- Email: [email protected]
Thematic role assignment in the posterior parietal cortex: A TMS study
Author’s Accepted Manuscript Thematic role assignment in the posterior parietal cortex: A TMS study Chiara Finocchiaro, Rita Capasso, Luigi Cattaneo, Arianna Zuanazzi, Gabriele Miceli www.elsevier.com/locate/neuropsychologia
PII: DOI: Reference:
S0028-3932(15)30141-X http://dx.doi.org/10.1016/j.neuropsychologia.2015.08.025 NSY5709
To appear in: Neuropsychologia Received date: 16 March 2015 Revised date: 4 August 2015 Accepted date: 24 August 2015 Cite this article as: Chiara Finocchiaro, Rita Capasso, Luigi Cattaneo, Arianna Zuanazzi and Gabriele Miceli, Thematic role assignment in the posterior parietal cortex: A TMS study, Neuropsychologia, http://dx.doi.org/10.1016/j.neuropsychologia.2015.08.025 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Title: Thematic role assignment in the posterior parietal cortex: A TMS study
Authors: Chiara Finocchiaro1, Rita Capasso3,4, Luigi Cattaneo2, Arianna Zuanazzi2, Gabriele Miceli1,2 1
Dipartimento di Psicologia e Scienze Cognitive, Università di Trento, Trento, Italy; 2Center for Mind/Brain Sciences (CIMeC), Trento, Italy; 3Fondazione Santa Lucia IRCCS, Roma, Italy; 4S.C.A. Associati, Roma, Italy
Affiliations institution(s) at which the work was performed: Università degli Studi di Trento Via Matteo del Ben 5/b, 38068 Rovereto (TN), Italy
Corresponding author: Chiara Finocchiaro Dipartimento di Psicologia e Scienze Cognitive Corso Bettini 31, 38068 Rovereto (TN), Italy e-mail: [email protected] Tel:++39 0464-808678; Fax:++39 0464-483554
Short title: Thematic roles in the parietal cortex
Key words: transcranial magnetic stimulation; brain mapping; natural language processing; posterior parietal lobe; thematic role assignment; reanalysis. Abstract: Verbs denote relations between entities acting a role in an event. Thematic roles are essential to the correct use of verbs and involve both semantic and syntactic aspects. 1
We used repetitive Transcranial Magnetic Stimulation (rTMS) to study the involvement of three different left parietal sites in the understanding of thematic roles. In a sentence-topicture matching task, twelve participants were asked to judge whether or not a given picture matched with a written sentence. Pictures represented simple reversible actions, and sentences were in the active or passive diathesis. Whereas both active and passive sentences require the correct encoding of thematic roles, passives also imply thematic reanalysis, as the canonical order of thematic roles is systematically reversed. The experiment was divided in three sessions. In each session a different parietal site (anterior, middle, posterior) was stimulated at 5 Hz in an event-related fashion, time-locked to the presentation of visual stimuli. Results showed increased accuracy for passive sentences following posterior parietal stimulation. The effect appeared to be (a) TMS-related, as no effect was observed in a control, no-TMS experiment with eighteen new participants; (b) independent from semantic processes involved in word-picture association, as no TMS-related effects were observed in a pictureword matching task. We interpret the results as showing that the posterior parietal site is specifically involved in the assignment of thematic roles, in particular when the correct interpretation of a sentence requires reanalysis of temporarily encoded thematic roles, as in passive reversible sentences.
