- Email: [email protected]
When space merges into language
Neuropsychologia 44 (2006) 556–565
When space merges into language M. Cristina Rinaldi a,∗ , Luigi Pizzamiglio a,b a
Centro Ricerche di Neuropsicologia, Fondazione I.R.C.C.S. Santa Lucia, Via Ardeatina 306, 00179 Rome, Italy b Dipartimento di Psicologia Universit` a “La Sapienza”, Rome, Italy Received 10 January 2005; received in revised form 25 May 2005; accepted 7 July 2005 Available online 30 August 2005
Abstract We present data from right brain-damaged patients, with and without spatial heminattention, which show the influence of hemispatial deficits on spoken language processing. We explored the findings of a previous study, which used an emphatic stress detection task and suggested spatial transcoding of a spoken active sentence in a ‘language line’. This transcoding was impaired in its initial portion (the subjectword) when the neglect syndrome was present. By expanding the original methodology, the present study provides a deeper understanding of the level of spoken language processing involved in the heminattentional bias. To ascertain the role played by syntactic structure, active and passive sentences were compared. Sentences comprised of musical notes and of a sequence of unrelated nouns were also compared to determine whether the bias was manifest with any sequence of events (not only linguistic ones) deployed over time, and with a sequence of linguistic events not embedded in a structured syntactic frame. Results showed that heminattention exerted an influence only when a syntactically structured linguistic input (=sentence with agent of action, action and recipient of action) was processed, and that it did not interfere when a sequence of non-linguistic sounds or unrelated words was presented. Furthermore, when passing from active to passive sentences, the heminattentional bias was inverted, suggesting that heminattention primarily involves the logical subject of the sentence, which has an inverted position in passive sentences. These results strongly suggest that heminattention acts on the spatial transcoding of the deep structure of spoken language. © 2005 Elsevier Ltd. All rights reserved. Keywords: Space; Language; Neglect
1. Introduction The spatial dimension has been so relevant in the evolution of the nervous system that sensory and motor processing cannot take place unless their spatial co-ordinates are coded. Coslett (1999) claimed that the influence of space also holds for higher-order more abstract functions (including language), which are less obviously related to space. Coslett’s (Coslett, Schwartz, Goldberg, Haas, & Perkins, 1993) considerations originated from clinical observations carried out on an aphasic patient, J.F., whose anomia became more or less severe depending on the hemifield stimuli were presented to. According to Coslett (1999), the importance of
∗
Corresponding author. E-mail address: [email protected] (M.C. Rinaldi).
0028-3932/$ – see front matter © 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2005.07.007
spatial information for every action we perform is so strong that each stimulus we perceive is automatically marked with reference to its co-ordinates in the egocentric space, even if spatial information does not seem relevant for the task at hand (“Spatial Registration Hypothesis”). Following this reasoning, any stimulus presented in the contralesional field (for patient J.F., the right hemispace) should activate the left hemisphere attentional system for the impaired right hemispace. This explains J.F.’s worse performance with stimuli in the contralesional field, while his performance in the left hemispace was supported by his intact right hemisphere (Coslett et al., 1993). This interpretation stresses the importance of the extrapersonal, egocentric space linguistic stimuli arise from (Coslett, 1999). Other authors hypothesize that any event expressed through language, whether spoken or written, automatically activates an “internal” spatial representation.
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
The latter explanation is postulated by Chatterjee, Maher, Gonzalez Rothi, and Heilman (1995b). They described an agrammatic patient, W.H., who assigned the thematic roles of agent and recipient of a given action on the basis of a spatial or temporal strategy. For example, in describing a picture, this patient was asked to report who did something to whom (agent and recipient of action). He tended to consider the figure located on the left as the agent of action (spatial strategy), or to match the first noun he heard with the figure located on the left (temporal strategy) (Chatterjee et al., 1995b). According to Chatterjee et al. (1995b), conceptual knowledge of thematic roles is actually independent of language, and is matched with linguistic content only subsequently, by means of the grammatical and syntactic algorithms specific to one’s own language. Due to a brain lesion, these linguistic algorithms were impaired in their agrammatic patient. Therefore, he based his thematic role detection exclusively on the automatic spatial representation of events, which proceeds naturally from left (agent of action) to right (recipient of action), with the verb being the vector defining the thematic roles and supporting the whole event frame. To further support this idea, Chatterjee, Maher, and Heilman (1995a) and Chatterjee, Southwood, and Basilico (1999) asked non brain-damaged subjects to draw an assigned thematic role. Their reasoning was that if there is a spatial representation underlying a spoken sentence, and an “implicit propensity” to conceive events as traversing space from left to right, with the agent of the action on the left of this representational space, the subjects should draw a given action proceeding from left to right. They showed that the majority of subjects located the agent closer to the left margin of the page, and the recipient of the action closer to the right margin of the page (Chatterjee et al., 1995a, 1999). For Chatterjee and colleagues, the presence of a spatial representation underlying spoken language reflects the primitive way the mind conceptualizes and represents events, a way that has been successively obscured—but not deleted— during the course of development by the evolution of language. Analogously, passing from phylogenesis to ontogenesis, Mandler (1996) observed that in the young child emerging linguistic concepts are mapped on a pre-existing conceptual system, which has space as a fundamental constituent dimension. If space and language interact at some still unidentified level, forming the spatio-linguistic representation outlined above, it is theoretically possible to observe representational abnormalities as a consequence of (1) a brain lesion modifying spatial processing and (2) a brain lesion modifying linguistic processing in two opposite, but converging, lines of action. Some evidence for the latter is provided by the two aphasic patients described above, i.e., by Coslett et al. (1993) and by Chatterjee et al. (1995b). Instead, some evidence for the former is provided by two patients with hemispatial neglect, one described by Baxter and Warrington (1983) and the other by Caramazza and Hillis (1990).
557
Patient O.R.F., described by Baxter and Warrington (1983), underwent two consecutive vascular accidents within a short period of time. The first parietal lesion caused conduction aphasia (the patient was left-handed), phonological dyslexia, and mild neglect dyslexia, without other signs of spatial neglect; the second vascular event, again striking the right parietal region, produced an additional impairment of word-spelling abilities that was not present before. The word–spelling deficit was observed only for letters at the beginning of words; for example, if the examiner spelled the word “b-e-g-g-a-r” aurally letter by letter, the patient reported “vinegar” (Baxter, & Warrington, 1983). Interestingly, this bias was manifested when spelling proceeded either from the beginning to the end of the word, or from the end to the beginning of the word. Baxter and Warrington (1983) hypothesized that the patient was representing the stimulus on an inner screen, and the symptoms resulted from an interaction between representational neglect and neglect dyslexia. The authors affirmed that the patient reported “reading” the aurally spelled words on an “inner screen”. Instead, patient N.J., described by Caramazza and Hillis (1990), exhibited severe neglect affecting the right half of space after a lesion to the left hemisphere; he had no language deficits. N.J. was submitted to several reading and spelling tests: horizontal reading, vertical reading, mirrorreverted reading, delayed copy, aural spelling, recognition of aural spelling, written spelling and backwards written spelling. When reading, the patient made errors exclusively at the end of words, irrespective of the form of the input (vertical, horizontal or mirror reversed). Also, on the spelling tasks errors appeared only at the right end of a word (N.J. reported “exceed” after the examiner spelled “e-x-c-e-s-s“), irrespective—also for spelling—of the form or order of the input (written and aural spelling, forward and backward spelling). According to Caramazza and Hillis (1990), reading and spelling both involve a common pre-lexical level of representation, which they call “grapheme description”. At this level of representation, the abstract identity of the single letters compounding the word is computed, regardless of the form/orientation of the input; therefore, damage to this format-independent level would impair both reading and spelling in a similar way and in conformity with the heminattentional deficit. Since the spelling tasks were performed aurally, the behavior of both patients strongly suggests a spatial transcoding of spoken language which, given its spatial format, undergoes the same bias observed in neglected patients for visuo-spatial stimuli in the contralesional hemispace. The clinical reports of these two patients document the influence of hemineglect on spoken language processing, but with reference to a very peculiar type of spoken language: spelling. Although spelling is performed aurally, in many respects it is actually very close to written language. This is probably why Baxter and Warrington defined the deficit described in their patient as a rare type of “dysgraphic
558
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
syndrome”, highlighting with the term “dysgraphic” the affinity of spelling and written language. With regard to the effects of hemineglect on genuine spoken language, a neglect patient -J.L.- was reported by Barbut and Gazzaniga (1987). As a consequence of a right parietal lesion, in addition to errors affecting the most leftward letters of words in reading, writing and spelling, J.L. also produced errors in spontaneous speech. For example, he often omitted the beginning of a word, saying “bulance” instead of “ambulance”, or “portant” instead of “important”. Most of his errors were omissions, but the examiner also observed that the patient often made substitutions, forming semantically unrelated words. According to Barbut and Gazzaniga (1987), the pattern of errors exhibited by their patient was consistent with the hypothesis that some aspects of language processing involve some spatial mechanisms responsible for the representation of what they call “conceptual space”. Recently, a decade after the literature reported above, a paper by Rinaldi, Marangolo, and Pizzamiglio (2003) added another piece of evidence to this field of investigation. The authors described a different type of hemineglect effect on the processing of spoken language than that previously observed. When asked to compare two spoken sentences (subject/verb/object) with regard to their emphatic stress, patients with right hemisphere lesions and spatial neglect made significantly more errors when the stress was placed at the beginning of the sentence (on the subject-word). The bias exhibited by neglected patients towards the initial word of the sentence suggests a transcoding of spoken language into a spatial format. The auditory trace left by the spoken sentence would be automatically transformed into an analog with an intrinsic spatial component. This analog would then be affected by the presence of hemineglect in the same way this deficit affects spatial information processing, which is on the contralesional side. In other terms, it appears that the word at the beginning of the sentence falls on the “left” side of the spatial transcoding of the sentence and is, therefore, neglected similar to the way left visuo-spatial information is neglected by these patients in the external space. The present investigation aims at expanding the original methodology and qualifying the nature of the interaction between the spatial and linguistic mechanisms underlying the observed effect. As in the previous experiment, the first question concerns the use of a unique format, i.e., active sentence structure. The replication or the change in the results using active and passive sentences will contribute to understanding whether the effect is related to the position of the stress in the auditory string of words or, instead, to its syntactic structure. The second question is whether the comparison of the stress location in the two sentences produces a bias in hemineglect patients only when linguistic strings are used, or whether, instead, it takes place also when two non-linguistic strings of acoustic events are compared. In other words, if the bias in hemineglect patients can also be observed when they compare the stress embedded in two musical sentences,
the effect can be attributed exclusively to the need to locate a target stimulus in a sequence of events spaced over time. Third, do the stimuli need to be meaningful sentences or can the same effect of neglect be observed with a sequence of non-related words? Does neglect affect the representation of the event expressed in the spoken sentence, where a verb supports the event frame and defines the thematic roles, or does it affect any sequence of three nouns, independently of the presence of a verb?
