Treatment of semantic verb classes in aphasia - Semantic Scholar

Clinical Linguistics & Phonetics, Early Online, 2010, 1–20

Treatment of semantic verb classes in aphasia: acquisition and generalization effects

Clin Linguist Phon Downloaded from informahealthcare.com by 173.66.17.61 on 03/24/11 For personal use only.

YASMEEN FAROQI-SHAH, & LAUREN E. GRAHAM Department of Hearing and Speech Sciences, University of Maryland, College Park, MD, USA (Received 10 June 2010; Accepted 4 December 2010)

Abstract Verb retrieval difficulties are common in aphasia; however, few successful treatments have been documented (e.g. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2006). Towards theory-driven therapies for aphasic verb impairments: A review of current theory and practice. Aphasiology, 20, 1159–1185). This study investigated the efficacy of a novel verb retrieval treatment in two individuals with aphasia who experience verb retrieval difficulty. It involved training verb classes with large (e.g. cut verbs) and limited (e.g. contact verbs) sets of semantic features. Based on action representation theories, semantically based training of cut verbs was predicted to generalize to retrieval of untrained cut and contact verbs. One participant improved on trained verbs whereas the other participant did not. Neither participant demonstrated within nor across-class generalization to untrained verbs. However, both participants significantly improved in verb naming as measured by An Object and Action Naming Battery, and their predominant error pattern changed from noun to verb substitutions. Therefore, both participants improved in overall verb retrieval strategies despite limited success with verbs trained in this treatment. Implications for the design of future treatments are discussed.

Keywords: aphasia, verb, treatment, semantic features, generalization

Introduction Difficulty in retrieving words is the most pervasive symptom of aphasia. Recent work on grammatical class differences suggests that verb retrieval deficits may be more common than noun retrieval deficits (Zingeser and Berndt, 1990; Berndt, Mitchum, Haendiges, and Sandson, 1997; Kim and Thompson, 2000; 2004; Luzzatti, Raggi, Zonca, Pistarini, Contardi, and Pinna, 2002; Mätzig, Druks, Masterson, and Vigliocco, 2009). For instance, in a naming study of a large group of aphasic patients, 52 of 58 individuals were either worse with verbs or comparably affected for nouns and verbs (Luzzatti et al., 2002). Several authors agree that the higher complexity of verbs relative to nouns at least partially accounts for their greater retrieval difficulty (Black and Chiat, 2003; Conroy, Sage, and Lambon-Ralph, 2006). Factors that influence greater verb complexity include low imageability, a looser concept, typically more grammatical morphology and higher syntactic weight with argument structure information (Black and Chiat, 2003). Greater retrieval difficulties have been observed with verbs requiring more arguments (dative versus intransitive Correspondence: Yasmeen Faroqi-Shah, Department of Hearing and Speech Sciences, University of Maryland, 0141F, Lefrak Hall, College Park, MD 20742, USA. Tel: 301-405-4229. Fax: 301-314-2023. E-mail: [email protected] ISSN 0269-9206 print/ISSN 1464-5076 online © 2011 Informa UK Ltd. DOI: 10.3109/02699206.2010.545964


2

Y. Faroqi-Shah & L. E. Graham

verbs) despite intactness of grammatical and syntactic knowledge of verbs (Kim and Thompson, 2000; 2004). Given that sentence structure is inherently dependent on the verb, unsurprisingly, moderate association has been reported between verb retrieval deficits and grammatical formulation deficits in non-fluent aphasia (Berndt, Mitchum, et al., 1997; Luzzatti et al., 2002). In addition to the above-mentioned syntactic aspects, semantic factors are also known to influence verb retrieval. It is reported that aphasic patients are better able to retrieve semantically complex ‘heavy’ verbs (e.g. run, bake) compared with general ‘light’ verbs (e.g. go, make; Berndt, Haendiges, Mitchum, and Sandson, 1997; Breedin, Saffran, and Schwartz, 1998; Kim and Thompson, 2004; Barde, Schwartz, and Boronat, 2006). Errors involving substitutions with semantically associated verbs (clean!wipe)/nouns (clean!soap) and difficulties differentiating subtle semantic differences between verbs are also reported (McCarthy and Warrington, 1985; Mitchum, Ritgert, Sandson, and Berndt, 1990; Breedin et al., 1998; Kemmerer and Tranel, 2000b). Therefore, although verbs with a greater number of semantic features seem to be more successfully retrieved, they may still be off-target by one or two specific semantic features. A few prior studies have targeted semantic aspects of verbs for the treatment of verb retrieval deficits (McNeil, Dolye, Spencer, Goda, Flores, and Small 1998; Wambaugh, Linebaugh, Doyle, Martinez, Kalinyak-Fliszar, and Spencer 2001; Raymer and Ellsworth, 2002; Wambaugh and Ferguson, 2007). Wambaugh et al. (2001) and Wambaugh, Cameron, Kalinyak-Fliszar, Nessler, and Wright (2004) examined the use of semantic cueing for verb retrieval. McNeil et al. (1998) focused on the generation of synonyms and antonyms of the verb being trained using a hierarchy of cues to aid the retrieval of unsuccessfully named items. Another semantically based treatment is semantic feature analysis (SFA), initially proposed by Boyle and Coelho (1995) for nouns. Wambaugh and Ferguson (2007) modified SFA for verb retrieval, where they required their participants to generate the semantic characteristics of each verb, such as the body part/tool used, agent of the action, its purpose and location. Outcomes of these verb naming treatments have been mixed, and improvements are training-specific. That is, although the naming of trained verbs typically improves, generalization to the retrieval of untrained verbs seldom occurs. Given that generalization to untrained stimuli is the gold standard in aphasia treatment, it is worth examining whether any other treatment approach might not only improve verb retrieval but also facilitate generalization to related untrained verbs. It should be noted that syntactic, phonological and gestural treatments for verb retrieval have also reported the same lack of generalization to untrained verbs (Raymer and Ellsworth, 2002; Schneider and Thompson, 2003; Raymer, Ciampitti, Holliway, Singletary, Blonder, Ketterson, Anderson, Lehnen, Heilman and Rothi 2007; Rose and Sussmilch, 2008; Conroy, Sage, and Lambon-Ralph, 2009; Edmonds, Nadeau, and Kiran, 2009; Boo and Rose, 2010; Links, Hurkmans, and Bastiaanse, 2010). Although the potency of any treatment approach is determined by participants’ impairment profile and other prognostic indicators, few aspects of previous studies could have accounted at least partly for the lack of generalization to untrained verbs. First, verbs that were used for treatment were often selected from a list of unsuccessfully named items on a naming test such as the Object and Action Naming Battery (e.g. Kim, Adingono, and Revoir, 2007; Wambaugh and Ferguson, 2007). This means that the verbs were often unrelated to each other in semantic, thematic or phonological features. Hence, during treatment, each verb likely stimulated (at least partially) non-overlapping semantic networks, diminishing the potential for neural plasticity (Kleim and Jones, 2008). Second, verbs used to test the generalization effects also often had little systematic relationship with the trained verbs, evidently yielding little treatment-related improvement (Marshall, Pring, and Chiat, 1998). Training of restricted semantic classes and testing of generalization to semantically related words have been examined for noun retrieval in


