The syntax-lexicon continuum: Explaining variation in aphasic ...

20
1 The syntax-lexicon continuum: Explaining variation in aphasic language Vitor C. Zimmerer, Ewa Dąbrowska, Rosemary A. Varley Abstract Generative “words and rules” theories have been at the basis of the most prominent approaches to ex- plaining grammatical disruption in aphasia. However, these have been challenged by a range of observa- tions in aphasia such as residual islands of grammatical speech, verb bias effects, and cases where a for- mally more complex grammatical construction is preferred over a simpler construction. These phenome- na are pervasive enough to motivate reconsideration of theoretical frameworks of language impairment. This article exemplifies these phenomena and introduces the framework of usage-based Construction Grammar which may account for these observations. One important innovation is the syntax-lexicon continuum which organizes all linguistic knowledge in terms of constructions which differ in degree of complexity and concreteness/abstractness. Another assumption is that grammatical constructions, just like single words, have meaning and are subject to frequency effects. We demonstrate how the frame- work explains hitherto challenging observations in aphasia and sketch out research questions for the future. We also discuss implications for diagnosis and intervention of aphasic language impairment. 1. Introduction Since the first systematic investigations of grammatical impairment in aphasia (Caramazza & Zurif, 1976; Zurif, Caramazza, & Myerson, 1972), a range of theories have emerged trying to explain how language representation is affected by neurological damage. One of the leading questions is motivated by an ob- served asymmetry in aphasic language capacity: Why do people with aphasia produce or comprehend some sentences and words more successfully than others? Early seminal studies were carried out at a time when generative grammar was becoming the dominant theory of language (e.g. Chomsky, 1965, 1981). These “word and rules” frameworks make one shared assumption: that linguistic knowledge es- sentially consists of stored word and morpheme representations, and a set of abstract grammatical pro- cedures (including structure-building rules and movement rules), which combine these representations and in so doing can generate any grammatical expression in a language. Approaches to aphasia made use of concepts specific to such theories, such as representation of traces (Grodzinsky, 2000; Mauner,

Transcript of The syntax-lexicon continuum: Explaining variation in aphasic ...

1

The syntax-lexicon continuum: Explaining variation in aphasic language

Vitor C. Zimmerer, Ewa Dąbrowska, Rosemary A. Varley

Abstract

Generative “words and rules” theories have been at the basis of the most prominent approaches to ex-

plaining grammatical disruption in aphasia. However, these have been challenged by a range of observa-

tions in aphasia such as residual islands of grammatical speech, verb bias effects, and cases where a for-

mally more complex grammatical construction is preferred over a simpler construction. These phenome-

na are pervasive enough to motivate reconsideration of theoretical frameworks of language impairment.

This article exemplifies these phenomena and introduces the framework of usage-based Construction

Grammar which may account for these observations. One important innovation is the syntax-lexicon

continuum which organizes all linguistic knowledge in terms of constructions which differ in degree of

complexity and concreteness/abstractness. Another assumption is that grammatical constructions, just

like single words, have meaning and are subject to frequency effects. We demonstrate how the frame-

work explains hitherto challenging observations in aphasia and sketch out research questions for the

future. We also discuss implications for diagnosis and intervention of aphasic language impairment.

1. Introduction

Since the first systematic investigations of grammatical impairment in aphasia (Caramazza & Zurif, 1976;

Zurif, Caramazza, & Myerson, 1972), a range of theories have emerged trying to explain how language

representation is affected by neurological damage. One of the leading questions is motivated by an ob-

served asymmetry in aphasic language capacity: Why do people with aphasia produce or comprehend

some sentences and words more successfully than others? Early seminal studies were carried out at a

time when generative grammar was becoming the dominant theory of language (e.g. Chomsky, 1965,

1981). These “word and rules” frameworks make one shared assumption: that linguistic knowledge es-

sentially consists of stored word and morpheme representations, and a set of abstract grammatical pro-

cedures (including structure-building rules and movement rules), which combine these representations

and in so doing can generate any grammatical expression in a language. Approaches to aphasia made use

of concepts specific to such theories, such as representation of traces (Grodzinsky, 2000; Mauner,

2

Fromkin, & Cornell, 1993) or particular levels of hierarchical projection (Avrutin, 2000; Friedmann &

Grodzinsky, 1997).

