Production of grammatical case in Polish aphasia

(1)

1 UNIVERSITEIT VAN AMSTERDAM

Graduate School of Humanities MA General Linguistics (Clinical Track)

MA Thesis

Production of grammatical case in Polish aphasia

Agnieszka Przeździak

Student ID: 11106921 Supervisors: Dr. Annemarie Kerkhoff Dr. Judith Rispens Amsterdam July 2016

(2)

2

Acknowledgements

In a short note, I would like to express my appreciation towards the people who contributed to the preparation and writing of this thesis. I am grateful to my supervisor, Annemarie Kerkhoff, for the time she took to help me write my thesis, her patience and useful comments. I would like to thank the staff and patients of the University Clinical Centre in Gdańsk and the Nursing and Healthcare Centre in Tczew, Poland, for their kind welcome and extremely helpful attitude while I collected my data. It must be mentioned that none of it would have been possible without my Grandfather – the time and effort he put into helping me contact and get around the hospitals were invaluable to the process. I owe a huge thank-you to my family as a whole, for their

unwavering support and encouragement not only during the writing of this paper, but throughout my entire time as a travelling student. Last but not least, I want to thank Erik for everything he did and still does to help me, computer-related or otherwise.

(3)

3

Aphasia is generally understood as a language disorder involving both production and comprehension, caused by brain damage and brought about by events such as strokes or traumatic brain injuries. Various types of aphasia can be distinguished based on the lesion site and the clinical picture. The specific terminology may vary from one classification system to another, e.g. Dronkers & Larsen (2001) list seven major types, consistent also with Goodglass & Kaplan’s (1972) classification: Broca’s, Wernicke’s, conduction, global, anomic, transcortical motor and transcortical sensory aphasias. A popular classification system among Polish speech and language therapists is that proposed by Alexander Luria (1967; Jastrzębowska 1998), who proposed a division into three major aphasia types: motor (expressive), acoustic (sensory) and semantic. Regardless of the choice of classification, it is important to note that the distinctions are in no way clear-cut from both the linguistic and neurological perspective, with variability within patient groups with regard to lesion site as well as the clinical picture. The two major aphasia types may be considered an example1_{: the syndrome known as Broca’s aphasia was}

traditionally associated with lesions to Broca’s area (Brodmann’s areas 44 and 45 in the left frontal lobe), and characterized by expressive difficulties, halting, effortful speech, and the inability to process complex syntactic structures (e.g. Goodglass & Kaplan 1972). Wernicke’s aphasia, on the other hand, was thought to be caused by lesions to Wenicke’s area (Brodmann’s area 22 in the left temporal lobe), and commonly associated with fluent speech and

comprehension difficulties, even at single-word level (Dronkers & Larsen 2001). However, it has been shown that a lesion to Broca’s area is not enough to cause the symptoms of Broca’s

aphasia, as damage to Broca’s area only usually results in transient mutism (ibid.). The

localization of lesions causing Wernicke’s syndrome has also been a matter of some controversy – some studies report lesions to post-Rolandic temporal and parietal areas with no involvement of anterior structures, while others found pre- as well as post-Rolandic lesions or exclusively anterior lesions (ibid.); in Dronkers & Larsen’s (2001) own study, participants identified as Wernicke’s aphasics did not have lesions in Wernicke’s area at all, with other areas of the temporal lobe damaged. Other studies have also shown a lack of one-to-one relation between

1_{‘Motor aphasia’ is taken to correspond to ‘non-fluent aphasia’ and ‘Broca’s aphasia’, while ‘sensory aphasia’} would correspond to ‘fluent’ or ‘Wernicke’s’ aphasia. When describing a study, the term provided by its authors is normally used. The current participants were diagnosed with either ‘motor’ or ‘sensory’ aphasia, but the terms ‘fluent’ and ‘non-fluent’ are preferred as they make the expected inter-group differences clearer.

(5)

5 lesion and syndrome, also underscoring the role of brain regions other than Broca’s and

Wernicke’s areas (e.g. Cappa 2012, in Bastiaanse & Thompson 2012; Thompson-Schill 2005; Dronkers et al. 2004; Dick et al. 2001), although, in some of these studies, certain areas have been implicated in language comprehension (Dick et al. 2001; Dronkers et al. 2004). Ullman et al. (1997) also found a dissociation in the localisation of regular and irregular morphology, as evidenced by their differential impairment in patients with lesions in anterior and posterior areas of the brain. The increasingly fluid distinction between the aphasia types has led some to use the

fluent and non-fluent labels, a distinction based on the aphasic patients’ speech rate. Fluent and

non-fluent aphasias correspond to posterior lesion sites and anterior lesion sites, respectively (Goodglass & Kaplan 1972; Berndt 2001; Jastrzębowska 1998; De Bleser 1987). A symptom commonly associated with non-fluent aphasia is agrammatism, generally characterized by omissions of function words, as well as a dissociation between single-word processing and sentence processing (Cappa 2012; Goodglass & Menn 1985). Fluent aphasia, on the other hand, is assumed to present with paragrammatism – the speech is fluent but full of errors of

substitution (e.g. Heeschen 1985; Kolk & Heeschen 1990). The concepts of agrammatism and paragrammatism themselves have their roots in German and French aphasiological literature from the beginning of the past century. De Bleser (1987) provides a detailed account of the German aphasiological tradition, discussing ideas put forward by Wernicke, Kleist, Goldstein, and others who had laid the groundwork for later research. Many modern works present the two syndromes as a dichotomy, although the early scholars apparently agreed that the distinction is not theoretically founded (De Bleser 1987). De Bleser also argues that the issues of agrammatic and paragrammatic utterances have been confounded with those of agrammatic and

paragrammatic patients due to the incomplete transfer of German literature into the modern academia. The original agrammatism/paragrammatism distinction, with its relation to particular lesion sites, was proposed by Karl Kleist in 1914. Further investigations revealed problems for such a clear syndrome/lesion association, which led Kleist (1916, in De Bleser 1987) to restate his definitions as ‘functionally different moments of the same underlying deficit’. The original debate will not be discussed here, but, as will hopefully be clear from the more modern research cited below, it is anything but solved. The research on agrammatism to date shows that it is difficult to identify a uniform pattern of impairments, even within one language or one study. Some findings include: the differential impairment of Tense and Agreement (e.g. Wenzlaff &

(6)

6 Clahsen 2004; Friedmann & Grodzinsky1997); the fact that omissions of bound morphemes rarely occur if they result in non-words and that substitutions generally occur within the same word class (e.g. Ułatowska & Sadowska 1988; Marková & Cséfalvay 2010; Lehečková 2001; Kertesz & Osman-Sagi 2001). These studies will be discussed in detail in the following sections. Although agrammatism and paragrammatism are often talked of as opposites – characterising aphasic syndromes that are thought of as members of a dichotomy – some studies have suggested that the underlying grammatical deficit may be the same, with the two patient groups

differentiated only by language behaviour (Heeschen 1985; Kolk & Heeschen 1990; 1992). In this view, ‘agrammatic’ production (i.e. omissions) is a symptom of adaptation, while

‘paragrammatic’ output (i.e. substitutions) is evidence of the true grammatical impairment of the aphasic patient.

