Spared syntax and impaired spell-out: The case of prepositions

Journal of Neurolinguistics 23 (2010) 354–382

Contents lists available at ScienceDirect

Journal of Neurolinguistics journal homepage: www.elsevier.com/locate/ jneuroling

Spared syntax and impaired spell-out: The case of prepositions Simone Ma¨tzig, Judit Druks*, Ad Neeleman, Gordon Craig Division of Psychology and Language Sciences, Department of Linguistics, University College London, Chandler House, 2 Wakefield Street, London WC1N 1PF, United Kingdom

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 October 2009 Received in revised form 1 February 2010 Accepted 5 February 2010

The objective of the study was to identify the factors that determine the preservation/impairment of prepositions in aphasia. Five parameters derived from previous research (Bennis, Prins, & Vermeulen, 1983, Friederici, 1982; Grodzinsky, 1988; Kean, 1977, 1979; Kreindler & Miha˜ilescu, 1970) were examined in a sentence completion task and three types of grammaticality judgement tasks using four subcategories of prepositions with 18 preposition tokens in a large number of test sentences. Prepositions were found impaired in both Broca’s and anomic aphasia. Most of the parameters could not account for the data, and some data were in the opposite direction to the predicted. No disproportionate impairments of meaningless prepositions were found and prepositions with syntactic function were best preserved in the majority of patients. Patients made predominately within-category substitution errors. The results are interpreted as evidence for preserved syntactic knowledge about prepositions. It is suggested that a deficit at the post syntactic level of (late) spell-out is the underlying reason for the preposition deficit. Ó 2010 Elsevier Ltd. All rights reserved.

Keywords: Aphasia Agrammatism Grammatical morphemes Prepositions Spell-out

* Corresponding author. E-mail address: [email protected] (J. Druks). 0911-6044/$ – see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jneuroling.2010.02.002

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

355

1. Introduction Our current knowledge of the availability of prepositions in aphasia is limited. Only a few past studies focused on prepositions with the aim of identifying the underlying reasons for their impairment1, mainly in the 1980s and 1990s, and these studies often used linguistic constructs that today are outdated. Importantly, they have rarely been cited, reviewed, or related to each other. The paucity of research on prepositions is in sharp contrast with the interest aphasia researchers have in verbs, nounverb differences and verb inflections. We counted, for example, 38 papers that have been published during the last 25 years that report noun and verb production in picture naming only of over two hundred patients (see Ma¨tzig, Druks, Masterson, & Vigliocco, 2009). The neglect of prepositions in aphasia research is surprising because prepositions are a particularly interesting grammatical class to study. The reason is that prepositions share properties of both lexical and functional categories. This is reflected in the ongoing debate among linguists as to how to characterize them (Grimshaw, 2005; van Riemsdijk, 1990; Rizzi, 1985; Svenonius, 2004, 2007). Prepositions, like functional heads, are caseless, and do not combine with tense or aspect morphology (in English and in many other languages). Some prepositions do not receive stress and their fixed number indicates that, like pronouns and determiners, they belong to closed-class words. On the other hand, prepositions have lexical features too. All prepositions are like lexical heads in that they mark case and some prepositions assign theta-roles, and have rich meaning and stress. A subset of prepositions assign clearly defined theta-roles to their complements (e.g., spatial: in/on/under/at the table; temporal: in two weeks/in 1999/at 3 pm; benefactor: he bought flowers for her; recipient: he showed the map to his mother; instrument: he opened the letter with a knife; etc.). Subcategorized prepositions are somewhat different. While they assign theta-roles to their complements, these are semantically opaque (e.g., to rely on friends, to suffer from headache, to be interested in something) (Neeleman, 1997). Another subset of prepositions fulfils a purely syntactic function. These prepositions assign case to their complements but not theta-roles. The English of is such a meaningless preposition. Of is inserted in order to satisfy the Case Filter when a phrase consists of nouns or adjectives2 that cannot assign case (e.g., Sue’s pride of her daughter, Sue is proud of her daughter). The infinitival to is another grammatical morpheme that heads verbal phrases in the form of bare infinitives (Hyde, 2000). The infinitival to – just like the syntactic preposition of – fulfils a syntactic role in the sentence. Similar to of-insertion, to-insertion circumvents violation of the Case Filter by assigning case to PRO, the (empty) subject of the verb because a non-finite verb cannot assign case to its subject3. A third preposition that fulfils a grammatical role is by that heads the by-phrase in passive sentences. In passives, the subject of the active sentence surfaces in the by-phrase, but maintains the theta-role assigned by the verb in the active sentence4. It appears, therefore, that by functions as a case assigner only5.

1 Additional studies that report preposition deficits in aphasia and acquired dyslexia (e.g., Druks & Froud, 2002; Friederici, 1981, 1985; Friederici et al., 1982; Froud, 2001; Goodglass et al., 1970; Kemmerer, 2005; Kemmerer & Tranel, 2000, 2003; Kolk & Friederici, 1985; Leikin, 2002; Lonzi & Luzzatti, 1995; Mack, 1981; Miceli et al., 1989; Mondini, Luzzatti, Zonca, Pistarini, & Semenza, 2004; Morton & Patterson, 1987; de Roo, Kolk, & Hofstede, 2003; Ruigendijk, 2002; Saffran, Schwartz, & Marin, 1980; Schwartz, Saffran, & Marin, 1980; Smith, 1974; Tesak & Hummer, 1994; Tranel & Kemmerer, 2004; Wales & Kinsella, 1981) are not discussed in the present paper, because their focus of interest was different from ours in so far that they did not compare performance on different subcategories of prepositions. 2 Only NP–P–NP constructions were tested here because AP–P–NP may select not only of but other prepositions too (e.g., proud of/interested in/good at/dependent on/susceptible to). 3 Not all linguists agree that the infinitival to is a preposition. Some classify it as a complementizer (e.g., Postal & Pullum, 1978), an inflection (e.g., Chomsky, 1981), or a modal auxiliary (e.g., Mittwoch, 1990). However, it is widely acknowledged that, historically, the infinitival to is derived from the preposition to (e.g., Haspelmath, 1989) and has characteristics of a preposition. 4 The following examples show that by does not contribute to the meaning of its complement: Susan was investigated by the CIA (agent); the window was broken by the storm (cause); the bread cannot be cut by an ordinary knife ( instrument) (from Svenonius, 2004, p. 25). 5 This, however, may be disputed by some linguists who may claim that theta-role marking by the verb to the complement of by is impossible because a complement must c-command the head from which it receives theta-roles (e.g. Chomsky, 1981), and the NP in the by-phrase in passive sentences does not c-command the verb that is argued to theta-mark it. If true, by must function as theta-marker in passive sentences.

356

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

Finally, some prepositions, usually those expressing the spatial meaning of path or endpoint, combine with verbs to form a phrasal verb (Lindstromberg, 1997). Superficially, phrasal verbs and verbs with subcategorized prepositions appear similar (e.g., fill in the form; believe in the idea). It is possible, however, to distinguish them by using a linguistic diagnostic. A preposition functioning as particle bonds with the verb and is permitted to occur on either side of the complement NP (e.g., fill the form in/fill in the form). When this is not permissible (i.e., believe in the idea/*believe the idea in), it is an indication that the preposition bonds with the noun complement and is subcategorized by the verb (Bolinger, 1971). 2. Prepositions in aphasia According to traditional assumptions, agrammatic Broca’s aphasic patients omit and Wernicke’s aphasic patients wrongly use grammatical morphemes, including prepositions, and all grammatical morphemes are affected to a similar extent (e.g., Kean, 1979; Pick, 1913). The latter claim, possibly, led researchers to focus on one subcategory of preposition, spatial prepositions. The production of spatial prepositions (Friederici, 1981; Friederici, Scho¨nle, & Garrett, 1982; Morton & Patterson, 1987; Smith, 1974), their comprehension (Friederici, 1981; Friederici et al., 1982; Goodglass, Gleason, & Hyde, 1970; Morton & Patterson, 1987; Smith, 1974), and the ability of acting them out (Mack, 1981; Morton & Patterson, 1987; Smith, 1974) was tested during the 1970s and early 1980s. All these studies reported deficits related to prepositions, though the severity of the impairment was dependent on the modality of the task (comprehension being less impaired than production) and the type of aphasia. Different studies differed in their conclusions. While Friederici and colleagues (Friederici, 1981; Friederici et al., 1982) found Broca’s aphasic patients more impaired than Wernicke’s aphasic patients in both comprehension and production of spatial prepositions, other studies reported that in comprehension, Broca’s aphasic patients as a group performed better than Wernicke’s aphasic patients (Goodglass et al., 1970; Mack, 1981), and better than anomic type patients (Smith, 1974). When the heterogeneous status of prepositions and its interest for research were acknowledged, researchers tried to show that not all types of prepositions were equally impaired. Different parameters that distinguish between (better) preserved and (more) impaired subcategories of prepositions in Broca’s and Wernicke’s aphasia have been identified and tested. Kean (1977, 1979) argued that the underlying reason for agrammatic Broca’s aphasic patients’ deficits related to grammatical morphemes is phonological in nature. She proposed that the reason for the reduced speech production (and the syntactic comprehension deficits) of agrammatic Broca’s aphasic patients is that they cannot produce (and perceive) unstressed morphemes, that is, the bulk of grammatical morphemes. Since the case of prepositions is special because some prepositions are like content words (i.e., two-syllabic and stressed: a’bove) and others are like function words (i.e., onesyllabic and unstressed: in), Kean’s theory predicts that producing unstressed, one-syllabic prepositions would be more difficult than producing two-syllabic prepositions that receive stress. ilescu as early as 1970 were interested in the effects of frequency on the use of Kreindler and Miha prepositions. They compared the production of prepositions in large samples of connected speech of 10 Romanian speaking aphasic patients and 10 non-brain-damaged control participants. They found that the rate of prepositions used by the patients was reduced in comparison to the controls (9.5 vs.13.1% of the total ilescu determined the number of words produced). In the absence of frequency counts, Kreindler and Miha frequency of individual prepositions by the frequency of their occurrence in the controls’ speech. They found that those prepositions that occurred frequently in the controls’ speech were more likely also to be ilescu’s used by the aphasic patients than those used less frequently by the controls. Kreindler and Miha results thus suggest that frequency of use of prepositions has an effect on their availability in aphasia. ilescu’s sample were prepositions Among the best preserved prepositions in Kreindler and Miha with syntactic function. It is well known that syntactic prepositions are of particular high frequency. For example, in Francis and Kucera’s database (1982), of, to, and by are among the 22 most frequent words in (written) English6. Most likely this is also true for other languages. Therefore, if frequency

6 Admittedly, Francis and Kucˇera’s database does not distinguish between the different usages of polysemic prepositions. Hence, the frequency value given for a preposition token such as of or by is the sum of the frequencies of their different usages.