2. Methods 2.1. Subjects This study was carried out on 26 right brain-damaged subjects admitted to the Santa Lucia Foundation I.R.C.C.S. in Rome for rehabilitation (mean age 61.7 years, S.D. 11.37; education 9.7 years, S.D. 4.19) (see Table 1 for individual patients’ data). For each patient, unilaterality of the brain lesion was documented by a CT or an MR scan. Each patient admitted to the hospital underwent a standard neuropsychological examination, which included a battery for the diagnosis of hemispatial neglect for personal and peripersonal space. The following tests comprised the battery: Cancellation Test (Albert, 1973), the Letter Cancellation Test (Diller et al., 1974), the Sentence Reading Test (Pizzamiglio et al., 1992), the Wundt–Jastrow Area Illusion Test (Massironi, Antonucci, Pizzamiglio, Vitale & Zoccolotti, 1988), and the Personal Neglect Test (Zoccolotti, Antonucci, & Judica, 1992). Patients who failed on at least three of the five tests for heminattention were classified as having hemispatial neglect. According to this criterion, patients were divided into two sub-groups, 12 patients with hemineglect (N+ Group: mean age 62 years, S.D. 8.33; mean education years 10, S.D. 4.31) and 14 patients without hemineglect (N− Group: mean age 62 years, S.D. 13.44; mean education years 10, S.D. 4.07). The neuropsychological examination also included an evaluation of phonological and visuo-spatial short-term memory (Wais Digit Span Test, Orsini et al., 1987; Corsi Span Test, Orsini et al., 1987). Phonological short-term memory was intact in all subjects examined, whereas visuo-spatial short-term memory, unimpaired in patients without hemineglect, was impaired in 4 of the 11 patients with neglect (since neglect was in part responsible for the fact that these patients scored below the cut-off). For the experimental condition described below as “Sequence of Three Nouns”, a different group of right braindamaged patients was tested, since this condition was added later. Respectively, 11 N− and 13 N+ patients were tested (N− group: mean age 63.6 years, S.D. 12.69; mean education years 10.53, S.D. 3.86; N+ group: mean age 60.3 years, S.D. 13.8; mean education years 10.61, S.D. 4.09). Patients with hearing deficits not previously acknowledged but manifested during the experimental sessions were
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
559
Table 1 Individual data of N− and N+ patients for sex, years of education, time post onset and locus of lesion Subjects (N = 26)
Age
Sex
Education (years)
Time post-onset (months)
Lesion site
N− group 1 2 3 4 5 6 7 8 9 10 11 12 13 14
74 58 52 62 48 57 52 39 57 52 76 79 71 82
F M M M M M F M M M M F M M
12 10 5 17 13 11 8 8 17 13 8 5 5 5
2 3 8 1 30 2 360 4.5 6 7 3 6 4 1.5
Right cortico-subcortical fronto-parietal lesion Right internal capsula + lateral thalamic lesion Right extern caps + nucl caud + white matt periventric parietal lesion Right temporo-parietal lesion Right subcortico- temporal lesion Right occipital lesion Right fronto-parietal lesion Right cortico-subcortical fronto-temporal + internal capsula Right parietal lesion Right thalamic lesion Right cortico-subcortico temporal lesion Right semioval center + frontal lobe lesion Right pons lesion Right temporal lesion
N+ group 1 2 3 4 5 6 7 8 9 10 11 12
60 62 71 43 64 54 59 75 53 68 63 75
M F F M M M M F F M M M
7 3 11 8 10 7 13 5 16 13 17 5
2 6 3.5 5 7 2 2 2 1 4.5 3 1.5
excluded from the study. The research was approved by the Ethical Committee and all patients gave their written consent. 2.2. Stimulus material Patients were requested to judge two spoken sentences emitted from a tape recorder in four different experimental conditions: (1) Active versus passive linguistic condition. (2) Musical condition. (3) Sequence of unrelated nouns condition. 2.2.1. Active–passive linguistic condition Stimuli consisted of spoken sentences recorded by a tape recorder; all sentences were in Italian and in the active or passive form, and only high frequency and highly familiar words were used. Each sentence was comprised of three elements: a subject, a verb and an object (e.g., “The boy reads the book”). The following description refers to the active sentences and it will be replicated for the passive sentences. Each item consisted of a pair of linguistically identical sentences, and with emphatic stress placed on the same or on a different word in each of the two sentences. The two sentences of the pair were separated by a two-second time interval. The emphatic
Right parietal lesion Right fronto-temporo-parietal lesion Right cortico-subcortical fronto-temporal lesion Right internal capsula lesion Right frontal lesion Right temporo-parietal lesion Right cortico-subcortical temporo-fronto-parietal lesion Right temporo-fronto-parietal lesion Right temporo-fronto-parietal lesion Right insular + temporo-parietal lesion Right parieto-temporal lesion Right insula + frontal lesion
stress could be placed either on the subject-, the verb- or on the object-word. Following is an example of stress placed on the same elements of a sentence pair: “The boy reads the book”/“The boy reads the book”; here, instead, are examples of stress placed on different elements of a sentence pair: “The wolf attacks the sheep”/”The wolf attacks the sheep”; “The thunderbolt struck the house”/“The thunderbolt struck the house”. The patients were asked to listen to each sentence pair and to judge whether the two sentences were stressed in the same way or not, or, more simply, if their intonation “sounded the same or not” (answer: “yes”/“no”). Each pair of sentences was preceded by an acoustic warning signal, and no time-limit for the answer was set by the examiner. Two separate sets of active and passive sentences (comprising two practice trials + 48 test trials each) were included in the test. For each experimental condition, a different set of 48 + 2 stimuli was used; however, in each condition the whole set of 48 stimuli was balanced according to the same schema, with half of the sentences requiring a “yes” answer and half a “no” answer. Most importantly, the whole set of stimuli was divided into two sub-groups according to the position of the emphatic stress: Type I, including all sentence pairs with the empathic stress on the subject word in at least one of the two sentences of the pair (e.g., “The boy reads the book” / “The
560
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Fig. 1. Schematic representation of item types and their separation into Type I and Type II.
boy reads the book”); and Type II, including all sentence pairs in which the emphatic stress was never on the subject word, but rather on the verb or the object word (e.g., “The teacher writes a letter” / “The teacher writes a letter”) (see Fig. 1). In each set of 48 stimuli, half of the sentence pairs were Type I and the other half, Type II. The set of 48 passive sentences had exactly the same characteristics as the active set. Here, too, there were 24 “same”sentence pairs with the stress located on the same element (e.g., “The ball is hit by the boy” / “The ball is hit by the boy”), and 24 “different” sentence pairs, with the stress located on different elements
(e.g., “The window is closed by the man” / “The window is closed by the man”). In the passive sentences set, the position of emphatic stress, the balancing and the separation of stimuli into Type I and Type II sub-groups were exactly the same as described for the active sentences. The two blocks of active and passive sentences were always presented in the active-first, passive-second sequence. 2.2.2. Musical condition In this condition, sentences formed by a sequence of musical notes instead of words were used. We substituted linguistic with musical stimulus material to ascertain whether
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
the bias of neglect on the spatial representation underlying spoken language affects (1) not language per se, but rather any sequence of elements deployed over time regardless of their linguistic nature, and (2) the so-called suprasegmental features of language, which include sentence stress as a component of linguistic prosody (in a musical sentence, the structure is analogous to the melodic contour of language). A professional musician played notes on a piano and they were recorded on a tape-recorder. We did not make the musical sentences by assembling the notes at random, which, according to the musician, would have been the same as meaningless linguistic babbling, but instead by taking fragments of real pieces of music which had an underlying structure with a meaning. Specifically, sentences consisted of one-voice thematic fragments extracted from a thematic catalogue of J.S. Bach (1685–1750). To form the musical sentences, only fragments with a ternary structure were chosen, so that the structure would be analogous to the threeelement structure of the linguistic sentences; therefore, also in the musical sentences an initial, a central and a final part of the sentence was detectable. Emphatic stress could be placed on each of these three components since, in this case, the stress was made by means of a marked increase in the loudness of the notes played on the keyboard. All sentences had been previously administered to a group of non brain-damaged control subjects to verify that the variations in emphasis could be detected, as foreseen by the research protocol. Here, too, sentences were presented in pairs (48 + 2). The notes comprising the two sentences of the pair were the same and the only variable element was the position of emphatic stress. So, apart from the nature of the stimuli, i.e., musical rather than verbal, the task was exactly the same as the other conditions described: listening to two “sentences” and saying whether they sounded the same or not. 2.2.3. Sequence of three nouns condition In this condition, sentences comprised of a sequence of three nouns were used instead of the canonical noun + verb + noun sentences. The verb was purposely excluded in order to ascertain its relevance—as claimed by linguists like Jackendoff and Gruber (quoted in Chatterjee et al., 1999)—in defining thematic roles and in supporting the whole spatial representation of linguistic events. In other words, if the verb is the element responsible for the spatial representation underlying spoken language, the effect of neglect on spoken language should not occur when the verb is not present. Nouns used were high frequency bi- and tri-syllabic words taken from the frequency lexicon of the Italian language “Barcellona Corpus” (Istituto di Linguistica Computazionale del CNR di Pisa, 1989). For this condition, 48 + 2 sentence pairs were created as well. Each sentence comprised three nouns in sequence, i.e. “dog–apple–pen”, repeated twice with the same or with a different emphatic stress (the same emphatic stress balancing used for the canonical noun + verb + noun sentences).