1. Introduction For a long time, a sharp distinction has been drawn between anterior and posterior perisylvian areas as regards their role in sentence comprehension. Traditionally, anterior areas of the left hemisphere, and in particular Brodmann areas 44 and 45, referred to jointly as Broca’s area, have been considered to be responsible for –or at least involved in– virtually all syntactic operations related to verb and sentence processing (e.g., Rogalski et al., 2008; Carreiras et al., 2007; Hagoort, 2005; Friederici, 2002; Grodzinsky, 2000). By contrast, more posterior areas 2
including superior and middle temporal gyri and the inferior parietal lobule have been deemed critical for a vast range of computations that include storage and retrieval of semantic representations as well as verbal working memory processes (e.g., Fiebach et al., 2007; Jonides et al., 1998; Hart and Gordon, 1990; Warrington, 1975). In the past few years, this dichotomy has been questioned by a number of neuroimaging studies showing that parietotemporal regions are involved in the access to verb argument structure, and that activation in these regions varies as a function of argument structure complexity (Thompson et al., 2007; Palti et al., 2007; Shetreet et al., 2007; Bornkessel et al., 2005; Ben-Shachar et al., 2003). Computing argument structure is essential for the correct production and understanding of a verb, and involves both semantic and syntactic aspects. As a pretheoretical cover term and independently of idiosyncratic conceptions of specific orientations, “argument structure” captures the idea that verbs denote relations between entities, and that their role in the event is completely independent of the specific words used (Andrews, 1988; Comrie, 1993; Alsina, 2006). For instance, the verb valuate presupposes someone who does the action (the agent) and someone/something receiving that action (the patient or theme). Therefore, the frame of the verb valuate would be:
(1)
a. subcategorization: valuate V [NP]
b. argument structure: valuate
Thus, the correct comprehension of a verb argument structure entails the correct computation of the syntactic frame in which the verb may appear and the assignment of thematic roles to the participants that may be involved in the specified event (i.e., if they are agent, experiencer, theme, goal).
3
A number of cues may help sentence comprehension and thus the correct processing of thematic roles, even when syntax is impaired. These cues may be semantic or probabilistic. Probabilistic cues mainly concern linear order. For instance, in English the agent often comes as the first element in a sentence. Consider as an example the sentence:
(2)
Linda eats the apple
In (2), even if the argument structure of the verb eat is not correctly retrieved, Linda can still be correctly identified as the agent, simply by appealing to a linear order cue (i.e., first element = agent). This strategy no longer works when the agent is in a different position. A typical case is that of passive sentences. The passive of (2) is:
(3)
The apple is eaten by Linda
The strategy first element = agent would lead to incorrectly interpreting the apple as the agent, which is clearly not the case. However, an interpretation along these lines is normally prevented by what is generally known about things in the world. That is, since we know that normally apples do not eat people, we end up with the correct interpretation of Linda as the agent of the action, even in the case of impaired syntax. Indeed, syntactic abilities are crucial for the identification of thematic roles when both probabilistic cues related to linear order and conceptual-semantic cues tied to encyclopedic knowledge cannot be used or when, if used, they would lead to misinterpreting the sentence. This condition is met by passive reversible sentences:
(4)
The girl is punched by the boy 4
Since in (4) both participants can in principle perform the punching action (this is the reason why sentences like (4) are called reversible), conceptual cues are unavailable. On the other hand, the linear order cue would lead to the wrong interpretation of the girl as the agent of punch. Note that the linear order cue could still work when reversible sentences are in the active diathesis. In:
(5)
The boy punches the girl
the boy is both the first element in the sentence and the agent of punch. In essence, if we were to represent sentence complexity on a scale, we would have:
(6)
Nonreversible Active > Reversible active > Nonreversible passive > Reversible
passive (Turner and Rommetveit, 1967)
There is clear evidence that subjects tend to interpret the first argument as the subject of the sentence and as the agent of the action denoted by the verb (Ferreira, 2003). Differently from active sentences, passive sentences systematically disconfirm the expectation set up by the order cue, thus forcing the listener to reanalyze temporarily encoded thematic roles. Since for reversible passives the reanalysis mechanism cannot be supported by semantic cues, the greater complexity of reversible passives accrues from the need to reanalyze thematic role assignment in the absence of either order or semantic cues. Linguistic theories vary with respect to the involvement of syntactic movement in passive sentences: Whereas movement-based accounts (e.g., Chomsky, 1981) state that identification of a passive structure requires the creation of a trace co-indexed with the grammatical subject, 5