Verb treatment for aphasia

3

aphasia, with successful outcomes (see Kiran and Thompson, 2003; Kiran, 2007). But, to our knowledge, the use of reinforcement of a cohesive set of semantic features within a single verb class has not been applied to verb retrieval deficits in aphasia. As described later, in this study, training and generalization verbs were selected on the basis of semantic relatedness. This study was motivated by theoretical frameworks of action representation such as the two-level theory, which assumes that verbs have two levels of semantic representation (Pinker, 1989; 2007; Levin, 1993; Levin and Rappaport Hovav, 2005; Wunderlich, 2006). One semantic level is suggested to be an event template generic to all verbs within a semantic class and includes information about argument and predicative structure. The second semantic level represents the unique features of each verb that differentiate it from other verbs of the same class. For example, the verb class cut in Levin’s (1993) classification includes numerous verbs, all of which share an event template – using a tool to break into pieces. Each member of this verb class – mince, chop, hack, saw, dice, etc. – possesses additional unique and idiosyncratic features that specify the manner of cutting motion and the shape/size of the resultant pieces (Figure 1). Psycholinguistic, neuroimaging and neuropsychological investigations support this two-level semantic representation (Kable, Lease-Spellmeyer, and Chatterjee, 2002; Kemmerer, 2003; Kemmerer, Castillo, Talavage, Patterson, and Wiley, 2008). Interestingly, a neuroimaging study in which participants made semantic judgements to verb triads (e.g. trudge-limp-stroll) found relatively distinct neural activations for semantic features such as action, motion and contact (Kemmerer et al., 2008). This study lends support to the neural instantiation of action semantics. Recent work on the mental representation of actions and their linguistic labels (i.e. verbs) strongly suggests that verbs trigger a complete or partial simulation of the action they refer to (Embodied Cognition; Tyler, Bright, Fletcher, and Stamatakis 2004; Bergen, 2007; Fisher and Zwaan, 2008; Kemmerer et al., 2008; Masson, Bub and Warren, 2008; see reviews by Faroqi-Shah, Wood, and Gassert (2010); Fernandino and Iacoboni, 2010; but see Hauk,

TEMPLATE

ROOT

CUT (superordinate class) to make into pieces +motion, +action, +contact, +tool use, +change of state

MINCE +results in minute pieces, +rapid repetitive motion

CHOP +results in large pieces, +repetitive motion

CRUSH +results in a smashed mass, +heavy tool +downward motion

SLICE +results in thin or flat pieces, +slow deliberate motion

Figure 1. Schematic illustration of the two-level theory of verb meaning using cut verbs in Levin’s (1993) taxonomy. Conceptual class-specific information is represented at the template level, whereas distinctive verb-specific features are represented at the root level.


4


Johnsrude and Pulvermuller (2004) for a different view). Although deficits in verb comprehension in aphasic individuals have been previously reported (Saygin, Wilson, Dronkers, and Bates, 2004; Fazio, Cantagallo, Craighero, Ausilio, Roy, Pozzo, Calzolari, Granieri, and Fadiga, 2009), it is noteworthy that at least some aspects of online action simulation are preserved in verb-impaired aphasic individuals (Faroqi-Shah et al., 2010). The implications of the two-level theory and the embodied cognition framework for the treatment of aphasic verb retrieval deficits are explored in this study. The assumption is that if verbs of the same semantic class are used for treatment, the same ‘template level’ semantic network should be repeatedly activated, and thus strengthened. Subsequently, retrieval of untrained members of this verb class should be facilitated because of the shared template features. SFA, a treatment approach mentioned earlier, has a format that can incorporate training of template and root-level semantic features of verbs. SFA uses explicit practice of conceptual attributes of words such as their location, function and physical attributes by asking patients to verbally generate these features for each target word. Participants also sort a given list of features into those which are and are not attributes of the target word. In this study, feature generation and sorting steps focused on template and root-level properties of verbs within a class (Levin, 1993). The complexity account of treatment efficacy (Thompson, Shapiro, Kiran, and Sobecks, 2003) posits that treatment effects generalize to linguistically related but less complex structures. Complexity in word retrieval is marked by (1) the extent of shared semantic features and (2) the atypicality of the item to its semantic class (Kiran, 2007). As per Levin’s (1993) taxonomy of verbs, distinct verb classes may overlap in a subset of semantic features. For instance, cut verbs have five template features (þaction, þmotion, þcontact, þtool use and þchange of state),of which the first three features also characterize contact verbs (e.g. nudge, tickle, kiss, bump; Table I). In contrast, some verb classes have no feature overlap (see verbs of non-verbal expression taken from Levin (1993) in Table I). As per the complexity account of treatment efficacy and the feature list in Table I, one might expect treatment of cut verbs to generalize to untrained contact verbs (more to less complex), but not vice versa. Treatment of cut or contact verbs should produce no change in non-verbal expression verbs as the template features do not coincide. Kiran and Thompson (2003) found that treatment of nouns with a larger number of (and less typical) semantic features improved naming of generic and more typical items that encoded fewer semantic features. That is, semantically oriented treatment with birds such as penguin and ostrich resulted in improved naming of robin (see also Kiran, 2007). To our knowledge, no study has utilized a complexity framework for semantically oriented verb treatment in aphasia (see Schneider and Thompson (2003), for a verb treatment based on argument structure complexity). This study The primary purpose of this study was to investigate verb retrieval following a novel treatment approach focusing on semantic features of verb classes. The treatment procedure was an

Table I. Verb classes and their semantic features. Verb class Cut Contact Non-verbal expression