561
Fig. 2. Comparison of the means and S.D.s of correct answers of N+ patients in the active linguistic vs. passive linguistic condition. Score ranges are shown under each histogram.
3. Results 3.1. Active–passive linguistic condition An ANOVA 2 × 2 × 2 was computed to compare the performance of the two groups of patients on the two sentence types, active and passive; factors used were “Group” (N−/N+ patients), “Sentence Type” (Type I/Type II) and “Condition” (active/passive). “Group” and “Condition” showed no main effects (p = 0.20 and p = 0.53); however, “Sentence Type” was significant (Type I: mean 19.16, S.D. 3.54; Type II: mean 18.34, S.D. 3.54) (F1,24 = 10.52, p = 0.003). The “Group” × “Sentence Type” interaction was not significant (p = 0.16), whereas the “Group” × “Condition” (interaction (p = 0.02) was significant; the “Condition” × “Sentence Type” interaction was also significant (p = 0.000); finally, the three-way “Group” × “Sentence Type” × “Condition” interaction was significant (p = 0.02). A Duncan post-hoc analysis revealed that: (a) the processing of Type I sentences by N+ patients differed significantly in the active versus the passive linguistic condition (p = 0.0001); (b) on the contrary, the processing of Type I sentences by N− patients did not differ in the two linguistic conditions (p = 0.24); N+ patients’ processing of Type II sentences also differed in the active versus the passive linguistic condition (p = 0.04). Therefore, as also shown in Figs. 2 and 3, N+ patients’ differential processing of Type I sentences distinguished the performance of the two groups of patients in active versus passive sentences. Specifically, in active sentences N+ patients made more errors when the stress was at the beginning of the sentence (Type I), and in passive sentences they made most of their errors when the stress was at the center or the end of the sentence (Type II).
Fig. 3. Comparison of the means and S.D.s of correct answers of N− patients in the active linguistic vs. passive linguistic condition. Score ranges are shown under each histogram.
562
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Fig. 4. Means and S.D.s of correct answers of patients with and without neglect in the two sentence types of the musical condition. Score ranges are shown under each histogram.
3.2. Musical condition The means of the correct answers obtained by N+ and N− patients in the musical condition are reported in Fig. 4. As shown by the graph, this condition proved to be quite difficult for both groups of patients compared to the other experimental conditions employed in this study. As before, the performance of the two groups of patients on the two types of musical sentences was compared by an ANOVA 2 × 2, with “Group” (N−/N+ patients) and “Sentence Type” (Type I, Type II) as factors. The results showed a main “Group” effect (F1,20 = 8.83, p = 0.007), since the N+ patients were more impaired than the N− patients on this task (mean of correct answers: N+ group = 28.2, S.D. 6.74; N− group = 36, S.D. 5.75). “Sentence Type” was also significant (F1,20 = 12.06, p = 0.002) because all patients tested found the Type II sentence type more difficult to process than the Type I sentence type (mean of correct answers: Type I = 17.27, S.D. 3.93; Type II = 15.18, S.D. 3.77). But, most importantly, the “Group” × “Sentence Type” interaction was not significant (F1,20 = 0.2, p = 0.64). This means that neglected patients do not process a sequence of non-linguistic sounds stressed in the initial part differently from sequences stressed in the central or final part. More specifically, this result suggests that the bias related to the presence of neglect does not affect (1) any sequence of events deployed over time, regardless of its nature (linguistic or non-linguistic), (2) a suprasegmental feature of language as prosody.
Fig. 5. Means and S.D.s of correct answers of patients with and without neglect in the two sentence types of the sequence of three nouns condition. Score ranges are shown under each histogram.
was not significant (F1,20 = 0.05, p = 0.82). Therefore, we can deduce that when there is no verb in the sentence the linguistic task is quite difficult for neglect patients. However, this is not due to greater impairment for the initial part of the linguistic string but seems to affect the different sections of the string in a similar way. An additional point concerns the possible relationship between the measures of neglect and the effect produced by the experimental manipulations (influence of stress on sentence comparison). We computed a correlation matrix between each test for neglect and the number of correct responses in our emphatic stress tasks. No correlations were significant: the highest correlations were found between the active/passive condition with the Wundt–Jastrow Illusion test (range from 0.32 to 0.52). The lowest correlations for the experimental conditions were found for the Personal neglect test and the Sentence reading test (range from 0.05 to 0.37). An overall measure of neglect severity was computed taking the average z-scores (computed on an independent pool of 120 N+ patients) of the five neglect tests. In this case, the correlation with the Passive linguistic condition was significant (−0.57, p < 0.05).
4. Discussion 4.1. Active versus passive linguistic condition
3.3. Sequence of unrelated nouns condition The means of the correct answers obtained by N+ and N− patients in the sequence of unrelated nouns condition are reported in Fig. 5. As shown by the graph, this condition—characterized by the absence of the verb—proved to be particularly difficult for the N+ group of patients. Based on the results of a 2 × 2 ANOVA (“Group” × “Sentence Type”), “Group” had a significant main effect (F1,20 = 22.44, p = 0.0001) (mean of correct answers for “Group”: N− group = 40.45, S.D. 6.12; N+ group = 29, S.D. 7.01), while “Sentence Type” did not (F1,20 = 1.24, p = 0.27) (mean of correct answers for “Sentence Type”: Type I = 17.73, S.D. 4.05; Type II = 18.41, S.D. 4.33); finally, the “Group” × “Sentence Type” interaction
First, the present results provide additional support for the results obtained in our previous study (Rinaldi et al., 2003), confirming that neglect patients have a specific impairment in processing the initial part of an active sentence (positional bias), namely, the subject-word. In the present study, a noteworthy additional result emerged from a comparison of the active versus the newly introduced passive condition, which in neglect patients is the inversion of the positional bias. This result goes beyond the reduction of the bias obtained in normal subjects by Chatterjee et al. (1995a, 1995b) passing from the active to the passive condition, since we observed an inversion, not a reduction, of the bias. Therefore, sentence structure has a role in determining the positional bias observed in N+ patients’ processing of spoken sentences.
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
An attempt to account for this inversion can be made by considering the distinction between the grammatical subject and the logical subject of a sentence. The grammatical subject is the one indicated by syntactic rules, and the logical subject is the one who actually performs the action in a sentence, i.e., the doer of the action. Although these two subjects coincide in active sentences (“The dog chases the ball” – logical/grammatical subject: “dog”), in passive sentences, instead, the logical subject appears at the end of the sentence under the grammatical guise of agent (e.g., “The ball is chased by the dog” – logical subject: “dog”; grammatical subject “ball”). Therefore, the inversion of the bias passing from active to passive structures suggests that the component specifically compromised by the presence of neglect is, namely, the doer of the action, the logical subject of the sentence. But, to reconnect this to the selective involvement of neglect for the left hemispace, we need an additional step, i.e., the consideration that active and passive sentences, though different in their “surface structures”, share the same “deep structure”, in Chomsky’s terms. In the deep structure, the logical subject (the doer of the action) always appears at the beginning, which means “on the left” in terms of the spatial transcoding of the spoken sentence hypothesized. In support of this view, that of an intrinsically spatial nature of the deep structure of language, as noted by O’Keefe (1996), has a long tradition in linguistics. According to this view, non-spatial propositions would be subsequently located in a parasitic way on intrinsically spatial prototypic structures. 4.2. Musical condition The results obtained in the musical condition provide us with important cues about the mechanisms that underlie the bias under investigation and the level of linguistic processing involved, because they show that the neglect bias is not present with any type/sequence of acoustic event (musical notes) utilized over time. Consequently, it is highly unlikely that the neglect bias acts at the level of the suprasegmental features of language such as prosody, as also suggested by the results of the active/passive condition, which, instead, point to a semantic involvement. Although both groups of patients made many errors in this condition, the errors were not distributed preferentially in any sentence type. Different factors may have acted together to determine this result. For instance, various studies have shown that the melodic contour of language is primarily processed by the right hemisphere (Bloomstein & Cooper, 1974). Therefore, the fact that patients with a right hemisphere lesion have major problems on a melodic comparison task is no surprise. Secondly, these studies have also shown that the absence of “segmental” cues in the acoustic input, which also characterized our stimuli, determined a higher number of errors. Third, in a study conducted on normal subjects Segalowitz and Plantery (1985) found that musical stimuli (pieces of orchestral music) orientated attention towards the left side
563
of space and verbal auditory stimuli towards the right. It is also possible that the attentional bias generated by musical stimuli towards the left hemispace makes up for neglect of the same hemispace, which characterizes heminattention, thus explaining the absence of a bias for Type I sentences in this experimental condition. 4.3. Sequence of unrelated nouns condition As in the case of the musical condition, this experimental condition also proved to be more difficult than expected, especially for our N+ patients, though not in terms of a differential error distribution (no bias for Type I sentences). The fact that when no verb was present in the linguistic string no positional bias emerged in N+ patients adds further support to the hypothesis (reported in Chatterjee et al., 1999) that the verb is the element responsible for genesis of the spatial representation underlying a spoken sentence (no verb = no event, no thematic roles defined, no elements to build a spatial representation). Therefore, without a verb the three nouns given are just unrelated concepts without any unifying structure. Thus, the presence of a left-to-right “vector-for-action”, represented by the verb and claimed by Chatterjee et al. (1999) to be the basis for the representation of events expressed by means of language, seems to be indispensable for manifestation of the positional bias exhibited by N+ patients in spoken language processing. Apart from the impossibility of building a structured representation of the “event”, this could also mean there is a greater demand on memory resources to carry out this task, thus explaining the lower performance of the patients tested. 4.4. Correlational results The low correlations between the experimental tasks and the measures of neglect, with the exception of the overall index of neglect severity and the Passive linguistic condition, require a final comment. First of all, it is known that neglect is not a unitary syndrome; for this reason, its diagnosis requires the use of more than one test since a patient may show neglect in one test but not in another. This clinical evidence indicates the existence of multiple spatial dimensions that can be separately or jointly affected by the neglect syndrome, dimensions which are not necessarily covered completely by the existing clinical tests, such as the ones used in our research. Additionally, it must be considered that while the tasks used to identify neglect deal with explicit visuo-spatial stimuli, the effect involved by the emphatic stress, inserted in different strings of acoustic signals, depends on a spatial transcoding of auditory traces. This transcoding, although effective in the group of neglect patients, may not be activated in all individuals. These individual differences in a related field of investigation have been described in research on the “number-line”, where a significant proportion of normal subjects did not show
564
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
the activation of spatial transcoding in different kinds of numerical processing (Dehaene, Bossini, & Giraux, 1993, for instance, found the number-line effects to be significant in students of scientific, but not always of literary studies). It may be suggested that although the presence of a spatial bias in neglect patients influences the activation of a “spatial language-line”, the use of spatial transcoding is associated more with individual differences and less with the intensity of neglect disorders.