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

357

indeed has an effect, syntactic prepositions are at an advantage. The problem, however, is that frequency and function of a preposition are inherently confounded. Therefore, it is difficult to tell if the relative preservation of syntactic prepositions is the effect of prepositional subcategory or frequency. From the 1980s onward the underlying deficit in agrammatism was thought to be syntactic by most researchers, and this assumption guided the formulation of research questions in later studies of prepositions. Friederici (1982) compared the availability of (meaningful) spatial prepositions and (meaningless) subcategorized prepositions in a sentence completion task, and found that Broca’s aphasic patients were more impaired at producing meaningless subcategorized prepositions. The majority of errors made were omissions and across-category substitutions. Friederici argued that because of an underlying syntactic deficit Broca’s aphasic patients are unable to assign syntactic structure during production. This impairment affects ‘syntactic’ (i.e., subcategorized) prepositions to a larger extent than ‘semantic’ (i.e., meaningful) prepositions. Wernicke’s aphasic patients in contrast performed better with subcategorized prepositions, and tended to make within-category substitution errors. This suggested to Friederici that the impairment of these patients is semantic. The problem with Friederici’s argumentation, however, is that subcategorized prepositions are not considered syntactic by all linguists. According to Neeleman (1997), they are theta-role assigners, though the semantic identity of the theta-role assigned is not transparent. If this is true, Friederici’s conclusion is based on the misclassification of subcategorized prepositions. Nevertheless, on the basis of these data (and other data unrelated to prepositions) Rizzi (1985; see also Ouhalla, 1993) made the generalization for agrammatism that linguistic units that assign/receive theta-roles, including (meaningful) prepositions (and verbs and nouns), are preserved and linguistic units that do not assign/ receive theta-roles (determiners, syntactic prepositions, etc.) are impaired. Bennis, Prins, and Vermeulen (1983) re-examined Friederici’s conclusion on the basis of linguistic criteria of the Extended Standard Theory (Chomsky, 1972; Jackendoff, 1972) that distinguish between prepositions that are lexically inserted (lexical prepositions) and prepositions whose insertion depends on the syntactic configuration of the sentence (syntactic prepositions). Among the lexical prepositions they included spatial and subcategorized prepositions and among the syntactic prepositions, the syntactic of and dative to. The authors found that Broca’s aphasic patients were better at producing lexical prepositions (that also included subcategorized prepositions that were found impaired in Friederici, 1982) than syntactic prepositions. The opposite pattern was found for Wernicke’s aphasic patients. Bennis et al. therefore concluded that Broca’s aphasic patients had a deficit in syntax-based processing with relatively well-preserved lexical processing, and Wernicke’s aphasic patients presented with the opposite pattern of impairment. A potential problem of the Bennis et al. study is related to the classification of the dative to which is controversial. Some argue, and this is the position of Bennis et al., that the dative to has a syntactic function only and is meaningless. According to this view, she gave me the cat is derived by dative shift from she gave the cat to me, a syntactic operation that changes only structure not meaning, similar to passivization (e.g., Larson, 1988). However, there are alternative views according to which the dative to is lexical and meaningful (see for example, Levin, 1993, who argues that the difference is due to give that may change its transitivity; and Neeleman, 1996, 1997, who shows that the dative to behaves very differently from other syntactic prepositions, such as of). If these arguments are correct, to was wrongly included by Bennis et al. among the syntactic prepositions. Another attempt to explain the impairment of prepositions in agrammatic Broca’s aphasia was that of Grodzinsky (1988) who used the linguistic constructs of government (Chomsky, 1981) to account for selective preposition deficits. According to his hypothesis, prepositions that are governed by the verb are impaired while prepositions that are ungoverned by the verb are spared. The distinction relates to whether a prepositional phrase (PP) is intricately related to the meaning of the verb, as in the case of arguments, or only accidentally, as in the case of adjuncts. Ungoverned PPs – adjuncts – are always optional, that is, the sentence is well-formed without the prepositional phrase (e.g., it was raining [on Sunday]). Governed PPs – arguments – on the other hand, are often obligatory (e.g., Sue relies [on her friend]), and sometimes optional (e.g., Adam stole £10 [from the old man]). Prepositional phrases headed by subcategorized prepositions (and particles) are always governed, and are always obligatory (e.g., *Sue relies [on her friend], *Sue filled [in the form]). Other preposition types, spatial prepositions, for example, can be either governed obligatory arguments (e.g., the key is [in the pocket]), governed

358

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

optional arguments (e.g., she sent a postcard [from Spain]) or ungoverned adjuncts of the verb (e.g. the team played hockey [in the hall]). Grodzinsky tested the effects of linguistic government on the availability of prepositions in aphasia by comparing the passive by (adjunct, ungoverned) with that of subcategorized prepositions (arguments, governed; e.g., my friend was very interested [in Jane]/Joe hopes [for a happy ending]) in a grammaticality judgement task. He found that Broca’s aphasic patients made more errors on sentences with governed than with ungoverned prepositions. Grodzinsky interpreted this to support his theory even though the evidence (present only in rejecting the ungrammatical sentences but not in accepting the grammatical sentences) was limited. There was another problem. In Grodzinsky’s stimuli the governed/ungoverned distinction overlapped with another distinction between lexical meaningless (subcategorized), and (depending on the linguistic analysis of the passive by, see above) syntactic meaningless or meaningful (passive by) prepositions. A few subsequent studies re-examined the governed/ungoverned distinction and its effect on the availability of prepositions in aphasia. Some supported Grodzinsky’s hypothesis (Lonzi & Luzzatti, 1995) others did not (Druks, 1991; Tesak & Hummer, 1994). Recently, Lonzi and colleagues (Lonzi, Luzzatti, & Vitolo, 2007) deviated from the trend of accounting for the preposition deficit in terms of syntax, and employed Optimality Theory (OT), a phonological theory, instead, to do so. Their claim is that while the pronunciation of grammatical morphemes in aphasia is regulated by the same principles as in healthy speakers, in the case of aphasia, some of these principles are re-ranked and as consequence they become unduly prominent, resulting in ungrammaticality. In particular, they refer to the Telegraph Constraint (Chomsky, 1981; Pesetsky, 1998), according to which function words are not to be pronounced. The Telegraph Constraint has low prominence in normality but becomes operational in aphasia resulting in the omitting of grammatical morphemes, including prepositions. What is omitted and what is maintained is constrained by the Recoverability Condition, according to which grammatical morphemes must be pronounced, unless their deletion can be recovered by a local antecedent. This divides prepositions into recoverable prepositions which are predicted to be omitted (e.g., subcategorized prepositions) and unrecoverable prepositions which are predicted to be preserved (e.g., syntactic and some meaningful prepositions). Lonzi et al. found that subcategorized prepositions were often omitted, and meaningful non-spatial prepositions and syntactic prepositions tended to be produced, in support of their predictions. Spatial prepositions, both recoverable and unrecoverable, were an exception in so far that they were both similarly impaired. This anomaly required additional stipulation in the form of ‘supplementary licensing conditions’ (p. 291) according to which all spatial prepositions are recoverable because all spatial prepositions can be predicted from a (non-linguistic) context due to ‘the obligatoriness of location for any material action’ (p. 291). The necessity for an extra stipulation, however, weakens the status of the Recoverability Condition (and Lonzi et al.’s account) that becomes either a grammatical or a pragmatic principle: in the case of non-spatial prepositions, recoverability functions as a grammatical principle, however, in the case of spatial prepositions, as a pragmatic principle. Finally, it is notable that the categorisation of the prepositions as suggested by Lonzi and colleagues largely overlaps with that of Grodzinsky, albeit on different grounds. Recoverable prepositions are all governed, and unrecoverable prepositions are all ungoverned (though with respect to spatial prepositions, the two theories make different claims. According to Grodzinsky, only governed spatial prepositions are impaired, Lonzi et al., however, argued that all spatial prepositions are recoverable and, therefore, all are prone to omission or substitution). Since Lonzi et al.’s predictions are largely overlapping with that of Grodzinsky, their parameter will not be further considered. To summarize, previous research resulted in the following conclusions: (a) phonologically stressed (longer) prepositions are better preserved than unstressed (shorter) ones; (b) high frequency prepositions are better preserved than low frequency prepositions; (c) meaningful prepositions are better preserved than meaningless prepositions; (d) lexical prepositions are better preserved than syntactic prepositions; and (e) ungoverned prepositions are better preserved than governed prepositions. Previous studies have a number of shortcomings: they did not test all preposition subcategories; used few exemplars of each subcategory; and most importantly, the classification of prepositions in the materials of some of the studies was unsafe. To improve upon the weaknesses evident in previous

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

359

studies, the present study tested the production and grammaticality judgement of all subcategories of prepositions7 (meaningful, subcategorized, syntactic prepositions and particles) with more exemplars of each type, and in classifying prepositions, we took into account the controversies related to their analyses in current linguistic theory. The objectives of the present study are threefold. Our first objective was to re-evaluate the parameters that were identified in past studies to account for the impairment/preservation of prepositions. Since we used more subcategories of prepositions (and more exemplars of each type) we were in a position to improve upon their tests. For example, when we tested the effects of meaningfulness on the availability of prepositions, instead of confining the comparison to spatial vs. subcategorized prepositions as Friederici (1982) did, the availability of a wide range of meaningful prepositions (spatial, temporal, benefactor, instrumental, source, goal, etc.) was compared with a range of meaningless prepositions (the infinitival to, the syntactic of). We similarly increased the number of preposition types that we included when we tested the effects of government by comparing many more governed and ungoverned prepositions than Grodzinsky (1988) did, and of lexicality (Bennis et al., 1983) by comparing all types of lexical prepositions with all types of syntactic prepositions. We are, therefore, in a better position than the original studies to assess the validity of their parameters. Our second objective was to reformulate the parameters identified by previous studies in conjunction with predominantly Broca’s aphasic patients, so that they make specific hypotheses not only for the Broca’s aphasic patients but also for the anomic patient in the present study. We did so by taking into consideration the traditionally held views that Broca’s aphasic patients have an underlying syntactic impairment (e.g., Berndt & Caramazza, 1980) and anomic aphasic patients, an underlying lexical impairment (e.g., Kay & Ellis, 1987). Our third objective was to test the hypothesis that the underlying reason for the preposition deficit in both Broca’s and anomic aphasia is not syntactic, and to put forward an alternative theoretical account. 2.1. The hypotheses The hypotheses directly derived from the parameters identified in previous studies are the following: (1) Since Broca’s aphasic patients are known to have phonological impairments, they are expected to be affected by the length of a preposition; in contrast, the length of a preposition should have no effect on anomic aphasic patients. (2) As lexical access is known to be influenced by frequency (see e.g., Forster & Chambers, 1973; Segui, Mehler, Frauenfelder, & Morton, 1982), and as anomic patients are assumed to have lexical impairments, frequency of a preposition is predicted to affect their performance, while it should have no effect on the performance of Broca’s aphasic patients. (3) As neither Broca’s nor anomic aphasic patients have semantic deficits, meaningfulness of a preposition is expected to be beneficial for both patient types. (4) Broca’s aphasic patients, having syntactic deficits, will be more impaired on syntactic prepositions, and anomic patients, having lexical deficits, on lexical prepositions. (5) Having syntactic deficits also predicts that Broca’s aphasic patients will be more impaired on governed than ungoverned prepositions. Government is expected to have no effect on the performance of anomic aphasic patients. Two more hypotheses were newly formulated for the present study. The first one, hypothesis (6), is a broad hypothesis related to all prepositions in Broca’s and anomic aphasia: since deficits related to closed-class words are associated with Broca’s aphasia, while anomic patients have predominantly noun deficits, it is predicted that the Broca’s aphasic patients will perform poorly on all prepositions, while anomic patients will not present with marked difficulties for prepositions. The second hypothesis (7) is derived from close reading of the research that repeatedly found that the preposition deficit is not confined to Broca’s aphasic patients (and by implications to syntactic impairments), but occurs in patients of all types of aphasia (see e.g., Bennis et al., 1983, for Broca’s and Wernicke’s aphasia; Goodglass et al., 1970, for conduction and global aphasia; Kemmerer, 2005,

7

With the exception of prepositional adverbials (e.g., she put the knife down).

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

360

Table 1 Demographical information about patients and matched control participants. Patient

Age

Years of education

Sex

Handedness

Lesion site

Matched control

Age

Years of education

Sex

BG

53

15

F

Right

SS

57

17

F

DC DOR

54 62

20.5 9.5

M M

Right Right

GC DGR

57 62

20 15

M M

EW

64

16

M

Left/ambidexter

WM

65

17

M

Right

No pathological findings evident in scan Not available Fronto-temporo-parieto-occipital extending subcortically, left Fronto-temporal (involving insula), left and fronto-occipital, right Not available

CM

36

20

F

Mean

55.4

17.8

–

TH

36

14

F

Mean

53.8

15

–

for transcortical motor aphasia; Leikin, 2002, for transcortical sensory aphasia; Smith, 1974, for anomic aphasia). These findings contradict hypothesis (6), and may also imply that the performance patterns predicted in hypotheses (4) and (5) would not be supported. A further motivation for hypothesis (7) is the prevalence of within-category substitution errors in Broca’s aphasic patients reported in some past studies (Bennis et al., 1983; Froud, 2001; Kemmerer & Tranel, 2000; Leikin, 2002; Lonzi et al., 2007; Morton & Patterson, 1987; Smith, 1974) that is likely to indicate preserved syntactic knowledge for prepositions, and preserved parsing. Consequently, this may imply that the source of the preposition deficit is not syntactic in nature not only in the anomic but also in the Broca’s aphasic patients. Thus, the finding of preposition deficits in anomic (and Wernicke’s) aphasic patients that are believed to have relatively well-preserved syntax,8 and the finding of within-category substitution errors in Broca’s aphasic patients that are said to have impaired syntax suggested hypothesis (7), that the preposition deficit in Broca’s and anomic aphasia is not due to failure of syntactic operations, but due to a failure to select the correct preposition after syntax, where ‘phonological factors [..] interact with syntactic ones’ as it has been argued as early as 1981 by Wales and Kinsella (p. 306). 3. Method 3.1. Case studies Five aphasic participants, four Broca’s type and one anomic patient (TH), and five healthy control participants matched for age and education participated in the study. Demographic information about the patients and control participants is given in Table 1. The Boston Diagnostic Aphasia Examination (BDAE, Kaplan, Goodglass, & Weintraub, 1983) and the Boston Naming Test (BNT, Kaplan, Goodglass, & Weintraub, 1978) were administered in order to obtain a general clinical profile for each aphasic patient. Background tests included the subtest cube analysis from the Visual Object and Space Perception Battery (VOSP, Warrington, 1991), copying of the Rey Complex Figure (Rey, 1941), the (three picture version of the) Pyramids and Palm Tree Test (PPT, Howard & Patterson, 1992), the Object and Action Naming battery (OAN, Druks & Masterson, 2000), the Noun and Verb Comprehension Test (NVC, Masterson & Druks, unpublished), and the Syntactic Comprehension Test (SCT, Froud & Druks, unpublished). Finally, for each aphasic and control participant samples of connected speech were recorded and transcribed. The results of the background tests are in Table 2. A summary of the patients’ clinical profile, including speech samples, is in Table 2 and Appendix I shows that DC, DOR and TH have naming deficits (TH, the anomic patient has the severest naming deficit), DC and DOR have syntactic comprehension deficits, BG, DC and DOR have reduced proportions of verbs in connected speech, all patients except BG have reduced proportions of prepositions, BG, DC, DOR and EW tend to omit or substitute grammatical

8 In contrast to agrammatism (and paragrammatism) in Broca’s aphasia, paragrammatism in Wernicke’s aphasia is not usually attributed to syntactic impairments (e.g., Jakobson, 1964).

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

361

Table 2 Results from neuropsychological background testing. BG

DC

DOR

EW

TH

Norms/control performance

Cube test

Test scores

9

10

9

1

9

–

Copy of complex Rey figure

Test scores

36

36

32

36

36

28  3

Pyramids and palm trees

Test scores

52

49

49

50

35

49–52

Object and action naming

Objects % Actions %

99 95

69 48

77 69

92 92

30 40

– –

Noun and verb comprehension

Nouns % Verbs %

97 95

94 88

97 95

98 98

87 84

–

Syntactic comprehension

Reversible %

100 100 100 100 100 100 100 100 90 90 100

70 15 90 80 30 80 90 10 60 50 83

70 45 60 60 60 60 80 10 90 50 80

100 95 90 100 70 100 100 100 90 100 93

95 95 100 100 85 100 100 100 70 90 93

– – – – – – – – – – –

878 .12 .07 .26

649 .03 .04 .52

452 .08 .04 .42

546 .10 .11 .36

887 .10 .13 .37

614–2476a .12–.17b .08–.14b .18–.43b

Irreversible %

Active Passive Truncated active Subject cleft Object cleft Active possible Active impossible Passive possible Passive impossible Truncated passive

Filler sentences % Description of spatial situations in connected speech a b

No of words produced Prepositions Verbs Auxiliaries/modals

Proportions correct Proportions correct Proportions correct

The range of words produced by controls. The normal range for proportions was calculated as two standard deviations below and above the controls’ mean.

morphemes, and TH performs below the norm on the Pyramids and Palm Trees test. The speech samples of all the Broca’s type patients show features of agrammatism such as short ungrammatical sentences, omission (or substitution) of grammatical morphemes, omission of lexical verbs, and problems with tense. The anomic patient’s speech consists of more elaborate sentence structures and features of mild paragrammatism. 3.2. Materials Four tasks, a sentence completion and three grammaticality judgement tasks were administered to evaluate the parameters put forward by previous studies to account for the differential impairment of different preposition types (i.e., phonology, Kean, 1977, 1979; frequency, Kreindler & Miha˜ilescu, 1970; meaningfulness, Friederici, 1982; Rizzi, 1985; lexicality, Bennis et al., 1983; and government, Grodzinsky, 1988). In total, 18 different preposition tokens were targeted. The prepositions in the study were either one or two syllables long (multiword prepositions were excluded). A list of all preposition tokens, their subcategories, the theta-roles they assign, and the number of times they were probed in each task is in Appendix II. 3.3. The preposition types employed in the different comparisons The effect of phonology was tested by comparing performance on two-syllabic and one-syllabic prepositions. In order to test the effects of frequency, frequency ratings were taken from the CELEX frequency database (Baayen, Piepenbrock, & Gulikers, 1995), and the preposition tokens used were

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

362

Table 3 The parameters, their predictions and prepositional subcategories used in the comparisons of the present study. Parameters

Source

Prediction

Comparisons

(i)

Phonology

Kean

two-syllabic prepositions vs. one-syllabic prepositions

(ii)

Frequency

Kreindler & Mih ailescu

stressed and longer > unstressed and shorter prepositions high frequency > medium frequency > low frequency prepositions

(iii)

Meaningfulness

Friederici/Rizzi/Ouhalla

meaningful > meaningless prepositions

(iv)

Lexicality

Bennis et al.

(v)

Government

Grodzinsky

lexical > syntactic prepositions ungoverned (adjunct) prepositions > governed (argument) prepositions

syntactic of, infinitival to vs. polysemic prepositions vs. non-polysemic spatial and temporal prepositions (a) spatial, temporal, ‘other theta-role assigning’ prepositions vs. infinitival to, syntactic of (b) spatial, temporal, ‘other theta-role assigning’ prepositions vs. subcategorized prepositions, particles subcategorized prepositions, particles vs. infinitival to, syntactic of ungoverned meaningful prepositions vs. governed meaningful prepositions

classified into high, medium and low frequency prepositions. In order to find out the effects of meaning, we compared all meaningful prepositions with prepositions with syntactic roles only. Crucially, we did not include subcategorized prepositions among the meaningless (syntactic) prepositions (as Friederici did) because of their controversial status. A second comparison to test the effects of meaningfulness followed the view that subcategorized prepositions are lexical (Neeleman, 1997): instead of comparing meaningful lexical versus meaningless syntactic prepositions, we contrasted meaningful and meaningless lexical prepositions. To test the effect of lexicality, and to constrain Bennis et al.’s (1983) comparison, we excluded meaningful prepositions from the comparison and compared only meaningless lexical prepositions (e.g., subcategorized prepositions, following Neeleman) with meaningless syntactic prepositions. As for testing Grodzinsky’s theory, instead of comparing the passive by with subcategorized prepositions, the effects of linguistic government were explored within one subcategory by contrasting governed and ungoverned meaningful prepositions. The parameters, the predictions they make and the preposition types used in their evaluation are summarized in Table 3.

3.4. Methodological considerations In both the production and grammaticality judgement tasks, data were analyzed in terms of accuracy. In the production tasks, no responses, substitutions, and alternative responses9 were considered as errors. Responding with synonymous prepositions, however, was not counted an error. A preposition was considered synonymous, if the four students involved in the stimulus rating10 used the target and its synonym interchangeably. This was the case for eight target prepositions (above 4 over, behind 4 after, besides 4 next to, into / in/into, onto / on/onto, under 4 below/underneath, to 4 towards (spatial) and (temporal) to / to/until). In the sentence completion task, only first responses entered the analysis. Control participants’ accuracy and latency and reaction time data were collected for all tasks.

9 Alternative responses are non-target responses, which, albeit grammatical, did not include the target preposition (e.g., Sue will arrive home late at night / tomorrow/in the night). They were considered errors, because they often were not single prepositions but multiword responses. These were rare errors. Most importantly, none of the controls produced them. 10 All sentences were presented to four students of UCL, whose mother tongue was English, for completion. A sentence was included in the test battery only if at least three out of four students provided the target preposition.

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

363

3.5. Statistical analysis Logistic regression of single case data with linear contrasts was applied to test specific hypotheses about differential impairments of subcategories of prepositions. The log odds ratios used in the regression analyses were calculated using the least impaired subcategory as reference category. Thus, the log odds ratio for each subcategory reflects the differential impairment of that category for a patient. The following example illustrates the statistical procedure used to test whether the predicted order of impairment has been obtained in the data or not: Friederici (1982) predicted that Broca’s aphasic patients perform better on meaningful (e.g., spatial, temporal, and ‘other theta-role assigning’ prepositions) than meaningless prepositions (e.g., the syntactic of, the infinitival to). The prediction was tested using PROC LOGISTIC procedure in SAS 9.1 statistical software (SAS Institute Inc., Cary, NC, USA) by performing a contrast where the log odds ratios involving meaningful prepositions, weighted for the number of subcategories (i.e., three subcategories of meaningful prepositions in this example were given coefficients of 2), are compared with the weighted log odds ratios involving meaningless prepositions (i.e., two subcategories in this example were given coefficients of 3). The results reveal if the data ‘behaved’ as predicted, that is, if there were indeed, proportionately more errors made on meaningless prepositions than meaningful prepositions. The alpha level was set at p ¼ .01 to correct for multiple comparisons. 4. The sentence completion task 4.1. Materials The following subcategories of preposition were included in sentence completion. Five prepositions with (homophonic) spatial and temporal meanings (at, from, in, on, to), five predominantly spatial prepositions (above, behind, into, onto, under), and five predominately temporal prepositions (after, before, for, since, until) were included. There were 48 sentence frames eliciting spatial prepositions and 60 eliciting temporal prepositions. Five prepositions assigning thematic roles other than spatial or temporal (‘other theta-role assigning’ prepositions) were also included (by, for, from, to, with). They assign theta-roles of benefactor (for the baby), recipient (to his father), instrument (he unscrewed the jar by hand), manner/degree (with care, with pleasure), substance (with sand), animate source (from John), comitative function (with a friend) or agent (the book was written by Jane Austen). There were 48 sentence frames eliciting prepositions assigning these theta-roles. Five subcategorized prepositions that are homophonic with spatial/temporal prepositions (at, from, in, on, to), one homophonic temporal preposition (for) and two non-spatial and non-temporal prepositions (of, with) were included. There were 48 sentence frames eliciting subcategorized prepositions. 24 sentence frames elicited the syntactic preposition of and 24 the infinitival to. Finally, two particles (on, in) homophonic with spatial and temporal prepositions were selected. Each sentence frame eliciting a particle was presented twice, once with the particle attached to the verb (he turned on the radio) and once with the particle and verb separated (he turned the radio on). There were 24 sentence frames eliciting particles, 12 of each type. Appendix II provides a complete list of the preposition tokens used, their subcategory, the theta-role they assign and the number of times they were probed. For each preposition a sentence was constructed which required the insertion of a preposition. In 193 items the presentation of the sentence frame sufficed for eliciting the missing target, in 26 items the sentence frame was accompanied by a picture, and in 57 items it was introduced by a lead-in sentence. Sentences were presented in written form and a line indicated the position of the missing preposition (see examples below). no context: John left _____ Wednesday. context: John is a very active child. He cannot sit still _____ the table. The 276 target sentences were administered in six individual blocks with 46 target sentences per block. The sentences were presented in pre-determined semi-randomized order with no more than three homophonic prepositions in a row. A separate practice block with 14 practice sentences was prepared to familiarize participants with the task.

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

364

picture:

The book is _____ the table.

4.2. Procedure Written sentence frames were presented on a laptop screen using PowerPoint. Pictures were presented above the sentences. Each sentence was read by the participants or, if necessary, by the experimenter and patients could ask for the sentence to be repeated. Participants were required to complete the sentences by producing the missing target words. The written stimulus was continuously available on the screen and there was no time limit to complete the sentence.

4.3. Results Table 4 gives the results in terms of both accuracy and latencies of the control participants. Since the controls performed well across all sentence types with no differences in relation to the subcategories of prepositions, their results are not discussed further. Table 5 gives the total number of correct responses and the number and type of errors produced by individual patients. Fig. 1 illustrates the individual patient’s results with respect to the five parameters tested. Table 5 shows that some patients made relatively few errors (BG: 23%, EW: 24%) while others made many (TH: 45%, DC: 66%, DOR: 82%). The majority of errors were within-category substitutions. Only DOR produced an equal number of no responses and within-category substitutions. This, however, must be considered with caution due to DOR’s tendency to perseverate on four prepositions, by, from, before, and for, which could account for the high number of within-category substitution errors. DOR’s behaviour illustrates the general inclination of most patients to use certain prepositions as substitutes for others. BG used of and in as replacement prepositions, TH preferred in as a substitute, and DC used mostly for, from, and on. Fig. 1 shows that only few parameters affected the patients’ performance: BG was better at high frequency prepositions than medium frequency and low frequency prepositions (see parameter (ii): trend [Wald c2 ¼ 4.00, df ¼ 1, p ¼ .05]). Syntactic prepositions were better preserved than meaningful lexical prepositions (see parameter (iii, a): BG: trend [Wald c2 ¼ 4.93, df ¼ 1, p ¼ .03]), however, meaningful lexical prepositions were better preserved than meaningless lexical prepositions (see parameter (iii, b): BG: trend [Wald c2 ¼ 6.07, df ¼ 1, p ¼ .014]; DOR: trend [Wald c2 ¼ 5.00, df ¼ 1, p ¼ .025]). Syntactic prepositions were produced more accurately than meaningless lexical prepositions (see parameter (iv): BG: [Wald c2 ¼ 10.47, df ¼ 1, p ¼ .001]; TH: [Wald c2 ¼ 10.44, df ¼ 1, p ¼ .001]; EW: trend [Wald c2 ¼ 5.27, df ¼ 1, p ¼ .02]). DOR’s performance was also significantly influenced by government in that he performed better when a meaningful preposition was governed by the verb (see parameter (v): [Wald c2 ¼ 10.98, df ¼ 1, p ¼ .0009]). None of the other parameters explained the patients’ performance patterns. 5. Grammaticality judgement tasks 5.1. Materials Three grammaticality judgement tasks were used: grammaticality judgement of single sentences, contrastive sentence pairs and forced choice judgement. The three tasks posed different task demands. It was expected that forced choice is the most difficult judgement task.

No. of Proportions/latencies for correct responses. prepositions correct (n ¼ 1360) /mean latency for correct responses Accuracy 1269 Latency 808

Spatial Other Temporal Subcategorized Syntactic Infinitival Particles One-syllabic Two-syllabic High Medium Low Governed Ungoverned theta-role of to frequency frequency frequency assigning

.91 803

.98 779

.90 856

.93 791

.93 808

.99 846

.90 748

.95 808

.88 810

.96 832

.94 800

.88 818

.89 791

.93 833

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

Table 4 Accuracy (proportions) and latencies (in milliseconds) of the control group in the sentence completion task.

365

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

366

Table 5 The overall number of prepositions correctly produced and the number and types of errors made by individual patients in the sentence completion task.

BG DC DOR EW TH

No. of prepositions correct (n ¼ 276)

Number and type of errors No responses

Across-category substitutions

Within-category substitutions

212 94 49 209 152

3 17 104 12 24

7 43 13 10 19

54 122 110 45 81

179 sentences, a subset of the items used in the production task were included in the three grammaticality judgement tasks. Items in the production task that required a picture (n ¼ 26) were not included. Also, in order to reduce the number of items in the tasks, 24 non-polysemic temporal prepositions, 29 ‘other theta-role assigning’ prepositions, six subcategorized prepositions, and 12 sentences eliciting a particle in a construction in which the particle is attached to the verb were further removed. For the grammaticality judgement of single sentences and sentence pairs, for each of the sentences an ungrammatical sentence was created which differed from the grammatical one only in the preposition used (see examples below). The erroneous prepositions were randomly selected from all the prepositions used in the judgement tasks. Particle violations were created by substitution with another particle (e.g., out, up). In the forced choice grammaticality judgement, for each of the 179 sentences two distracters were randomly selected from all the prepositions used in the judgement tasks (one of them was taken from the ungrammatical utterances used in grammaticality judgement of sentence pairs).

Fig. 1. Individual patient’s results in sentence completion with respect to the five parameters (length, frequency, meaningfulness, lexicality, government) tested (marked with stripes are significant results of p  .01). Mean correct performance of the group is indicated by a dot (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

spatial prepositions temporal prepositions ‘other theta-role assigning’ prepositions subcategorized prepositions particles syntactic of infinitival to

367

Jeremy is currently at/*to work. Sue will arrive home late at/*of night. Erin sold the ring to/*on her friend. James is late but Helen will wait for/*at him. Sandra filled in/*at the form. Allan mourned the loss of/*in his brother. Babies scream to/*at get attention.

Stimulus sentences in all grammaticality judgement tasks were spoken by a female native speaker of English in a slowed down but still natural rate and were recorded using Audacity (http://audacity. sourceforge.net/). The tasks were programmed using Visual Basics (Microsoft). 5.2. Procedure In grammaticality judgement of single sentences, participants were presented with a sentence in the upper half of a laptop screen. After a delay of 1000 ms the sentence was read aloud and highlighted by a coloured frame. After spoken presentation of the sentence, yes? and no? buttons were activated on the screen and participants were asked to judge the grammaticality of the sentence by pressing one of the buttons using a touch screen. There was a time limit of 10 s in which a response had to be made, after which the screen went blank. The 358 test sentences were distributed over eight blocks with 45 items in six and 44 items in two blocks. The number of grammatical and ungrammatical sentences in each block varied between 20 and 24. The grammatical and ungrammatical version of a sentence never occurred within the same block. Within blocks, all sentences were arranged in a pre-determined semi-random order with no more than three yes or no responses consecutively. A block of 28 practice sentences was administered to familiarize the participants with the task. In contrastive grammaticality judgement of sentence pairs, participants were presented with two sentences of identical structure that differed only in the prepositions used. Participants had to identify the grammatical sentence. The first sentence of a pair was presented in writing in the upper half of a laptop screen. After a delay of 1000 ms the sentence was read aloud and highlighted by a coloured frame. Another 1000 ms later, the second sentence appeared in the lower half of the screen. 1000 ms later, the sentence was read aloud and highlighted. At this stage, response buttons were activated. The buttons, placed underneath each sentence, were labeled with good?. Participants selected a sentence by pressing the corresponding good? button using the touch screen. Selecting one of the sentences triggered the appearance of a blank screen displaying only the next button. Pressing next started a new trial. After a time limit of 10 s, if no response was made, the screen went blank. The 179 sentence pairs were presented in three blocks with 45 trials and one block with 44 trials. Presentation was in a predetermined semi-random order with no more than three grammatical/ungrammatical sentences consecutively in the same position. 14 practice sentences were administered prior to testing proper. In the forced choice task participants were presented with a sentence frame with a missing preposition and a choice of three prepositions, only one of which correctly completed the sentence frame. The task was to select the preposition that correctly completed the sentence. The sentence frame was presented in the upper centre of a laptop screen. The missing word was represented by a line. 1000 ms following the appearance of the written sentence frame, it was read out aloud and highlighted by a coloured frame. For the spoken presentation of a sentence frame, the target word was cut out from the sound wave and a gap was inserted lasting for 2 s. The 2 s consisted of a .5 s silence followed by a 1s sine noise of 330.00 Hz and an amplitude of .5 and another .5 s silence. On the lower part of the screen, three buttons were displayed in a row each reading this one?. 1000 ms following the spoken sentence frame presentation, the first preposition that may complete the gap appeared in writing above the leftmost button, and 1000 ms later it was read aloud. 1000 ms later, the second preposition appeared above the middle button followed by its spoken presentation 1000 ms later. The same procedure applied to the third preposition. All this one? buttons were activated once the third preposition was

368

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

presented aurally. Participants responded by pressing one of the this one? buttons using a touch screen. There was a time limit of 10 s within which participants had to respond otherwise the screen went blank. 14 practice sentences were administered prior to testing proper. 5.3. Results Table 6a–c give the results in terms of accuracy and latencies of the control participants. These tables show that controls were at ceiling on this task and even the reaction times did not distinguish between the different prepositions types. This is often the case in language tasks in which control participants perform at ease and with no errors. Table 7 gives the total number of correct responses and the number and type of errors produced by individual patients in each grammaticality judgement task, and Fig. 2 illustrates individual patients’ results with respect to the five parameters tested for each grammaticality judgement task. Table 7 shows that two out of the three grammatical judgement tasks were too easy for some of the patients too. BG made only a few errors across tasks (accuracy ranged from 92 to 98% correct) while most patients made many errors in some but not all tasks (DC’s accuracy ranged from 72 to 91% correct, DOR’s from 39 to 90% correct, EW’s from 55 to 92% correct, and TH’s from 77 to 89% correct). Most errors were made in forced choice grammaticality judgement. Only DC and TH made as many errors in grammaticality judgement of single sentences as in forced choice grammaticality judgement. The majority of errors across tasks were erroneous responses. However, more no responses than erroneous responses were made by BG, DC, and EW in forced choice grammaticality judgement and by DC in contrastive grammaticality judgement of sentence pairs. Fig. 2 shows that only some of the parameters could account for the performance patterns of the patients. DC, DOR, and TH showed a negative frequency effect (see parameter (ii) in forced choice grammaticality judgement: DOR: trend [Wald c2 ¼ 4.98, df ¼ 1, p ¼ .03]11), in contrastive grammaticality judgement (DOR: [Wald c2 ¼ 6.62, df ¼ 1, p ¼ .01]; DC: trend [Wald c2 ¼ 4.93, df ¼ 1, p ¼ .0312), and in grammaticality judgement of single sentences (TH: trend [Wald c2 ¼ 3.85, df ¼ 1, p ¼ .05]). DOR performed significantly better on (meaningful and meaningless) lexical prepositions than syntactic prepositions (see parameter (iii, a): [Wald c2 ¼ 7.70, df ¼ 1, p ¼ .006] and parameter (iv): [Wald c2 ¼ 7.88 df ¼ 1, p ¼ .005]) in force choice grammaticality judgement. Since DOR did not show a difference between meaningless and meaningful lexical prepositions (see parameter (iii, b), it seems that his performance was not influenced by meaningfulness as such. BG, DC, and TH performed better on meaningful than meaningless lexical prepositions (see parameter (iii, b) in grammaticality judgement of single sentences: DC: [Wald c2 ¼ 7.41, df ¼ 1, p ¼ .007]; TH: [Wald c2 ¼ 13.8, df ¼ 1, p ¼ .0002])), and in forced choice grammaticality judgement (TH: [Wald c2 ¼ 9.68, df ¼ 1, p ¼ .002]; BG: trend [Wald c2 ¼ 6.46, df ¼ 1, p ¼ .011]). EW and TH performed better on syntactic than meaningless lexical prepositions (see parameter (iv) in contrastive grammaticality judgement: EW: trend [Wald c2 ¼ 4.19, df ¼ 1, p ¼ .04]) and grammaticality judgement of single sentences (TH: trend [Wald c2 ¼ 4.90, df ¼ 1, p ¼ .03]). None of the other parameters accounted for the patients’ performance patterns. 6. Discussion The aim of this study was to identify the factors that contribute to the preservation and impairment of prepositions. To achieve this, our own data of sentence completion and grammaticality judgement were used to test the parameters made in previous studies to account for the performance patterns of aphasic patients in relation to prepositions. These parameters constituted the basis for our own hypotheses. They were reformulated so that they make specific predictions for Broca’s and anomic

11 Statistical calculations on DOR’s original data were impossible due to ceiling effects for one of the three categories (low frequency prepositions). In this case, an error was added (to avoid ceiling performance) in order to carry out the statistical analysis. After doing so, the comparison approached significance. 12 DC performed at ceiling on low frequency prepositions. In order to be able to carry out a statistical comparison, the procedure described in footnote 11 was applied.

Table 6a Accuracy (proportions) and latencies (in milliseconds) of the control group in the grammaticality judgement of single sentences. No. of prepositions correct (n ¼ 895) and mean reaction times (in ms) for correct responses in grammatical condition Accuracy 886 RTs 901

No. of prepositions Proportions/reaction times correct for. correct (n ¼ 895) Spatial Other Temporal Subcategorized Syntactic Infinitival Particles High Medium Low Governed Ungoverned and mean reaction theta-role of to frequency frequency frequency times (in ms) for assigning correct responses in ungrammatical condition 889 843

.99 834

1 891

.99 859

1 865

.98 882

1 830

.98 1029

.99 856

.99 878

1 929

.98 870

.99 843

Accuracy RTs

No. of prepositions correct (n ¼ 895) and mean reaction times (in ms) for correct responses

Proportions/reaction times correct for. Spatial

Other theta-role assigning

Temporal

Subcategorized

Syntactic of

Infinitival to

Particles

High frequency

Medium frequency

Low frequency

Governed

Ungoverned

892 728

.99 723

1 772

1 714

1 732

.98 683

1 719

1 788

.99 701

1 735

1 768

1 725

1 737

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

Table 6b Accuracy (proportions) and latencies (in milliseconds) of the control group in the grammaticality judgement of sentence pairs.

Table 6c Accuracy (proportions) and latencies (in milliseconds) of the control group in the forced choice grammaticality judgement. No. of prepositions Proportions correct for. correct (n ¼ 895) Spatial Other Temporal Subcategorized Syntactic of Infinitival to Particles High Medium Low Governed Ungoverned and mean reaction meaningful frequency frequency frequency times (in ms) for correct responses .99 1048

1 1033

1 957

1 938

.98 1020

.99 866

1 1112

.99 943

.99 973

1 1074

.99 1029

1 999

369

Accuracy 891 RTs 999

370

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

Table 7 The number of correct judgements and the number and types of errors made by individual patients in the three grammaticality judgement tasks. BG

DC

DOR

EW

TH

344 7 7

263 16 79

290 9 59

329 4 25

277 23 58

Contrastive grammaticality judgement of sentence pairs No. of prepositions correct (n ¼ 179) 176 No. responses 1 Erroneous responses 2

162 11 6

161 5 13

163 0 16

159 1 19

Forced choice grammaticality judgement No. of prepositions correct (n ¼ 179) No. responses Erroneous responses

129 35 15

69 15 95

99 74 6

142 8 29

Grammaticality judgement of single sentences No. of prepositions correct (n ¼ 358) No. responses Erroneous responses

165 13 1

aphasic patients in the present study. Below we discuss the results in relation to our hypotheses (see Table 8 for a summary of the results). The first parameter proposed that phonological properties of a preposition determine its preservation/impairment (Kean, 1977, 1979). According to hypothesis (1), therefore, it was expected that the Broca’s aphasic patients in the study would perform poorly on one-syllabic but not on two-syllabic prepositions in sentence completion13. For the anomic patient no difference was predicted. There was no support for this hypothesis for any of the Broca’s aphasic patients: their performance was not influenced by length. This was also true for the anomic patient (in her case, in line with the hypothesis). Since frequency is known to be an important psycholinguistic factor that affects lexical retrieval (see e.g., Segui et al., 1982), and previous research has demonstrated its effect on the production of prepositions (Kreindler & Miha˜ilescu, 1970), the possibility that the frequency of prepositions would contribute to their preservation/impairment in aphasia was considered in the second parameter. Hypothesis (2), therefore, predicted that the anomic aphasic patient, due to her presupposed lexical deficit, would show frequency effects in preposition use. Broca’s aphasic patients, because their underlying deficit is believed not to be lexical, would not be affected by frequency. Although it was somewhat unexpected to find frequency effects in grammaticality judgement, frequency was shown to be a significant factor for DOR in contrastive grammaticality judgement (and forced choice grammaticality judgement), albeit in the opposite direction to the predicted. Thus, less frequently occurring prepositions were judged more accurately. The interpretation of effects of frequency, however, is difficult because frequency differences coincide with prepositional subcategories. The effect found for DOR may mean that he was impaired in judging the grammaticality of sentences with syntactic prepositions, while he performed better in sentences with (non-polysemic) meaningful prepositions that belong to the least frequently used prepositions. This is in line with his general performance pattern of performing somewhat better on meaningful prepositions. For the rest of the patients, no frequency effects were found in any of the tasks. The third parameter claimed that meaningfulness facilitates the production of prepositions in Broca’s aphasia (e.g., Friederici, 1982; Friederici et al., 1982). Hypothesis (3), therefore predicted that since both Broca’s and anomic aphasic patients have no semantic deficits, meaningfulness will be beneficial for both. We found, however, that meaningfulness did not affect performance of any of the patients in sentence completion, though in some of the grammaticality judgement tasks DC, DOR, and TH, the anomic patient, performed as predicted. The sporadic presence of the effect in grammaticality judgement in the case of the two most impaired Broca’s aphasic patients suggest that meaning might have a role in comprehension, and/or in tasks in which responses have to be made within a short time frame (as in all the grammaticality judgement tasks). As for the anomic patient, she had severe short

13

The parameter related to phonological properties was not considered applicable to grammaticality judgement.

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

371

Fig. 2. Individual patient’s results in the grammaticality judgement tasks with respect to the four parameters (frequency, meaningfulness, lexicality and government) tested (marked with stripes are significant results of p  .01). Mean correct performance of the group is indicated by a dot (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

372

Table 8 The results for individual patients in relation to the predictions of the parameters in previous studies. Parameters

BG

Sentence completion Judgement of single sentences Contrastive judgement of sentence pairs Forced choice judgement

DC

Sentence completion Judgement of single sentences Contrastive judgement of sentence pairs Forced choice judgement

(i) Phonology (ii) Frequency (iii) Meaningfulness (iv) Lexicality (v) Government

Sentence completion Judgement of single sentences DOR Contrastive judgement of sentence pairs Forced choice judgement

EW

Sentence completion Judgement of single sentences Contrastive judgement of sentence pairs Forced choice judgement

TH

Sentence completion Judgement of single sentences Contrastive judgement of sentence pairs Forced choice judgement

–

þ

– – þ

þ

þ þ

þ

Parameters significantly supported are marked þ. Parameters significantly rejected are marked –.

term memory problems which may explain her reliance on the meaning of prepositions when judging the grammaticality of the sentences. However, since meaningfulness did not consistently facilitate performance, and in particular not in the more demanding sentence completion task, it may be concluded that the contribution of meaningfulness of prepositions to their production and judgement is limited. According to the fourth parameter, (Bennis et al., 1983) syntactic prepositions are expected to be impaired in the Broca’s aphasic patients, while lexical prepositions (meaningful and meaningless) are expected to be spared. Hypothesis (4) predicted the opposite pattern for the anomic patient. The results showed that DOR’s performance in forced choice grammaticality judgement was as predicted, however, BG and TH, two relatively high level patients showed the opposite pattern, that is, better performance for syntactic than lexical prepositions in sentence completion. This is according to what is predicted for TH, the anomic patient, but not for BG, a Broca’s aphasic patient. Thus, hypothesis (4) was supported by the anomic patient in production but not by any of the other patients (except DOR, in grammaticality judgement) and was contradicted by BG. The fifth parameter referred to the effects of government on preposition use in Broca’s aphasia. This is related to their syntactic impairments (Grodzinsky, 1988). Accordingly, hypothesis (5) predicted that government affects the availability of prepositions in Broca’s but not anomic aphasic patients, who have no syntactic deficits. This was rejected because none of the Broca’s aphasic patients of the present study performed better on ungoverned than governed prepositions (DOR was affected by government in the opposite direction, in sentence completion). It seems, therefore, that the use of prepositions by aphasic patients is not affected by syntactic government. On the basis of the results related to the first five hypotheses we may conclude that (i) none of the parameters obtained overall support from all the patients, or from an individual patient in all tasks, though meaning and lexicality had some effect on some patients and in some tasks, and (ii) the use of prepositions was impaired in both Broca’s and anomic aphasic patients and that, contrary to

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

373

hypotheses (1), (2), (4), and (5), the performance patterns of Broca’s aphasic patients and the anomic aphasic patient did not differ, rejecting hypothesis (6). Admittedly, the interpretation of the results of the present study should take into consideration the relative small number of patients studied, and the variability in their overall language abilities and their performance on the experimental tasks. Both these factors have the potential to reduce our confidence in the force of our conclusions. As for the number of subjects, we compared patient numbers in our study and in previous studies of prepositions. Friederici (1982) who claimed that meaningfulness is an important factor worked with the largest number of patients (12 Broca’s aphasics and 12 fluent type aphasics); Bennis et al. (1983) who suggested that lexical prepositions are the best preserved ones worked with seven Broca’s aphasics and seven anomic patients; Grodzinsky who tested the hypothesis of government worked with four Broca’s aphasic and four fluent type patients; Druks (1991) who found that longer prepositions are better preserved than shorter prepositions based her conclusions on a single patient; and finally, the effects of frequency of prepositions were tested on 10 patients with unspecified forms of aphasia by Kreindler and Miha˜ilescu (1970). Thus the existing data base of research on the availability of prepositions in aphasia in general, and in support of the specific hypotheses, is not large, especially, if we consider that previous research studied fewer types of prepositions, and probed fewer examplars of each, in smaller scale studies. The strength of the present study is its comprehensive exploration of prepositions and the large number of stimuli it used in each condition. Nevertheless, our conclusions here will have to be further validated with larger number of patients in future studies. Variability across patients of similar clinical profile is often reported in aphasia research. For example, Miceli, Silveri, Romani, and Caramazza (1989), who studied the use of different free and bound grammatical morphemes (including prepositions) in the spontaneous speech of a group of agrammatic patients, found a great deal of variability across patients, across grammatical morphemes, and in the pattern of their omission and substitution. The patients in the present study presented with similar variable performance. In particular, two of the Broca’s aphasic patients were more severely impaired and two of them considerably less so. In order to deal with this form of variability, we assume that if a certain performance pattern is predicted to be present in Broca’s aphasia (impairment of syntactic prepositions, for example), one would expect to observe this in high level Broca’s aphasic patients too, albeit to a lesser extent than in the case of more severely impaired Broca’s aphasic patients. Despite the variability across patients, however, in the present study, all the patients, irrespective of clinical diagnosis or severity, performed strikingly similarly in terms of error types: the majority of errors of all the patients were within-category substitution errors. Omission errors were rare in comparison. This finding is important for our arguments in support of hypothesis (7). In support for hypothesis (7), we present four arguments to make the claim that the antecedent of the preposition deficit is not syntactic. (i) In order to correctly produce a preposition in a sentence completion task the patient has to parse the sentence, that is, to assign syntactic structure to it. This process is generally believed to be disrupted in Broca’s aphasia. However, the majority of errors in the present study were errors of substitution, and in substituting, patients usually maintained word category. This implies that they were able to syntactically parse the sentences, identify the word class that is missing, and retrieve, if not the right preposition, another preposition (see Friederici, 1982, who interpreted the prevalence of within-category substitutions in Wernicke’s aphasic patients as indication of their preserved syntactic knowledge; see also Bennis et al., 1983). (ii) The omission of syntactic prepositions in speech and sentence completion was in the past taken as indication of Broca’s aphasic patients’ syntactic deficit (see Bennis et al., 1983; Friederici, 1982). However, the patients in the present study performed relatively well on syntactic prepositions. (iii) More generally, in connected speech (which is not reported here but see Ma¨tzig, 2009, and speech samples in the Appendix III), word order errors were very rare, and, when present, they mostly involved the misplacing of an adverbial. Crucially, patients never placed prepositions in illegal positions in a sentence, and thus demonstrated awareness that prepositions are heads of phrases and that they take complements. They also showed sensitivity for the syntactic property of case assignment by prepositions (and verbs) in that there were no case violations of pronouns in connected speech (see

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

374

Appendix III for speech examples). (iv) Finally, the Broca’s and anomic aphasic patients in the present study behaved similarly in relation to prepositions: they were neither influenced by government, nor length, and syntactic prepositions in comparison to other prepositions were relatively well preserved in most patients. All patients produced mostly within-category substitution errors. Thus, the Broca’s aphasic patients performed in relation to preposition as it is expected of patients with a lexical type deficit. Our study, therefore, suggests that the preposition deficit is not due to syntactic impairments (even in patients with general syntactic impairments), and that preposition deficits cannot be used as evidence for syntactic impairments in agrammatic Broca’s aphasia, as it had been argued previously (e.g., Bennis et al., 1983; Friederici, 1982; Grodzinsky, 1988). Our results are compatible with hypothesis (7) which holds that the preposition deficit in the Broca’s and anomic aphasic patients is due to a failure to select the correct preposition after syntax at spell-out. The following two sections explain how spellout operates and outlines evidence in favour of the view that the preposition deficit takes place at this late stage of the derivation. 7. Spell-out We follow the framework of late vocabulary insertion (Halle & Marantz, 1993, henceforth ‘spell-out’) in the Minimalist Program (Chomsky, 1995), which assumes that terminal nodes in a syntactic tree consist of feature bundles (e.g., [pronoun PERSON: 3rd, NUMBER: sing, GENDER: male, CASE: nom]), not lexical material (e.g., he) and that the features contain morpho-syntactic and semantic but not phonological information. In order to mediate between syntax and phonology, the grammar uses a set of correspondence rules – spell-out rules – that specify that a syntactic terminal X corresponds to phonological form Y (e.g., Halle & Marantz, 1993; Jackendoff, 1997). While the spell-out system is now a well-established component of the grammar, and its operation has been worked out in some detail for pronouns (e.g., Harley & Ritter, 2002) and verb inflections (Chomsky, 1995), its operation in relation to prepositions has not yet been fully worked out. Therefore, the following account of the decomposition of prepositions is still speculative, though there is a wealth of work on the semantics of prepositions, which can be used for speculating about the likely featural make-up of prepositions (see e.g., Svenonius, 2004, 2007; Tyler & Evans, 2003). If we assume that prepositions are syntactically encoded through a set of features, each preposition can be conceived of as a feature bundle that will partially overlap with other prepositions (i.e., those with partially shared meaning/function). For example, all prepositions would have the feature of [P] (preposition), most spatial prepositions would have the feature of [PLACE], all directional prepositions would also have the feature of [PATH], and so on. Below is a potential example of possible syntactic feature bundles for of, in, into, and locational and directional under (see (1)). [P], [PATH] and [PLACE] are features that are part of a preposition’s structural description, [IN], [TO], and [UNDER] are values that represent a preposition’s distinct properties such as containment (for the locational in), path (for directional prepositions like into and to), contact (for the spatial on), and so on. (1) Syntactic features [P] (‘of’) [P PLACE: IN] (‘in’) [P PATH: TO, PLACE: IN] (‘into’) [P PLACE: UNDER] (locational ‘under’) [P PATH: TO, PLACE: UNDER] (directional ‘under’) It may be speculated that these syntactic feature bundles are spelled out as illustrated in (2). (2) Spell-out rules [P] 4 /of/ [P PLACE: IN] 4 /in/ [P PATH: TO, PLACE: IN] 4 /into/ [P PLACE: UNDER] 4 /under/

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

375

The spell-out rules in (2) show that locational and directional under receive the same phonological form, because the spell-out rule that realizes them is underspecified for the feature [PATH]. This analysis maintains that the two prepositions are syntactically and semantically distinct, but have the same phonological form. Evidence for it comes from languages such as German in which homophonic prepositions assign different types of overt case: for example, locational in and under assign dative case and directional in and under assign accusative case. This shows that two prepositions are syntactically distinct despite having the same phonological form. According to this analysis, there is a great deal of featural overlap in the characterization of individual prepositions. For example, any preposition could be spelled out as /of/ as all prepositions have in their specification the feature [P]. Similarly, the directional into could be realized as /in/ as a result of the shared feature [PLACE: IN]. The phonological realization of prepositions via the spell-out rules is, therefore, regulated (as in the case of pronouns and verb inflections, presumably) by the Subset Principle and the Elsewhere Principle. The Subset Principle ensures that the set of competing spell-out rules for a target morpheme contains only those spell-out rules that share in their specification a subset of the target’s syntactic features (e.g., Halle, 1997; Halle & Marantz, 1993). For example, if the preposition to be pronounced is into, then all spell-out rules that contain the features [P], [PLACE] and/or [PATH] and the values IN and/or TO will be included in the subset. Spell-out rules which contain features or values (i.e., [UNDER]) that are not shared by the target will not be included in the subset (see (3)). (3) Subset of competing spell-out rules for the phonological realization of ‘into’ [P] 4 /of/ [P PLACE: IN] 4 /in/ [P PATH: TO, PLACE: IN] 4 /into/ The Elsewhere Principle guaranties that the most specific spell-out rule is selected from the subset of competing candidates. It has the effect that into will not be realized as /of/ or /in/ as there is another, more specific spell-out rule that better matches the syntactic features of into namely the one that mentions all required features [P PATH: TO, PLACE: IN]. 8. Impaired selection at spell-out What effect does a deficit in spell-out have on preposition use in aphasia? In order to produce a preposition (say, into), the Subset Principle must be applied, that is, those spell-out rules must be identified whose structural description mentions the exact features of the target morpheme or a subset of them (e.g., the spell-out rules for /of/, /in/, and /into/). Then, in order to ensure that the target preposition is produced, the most specific spell-out rule must be selected through application of the Elsewhere Principle. An overall deficit in accessing the syntactic features of prepositions would cause omission errors or unrelated or across-category substitution errors. Within-category substitution errors, which were the majority of errors for the patients in this study, show that access to the syntactic features of prepositions, crucially, the feature P, is possible, and, therefore, the deficit is likely to be in applying the Subset Principle or the Elsewhere Principle. A deficit in selecting an appropriate subset may result in making substitution errors that are more specific than the target (e.g., in / into). The reason is that the subset selected would wrongly include into in the set to produce in, in which case the Elsewhere Principle would (appropriately) select into as the most specific preposition of the competing subset. An alternative deficit, a deficit of the Elsewhere Principle, would result in substitution errors that are less specific than the target (e.g., into / in; in / of). The reason is that the malfunctioning Elsewhere Principle fails to select the most specific spell-out rule from the subset of competitor spell-out rules. Clearly much more work still has to be carried out both in terms of linguistic analysis and experimental work with aphasic patients. As soon as the linguistic analysis provides a thorough decompositional analysis of prepositions and their corresponding spell-out rules, the quality of the patients’ within-category substitution errors will be more fully understood. Errors that are consistently in one direction (i.e., errors that are either consistently more or consistently less specific) confirm that the

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

376

within-category substitution errors are not simply random due to semantic proximity but an indication of impaired application of either the Subset Principle or the Elsewhere Principle, and would constitute supporting evidence for hypothesis (7), that the preposition deficit is due to impaired selection at spellout. Since a fully worked out analysis of prepositions in terms of featural decomposition and spell-out is not yet available (though work in this direction is on its way), it is not possible at this stage to distinguish among deficits in the application of the Subset Principle and deficits in the application of the Elsewhere Principle. One way of exploring this issue in the present data is however to consider syntactic prepositions. Prepositions such as of have the least specific set of syntactic features and are therefore most likely to be included in the set of competing spell-out rules initially selected for all prepositions. This means that they are also more likely to be used as a replacement for other prepositions in case the Elsewhere Principle is not applied properly, while they are themselves less likely to be replaced. Thus, they are candidates for patients using them as default preposition. BG indeed uses of in this way. Abstracting away from such speculative details, the hypothesis that problems with spell-out underlie the behaviour found in our patients is promising in that it is able to reconcile demonstrated knowledge of syntax with impaired use of prepositions and it opens up new avenues for research not only in relation prepositions but also in relation to other grammatical morphemes. Appendix I Case studies of the patients participating in the study BG suffered a subarachnoid haemorrhage in 2000. Initially, BG was diagnosed as non-fluent agrammatic Broca’s aphasic without dysarthria. Eight years post-onset, her language has recovered largely which is reflected in her mixed profile on the BDAE now. BG is still rated as Broca’s aphasic in terms of articulatory agility, paraphasia, word finding and auditory comprehension. However, her phrase length, melodic line and repetition are in line with anomic aphasia. Her use of grammatical form is better than that of typical Broca’s aphasics but still not within the normal range and she still shows agrammatic features in speech such as omissions/substitutions of grammatical morphemes and reduced proportion of verbs. Her proportion of correctly produced prepositions in connected speech is within the normal range, her syntactic comprehension and object and action naming are good. She has unimpaired spatial processing and semantics. Below is an example of BG’s description of a set of pictures (Dinner Party). Her description is adequate, however it has numerous features of agrammatism such as omission of grammatical morphemes, tense errors, and verb omissions (e.g., he is write(ing) it down /the woman (is) cooking/she (is) stir(ing) the food/(in) the back of the frontal picture). a man calls another man to invite him for a dinner/the recipients of the party.he is write it down the information on a piece of paper with a pen/the woman and man are cooking in their kitchen/the man doing the washing up/the woman cooking on top of the stove/she stir the food in the sauce pan/the back of the frontal picture is a fish/this is Smith/the woman and man are in the dining room/they set the table in the ./plate and knife and fork and glasses and candles and two bottles of wine and the fish and potatoes/there a dog under the table/you can see his tail and back legs/the Smiths are in their bedroom getting ready for the party/the woman have long dress/she doing her necklace to do it up/the man have tie and suits/he doing up his tie/they ’re in the hall now greeting their guests/the woman have big bunch of flowers with her hand/the man shaking the other man hands/I now see/they’re in the dining room/just coming the dining room they saw the fish is not there/they all then are horrified/the Smith man look two fingers one eye to look what happened/Mrs. Smith is crying/she’s have her hands over her face/the woman guest pampering her while Mr. Smith running out the door/the other picture says he’s gone and buy fish and chips for guests and himself/now they’re sitting/they finished dinner in dining room/ they’re quite happy now/you can see the cat happy as well/this cat licking his paw with his tongue

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

377

DC suffered a subarachnoid haemorrhage in 2002. The haemorrhage left him with severe language problems without dysarthria. Six years post-onset, DC presents with a mixed, predominately, Broca’s aphasic’s profile on the BDAE. He has word finding difficulties, auditory comprehension deficits and he uses only simple grammatical constructions, in line with Broca’s aphasia. Yet, his BDAE scores on melodic line, phrase length, articulatory agility and repetition are more in line with anomic aphasia. DC produces many paraphasias and his body part identification and comprehension of commands is poorer than what is expected of both Broca’s aphasic and anomic aphasic patients. He has severe difficulties in naming objects in the BNT (40% correct). DC’s speech is effortful, non-fluent and agrammatic characterized by reduced verb and preposition production and many omissions/substitutions of lexical and grammatical morphemes. DC shows verb deficits in production and comprehension and has a severe comprehension deficit for syntactically complex sentences such as passives and object cleft sentences. His spatial and semantic-associative processing is intact. Below is an example of DC’s description of a set of pictures (Dinner Party). He prefers to use direct speech (e.g., ‘‘I’ll see you in about an hour or six hours’’/‘‘how do you do’’) and proper, instead of common nouns (e.g., Mickey and Jane), and he omits grammatical morphemes and lexical verbs (e.g., (the) men and women they are married/(he) put the phone down/the man and woman (are) starting to (prepare) the dinner and (the) fish/the kitten is (smelling/looking at) the fish). it’s four guys/two men and women/men and women they are married/that’s two marriages/two of them is phoning them/six o’clock in the evening/how do you want dinner/this is the guy making out/I’ll see you in about an hour or six hours/put the phone down/the man and woman starting to the dinner and fish/the kitten is the fish/he is cooking/the man is drying the plates/ that’s Mickey and June/it is lighting the stars/two wine and potatoes and fish/that’s four dining places/is ready for the cooking/the cat is slowly creeping under this table/this is the bedroom/ Jane’s black/she is: come on let’s go/the white here/that is fantastic/I’m going to the dining room/ Mickey and Jane and: how do you do/Phil/How do you do Phil/Sue, God Sue fantastic/flowers for you/Jane is fantastic/suddenly the dining room/where is the fish/Mickey and Jane: It’s gone/ where is it gone/Sue and Phil said where is the fish/Jane said: Come on Sue, is alright/Mickey sort of getting fish and chips/he’s got the fish and chips/got four fish/Mickey and Jane and Sue and Phil have got the fish and chips/how do the cat/cattle eaten the fish DOR has a long history of neurological impairments staring in 1986. In 2007, a structural scan was obtained. The brain scan showed a large left hemisphere lesion involving the frontal, parietal, temporal, and occipital lobes extending subcortically. 22 years post-onset, DOR presents with a mixed diagnostic profile on the BDAE. Subtests tapping semantic and lexical knowledge identified him as Broca’s aphasic, but his scores on utterance length, melodic line, and articulatory agility were better than that of typical Broca’s aphasics. DOR’s object naming is severely impaired (40% correct on the BNT) and he produces effortful, non-fluent, non-dysarthric, but agrammatic speech with only few verbs and function words including prepositions and a large number of nouns, adverbs and adjectives in copula constructions. DOR’s syntactic comprehension is severely impaired. He is severely dyslexic and dysgraphic. He has unimpaired spatial processing and semantics. Below is an example of DOR’s description of a set of pictures (Dinner Party). He tends to use many copula constructions (e.g., it’s loads and loads of pots/it’s a candle/it’s a bottle/it’s two candles/it’s like a dining room) in place of lexical verbs, and omits grammatical morphemes (e.g., the two blokes (are) on the phone/he’s writing (it) down). he’s on the phone/the two blokes on the phone/he’s writing down/he’s in the kitchen and lady and man and a bloody fish/it’s crazy/it’s loads and loads of pots/in the washing up is a bloke/it’s definitely a large fish/with salt and pepper and pots and plates/and small plates and large plates/ it’s a candle/it’s a bottle/it’s two candles/that’s four candles/like a feast/dog/I’m not sure if it’s a dog or a cat/maybe a dog/napkins and a gravy boat and a fish/it’s a large fish/I don’t like fish/it’s like a dining room/not a going out to dinner/very posh/it just maybe a dinner/it’s two ladies and two men/it’s a bunch of flowers here/it’s definite to dinner/it’s gone/any money like the dog’s taken it/dog ate it/it’s the cat/it’s licking the lips/fish and chips/is very sad/two blokes and two ladies in fish and chips/now fish and chips

378

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

EW was involved in a bike accident in 2000 (wearing no helmet) in which he contracted a severe traumatic brain injury. A recent brain scan (2007) shows damage to both left and right hemispheres. An extensive frontal lesion is evident involving the anterior parts of the superior, middle and inferior frontal gyri of the left hemisphere. It extends ventrally into the medial orbito-frontal cortex and laterally and posteriorly into the insula. The left temporal lesion involves the anterior temporal pole (middle and inferior temporal gyri). The primary auditory cortex appears to be spared. There is evidence of occipital damage on the right hemisphere due to a larger posterior horn of the lateral ventricle on the right and a right medial frontal contusion. EW’s large lesion caused severe expressive aphasia, dyspraxia and dysarthria. Eight years post-onset, EW is diagnosed as Broca’s aphasic on the BDAE, however, his excellent repetition skills are more in line with anomic aphasia. Naming objects in the BNT is poor (58% correct). EW has very little self-initiated speech which, if produced, is effortful, non-fluent and agrammatic with a reduced number of function words (i.e., prepositions, determiners). His speech and naming data suggest that EW does not have a selective verb impairment. His proportion of verbs (and nouns) in spoken language is higher than that of controls, which indicates that his speech mainly consists of content words (for a detailed analysis of EW’s and all other patients’ speech see Ma¨tzig, 2009). Picture naming of objects and actions is, if at all, mildly impaired with no difference between nouns and verbs. EW’s agrammatism is confined to production as his syntactic comprehension is intact. His word comprehension and semantics are also good. EW shows a deficit in three dimensional spatial processing. Below is an example of EW’s description of a set of pictures (Dinner Party). He produced very little speech with short ungrammatical utterances with missing constituents, in particular, missing (or wrongly used) grammatical morphemes (e.g., writing (down) the appointment/(he is) cooking (a) fish/ (the) cat is under the table/it (someone/he) is phoning and writing (down) the appointment). the tea time party/it’s phoning and writing the appointment/cooking fish/he’s just washing up/cat under the table/is going to sit there and call/is dressing up and fitting the necklace on/this is shirt and tie/they’ve got some flowers/is shaking hands/she smiling/person has stolen the piece/fish and chips/ is just running/that cat that’s licking out paws/it finished the fish/that’s looking all right/chatting TH has a family history of diabetes and in 1994 she was diagnosed with diabetes type II. In 2001, she suffered an episode of severe hypoglycaemia which led to diffuse brain damage (no lesion information available). Following the brain injury, TH became epileptic. The brain injury left TH with severe memory problems and fluent aphasia. Seven years post-onset, TH, according to the BDAE, is diagnosed as a fluent, predominantly anomic patient. Her rich variety of grammatical constructions, melodic line, phrase length, word finding difficulties, severely impaired oral object naming (10% correct on BNT) and well persevered repetition abilities identify her as an anomic aphasic. TH however also presents with mild paragrammatism, many paraphasias in her spoken language and moderately impaired auditory comprehension. These are typical features of a Wernicke’s aphasic patient. TH produces few nouns (and determiners) in connected speech while her use of verbs (and prepositions) is within the norm. Her object and action naming is severely impaired but more so for objects. TH’s comprehension of complex and long sentences is mildly impaired, possibly due to severe memory problems. Comprehension of nouns and verbs is moderately impaired with verbs being slightly more affected than nouns and she has moderate semantic difficulties. No impairment of spatial processing was found. Below is an example of TH’s description of a set of pictures (Dinner Party). TH is able to produce fluent, syntactically complex and often grammatically correct sentences. At times she uses too many grammatical morphemes (e.g., it’s very unusual but the dad are be (is) washing up/they go to the shop to get the fish and chips) and shows word finding difficulties for nouns which she can often correct herself (e.g., it’s where we actually sit down and eat. our food/there’s a little cat or dog running under it (the table)). somebody’s having a party/one of them is calling the other man to say would you like to come to the party/this is normal/mummy has to be cooking/it’s very unusual but the dad are be washing up/it’s. fish she is cooking/it’s where we actually sit down and eat... our food/ they’re getting the table ready/obviously they’ll be having friends coming around to have

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

379

dinner with them/the drinks and everything of course/lucky people/there’s a little cat or dog running under it/there’s pictures on the wall/everything is more or less what we all do when we’re having our dinner/obviously they are going out/the lady is getting dressed/obviously the man is getting dressed as well... in their bedroom/here come the friends/they are having somewhere out for their dinner/these it must be their two friends coming in/man and woman as always/she’s holding flowers.. probably to give to the other lady that’s cooking . as a present/something’s gone wrong/I don’t know what happened/there’s no fish/why would I put fish up/I personally wouldn’t do that/they go to the shop to get the fish and chips/they can take it home to the friends they’ve invited.for dinner/now the four are just sitting there eating

Appendix II Prepositions tokens, their properties and the number of times they were probed in sentence completion and grammaticality judgement

Experimental tasks Preposition subcategories

Theta-role assigned

above

spatial

spatial

6

after

temporal

temporal

6

–

at

spatial temporal subcategorized

spatial temporal

6 6 6

6 6 6

before

temporal

temporal

6

–

behind

spatial

spatial

6

–

by

other theta-role assigning preposition other theta-role assigning preposition

agent

6

–

instrument, manner

6

–

for

temporal other theta-role assigning preposition subcategorized

temporal benefactor

6 6

6 6

from

spatial temporal other theta-role assigning preposition subcategorized

spatial temporal animate source

in

spatial temporal subcategorized particle

spatial temporal

spatial

into

spatial

of

syntactic subcategorized

on

spatial temporal subcategorized particle

spatial temporal

Sentence completion

Grammaticality judgement tasks –

6

6

6 6 6

6 6 6

6

6

3 6 6 12

1 6 6 6

3

2

24 6

24 6

3 6 6 12

– 6 6 6 (continued on next page)

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

380 (continued)

Experimental tasks Preposition subcategories

Theta-role assigned

onto

spatial

spatial

3

since

temporal

temporal

6

–

to

spatial temporal other theta-role assigning preposition subcategorized infinitival

spatial temporal recipient

6 6 6

6 6 6

6 24

6 24

under

spatial

spatial

6

–

until

temporal

temporal

6

–

with

other theta-role assigning preposition

instrument (n ¼ 6), comitative (n ¼ 6), degree (n ¼ 3), substance (n ¼ 3)

18

–

6

–

276

179

subcategorized Total per task

Sentence completion

Grammaticality judgement tasks 2

Appendix III Examples of connected speech by the patients Utterances that show the preservation of word order (interpretation in brackets) BG:

DC: DOR: EW:

TH:

four years ago before I had my haemorrhage I was elected branch secretary for the union I do more than represent for members for people employment tribunal (I did more than represent the members at the people’s employment tribunal) Saturday Sunday and Wednesday I am off for the cross country skiing (on Saturdays, Sundays, and Wednesdays, I went off to do cross country skiing) now it’s about two hundred thimbles (DOR explains that his collection of thimbles has 200 exemplars by now) copying it’s no problem but I can’t read like the words two days it’s closed down (for two days it (the airport) was closed down) I learned some maths geography and household stuff. I changed the photography (EW explains how he was a teacher of maths, geography and household stuff but later he changed to photography and became a photographer) it was fairly bad but what it was actually called I don’t know (TH explains what happened the morning when she suffered an episode of severe hypoglycaemia) unfortunately I went down to my mother’s in Basilstroke

Utterances that show that prepositions are placed in correct position: BG: DC: DOR:

EW: TH:

I was office for the union long time ago (I worked for the union long time ago’) all the time we have negotiation with the union February the 7th right through to the 20th I am going to see Malaga it’s going there by coach (we are going there by coach) tears like drop off the head (DOR circumscribes crying) the stroke club like south end or bit before (the stroke club is at the south end or a bit before the south end) I was on bicycle and winding quite quickly on the road it did see me but it walks out on me (the pedestrian saw me but walked onto the street) obviously they go on to their dad and into hospital and everything (the children stayed with their dad and visited me in hospital) I wanted to be home around my kids when they are growing up obviously my kids are just down the road for me

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

381

Utterances that show sensitivity for the syntactic property of case assignment by prepositions (and verbs): BG: DC: DOR: EW: TH:

a man calls another man to invite him for a dinner for me that’s finished two of them is phoning them (two men are phoning each other) for me it’s ninety percent it’s rain (according to my opinion, it rains most of the time) I worked for him three years (I worked for him for three years) I haven’t spoke to her for a quite a few years now because she’s gone obviously they’ll be having friends coming around to have dinner with them

References Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX lexical database (release 2) [CD-ROM]. Philadephia, PA: Linguistic Data Consortium, University of Pennsylvania [distributor]. Bennis, H., Prins, R., & Vermeulen, J. (1983). Lexical-semantic versus syntactic disorders in aphasia: the processing of prepositions. Publikaties-van-het-Instituut-voor-Algemene-Taalwerenschap, 40, 1–32. Berndt, R. S., & Caramazza, A. (1980). A redefinition of Broca’s aphasia: implications for a neuropsychological model of language. Applied Psycholinguistics, 1, 225–278. Bolinger, D. (1971). The phrasal verb in English. Cambridge, MA: Harvard University Press. Chomsky, N. (1972). Studies on semantics in generative grammar. The Hague: Mouton. Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris Publications. Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press. Druks, J. (1991). A study of agrammatism with special reference to Hebrew. Unpubl. doctoral thesis. Milton Keyes, UK: The Open University. Druks, J., & Froud, K. (2002). The syntax of single words: evidence from a patient with a selective function word reading deficit. Cognitive Neuropsychology, 19, 207–244. Druks, J., & Masterson, J. (2000). An object and action naming battery. Hove: Psychology Press. Forster, K. I., & Chambers, S. M. (1973). Lexical access and naming time. Journal of Verbal Learning and Verbal Behavior, 12, 627–635. Francis, W. N., & Kucera, H. (1982). Frequency analysis of English usage: Lexicon and grammar. Boston: Houghton Mifflin. Friederici, A. D. (1981). Production and comprehension of prepositions in aphasia. Neuropsychologia, 19, 191–199. Friederici, A. D. (1982). Syntactic and semantic processes in aphasic deficits: the availability of prepositions. Brain and Language, 15, 249–258. Friederici, A. D. (1985). Levels of processing and vocabulary types: evidence from on-line comprehension in normals and agrammatics. Cognition, 19, 133–166. Friederici, A. D., Scho¨nle, P. W., & Garrett, M. F. (1982). Syntactically and semantically based computations: processing of prepositions in agrammatism. Cortex, 18, 525–534. Froud, K. (2001). Prepositions and the lexical/functional divide: aphasic evidence. Lingua, 111, 1–28. Froud, K., & Druks, J. Syntactic comprehension test. (unpublished). Goodglass, H., Gleason, J. B., & Hyde, M. R. (1970). Some dimensions of auditory language comprehension in aphasia. Journal of Speech and Hearing Research, 13, 595–606. Grimshaw, J. (2005). Words and structure. Stanford: CSLI Publications. Grodzinsky, Y. (1988). Syntactic representations in agrammatic aphasia: the case of prepositions. Language and Speech, 31, 115–134. Halle, M. (1997). Distributed morphology: impoverishment and fission. MIT Working Papers in Linguistics, 30, 425–449. Halle, M., & Marantz, A. (1993). Distributed morphology and the pieces of inflection. In K. Hale, & S. J. Keyser (Eds.), The view from building 20. Cambridge, Massachusetts: MIT Press. Harley, H., & Ritter, E. (2002). Person and number in pronouns: a feature-geometric analysis. Language, 78, 482–526. Haspelmath, M. (1989). From purposive to infinitive – a universal path of grammaticization. Folia Linguistica Historica, X, 287–310. Howard, D., & Patterson, K. (1992). Pyramids and palm trees: A test of semantic access from pictures and words. Bury St. Edmunds, UK: Thames Valley Test Company. Hyde, B. (2000). The structures of the to-infinitive. Lingua, 110, 27–58. Jackendoff, R. (1972). Semantic interpretation in generative grammar. Cambridge, MA: The MIT Press. Jackendoff, R. (1997). The architecture of the language faculty. Cambridge, Massachusetts: The MIT Press. Jakobson, R. (1964). Toward a linguistic typology of aphasic impairments. In A. De Reuck, & M. O’Connor (Eds.), Disorders of language. London: Churchill. Kaplan, E. F., Goodglass, H., & Weintraub, S. (1978). The Boston naming test. Philadelphia: Lea & Febinger. Kaplan, E., Goodglass, H., & Weintraub, S. (1983). The Boston diagnostic aphasia examination. Philadelphia: Lea & Febinger. Kay, J., & Ellis, A. W. (1987). A cognitive neuropsychological case study of anomia: implications for psychological models of word retrieval. Brain, 110, 613–629. Kean, M. L. (1977). The linguistic interpretation of aphasic syndromes: agrammatism in Broca’s aphasia, an example. Cognition, 5, 9–46. Kean, M. L. (1979). Agrammatism, a phonological deficit? Cognition, 7, 69–83. Kemmerer, D. (2005). The spatial and temporal meanings of English prepositions can be independently impaired. Neuropsychologia, 43, 797–806. Kemmerer, D., & Tranel, D. (2000). A double dissociation between linguistic and perceptual representations of spatial relationships. Cognitive Neuropsychology, 17, 393–414.

382

S. Ma¨tzig et al. / Journal of Neurolinguistics 23 (2010) 354–382

Kemmerer, D., & Tranel, D. (2003). A double dissociation between the meanings of action verbs and locative prepositions. Neurocase, 9, 421–435. Kolk, H., & Friederici, A. D. (1985). Strategy and impairment in sentence understanding by Broca’s and Wernicke’s aphasics. Cortex, 21, 47–67. Kreindler, A., & Miha˜ilescu, L. (1970). Operational words in aphasiacs. Journal of the Neurological Sciences, 11, 275–284. Larson, R. K. (1988). On the double object construction. Linguistic Inquiry, 19, 335–391. Leikin, M. (2002). Locative prepositions in language acquisition and aphasia. In S.Feigenbaum., & D. Kurzon (Eds.), Prepositions in their syntactic, semantic, and pragmatic context. Amsterdam: J. Benjamins. Levin, B. (1993). English verb classes and alternations: A preliminary investigation. Chicago: Chicago University Press. Lindstromberg, S. (1997). English prepositions explained. Amsterdam/Philadelphia: John Benjamins Publishing Company. Lonzi, L., & Luzzatti, C. (1995). Omission of prepositions in agrammatism and the universal grammar constraint of recoverability. Brain and Language, 51, 129–132. Lonzi, L., Luzzatti, C., & Vitolo, F. (2007). Recoverability of deletion in agrammatic production: the omission of prepositions. Italian Journal of Linguistics, 18, 249–277. Mack, J. L. (1981). The comprehension of locative prepositions in nonfluent and fluent aphasia. Brain and Language, 14, 81–92. Masterson, J., & Druks, J. Noun and verb comprehension test. (unpublished). Ma¨tzig, S. (2009). Spared syntax and impaired spell-out: the case of prepositions in Broca’s and anomic aphasia. Unpublished doctoral thesis. University College London. Ma¨tzig, S., Druks, J., Masterson, J., & Vigliocco, G. (2009). Noun and verb differences in picture naming: past studies and new evidence. Cortex, 45, 738–758. Miceli, G., Silveri, M. C., Romani, C., & Caramazza, A. (1989). Variation in the pattern of omissions and substitutions of grammatical morphemes in the spontaneous speech of so-called agrammatic patients. Brain and Language, 36, 447–492. Mittwoch, A. (1990). On the distribution of bare infinitive complements in English. Journal of Linguistics, 26, 103–131. Mondini, S., Luzzatti, C., Zonca, G., Pistarini, C., & Semenza, C. (2004). The mental representation of verb–noun compounds in Italian: evidence from a multiple single-case study in aphasia. Brain and Language, 90, 470–477. Morton, J., & Patterson, K. (1987). Little words – no! In M. Coltheart, K. Patterson, & J. C. Marshall (Eds.), Deep dyslexia London: Routledge & Kegan Paul. Neeleman, A. (1996). PP-complements in Dutch. In J. Camacho (Ed.), Proceedings of WCCFL 14 (pp. 383–397). Stanford: CSLI. Neeleman, A. (1997). PP-complements. Natural Language and Linguistic Theory, 15, 89–137. Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3–36. Pesetsky, D. (1998). Some optimality principles of sentence pronunciation. In P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis, & D. Pesetsky (Eds.), Optimality and competition in syntax. Cambridge, Mass.: The MIT Press. ¨ rungen: Studien zur psychologischen Grundlegung der Aphasielehre. Berlin: Springer. Pick, A. (1913). Die agrammatischen Sprachsto Postal, P. M., & Pullum, G. K. (1978). Traces and the descriptions of English complementizer contraction. Linguistic Inquiry, 9, 1–29. Rey, A. (1941). L’examen psychologique dans les cas d’encephalopathie traumatique. (Les problems). Archives de Psychologie, 28, 215–285. van Riemsdijk, H. (1990). Functional prepositions. In H. Pinkster, & I. Gene (Eds.), Unity in diversity. Dordrecht: Foris. Rizzi, L. (1985). Two notes on the linguistic interpretation of aphasia. In M. L. Kean (Ed.), Agrammatism. New York: Academic Press. de Roo, E., Kolk, H., & Hofstede, B. (2003). Structural properties of syntactically reduced speech: a comparison of normal speakers and Broca’s aphasics. Brain and Language, 86, 99–115. Ruigendijk, E. (2002). Case assignment in agrammatism: A cross-linguistic study. Saffran, E. M., Schwartz, M. F., & Marin, O. S. M. (1980). The word order problem in agrammatism: II. Production. Brain and Language, 10, 263–280. Schwartz, M. F., Saffran, E. M., & Marin, O. S. M. (1980). The word order problem in agrammatism: I. Comprehension. Brain and Language, 10, 249–262. Segui, J., Mehler, J., Frauenfelder, U., & Morton, J. (1982). The word frequency effect and lexical access. Neuropsychologia, 20, 615–627. Smith, M. D. (1974). On the understanding of some relational words in aphasia. Neuropsychologia, 12, 377–384. Svenonius, P. (2004). Adpositions, particles, and the arguments they introduce. http://ling.auf.net/lingBuzz/000042. Svenonius, P. (2007). Spatial prepositions in English. http://ling.auf.net/lingBuzz/000001. Tesak, J., & Hummer, P. (1994). A note on prepositions in agrammatism. Brain and Language, 46, 463–468. Tranel, D., & Kemmerer, D. (2004). Neuroanatomical correlates of locative prepositions. Cognitive Neuropsychology, 21, 719–749. Tyler, A., & Evans, V. (2003). The semantics of English prepositions. Spatial scenes, embodied meaning and cognition. Cambridge: Cambidge University Press. Wales, R., & Kinsella, G. (1981). Syntactic effects in sentence completion by Broca’s aphasics. Brain and Language, 13, 301–307. Warrington, E. K. (1991). The visual object and space perception battery. Bury St. Edmunds: Thames Valley Test Company.