There is, however, a building interest in phenomena that are not traditionally captured by generative

approaches to language. We review how studies on formulaic language, lexical bias and non-

conventional profiles have strained conventional theory, and how some authors have favoured an alter-

native, namely the Construction Grammar framework. Construction Grammar has greatly influenced

other domains of research such as language typology (Croft, 2001), first and second language acquisition

(Ellis, Römer, & O’Donnell, 2016; Tomasello, 2003), and understanding of individual differences

(Dąbrowska, 2015; Street & Dąbrowska, 2014). Increasingly, researchers are beginning to apply some of

its concepts to aphasia, but so far, no explicit case has been made for viewing acquired language disor-

ders through the prism of Construction Grammar. We explain the framework and put focus on the “syn-

tax-lexicon continuum” (SLC), which is where Construction Grammar presents its biggest departure from

generative approaches.

Linguistic representation, so the theory suggests, does not consist of words and rules, but is embodied in

a “constructicon”, i.e., a structured network of constructions, the forms of which differ in size and ab-

stractness. Constructions include single words and morphemes, but also lexicalized utterances, partially

lexicalized frames with one or more slots for additional construction insertion (e.g. What’s X?), as well as

fully abstract constructions such as the transitive active (X [VERB] Y). Together, they form the SLC. We

show how it can explain observations from aphasia which have been difficult to integrate into words and

rules approaches, and pay particular attention to the passive construction in aphasia, which has been

crucial for theory development in the past. Finally, we explore consequences for future research and

clinical practice.

2. Aphasia: Exceptions to the rules

Generative approaches have produced explanations for how certain hypothesized syntactic transfor-

mations, such as movement, or types of suffixation, can become disrupted or unavailable. Such impair-

ment is assumed to affect entire categories of constructions, such as non-canonical structures (including

passives) in the case of movement, or regular plural forms or past tense in the case of suffixation. Chal-

lenges to these accounts present when data show variations, i.e. when some forms within a category are

available, but not others. Some of this heterogeneity may be explained by noise in the data, as perturba-

tion of an intricate system (language) is affected not only by individual diversity but also variation in site

3

and extent of neurological damage. However, a number of studies have found these variations to be

systematic and have drawn on alternative theories for explanation. We first review these studies.

2.1. Formulaic language

One core feature of aphasia is disruption of grammatical processing, and while most strongly associated

with non-fluent, Broca’s type aphasia (Caramazza & Zurif, 1976; Goodglass, Christiansen, & Gallagher,

1994), some degree of grammatical impairment can be found in other aphasias as well (Bates &

Goodman, 1997). In production, these manifest as agrammatic (“telegraphic”), or paragrammatic utter-

ances. However, even agrammatic speech contains some islands of fluency and grammatically complete

structures, as in the following sample (underlined; Varley & Zimmerer, 2018, p884):

(1) It’s alright. Yes. Mate ... Jack comes and all … but … oh dear ... Jack … er ... old … er seventy. No.

Sixty eight Jack … but swim. Me like this [gesture] … swimming … er … I can’t say it …

Paul Broca’s historical, severely aphasic case “Leborgne”, known for his repeated use of the neologistic

form tan, also could produce one complex noun phrase, namely the expletive sacré nom de dieu (Code,

2013). The properties of that phrase as well as of utterances such as It’s alright, oh dear, like this and I

can’t say it in (1) are notable. First, they all appear to be fixed phrases with specific discourse functions.

Second, , they can contain a high proportion of function words: 80% in the case of I can’t say it and

100%. While omission of grammatical morphology is a symptom of agrammatic production, and can be

observed in other utterances, use of fixed expressions can result in samples from speakers with non-

fluent aphasia containing a similar rate of function words to normative samples (Zimmerer, Newman,

Thomson, Coleman, & Varley, 2018).

The expressions above are part of what is regarded as formulaic language, i.e. idiomatic, familiar and/or

common language forms. Formulaic language is not a marginal phenomenon but makes up a substantial

portion of discourse in healthy speakers (Erman & Warren, 2000; van Lancker-Sidtis & Rallon, 2004;

Wray, 2012). The proportion of formulaic output is higher in both fluent and non-fluent aphasia (Bruns et

al., 2018; Code, 1982, 1989; Van Lancker Sidtis & Postman, 2006; Zimmerer et al., 2018). Formulaic lan-

guage present a challenge to models that solely consist of single words and grammatical rules as these

would predict that an individual who can produce I can’t say it could also produce utterance with a simi-

lar structure, such as You can’t see Jack. But this is not the case as especially in severe aphasia, language

formulas see little or no variation. Formulas are sometimes excluded from analysis of aphasic production

and labelled as automatic, non-propositional speech; this view can be traced back to Hughlings-Jackson

(1874). However, one analysis of production of the formula I don’t know in aphasic discourse suggested

4

that at least speakers with non-fluent aphasia use them in appropriate contexts (Bruns et al., 2018).

Comprehension of formulas may be equally preserved (Sidtis & Yang, 2017).

Formulaic expressions are regarded as single units which represent sequences of strongly collocated

individual words, or storage of the entire forms as single morphemes (Siyanova-Chanturia, 2015). Repre-

sentation may depend on additional properties such as meaning, length, usage frequency and age of

acquisition. The holistic nature of formulas explains why these constructions are easier to process for

healthy speakers (Conklin & Schmitt, 2012), and retained in the face of lexical or grammatical impair-

ment. While processing of less familiar or novel utterances require activation of multiple lexical forms

together with more abstract units, a language formula facilitates lexical activation and combination ei-

ther through strong co-activation patterns, or, in case of more lexicalized expressions, by reducing the

number of words and combinatorial steps.

In principle, the two types of representation (strong collocations and single form storage) are not exclu-

sive. A phrase or utterance can be represented as both even within the same individual. For example, a

child may learn a formula as a single form early in life (e.g. What’s that?), later learn how it is segmented

(what, is, that), but retain the lexicalized form. In healthy speakers, lexicalized formulas can activate their

formal constituents (Siyanova-Chanturia, 2015). It is possible that this process is reversed in aphasia, as

individuals lose the multi-word processing route and have to rely on the lexicalized form.

2.2. Plurals

Application of grammatical rules typically makes a structure more complex. Agrammatism or grammati-

cal simplification in aphasia is typically explained by impairment to a rule application process. One exam-

ple of this is plural inflection of regular nouns, which in English involves adding the plural morpheme -s

to the noun stem (e.g. Marcus, 1995). A postulated impairment of the pluralization rule in aphasia would

then lead to number errors in aphasia and a strong bias towards singular forms where the plural is re-

quired. Such singularization errors can be observed in aphasia. At the same time, error rates are affected

by whether the noun in question occurs more often in its singular form (e.g. cat) or its plural form (e.g.

ears; Biedermann, Lorenz, Beyersmann, & Nickels, 2013). Importantly, overuse of the plural becomes

more likely with plural dominant nouns. In a Cinderella story retelling task, people with aphasia errone-

ously produced the plurals shoes and slippers when the story required the singular (Hatchard, 2015;

Hatchard & Lieven, under review). In these cases, the formally more complex plural is preferred over the

unsuffixed and therefore formally simpler singular, contradicting accounts of rule-based inflection im-

pairment.

5

2.3. Verb frame bias and the passive

Frequency effects occur not only at the single word level, but also at the level of word combinations.

Processing difficulty in aphasia is increased when a verb does not appear with its “preferred” argument

structure frame. In a self-paced reading task, DeDe (2013a) presented sentential complements (e.g. The

ticket agent admitted that the airplane had been late taking off) to participants with aphasia and con-

trols. Sentences contained verbs that were either biased towards sentential complements (e.g. admit) or

towards direct objects (e.g. accept). The complementizer (that) was either present or absent, and with-

out the complementizer, the initial noun phrase in the embedded sentence (e.g. the airplane) can also be

interpreted as a direct object at the time it is presented. This ambiguity adds processing cost which re-

sults in increased reading time. In controls, reading speed was indeed mostly affected by the presence or

absence of the complementizer. Because syntactic interpretation of function words is often impaired in

aphasia, reading speed was less affected by the presence of absence of the complementizer, and rather

depended on verb bias violation. Where there was a mismatch between verb bias and sentence struc-

ture (e.g. The talented photographer accepted that the fire could have been prevented), reading speed

declined. When using the same paradigm to investigate reading of transitive and intransitive sentences,

DeDe (2013b) found that processing was slower in intransitive sentences when the verb had a transitive

bias.

Verb bias effects extend to passive sentences. The typical profile of a listener with syntactic impairment

is reported to be intact processing of canonical actives (The man kills the lion), but impaired processing

of passives (The lion is killed by the man). Common theories of aphasia regard non-canonical sentences,

such as the passive, as more difficult since they are transformed from an underlying canonical order (e.g.

Grodzinsky, 2000; Mauner et al., 1993). However, in plausibility judgment tasks (Gahl, 2002; Gahl et al.,

2003) participants with aphasia made fewer errors on passives and more errors on transitive actives

when stimuli contained verbs which were passive-dominant, e.g. elect or injure. This observation chal-

lenges rule-based approaches, since these would predicted all passive forms become unavailable. Fre-

quency can also apply to differences between passive processing across speakers of different languages.

In Standard Indonesian, where the passive is very frequent in comparisons to languages such as English,

individuals with Broca’s type aphasia are less likely to show a selective impairment of passive processing

(Jap, Martinez-Ferreiro, & Bastiaanse, 2016).

Not all aphasic speakers with grammatical impairment have more difficulties with actives than with ca-

nonical sentences impairment of passives in relation to canonical sentences (Berndt & Caramazza, 1999;

Caramazza, Capasso, Capitani, & Miceli, 2005; Caramazza, Capitani, Rey, & Berndt, 2001). In some indi-

6

viduals passives have been found to be processed more successfully. In their review of several sentence-

picture matching studies, Berndt, Mitchum and Haendiges (1996) identified a number of individuals with

aphasia who performed better at matching passive sentences to their corresponding pictures than active

sentences. While in some of these cases the pattern may be explained by random effects in the data,

particularly when trial numbers are small (Drai & Grodzinsky, 2006), two case studies have documented

this profile over multiple experiments. Druks and Marshall (1995) report the case of BM, a man with

post-stroke aphasia and, over several sentence-picture matching tasks, performed at chance on simple

actives, active questions and active existentials, but significantly above chance on their passive counter-

parts. Another case is WR, a man with primary progressive aphasia, who performed at chance on simple

actives and truncated actives with an auxiliary (e.g. The man is washing), but at ceiling on passives and

truncated passives (Zimmerer, Dąbrowska, Romanowski, Blank, & Varley, 2014). If passives represent a

canonical structure transformed into a non-canonical structure by syntactic movement, it should not be

possible for an individual who cannot interpret the canonical form to display residual capacity on the

non-canonical derivative.

3. The syntax-lexicon continuum (SLC)

The evidence is strong that these phenomena are not merely noise, but rather point towards systematic

properties of aphasic language which have not been sufficiently addressed by generative words-and-

rules approaches. This tension has been noted in virtually all of the articles reviewed above, and some

solutions have been proposed. Van Lancker-Sidtis and colleagues suggest a dual processing model (Van

Lancker Sidtis, 2012; Van Lancker Sidtis & Yang, 2016), which consists of a generative words-and-rules

processing route underpinned by left-hemisphere mechanisms, and a second for “prefabricated” (i.e.

formulaic) utterances which involves right-hemisphere and subcortical networks. While such dual-model

might explain formulaic language, it does not account for phenomena such as verb frame bias which is

neither explained by words and rules, nor by whole-form storage. Others have suggested an integrated

processing model, and Construction Grammar frameworks in particular (Conklin & Schmitt, 2012;

Hatchard & Lieven, under review; Wray, 2012; Zimmerer et al., 2018; Zimmerer et al., 2014). However,

where research reports have described the framework, they did it only in relation to specific research

questions of the study. We provide a broader overview with a specific focus on the SLC, which lies at the

core of the model.

7

The SLC was first suggested to accommodate idiomatic constructions within theories of language pro-

cessing. Idioms such as kick the bucket have for a long time been a challenge to generative frameworks

since their meaning (e.g. “to die”) cannot be derived from the meaning of the individual words combined

with a phrase structure rule. At a syntactic level, many do not allow any transformation without losing

their idiomatic meaning (e.g. the bucket was kicked), and sometimes they are not grammatically “well-

formed” in a formal sense (e.g. all of a sudden; see Chafe, 1968). However, more productive construc-

tions have also been identified as challenging: Fillmore, Kay and O’Connor (1988) found that the let alone

construction (e.g. I couldn’t get up to eat lunch, let alone cook breakfast) is specific enough in its syntac-

tic and semantic structure not to be definable by general phrase structure rules, and yet productive

enough not to be an idiom. Their conclusion was that it must be represented by a “mini-grammar” with

some idiomaticity but also remarkable generative power. They listed further examples for specific, but

productive constructions, such as the Xer, the Yer (e.g. the more, the better).

As in aphasia, the question following from such observations is whether such cases are rare exceptions,

or whether they are systematic and prevalent enough to justify the search for an alternative to words

and rules theories. Construction Grammar (Croft, 2001; Fillmore, Kay, Michaelis, & Sag, 2003; Goldberg,

2003, 2006, 2019) is such an alternative, and its introduction of the SLC is a means to capture and de-

scribe all observable language. The SLC is constructional space, commonly characterized along two prop-

erty dimensions. The first spans from simple to complex, with single words and morphemes at one end

of the spectrum and complex multiword phrases or utterances at the other. The other dimension ranges

from concrete to abstract and describes the degree to which a construction is lexically specific. One ex-

ample of a simple and concrete construction is a single word (e.g. house), while a simple and abstract

construction might be a grammatical category such as noun. Complex and concrete utterances are lexical

idioms such as get a life, or conversational formulas like I don’t know. Complex and abstract construc-

tions involve schema with open slots for insertion of words or phrases, such as the Xer, the Yer, which

can be filled by comparative adjectives or adjectival phrases, e.g. The earlier we arrive, the better, or Will

you X, which can be filled by any verb phrase headed by a bare verb e.g. Will you join us? Fully abstract

complex constructions such as the transitive (SVO), the passive and other argument structure construc-

tions consist entirely of slots. Because of the two dimensions of complexity and abstractness, we present

the continuum as a coordinate system (Figure 1).

Each construction is a form-meaning pairing, that is to say that not only words and idioms have semantic

and pragmatic functions, but also abstract constructions. The abstract transitive construction, for exam-

ple, has a very general meaning. It prototypically describes a causative event (Hopper & Thompson,

8

1980; Lakoff, 1977; Rice, 1987) involving a volitional agent acting and affecting a patient, while the func-

tion of the passive is to highlight the patient and/or to mask the agent. Construction Grammar therefore

offers a unifying account of lexicon, grammar, semantics and pragmatics.

One core property of language is its creativity, an in particular the ability to generate an infinite number

of novel expressions. In Construction Grammar this happens by combining constructions, and two com-

binatorial operations have been proposed (Dąbrowska, 2014; Dąbrowska & Lieven, 2005):

(1) Superimposition fuses constructions. This can involve inserting words into slots, e.g. the noun

glass into the where’s the NOUN construction. It can apply recursively: for example, buy some

NOUN and the noun milk can be superimposed to create buy some milk, and the result can be

superimposed over the VP slot in Do you want to VP to produce the expression Do you want to

buy some milk?

(2) Juxtaposition involves coordination of constructions without fusion, e.g. combining where are

you? and baby to Where are you, baby? or Baby, where are you?

These are more concrete examples, but note that constructions can be complex and abstract (Figure 1),

allowing more creative utterances such as She smiled herself an upgrade (Douglas Adams as cited in

Goldberg, 2003). However, typical speech is mostly formulaic, suggesting that speakers are biased to-

wards concrete constructions (Dąbrowska, 2014).

9

Figure 1. Examples for constructions arranged along the syntax-lexicon continuum. Constructions vary in

size along the x-axis, and in abstraction along the y-axis. Single words are concrete and simple; lexicalized

(formulaic) expressions and inflected forms (including some regular plurals) are concrete and more com-

plex. Word classes are abstract; Schemas with slots are abstract and more complex. For example, the

idiom The die is cast and [SubjNP] [aux] [VPP] are instances of the passive with different levels of ab-

stractness.

There are formal criteria for determining whether a particular expression or pattern is a construction: it

may be idiomatic in a semantic sense, or their grammatical structure can be exceptional. However, prop-

erties such as frequency, age of acquisition and conversational functions are other important factors.

Utterances which are common and/or learned early in life are more likely to be lexically specific (Arnon,

McCauley, & Christiansen, 2017; Siyanova-Chanturia, 2015; Siyanova-Chanturia, Conklin, