1998; Nettleton & Lesser, 1991). However, there is yet no conclusive evidence of interactions between the nature of the naming deficits (semantically or phonologically based) and the effects of treatments that target specifically semantic or phonological processing. Sometimes it is difficult to ascertain the precise mechanisms by which the observed effects of a treatment have been achieved (Nickels & Best, 1996). Here, we report a study that is part of an ongoing investigation of possible treatment–deficit interactions. We will use the CP technique of priming by repeating a group of target words, but varying the relationship among the words (semantic, phonological, and no relation). This feature of the technique, along with use of a well-defined model as a framework allows us to draw conclusions about why and at which level observed effects occur. Employment of repetition and semantic/phonological contexts in the CP technique is by no means a totally new approach. Repetition is widely used in almost all speech therapy tasks, and it has been studied by Weigl (1961), who used the so called Ôde-blockingÕ technique, and more recently by Nettleton and Lesser (1991) and Miceli, Amitrano, Capasso, and Caramazza (1996). In these and other studies, repetition has been combined with other phonologically based methods of stimulating word retrieval, such as phonological cueing and reading aloud, as well as semantically based methods such as word-topicture matching and producing semantic features and synonyms (e.g., Basso, Marangolo, Piras, & Galluzzi, 2001; Best, Howard, Bruce, & Gatehouse, 1997; Greenwald, Raymer, Richardson, & Rothi, 1995; Hillis, 1991, 1998; Hillis & Caramazza, 1994; Raymer, Thompson, Jacobs, & Le Grand, 1993). Other studies have concentrated on tasks tapping primarily the semantic level processing (e.g., Coelho, McHugh, & Boyle, 2000; Kiran & Thompson, 2003; Visch-Brink, Bajema, & Van De Sandt-Koenderman, 1997). The CP technique combines repetition and naming contexts in a way that each participant gets both semantic and phonological level treatment in a more or less implicit fashion. Still, as we will argue later in this section, it is possible to detect the separate effects of the semantic and phonological priming as they are contrasted to the condition in which there is no relation among words being trained. The locally interactive models such as that of Dell (1986); Dell and OÕSeaghdha (1992); Martin, Dell, Saffran, and Schwartz (1994); Dell, Schwartz, Martin, Saffran, and Gagnon (1997); Foygel and Dell (2000) serve as a framework for the CP technique. According to these models, lexical retrieval is expected to involve two steps (lexical-semantic and lexical-phonological) in a network consisting of three layers of nodes (semantic features, words, and phonemes). These models assume that activation spreads among the related semantic and phonological nodes and that all more or less active

lexical-semantic representations become also phonologically encoded. Moreover, the connections between the layers are not only one-way but bidirectional and excitatory. Based on these basic features of the models, it is possible to explain, for example, why normal adults make more mixed semantic-phonological errors than would occur by chance (see Martin, Weisberg, & Saffran, 1989). Similarly, as we will explain next, it can be understood why both semantic and phonological relatedness can have effects on aphasic speakersÕ naming. The facilitation of naming in the CP treatment is expected to arise from two sources explained by the afore mentioned two-step interactive models (Dell, 1986; Dell & OÕSeaghdha, 1992; Dell et al., 1997; Foygel & Dell, 2000; Martin et al., 1994). The first source of facilitation is repetition priming which has been shown to have long-lasting effects in word production (for normal adults see, e.g., Cave, 1997, for anomic patients see, e.g., Davis & Pring, 1991; Miceli et al., 1996). Repetition of words is assumed to increase the activation levels of the corresponding word forms and thus help in retrieving those items. The other sources of facilitation are the semantic and phonological priming effects. This facilitation mediates activation spreading amongst the related items in a picture set, and represents a fundamental feature of the interactive word-production models. Based on the models, the target items are assumed to activate several meaning-related items which share partly overlapping semantic features. The semantic activation in turn feeds forward to corresponding lexical nodes. Continuous feedforward–feedback activation between semantic and lexical nodes serves to strengthen the semantic relatedness effect so that the target and its semantic competitors have higher activation levels than items that are not semantically related to targets. This is why the items in the semantic contexts are assumed to become more available in the CP treatment. In the phonological contexts, the activation boost stems from shared sublexical representations which feed back activation to the corresponding lexical nodes. The repeated feedforward–feedback activation cycles increase activation levels of the target and its phonological competitors. Thus, the activation boost amongst related items via activation spreading may help an aphasic individual to name the pictures in both semantically and phonologically related sets better than those in an unrelated set. In addition to the direct facilitation effects, i.e., the increased number of correct naming responses, spreading activation also brings related nontarget representations closer to the selection threshold and can give rise to contextual errors (word substitution errors within a picture set) and noncontextual errors (semantically or phonologically related word errors from outside the target set). We have observed these errors in prior CP studies although in some patients they are very rare

K. Renvall et al. / Brain and Language 95 (2005) 327–341

(Laine & Martin, 1996; Martin et al., 2004a, 2004b; Martin & Laine, 2000; Renvall et al., 2003). The contextual errors reflect interference which seems to be shortlasting as these errors have mainly been observed in naming attempts that follow soon after the repetition priming of the same items (Martin et al., 2004a, 2004b). The interactive activation models predicts that the immediate interference, due to more available lexical competitors, shifts to short-term facilitation after some decay has occurred. The recent CP studies have indeed brought evidence of this shift in aphasic patients although the exact time intervals for the interference and facilitation effects are not yet known (Martin et al., 2004a, 2004b). The noncontextual word-errors that are either semantically or phonologically related to the target items but not within the target sets provide further evidence for activation spreading amongst multiple lexical candidates. In aphasic patients, the noncontextual related word errors are direct evidence for activation spreading within semantic or phonological lexical neighbourhood. Thus, both contextual and noncontextual errors can reveal important aspects of the mechanisms underlying the CP treatment. In the CP, specific effects of the repetition priming and spreading activation amongst related items can be distinguished by analyses of the patientsÕ responses in the different task conditions. This is important as most of the tasks employed in anomia therapy involve priming in some form but the effects of semantic and phonological priming as well as their efficacy for different aphasia types still remain unclear. Repetition priming should affect all picture sets, whereas semantic/phonological priming effects should be observed in the related picture sets only. A higher rate of correct responses and contextual errors in the semantic or phonological conditions as compared with the unrelated condition indicates that the contextual component (i.e., activation spreading amongst related items) is having an effect over and above that of repetition priming. The present case study continues the series of CP studies exploring the effects of this technique on the naming of aphasic individuals. Thus far, there are three smaller-scale facilitation studies (Laine & Martin, 1996; Martin et al., 2004b; Martin & Laine, 2000) and two full-scale treatment studies (Martin et al., 2004a; Renvall et al., 2003) reported with the CP technique.1 In addition, CP with semantic contexts only has been used to treat the naming abilities of three chronic lefthemisphere-damaged aphasics in a functional brain 1 The cited facilitation and treatment studies differ in several ways. First, the facilitation studies have included fewer treated items, fewer treatment sessions, and fewer (if any) separate naming measurements with the target items than the full-scale treatment studies. In addition, the first two facilitation studies (Laine & Martin, 1996; Martin & Laine, 2000) did not include nontreated control items.