Much of the theory behind the impairments found in fluent and non-fluent aphasia is based on data from English, a language with little inflectional morphology. However, cross-linguistic research in aphasia has shown that the same deficits can manifest differently in

different languages, depending on how various features are marked (Paradis 2000): the pattern of impairment depends on factors such as word order, the existence of articles, affixes marking case, tense, and so on, and the importance of these features in a given language. For example, German cases are marked primarily on articles, which means the latter carry a lot of syntactic information – German aphasics have been found not to omit articles, as opposed to language-impaired speakers of Dutch, in which articles do not carry equally important information (Bastiaanse et al. 2003). In morphologically rich languages such as Polish, case, tense, aspect and agreement are marked using a system of prefixes, roots and suffixes, which means bound morphemes are especially important for expressing meaning, making it an interesting language to study in terms of grammatical impairment. The present investigation will focus on the qualitative differences between fluent and non-fluent aphasics, in light of previous research suggesting that these differences may not be so dramatic in languages with developed

inflectional systems (e.g. Tomaszewska-Volovici 1976; Slobin 1991; MacWhinney & Osman-Sagi 1991). Tomaszewska-Volovici’s (1976) study is of special interest here, as the author

compared the production of nominal morphology in groups of fluent and non-fluent aphasics and found no qualitative difference between the two and no particular pattern of impairment. Polish has seven cases, some with more than one ending depending on the noun’s grammatical gender,

(7)

7 making it an interesting language to examine in terms of potential morphological deficits in aphasic production. This paper is structured as follows: a description of some concepts regarding the deficits found in aphasia are described, followed by a discussion of previous findings

regarding the production of case and the predictions for the present study. The data presented afterwards was collected with a focus on the questions: 1) To what extent is case marking impaired in Polish aphasics? Here, the difficulties observed in the currently tested patients will be of interest, as well as the question whether a particular case marking is the most vulnerab le, and how the results compare to previous findings from Polish and other languages; and 2) Are there qualitative differences in error patterns between fluent and non-fluent aphasia? In order to answer these questions, fluent and non-fluent Polish-speaking aphasics were tested using a picture description task and a case elicitation task.

(8)

8

2. Background

2.1. Relevant features of Polish

There are seven cases in Polish, of which six will be discussed2_{.Polish case is a rather}

complex phenomenon – it is marked on nouns and adjectives, each with paradigms of their own, as well as numerals, pronouns, gerunds and adjectival participles. This paper, due to its relatively small scope, will concern itself only with case assignment to nouns. The role of case is,

primarily, to indicate syntactic relations in the sentence. It can be said to have a semantic

function, in that specific cases play specific semantic roles, i.e. the Nominative is ascribed to the agent performing an action, the Genitive and Accusative to the object, Dative often indicates the person affected by the action, Instrumental – the instrument with which the action was

performed, and the Locative – where the event took place (Bańko 2007: 145). The Locative is the only case that always requires a preposition. However, the case marking system itself is complex and occasionally irregular. The same case can be signalled by up to five endings, and the same ending can be an exponent of more than one case. It is marked by inflectional suffixes, sometimes accompanied by stem alternations, which conflate case and number. The case marking system is illustrated in Tables 1 and 2 (after Dąbrowska 2004 and Jagodziński 2008).

Case Feminine Masculine Neuter

Nominative -a (-Ø, -i) -Ø (-a, -o) -o, -e, -ę

Genitive -i/-y -a, -u (-i/-y) -a

Dative -‘e,a_-i/-y _{-owi (-u, -‘e, -i/-y)} _-u

Accusative -ę (-Ø) -Ø, -a (-ę, -o) -o, -e, -ę

Instrumental -ą -em (-ą) -em

Locative -‘e,a_-i/-y _-‘e,a_{-u (-i/-y)} _-‘e,a_-u

Vocative -o, -u, -i/-y, (-Ø) -‘e,a_{-u, (-o)} _-‘e,a_{-u, (-o)}

Table 1. The Polish case marking system (the singular endings). a The symbol ‘ indicates that the preceding consonant is

palatalised. Infrequent endings are parenthesized.

2_{The seventh case, Vocative, indicates a noun that is being directly addressed and is used relatively little in} everyday speech. Moreover, the Vocative is special in that it is not involved in any syntactic relations (Bańko 2007: 57).

(9)

9

Case Feminine Masculine Neuter

Nominative (=VOC) -y (-i), -e -i (-y), -e, -owie, -a -a

Genitive -ów, -i (-y), -Ø -ów, -i (-y), -Ø -ów, -i (-y), -Ø Dative -om -om -om

Accusative -y (-i), -e -ów, -i (-y), -Ø -a

Instrumental -ami, -mi -ami, -mi -ami, -mi

Locative -ach -ach -ach

Table 2. The Polish case system (plural endings). Infrequent endings are parenthesized.

The most important factor in the choice of an ending is grammatical gender, which can be predicted from the phonological form of the nominative for most nouns. Where more than one case ending is available, additional criteria are applied, e.g. masculine mass and collective nouns, abstract nouns and nouns referring to locations tend to take –u in Genitive, while those

designating tools, names of months or body parts are likely to take –a. Morphological criteria also exist, e.g. derivational affixes –arz, -acz, -ek take –a in the Genitive, while –unek takes –u. Phonology may play a role as well – for instance, nouns ending in a palatalised consonant are usually inflected with –a. These criteria are by no means definitive, however, and can often turn out to contradict one another (Dąbrowska 2004). The extent of the complexities observed in the Polish inflectional system is not directly relevant to the present investigation, but it may be useful in later discussion. As can be seen in Table 3, ambiguities arise between the cases, but the gender of the noun determines where they occur. Thus for the masculine virile3_{noun, the}

Genitive form overlaps with the Accusative, while for the non-virile noun the Accusative is the same as the Nominative, and stem alternations are present. For the feminine and neuter nouns, the case marking is shared in the Dative and Locative. The neuter noun also does not change

3_{‘Virility’ in terms of grammatical gender denotes nouns that are animate, such as people and some animals; ‘virile’} masculine nouns have different declensions that ‘non-virile’ ones. However, a noun’s classification as ‘virile’ does not always reflect reality, e.g. trup ‘corpse’ is declined like virile masculine nouns. This distinction is interesting to make a note of when describing the Polish declension system, but it was not a factor in the present investigation.