lexical-thematic accounts (e.g., Pollard and Sag, 1994; Bresnan, 2000) do not. However, both accounts share the idea that noncanonical sentences like passive sentences trigger a process of thematic reanalysis, leading to a revision of an initial mapping of thematic roles. Reversible sentences, especially when the diathesis is passive, have been used to isolate the syntactic component in both neuropsychological and neuroimaging studies, in order to verify the functional organization and the neural correlates of the syntactic processes involved in sentence comprehension. A selective deficit in thematic role assignment was documented for the first time by Caramazza and Miceli (1991). Following vascular damage to the upper portion of the parietal lobe, patient EB showed a particularly pure dissociation between impairment of thematic role assignment and spared ability to process the morphological structure of sentences. In particular, both comprehension and production of semantically reversible sentences were impaired, significantly more for passive than active sentences. While Caramazza and Miceli (1991) focused on the contrast between the ability to assign thematic roles and to process morphosyntactic information in reversible sentences, other behavioral studies, involving brain-damaged or healthy participants, compared thematic role assignment accuracy by contrasting reversible sentences and other sentence types. Results showed that both aphasic and non-aphasic participants are slower and less accurate with passive than active reversible sentences in a sentence-picture matching task (Brookshire and Nicholas, 1980). More recently, Ferreira (2003) asked unimpaired participants to identify thematic roles in aurally presented sentences (e.g., Who was the doer?). Participants took longer and were more prone to misinterpret semantically reversible sentences, especially when grammatical structure deviated from the canonical subject-verb-object order (as is the case in reversible passives). Meyer et al. (2012), using the eye tracking-while-listening paradigm, asked aphasic and control participants to perform a sentence-picture matching task. 6
Participants listened to active and passive sentences and were presented with two pictures, the correct target and an alternative in which thematic roles were reversed. While participants selected the picture, their eye movements were monitored. The authors found that control participants showed an initial agent-first processing bias, followed by fixation on the correct picture in the vicinity of the verb in both active and passive sentences. Stated differently, in passive sentences, upon hearing N1 participants immediately looked at the picture in which N1 was the agent. A growing number of neuroimaging studies sheds light on the neurofunctional correlates of thematic role assignment. Some investigations found increased activations to verbs characterized by a more complex thematic structure or associated to more than one thematic grid. For example, Thompson et al. (2010), using a lexical decision task that included oneargument, two-argument, and three-argument verbs, found that a posterior perisylvian network is crucially engaged in processing information related to verb argument structure. Similar results were obtained by den Ouden et al. (2009), who found increased activity in posterior regions for naming transitive vs. intransitive verbs. In a lexical decision task, Meltzer-Asscher et al. (2013) compared simple verbs associated to one thematic grid, to verbs with alternating transitivity, which may be used transitively (with two thematic roles) or intransitively (with one thematic role), showing that alternating transitivity verbs yield significantly greater bilateral activation in the angular and supramarginal gyri, extending to the posterior superior temporal gyri. Other neuroimaging studies investigated thematic role assignment using sentences rather than single words. For instance, Richardson et al. (2010) found greater activation for reversible sentences in the left temporo-parietal boundary when participants silently listened to or read reversible and non-reversible sentences. In a sentencepicture matching task, Mack et al. (2013) found significantly greater activation for passive relative to active sentences in the inferior frontal gyrus bilaterally as well as in left temporo7
occipital regions. In a study of voxel-based lesion-symptom mapping in aphasia, Thothathiri et al. (2012) found a significant association between damage to temporo-parietal cortex and impaired sentence comprehension, even when controlling for phonological working memory. Thompson et al. (2013) used both verb naming and sentence generation tasks to train agrammatic patients in exploiting the argumental properties of verbs. They observed an improvement of verb production, that generalized to untrained verbs and recruited the same regions (posterior perisylvian, parietal and sensory-motor cortices bilaterally) as those associated with complex argument structure in healthy controls. Overall, despite different methods and results, available fMRI data converge with the neuropsychological literature in showing the involvement of left temporo-parietal regions in the assignment of thematic roles in reversible and passive sentences. However, even though fMRI provides neuroanatomical evidence with optimal spatial resolution, it only represents an indirect measure of brain activity. For this reason, in the present study we used Transcranial Magnetic Stimulation (TMS). This technique has an immediate effect on the activity of underlying cortical structures. Therefore, any behavioral modification that follows TMS is likely to be causally linked to changes in the activity of the