329

imaging study (Cornelissen, Laine, Tarkiainen, Ja¨rvensivu, Martin, & Salmelin, 2003). With regard to the efficacy of the CP treatment, the results of the prior studies have been promising as all of the patients who have participated in the CP facilitation studies or full-scale treatment programmes have shown at least short-term improvement on naming the trained items in one or more contextual conditions. In addition, long-term improvement has been observed in two of three patients of the treatment studies while all of the three patients have shown at least some evidence of generalisation of treatment to untrained words. While no consistent relationship between the primary source of naming impairments and the facilitating contexts has been observed, the recent results of Martin et al. (2004a, 2004b) imply that the procedure itself could be most effective when semantic abilities are relatively spared. Another interesting phenomenon observed in the recent CP studies is that an immediate interference surfacing up as contextual word errors can turn into short-term facilitation of naming the same items (see Martin et al., 2004a, 2004b), as predicted by an interactive activation model of word retrieval (e.g., Dell & OÕSeaghdha, 1992). Moreover, Martin et al. (2004b) suggest that more interference, i.e., increased rates of contextual and noncontextual errors, occurs when semantic context is used for semantically impaired persons and phonological context for phonologically impaired persons than when the treated context and the underlying impairment do not match. Here, we report the third full-scale treatment study conducted with the CP technique and the second one used with a patient suffering from semantically based anomia. Our aim was to acquire further evidence of the efficacy of the CP treatment when the patient suffers from a semantic deficit. We had four questions: (1) Does contextual priming facilitate or interfere with naming of hard-to-name targets in our semantic anomia case? (2) If there is facilitation and/or interference, do the effects differ between the different contextual conditions (semantic, phonological)? (3) If there is a facilitatory effect, is it enduring? (4) Does the treatment facilitate naming of the untrained control items? The first aphasic with a semantic deficit for whom a CP treatment with several contextual conditions was conducted, AS (Martin et al., 2004a), suffered also to some extent from a phonological deficit. For this reason, we were especially interested to carry out CP treatment with an aphasic having a more circumscribed semantic deficit. The participant in the present study, PH, had a more circumscribed semantic deficit. Before starting the treatment for PH, we predicted that he would still

330

K. Renvall et al. / Brain and Language 95 (2005) 327–341

show a similar pattern of treatment effects as observed in the case of AS (Martin et al., 2004a) and those patients with semantic deficits in the earlier facilitation studies (Martin et al., 2004b; Martin & Laine, 2000). Thus, we expected to find an effect of immediate interference in the semantic condition in particular and at most modest and short-term improvement in the trained items. Whether we could get evidence of facilitation or lack of it was the most important goal in this study as a repeated negative finding of the CP treatment in semantic anomia would imply that CP treatment in its current form is not enough for this group of anomics. Since the participants of the prior CP studies have not shown consistent naming patterns in the different contexts after treatment, no predictions about the facilitation in the semantic vs phonological conditions were made. In theory, facilitation due to contextual effects could be possible in both semantic and phonological conditions, as explained above. If the improvement in the unrelated condition is comparable to that in the related context conditions, we would assume that the facilitatory effect is due to repetition priming.

2. Case description At the time of the study, PH was a 73-year-old righthanded male who had suffered from an extensive left middle cerebral artery infarction affecting also nucleus lentiformis and the periventricular white matter and subsequent right-sided hemiparesis over 2 years prior to this study. Prior to the infarction, he had suffered from rheumatoid arthritis, and his right wrist was ankylosed (stiffened), making it difficult for PH to use his right hand in writing. After the infarction, PH had received intensive speech therapy for 1 year, after which one short speech therapy period and a few follow-up visits were arranged. At the time of the present study, he was not receiving speech therapy. Apart from language problems, no major cognitive defects had been observed in the clinical neuropsychological examinations. PH had an academic degree in agronomy and had worked as a director of a regional centre of agriculture. At the time of the present study, PH had already been retired for 11 years.

3. Background testing of language functions Prior to the training, PH underwent a comprehensive background testing including the Finnish version of the Boston Diagnostic Aphasia Examination (BDAE; Laine, Niemi, Koivuselka¨-Sallinen, & Tuomainen, 1997b), the Boston Naming Test (BNT; Laine, Koivuselka¨-Sallinen, Ha¨nninen, & Niemi, 1997a), and tasks probing the semantic and phonological processing of

words. The Appendix A gives further details about the experimental pre- and post-treatment tasks administered for PH. The performance on the BDAE (see Table 1) showed that PH had some difficulty in the auditory comprehension of words and phrases, although this was not obvious in informal conversations with him. Repetition of words as well as oral reading of words and sentences was intact. Reading comprehension was good with only

Table 1 PHÕs scores on the standardised Finnish version of the Boston Diagnostic Aphasia Examination (Laine et al., 1997b) before treatment Task

Before training

Severity rating

2

Fluency Articulation rating Phrase length Melodic line Verbal agility

7/7 5/7 5/7 10/14

Auditory comprehension Word discrimination Body-part identification Commands Complex ideational material

65.5/72 20/20 8/15 10/12

Naming Responsive naming Confrontation naming Animal naming

21/30 106/114 6

Oral reading Word reading Oral sentence reading

30/30 10/10

Repetition Repetition of words Repetition of high-probability sentences Repetition of low-probability sentences

10/10 8/8 8/8

Paraphasia Neologistic Literal Verbal Extended

1 1 3 0

Automatic speech Automatized sequences Reciting

7/8 2/2

Reading comprehension Symbol discrimination Word recognition Comprehension of oral spelling Word-picture matching Reading sentences and paragraphs

10/10 8/8 8/8 8/10 9/10

Writing Mechanics Serial writing Primer-level dictation Spelling to dictation Written confrontation naming Sentences to dictation Narrative writing

3/5 Not Not Not Not Not Not

attempted attempted attempted attempted attempted attempted

K. Renvall et al. / Brain and Language 95 (2005) 327–341

Table 2 PHÕs performance on the lexical-semantic tasks before treatment Task

Number correct

Proportion correct

50-picture classification task (Laine et al., 1992) Odd-one-out task 1 (Laine et al., 1992) Pictures Words Odd-one-out task 2 (Laine et al., 1992) Pictures Words Odd-one-out task 3 (Laine et al., 1992) Pictures Words Word-picture matching task (Laine et al., 1992) Category-specific word-picture matching task (Laine et al., 1998a) Category-specific odd-one-out task (Laine et al., 1998b) Pictures (living 5/8, nonliving 5/8) Words (living 7/8, nonliving 5/8) Spoken word-picture matching (PALPA 47; translation from Kay et al., 1992) Written word-picture matching (PALPA 48; translation from Kay et al., 1992) Auditory synonym judgements (PALPA 49; translation from Kay et al., 1992) Written synonym judgements (PALPA 50; translation from Kay et al., 1992)