(10)

10 form in the Accusative. This table is only an example of a case paradigm in Polish and is by no means definitive; as was mentioned above, a variety of endings is possible depending on a noun’s gender and phonological properties.

Case/Gender Masc. virile Masc.

non-virile

Feminine Neuter

Nominative chłop-iec ‘boy’ stół ‘table’ dziewczyn-a

‘girl’

dzieck-o ‘child’

Genitive chłop-ca stoł-u dziewczyn-y dzieck-a

Dative chłop-cu stoł-owi dziewczyn-ie dzieck-u

Accusative chłop-ca stół dziewczyn-ę dzieck-o

Instrumental chłop-cem stoł-em dziewczyn-ą dzieck-iem

Locative chłop-cu stol-e dziewczyn-ie dzieck-u

Table 3. Example case paradigm for singular masculine, feminine and neuter nouns.

A this stage of the discussion, a short note on healthy speakers’ performance on noun morphology seems in order. Tested on the Genitive and Dative, adults use case inflections productively on nonce words, suggesting they are sensitive to phonological properties of the stem, as well as the nouns’ semantics indicated by the context (Dąbrowska 2004). Children produce explicitly marked Genitives early, and the earliest produced markings are the most frequent ones (masculine –a and feminine –i/-y). All three Genitive endings are overgeneralised, albeit to a small extent, and most errors involve confusions of the two masculine endings, suggesting there is no ‘default’ Genitive masculine (ibid.). Another study testing children’s ability to produce Genitive, Dative and Accusative inflections found an interaction of case and gender (Dąbrowska & Szczerbiński 2006). Dative was more difficult than Genitive regardless of gender, but proved significantly harder than Accusative only for feminine and neuter genders. Accusative, in turn, was more difficult than Genitive for masculine and feminine but not neuter. Both studies found that token frequency, not regularity, was the best predictor of productivity. Apart from the Nominative, the Genitive and Accusative inflections have the highest type and token frequencies, while Dative has the lowest (Dąbrowska & Szczerbiński 2006).

(11)

11

2.2. Agrammatis m vs. paragrammatism and dissociations in aphasic syndromes

Generally, agrammatism is the more widely investigated syndrome. A number of models and theories have been put forward to account for the deficits observed in agrammatic

production. One proposal that seems to have gained traction is the Tree Pruning Hypothesis (Friedmann & Grodzinsky 1997). The TPH assumes that the C-, Tense and/or Agreement nodes are underspecified, resulting in the impairment of morphology depending on the severity of aphasia (i.e. the more severe the aphasia, the higher the impairment on the syntactic tree). Friedmann and Grodzinsky (1997) present a case study of a Hebrew-speaking patient with agrammatic aphasia, who showed a severe impairment in verbal tense, copula (located in the T node) and Wh-questions as well as complementizers (which are thought to reside in the CP node), but no impairment in verbal, adjectival or nominal agreement. It should be emphasized that these results only refer to production, with the grammaticality judgement task showing almost no errors. Support for the TPH was found in studies in Hebrew (Goral 2001; Friedmann 2006), where the production of Tense was shown to be impaired with the preservation of Agreement, but never the opposite, and Russian (Friedmann et al. 2010), where agrammatic aphasics were unable to comprehend Wh-movement-derived sentences, even with the availability of morphological cues such as case. Many other studies disproved the TPH or supported it only partially, e.g. in German (Wenzlaff & Clahsen 2004; Burchert et al. 2005), English (Clahsen & Ali 2008) or Swahili (Abuom et al. 2011). This hypothesis was modified by Wenzlaff and Clahsen (2004) and presented as Tense Underspecification Hypothesis to account for their findings in German agrammatism. Burchert and colleagues (2005) expanded it further as Tense/Agreement Underspecification Hypothesis to explain their own results. Another proposal was put forward by Bastiaanse and colleagues (2011) in the form of PAst DIscourse LInking Hypothesis (PADILIH), which does not rely on the syntactic tree and seeks to explain the difficulties some agrammatic aphasic have with reference to the past. Bastiaanse et al. (2011) found that native speakers of Chinese, English and Turkish performed significantly worse on past than present and future tenses in both production and comprehension. In Russian, where time-reference interacts with aspect, aphasics did not show a significant impairment of the perfective aspect compared to the imperfective, but a non-past advantage was found in the production of the imperfective (Dragoy & Bastiaanse 2013). Interestingly, such an impairment

(12)

12 was not found in Polish, with the errors consisting solely of adding a wrong prefix to the

perfective (Ułatowska et al. 2011).

Based on studies of fluent aphasia, attempts at identifying the underlying deficit of ‘paragrammatic’ production have also been made. Butterworth and Howard (1987) suggested that paragrammatism is a deficit in control. The fluent aphasics in their study made qualitatively similar errors to their controls, leading the authors to hypothesize that paragrammatisms are normal ‘slips of the tongue’ found in healthy individuals; aphasics simply produce more of them because they lack the appropriate amount of cognitive control to supress them. In this framework the rate of paragrammatisms does not correlate with the production of neologisms or

comprehension deficits. Butterworth & Howard’s data also reinforces the decades-old observation (e.g. Isserlin 1922, cited in De Bleser 1987) that agrammatic and paragrammatic phenomena can co-occur in the same patient. They found frequent omissions of open- and closed-class words but not of inflectional endings, also leading to a suspicion that agrammatism and paragrammatism may be the opposite ends of a continuum of error processes. In any case, Butterworth & Howard (1987) conclude that paragrammatism is not a distinct pattern of aphasic impairment. Based on an extensive case study of a Wernicke’s aphasic, Ellis et al. (1983) claim, on the other hand, that if a central conceptual deficit causes paragrammatic output, then it should also affect comprehension. Indeed, the Wernicke’s aphasic tested by Ellis et al. was impaired in speech perception – the authors conclude that the patient’s inability to monitor his speech leads to the production of neologisms and paraphasias.