Examples

Contact

Motion

Action

Tool use

Change of state

Dice, chop Bump, scratch Yawn, smile

þ þ

þ þ

þ þ

þ

þ



5

adaptation of SFA. Two verb classes, cut and contact, were selected for the treatment because of partial overlap in their semantic features (see Table I). The second purpose was to determine whether treatment effects would generalize to untrained verbs (1) within the same class that shares all features of trained verbs (trained cut to untrained cut); (2) from a different class that shares some, but not all, of the trained semantic features (trained cut to untrained contact; trained contact to untrained cut); and (3) in a different class that shares no semantic features with the trained verbs (trained cut to non-verbal expression; trained contact to non-verbal expression). It was predicted that training of cut verbs would generalize to contact verbs (Pashek, 1998; Wambaugh, Doyle, Martinez, and Kalinyak-Fliszar, 2002; Thompson et al., 2003), but generalization of trained contact verbs to cut verbs was not expected because of the larger number of semantic features encoded by cut verbs. Improvement of non-verbal expression verbs after cut or contact verb training was not predicted due to the lack of feature overlap with the trained verb category. The additional purpose of including verbs of non-verbal expression was to establish the specificity of treatment effects to trained and semantically related verbs (as per a multiple baseline design). An additional exploratory question was whether there would be change in verb retrieval in other measures, particularly the Object and Action Naming Battery (OANB) (Druks and Masterson, 2000). Methods Participants Two male participants with aphasia were recruited for the study. Participants provided informed consent before participation, in accordance with the ethical standards set forth by the Declaration of Helsinki. P1, a 62-year-old male who was a native Chinese speaker, was also premorbidly fluent in English for 30 years. P2, a 47-year-old male, was a native speaker of English. Both participants had developed aphasia consequent to a single left-hemisphere cerebrovascular accident of the middle cerebral artery territory, were at least one-year post-onset, had at least high school education and had no premorbid history of psychiatric, neurological, cognitive or speech-language deficits. Both participants passed binaural pure-tone audiometric screening (P1: aided; P2: unaided) at 500, 1000 and 2000 Hz at 25 dBHL (ANSI: 1969) and passed a vision screen (at least 20/40 corrected or uncorrected vision and the absence of spatial neglect and visual field deficits). Similarly, both participants demonstrated adequate reading for single words and short phrases, as determined by a screening test before initiation of treatment. Neither participant showed significant signs of verbal apraxia (as per the criteria listed in Apraxia Battery for Adults, 2nd ed., Dabul (2000)). Demographic details are given in Table II. Both participants met the following inclusionary language criteria, which are given in Table II: (1) A language profile consistent with Broca’s aphasia as per the Western Aphasia Battery (WAB; Kertesz, 1982). (2) Verb retrieval difficulty in narrative speech (evidenced by use of 0.05). Two 5-s long videos were developed to illustrate each of the 35 verbs so that different videos could be used during the treatment steps and for testing the acquisition of treatment verbs. All videos showed an actor performing the target action on a recipient. For example, the verb chop was filmed with a male actor chopping celery in one video and a female actor chopping onions in another video. Naming accuracy for the videos was obtained by showing the videos to 15 unimpaired non–brain-damaged, native English-speaking volunteers (8 males and 7 females; age mean ¼ 54.87 years, education mean ¼ 16.13 years). These participants were asked to provide a name for the action and were prompted to provide a synonym in the case that the target verb was not elicited. Videos with less than 80% naming accuracy were re-filmed and normed again. Verbs for which video naming accuracy failed to reach 80% following two norming procedures were not used in the study. Still images of the actions obtained from each



7

Figure 2. Images of the verbs (a) crush (a cut verb), and (b) nudge (a contact verb).

video were used during two treatment steps (generation and analysis of semantic features) as a visual reminder of the action (Figure 2). Treatment Overall design. A multiple baseline alternating treatments (ABACA) single-participant design was planned. For each participant, baseline testing (A), in which verb naming was elicited without any direct treatment, was performed over two testing sessions to establish the lack of spontaneous improvement. Hence each participant served as his own experimental control. This was followed by a phase of treatment (B or C) during which participants received four 1-h


8


treatment sessions every week (1 h/day). Naming of treatment verbs (henceforth called treatment probes) was tested every treatment session to monitor the change in accuracy of trained verbs. Testing for accuracy of untreated verbs (henceforth generalization probes) was done once every three sessions to minimize improvement due to repeated exposure (Fink, Schwartz, Sobel, and Myers, 1997). Three a priori criteria were established for cessation of treatment: (1) accuracy of 6 of 7 on three consecutive treatment probes; (2) less than 30% increase from baseline treatment verb naming accuracy after eight sessions; or (3) maximum of 15 treatment sessions. Posttreatment testing, including treatment probes, generalization probes, the WAB (Kertesz, 1982) and OANB (Druks and Masterson, 2000), was completed upon cessation of treatment to determine the changes in language measures. Four weeks following the cessation of treatment, maintenance (A) of verb naming for trained and untrained stimuli was tested. To investigate whether the verb class used for treatment would influence the pattern of generalization within and across verb classes, the study was planned such that each participant received treatment with a different verb class during the first treatment phase (P1 ¼ contact verbs; P2 ¼ cut verbs), followed by the other verb class during the second treatment phase (i.e. P1 ¼ cut verbs; P2 ¼ contact verbs). The second treatment phase was planned to be initiated 2 weeks later to ensure the lack of carry-over from the previous treatment phase and all verb naming was tested before the onset of this second phase. Unfortunately, P2 was unable to participate in his second treatment phase due to difficulties with transportation. Hence, P1 received an ABACA design, whereas P2 received an ACA design (B ¼ contact verbs, C ¼ cut verbs). During phase 1, P1 received treatment with 7 contact verbs, whereas generalization was tested to untrained contact, cut and non-verbal expression verbs. The treatment and generalization verbs for each participant are listed in Appendix 3. Treatment protocol. Every treatment session began with the administration of treatment probes, in which the action names of the seven treatment verbs were elicited using videos. As mentioned earlier, generalization probes were administered once in every third session. Whenever the participant failed to produce the target verb during treatment and generalization probes, he was allowed to write the action name. If participants produced a more general verb (e.g. cut for slice), they were prompted to be more specific. No other feedback or prompting was given during probes. Following elicitation of probes, four treatment steps (modified from SFA, Boyle and Coelho (1995)) were used for each of the seven treatment verbs. (1)

(2)

Naming of the action in a video: Incorrect responses were redirected with verbal corrections and feedback until the correct response was produced. The next two treatment steps focused on delineating template and root-level semantic features (as per the two-level theory described in the Introduction section). Generation of semantic features: A still image of the action was displayed, and the participant was instructed to independently generate three features for the target verb. This was initially modelled to the participants so they had a clear understanding of the kinds of responses that were needed. The following are the examples of features: This action requires a tool/knife, or This action results in small/large pieces. If the participant was unable to independently produce three features, the researcher provided verbal yes/no prompts such as, Does it require a tool? The participant repeated these prompted features to strengthen his repertoire of semantic characteristics for each treatment verb.