47/50

0.94

7/11 8/11

0.63 0.72

8/12 9/12

0.67 0.75

11/20 12/20 19/20

0.55 0.60 0.95

47/48

0.98

10/16 12/16 39/40

0.63 0.75 0.96

36/40

0.90

51/60

0.85

50/60

0.83

Table 3 PHÕs performance on the naming and repetition tasks before treatment Task Naming tasks Boston Naming Test Action naming task 90-picture naming task (Laine et al., 1992) Category-specific naming task (Laine et al., 1998c) living 18/24, nonliving 22/24 Repetition tasks Word repetition task Nonword repetition task

Number correct

Proportion correct

33/60 43/60 68/90 40/48

0.55 0.72 0.76 0.83

90/90 88/90

1.00 0.98

the BNT that included at least two phonemes and were made within 20 s. Table 4 shows that the predominant error types in naming were semantically related words and (informative and uninformative) circumlocutions, as well as nonwords. On the BNT as well as on the other picture naming tasks, it was typical for PH to quickly produce a word which he then often correctly (but sometimes also incorrectly) rejected and started to search for another word, e.g., harp fi ‘‘cello is not a cello (pause) it looks like a cello but it is not a cello (pause) what is that (pause) it looks like a cello but something else (pause)’’ (translation). Sometimes the first response bore only relatively distant semantic relation to the target word although it belonged to the same broad superordinate category like animals, e.g., camel fi ‘‘dolphins,’’ sea horse fi ‘‘fly.’’ The nonwords were often short productions that did not resemble the target words (e.g., MAI¨ HKA ¨ ‘‘corn’’ fi [k ne]).2 In addition, PH SSINTA produced occasional unrelated word responses, e.g., TENNISMAILA ‘‘tennis racket’’ fi [peitto] ‘‘blanket.’’ To examine the familiarity effect on PHÕs naming performance, we further analysed the naming responses produced by PH in the 90-item picture naming task. A

one error in reading sentences and paragraphs. The performance on the easy 50-picture categorisation task and on the two word-picture matching tasks was again relatively good with 90–98% correct whereas problems occurred in all of the odd-one-out tasks (see Table 2). The semantic multiple-choice tasks administered in conjunction with the 90-picture naming task and requiring PH to provide semantic information of the unnamed targets, indicated that semantic knowledge of the targets was mostly available for PH but occasional problems did occur (choice of superordinate 91%; choice of specific semantic feature 82%). Compared with the preliminary normative data available for the 50-picture classification task, the odd-one-out tasks, the word-picture matching task, and the semantic multiple-choice tasks, PHÕs performance was worse than that of the normal controls, suggesting an impairment at the semantic level (see the Appendix A for further details of the normative data). PHÕs spontaneous speech was fluent, well-articulated, and grammatically correct but lacked informational content and was frequently interrupted by word-finding difficulties. His moderate naming difficulty was evident in the BDAE (see Table 1), as well as in the other naming tasks on which PH achieved 55–83% correct (see Table 3). To get a picture of the error types in PHÕs naming, we further analysed all his naming attempts in

331

2 To obtain a more exact view on the phonological proximity of PHÕs nonword responses, we calculated an index of the phonological correspondence between the targets and the nonword responses. The average phonological correspondence of the nonword errors to the targets in the BNT was 20%. The index was obtained by counting the number of the correct target-response phoneme correspondences using the following criteria: In the first syllable, one point was given if the nonword started with the correct phoneme and further points were given for additional (not necessarily successive) correct phonemes produced in their correct positions. In other syllabic positions, only the totally correctly produced syllables were counted and as many points were given as there were phonemes; additional points were, however, given for correct phonemes immediately preceding or following a correctly produced syllable. For example, the nonword response ‘‘popvi-ka’’ for the target ‘‘pap-ri-ka’’ (pepper) would give two points for the first syllable, two points for the last syllable, and also one point for the phoneme /i/ in the second syllable because it immediately precedes a correctly produced last syllable. This response would thus yield 5/7 points, i.e., 71% phonological correspondence between the target and the nonword response.

K. Renvall et al. / Brain and Language 95 (2005) 327–341

Occurrences

Semantically related words Circumlocutions Nonwords and partial words Semantic-then-phonological errors Formal paraphasias Unrelated words No responses Other

15 13 21 5 2 4 3 1

a The following criteria were used in the BNT and in all other naming measurements as well: (1) a naming response was considered correct if it was completely phonologically accurate or if there was a simplification of a consonant cluster (e.g., TRUMPETTI ‘‘trumpet’’ fi [rumpetti]) or a single consonant (e.g., KENGURU ‘‘kangaroo’’ fi [kenkuru]) usual for normal Finnish-speaking people; (2) nonwords and partial words included all phonological strings which were (a) not legal words and which were sharing less than 50% of the phonemes with a semantically target related word or (b) legal words as such but in relation to the expected target responses were partial productions of either target compound or target derived words e.g., RUOHONLEIKKURI ‘‘lawn-mower’’ fi [leikkuri], KARATEKA ‘‘karateka’’ fi [karate] ‘‘karate’’; (3) semantically related words were superordinates, category members and semantically associated words; (4) circumlocutions were descriptive phrases in place of correct target names; (5) semantic-then-phonological errors shared more than 50% of the phonemes with a semantically target related word; (6) formal paraphasias were real words resembling a target word phonologically by fulfilling at least two out of three criteria (the same first phoneme, the same phonemic structure in the first syllable, the same number of syllables); (7) unrelated words were words bearing neither semantic nor phonological relation to the target words; (8) no responses included situations where there was no response at all, when the response was the type ‘‘I do not know,’’ and where verbal responses consisted of a single phoneme only; (9) the other category included all the other erroneous words.

PH managed to name 28/30 of the high-familiarity pictures, 24/30 of the medium-familiarity pictures, and 16/ 30 of the low-familiarity pictures, showing a statistically significant familiarity effect (v2 = 13.48, df = 2, p < .005). In the multiple-choice tasks carried out in conjunction with the 90-picture naming task, PH could only tell the number of letters and syllables of those target words which he spontaneously named after deadline (target word letter number 9% correct; target word syllable number 23% correct). He also had difficulty selecting the first syllable of the to-be-named target among four written and spoken alternatives (four-choice word initial syllable recognition 59% correct) and some difficulty selecting the correct word among four alternatives that were simultaneously presented in written and spoken form (82% correct). Phonological cues like word initial syllable in the BNT (15/27 correct) and word initial phoneme in the 90-picture naming task (13/22 correct) were sometimes effective in cueing retrieval of the target word. Quite often though the phonological cues led PH