In their paper describing a systematic method for analysing aphasic speech, Saffran and colleagues (1989) compared the performance of agrammatic aphasics, fluent

non-agrammatic aphasics and normal controls on a re-telling of the Cinderella story. They found that the agrammatic group showed a significant reduction in the use of free and bound grammatical morphemes, while the production of morphology by the other aphasic group do not significantly differ from that of normal controls. An inconsistency was also observed in the impairment of free and bound morphemes among the agrammatic aphasics, suggesting that the two do not

necessarily co-occur. The agrammatic group was also characterised by their difficulty in producing propositional utterances and structural simplicity of the utterances that they do

(13)

13 in producing them and the proportion of well-formed sentences was higher. While the

impairments found in the agrammatic group had been predicted, the reduced sentence

complexity among the non-agrammatic aphasics was unexpected. The authors suggest that the deficit underlying this reduction may be of lexical or phonological nature, since there is evidence that these factors affect structures produced by normal speakers (Saffran et al. 1989). Bird and Franklin (1996) largely replicated Saffran et al.’s (1989) results in a study comparing the retelling of the Cinderella story by non-fluent agrammatic, non-fluent non-agrammatic, and fluent English-speaking aphasics. All non-agrammatic patients exhibited a similar pattern of deficits in well-formedness, elaboration and word class measures, with non-fluency acting as an additional symptom rather than a factor in other deficits. The agrammatic patient was visibly impaired in bound morphology and omitted articles and pronouns completely. This participant also produced many more nouns than verbs, which the authors suspect may be an effect of imageability and not just word class, but this conclusion is not presented as definitive. In any case, Bird & Franklin (1996) conclude that the reduction of complexity and well-formedness in aphasic production is, indeed, likely to be an effect of word-finding difficulties, although that conclusion seems less certain when it comes to the one agrammatic patient. This lexical account might suggest that agrammatism is an optionally occurring symptom not necessarily tied to non-fluent aphasia, and that the distinction between non-fluent and non-non-fluent aphasia is not clear cut even in morphologically impoverished languages such as English.

Bastiaanse and colleagues (1996) compared the spontaneous speech of English, Dutch, and Hungarian fluent aphasics to normal controls. Some of the aphasics displayed a reduction in the ability to use complex grammatical structures while still producing matrix clauses and phrases with the same frequency as normal speakers. The aphasic group’s lexical deficit was demonstrated in the reduced verb frequency and diversity. The authors concluded, however, that at least some of the fluent aphasics have grammatical limitation which cannot be explained by a lexical deficit. The study was elaborated on by Edwards & Bastiaanse (1998), using larger groups of Dutch- and English-speaking aphasics and controls. Both English and Dutch aphasics displayed difficulties with nouns as well as verbs (both groups’ verbs were reduced in range, but Dutch speakers still used verbs with a frequency similar to that of controls’). The English

aphasics produced fewer embedded clauses both than controls and Dutch aphasics. The data from all participants cannot be explained with the lexical account (since noun production

(14)

14 problems did not always correlate with deficits in production of complex syntactic structures). A processing account (according to which aphasics produce simple syntax because of a deficit in processing; see e.g. Niemi (1990), discussed in section 2.4) was also not considered an adequate explanation by the authors, as the aphasics did not compromise on the range of argument

structures used, although the authors allow for the possibility that some fluent aphasics rely on a restricted range of verbs to achieve the apparently normal range of argument structures. The authors came to the same conclusion as in the previous study (Bastiaanse et al. 1996).

The aphasic patients’ problem with verb production, mentioned above, was explored in studies on the dissociation of verbs and nouns in aphasia – although, as with everything else, the pattern of impairment is by no means uniform. Miceli et al. (1988) gave an object and action naming task to agrammatic (non-fluent) and anomic (fluent) aphasics and found that the former group performed worse at naming actions than objects, while the opposite was true for the anomic patients. Moreover, there was a dissociation between production and comprehension, i.e. the deficits found in production were not predictive of comprehension. The authors interpret these results as evidence that word class is a dimension of lexical organisation and that these dimensions are likely independent of one another (conforming to the idea that the deficits observed in aphasic production have an underlying lexical cause) . Luzzatti and colleagues (2002) also compared the two groups’ performance on naming tasks; the ‘fluent’ group included Wernicke’s and anomic aphasics, while the ‘non-fluent’ aphasics were split into agrammatic and non-agrammatic. In the non-fluent group, a dissociation in one direction only was found – performance was decidedly better on nouns, primarily among the agrammatic aphasics. Both dissociations were found in the fluent group, albeit within separate subgroups – Wernicke’s aphasics performed better on nouns, while anomic aphasics showed superior performance for verbs. Effects of frequency were found for verbs and imageability for nouns, and the advantage partially disappeared when these effects were removed. In an analysis of the spontaneous speech of agrammatic and anomic aphasics, Bastiaanse et al. (1996) found that both groups produce fewer verbs than normal controls, although the agrammatic patients had normal type-token ratios, as opposed to the anomic group. This is not very surprising, since the main characteristic of anomic aphasia is word-finding difficulty, and the anomic participants in this study were reported to use a strategy to avoid producing verbs. Bastiaanse et al. only focused on the production of verbs, so while it shows that aphasics produce fewer verbs, it cannot exactly be

(15)

15 taken as evidence for a dissociation between nouns and verbs in aphasia. The evidence for such a dissociation was presented by Bates et al. (1991a), who tested Chinese (Mandarin- and

Taiwanese-speaking) aphasic patients. The Chinese language is useful for such an investigation for two reasons. Individual nouns and verbs are compounds made up of elements that carry their own noun or verb meaning, and there is virtually no noun or verb morphology, which makes it less likely that one word class will be more difficult to process than the other.4_{Bates and}

colleagues’ participants with Broca’s aphasia also performed worse on action than object naming, while the result was opposite for the Wernicke’s aphasics. The authors do not discuss whether the effects of imageability were found, which can also influence the difficulty with which the two word classes are retrieved. Bates et al. argue that the most plausible explanation is a semantic-conceptual one, according to which the names of actions and objects are processed in different areas of the brain (motor and sensory, respectively).

Agrammatism and paragrammatism have also been approached as symptoms of

adaptation and impairment (Kolk & Heeschen 1990). The omissions of grammatical morphology observed in agrammatism are seen as an adaptation, in that they result from overuse of a

particular linguistics form. Kolk and Heeschen argue that in normal speakers legal omissions result in contextual ellipsis, and that there are similarities with agrammatic speech: frequent omissions but few substitutions, and rare omissions of obligatory suffixes (e.g. plurals). Other similarities involve word order relative to the non-finite verb and articles – features specific to Dutch and German, the languages on which Kolk and Heeschen based their proposal. Since adaptation is optional, the authors assumed that in elicitation tasks, which limit the freedom to omit, the shift to elliptical style will be less likely than in spontaneous speech and substitutional errors will increase. Therefore paragrammatism – characterised primarily by errors of

substitution – would be the real symptom of a patient’s impairment in the authors’ view. Kolk and Heeschen suggest that Wernicke’s aphasics are unaware of the errors in their speech which causes them to be less concerned about their disorder and fail to adopt such strategy – task manipulation should therefore not result in increased substitutions among those patients. Heeschen’s (1985) data indeed supports this expectation.

4_{One proposed explanation for the dissociation is that verbs are more difficult to process because 1) they often} require a heavier load of grammatical marking, and 2) they carry more syntactic information that nouns, such as the argument structure (e.g. Bates et al. 1991a; Bastiaanse et al. 1996).

(16)

16

2.3. Accessing linguistic information

Another widely debated issue is how grammar is accessed and processed within the brain. When discussing grammatical morphology, it is interesting to consider the distinction between single and dual processing mechanisms. The discussion originally centred on the past tense of verbs in English (Marcus et al. 1995; Joanisse & Seidenberg 1999, etc.). German plural

inflections also provided a fertile ground for accounts of both mechanisms (Marcus et al. 1995; Clahsen 1999, etc.).

According to dual-mechanism models, irregular inflections are stored in memory, and regular, ‘default’, ones are applied using a separate system which processes rules. English irregular verbs, for instance, form their past tense in idiosyncratic ways, and there is a fixed number of them, which suggests that they are stored in memory as a list. The productive –ed ending, on the other hand, is applied in all other circumstances, including novel words (Marcus et al. 1995); evidence for this also comes from overregularizations produced by

English-acquiring children (in circumstances when the correct form could not be found in memory) as well as the classic ‘wug experiment’. The memory and rule components interact, therefore the rule is blocked if an irregular form can be retrieved from memory. Marcus and colleagues (1995) provide a list of circumstances in which memory cannot be accessed, which includes, but is not limited to, the aforementioned nonce words, low-frequency words, competing roots (e.g. homophones), and derivations, as well as memory failures observed in e.g. Alzheimer’s disease or anomic aphasia. If an inflection is applied in these situations, it is a ‘default’, i.e. an operation that can apply to any item as long as there is not a pre-existing form listed for that item. Clahsen (1999) provided evidence for the dual approach in an extensive study of German plural and past tense inflections, basing their classification as regular and irregular on Marcus et al.’s model. Clahsen examined the German morphology within the so-called Minimalist Morphology model, which maintains the basic distinction between lexicon and rules, with the irregular (lexically restricted) inflection encoded in the lexical items themselves. The ‘irregular’ words were found to show frequency and similarity effects, while the ‘regular’ ones were apparently computed as ‘stem + affix’. Children as well as adults generalised regulars and extended irregulars only by analogy, which is interpreted by Clahsen as evidence for the dual mechanism. Moreover, a dissociation between regular and irregular inflections was found in an experiment employing event-related potentials. The assumption that inflections are processed by two separate systems

(17)

17 also underlies the Declarative/Procedural Model (Ullman et al. 1997; Ullman 2004), according to which irregular and regular forms are produced using two distinct kinds of memory systems. Declarative memory is thought to be responsible for learning and storage of facts and events and thus be responsible for irregulars, while Procedural memory is used to learn and process motor, perceptual and cognitive skills and, consequently, apply rules to words. Ullman and others (1997) assume that Declarative memory is located in temporo-parietal regions, and Procedural memory in frontal regions, and should therefore be selectively impaired in case of brain damage. Indeed, Ullman et al.’s English-speaking participant with posterior aphasia was shown to

produce regular forms more reliably and overregularize frequently in a sentence completion task. On the other hand, the patient with anterior aphasia tested in this study was better at producing irregular forms and never overregularized. This finding is by no means definitive – Penke and Westermann (2006), for example, tested German- and Dutch-speaking Broca’s aphasics using a similar elicitation task and did not find the selective deficit with regular inflections. In fact, most participants struggled to produce correct irregular inflections. The authors argue that this

difference can partly be ascribed to the confound between an inflectional suffix and regularity present in English but not in all languages.

Psycholinguistic theories supporting the single mental mechanism assume that inflectional forms are supplied using one mental mechanism based on patterns found in the environment. There is no clear separation between lexicon and grammar, and forms are accessed in memory using the phonological and semantic information. One example of such an account is Joanisse and Seidenberg’s (1999) connectionist model, used to evaluate claims regarding the dual mechanism made in Ullman et al. (1997). The model was trained to analyse semantic, phonological and morphological information in ways which would normally be required in speaking, hearing, repeating and transforming (e.g. from present to past tense). It was also given the phonological code and past-tense semantics of Ullman et al.’s (1997) nonce words. Once the model ‘learned’ these dependencies, it was given a phonological and a semantic deficit. The phonological deficit affected regular, irregular and nonce verbs, but the effect was greatest for the last category, while damage to semantics affected irregular verbs the most. Joanisse and Seidenberg take these results as evidence against the dual model and in favour of a network which includes distinct phonological and semantic representations. Since the representations have different realisations in the brain, they can be dissociated in neuropathology. This

(18)

18 connectionist model was supported in another study (Patterson & Holland 2014), in which

Broca’s aphasics were given the standard present-to-past verb inflection task as well as the opposite. The participants had most difficulty in transforming regular verbs from present to past and irregular verbs from past to present. Moreover, when asked to determine whether a given verb refers to a past or present action, the participants correctly classified almost all regular verbs but were at chance for irregulars. These results suggest that the impairments in inflectional

morphology must stem from something more than the inability to produce inflected verbs, and point to the importance of semantic knowledge. Bates and colleagues (1991b) are also

proponents of a connectionist model, which involves two basic principles: ‘cue validity’, which reflects the information value of a given form within a particular language, and ‘cue cost’, which refers to the amount and type of processing required with the activation of that form. Linguistic information is thus represented in a network of probabilistic connections between forms and meanings. The model predicts variation primarily based on language type (i.e. depending on which features are important in a given language), and rules are represented in form-meaning and form-form mappings, the strength of which can vary cross-linguistically. Bates et al. admit, however, that the model does not contain principles that would predict qualitative differences between patient groups when applied to aphasia. Another version of a single processing model is the ‘network’ model (Bybee 1995), which emphasizes the importance of actual usage of forms in terms of type and token frequency for establishing and maintaining representations. An

important feature of the elements in the lexicon is ‘lexical strength’, which stems from words’ token frequency and determines the ease with which they are accessed. Lexical strength explains why English irregular forms have high token frequency – their availability in the input makes them easy to memorize, while less frequent forms are regularised. Connections between forms are made through finding similar phonological and semantic features. The type and number of shared features determine the strength of these lexical connections; words with similar patterns of semantic and phonological connections create generalisations which Bybee dubs ‘schemas’. The higher the type frequency of the pattern in the schema, the more likely it is that it will be applied to new items. The main predictor of morphological productivity in the single mechanism models is frequency, with the regular and irregular forms treated in the same way (i.e. as

(19)

