نقش نحو اسم در تولید کلمه گفتاری: شواهدی از زبان پریشی

We describe MH who presents with agrammatic aphasia and anomia, and who produces semantic errors in the absence of a central semantic impairment. This pattern of performance implies damage to syntactic processes operating between semantics and phonological output. Damage here may lead to lexical selection errors and a deficit in combining words to form phrases. We investigated MH's knowledge and processing of noun syntax in mass and count nouns. She produced more count nouns than mass nouns. She showed impaired knowledge of noun syntax in judgement tasks and production tasks, with mass noun syntax being more impaired than count. We interpret these results in terms of a two-stage model of lexical retrieval. We propose that syntactic information represented at the lemma level is activated even in bare noun production, and can be differentially impaired across noun categories. That same damage can lead to semantic errors in production. For MH limited syntactic options are available to support production, and these favour count noun production. The data provide a new account of output semantic errors.

Theories of spoken word production need to account for the patterns of impaired and intact linguistic processes found in people with aphasia. One account of spoken word production which has been very influential in this regard holds that a direct mapping is achieved from a semantic level, incorporating the meaning of the word, to a phonological level, where the word's sounds are stored (e.g., Caramazza, 1997 and Morton, 1985). This model has dominated research and clinical work in aphasia, defining our understanding of aphasia and influencing the way in which people with aphasia are assessed and treated. There are many published case studies of people with anomia who present with a purely semantic deficit (e.g., JCU: Howard and Orchard-Lisle, 1984), those who present with a deficit in the phonological output lexicon (e.g., EE: Howard, 1995), those with a deficit in mapping between the two levels (e.g., GM and JS: Lambon Ralph et al., 2000; Cuetos et al., 2005), and those with a post-lexical deficit (e.g., NC: Martin et al., 1994 and Martin and Saffran, 1992). This model has influenced the selection of therapy methods for both adults and children (e.g., Nickels, 2002 and Best, 2005). In addition the model accounts for error production in aphasia by postulating that semantic paraphasias arise due to damage at the semantic processing level and phonological paraphasias arise at a phonological processing level. The deficit in mapping between the two levels of processing is of interest here. Patients with pure anomia, such as those described by Lambon Ralph et al. (2000) make very few semantic errors, readily reject the small number they do make, and, when they cannot name an item, they produce no response, or they attempt to describe the item. In other words, the deficit in mapping between the two levels does not result in significant numbers of semantic errors in such cases. The same hypothetical level of deficit has also been claimed however, as the source of output semantic errors. Caramazza and Hillis (1990) described RGB and HW, who made semantic errors in oral naming, while showing no impairment on input tests of semantics. Their breakdown in spoken word production is thus located in the same functional site as that of people with pure anomia, i.e., between semantic and phonological processing, and yet results in a different pattern of output. Caramazza and Hillis (1990) explained such errors in terms of a response blocking mechanism, which prohibits access to the target phonological form, but allows access to a near semantic neighbour. If this is the case the question arises as to why such errors do not occur in pure anomia. The data from RGB and HW and from the similar cases reported since then (e.g., PH and NK: Herbert et al., 2003 and Hickin et al., 2002) have placed a strain on the descriptive power of such models. Two influential groups of researchers claim that an intermediate stage exists between semantic and phonological representations (Dell et al., 1997 and Levelt et al., 1999). This has been labelled the lemma level (Kempen and Huijbers, 1983). According to the fullest articulation of their theory (Indefrey and Levelt, 2004 and Levelt et al., 1999) the lemma node provides access to information about a word's syntax. In this account access to phonology from semantics is dependent upon prior access to the syntactic level. The importance of this intermediate lexical level for understanding aphasic word retrieval deficits is as follows: if there is obligatory access to this level in normal spoken word production, and this level can be damaged in aphasia, it follows that a specific pattern of deficit will result from the damage.1 Incorporation of the lemma level into explanations of aphasic speech production offers three further possible deficit patterns: a deficit in access to the lemma, a deficit within the lemma level itself, and a deficit in accessing phonology from an intact lemma level. According to the theory a particular pattern of performance is associated with each level of deficit. There are published reports of the third functional lesion in processing nouns (e.g., Badecker et al., 1995), and verbs (Lee and Thompson, 2004), but few investigations have looked directly at lemma processing in aphasia. At the lemma level lexical selection takes place, bound by semantic constraints from input processes and by syntactic constraints operating within the level. It is possible that the pattern of semantic output errors without a semantic processing deficit, seen in patients such as RGB and HW (Caramazza and Hillis, 1990), results from damage to the lemma level. The lemma information for nouns includes the following: the syntactic class (noun, verb, adjective and so on); whether it is a count noun such as ‘cat’ or a mass noun such as ‘rice’; and in certain languages the grammatical gender of the noun (see e.g., Bock and Griffin, 2000). This information allows a noun to be combined correctly with other word classes in a noun phrase. The lemma node thus provides access to syntactic features, which specify the combinatorial information necessary for the word's appropriate syntactic use, including access to noun phrase structures in syntax (see Cleland and Pickering, 2003, for an account of the interface between the lemma and syntactic structures). Investigations into lemma level knowledge involve the experimental manipulation of these syntactic features. Two aspects of noun lemma information have been exploited in previous studies: grammatical gender, and the mass/count status of nouns. English no longer marks nouns for grammatical gender, however, there is a distinction between mass and count nouns, which is reflected in noun syntax. Syntactic rules govern the ways in which the two types of nouns combine with other words to make phrases. Count nouns can take a plural form (cat, cats), can be combined with both definite and indefinite articles (the cat, a cat), with denumerable quantifiers such as many and few, and with numerals (three cats), but not with non-denumerable quantifiers (such as much). Mass nouns cannot be pluralized (*rices), cannot take indefinite articles (*a rice), cannot be used with certain quantifiers such as many, or with numerals (*three rices), but can be combined with quantifiers such as much. 2 These differences can be exploited to investigate participants' access to lemma information. There is evidence from normal speakers and from those with aphasia, of retained access to syntactic information, when phonology is not available. Normal subjects in tip of the tongue states are frequently able to report the grammatical gender of items they cannot name, knowledge of the words' phonological form being available partially or not at all (Caramazza and Miozzo, 1997 and Vigliocco et al., 1997). Furthermore Vigliocco et al. (1999) report that normal participants have knowledge of the count-mass status of items for which they are in a tip of the tongue state. A small number of neuropsychological studies support the ToT findings. GM (Henaff Gonon et al., 1989), Dante (Badecker et al., 1995), and BA (Macoir and Béland, 2004) had marked word finding problems, but were able to correctly identify the grammatical gender of items they could not name. Vigliocco et al. (1999) describe the performance of MS who was able to identify whether an item was a count or mass noun even though he was unable to name the item in a picture naming task. All four cases support the contention that syntax and phonology are independently represented. These results have been interpreted as support for the existence of an intermediate stage of processing at which syntactic information is available. Lexically represented syntactic features are available when the full word form is not. Although comprehensive test details are not given in all cases, all four participants with aphasia appear to have had good semantic processing, and none of the four produced significant numbers of semantic errors. They are similar in this respect to GM and JS (Lambon Ralph et al., 2000), who presented with a pure anomia. Two-stage models incorporating an intermediate level can explain the pattern found in these three cases, and in cases of pure anomia, as a breakdown in processing after an intact lemma level has been accessed. In most of these cases the errors were predominantly failures to respond and circumlocutions, and BA (Macoir and Béland, 2004) produced mainly neologisms. This leaves open the possibility that damage to the lemma level itself may result in output semantic errors, i.e., errors of lexical selection. There are a number of case reports investigating mass and count noun processing in people with acquired language deficits. In these studies the authors were concerned with a possible dissociation between mass and count noun processing, and thus the data are reported in this light. FA (Semenza et al., 1997) could name mass and count nouns equally well, and had intact sentence syntax on testing. Tests of noun phrase syntax (e.g., grammaticality judgement) showed an advantage for processing count nouns over mass nouns however. CN (Semenza et al., 2000) showed an advantage for naming mass nouns over count. The two cases demonstrate that count and mass noun processing including syntactic processing, can be differentially affected in aphasia, with implications for models of production and for theories of lexical representation. The findings from the studies outlined above can be summarised as follows: semantic errors can arise when semantic processing is intact, but in none of the reports of these cases so far has noun syntax been formally investigated (Caramazza and Hillis, 1990 and Herbert et al., 2003). In cases where syntactic information is available when phonological access is not, semantic errors are rare, and they are often readily rejected by the person with aphasia ( Badecker et al., 1995, Henaff Gonon et al., 1989, Macoir and Béland, 2004 and Vigliocco et al., 1999). In studies investigating the mass/count noun distinction directly, a dissociation between the two sets may be present, which is reflected in either input processing in tests such as grammaticality judgement, or in output such as oral naming ( Semenza et al., 1997 and Semenza et al., 2000). Of interest therefore is the status of noun syntactic knowledge in people with aphasia presenting with semantic errors in output only. We address this issue in the data reported here.

In summary, MH's data offer the first detailed case report of anomia due primarily to a deficit in the syntax of nouns. Testing of word retrieval showed that both mass and count nouns were affected, but mass nouns more so. This was shown in her overall performance and her error pattern. This is the first account of the influence of determiner cues on production, and on semantic error type. We propose that this is because the syntactic information available for mass nouns was differentially affected relative to that for count nouns, and that this related directly to determiner availability which is determined by frequency. We propose also that the evidence provided here supports the claim that syntactic information influences production even for bare nouns, at least for people with aphasia. For MH, damage to the lemma level gives rise to semantic errors in spoken output. This explanation provides a principled account of this pattern of impairment, and has implications for both assessment and intervention strategies.