AA

Response typea

astray and he produced a semantic relative or a formal error, i.e., MERIHEVONEN ‘‘sea horse’’ fi [meduus ] ‘‘medusa,’’ MUSTEKALA ‘‘octopus’’ fi [muslimi] ‘‘muslim,’’ TURBAANI ‘‘turban’’ fi [tur-koosi] ‘‘turquoise.’’ PH was occasionally even led astray by his own erroneous productions, e.g., KAULIN ‘‘rolling pin’’ fi [k veli k peli] (‘‘kaaveli’’ is a nonword and ‘‘kaapeli’’ corresponds to ‘‘cable’’). With regard to the functional locus of PHÕs anomia, the data just presented point to a mainly semantic deficit. The main evidence is his impaired performance on a variety of semantic tasks, with more pronounced problems on the more difficult odd-one-out tasks. Furthermore, over 30% of PHÕs naming responses in the third baseline measurement (administered after the background testing but before treatment) were semantic errors. In fact, such a high percentage of semantic errors in the absence of some degree of formal and unrelated word errors is incompatible with the computational implementations of the interactive activation models of Dell et al. (1997) and Foygel and Dell (2000), as they model semantic paraphasias arising from faulty semantic-to-lexical mappings and not from malfunctioning semantic representations. Moreover, production of semantically related words to phonemic cues indicates a semantic problem. As PHÕs repetition of words and nonwords as well as oral reading of words and phrases was intact (see Tables 1 and 3), input and output phonology as such were functional and could not account for his naming difficulties. However, not all of PHÕs semantic naming errors had to be of central origin as he had a tendency to quickly reject his own errors (albeit sometimes also correct responses that suggests an imprecision in lexical-semantic choice), a feature that has been linked to a functional disconnection between semantics and phonology in word production (Lambon Ralph, Sage, & Roberts, 2000). This disconnection would affect an interplay between semantics and phonology that is needed to narrow down lexical choice amongst competing candidates to a single target. It is notable that PH did also produce nonword errors but they appeared only in semantically driven tasks such as naming and not in repetition or oral reading. This could be interpreted as a phonological access problem due to underspecified information from the semantic level. PHÕs nonword errors were often far from the target, indicating that access to the phonological form was disturbed at a very early point in phonological encoding. Does PHÕs rather mild but clear cut semantic impairment represent an access or a storage deficit? While it is difficult to give a definitive answer to this fundamental question, certain features of his performance point to an access rather than a storage problem. Between tasks, item-by-item analysis revealed performance variability in 44% of his answers on the verbal and pictorial AA

Table 4 Distribution of PHÕs erroneous responses across response type in the Finnish version of the Boston Naming Test (Laine et al., 1997a) before treatment

A

332

K. Renvall et al. / Brain and Language 95 (2005) 327–341

versions of the odd-one-out tasks 1, 2, and 3. Within tasks, he evidenced performance variability on 32% of the BNT and action naming items when the pre- vs post-treatment naming success was analysed.

4. Treatment programme 4.1. Stimuli To select suitable hard-to-name stimulus pictures for training and for control purposes, PH was asked to name a pool of approximately 700 black-and-white pictures of objects obtained from various sources. This was done twice during the background testing period with an approximately 2-week interval between the two separate naming attempts of each item. Pictures that PH had named correctly on both test occasions within 10 s were considered easy-to-name items (used as fillers, see below) and pictures that he could not name correctly on either occasion within 10 s were considered hard-to-name objects. To rule out the possibility that the naming difficulty for the hard-to-name objects was due to problems in recognising the objects, we asked PH to provide (verbally or mimicking) semantic information on each object he could not name, e.g., information as to how the objects are used or what they are used for. Those items for which he could not provide semantic information were discarded. Through a careful selection of 120 hard-to-name pictures (60 trained and 60 untrained), we created five picture sets for each of the following three conditions: (1) Semantic condition: In each picture set there were four target items and four semantically close untrained control items, e.g., the target words were KIRSIKKA ‘‘cherry,’’ PERSIKKA ‘‘peach,’’ ANANAS ‘‘pineapple,’’ KARPALO ‘‘cranberry’’; ¨ LE ‘‘grape,’’ the control words VIINIRYPA APRIKOOSI ‘‘apricot,’’ MELONI ‘‘melon,’’ KIIVI ‘‘kiwi.’’ We tried to select items that would not be phonologically related but there are occasional items in the sets that share the first phoneme. (2) Phonological condition: In each picture set, there were four phonologically related target items which shared at least the same initial CV, and four control items that were phonologically as close as possible. The trained target items and the untrained control items shared at least the first phoneme and were semantically unrelated, e.g., the target words KARATEKA ‘‘karateka,’’ KALOTTI ‘‘skullcap,’’ KALEIDOSKOOPPI ‘‘kaleidoscope,’’ KALOSSI ‘‘galosh’’; the control words KAAPPI ‘‘cupboard,’’ KASETTI ‘‘cassette,’’ KAULAKORU ‘‘necklace,’’ KAKKU ‘‘cake.’’ (3) Unrelated condition: In each picture set, there were four target items and four control items. In

333

this condition, the items were neither semantically nor phonologically related to each other, e.g., the target words MOKKASIINI ‘‘moccasin,’’ PELLE ‘‘clown,’’ SUPPILO ‘‘funnel,’’ LEIMASIN ‘‘rubber stamp’’; the control words AKVAARIO ‘‘aquarium,’’ JEEPPI ‘‘jeep,’’ MIKROSKOOPPI ‘‘microscope,’’ VIIRI ‘‘streamer.’’ Lemma frequency, word length, and number of syllables were matched across the different conditions and also across the target and control items within each condition. Lemma frequencies were derived from a massive newspaper corpus using a specially designed lexical search program (Laine & Virtanen, 1999).

5. Treatment design The treatment period started with three baseline measurements (Baseline 1, 2, and 3) each administered at 1week intervals. Six days after the third baseline measurement, the treatment was started. There were altogether 12 separate treatment sessions (2–3 sessions per week) which were carried out within 5 weeks. Five days after the last treatment session, the first post-treatment measurement (Post 1) was administered followed by the second post-treatment measurement (Post 2) 5 days after the first post-treatment measurement. To study the endurance of the treatment effects, one follow-up measurement was administered 1.5 months after the treatment had been finished (1.5 months follow-up). The naming measurements included the 120 hard-to-name pictures (described above). In addition, 30 easy-to-name fillers were employed in the naming measurements to make the tests more comfortable for PH. The pictures were shown on a computer screen, and PH was given 5 s to name each picture. After this the examiner changed the picture to a blank screen, and then to a new picture when PH indicated that he was ready to move on to the next item. No feedback on performance was given. The presentation order of the pictures was randomised separately for each measurement. All naming attempts within 5 s were fully transcribed and scored. The contextual priming method used for PH is described in Table 5. Instead of training all the 15 picture sets within a treatment session, picture sets were cycled so that each picture set was treated in nine sessions. The overall duration of a treatment session usually was 45–60 min.