Verb treatment for aphasia (3)


(4)

9

SFA: This step required the participant to determine whether a given set of semantic features belonged or did not belong to the target verb. Four cards, each containing a semantic feature that was relevant or irrelevant to the verb, were placed in a row in front of the participant. In general, the semantic features listed in each of the four cards were a template-level feature that characterized the entire treatment verb class (e.g. use a tool for cut verbs), a root-level feature unique to the specific target verb (e.g. results in small pieces for mince), one feature of another verb in the same class but that did not apply to the target verb (e.g. results in large pieces for mince) and one irrelevant feature (e.g. uses one’s legs for mince). The participant read each feature and responded by placing each card in a YES or NO column. Sentence generation: The video of the target action was replayed and the participant was instructed to produce a sentence using the target verb (such as The man is mincing the onion). When responses did not contain the target verb or were incomplete phrases, the researcher prompted the participant to use the target verb or use a complete sentence. The purpose of sentence generation was to strengthen the semantic network by activating the noun associates of the action (Edmonds, Nadeau, and Kiran, 2009). If the participant was still unable to provide a complete sentence using the target verb, the researcher provided a sentence model.

The use of still images of the action in steps 2 and 3 is consistent with the procedures of SFA and is intended to enable visual imagery of the action while the participant generates and analyses semantic features. In steps 1 and 4, videos of the action were used to facilitate retrieval of the verb label and the corresponding verb arguments. These steps were repeated for each of the seven treatment verbs. The order of the treatment verbs was randomized each session. Treatment was terminated whenever one of the three aforementioned predetermined criteria was met. Scoring and data analysis Verb naming responses were scored as accurate if the target verb was produced either orally or in written form. Minor phonemic paraphasias and spontaneous self-corrections within 10 s were accepted if the response was unambiguous. Verb naming errors (for probes and OANB) were classified as verb substitutions, noun substitutions, phonemic errors and others (I don’t know or no response). Given the single-subject design of this study, participants served as their own experimental control. Thus, the effects of treatment were determined by comparing each participant’s preand post-treatment performance in verb naming accuracy and other language measures. Effect sizes were calculated to determine the magnitude of treatment and generalization gains for treatment and generalization probes using the following formula (Busk and Serlin, 1992): Posttreatment naming accuracy Mean pretreatment naming accuracy Standard deviation of pretreatment naming accuracy

Reliability Every treatment session was audio- and video-recorded with participant consent for reliability scoring. A trained research assistant checked whether the predetermined treatment steps were accurately followed (independent variable) and scored the probes (dependent measures). Reliability was established for adherence to treatment procedures in 9 of 19 treatment sessions

10


at 100%. Reliability for scoring of probes was completed for seven sessions and this exceeded 90% (Cohen’s Kappa ¼ 0.97). Scoring differences between the original and reliability scorers were resolved by listening to audio-recordings of the probes and arriving at a consensus.


Results Both participants performed the treatment steps rapidly and typically completed two or more iterations of the training verb set every treatment session. Both participants’ ability to generate each verb’s semantic features, sort semantic features and produce the trained verbs in a sentence improved dramatically during the course of treatment, and they made no errors on these treatment steps after the first two to three treatment sessions. However, success in retrieving each verb’s name, both during the treatment steps and in probes, was different for P1 and P2. The response of each participant to treatment is presented separately in the following sections. Participant P1 P1 was trained on contact verbs during the first treatment phase and his naming accuracy of treatment and generalization probes over different sessions is shown in the left side of Figure 3. His treatment probe accuracy changed from a mean pre-treatment accuracy of 0 of 7 to 7 of 7 in five treatment sessions totalling 3.75 treatment hours (McNemar’s change test, p < 0.05). There was no generalization to untrained contact, cut or non-verbal expression verbs (pre- to posttreatment score changes are 0/7 to 1/7, 1.5/7 to 3/7 and 2/7 to 2/7, respectively). The right side of Figure 3 shows P1’s response to the second treatment phase with cut verbs, which was initiated after a washout period of 2 weeks after treatment with contact verbs ended. P1 achieved the criterion of 6 of 7 verbs in six treatment sessions (total of 4.5 h of treatment). P1 was significantly more accurate with naming of trained cut verbs in post-treatment testing compared with baseline (McNemar’s change test, p < 0.05; effect size ¼ 2.85). There were no statistically significant changes for untrained cut verbs (pre-treatment ¼ 2/7, post-treatment 3/7), untrained contact verbs (pre-treatment ¼ 0/7, post-treatment ¼ 1/7) or for verbs of non-

Contact-Tx Cut-Tx Contact-Gen Cut-Gen Non-verbal expresssion

Sessions Figure 3. Naming accuracy of treatment and generalization probes for P1. Sessions T1–T5 show response to contact verb treatment and sessions T6–T11 represent response to cut verb training. Treatment probes are connected by a solid line. Generalization probes were administered every third session (see text for details).


11

verbal expression (pre-treatment ¼ 2/7, post-treatment ¼ 2/7) (McNemar’s change test, p > 1 for all comparisons). P1 demonstrated maintenance of trained cut verbs when tested 4 weeks after treatment, with accuracy of 5 of 7. To summarize, P1 improved significantly in naming of trained verbs during both phases of treatment in a relatively short number of sessions and maintained these effects. However, generalization to untrained verbs was not found.


Errors on treatment probes. P1’s errors were primarily semantically related verb substitutions (17/24 errors; compared with 3/24 noun substitutions and 4/24 other responses). Verb substitutions were distributed between substitutions of the superordinate verb class name (e.g. cut or contact) and within-class substitutions (e.g. crush for hack, punch for chip). The pre-treatment pattern of noun substitutions changed to verb substitutions (Table III). Participant P2 P2 participated in cut verb treatment only and his performance on treatment and generalization probes is shown in Figure 4. P2 failed to reach the criterion of 6 of 7 verbs by the end of eight sessions and hence his treatment was terminated. He participated in a total of 8 h of direct treatment. There was no significant change in verb naming accuracy from baseline to post-treatment of any verb category. Non-verbal expression verbs were produced at a higher accuracy than other verbs even during baseline, and this category remained highly accurate (pre-treatment accuracy ¼ 5/7, post-treatment accuracy ¼ 6/7). Errors on treatment probes. P2’s errors were primarily substitutions by the superordinate verb class name (cut, 7/21 instances), or overuse of a single verb target (dice for slit, chip, shred, 11/ 21 instances), or no responses (see Table III). Other language measures Post-treatment scores of the WAB and OANB are given in Table II. Both participants increased significantly in the WAB aphasia quotient post-treatment, specifically in the spontaneous speech subtest. The pre- and post-treatment WAB narrative samples are given in Appendix 1. Both participants also increased significantly for action naming on the OANB (McNemar’s (1969) change test; p < 0.05). There was no change in object naming on the OANB for either participant. As seen in Table III, both participants’ errors showed a similar change from pre- to posttreatment, namely, an increase in verb substitutions and a decrease in noun substitutions. Table III. Distribution of errors for each participant for the Object and Action Naming Battery (Druks and Masterson, 2000) responses and during treatment probes. Error categories Substitutions

P1

P2

Object and Action Naming Battery (Pre-Tx) Object and Action Naming Battery (Post-Tx) Treatment probes Object and Action Naming Battery (Pre-Tx) Object and Action Naming Battery (Post-Tx) Treatment probes

Verb

Noun

Phonemic

Other

21/39 16/16 17/24 18/36 18/23 35/43

17/39

1/39

3/24 11/36 2/23

4/24 7/36 3/23 8/43

12



Cut-Tx Contact-Gen Cut-Gen Non-verbal expresssion

Sessions Figure 4. Naming accuracy of treatment and generalization probes for P2. P2 received training with cut verbs only (sessions T1–T8). Treatment probes are connected by a solid line. Generalization probes were administered every third session (see text for details).

Discussion This study investigated the efficacy of a verb retrieval treatment focusing on the identification of template- and root-level semantic features unique to two verb classes, cut and contact verbs, in two aphasic individuals. Acquisition of trained verbs and generalization to untrained verbs within and across verb classes were investigated. Training cut verbs was predicted to generalize the retrieval of contact verbs, but not vice versa. No improvement in unrelated non-verbal expression verbs was expected. This treatment was novel because exemplars of a specific verb class were trained, and training emphasized recognition and generation of class-general and verb-specific semantic features. A single-participant alternating treatments design was planned. P1 was trained on contact followed by cut verbs. He improved in both although generalization was negligible. P2 could participate in only cut verb treatment, and his retrieval of trained and generalization stimuli remained unchanged. This mixed success has been previously reported (Kim et al., 2007; Wambaugh et al., 2004) and will be discussed later. Interestingly, both participants improved in overall verb retrieval measured by naming accuracy in the OANB. And, their predominant error pattern changed from noun to verb substitutions. The implications of these findings are discussed in the following sections. Treatment effects P1’s improvement on trained cut and contact verbs indicates that this semantically oriented treatment may be appropriate for some individuals with aphasia. On the basis of the two-level theory of action representation, we can suggest that strengthening of the repeatedly practised event template and verb-specific features may underlie this improvement (Pinker, 1989; 2007; Levin, 1993; Levin and Rappaport Hovav, 2005; Wunderlich, 2006). However, P2’s failure to improve was unexpected because he appeared to be a good candidate for the



13

treatment based on spared semantic knowledge of verbs, adequate auditory comprehension, absence of severe phonological paraphasias and young age. There are several possible explanations for P2’s response. First, post hoc analysis of P2’s responses in baseline and treatment probes revealed that he consistently produced errors on the same items but demonstrated high accuracy on their treatment steps (e.g. chip, crush). This could suggest loss of phonological representations of these verbs as a result of brain damage (Butterworth, 1992). Verbs whose phonological representations have been erased are less likely to respond to treatment because the phonological label has to be re-learned, rather than just reaccessed. P1, in contrast, was inconsistent in errored items, suggesting impaired access, hence accounting for his rapid response to treatment. Or, the source of P2’s difficulty with verbs may have generally impaired access to phonological representations. Because phonological cueing or practice was not a part of this study, P2 may not have benefited from treatment. A second plausible explanation is that P2 was unfamiliar with the training verbs because of lower premorbid education level than P1 (Table II). Even though mere exposure to verbs (with no explicit treatment) has been shown to improve verb retrieval (Fink, Martin, Schwartz, Saffran, and Myers,1992; Fink et al., 1997), P2 demonstrated no such facilitatory effect suggesting that both phonological limitations and premorbid unfamiliarity with the vocabulary are plausible explanations for poor success. The mixed treatment outcomes of this study are consistent with the variable outcomes of prior verb treatment studies (e.g. Wambaugh et al., 2004; Kim et al., 2007). For instance, in Wambaugh et al.’s (2004) investigation of phonological and semantic cueing verb treatments in five aphasic individuals, two participants showed dramatic improvement with treatment, whereas two others showed only modest effects and one participant was unresponsive to treatment. Generalization effects The second question posed in this study was whether this treatment would facilitate the retrieval of closely related verbs within and across trained verb classes. Neither P1 nor P2 improved on generalization verbs, although P1’s naming of untrained contact verbs showed an improving trend (4/7 during session T11 versus 0/7 at baseline; Figure 3). Hence, the prediction of spread of treatment was not supported. This lack of generalization is consistent with previous verb treatment studies (Raymer and Ellsworth, 2002; Kim et al., 2007; Wambaugh and Ferguson, 2007; Edmonds, Nadaeu, and Kiran, 2009). SFA is shown to improve the retrieval of untrained nouns but not verbs (Wambaugh and Ferguson, 2007). In this study, there are at least three explanations for this lack of generalization. First, practice with class-general semantic features (event template) appears insufficient in facilitating retrieval of specific exemplars within that verb class. That is, strengthening of features such as tool use and results in small pieces with treatment of shred is insufficient in facilitating the retrieval of other verbs with the same event template, such as grind and dice. Based on the twolevel theory of verb representation, P1’s successful acquisition of trained verbs and lack of generalization to very similar untrained verbs implies that each verb’s specific features may need to be associated with its respective verb label for improvement in verb naming to occur. A second explanation for the lack of generalization may be the content of the semantic feature treatment itself, specifically the extent to which contrasts between verb-specific features were (not) emphasized. The two crucial treatment steps that were aimed at strengthening action representation were semantic feature generation and SFA. Perhaps, the number of features generated (three in this study) and the number of features analysed (four in this study) were insufficient to boost all relevant verb-specific features, resulting in limited strengthening of features crucial for generalization verbs. Aspects of action representation,