19 Polish-speaking adults and children (Dąbrowska 2004; Dąbrowska & Szczerbiński 2006)

supports this usage-based view, rather than the dual mechanism. 2.4. Previous studies of morphology in aphasia

The finding that inflectional morphology is often retained in aphasia is presented in studies of some morphologically rich languages. In a study of Hungarian, MacWhinney and Osman-Sagi (1991) analysed the speech of 9 Broca’s and 5 Wernicke’s aphasics during a picture description task. Both patient groups omitted subjects (grammatical in this language) much more frequently than other constituents. However, Broca’s aphasics tended to omit indirect objects more than Wernicke’s aphasics, while the reverse was true for direct objects omissions. Since indirect objects are more optional, this trend suggests that Broca’s aphasics are more sensitive to the language rules. Interestingly, this pattern was also reflected in the omissions of inflections (which are synonymous with substitutions with the Nominative in Hungarian) – Broca’s aphasics omitted more Accusative inflections while Wernicke’s aphasics tended to omit the Datives. Substitutions with the Nominative were quite high, while substitutions using other inflections were rather rare; some were interpreted by the authors as strategies to simplify

discourse. Nominal morphology was also largely preserved in Turkish-speaking aphasics (Slobin 1991). The difference between patients with Broca’s and Wernicke’s aphasia was found

primarily in terms of fluency – the former group produced simple, well-formed structures while the latter’s speech involved a lot of ‘extraneous elaboration’ (rambling, groping speech). Wernicke’s aphasics also tended to omit direct and indirect objects, while verb omissio ns were found in Broca’s aphasia. Omissions of case inflections (resulting in Nominatives in Hungarian) were rare and more frequently found in Broca’s aphasia. Almost no other case substitutions were found. Interestingly, when Broca’s aphasics had word-finding difficulties they still produced correctly-inflected pronouns. In another study, Niemi (1990) found that the syntactic structures produced by Finnish-speaking Wernicke’s aphasics were simplified in comparison to healthy speakers. The aphasic group produced fewer complex subjects, fewer embedded clauses, and showed a tendency to resort to canonical word order. Moreover, the Finnish aphasics seemed to use the case markings in a way that results primarily in a canonical sentence pattern. Finnish is a morphologically rich language, and surface case marking on subjects and direct objects of sentences is governed by aspect, quantification, definiteness, and negation information.

(20)

20 participants tended to produce Nominative subjects as opposed to Genitive and Partitive (which are morphologically complex) to a greater extent than controls. On the other hand, Nominative is underused in direct objects, while the Partitive and Genitive cases are overused. In normal

speech, oblique cases also account for the majority of case markings on direct object. Niemi (1990) argues that since a canonical sentence pattern in Finnish is one where the object is morphologically marked and the subject unmarked, the production patterns in fluent aphasics should be seen as a strategy. The author also ascribes the low verb ellipsis rates to the

psychological and computational significance of the verb, which assigns the argument structure. In sum, the overall simplification of structures produced by fluent aphasics in this study, as well as the strategies suggesting reliance on canonical forms, are to be seen as evidence for a

processing deficit (rather than loss of linguistic knowledge). In German, which marks case on articles rather than nouns, agrammatic Broca’s aphasics omit a low percentage of articles (Bastiaanse et al. 2003). The rate of substitutions was higher and the substitutions primarily involved case errors. These errors predominantly involved Dative and Accusative exchanges, with the Datives substituted more often. Dutch-speaking agrammatic aphasics in the same study tended to omit rather than substitute articles which carry information about definiteness but not case. In a different study (De Bleser et al. 1996), German Broca’s aphasics were able to assign case correctly in sentences with a narrow context, i.e. in the presence of a preposition or another NP. Performance was significantly poorer in more ambiguous sentences, i.e. when the missing case morphology was only recoverable based on the complementary distribution between the Nominative and the objective cases, Accusative and Dative.

It is also interesting to look at findings from Slavic languages, of which many have morphological systems similar to Polish. Lehečková (2001) examined the speech of 17 Czech-speaking motor aphasics in conversation as well as a story-telling and a picture description tasks. The patients made significantly more errors of substitution than omission overall. Omissions involved free grammatical morphemes as well as many content words. Almost all substitutions were made with existing forms, and most involved case, tense, and mood markings. Most substitutions of case involved producing the Nominative instead of other inflections; the participants substituted more than half of the possible case markings. In the same paper, Lehečková presents a case study of a non-fluent aphasic. The patient produced a relatively low number of ungrammatical utterances (89/824). Omissions constituted the majority of errors, and

(21)

21 most involved content words (primarily verbs and participles, and only 3 nouns). Substitutions involved a variety of inflections, but mostly resulted in the production of present instead of past tense and of Nominative instead of Locative, Accusative, and Genitive case markings. The author points out that the most frequent forms were most likely to be retained. Marková & Cséfalvay (2010) analysed the spontaneous speech of 7 patients with anomic aphasia and 4 patients with non-fluent aphasia. The non-fluent aphasics were further subdivided into mild and moderate groups according to fluency. There was a high rate of noun substitution in the more severe non-fluent group, and all prepositions were omitted. This group also replaced all case inflections with the Nominative; the Instrumental case was an exception as it was substituted to a lesser extent. The milder aphasic group made significantly fewer case errors; Genitive and

Instrumental were substituted the most with the Nominative, followed by Locative. Non-Nominative substitutions were also found in this group and primarily involved substituting the Instrumental with other inflections. The more severe group also made many errors with verbs, while the mild group produced most of them correctly; both groups tended to omit verbs in their speech. Omissions of bound grammatical morphemes were not found at all; free morphemes were omitted rather than substituted, and there was a slightly higher number of such omissions in the mild aphasic group. The anomic patients exhibited virtually no difficulties in their use of grammatical morphemes and they did not omit verbs. However, they omitted nouns (albeit to a small extent) and prepositions. In a study of case assignment in Russian aphasia, Ruigendijk (2002) tested 7 non-fluent agrammatic aphasics on free speech as well as case elicitation tasks. Overall, the number of case errors was quite low. The most impaired inflections were Dative and Instrumental; errors consisted mainly of substitutions and Nominative was overgeneralized to some extent in the elicitation tasks. In both free speech and the elicitation tasks, patients produced more correct inflections when a case-assigning category (verb or preposition) was present. In free speech, the participants tended to use the default Nominative case when no case assigning category was present; in the other tasks, errors were primarily substitutions.