stimulated brain structures. In a sentence-to-picture matching task, participants were asked to judge whether or not a picture matched with a sentence written below it. Pictures represented simple reversible actions involving humans or animals. In order to avoid the strategy first noun = subject, and force participants to analyze the verb before responding, sentences were presented without the first argument (e.g., baciano le figlie “[they] kiss the daughters”). This was possible because Italian, differently from English, is a pro-drop language (i.e., the subject may be omitted). The two roles in each sentence shared morphosyntactic features such as gender and number. In other words, even if the verb was the first element, its morphological marking may in principle apply to both participants. Mismatching sentences always used the same words as 8
the matching sentences, but the verb did not agree with the subject of the scene as represented in the picture (e.g., bacia “kisses” instead of baciano “[they] kiss”), or the thematic roles of agent and theme were reversed (e.g., baciano le mamme “[they] kiss the mothers” paired to a picture in which the mothers perform the kissing action on the daughters). Both matching and mismatching sentences could be in the active or in the passive diathesis. Thus, in order to perform the task correctly, participants had to identify the entities involved in the represented action and assign their thematic roles. Based on lesion data and on neuroimaging studies on healthy individuals, we predicted that TMS over parietal sites would affect performance in the experimental task. We had no specific expectations regarding which sites in the parietal lobe would yield an effect, because lesion data show the involvement of a large slab of parietal cortex, and neuroimaging studies do not provide univocal indications, especially because the specific issues addressed by each study and, consequently, methods and procedures show a large variability. Therefore, we decided to stimulate three sites along the intraparietal sulcus: If the parietal site stimulated in a given session is crucial for the task, performance after its stimulation should differ from that observed after stimulation of other parietal sites or during a control condition, in which (a different group of) participants performed the task with no TMS. A word-picture matching task was also administered. In this task, participants had to judge whether or not the picture of an object matched the noun written below it. Nouns could be correct or semantically related (e.g., APPLE- orange). This task was chosen because it requires lexical-semantic processes but not thematic role assignment. Performance accuracy in this task should not be affected by TMS, as available evidence from lesion and neuroimaging studies does not assign the parietal lobe a significant role in the lexical-semantic processes required by single-word comprehension.
9
2. Methods 2.1. Participants A total of thirty right-handed (Oldfield, 1971) Italian native speakers took part in the study. Of these, twelve (7 male; mean age: 25.8; range: 19-37) were assigned to the TMS experiment, and eighteen (4 male; mean age: 22.2; age range: 19-36) to the no-TMS (control) experiment. All participants had normal hearing and normal or corrected-to-normal vision. They were not informed of the purpose of the study until the end of the experiment. None had a history of neurological or psychiatric disorders. Written consent was obtained from each participant. Participants in the TMS group were preliminarily screened for any relative or absolute contraindication to TMS. The study was approved by the local ethics committee.
2.2. Materials 2.2.1. Word-picture matching task A set of 40 pictures of objects was selected (Snodgrass and Vanderwart, 1980). Each picture was presented twice, once with its correct name (matching condition), once with a semantically-related noun (mismatching condition). Participants had to judge by pressing two different keys whether or not noun and picture matched. Four blocks were created in which each picture appeared only once and the same number of matching and mismatching trials was presented.
2.2.2. Sentence-picture matching task Thirty-two pictures representing a reversible action were selected (Miceli et al. 2006; Capasso and Miceli, 2001). For an example, see Fig. 1. The thirty-two verbs used in the sentences were all transitive obligatory 2-place verbs (e.g., baciare ‘to kiss’) with the exception of three 10
optional transitives (applaudire ‘to applaud’, fischiare ‘to whistle’, sognare ‘to dream’). As we used reversible sentences, virtually all verbs allow for a reciprocal use (e.g., baciarsi ‘to kiss reciprocally’) but only four verbs may be used as direct reflexives (uccidere/uccidersi ‘to kill/to commit suicide’, vestire/vestirsi ‘to dress/to get dressed’, bagnare/bagnarsi ‘to bathe’, asciugare/asciugarsi ‘to dry’). All the verbs used are listed in Appendix A.
Each picture was presented four times: twice with a matching sentence (one active, one passive), twice with a mismatching sentence (one active, one passive), yielding a total of 128 picture-sentence trials divided in four blocks. Each picture appeared only once per block, each time paired with a different sentence. In each block, there were an equal number of conditions. Mismatching trials could be in the active or in the passive diathesis and included two equally frequent types of mismatch: sentences with a morphological foil or sentences with a thematic foil. The trials associated to Figure 1 are listed as an example in Table 1.
PII: DOI: Reference:
S0028-3932(15)30141-X http://dx.doi.org/10.1016/j.neuropsychologia.2015.08.025 NSY5709
To appear in: Neuropsychologia Received date: 16 March 2015 Revised date: 4 August 2015 Accepted date: 24 August 2015 Cite this article as: Chiara Finocchiaro, Rita Capasso, Luigi Cattaneo, Arianna Zuanazzi and Gabriele Miceli, Thematic role assignment in the posterior parietal cortex: A TMS study, Neuropsychologia, http://dx.doi.org/10.1016/j.neuropsychologia.2015.08.025 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Title: Thematic role assignment in the posterior parietal cortex: A TMS study
Authors: Chiara Finocchiaro1, Rita Capasso3,4, Luigi Cattaneo2, Arianna Zuanazzi2, Gabriele Miceli1,2 1
Dipartimento di Psicologia e Scienze Cognitive, Università di Trento, Trento, Italy; 2Center for Mind/Brain Sciences (CIMeC), Trento, Italy; 3Fondazione Santa Lucia IRCCS, Roma, Italy; 4S.C.A. Associati, Roma, Italy
Affiliations institution(s) at which the work was performed: Università degli Studi di Trento Via Matteo del Ben 5/b, 38068 Rovereto (TN), Italy
Corresponding author: Chiara Finocchiaro Dipartimento di Psicologia e Scienze Cognitive Corso Bettini 31, 38068 Rovereto (TN), Italy e-mail: [email protected] Tel:++39 0464-808678; Fax:++39 0464-483554
Short title: Thematic roles in the parietal cortex
Key words: transcranial magnetic stimulation; brain mapping; natural language processing; posterior parietal lobe; thematic role assignment; reanalysis. Abstract: Verbs denote relations between entities acting a role in an event. Thematic roles are essential to the correct use of verbs and involve both semantic and syntactic aspects. 1
We used repetitive Transcranial Magnetic Stimulation (rTMS) to study the involvement of three different left parietal sites in the understanding of thematic roles. In a sentence-topicture matching task, twelve participants were asked to judge whether or not a given picture matched with a written sentence. Pictures represented simple reversible actions, and sentences were in the active or passive diathesis. Whereas both active and passive sentences require the correct encoding of thematic roles, passives also imply thematic reanalysis, as the canonical order of thematic roles is systematically reversed. The experiment was divided in three sessions. In each session a different parietal site (anterior, middle, posterior) was stimulated at 5 Hz in an event-related fashion, time-locked to the presentation of visual stimuli. Results showed increased accuracy for passive sentences following posterior parietal stimulation. The effect appeared to be (a) TMS-related, as no effect was observed in a control, no-TMS experiment with eighteen new participants; (b) independent from semantic processes involved in word-picture association, as no TMS-related effects were observed in a pictureword matching task. We interpret the results as showing that the posterior parietal site is specifically involved in the assignment of thematic roles, in particular when the correct interpretation of a sentence requires reanalysis of temporarily encoded thematic roles, as in passive reversible sentences.
1. Introduction For a long time, a sharp distinction has been drawn between anterior and posterior perisylvian areas as regards their role in sentence comprehension. Traditionally, anterior areas of the left hemisphere, and in particular Brodmann areas 44 and 45, referred to jointly as Broca’s area, have been considered to be responsible for –or at least involved in– virtually all syntactic operations related to verb and sentence processing (e.g., Rogalski et al., 2008; Carreiras et al., 2007; Hagoort, 2005; Friederici, 2002; Grodzinsky, 2000). By contrast, more posterior areas 2
including superior and middle temporal gyri and the inferior parietal lobule have been deemed critical for a vast range of computations that include storage and retrieval of semantic representations as well as verbal working memory processes (e.g., Fiebach et al., 2007; Jonides et al., 1998; Hart and Gordon, 1990; Warrington, 1975). In the past few years, this dichotomy has been questioned by a number of neuroimaging studies showing that parietotemporal regions are involved in the access to verb argument structure, and that activation in these regions varies as a function of argument structure complexity (Thompson et al., 2007; Palti et al., 2007; Shetreet et al., 2007; Bornkessel et al., 2005; Ben-Shachar et al., 2003). Computing argument structure is essential for the correct production and understanding of a verb, and involves both semantic and syntactic aspects. As a pretheoretical cover term and independently of idiosyncratic conceptions of specific orientations, “argument structure” captures the idea that verbs denote relations between entities, and that their role in the event is completely independent of the specific words used (Andrews, 1988; Comrie, 1993; Alsina, 2006). For instance, the verb valuate presupposes someone who does the action (the agent) and someone/something receiving that action (the patient or theme). Therefore, the frame of the verb valuate would be:
(1)
a. subcategorization: valuate V [NP]
b. argument structure: valuate
Thus, the correct comprehension of a verb argument structure entails the correct computation of the syntactic frame in which the verb may appear and the assignment of thematic roles to the participants that may be involved in the specified event (i.e., if they are agent, experiencer, theme, goal).
3
A number of cues may help sentence comprehension and thus the correct processing of thematic roles, even when syntax is impaired. These cues may be semantic or probabilistic. Probabilistic cues mainly concern linear order. For instance, in English the agent often comes as the first element in a sentence. Consider as an example the sentence:
(2)
Linda eats the apple
In (2), even if the argument structure of the verb eat is not correctly retrieved, Linda can still be correctly identified as the agent, simply by appealing to a linear order cue (i.e., first element = agent). This strategy no longer works when the agent is in a different position. A typical case is that of passive sentences. The passive of (2) is:
(3)
The apple is eaten by Linda
The strategy first element = agent would lead to incorrectly interpreting the apple as the agent, which is clearly not the case. However, an interpretation along these lines is normally prevented by what is generally known about things in the world. That is, since we know that normally apples do not eat people, we end up with the correct interpretation of Linda as the agent of the action, even in the case of impaired syntax. Indeed, syntactic abilities are crucial for the identification of thematic roles when both probabilistic cues related to linear order and conceptual-semantic cues tied to encyclopedic knowledge cannot be used or when, if used, they would lead to misinterpreting the sentence. This condition is met by passive reversible sentences:
(4)
The girl is punched by the boy 4
Since in (4) both participants can in principle perform the punching action (this is the reason why sentences like (4) are called reversible), conceptual cues are unavailable. On the other hand, the linear order cue would lead to the wrong interpretation of the girl as the agent of punch. Note that the linear order cue could still work when reversible sentences are in the active diathesis. In:
(5)
The boy punches the girl
the boy is both the first element in the sentence and the agent of punch. In essence, if we were to represent sentence complexity on a scale, we would have:
(6)
Nonreversible Active > Reversible active > Nonreversible passive > Reversible
passive (Turner and Rommetveit, 1967)
There is clear evidence that subjects tend to interpret the first argument as the subject of the sentence and as the agent of the action denoted by the verb (Ferreira, 2003). Differently from active sentences, passive sentences systematically disconfirm the expectation set up by the order cue, thus forcing the listener to reanalyze temporarily encoded thematic roles. Since for reversible passives the reanalysis mechanism cannot be supported by semantic cues, the greater complexity of reversible passives accrues from the need to reanalyze thematic role assignment in the absence of either order or semantic cues. Linguistic theories vary with respect to the involvement of syntactic movement in passive sentences: Whereas movement-based accounts (e.g., Chomsky, 1981) state that identification of a passive structure requires the creation of a trace co-indexed with the grammatical subject, 5
lexical-thematic accounts (e.g., Pollard and Sag, 1994; Bresnan, 2000) do not. However, both accounts share the idea that noncanonical sentences like passive sentences trigger a process of thematic reanalysis, leading to a revision of an initial mapping of thematic roles. Reversible sentences, especially when the diathesis is passive, have been used to isolate the syntactic component in both neuropsychological and neuroimaging studies, in order to verify the functional organization and the neural correlates of the syntactic processes involved in sentence comprehension. A selective deficit in thematic role assignment was documented for the first time by Caramazza and Miceli (1991). Following vascular damage to the upper portion of the parietal lobe, patient EB showed a particularly pure dissociation between impairment of thematic role assignment and spared ability to process the morphological structure of sentences. In particular, both comprehension and production of semantically reversible sentences were impaired, significantly more for passive than active sentences. While Caramazza and Miceli (1991) focused on the contrast between the ability to assign thematic roles and to process morphosyntactic information in reversible sentences, other behavioral studies, involving brain-damaged or healthy participants, compared thematic role assignment accuracy by contrasting reversible sentences and other sentence types. Results showed that both aphasic and non-aphasic participants are slower and less accurate with passive than active reversible sentences in a sentence-picture matching task (Brookshire and Nicholas, 1980). More recently, Ferreira (2003) asked unimpaired participants to identify thematic roles in aurally presented sentences (e.g., Who was the doer?). Participants took longer and were more prone to misinterpret semantically reversible sentences, especially when grammatical structure deviated from the canonical subject-verb-object order (as is the case in reversible passives). Meyer et al. (2012), using the eye tracking-while-listening paradigm, asked aphasic and control participants to perform a sentence-picture matching task. 6
Participants listened to active and passive sentences and were presented with two pictures, the correct target and an alternative in which thematic roles were reversed. While participants selected the picture, their eye movements were monitored. The authors found that control participants showed an initial agent-first processing bias, followed by fixation on the correct picture in the vicinity of the verb in both active and passive sentences. Stated differently, in passive sentences, upon hearing N1 participants immediately looked at the picture in which N1 was the agent. A growing number of neuroimaging studies sheds light on the neurofunctional correlates of thematic role assignment. Some investigations found increased activations to verbs characterized by a more complex thematic structure or associated to more than one thematic grid. For example, Thompson et al. (2010), using a lexical decision task that included oneargument, two-argument, and three-argument verbs, found that a posterior perisylvian network is crucially engaged in processing information related to verb argument structure. Similar results were obtained by den Ouden et al. (2009), who found increased activity in posterior regions for naming transitive vs. intransitive verbs. In a lexical decision task, Meltzer-Asscher et al. (2013) compared simple verbs associated to one thematic grid, to verbs with alternating transitivity, which may be used transitively (with two thematic roles) or intransitively (with one thematic role), showing that alternating transitivity verbs yield significantly greater bilateral activation in the angular and supramarginal gyri, extending to the posterior superior temporal gyri. Other neuroimaging studies investigated thematic role assignment using sentences rather than single words. For instance, Richardson et al. (2010) found greater activation for reversible sentences in the left temporo-parietal boundary when participants silently listened to or read reversible and non-reversible sentences. In a sentencepicture matching task, Mack et al. (2013) found significantly greater activation for passive relative to active sentences in the inferior frontal gyrus bilaterally as well as in left temporo7
occipital regions. In a study of voxel-based lesion-symptom mapping in aphasia, Thothathiri et al. (2012) found a significant association between damage to temporo-parietal cortex and impaired sentence comprehension, even when controlling for phonological working memory. Thompson et al. (2013) used both verb naming and sentence generation tasks to train agrammatic patients in exploiting the argumental properties of verbs. They observed an improvement of verb production, that generalized to untrained verbs and recruited the same regions (posterior perisylvian, parietal and sensory-motor cortices bilaterally) as those associated with complex argument structure in healthy controls. Overall, despite different methods and results, available fMRI data converge with the neuropsychological literature in showing the involvement of left temporo-parietal regions in the assignment of thematic roles in reversible and passive sentences. However, even though fMRI provides neuroanatomical evidence with optimal spatial resolution, it only represents an indirect measure of brain activity. For this reason, in the present study we used Transcranial Magnetic Stimulation (TMS). This technique has an immediate effect on the activity of underlying cortical structures. Therefore, any behavioral modification that follows TMS is likely to be causally linked to changes in the activity of the stimulated brain structures. In a sentence-to-picture matching task, participants were asked to judge whether or not a picture matched with a sentence written below it. Pictures represented simple reversible actions involving humans or animals. In order to avoid the strategy first noun = subject, and force participants to analyze the verb before responding, sentences were presented without the first argument (e.g., baciano le figlie “[they] kiss the daughters”). This was possible because Italian, differently from English, is a pro-drop language (i.e., the subject may be omitted). The two roles in each sentence shared morphosyntactic features such as gender and number. In other words, even if the verb was the first element, its morphological marking may in principle apply to both participants. Mismatching sentences always used the same words as 8
the matching sentences, but the verb did not agree with the subject of the scene as represented in the picture (e.g., bacia “kisses” instead of baciano “[they] kiss”), or the thematic roles of agent and theme were reversed (e.g., baciano le mamme “[they] kiss the mothers” paired to a picture in which the mothers perform the kissing action on the daughters). Both matching and mismatching sentences could be in the active or in the passive diathesis. Thus, in order to perform the task correctly, participants had to identify the entities involved in the represented action and assign their thematic roles. Based on lesion data and on neuroimaging studies on healthy individuals, we predicted that TMS over parietal sites would affect performance in the experimental task. We had no specific expectations regarding which sites in the parietal lobe would yield an effect, because lesion data show the involvement of a large slab of parietal cortex, and neuroimaging studies do not provide univocal indications, especially because the specific issues addressed by each study and, consequently, methods and procedures show a large variability. Therefore, we decided to stimulate three sites along the intraparietal sulcus: If the parietal site stimulated in a given session is crucial for the task, performance after its stimulation should differ from that observed after stimulation of other parietal sites or during a control condition, in which (a different group of) participants performed the task with no TMS. A word-picture matching task was also administered. In this task, participants had to judge whether or not the picture of an object matched the noun written below it. Nouns could be correct or semantically related (e.g., APPLE- orange). This task was chosen because it requires lexical-semantic processes but not thematic role assignment. Performance accuracy in this task should not be affected by TMS, as available evidence from lesion and neuroimaging studies does not assign the parietal lobe a significant role in the lexical-semantic processes required by single-word comprehension.
9
2. Methods 2.1. Participants A total of thirty right-handed (Oldfield, 1971) Italian native speakers took part in the study. Of these, twelve (7 male; mean age: 25.8; range: 19-37) were assigned to the TMS experiment, and eighteen (4 male; mean age: 22.2; age range: 19-36) to the no-TMS (control) experiment. All participants had normal hearing and normal or corrected-to-normal vision. They were not informed of the purpose of the study until the end of the experiment. None had a history of neurological or psychiatric disorders. Written consent was obtained from each participant. Participants in the TMS group were preliminarily screened for any relative or absolute contraindication to TMS. The study was approved by the local ethics committee.
2.2. Materials 2.2.1. Word-picture matching task A set of 40 pictures of objects was selected (Snodgrass and Vanderwart, 1980). Each picture was presented twice, once with its correct name (matching condition), once with a semantically-related noun (mismatching condition). Participants had to judge by pressing two different keys whether or not noun and picture matched. Four blocks were created in which each picture appeared only once and the same number of matching and mismatching trials was presented.
2.2.2. Sentence-picture matching task Thirty-two pictures representing a reversible action were selected (Miceli et al. 2006; Capasso and Miceli, 2001). For an example, see Fig. 1. The thirty-two verbs used in the sentences were all transitive obligatory 2-place verbs (e.g., baciare ‘to kiss’) with the exception of three 10
optional transitives (applaudire ‘to applaud’, fischiare ‘to whistle’, sognare ‘to dream’). As we used reversible sentences, virtually all verbs allow for a reciprocal use (e.g., baciarsi ‘to kiss reciprocally’) but only four verbs may be used as direct reflexives (uccidere/uccidersi ‘to kill/to commit suicide’, vestire/vestirsi ‘to dress/to get dressed’, bagnare/bagnarsi ‘to bathe’, asciugare/asciugarsi ‘to dry’). All the verbs used are listed in Appendix A.
Each picture was presented four times: twice with a matching sentence (one active, one passive), twice with a mismatching sentence (one active, one passive), yielding a total of 128 picture-sentence trials divided in four blocks. Each picture appeared only once per block, each time paired with a different sentence. In each block, there were an equal number of conditions. Mismatching trials could be in the active or in the passive diathesis and included two equally frequent types of mismatch: sentences with a morphological foil or sentences with a thematic foil. The trials associated to Figure 1 are listed as an example in Table 1.