14


such as the path, direction and speed of motion, could be more relevant than the features actually generated in this treatment (Wu, Morganti, and Chatterjee, 2008). Likewise, the treatment and generalization verbs did not entirely overlap for the body part used, a feature that is increasingly being recognized as relevant for action representation (Pulvermuller, 2005; Faroqi-Shah et al., 2010). The importance of training verb-specific features is consistent with prior findings that individuals with aphasia experience difficulty in discerning subtle semantic differences among verbs in the same class (Kemmerer and Tranel, 2000a; b). A third explanation may have been the subtlety of differences between various trained and untrained verbs, both in terms of semantic features and filmability, especially for the cut verbs. However, this does not quite explain the lack of within-class generalization for contact verbs, which were more distinct from one another (e.g. bump and scratch). Another generalization question that this study asked was whether there would be any change in verb naming as measured by the OANB. Both participants improved significantly in verb, but not object naming, after treatment. Although prior research has shown that some aphasic participants improve in naming with repeated exposure (Fink et al., 1992), the specificity of improvement in verb rather than noun naming makes this an unlikely explanation of P1’s and P2’s improvements in OANB. The semantic feature treatment may have facilitated more effective use of some general retrieval strategies (Boyle and Coelho, 1995; Wambaugh and Ferguson, 2007). Interestingly, the trained cut verb class encodes five features (Table I) that overlap with several verb classes, and especially with some verbs in the OANB. For example, pouring is characterized as þmotion and þaction (as well as being a hand verb, like all cut verbs). It is noteworthy that several OANB verbs that were accurately named by P1 and P2 posttreatment (P1: raking, ringing, watering, pouring; P2: waving, knitting, pulling, drawing) share features with cut verbs (þhand motion, þtool use, þaction, þmotion). The only exception is waving, which is described as -tool use. Perhaps what were learned during treatment were not necessarily the names of specific treatment verbs, but rather a strategy for accessing semantic features to facilitate verb retrieval. The change in error patterns from pre- to post-treatment testing is noteworthy and further supports an improved verb retrieval strategy after treatment. Limitations of the study and implications for future verb retrieval treatments This is an exploratory Phase I treatment study, intended to test hypotheses about treatment efficacy (Robey, 2004). Given the partial success of the treatment, future research for further replication with more participants is warranted. It is also crucial to tease apart effects of semantic feature strengthening from those of mere repetition of the phonological form of the verb, which inadvertently occurs during naming trials. Although the improvements in OANB are promising, given the single treatment experimental design, it is yet to be determined whether other treatment approaches (e.g. phonological cueing or verb argument structure therapy) using these same stimuli would have produced similar effects. In interpreting the verb-specific improvement in OANB scores, we note the possibility that noun naming may have had little room for improvement because of initial high performance on noun naming. P2’s unexpected failure to respond to treatment indicates the need to include more extensive pre-testing to understand candidacy for this treatment. Furthermore, P2 was unable to complete the second half of the alternating treatments design, providing an incomplete picture of this response to treatment. Future research with a larger group of participants and using an alternating treatments design (where effects of different treatments are compared within the same participants) would address concerns about the replicability of these treatment outcomes and improve the generalizability of the findings.


15


Treatment outcomes may have also been impacted by certain methodological aspects of the study. As already alluded, an aspect that could be improved in future research is the use of more extensive semantic analysis and feature generation, thereby tapping more aspects of action representation. A limitation of the study was the relatively small number of stimuli in each verb category, making tests of statistical significance rather unreliable and limiting the ability to observe clear generalization effects. As described earlier, the small numbers were unavoidable due to the specificity of the verb classes being tested. However, there are other published studies of small number of treatment verbs (Fink et al., 1997). Extensive testing of linguistic and non-linguistic variables, such as premorbid language proficiency, digit span, motivation, access to phonological versus semantic representations and learning style, would also be useful in interpreting treatment outcomes. Conclusions The results of this study suggest that selecting and training verbs based on semantic features may improve verb retrieval for some individuals with aphasia. This training may also facilitate naming of untrained stimuli due to an improved strategy for verb retrieval, as evidenced by increase in the number of verb paraphasias. However, there are participant variables that may influence treatment outcomes, such as the nature of the verb deficit and premorbid familiarity with the stimuli, which warrant cautious interpretation of the generalizability of these findings to other aphasic individuals. These need careful consideration to determine candidacy for treatment. This study reiterates the findings of prior research that generalization effects are more evasive with verb retrieval treatments when compared with similar noun retrieval treatments (e.g. Kiran and Thompson, 2003). This could be attributed to the widely acknowledged representational complexity of verbs compared with nouns (Gentner, 1981; Kemmerer and Tranel, 2000a; b; Conroy et al., 2006; Mätzig et al., 2009; Vigliocco, Vinson, Druks, Barber, and Cappa, in press). When Kemmerer and Tranel (2000a) examined the influence of a variety of psycholinguistic variables such as argument structure and instrumentality in the verb retrieval of 53 individuals with aphasia, they found high individual variability among the influence of these factors. Predictions of treatment outcomes are likely to improve in the future with improved knowledge of the nature of verb deficit across participants. Acknowledgements This study was completed as Lauren Graham’s master’s thesis at the University of Maryland, College Park. The authors thank Amanda Peterson and Mohan Singh for help with reliability scoring and Erin Larter for help with error analysis. The authors are also grateful to the individuals with aphasia and their families for their participation in this research. Declaration of interest: The authors report no conflict of interest. The authors alone are responsible for the content and writing of this paper. Note 1. Normative data were obtained from descriptions of Western Aphasia Battery’s (Kertesz, 1982) picnic picture by 12 age-matched unimpaired volunteers. These individuals produced an average of 14.58 verbs while describing the picnic picture (range: 5–21).

16



References Baayen, R., Piepenbrock, R., & van Rijn, H. (1993). The CELEX lexical database (Release 1). [CD-ROM]. Philadelphia: Linguistic Data Consortium, University of Pennsylvania [Distributer]. Barde, L. H. F., Schwartz, M. F., & Boronat, C. B. (2006). Semantic weight and verb retrieval in aphasia. Brain and Language, 97, 266–278. Bergen, B. (2007). Experimental methods for simulation semantics. In M. Gonzalez-Marquez, I. Mittleberg, S. Coulson, & M. Spivey (Eds.), Methods in cognitive linguistics (pp. 276–301). Amsterdam: John Benjamins. Berndt, R. S., Haendiges, A. N., Mitchum, C. C., & Sandson, J. (1997). Verb retrieval in aphasia: 2. Relationship to sentence processing. Brain and Language, 56, 107–137. Berndt, R. S., Mitchum, C. C., Haendiges, A. N., & Sandson, J. (1997). Verb retrieval in aphasia: 1. Characterizing single word impairments. Brain and Language, 56, 68–106. Black, M., & Chiat, S. (2003). Noun-verb dissociations: a multi-faceted phenomenon. Journal of Neurolinguistics, 16, 231–250. Boo, M., & Rose, M. (2010). The effects of repetition, semantic, and gesture treatments for verb retrieval and use in Broca’s aphasia. Aphasiology, First published on: 13 August 2010 (iFirst), doi: 10.1080/02687031003743789. Boyle, M., & Coelho, C. A. (1995). Application of semantic feature analysis as a treatment for aphasic dysnomia. American Journal of Speech-Language Pathology, 4, 94–98. Breedin, S. D., Saffran, E. M., & Schwartz, M. F. (1998). Semantic factors in verb retrieval: an effect of complexity. Brain and Language, 63, 1–31. Busk, P. L., & Serlin, R. C. (1992). Meta-analysis for single case research. In T. R. Kratochwill and J. R. Levin (Eds.), Single Case Research Design and Analysis (pp. 197–198). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Butterworth, B. (1992). Disorders of phonological encoding. Cognition, 42, 261–286. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2006). Towards theory-driven therapies for aphasic verb impairments: a review of current theory and practice. Aphasiology, 20, 1159–1185. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2009). A comparison of word versus sentence cues as therapy for verb naming in aphasia. Aphasiology, 23, 462–482. Dabul, B. L. (2000). Apraxia battery for adults (2nd ed.). Austin, TX: Pro-Ed Inc. Druks, J., & Masterson, J. (2000). An object and action naming battery. Philadelphia, PA: Taylor & Francis. Edmonds, L. A., Nadeau, S. E., & Kiran, S. (2009). Effect of Verb Network Strengthening Treatment (VNeST) on lexical retrieval of content words in sentences in persons with aphasia. Aphasiology, 23, 402–424. Faroqi-Shah, Y., Wood, E., & Gassert, J. (2010). Verb impairment in aphasia: a priming study of body part overlap. Aphasiology, 24, 1377–1388. doi: 10.1080/02687030903515362. Fazio, P., Cantagallo, A., Craighero, L., Ausilio, A. D., Roy, A. C., Pozzo, T., Calzolari, F., Granieri, E., & Fadiga, L. (2009). Encoding of human action in Broca’s area. Brain, doi:10.1093/brain/awp118. Fernandino, L., & Iacoboni, M. (2010). Are cortical motor maps based on body parts or coordinated actions? Implications for embodied semantics. Brain and Language, 112, 44–53. Fink, R. B., Martin, N., Schwartz, M. F., Saffran, E. M., & Myers, J. L. (1992). Facilitation of verb retrieval in aphasia: a comparison of two approaches. Clinical Aphasiology, 21, 263–275. Fink, R. B., Schwartz, E. M., Sobel, P. R., & Myers, J. L. (1997). Effects of multilevel training on verb retrieval: is more always better?. Brain and Language, 60, 41–44. Fisher, M. H., & Zwaan, R. A. (2008). Embodied language: a review of the role of the motor system in language comprehension. The Quarterly Journal of Experimental Psychology, 61, 825–850. Gentner, D. (1981). Some interesting differences between verbs and nouns. Cognition and Brain Theory, 4, 161– 178. Hauk, O., Johnsrude, I., & Pulvermuller, F. (2004). Somatotopic representation of action words in human motor and premotor cortex. Neuron, 41, 301–307. Helm-Estabrooks, N., & Nicholas, M. (2003). Narrative story cards (6th ed.). Austin, TX: PRO-ED, Inc. Kable, J. W., Lease-Spellmeyer, J., & Chatterjee, A. (2002). Neural substrates of action event knowledge. Journal of Cognitive Neuroscience, 14, 795–805. Kemmerer, D. (2003). Why can you hit someone on the arm, but not break someone on the arm? A neuropsychological investigation of the English body part possessor ascension construction. Journal of Neurolinguistics, 16, 13–36. Kemmerer, D., Castillo, J. G., Talavage, T., Patterson, S., & Wiley, C. (2008). Neuroanatomical distribution of five semantic components of verbs: evidence from fMRI. Brain and Language, 107, 16–43. Kemmerer, D., & Tranel, D. (2000a). Verb retrieval in brain-damaged subjects: 1. Analysis of stimulus, lexical, and conceptual factors. Brain and Language, 73, 347–392.



17

Kemmerer, D., & Tranel, D. (2000b). Verb retrieval in brain-damaged subjects: 2. Analysis of errors. Brain and Language, 73, 393–420. Kertesz, A. (1982). The Western aphasia battery. New York: Grune Stratton. Kim, M., Adingono, M. F., & Revoir, J. S. (2007). Argument structure enhanced verb naming treatment: two case studies. Contemporary Issues in Communication Science and Disorders, 34, 24–36. Kim, M., & Thompson, C. K. (2000). Patterns of comprehension and production of nouns and verbs in agrammatism: Implications for lexical organization. Brain and Language, 74, 1–25. Kim, M., & Thompson, C. K. (2004). Verb deficits in Alzheimer’s disease and agrammatism: implications for lexical organization. Brain and Language, 88, 1–20. Kiran, S. (2007). Complexity in the treatment of naming deficits. American Journal of Speech Language Pathology, 16, 18–29. Kiran, S., & Thompson, C. K. (2003). The role of semantic complexity in treatment of naming deficits. Journal of Speech, Language, and Hearing Research, 46, 608–622. Kleim, J. A., & Jones, T. A. (2008). Principles of experience-dependent neural plasticity: implications for rehabilitation after brain damage. Journal of Speech Language Hearing Research, 51, S225–S239. Levin, B. (1993). English verb classes and alternations: A preliminary investigation. Chicago, IL: University of Chicago Press. Levin, B., & Rappaport Hovav, M. (2005). Argument realization. Cambridge, UK: Cambridge University Press. Links, P., Hurkmans, J., & Bastiaanse, R. (2010). Training verb and sentence production in agrammatic Broca’s aphasia. Aphasiology, 24, 1303–1325. Luzzatti, C., Raggi, R., Zonca, G., Pistarini, C., Contardi, A., & Pinna, G. D. (2002). Verb-noun double dissociation in aphasic lexical impairments: the role of word frequency and imageability. Brain and Language, 81, 432–444. Marshall, J., Pring, T., & Chiat, S. (1998). Verb retrieval and sentence production in aphasia. Brain and Language, 63, 159–183. Masson, M. E. J., Bub, D. N., & Warren, C. M. (2008). Kicking calculators: contribution of embodied cognition to sentence comprehension. Journal of Memory and Language, 59, 256–265. Mätzig, S., Druks, J., Masterson, J., & Vigliocco, G. (2009). Noun-verb differences in picture naming: past studies and new evidence. Cortex, 45, 738–758. McCarthy, R. A., & Warrington, E. K. (1985). Category specificity in an agrammatic patient: the relative impairment of verb retrieval and comprehension. Neuropsychologia, 23, 709–727. McNeil, M. R., Doyle, P. J., Spencer, K., Goda, A. J., Flores, D. K., & Small, S. L. (1998). Effects of training multiple form classes on acquisition, generalization, maintenance of word retrieval in a single subject. Aphasiology, 12, 385–400. McNemar, Q. (1969). Psychological statistics (4th ed.). New York: J. Wiley. Mitchum, C. C., Ritgert, B. A., Sandson, J., & Berndt, R. S. (1990). The use of response analysis in confrontation naming. Aphasiology, 4, 261–280. Pashek, G. (1998). Gestural facilitation of noun and verb retrieval in aphasia: a case study. Brain and Language, 65, 177–180. Pinker, A. (1989). Learnability and cognition: The acquisition of argument structure. Cambridge, MA: MIT Press. Pinker, S. (2007). The stuff of thought. New York: Viking. Pulvermuller, F. (2005). Brain mechanisms linking language and action. Nature, 6, 576–582. Raymer, A. M., Ciampitti, M., Holliway, B., Singletary, F., Blonder, L. X., Ketterson, T., Anderson, S. Lehnen, J., Heilman, K. M., & Rothi, L. J. G. (2007). Semantic-phonologic treatment for noun and verb retrieval impairments in aphasia. Neuropsychological Rehabilitation, 17, 244–270. Raymer, A. M., & Ellsworth, T. A. (2002). Response to contrasting verb treatments: a case study. Aphasiology, 16, 1031–1045. Robey, R. R. (2004). A five-phase model for clinical-outcome research. Journal of Communication Disorders, 37, 401–410. Rose, M., & Sussmilch, G. (2008). The effects of semantic and gesture treatments on verb retrieval and verb use in Broca’s aphasia. Aphasiology, 22(7/8), 691–706. Saygin, A., Wilson, B., Dronkers, N., & Bates, E. (2004). Action comprehension in aphasia: linguistic and nonlinguistic deficits and their lesion correlates. Neuropsychologia, 42, 1788–1804. Schneider, S. L., & Thompson, C. K. (2004). Verb production in agrammatic aphasia: the influence of semantic class and argument structure properties on generalization. Aphasiology, 17, 213–241. Thompson, C. K., Shapiro, L. P., Kiran, S., & Sobecks, J. (2003). The role of syntactic complexity in treatment of sentence deficits in agrammatic aphasia: the complexity account of treatment efficacy (CATE). Journal of Speech, Language & Hearing Research, 46, 591–607. Tyler, L. K., Bright, P., Fletcher, P., & Stamatakis, E. A. (2004). Neural processing of nouns and verbs: the role of inflectional morphology. Neuropsychologia, 42, 512–523.


18


Vigliocco, G., Vinson, D. P., Druks, J., Barber, H., & Cappa, S. F. (in press). Nouns and verbs in the brain: a review of behavioural, electrophysiological, neuropsychological and imaging studies. Neuroscience & Biobehavioral Reviews, doi:10.1016/j.neubiorev.2010.04.007. Wambaugh, J. L., Cameron, R., Kalinyak-Fliszar, M., Nessler, C., & Wright, S. (2004). Retrieval of action names: effects of two cueing treatments in aphasia. 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 Development, 39, 455–466. Wambaugh, J. L., & Ferguson, M. (2007). Application of semantic feature analysis to retrieval of action names in aphasia. Journal of Rehabilitation Research and Development, 44, 381–394. Wambaugh, J. L., Linebaugh, C. W., Doyle, P. J., Martinez, A. L., Kalinyak-Fliszar, M., & Spencer, K. A. (2001). Effects of two cueing treatments on lexical retrieval in aphasic speakers with different levels of deficit. Aphasiology, 15, 933–950. Wu, D., Morganti, A., & Chatterjee, A. (2008). Neural representation of processing path and manner information of a moving event. Neuropsychologia, 46, 704–713. Wunderlich, D. (2006). Towards a structural typology of verb classes. In D. Wunderlich (Ed.), Advances in the theory of the lexicon (pp. 58–166). Berlin: Mouton de Gruyter. Zingeser, L. B., & Berndt, L. S. (1990). Retrieval of verbs and nouns in agrammatism and anomia. Brain and Language, 39, 14–32.

Appendix 1: Pre- and post-treatment narratives elicited from picture description of the WAB Participant 1 Pre-treatment. Vacation study . . . make a picnic . . . son take it for a fly to watch the dog. Another . . . it’s another guy is fishing. Another guy . . . the two person . . . the . . . rises some boat. And the . . . um . . . the weather is beautiful . . . has a sun. The tree. Post-treatment. The couple have a picnic . . . The boy flies, fly the flag. A dog. The . . . another girl is fishing. The boy . . . play the toy. And the other two people play the boat. And the weather is good. They have a picnic . . . family picnic. Participant 2 Pre-treatment. I see reading the book book. Pourin’ the drink . . . kite flying . . . the flag. . . . the tree . . . the car . . . houses . . . the dog . . . um . . . oh my . . . I do/n’t know {laughs} . . . Okay . . . (7 seconds pause) . . . a boat (8 s pause) . . . That’s it . . . Okay. Post-treatment. A tree. A car. A kite. A dog. A boat. A book. A flag. Some trees. A car. A radio. A blanket. A book. Some sage? A dog. A boat. Appendix 2: Verbs used in the study with their lemma frequencies based on CELEX database (Baayen et al., 1993)

Verb class

Stimuli

Frequency per million of verb form

Frequency per million of noun form

Cut Chip Chop Crush Cube

6 19 21 2

15 6 3 9 (Continued)


19

Appendix 2: (Continued) Verb class

Frequency per million of verb form

Frequency per million of noun form

1 27 6 1 10 12 4 12 3 2 9

2 2 2 NA 7 1 4 18 5 12 6.61

Bite Bump Kiss Knock Lick Nudge Pat Pinch Rap Scratch Stroke Tap Tickle Touch

27 11 59 54 11 5 17 9 4 24 19 25 4 110 27

15 5 17 8 1 1 2 4 2 7 25 20 0 57 11.7

Cough Gasp Smile Sneeze Snore Whistle Yawn

12 16 161 3 4 13 8 31

12 5 83 1 1 9 2 16

Stimuli


Dice Grind Hack Perforate Punch Scrape Shred Slice Slit Spear Mean frequency Contact

Mean frequency Non-verbal expression

Mean frequency

Appendix 3: Treatment and generalization verbs used in this study, sorted by participant and treatment phase Participant 1 PHASE 1 Treatment verbs

Contact verbs 1. Bite 2. Knock 3. Lick 4. Nudge

Participant 2

Cut verbs 1. Chip 2. Crush 3. Hack 4. Dice (Continued)

20



Generalization verbs

PHASE 2 Treatment verbs

Generalization verbs

5. Pinch 6. Tickle 7. Touch Non-verbal expression verbs 1. Cough 2. Gasp 3. Smile 4. Sneeze 5. Snore 6. Whistle 7. Yawn Cut verbs 1. Chop 2. Cube 3. Grind 4. Perforate 5. Scrape 6. Slice 7. Spear Contact verbs 1. Bump 2. Kiss 3. Pat 4. Rap 5. Scratch 6. Stroke 7. Tap Cut verbs 1. Chip 2. Crush 3. Hack 4. Dice 5. Punch 6. Shred 7. Slit Non-verbal expression Verbs 1. Cough 2. Gasp 3. Smile 4. Sneeze 5. Snore 6. Whistle 7. Yawn Cut verbs 1. Chop 2. Cube 3. Grind 4. Perforate 5. Scrape 6. Slice 7. Spear Contact verbs 1. Bump 2. Kiss 3. Pat 4. Rap 5. Scratch 6. Stroke 7. Tap

5. Punch 6. Shred 7. Slit Non-verbal expression verbs 1. Cough 2. Gasp 3. Smile 4. Sneeze 5. Snore 6. Whistle 7. Yawn Cut verbs 1. Chop 2. Cube 3. Grind 4. Perforate 5. Scrape 6. Slice 7. Spear Contact verbs 1. Bump 2. Kiss 3. Pat 4. Rap 5. Scratch 6. Stroke 7. Tap N/A