6. Data analysis and scoring Five kinds of treatment-related data were analysed: (1) the number of correct responses on the six naming measurements; (2) the distribution of naming responses

334

K. Renvall et al. / Brain and Language 95 (2005) 327–341

Table 5 Experimental protocol for PH Step 1: Step 2:

Step 3: Step 4: Step 5:

One matrix of four pictures from one of the contextual conditions (semantic, phonological, and unrelated) is displayed and the examiner points to four pictures in a random order. PH tries to name the pictures on each occasion; (a) if PH manages to name all the pictures correctly within 5 s, no contextual priming training is started but another matrix of four pictures is displayed and the procedure is restarted from Step 1. (b) if PH produces 1–4 naming errors, contextual priming training is started from Step 3. Examiner points to four pictures in a random order and at the same time names the pictures altogether four times. PH repeats the names on each occasion. After this, the procedure is restarted from naming attempts in Step 1 and continued to Step 4 if naming errors occur during the naming attempts (in Step 2). The naming-repetition cycle is repeated up to five times after which another matrix from one of the contextual conditions is displayed even though PH had not been able to name all the previous four pictures.

in the four naming measurements (Baseline 3, Post 1, Post 2, 1.5 months follow-up); (3) the number of contextual word errors (from within the target set) within the treatment sessions; (4) the number of noncontextual word errors (from outside the target set holding a semantic or phonological relation to the target names) within the treatment sessions; (5) the number of correct responses within the treatment sessions. In all analyses, all naming attempts produced within 5 s were transcribed and scored. The criteria for different response types were the same as those used in the pre-treatment tasks (see the footnote for Table 4) except for the within-treatment response analyses which distinguished between contextual and noncontextual word errors on the basis of whether the erroneous response was a substitution of another item from within a picture set (contextual error) or from outside the picture set but semantically and/or phonologically related to the target word (noncontextual error). A noncontextual word error was classified as only phonologically related to the target word if it shared at least the first two phonemes with the target word and was not semantically related to the target word. A noncontextual word error was

considered both semantically and phonologically related to the target word (a ÔmixedÕ error) if it was both semantically related to the target word and fulfilled at least two out of the following three phonological criteria: the same first phoneme, the same phonemic structure in the first syllable, the same number of syllables.

7. Results 7.1. Naming measurements The most important goal in this study was to explore whether treatment would have a facilitative effect on PHÕs naming. The percent of totally correct responses in the six naming measurements before and after the treatment are shown in Fig. 1. To get an idea of PHÕs performance during the treatment period, we have also included in the figure his performance on naming the trained items within the treatment sessions. As we did not administer separate naming measurements during the treatment period, we calculated the percent of correct naming responses each time a new picture set was

Fig. 1. PHÕs naming performance on the trained and control items in the three pre- and three post-treatment measurements and on the trained items within the treatment sessions. The nine within-treatment scores represent the percent of the correct naming responses calculated each time a new picture set was displayed and PH was asked to name the pictures for the first time within each session.

K. Renvall et al. / Brain and Language 95 (2005) 327–341

displayed and PH was asked to name the pictures for the first time within each session. Thus, although the withintreatment results are not directly comparable to those of the pre- and post-naming measurements where pictures were presented in isolation, they provide an indication of the trajectory of PHÕs response to the treatment over the course of intervention. The number of correct responses during treatment is further analysed in the light of context sensitivity (see below). We present nine within-treatment measurement points as each picture set was cycled and treated in nine sessions although the overall number of the treatment sessions was 12. To analyse the naming success before and after the treatment we used the McNemar and binomial tests (e.g., Siegel, 1956). They revealed that there was a significant change in the naming success on the trained items between the Baseline 3 (3/60 correct) and the Post 1 measurement (18/60 correct; McNemar test with continuity correction, v2 = 11.53, df = 1, p < .0001) and also between the Baseline 3 and the Post 2 measurement (14/ 60 correct; McNemar test with continuity correction, v2 = 7.69, df = 1, p < .001). However, this change was no longer observed when the Baseline 3 measurement was compared with the follow-up measurement administered 1.5 months after the treatment (8/60 correct). Also, no significant change was observed between pre- and post-tests for the control items. Interestingly, when the conditions were analysed separately (see Fig. 2), the semantic condition was the only condition in which we found significant improvement on naming the trained items. This occurred when the Baseline 2 (1/20 correct) and Baseline 3 (0/20) were compared with the Post 1 measurement (7/20 correct, binomial test, Baseline 2 vs Post 1, p < .032, Baseline 3 vs Post 1, p < .016). In all conditions, the best performance was either in the Post 1 or 2 measurement although this was still relatively low (35% correct in the semantic, 30% correct in the phonological, 30% correct in the unrelated). Performance dropped to the baseline level in the follow-up measurement. 7.2. Distribution of naming responses in the four naming measurements We analysed the naming response distributions on Baseline 3, Post 1, Post 2, and 1.5 monthsÕ follow-up (Fig. 3) to detect possible changes in the error-type distribution and in the target-response relationships of the nonword errors. First, a similar trend is seen in both correct responses and nonwords/partial words, both response types yielding highest proportions right after the treatment was completed (in the Post 1 measurement), dropping in a week (in the Post 2 measurement), and staying approximately at the same level 1.5 months after the treatment as before the treatment. As for nonwords/ partial words, we again calculated the target-response

335

phonological correspondence (see Footnote 2) which was low in the Baseline 3 (11.6%), but somewhat increased in the Post 1 measurement (19.9%) while going down to the base level in the Post 2 (12.1%) and the Post 1.5 months (13%). An opposite trend is seen in the rate of semantically related words as their proportion was lowest in the Post 1 measurement but yielding again almost the same level in the Post 2 measurement as in the Baseline 3. The proportion of circumlocutions decreased in the Post 1 and 2 measurements but increased again in the follow-up measurement. These findings bring further evidence of only a short-term effect of the CP treatment on PHÕs naming. 7.3. Context sensitivity during treatment We looked at three types of responses in the naming attempts during treatment: (1) contextual word errors (names of other items in the same picture set); (2) noncontextual word errors (words that are not included in the picture set but hold a semantic or phonological relation to the target); (3) correct responses. We found that PH produced many contextual word errors (137 errors) and noncontextual word errors (114 errors). The rates of the contextual errors and all the other responses in the semantic condition were then compared to those in the phonological and in the unrelated conditions using v2 analyses. This revealed that more contextual errors occurred in the semantic condition (75 of 687 attempts to name) than in the phonological condition (34 of 626 attempts to name, v2 = 12.24, df = 1, p = .0005, corrected for continuity) or in the unrelated condition (28 of 641 attempts to name, v2 = 18.97, df = 1, p < .0001, corrected for continuity). As for the noncontextual word errors, there were more such errors in the unrelated condition (52 errors) than in the semantic condition (26 errors, v2 = 10.46, df = 1, p < .005, corrected for continuity), while no significant difference in the rates of noncontextual errors and other responses were found between the unrelated and the phonological condition (33 errors). Most of the noncontextual errors were semantically related to the target words (87 errors), whereas 16 errors were both semantically and phonologically related to the target words (of which 12 occurred on one item in the unrelated condition) and eight errors were only phonologically related to target words. No difference was found in the number of correct and erroneous responses across the different conditions during the treatment period (semantic: 361 correct, 326 erroneous; phonological: 325 correct, 301 erroneous; and unrelated: 332 correct, 309 erroneous). 7.4. Post-treatment testing with the background tests Post-treatment testing, administered immediately after the treatment had finished, included the Finnish