In studies of Polish-speaking aphasics, an unclear picture of impairments emerges. In one case study of a motor aphasic (Jarema & Kądzielawa 1987), a syntactic rather than

morphological impairment was observed. The patient was tested on a picture description task as well as on sentence completion. Massive omissions of grammatical morphology – traditionally expected in aphasia – were not observed, although the patient did omit obligatory pronouns and

(22)

22 substitute tense and aspect within verb morphology. A severe reduction of speech rate and phrase length was observed in spontaneous speech (MLU in words 2.48, compared to the control’s 10.14). More substitutions than omissions were found in the elicitation task (which corresponds Kolk & Heeschen’s (1990) idea that substitutions occur when the patient is unable to omit structures as a strategy). Case inflections were unimpaired, although the authors provide one example of a substitution of Accusative with the Dative. The patient was unable to process embedded clauses, struggled to produce pronouns which involved co-reference and tended to overuse simple, canonical constructions. The same features were impaired in comprehension. An interesting finding from a sentence-picture matching task is that 5 out 6 case errors were

misinterpretations of the Genitive for the Dative case. Ułatowska and Sadowska’s (1988) fluent aphasic, on the other hand, was severely impaired in inflectional morphology, while derivational morphology was relatively well-preserved. The suffixes marking Nominative case and third person in present tense were best preserved. With regard to the remaining case markings, Genitive was the least and Dative was the most impaired. Grammatical gender of nouns did not play a role. Inflections requiring morphophonemic alternations yielded many errors. Verb inflections were severely impaired in gender, person, tense, and aspect paradigms. All inflectional errors involved substitutions rather than omissions and the substitutions never crossed word classes. The patient’s sentences were reduced in complexity and produced primarily in the canonical SVO order. Omissions of function words were very rare and mostly involved reflexive particles. Tomaszewska-Volovici (1976) compared 10 fluent and 10 non-fluent aphasics on a picture description task and a sentence completion task (similar to the one used in the current study). She also found that the impairment in noun morphology was exhibited by substitutions rather than omissions of suffixes. No specific pattern of case impairment was found in this study, but, as in Ułatowska & Sadowska (1988), noun stems which undergo morphophonemic alternations were the most problematic. Case errors were found, but most inflections were produced correctly: motor aphasics provided 1211 correct answers, out of 1600 (75%); for sensory aphasics, the correct answers constituted 899 out of 1500 answers (60%). Tomaszewska-Volovici also noted that the plural Dative inflection was apparently the most difficult one for the motor aphasics, which she interprets as a frequency effect; it is pointed out, however, that the difference is not great enough to be considered significant. The author

(23)

23 fluent aphasics made more errors overall. This group also exhibited word-finding difficulties, produced perseverations and tended to substitute prepositions and suffixes in spontaneous speech, as opposed to the non-fluent aphasics who mainly omitted free morphemes. This last finding is, in fact, in line with ‘traditional’ views of fluent and non-fluent aphasic speech. Tomaszewska-Volovici also noted that both patient groups produce many phonological

paraphasias at the end of words, which makes the interpretation of some errors difficult: as can be seen in Tables 1 and 2, the majority of case endings are vowels, so mispronunciations can often lead to ambiguities.

2.5. Summary and predictions

The focus of this paper is on case production in Polish aphasia – to what extent it is impaired and whether the fluent and non-fluent aphasics can be said to differ qualitatively. To sum up the background, paraphasias and neologisms are thought to be found in fluent rather than non-fluent aphasia, due to the more significant control deficits found in the former (Butterworth & Howard 1987; Ellis et al. 1983). Aphasic patients tend to produce fewer content words than normal controls, with some studies suggesting that non-fluent aphasics struggle with the production of verbs more than nouns, and the opposite being true for fluent aphasics. The majority of the studies in morphologically rich languages show that both the fluent and non-fluent aphasics tend to retain their ability to produce inflections. Case studies from, for instance, Polish, show pictures of impairment that do not correspond to the ‘traditional’ view of deficits found in different types of aphasia, i.e. omissions of grammatical morphology in motor aphasia and substitutions in sensory aphasia. What is more, the studies comparing the two patient groups rarely find a qualitative difference in terms of morphology. A finding that the studies described above have in common is that the Dative case is the most impaired. Dative is also the least frequent case in Polish; we also saw in studies of healthy adults and children that frequency plays an important role in productivity. It is, therefore, reasonable to expect that the Dative inflection will be the most problematic. Moreover, although omissions of bound morphemes are unlikely to occur as they often result in non-words, the Nominative is the least marked form, the citation form, so it would not be surprising if participants substituted other inflections with it. More errors can be expected in the elicitation task, where the participants cannot use a strategy to avoid a difficult structure (Kolk & Heeschen 1990). The control group is not expected to make grammatical errors. The fluent and non-fluent patient groups are not expected to differ

(24)

24 qualitatively in terms of inflectional morphology. The aphasic groups are expected to differ in terms of fluency, with the motor aphasics more dysfluent. The participants with sensory aphasia are expected to produce more paraphasias and neologisms (cf. also Tomaszewska-Volovici 1976). The aphasics’ speech will also be examined for apparent difficulties with verbs and nouns, although it should be kept in mind that the tasks used in this study (picture description and case marking elicitation) do not allow for a precise assessment of the potential difficulties. Finally, the data will be discussed in light of the different accounts of agrammatism and paragrammatism discussed above.

(25)

25

3. Method 3.1. Participants

The participants were tested at the rehabilitation ward of the University Clinical Centre in Gdańsk and at the Nursing and Healthcare Centre in Tczew, Poland. The primary criterion for inclusion was a left-hemisphere lesion and aphasia diagnosed by a speech therapist. Due to limited time and resources available for this study, the matching for factors such as age,

education and time post-onset both between and within the participant groups is less than perfect. However, considering that the primary aim of this paper is to provide a qualitative description of Polish-speaking aphasics, it was decided to include all patients meeting the criteria and capable of performing the tasks discussed in section 3.2. Ten participants took part in the study, of which 8 were diagnosed with motor (non-fluent) aphasia and 2 with sensory (fluent) aphasia. According to both institutions’ speech therapists and clinical psychologists, the standard measures used to diagnose aphasia are Jadwiga Szumska’s diagnostic battery (Szumska, 1980), the Token Test and the Verbal Fluency Test. The participants ages ranged from 56 to 80 (mean age 66.9, SD=8.21) and there were 6 women and 4 men. The patients varied in their times post-onset, ranging from 1 month to 8 months and 25 days at the time of testing. All patients had suffered a stroke: the aetiologies consisted of 3 haemorrhagic and 7 ischaemic strokes. Detailed information regarding the age, aetiology, lesions and diagnosis of all participants is presented in Table 4. Five

additional patients were included as controls, of whom 3 were diagnosed as dysarthric and 2 had had transient aphasia as a result of stroke (n=4) or surgery (n=1). The controls were aged 41-80 (mean age 62.6, SD=13.9). Their times post-onset ranged between 21 days and 3 months, and there were 3 women and 2 men. The control patients were comparable to the aphasics in terms of age and education (Table 4). These patients make for an interesting control group since they had all suffered from left-hemisphere lesions but had no language deficit at the time of testing.5_Note

that this last conclusion is based solely on these participants’ patient histories; no results from the tests mentioned above had been provided and the participants were not evaluated using any additional measures prior to testing. Unfortunately, two of dysarthric patients, TR and RK, could only be included as controls for the case elicitation task. Their articulation problems proved to be

5_{As will become apparent in later discussion, this diagnosis may not be completely accurate. Of TR’s 27 utterances,} 8 were partially or fully impossible to transcribe. While this may seem a relatively small propo rtion, the missing utterances were rather long and thus the comparison would still be skewed. Of RK’s 11 utterances, 8 could not be fully transcribed.