5. General discussion Our data show that in brain-damaged patients an interaction exists between the presence of neglect and the type of task administered. The task consisted of comparing the stress positions in two heard sentences, a comparison, which directly involved the acoustic traces left by the spoken sentences in the working memory. When they were given two sequences of words to compare, patients with unilateral neglect made more errors when the cue (the stress) appeared in the initial part than in the central or the final part of the sentence. This type of effect exerted by neglect on the task employed is strongly suggestive of a spatial transcoding of the auditory traces left in the working memory by the heard sentence, as already postulated in our previous study. Our data from the previous and the present study suggest that spoken language may activate a “spatial language line’, which is a spatial transcoding of linguistic input. This transcoding is then perturbed in a spatially predictable way by the presence of a spatial bias such as the heminattentional syndrome. With respect to our previous work, we are now able to better define the conditions required for the effect to take place. The effect is not found for all auditory input, but only in specific circumstances. In fact, we found that neglect had an effect on the linguistic task in the active/passive condition, but not in the musical or in the three-noun condition. We also found that the effect was inverted in the positive condition, the active /passive condition, when passing from the active to the passive sentence structure. These results mainly suggest the following: (1) neglect interacts with auditory linguistic input; e.g. a sequence of musical notes, although forming a “musical sentence”, does not produce the same spatial bias; (2) neglect interacts with a syntactically structured auditory linguistic input (agent/action/recipient of action), and not with just any auditory linguistic input (noun /noun/ noun); (3) the verb has a primary role in determining the presence of a structured spatial representation in linguistic input, defining the thematic roles and providing the frame for the event narrated in the sentence; (4) neglect acts at the level of the deep structure of the sentence, which in the case of active sentences coincides with surface structures. More exactly, we hypothesize that neglect acts on the spatial transcoding of the deep structure of linguistic input; the deep structure is a representation of “who does
something to whom”, and in its spatial transcoding the logical subject –the agent of the action, the “who”—consequently appears “on the left” (in both active and passive sentences). We see, therefore, that the absence of a structure underlying linguistic input, determined by the absence of the verb, was a critical factor, and this absence characterized both the three-noun and the musical condition employed in this study in which the neglect bias was not observed. In the musical condition, it can be objected that the stimuli employed were pieces of real music, with an actual underlying structure. However, since none of the subjects tested in our study had had any previous musical education, it is highly unlikely they could have perceived the underlying structure. It is possible that a different result would have emerged with professional musicians; but, unfortunately, very few musicians with neglect are described in the literature (we know of only one case reported by Landi et al., 1997). In the case of the three-noun sequence, instead, stimuli were actually processed as linguistic and had a semantic content (each noun is a concept). But, also here, semantic content per se did not seem to be enough for the neglect bias to appear. Semantic content must, in fact, be engraved into a carrying structure, defined by the verb, which establishes which concept acts as agent and which as recipient of action within the whole sentence frame. To conclude, we would like to present some considerations about the entity of the bias exerted by neglect on spoken language. As estimated from the active sentences, the bias was not present in every N+ patient tested but in 8/12 patients, and in a different strength in each one. The fact that the effect of neglect on spoken language was not universal conforms well to results obtained by other authors who investigated the relationship between language and space. Chatterjee, for instance, admits that in some of his experiments with normal subjects the spatial bias was evaluated only in the limited percentage of subjects who constantly placed the agent of the action on the same side of the page (in one experiment, only 37% of the subjects) (Chatterjee et al., 1999), and that the others were not included in the analysis. Also, Coslett (1999) found his hemifield effects only in patients with a specific lobe involvement (parietal lobe, independently of the presence of neglect); but, here too, not every patient exhibited these effects. Leaving the language domain, other authors who investigated the spatial transcoding of numerical quantities (the number-line) did not find this relationship to be ubiquitous (Dehaene et al., 1993). A. Kim too found the number-line in only 63% of the subjects tested (A. Kim, personal communication). This spatial transcoding of numerical quantities, i.e., the number-line, represents a good piece of evidence showing, as in the study we described, how spatial transcoding of a different domain (numbers) can be affected by the presence of a spatial deficit, i.e., neglect, undergoing a shift towards the ipsilesional side when this syndrome is present (Zorzi, Priftis, & Umilt`a, 2002; Vuilleumier, Ortigue, & Brugger, 2004).
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Acknowledgement This research has been supported by grant of Ministero Italiano della Sanit`a. References Albert, M. L. (1973). A simple test of visual neglect. Neurology, 23, 658–664. Barbut, D., & Gazzaniga, M. S. (1987). Disturbances in conceptual space involving language and speech. Brain, 110(6), 1487–1496. Baxter, D. M., & Warrington, E. K. (1983). Neglect disgraphia. Journal of Neurology, Neurosurgery and Psychiatry, 46, 1073–1078. Behrens, S. J. (1985). The perception of stress and lateralization of prosody. Brain and Language, 26, 332–348. Bloomstein, S., & Cooper, W. (1974). Hemispheric processing of intonation contours. Cortex, 10, 146–158. Caramazza, A., & Hillis, A. E. (1990). Levels of representation, coordinate frames, and unilateral neglect. Cognitive Neurospychology, 7, 391–445. Chatterjee, A., Maher, L., & Heilman, K. (1995). Spatial charachteristics of thematic role representation. Neuropsychologia, 33, 643–648. Chatterjee, A., Maher, L. M., Gonzalez Rothi, L. J., & Heilman, K. M. (1995). Asyntactic thematic role assignment: The use of a temporalspatial strategy. Brain and Language, 49, 125–139. Chatterjee, A., Southwood, M. H., & Basilico, D. (1999). Verbs, events and spatial representations. Neuropsychologia, 37, 395–402. Coslett, H. B. (1999). Spatial influences on motor and language function. Neuropsychologia, 37, 695–706. Coslett, H. B., Schwartz, M. F., Goldberg, G., Haas, D., & Perkins, J. (1993). Multimodal hemispatial deficits after left hemisphere stroke. Brain, 116, 527–554. Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental representation of parity and number magnitude. Journal of Experimental Psychology, 122, 371–396. Diller, L., Weinberg, J., Piasetsk, E., Ruckdeschel-Hibbard, M., Egelko, S., Scotzin, et al. (1974). Methods for the evaluation and treatment of the visual perceptual difficulties of right brain damaged individuals.
565
New York University Medical Center, supplement of the 8th Annual Workshop for Rehabilitation Professionals. Istituto di Linguistica Computazionale del CNR di Pisa (1989). Barcellona Corpus di Italiano Scritto Contemporaneo. Unpublished manuscript. Landi, F., Zuccala, G., Cocchi, A., Bernabei, R., Tafani, A., & Carbonin, P. (1997). Playing piano in visuo-spatial neglect: A case study. Journal of Neurology, Neurosurgery and Psychiatry, 62(5), 543– 544. Mandler, J. M. (1996). Preverbal representation and language. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space (pp. 365–384). Cambridge, MA: MIT Press, Cap. 9. Massironi, M., Antonucci, G., Pizzamiglio, L., Vitale, M. V., & Zoccolotti, P. (1988). The Wundt-Jastrow illusion in the study of spatial hemiinattention. Neuropsychologia, 26, 1661–1666. O’Keefe, J. (1996). The spatial preposition in English, vector grammar, and the cognitive map theory. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space (pp. 365–384). Cambridge, MA: MIT Press, Cap. 9. Orsini, A., Grossi, D., Capitani, E., Laiacona, M., Papagno, C., & Vallar, G. (1987). Verbal and spatial immediate memory span: Normative data from 1355 adults and 1112 children. Italian Journal of Neurological Sciences, 8, 539–548. Pizzamiglio, L., Antonucci, G., Guariglia, C., Judica, A., Montenero, P., Razzano, C., et al. (1992). Cognitive rehabilitation of the hemineglect disorders in chronic patients with unilateral brain damage. Journal of Clinical and Experimental Neuropsychology, 14, 901–923. Rinaldi, C., Marangolo, P., & Pizzamiglio, L. (2003). Between language and space. A cross-domain interaction. Neuroreport, 14(10), 1381–1383. Segalowitz, S. J., & Plantery, P. (1985). Music draws attention to the left and speech draws attention to the right. Brain and Cognition, 4(1), 1–6. Vuilleumier, P., Ortigue, S., & Brugger, P. (2004). The number space and neglect. Cortex, 40, 399–410. Zoccolotti, P., Antonucci, G., & Judica, A. (1992). Psychometric characteristics of two semi-structured scales for the functional evaluation of hemi-inattention in extra-personal and personal space. Neuropsychological Rehabilitation, 2, 179–191. Zorzi, M., Priftis, C., & Umilt`a, C. (2002). Brain damage: Neglect disrupts the mental number line. Nature, 417, 138–139.
When space merges into language M. Cristina Rinaldi a,∗ , Luigi Pizzamiglio a,b a
Centro Ricerche di Neuropsicologia, Fondazione I.R.C.C.S. Santa Lucia, Via Ardeatina 306, 00179 Rome, Italy b Dipartimento di Psicologia Universit` a “La Sapienza”, Rome, Italy Received 10 January 2005; received in revised form 25 May 2005; accepted 7 July 2005 Available online 30 August 2005
Abstract We present data from right brain-damaged patients, with and without spatial heminattention, which show the influence of hemispatial deficits on spoken language processing. We explored the findings of a previous study, which used an emphatic stress detection task and suggested spatial transcoding of a spoken active sentence in a ‘language line’. This transcoding was impaired in its initial portion (the subjectword) when the neglect syndrome was present. By expanding the original methodology, the present study provides a deeper understanding of the level of spoken language processing involved in the heminattentional bias. To ascertain the role played by syntactic structure, active and passive sentences were compared. Sentences comprised of musical notes and of a sequence of unrelated nouns were also compared to determine whether the bias was manifest with any sequence of events (not only linguistic ones) deployed over time, and with a sequence of linguistic events not embedded in a structured syntactic frame. Results showed that heminattention exerted an influence only when a syntactically structured linguistic input (=sentence with agent of action, action and recipient of action) was processed, and that it did not interfere when a sequence of non-linguistic sounds or unrelated words was presented. Furthermore, when passing from active to passive sentences, the heminattentional bias was inverted, suggesting that heminattention primarily involves the logical subject of the sentence, which has an inverted position in passive sentences. These results strongly suggest that heminattention acts on the spatial transcoding of the deep structure of spoken language. © 2005 Elsevier Ltd. All rights reserved. Keywords: Space; Language; Neglect
1. Introduction The spatial dimension has been so relevant in the evolution of the nervous system that sensory and motor processing cannot take place unless their spatial co-ordinates are coded. Coslett (1999) claimed that the influence of space also holds for higher-order more abstract functions (including language), which are less obviously related to space. Coslett’s (Coslett, Schwartz, Goldberg, Haas, & Perkins, 1993) considerations originated from clinical observations carried out on an aphasic patient, J.F., whose anomia became more or less severe depending on the hemifield stimuli were presented to. According to Coslett (1999), the importance of
∗
Corresponding author. E-mail address: [email protected] (M.C. Rinaldi).
0028-3932/$ – see front matter © 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2005.07.007
spatial information for every action we perform is so strong that each stimulus we perceive is automatically marked with reference to its co-ordinates in the egocentric space, even if spatial information does not seem relevant for the task at hand (“Spatial Registration Hypothesis”). Following this reasoning, any stimulus presented in the contralesional field (for patient J.F., the right hemispace) should activate the left hemisphere attentional system for the impaired right hemispace. This explains J.F.’s worse performance with stimuli in the contralesional field, while his performance in the left hemispace was supported by his intact right hemisphere (Coslett et al., 1993). This interpretation stresses the importance of the extrapersonal, egocentric space linguistic stimuli arise from (Coslett, 1999). Other authors hypothesize that any event expressed through language, whether spoken or written, automatically activates an “internal” spatial representation.
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
The latter explanation is postulated by Chatterjee, Maher, Gonzalez Rothi, and Heilman (1995b). They described an agrammatic patient, W.H., who assigned the thematic roles of agent and recipient of a given action on the basis of a spatial or temporal strategy. For example, in describing a picture, this patient was asked to report who did something to whom (agent and recipient of action). He tended to consider the figure located on the left as the agent of action (spatial strategy), or to match the first noun he heard with the figure located on the left (temporal strategy) (Chatterjee et al., 1995b). According to Chatterjee et al. (1995b), conceptual knowledge of thematic roles is actually independent of language, and is matched with linguistic content only subsequently, by means of the grammatical and syntactic algorithms specific to one’s own language. Due to a brain lesion, these linguistic algorithms were impaired in their agrammatic patient. Therefore, he based his thematic role detection exclusively on the automatic spatial representation of events, which proceeds naturally from left (agent of action) to right (recipient of action), with the verb being the vector defining the thematic roles and supporting the whole event frame. To further support this idea, Chatterjee, Maher, and Heilman (1995a) and Chatterjee, Southwood, and Basilico (1999) asked non brain-damaged subjects to draw an assigned thematic role. Their reasoning was that if there is a spatial representation underlying a spoken sentence, and an “implicit propensity” to conceive events as traversing space from left to right, with the agent of the action on the left of this representational space, the subjects should draw a given action proceeding from left to right. They showed that the majority of subjects located the agent closer to the left margin of the page, and the recipient of the action closer to the right margin of the page (Chatterjee et al., 1995a, 1999). For Chatterjee and colleagues, the presence of a spatial representation underlying spoken language reflects the primitive way the mind conceptualizes and represents events, a way that has been successively obscured—but not deleted— during the course of development by the evolution of language. Analogously, passing from phylogenesis to ontogenesis, Mandler (1996) observed that in the young child emerging linguistic concepts are mapped on a pre-existing conceptual system, which has space as a fundamental constituent dimension. If space and language interact at some still unidentified level, forming the spatio-linguistic representation outlined above, it is theoretically possible to observe representational abnormalities as a consequence of (1) a brain lesion modifying spatial processing and (2) a brain lesion modifying linguistic processing in two opposite, but converging, lines of action. Some evidence for the latter is provided by the two aphasic patients described above, i.e., by Coslett et al. (1993) and by Chatterjee et al. (1995b). Instead, some evidence for the former is provided by two patients with hemispatial neglect, one described by Baxter and Warrington (1983) and the other by Caramazza and Hillis (1990).
557
Patient O.R.F., described by Baxter and Warrington (1983), underwent two consecutive vascular accidents within a short period of time. The first parietal lesion caused conduction aphasia (the patient was left-handed), phonological dyslexia, and mild neglect dyslexia, without other signs of spatial neglect; the second vascular event, again striking the right parietal region, produced an additional impairment of word-spelling abilities that was not present before. The word–spelling deficit was observed only for letters at the beginning of words; for example, if the examiner spelled the word “b-e-g-g-a-r” aurally letter by letter, the patient reported “vinegar” (Baxter, & Warrington, 1983). Interestingly, this bias was manifested when spelling proceeded either from the beginning to the end of the word, or from the end to the beginning of the word. Baxter and Warrington (1983) hypothesized that the patient was representing the stimulus on an inner screen, and the symptoms resulted from an interaction between representational neglect and neglect dyslexia. The authors affirmed that the patient reported “reading” the aurally spelled words on an “inner screen”. Instead, patient N.J., described by Caramazza and Hillis (1990), exhibited severe neglect affecting the right half of space after a lesion to the left hemisphere; he had no language deficits. N.J. was submitted to several reading and spelling tests: horizontal reading, vertical reading, mirrorreverted reading, delayed copy, aural spelling, recognition of aural spelling, written spelling and backwards written spelling. When reading, the patient made errors exclusively at the end of words, irrespective of the form of the input (vertical, horizontal or mirror reversed). Also, on the spelling tasks errors appeared only at the right end of a word (N.J. reported “exceed” after the examiner spelled “e-x-c-e-s-s“), irrespective—also for spelling—of the form or order of the input (written and aural spelling, forward and backward spelling). According to Caramazza and Hillis (1990), reading and spelling both involve a common pre-lexical level of representation, which they call “grapheme description”. At this level of representation, the abstract identity of the single letters compounding the word is computed, regardless of the form/orientation of the input; therefore, damage to this format-independent level would impair both reading and spelling in a similar way and in conformity with the heminattentional deficit. Since the spelling tasks were performed aurally, the behavior of both patients strongly suggests a spatial transcoding of spoken language which, given its spatial format, undergoes the same bias observed in neglected patients for visuo-spatial stimuli in the contralesional hemispace. The clinical reports of these two patients document the influence of hemineglect on spoken language processing, but with reference to a very peculiar type of spoken language: spelling. Although spelling is performed aurally, in many respects it is actually very close to written language. This is probably why Baxter and Warrington defined the deficit described in their patient as a rare type of “dysgraphic
558
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
syndrome”, highlighting with the term “dysgraphic” the affinity of spelling and written language. With regard to the effects of hemineglect on genuine spoken language, a neglect patient -J.L.- was reported by Barbut and Gazzaniga (1987). As a consequence of a right parietal lesion, in addition to errors affecting the most leftward letters of words in reading, writing and spelling, J.L. also produced errors in spontaneous speech. For example, he often omitted the beginning of a word, saying “bulance” instead of “ambulance”, or “portant” instead of “important”. Most of his errors were omissions, but the examiner also observed that the patient often made substitutions, forming semantically unrelated words. According to Barbut and Gazzaniga (1987), the pattern of errors exhibited by their patient was consistent with the hypothesis that some aspects of language processing involve some spatial mechanisms responsible for the representation of what they call “conceptual space”. Recently, a decade after the literature reported above, a paper by Rinaldi, Marangolo, and Pizzamiglio (2003) added another piece of evidence to this field of investigation. The authors described a different type of hemineglect effect on the processing of spoken language than that previously observed. When asked to compare two spoken sentences (subject/verb/object) with regard to their emphatic stress, patients with right hemisphere lesions and spatial neglect made significantly more errors when the stress was placed at the beginning of the sentence (on the subject-word). The bias exhibited by neglected patients towards the initial word of the sentence suggests a transcoding of spoken language into a spatial format. The auditory trace left by the spoken sentence would be automatically transformed into an analog with an intrinsic spatial component. This analog would then be affected by the presence of hemineglect in the same way this deficit affects spatial information processing, which is on the contralesional side. In other terms, it appears that the word at the beginning of the sentence falls on the “left” side of the spatial transcoding of the sentence and is, therefore, neglected similar to the way left visuo-spatial information is neglected by these patients in the external space. The present investigation aims at expanding the original methodology and qualifying the nature of the interaction between the spatial and linguistic mechanisms underlying the observed effect. As in the previous experiment, the first question concerns the use of a unique format, i.e., active sentence structure. The replication or the change in the results using active and passive sentences will contribute to understanding whether the effect is related to the position of the stress in the auditory string of words or, instead, to its syntactic structure. The second question is whether the comparison of the stress location in the two sentences produces a bias in hemineglect patients only when linguistic strings are used, or whether, instead, it takes place also when two non-linguistic strings of acoustic events are compared. In other words, if the bias in hemineglect patients can also be observed when they compare the stress embedded in two musical sentences,
the effect can be attributed exclusively to the need to locate a target stimulus in a sequence of events spaced over time. Third, do the stimuli need to be meaningful sentences or can the same effect of neglect be observed with a sequence of non-related words? Does neglect affect the representation of the event expressed in the spoken sentence, where a verb supports the event frame and defines the thematic roles, or does it affect any sequence of three nouns, independently of the presence of a verb?
2. Methods 2.1. Subjects This study was carried out on 26 right brain-damaged subjects admitted to the Santa Lucia Foundation I.R.C.C.S. in Rome for rehabilitation (mean age 61.7 years, S.D. 11.37; education 9.7 years, S.D. 4.19) (see Table 1 for individual patients’ data). For each patient, unilaterality of the brain lesion was documented by a CT or an MR scan. Each patient admitted to the hospital underwent a standard neuropsychological examination, which included a battery for the diagnosis of hemispatial neglect for personal and peripersonal space. The following tests comprised the battery: Cancellation Test (Albert, 1973), the Letter Cancellation Test (Diller et al., 1974), the Sentence Reading Test (Pizzamiglio et al., 1992), the Wundt–Jastrow Area Illusion Test (Massironi, Antonucci, Pizzamiglio, Vitale & Zoccolotti, 1988), and the Personal Neglect Test (Zoccolotti, Antonucci, & Judica, 1992). Patients who failed on at least three of the five tests for heminattention were classified as having hemispatial neglect. According to this criterion, patients were divided into two sub-groups, 12 patients with hemineglect (N+ Group: mean age 62 years, S.D. 8.33; mean education years 10, S.D. 4.31) and 14 patients without hemineglect (N− Group: mean age 62 years, S.D. 13.44; mean education years 10, S.D. 4.07). The neuropsychological examination also included an evaluation of phonological and visuo-spatial short-term memory (Wais Digit Span Test, Orsini et al., 1987; Corsi Span Test, Orsini et al., 1987). Phonological short-term memory was intact in all subjects examined, whereas visuo-spatial short-term memory, unimpaired in patients without hemineglect, was impaired in 4 of the 11 patients with neglect (since neglect was in part responsible for the fact that these patients scored below the cut-off). For the experimental condition described below as “Sequence of Three Nouns”, a different group of right braindamaged patients was tested, since this condition was added later. Respectively, 11 N− and 13 N+ patients were tested (N− group: mean age 63.6 years, S.D. 12.69; mean education years 10.53, S.D. 3.86; N+ group: mean age 60.3 years, S.D. 13.8; mean education years 10.61, S.D. 4.09). Patients with hearing deficits not previously acknowledged but manifested during the experimental sessions were
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
559
Table 1 Individual data of N− and N+ patients for sex, years of education, time post onset and locus of lesion Subjects (N = 26)
Age
Sex
Education (years)
Time post-onset (months)
Lesion site
N− group 1 2 3 4 5 6 7 8 9 10 11 12 13 14
74 58 52 62 48 57 52 39 57 52 76 79 71 82
F M M M M M F M M M M F M M
12 10 5 17 13 11 8 8 17 13 8 5 5 5
2 3 8 1 30 2 360 4.5 6 7 3 6 4 1.5
Right cortico-subcortical fronto-parietal lesion Right internal capsula + lateral thalamic lesion Right extern caps + nucl caud + white matt periventric parietal lesion Right temporo-parietal lesion Right subcortico- temporal lesion Right occipital lesion Right fronto-parietal lesion Right cortico-subcortical fronto-temporal + internal capsula Right parietal lesion Right thalamic lesion Right cortico-subcortico temporal lesion Right semioval center + frontal lobe lesion Right pons lesion Right temporal lesion
N+ group 1 2 3 4 5 6 7 8 9 10 11 12
60 62 71 43 64 54 59 75 53 68 63 75
M F F M M M M F F M M M
7 3 11 8 10 7 13 5 16 13 17 5
2 6 3.5 5 7 2 2 2 1 4.5 3 1.5
excluded from the study. The research was approved by the Ethical Committee and all patients gave their written consent. 2.2. Stimulus material Patients were requested to judge two spoken sentences emitted from a tape recorder in four different experimental conditions: (1) Active versus passive linguistic condition. (2) Musical condition. (3) Sequence of unrelated nouns condition. 2.2.1. Active–passive linguistic condition Stimuli consisted of spoken sentences recorded by a tape recorder; all sentences were in Italian and in the active or passive form, and only high frequency and highly familiar words were used. Each sentence was comprised of three elements: a subject, a verb and an object (e.g., “The boy reads the book”). The following description refers to the active sentences and it will be replicated for the passive sentences. Each item consisted of a pair of linguistically identical sentences, and with emphatic stress placed on the same or on a different word in each of the two sentences. The two sentences of the pair were separated by a two-second time interval. The emphatic
Right parietal lesion Right fronto-temporo-parietal lesion Right cortico-subcortical fronto-temporal lesion Right internal capsula lesion Right frontal lesion Right temporo-parietal lesion Right cortico-subcortical temporo-fronto-parietal lesion Right temporo-fronto-parietal lesion Right temporo-fronto-parietal lesion Right insular + temporo-parietal lesion Right parieto-temporal lesion Right insula + frontal lesion
stress could be placed either on the subject-, the verb- or on the object-word. Following is an example of stress placed on the same elements of a sentence pair: “The boy reads the book”/“The boy reads the book”; here, instead, are examples of stress placed on different elements of a sentence pair: “The wolf attacks the sheep”/”The wolf attacks the sheep”; “The thunderbolt struck the house”/“The thunderbolt struck the house”. The patients were asked to listen to each sentence pair and to judge whether the two sentences were stressed in the same way or not, or, more simply, if their intonation “sounded the same or not” (answer: “yes”/“no”). Each pair of sentences was preceded by an acoustic warning signal, and no time-limit for the answer was set by the examiner. Two separate sets of active and passive sentences (comprising two practice trials + 48 test trials each) were included in the test. For each experimental condition, a different set of 48 + 2 stimuli was used; however, in each condition the whole set of 48 stimuli was balanced according to the same schema, with half of the sentences requiring a “yes” answer and half a “no” answer. Most importantly, the whole set of stimuli was divided into two sub-groups according to the position of the emphatic stress: Type I, including all sentence pairs with the empathic stress on the subject word in at least one of the two sentences of the pair (e.g., “The boy reads the book” / “The
560
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Fig. 1. Schematic representation of item types and their separation into Type I and Type II.
boy reads the book”); and Type II, including all sentence pairs in which the emphatic stress was never on the subject word, but rather on the verb or the object word (e.g., “The teacher writes a letter” / “The teacher writes a letter”) (see Fig. 1). In each set of 48 stimuli, half of the sentence pairs were Type I and the other half, Type II. The set of 48 passive sentences had exactly the same characteristics as the active set. Here, too, there were 24 “same”sentence pairs with the stress located on the same element (e.g., “The ball is hit by the boy” / “The ball is hit by the boy”), and 24 “different” sentence pairs, with the stress located on different elements
(e.g., “The window is closed by the man” / “The window is closed by the man”). In the passive sentences set, the position of emphatic stress, the balancing and the separation of stimuli into Type I and Type II sub-groups were exactly the same as described for the active sentences. The two blocks of active and passive sentences were always presented in the active-first, passive-second sequence. 2.2.2. Musical condition In this condition, sentences formed by a sequence of musical notes instead of words were used. We substituted linguistic with musical stimulus material to ascertain whether
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
the bias of neglect on the spatial representation underlying spoken language affects (1) not language per se, but rather any sequence of elements deployed over time regardless of their linguistic nature, and (2) the so-called suprasegmental features of language, which include sentence stress as a component of linguistic prosody (in a musical sentence, the structure is analogous to the melodic contour of language). A professional musician played notes on a piano and they were recorded on a tape-recorder. We did not make the musical sentences by assembling the notes at random, which, according to the musician, would have been the same as meaningless linguistic babbling, but instead by taking fragments of real pieces of music which had an underlying structure with a meaning. Specifically, sentences consisted of one-voice thematic fragments extracted from a thematic catalogue of J.S. Bach (1685–1750). To form the musical sentences, only fragments with a ternary structure were chosen, so that the structure would be analogous to the threeelement structure of the linguistic sentences; therefore, also in the musical sentences an initial, a central and a final part of the sentence was detectable. Emphatic stress could be placed on each of these three components since, in this case, the stress was made by means of a marked increase in the loudness of the notes played on the keyboard. All sentences had been previously administered to a group of non brain-damaged control subjects to verify that the variations in emphasis could be detected, as foreseen by the research protocol. Here, too, sentences were presented in pairs (48 + 2). The notes comprising the two sentences of the pair were the same and the only variable element was the position of emphatic stress. So, apart from the nature of the stimuli, i.e., musical rather than verbal, the task was exactly the same as the other conditions described: listening to two “sentences” and saying whether they sounded the same or not. 2.2.3. Sequence of three nouns condition In this condition, sentences comprised of a sequence of three nouns were used instead of the canonical noun + verb + noun sentences. The verb was purposely excluded in order to ascertain its relevance—as claimed by linguists like Jackendoff and Gruber (quoted in Chatterjee et al., 1999)—in defining thematic roles and in supporting the whole spatial representation of linguistic events. In other words, if the verb is the element responsible for the spatial representation underlying spoken language, the effect of neglect on spoken language should not occur when the verb is not present. Nouns used were high frequency bi- and tri-syllabic words taken from the frequency lexicon of the Italian language “Barcellona Corpus” (Istituto di Linguistica Computazionale del CNR di Pisa, 1989). For this condition, 48 + 2 sentence pairs were created as well. Each sentence comprised three nouns in sequence, i.e. “dog–apple–pen”, repeated twice with the same or with a different emphatic stress (the same emphatic stress balancing used for the canonical noun + verb + noun sentences).
561
Fig. 2. Comparison of the means and S.D.s of correct answers of N+ patients in the active linguistic vs. passive linguistic condition. Score ranges are shown under each histogram.
3. Results 3.1. Active–passive linguistic condition An ANOVA 2 × 2 × 2 was computed to compare the performance of the two groups of patients on the two sentence types, active and passive; factors used were “Group” (N−/N+ patients), “Sentence Type” (Type I/Type II) and “Condition” (active/passive). “Group” and “Condition” showed no main effects (p = 0.20 and p = 0.53); however, “Sentence Type” was significant (Type I: mean 19.16, S.D. 3.54; Type II: mean 18.34, S.D. 3.54) (F1,24 = 10.52, p = 0.003). The “Group” × “Sentence Type” interaction was not significant (p = 0.16), whereas the “Group” × “Condition” (interaction (p = 0.02) was significant; the “Condition” × “Sentence Type” interaction was also significant (p = 0.000); finally, the three-way “Group” × “Sentence Type” × “Condition” interaction was significant (p = 0.02). A Duncan post-hoc analysis revealed that: (a) the processing of Type I sentences by N+ patients differed significantly in the active versus the passive linguistic condition (p = 0.0001); (b) on the contrary, the processing of Type I sentences by N− patients did not differ in the two linguistic conditions (p = 0.24); N+ patients’ processing of Type II sentences also differed in the active versus the passive linguistic condition (p = 0.04). Therefore, as also shown in Figs. 2 and 3, N+ patients’ differential processing of Type I sentences distinguished the performance of the two groups of patients in active versus passive sentences. Specifically, in active sentences N+ patients made more errors when the stress was at the beginning of the sentence (Type I), and in passive sentences they made most of their errors when the stress was at the center or the end of the sentence (Type II).
Fig. 3. Comparison of the means and S.D.s of correct answers of N− patients in the active linguistic vs. passive linguistic condition. Score ranges are shown under each histogram.
562
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Fig. 4. Means and S.D.s of correct answers of patients with and without neglect in the two sentence types of the musical condition. Score ranges are shown under each histogram.
3.2. Musical condition The means of the correct answers obtained by N+ and N− patients in the musical condition are reported in Fig. 4. As shown by the graph, this condition proved to be quite difficult for both groups of patients compared to the other experimental conditions employed in this study. As before, the performance of the two groups of patients on the two types of musical sentences was compared by an ANOVA 2 × 2, with “Group” (N−/N+ patients) and “Sentence Type” (Type I, Type II) as factors. The results showed a main “Group” effect (F1,20 = 8.83, p = 0.007), since the N+ patients were more impaired than the N− patients on this task (mean of correct answers: N+ group = 28.2, S.D. 6.74; N− group = 36, S.D. 5.75). “Sentence Type” was also significant (F1,20 = 12.06, p = 0.002) because all patients tested found the Type II sentence type more difficult to process than the Type I sentence type (mean of correct answers: Type I = 17.27, S.D. 3.93; Type II = 15.18, S.D. 3.77). But, most importantly, the “Group” × “Sentence Type” interaction was not significant (F1,20 = 0.2, p = 0.64). This means that neglected patients do not process a sequence of non-linguistic sounds stressed in the initial part differently from sequences stressed in the central or final part. More specifically, this result suggests that the bias related to the presence of neglect does not affect (1) any sequence of events deployed over time, regardless of its nature (linguistic or non-linguistic), (2) a suprasegmental feature of language as prosody.
Fig. 5. Means and S.D.s of correct answers of patients with and without neglect in the two sentence types of the sequence of three nouns condition. Score ranges are shown under each histogram.
was not significant (F1,20 = 0.05, p = 0.82). Therefore, we can deduce that when there is no verb in the sentence the linguistic task is quite difficult for neglect patients. However, this is not due to greater impairment for the initial part of the linguistic string but seems to affect the different sections of the string in a similar way. An additional point concerns the possible relationship between the measures of neglect and the effect produced by the experimental manipulations (influence of stress on sentence comparison). We computed a correlation matrix between each test for neglect and the number of correct responses in our emphatic stress tasks. No correlations were significant: the highest correlations were found between the active/passive condition with the Wundt–Jastrow Illusion test (range from 0.32 to 0.52). The lowest correlations for the experimental conditions were found for the Personal neglect test and the Sentence reading test (range from 0.05 to 0.37). An overall measure of neglect severity was computed taking the average z-scores (computed on an independent pool of 120 N+ patients) of the five neglect tests. In this case, the correlation with the Passive linguistic condition was significant (−0.57, p < 0.05).
4. Discussion 4.1. Active versus passive linguistic condition
3.3. Sequence of unrelated nouns condition The means of the correct answers obtained by N+ and N− patients in the sequence of unrelated nouns condition are reported in Fig. 5. As shown by the graph, this condition—characterized by the absence of the verb—proved to be particularly difficult for the N+ group of patients. Based on the results of a 2 × 2 ANOVA (“Group” × “Sentence Type”), “Group” had a significant main effect (F1,20 = 22.44, p = 0.0001) (mean of correct answers for “Group”: N− group = 40.45, S.D. 6.12; N+ group = 29, S.D. 7.01), while “Sentence Type” did not (F1,20 = 1.24, p = 0.27) (mean of correct answers for “Sentence Type”: Type I = 17.73, S.D. 4.05; Type II = 18.41, S.D. 4.33); finally, the “Group” × “Sentence Type” interaction
First, the present results provide additional support for the results obtained in our previous study (Rinaldi et al., 2003), confirming that neglect patients have a specific impairment in processing the initial part of an active sentence (positional bias), namely, the subject-word. In the present study, a noteworthy additional result emerged from a comparison of the active versus the newly introduced passive condition, which in neglect patients is the inversion of the positional bias. This result goes beyond the reduction of the bias obtained in normal subjects by Chatterjee et al. (1995a, 1995b) passing from the active to the passive condition, since we observed an inversion, not a reduction, of the bias. Therefore, sentence structure has a role in determining the positional bias observed in N+ patients’ processing of spoken sentences.
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
An attempt to account for this inversion can be made by considering the distinction between the grammatical subject and the logical subject of a sentence. The grammatical subject is the one indicated by syntactic rules, and the logical subject is the one who actually performs the action in a sentence, i.e., the doer of the action. Although these two subjects coincide in active sentences (“The dog chases the ball” – logical/grammatical subject: “dog”), in passive sentences, instead, the logical subject appears at the end of the sentence under the grammatical guise of agent (e.g., “The ball is chased by the dog” – logical subject: “dog”; grammatical subject “ball”). Therefore, the inversion of the bias passing from active to passive structures suggests that the component specifically compromised by the presence of neglect is, namely, the doer of the action, the logical subject of the sentence. But, to reconnect this to the selective involvement of neglect for the left hemispace, we need an additional step, i.e., the consideration that active and passive sentences, though different in their “surface structures”, share the same “deep structure”, in Chomsky’s terms. In the deep structure, the logical subject (the doer of the action) always appears at the beginning, which means “on the left” in terms of the spatial transcoding of the spoken sentence hypothesized. In support of this view, that of an intrinsically spatial nature of the deep structure of language, as noted by O’Keefe (1996), has a long tradition in linguistics. According to this view, non-spatial propositions would be subsequently located in a parasitic way on intrinsically spatial prototypic structures. 4.2. Musical condition The results obtained in the musical condition provide us with important cues about the mechanisms that underlie the bias under investigation and the level of linguistic processing involved, because they show that the neglect bias is not present with any type/sequence of acoustic event (musical notes) utilized over time. Consequently, it is highly unlikely that the neglect bias acts at the level of the suprasegmental features of language such as prosody, as also suggested by the results of the active/passive condition, which, instead, point to a semantic involvement. Although both groups of patients made many errors in this condition, the errors were not distributed preferentially in any sentence type. Different factors may have acted together to determine this result. For instance, various studies have shown that the melodic contour of language is primarily processed by the right hemisphere (Bloomstein & Cooper, 1974). Therefore, the fact that patients with a right hemisphere lesion have major problems on a melodic comparison task is no surprise. Secondly, these studies have also shown that the absence of “segmental” cues in the acoustic input, which also characterized our stimuli, determined a higher number of errors. Third, in a study conducted on normal subjects Segalowitz and Plantery (1985) found that musical stimuli (pieces of orchestral music) orientated attention towards the left side
563
of space and verbal auditory stimuli towards the right. It is also possible that the attentional bias generated by musical stimuli towards the left hemispace makes up for neglect of the same hemispace, which characterizes heminattention, thus explaining the absence of a bias for Type I sentences in this experimental condition. 4.3. Sequence of unrelated nouns condition As in the case of the musical condition, this experimental condition also proved to be more difficult than expected, especially for our N+ patients, though not in terms of a differential error distribution (no bias for Type I sentences). The fact that when no verb was present in the linguistic string no positional bias emerged in N+ patients adds further support to the hypothesis (reported in Chatterjee et al., 1999) that the verb is the element responsible for genesis of the spatial representation underlying a spoken sentence (no verb = no event, no thematic roles defined, no elements to build a spatial representation). Therefore, without a verb the three nouns given are just unrelated concepts without any unifying structure. Thus, the presence of a left-to-right “vector-for-action”, represented by the verb and claimed by Chatterjee et al. (1999) to be the basis for the representation of events expressed by means of language, seems to be indispensable for manifestation of the positional bias exhibited by N+ patients in spoken language processing. Apart from the impossibility of building a structured representation of the “event”, this could also mean there is a greater demand on memory resources to carry out this task, thus explaining the lower performance of the patients tested. 4.4. Correlational results The low correlations between the experimental tasks and the measures of neglect, with the exception of the overall index of neglect severity and the Passive linguistic condition, require a final comment. First of all, it is known that neglect is not a unitary syndrome; for this reason, its diagnosis requires the use of more than one test since a patient may show neglect in one test but not in another. This clinical evidence indicates the existence of multiple spatial dimensions that can be separately or jointly affected by the neglect syndrome, dimensions which are not necessarily covered completely by the existing clinical tests, such as the ones used in our research. Additionally, it must be considered that while the tasks used to identify neglect deal with explicit visuo-spatial stimuli, the effect involved by the emphatic stress, inserted in different strings of acoustic signals, depends on a spatial transcoding of auditory traces. This transcoding, although effective in the group of neglect patients, may not be activated in all individuals. These individual differences in a related field of investigation have been described in research on the “number-line”, where a significant proportion of normal subjects did not show
564
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
the activation of spatial transcoding in different kinds of numerical processing (Dehaene, Bossini, & Giraux, 1993, for instance, found the number-line effects to be significant in students of scientific, but not always of literary studies). It may be suggested that although the presence of a spatial bias in neglect patients influences the activation of a “spatial language-line”, the use of spatial transcoding is associated more with individual differences and less with the intensity of neglect disorders.
5. General discussion Our data show that in brain-damaged patients an interaction exists between the presence of neglect and the type of task administered. The task consisted of comparing the stress positions in two heard sentences, a comparison, which directly involved the acoustic traces left by the spoken sentences in the working memory. When they were given two sequences of words to compare, patients with unilateral neglect made more errors when the cue (the stress) appeared in the initial part than in the central or the final part of the sentence. This type of effect exerted by neglect on the task employed is strongly suggestive of a spatial transcoding of the auditory traces left in the working memory by the heard sentence, as already postulated in our previous study. Our data from the previous and the present study suggest that spoken language may activate a “spatial language line’, which is a spatial transcoding of linguistic input. This transcoding is then perturbed in a spatially predictable way by the presence of a spatial bias such as the heminattentional syndrome. With respect to our previous work, we are now able to better define the conditions required for the effect to take place. The effect is not found for all auditory input, but only in specific circumstances. In fact, we found that neglect had an effect on the linguistic task in the active/passive condition, but not in the musical or in the three-noun condition. We also found that the effect was inverted in the positive condition, the active /passive condition, when passing from the active to the passive sentence structure. These results mainly suggest the following: (1) neglect interacts with auditory linguistic input; e.g. a sequence of musical notes, although forming a “musical sentence”, does not produce the same spatial bias; (2) neglect interacts with a syntactically structured auditory linguistic input (agent/action/recipient of action), and not with just any auditory linguistic input (noun /noun/ noun); (3) the verb has a primary role in determining the presence of a structured spatial representation in linguistic input, defining the thematic roles and providing the frame for the event narrated in the sentence; (4) neglect acts at the level of the deep structure of the sentence, which in the case of active sentences coincides with surface structures. More exactly, we hypothesize that neglect acts on the spatial transcoding of the deep structure of linguistic input; the deep structure is a representation of “who does
something to whom”, and in its spatial transcoding the logical subject –the agent of the action, the “who”—consequently appears “on the left” (in both active and passive sentences). We see, therefore, that the absence of a structure underlying linguistic input, determined by the absence of the verb, was a critical factor, and this absence characterized both the three-noun and the musical condition employed in this study in which the neglect bias was not observed. In the musical condition, it can be objected that the stimuli employed were pieces of real music, with an actual underlying structure. However, since none of the subjects tested in our study had had any previous musical education, it is highly unlikely they could have perceived the underlying structure. It is possible that a different result would have emerged with professional musicians; but, unfortunately, very few musicians with neglect are described in the literature (we know of only one case reported by Landi et al., 1997). In the case of the three-noun sequence, instead, stimuli were actually processed as linguistic and had a semantic content (each noun is a concept). But, also here, semantic content per se did not seem to be enough for the neglect bias to appear. Semantic content must, in fact, be engraved into a carrying structure, defined by the verb, which establishes which concept acts as agent and which as recipient of action within the whole sentence frame. To conclude, we would like to present some considerations about the entity of the bias exerted by neglect on spoken language. As estimated from the active sentences, the bias was not present in every N+ patient tested but in 8/12 patients, and in a different strength in each one. The fact that the effect of neglect on spoken language was not universal conforms well to results obtained by other authors who investigated the relationship between language and space. Chatterjee, for instance, admits that in some of his experiments with normal subjects the spatial bias was evaluated only in the limited percentage of subjects who constantly placed the agent of the action on the same side of the page (in one experiment, only 37% of the subjects) (Chatterjee et al., 1999), and that the others were not included in the analysis. Also, Coslett (1999) found his hemifield effects only in patients with a specific lobe involvement (parietal lobe, independently of the presence of neglect); but, here too, not every patient exhibited these effects. Leaving the language domain, other authors who investigated the spatial transcoding of numerical quantities (the number-line) did not find this relationship to be ubiquitous (Dehaene et al., 1993). A. Kim too found the number-line in only 63% of the subjects tested (A. Kim, personal communication). This spatial transcoding of numerical quantities, i.e., the number-line, represents a good piece of evidence showing, as in the study we described, how spatial transcoding of a different domain (numbers) can be affected by the presence of a spatial deficit, i.e., neglect, undergoing a shift towards the ipsilesional side when this syndrome is present (Zorzi, Priftis, & Umilt`a, 2002; Vuilleumier, Ortigue, & Brugger, 2004).
M.C. Rinaldi, L. Pizzamiglio / Neuropsychologia 44 (2006) 556–565
Acknowledgement This research has been supported by grant of Ministero Italiano della Sanit`a. References Albert, M. L. (1973). A simple test of visual neglect. Neurology, 23, 658–664. Barbut, D., & Gazzaniga, M. S. (1987). Disturbances in conceptual space involving language and speech. Brain, 110(6), 1487–1496. Baxter, D. M., & Warrington, E. K. (1983). Neglect disgraphia. Journal of Neurology, Neurosurgery and Psychiatry, 46, 1073–1078. Behrens, S. J. (1985). The perception of stress and lateralization of prosody. Brain and Language, 26, 332–348. Bloomstein, S., & Cooper, W. (1974). Hemispheric processing of intonation contours. Cortex, 10, 146–158. Caramazza, A., & Hillis, A. E. (1990). Levels of representation, coordinate frames, and unilateral neglect. Cognitive Neurospychology, 7, 391–445. Chatterjee, A., Maher, L., & Heilman, K. (1995). Spatial charachteristics of thematic role representation. Neuropsychologia, 33, 643–648. Chatterjee, A., Maher, L. M., Gonzalez Rothi, L. J., & Heilman, K. M. (1995). Asyntactic thematic role assignment: The use of a temporalspatial strategy. Brain and Language, 49, 125–139. Chatterjee, A., Southwood, M. H., & Basilico, D. (1999). Verbs, events and spatial representations. Neuropsychologia, 37, 395–402. Coslett, H. B. (1999). Spatial influences on motor and language function. Neuropsychologia, 37, 695–706. Coslett, H. B., Schwartz, M. F., Goldberg, G., Haas, D., & Perkins, J. (1993). Multimodal hemispatial deficits after left hemisphere stroke. Brain, 116, 527–554. Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental representation of parity and number magnitude. Journal of Experimental Psychology, 122, 371–396. Diller, L., Weinberg, J., Piasetsk, E., Ruckdeschel-Hibbard, M., Egelko, S., Scotzin, et al. (1974). Methods for the evaluation and treatment of the visual perceptual difficulties of right brain damaged individuals.
565
New York University Medical Center, supplement of the 8th Annual Workshop for Rehabilitation Professionals. Istituto di Linguistica Computazionale del CNR di Pisa (1989). Barcellona Corpus di Italiano Scritto Contemporaneo. Unpublished manuscript. Landi, F., Zuccala, G., Cocchi, A., Bernabei, R., Tafani, A., & Carbonin, P. (1997). Playing piano in visuo-spatial neglect: A case study. Journal of Neurology, Neurosurgery and Psychiatry, 62(5), 543– 544. Mandler, J. M. (1996). Preverbal representation and language. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space (pp. 365–384). Cambridge, MA: MIT Press, Cap. 9. Massironi, M., Antonucci, G., Pizzamiglio, L., Vitale, M. V., & Zoccolotti, P. (1988). The Wundt-Jastrow illusion in the study of spatial hemiinattention. Neuropsychologia, 26, 1661–1666. O’Keefe, J. (1996). The spatial preposition in English, vector grammar, and the cognitive map theory. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space (pp. 365–384). Cambridge, MA: MIT Press, Cap. 9. Orsini, A., Grossi, D., Capitani, E., Laiacona, M., Papagno, C., & Vallar, G. (1987). Verbal and spatial immediate memory span: Normative data from 1355 adults and 1112 children. Italian Journal of Neurological Sciences, 8, 539–548. Pizzamiglio, L., Antonucci, G., Guariglia, C., Judica, A., Montenero, P., Razzano, C., et al. (1992). Cognitive rehabilitation of the hemineglect disorders in chronic patients with unilateral brain damage. Journal of Clinical and Experimental Neuropsychology, 14, 901–923. Rinaldi, C., Marangolo, P., & Pizzamiglio, L. (2003). Between language and space. A cross-domain interaction. Neuroreport, 14(10), 1381–1383. Segalowitz, S. J., & Plantery, P. (1985). Music draws attention to the left and speech draws attention to the right. Brain and Cognition, 4(1), 1–6. Vuilleumier, P., Ortigue, S., & Brugger, P. (2004). The number space and neglect. Cortex, 40, 399–410. Zoccolotti, P., Antonucci, G., & Judica, A. (1992). Psychometric characteristics of two semi-structured scales for the functional evaluation of hemi-inattention in extra-personal and personal space. Neuropsychological Rehabilitation, 2, 179–191. Zorzi, M., Priftis, C., & Umilt`a, C. (2002). Brain damage: Neglect disrupts the mental number line. Nature, 417, 138–139.