336

K. Renvall et al. / Brain and Language 95 (2005) 327–341

Fig. 2. PHÕs naming performance on the trained and control items of the three pre- and three post-treatment measurements and on the trained items within the treatment sessions in the different conditions.

version of the BNT, the experimental action-naming task, and three selected lexical-semantic tasks which were also used in the pre-treatment testing (odd-oneout 1 with words, odd-one-out 2 with pictures, category-specific odd-one-out with pictures; see Table 2 for pre-treatment scores). Comparison of the scores in the BNT and in the experimental action naming task before and after treatment (BNT pre-treatment 33/60 vs posttreatment 26/60; action naming pre-treatment 42/60 vs post-treatment 40/60) reveals practically no change in

PHÕs naming performance on these tasks after treatment. Also the distribution of the different responses in the BNT before (see Table 4) and after treatment (see Table 6) are almost alike; the few more errors after treatment were mostly semantically related words and semanticthen-phonological errors. The phonological proximity of the nonword/partial word responses (see Footnote 1) in the BNT decreased from the pre-treatment 20% to post-treatment 7%. Similarly to the naming tasks, no change was observed in the pre- vs post-treatment

K. Renvall et al. / Brain and Language 95 (2005) 327–341

337

Fig. 3. Distribution of the naming responses in the Baseline 3, Post 1, Post 2, and 1.5 months follow-up measurements.

Table 6 Distribution of PHÕs erroneous responses across response type in the Finnish version of the Boston Naming Test (Laine et al., 1997a) after treatment Response typea

Occurrences

Semantically related words Circumlocutions Nonwords Semantic-then-phonological errors Formal paraphasias Unrelated words No responses Other

21 13 23 9 4 7 2 3

a

See the criteria for different response types in Table 4.

lexical-semantic tasks (odd-one-out 1 with words, pre 8/11 vs post 8/11; odd-one-out 2 with pictures, pre 8/12 vs post 4/12; category-specific odd-one-out with pictures, pre 10/16 vs post 10/16 with 5/8 living and 5/8 nonliving in both).

8. Discussion PH is the second semantically impaired aphasic who has now been treated with the contextual priming technique. The first such aphasic, AS in Martin et al.Õs (2004a) study, suffered not only from a semantic deficit but also to some extent a phonological deficit. Compared with AS, PH presented with a purer semantic deficit, as his performance on a variety of semantic tasks was more errorful and repetition intact. Thus, it was of interest to determine whether PH would show a similar pattern of treatment effects as AS, i.e., immediate interference in the semantic condition and at most short-term gains. Such a result would help to delineate the limitations of the CP technique in its current form (repetition priming plus context), i.e., that it would be

most effective when semantic processing of words is relatively spared (Martin et al., 2004a). This hypothesis gained support in the present case study. Similarly to AS in the Martin et al. (2004a) study, PH showed short-term gains of treatment, as the naming of the trained items improved in the measurements administered 5 and 10 days after the treatment. However, no long-term improvement was observed and the improvement on trained items was at most modest, as in the case of AS. The naming response distributions on the four naming measurements showed similar hints of shortterm effects; an increase in the proportion of correct responses and nonwords/partial words in the Post 1 measurement after which their proportion went down again while the proportion of semantically related words and circumlocutions decreased in the Post 1 measurement and increased almost to the baseline level in the 1.5 monthsÕ follow-up. Finally, facilitation in naming the untrained items, i.e., generalisation, was not found. In addition, two interesting phenomena were observed. First, when looking at the different contexts, significant short-term facilitation of trained items was only found in the semantic condition (between the two last baseline measurements and the Post 1 measurement). Second, PH made more contextual word errors in the semantic condition compared with the phonological and unrelated conditions, indicating immediate interference, again, in the semantic condition and thus at the level of PHÕs deficit. The latter finding is also consistent with the error data obtained by Martin et al. (2004a) and supports the suggestion that interference during training is more likely when the context and underlying source of the lexical processing impairment match. These two findings together imply that early interference does not necessarily preclude some benefit from a treatment. We also looked at the noncontextual word errors as they can be one indication of spreading activation and

338

K. Renvall et al. / Brain and Language 95 (2005) 327–341

the mechanism enabling generalisation of CP treatment to untrained, related entries in the lexicon (see e.g., Laine & Martin, 1996; Martin et al., 2004a, 2004b; Renvall et al., 2003). The great number of semantically related noncontextual word errors, compared with only a few phonologically related noncontextual errors, is not surprising as PH produced semantically related errors in all naming tasks, a pattern that is typical for semantically impaired aphasics in general (see e.g., Caramazza & Hillis, 1990). It should be noted that in the unrelated condition there was one item causing more noncontextual word errors than any other item (the erroneous word being related to the target item phonologically, semantically, and also visually). Therefore, it is likely that the noncontextual errors reflect PHÕs anomia type and the array of items used rather than a treatment-specific increase in the activation spreading and generalisation effects. Looking at the results of both this and the earlier Martin et al. (2004a) study, it seems that despite its positive short-term effects, in its current form CP treatment is not sufficient for those aphasics who have a semantic deficit underlying their anomia. This might be because via automatic spreading activation, the context in CP serves to boost a whole semantic (or phonological, depending on context employed) neighbourhood. If the patientÕs main problem is in activating the correct semantic representation in a semantically driven task such as naming, this contextual boost is not channeled to the specific item but instead spreads more diffusely in the neighbourhood, maintaining the abnormal competition between lexical items. Thus, it fails to strengthen the item-specific semantic–phonological connections. In this situation, the ‘‘bottom–up’’ repetition priming component may be not be particularly helpful either, as it does not help in sharpening the semantic representations that would be needed to overcome the problems with ‘‘top–down’’ activation, i.e., conceptually driven lexical access. It is important to note that the most promising results with CP treatment have been achieved with those participants whose semantic processing has been relatively intact (see Martin et al., 2004a; Renvall et al., 2003). To develop the CP method further, one should examine ways to modify the method so that it yields longer-lasting positive effects even in semantically impaired patients. One approach could be to provide more item-specific semantic support in order to stimulate meaning-form links for each trained item in a more focused fashion. Semantic anomia patients like PH might also benefit more from a treatment that primes the connections between semantics and phonological representations in the top–down direction. There are some treatment techniques that presumably stimulate top– down processing, e.g., reteaching distinctions between semantically related items (Hillis, 1991, 1998), semantic

feature analysis (e.g., Coelho et al., 2000), and self-generated semantic cues (e.g., Marshall, Freed, & Karow, 2001). These have met with some success in aphasia treatment. Finally, it would be important to track the time course from interference to facilitation more exactly as our study suggests that despite immediate interference treatment may reach some benefit. This is a phenomenon which may be important to take into account in clinical settings too when selecting appropriate treatment techniques. These lines of research, we hope, will help to increase the gains from CP treatment in general and its efficacy for semantic anomia patients in particular.

Appendix A. Details of the experimental tasks used in assessing PHÕs naming, repetition, and lexical-semantic knowledge Lexical-semantic tasks: (1) 50-picture classification task (Laine, Kujala, Niemi, & Uusipaikka, 1992). The task measures the ability to make simple semantic categorization. A participant is asked to classify 50 pictures into five semantic categories (pieces of clothes, fruits and vegetables, parts of body, home equipments, and animals) by placing the object pictures under the written names of the categories (also spoken before the task performance). The items are selected from the same group of pictures as is used in the 90-picture naming task (see below). Laine et al. (1992) report that the task has been administered for five neurologically intact subjects (age 50–70 years) and their performance was flawless. (2) Odd-one-out tasks 1, 2, and 3 (Laine et al., 1992). In the odd-one-out tasks 1 and 2, the task is to decide which of the two items (pictures or words) would go together best with the target (e.g., cowhorse, lion). In the odd-one-out task 3, the participant is shown five items belonging to the same semantic category. The task is to select the item that differs most from the other four (e.g., gorilla–rhinoceros–kangaroo–crocodile–cat). All three tasks can be performed with pictures and words (written and spoken). Laine et al. (1992) present the preliminary control data of five neurologically intact subjects on the odd-one-out task 1 (mean 10.6, range 10–11, maximum 11). To get an idea of how neurologically intact subjects perform on the odd-one-out tasks 2 and 3, we have tested nine 61–79-year-old controls (mean age 67.1). In the odd-one-out task 2, the average number of correct responses was 11.4 (range 10–12. maximum 12) based on pictorial stimuli (n = 5) and 11.5 (range

K. Renvall et al. / Brain and Language 95 (2005) 327–341

(3)

(4)

(5)

(6)

(7)

11–12, maximum 12) based on words (n = 4). In the odd-one-out task 3, the average number of correct responses was 16.8 (range 14–18, maximum 20) based on pictorial stimuli (n = 4) and 19.0 (range 18–20, maximum 20) based on words (n = 5). Word-picture matching task (Laine et al., 1992). The names of 20 items were presented simultaneously orally and in written form. PH was asked to choose the corresponding picture out of five alternatives. Laine et al. (1992) report that the task has been administered for five neurologically intact subjects (age 50–70 years) and their performance was flawless. Category-specific word-picture matching task (Laine, Schmied, & Trefzer, 1998a). PH was asked to show the picture out of eight alternatives that was spoken and written to him. There are 48 target pictures, half of the pictures representing living and the other half nonliving objects. Category-specific odd-one-out task (Laine, Schmied, & Trefzer, 1998b). PH was shown three pictures, the task being to decide which of the two pictures would go together best and which one is left out (e.g., bus-car, bicycle). The task was performed with pictures as well as with words (which PH both saw and heard at the same time) as stimuli. A half of the pictures (24) represent living and the other half (24) nonliving objects. Spoken and written word-picture matching tasks (PALPA 47 and 48) The Finnish version of the task has been translated from the original English version of PALPA (Kay, Lesser, & Coltheart, 1992). Two pictures have been re-drawn so that they would better resemble the prototypes of the corresponding words (pictures of a ‘‘plug’’ and a ‘‘stamp’’). Auditory and Written synonym judgements tasks (PALPA 49 and 50). As regards the Finnish version of these synonym judgement tasks, it should be noted that most of the stimulus words were translations from the words used in the original English version of PALPA (Kay et al., 1992) but the imageability of the Finnish translation equivalents has not been evaluated separately. Hence, only preliminary conclusions from imageability can be drawn based on these tasks.

Naming tasks: (8) Action naming task. The task includes 126 action pictures which we have chosen from a pool of approximately 200 new black-and-white outline drawings acquired for the construction of a Finnish action naming task. Since this work is in progress, the stimuli in the present set are not yet

339

controlled for factors such as name agreement, lexical frequency, image agreement, or visual complexity. This is why PHÕs performance can inform us about his overall verb naming ability only. (9) 90-picture naming task. The task is shortened version of that of Laine et al. (1992). When a participant can not name a target picture he/she is asked five subsequent questions which are presented both spoken and written: First, he/she is asked to identify the subordinate semantic category of the target from the choice of four representing either animals, pieces of clothes, pieces of furniture, parts of the body, fruits and vegetables, tools, vehicles, or home equipments. Second, he/ she is asked to identify the specific semantic feature belonging to the target from the choice of four, e.g., the four choices for the target ÔcrocodileÕ were ‘‘it can fly,’’ ‘‘is a pet,’’ ‘‘you can ride on it,’’ ‘‘it is dangerous.’’ Third, he/she is asked to estimate whether the sought word is short (arbitrarily defined as six phonemes or less) or long (over six phonemes). Fourth, he/she is presented with four initial syllables of words that were semantically closely related to the target. One of them was correct, two were the first syllables of words that were semantically closely related to the target, and one was from a word that was neither semantically nor phonologically related to the target. Fifth, he/she is asked to recognise the target out of four words including the target word, two semantically related words, and one phonologically related word (sharing the initial letter/phoneme with the target). In the syllable and word recognition tasks the items were presented simultaneously in written form and orally. The original 106-picture naming task of Laine et al. (1992) has been administered for five neurologically intact subjects (age 50–70 years). They performed the choice of the superordinate and the choice of the specific semantic feature 100% correct. (10) Category-specific naming task (Laine, Schmied, & Trefzer, 1998c). The task includes 48 relatively easy to-be-named pictures of which half represent living (fruits, vegetables, and animals) and the other half nonliving items (tools, furniture, clothes, and vehicles). Repetition tasks: (11) Word repetition task. PH was asked to repeat 90 words that were divided into three frequency categories (low-frequency, medium frequency, and high-frequency) based on a computerised newspaper corpus including 22.7 million Finnish word tokens (Laine & Virtanen, 1999). Lemma fre-

340

K. Renvall et al. / Brain and Language 95 (2005) 327–341

quency was 4.7 ± 3.0 for the low-frequency words, 50.3 ± 20.7 for the medium-frequency words, and 208.3 28.6 for the high-frequency words. In addition, the words were divided into three categories based on the phoneme length (4, 6, and 8 letters) so that each frequency category included 10 exemplars from each length-category. (12) Nonword repetition task. PH was asked to repeat 90 nonwords that were created from the real words used in the word repetition task (see above) by changing one to three phonemes (mean 2.2 phonemes). Thus, there were an equal number of nonwords with 4, 6, and 8 phonemes. References Basso, A., Marangolo, P., Piras, F., & Galluzzi, C. (2001). Acquisition of new ‘‘words’’ in normal subjects: A suggestion for the treatment of anomia. Brain and Language, 77, 45–59. Bastiaanse, R., Bosje, M., & Franssen, M. (1996). Deficit-oriented treatment of word-finding problems: Another replication. Aphasiology, 10, 363–383. Best, W. M., Howard, D., Bruce, C., & Gatehouse, C. E. P. (1997). Cueing the words: A single case study of treatments for anomia. Neuropsychological Rehabilitation, 7, 105–141. Caramazza, A., & Hillis, A. E. (1990). Where do semantic errors come from? Cortex, 26, 95–122. Cave, C. B. (1997). Very long-lasting priming in picture naming. Psychological Science, 8, 322–325. Coelho, C. A., McHugh, R. E., & Boyle, M. (2000). Semantic feature analysis as a treatment for aphasic dysnomia: A replication. Aphasiology, 14, 133–142. Cornelissen, K., Laine, M., Tarkiainen, A., Ja¨rvensivu, T., Martin, N., & Salmelin, R. (2003). Adult brain plasticity elicited by anomia treatment. Journal of Cognitive Neuroscience, 15, 444–461. Davis, A., & Pring, T. (1991). Therapy for word-finding deficits: More on the effects of semantic and phonological approaches to treatment with dysphasic patients. Neuropsychological Rehabilitation, 1, 135–145. Dell, G. S. (1986). A spreading-activation theory of retrieval in sentence production. Psychological Review, 93, 283–321. Dell, G. S., & OÕSeaghdha, P. G. (1992). Stages of lexical access in language production. Cognition, 42, 287–314. Dell, G. S., Schwartz, M. F., Martin, N., Saffran, E. M., & Gagnon, D. A. (1997). Lexical access in aphasic and nonaphasic speakers. Psychological Review, 104(4), 801–838. 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. Foygel, D., & Dell, G. S. (2000). Models of impaired lexical access in speech production. Journal of Memory and Language, 43, 182–216. 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. Hickin, J., Herbert, R., Best, W., Howard, D., & Osborne, F. (2002). Phonological therapy for word finding difficulties: A re-evaluation. Aphasiology, 16, 981–999. Hillis, A. E. (1991). Effects of separate treatments for distinct impairments within the naming process. In T. Prescott (Ed.). Clinical aphasiology (Vol. 19, pp. 255–265). Austin, TX: Pro-Ed. Hillis, A. E. (1998). Treatment of naming disorders: New issues regarding old therapies. Journal of the International Neuropsychological Society, 4, 648–660.

Hillis, A. E., & Caramazza, A. (1994). Theories of lexical processing and rehabilitation of lexical deficits. In M. J. Riddoch & G. W. Humphreys (Eds.), Cognitive neuropsychology and cognitive rehabilitation (pp. 449–484). Hove: Lawrence Erlbaum. Kay, J., Lesser, R., & Coltheart, M. (1992). Psycholinguistic assessments of language processing in aphasia (PALPA). Hove: Psychology Press. 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, and Hearing Research, 46, 773–787. Laine, M., Koivuselka¨-Sallinen, P., Ha¨nninen, R., & Niemi, J. (1997a). Bostonin nimenta¨testi [Finnish version of the Boston Naming Test]. Helsinki: Psykologien Kustannus. Laine, M., Kujala, P., Niemi, J., & Uusipaikka, E. (1992). On the nature of naming difficulties in aphasia. Cortex, 28, 537– 554. Laine, M., & Martin, N. (1996). Lexical retrieval deficit in picture naming: Implications for word-production models. Brain and Language, 53, 283–314. Laine, M., Niemi, J., Koivuselka¨-Sallinen, P., & Tuomainen, J. (1997b). Bostonin diagnostinen afasiatestisto¨ [The standardised Finnish version of the Boston Diagnostic Aphasia Examination]. Helsinki: Psykologien Kustannus. Laine, M., Schmied, W., & Trefzer, K. (1998a). Category-specific naming test (for research purposes only). Department of Psychology, University of Turku. Laine, M., Schmied, W., & Trefzer, K. (1998b). Category-specific wordpicture matching test (for research purposes only). Department of Psychology, University of Turku. Laine, M., Schmied, W., & Trefzer, K. (1998c). Category-specific oddone-out test (for research purposes only). Department of Psychology, University of Turku. Laine, M., & Virtanen, P. (1999). WordMill lexical search program. Centre for Cognitive Neuroscience, University of Turku. Lambon Ralph, M. A., Sage, K., & Roberts, J. (2000). Classical anomia: A neuropsychological perspective on speech production. Neuropsychologia, 38, 186–202. Marshall, R. C., Freed, D. B., & Karow, C. M. (2001). Learning of subordinate category names by aphasic subjects: A comparison of deep and surface-level training methods. Aphasiology, 15, 585– 598. Marshall, J., Robson, J., Pring, T., & Chiat, S. (1998). Why does monitoring fail in jargon aphasia? Comprehension, judgement, and therapy evidence. Brain and Language, 63, 79–107. Martin, N., Dell, G. S., Saffran, E. M., & Schwartz, M. F. (1994). Origins of paraphasias in deep dysphasia: Testing the consequences of a decay impairment to an interactive spreading activation model of lexical retrieval. Brain and Language, 47, 609–660. Martin, N., Fink, R., & Laine, M. (2004a). Treatment of word retrieval deficits with contextual priming. Aphasiology, 18, 457–471. Martin, N., Fink, R., Laine, M., & Ayala, J. (2004b). Immediate and short-term effects of contextual priming on word retrieval in aphasia. Aphasiology, 18, 867–898. Martin, N., & Laine, M. (2000). Effects of contextual priming on impaired word retrieval. Aphasiology, 14, 53–70. Martin, N., Weisberg, W., & Saffran, E. M. (1989). Variables influencing the occurrence of naming errors: implications for models of lexical retrieval. Journal of Memory and Language, 28, 462–485. Miceli, G., Amitrano, A., Capasso, R., & Caramazza, A. (1996). The treatment of anomia resulting from output lexical damage: Analysis of two cases. Brain and Language, 52, 150–174. Nettleton, J., & Lesser, R. (1991). Therapy for naming difficulties in aphasia: Application of a cognitive neuropsychological model. Journal of Neurolinguistics, 6, 139–157.

K. Renvall et al. / Brain and Language 95 (2005) 327–341 Nickels, L. A. (1997). Spoken word production and its breakdown in aphasia. Hove, UK: Psychology Press. Nickels, L. (2002). Therapy for naming disorders: Revisiting, revising, and reviewing. Aphasiology, 16(10/11), 935–979. Nickels, L., & Best, W. (1996). Therapy for naming disorders (Part I): Principles, puzzles and progress. Aphasiology, 10(1), 21–47. Raymer, A. M., Thompson, C. K., Jacobs, B., & Le Grand, H. R. (1993). Phonological treatment of naming deficits in aphasia: Model-based generalization analysis. Aphasiology, 7(1), 27– 53.

341

Renvall, K., Laine, M., Laakso, M., & Martin, N. (2003). Anomia treatment with contextual priming: A case study. Aphasiology, 17, 305–328. Siegel, S. (1956). Nonparametric statistics for the behavioral sciences. Tokyo: McGraw Hill Kogakusha. Visch-Brink, E. G., Bajema, I. M., & Van De Sandt-Koenderman, M. E. (1997). Lexical semantic therapy: BOX. Aphasiology, 11, 1057–1115. Weigl, E. (1961). The phenomenon of temporary deblocking in aphasia. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung, 14, 337–364.