(26)

26 so severe at times that large parts of the recording could not be transcribed. The comparison between those two patients and the rest of the participants would therefore be skewed, especially in terms of relative measures, such as MLU or relative dysfluency, discussed in section 3.4. TR and RK’s performance in the first task will be discussed separately in the results section. Details regarding the participants from the control group are also included in Table 4. The aphasia diagnoses and education levels provided are stated as found in the patient histories.

(27)

27

PatientID Gender Age Onset Aetiology Lesions Diagnosis Education/Work

KM (1) M 76 1m 19d

ischaemic stroke; previously also 2005 (L) and 2009 (R)

left parietal lobe, massive cortical & subcortical deficiencies, paraventricular white matter hypodensity; old foci in left occipital lobe

sensory aphasia; episodic memory

deficits; word-finding difficulties higher

ZG (1) M 66 1m 29d ischaemic stroke left (no details) sensory a., anomia secondary/carpenter

AN (2) F 56 3m 14d

subarachnoid haemorrhage from AcoA

no details (likely frontal left) mixed aphasia, mainly motor secondary/railway

BS (2) F 67 2m 5d haemorrhagic

stroke left fronto-parietal region motor aphasia secondary

EO (2) M 64 3m 18d ischaemic stroke

deep structures of L. hemisphere; irregular

hypodensity between the head of left caudate nucleus and frontal part of semioval centre

motor a.; articulation difficulties;

word-finding difficulties; anomia higher/engineer

JA (2) M 58 5m ischaemic stroke left (no details) non-fluent motor a., dysarthria higher/lawyer

JJ (2) F 80 1m 1d

ischaemic stroke; previously in 2011

L. subcortical occipital-temporo-parietal region & frontal lobe; previously subcortical parieto-occipital

motor aphasia no information available

KD (2) F 76 1m 17d ischaemic stroke

malacia in left occipital lobe, ischaemic sites in left parietal lobe

mixed a., mainly motor secondary

LP (2) F 59 8m 25d

subarachnoid haemorrhage from AcoA

no details (likely frontal left) transcortical motor a.; anomia,

(28)

28

Table 4. Patient data. 1 – fluent aphasia, 2 – non-fluent aphasia, 3 – dysarthria (control), 4 – transient aphasia (control); L – left, R – right; AcoA – anterior communicating artery.

PatientID Gender Age Onset Aetiology Lesions Diagnosis Education/Work

TC (2) F 67 1m 20d ischaemic stroke

acute ischaemic focus in M2CA; malacia in left frontal lobe

mixed aphasia, mainly motor secondary

GK (3) F 80 21d ischaemic stroke

deep structures of L. hemisphere; small malacia sites in both temporal lobes & near occipital ventricles

dysarthria higher/teacher

RK (3) M 65 3m 5d ischaemic stroke

malacia in both sides of the cerebellum; ischaemic foci in left occipital lobe and pons

dysarthria secondary/decorator

TR (3) M 65 1m 16d ischaemic stroke

R. cerebellum, smaller in the L. cortex of the cerebellum; centrum semiovale & deep structures of both

hemispheres

dysarthria secondary/shipyard mechanic

BK (4) F 41 2m 17d surgery: removal

of meningioma left (no details)

transient mixed aphasia; transient

right hemiparesis secondary/housewife

RS (4) F 62 1m 9d ischaemic stroke

ischaemic focus in deep structures of L. hemisphere; cortical & subcortical hypodensity: fronto-temporo-parietal region

transient aphasia; transient right

(29)

29

3.2. Materials

The test consisted of two parts, eliciting semi-spontaneous speech and specific case inflections, respectively. Semi-spontaneous speech was elicited using the Cookie Theft

Picture, taken from the Boston Diagnostic Aphasia Examination (Goodglass & Kaplan 1972). It was deemed unlikely that the picture will be well-known to Polish patients, since the

battery is rarely used by speech therapists. Its other advantages are that it offers plenty of context for the patient to provide an ample speech sample, and its widespread use in

academic research makes potentially comparing the results with existing literature possible. The second task consisted of 15 items eliciting 5 out of the 7 Polish cases – the Vocative had been left out as it is only used in spoken language, and specifically when addressing another person. The task involved sentence completion, with the target noun provided in Nominative form underneath the gap. In all but item one the gap was at the end of the sentence. The participants were expected to fill in the sentence with the noun in the correct form. Each case was elicited with two nouns in singular form and one in plural form, in order to target a variety of inflectional endings. The items were chosen from a frequency list of Polish lexemes (Kazojć 2011) and included only frequently occurring nouns in order to ensure that all patients were familiar with them. The items were arranged in a random order. The full list of items used in the elicitation task can be found in the Appendix.

3.3. Procedure

The participants were tested individually in a quiet room. No breaks were planned as the test took around 10 minutes to administer, depending on the patient. Each session was recorded. Before the beginning of the test, the participants were given a brief explanation of the project’s aims and asked to sign a statement previously prepared by the examiner,

confirming that they agree to participate in the study and to be recorded. If a participant was unable to provide a signature due to e.g. hemiparesis, he or she was asked to state their name on the recording. The Cookie Theft Picture was administered first, and the participants were asked to ‘describe what is happening in the picture in as much detail as possible’. If a patient struggled with the description, he or she was given prompts such as: ‘what is happening here? [point]’, ‘what is this woman doing?’, ‘what are the children doing?’, etc. The elicitation task was administered immediately after. A brief explanation was provided along with the

instruction, ‘A word is missing from the sentence. Please fill in the sentence using they key word [point] in the correct form.’. The examiner first read out the key word in the

Nominative form and then the sentence, indicating the gap with prosody. The items had also been printed in large font and with plenty of line spacing to avoid confusion, and presented to

Production of grammatical case in Polish aphasia

MA Thesis

Production of grammatical case in Polish aphasia

Agnieszka Przeździak

Acknowledgements

Contents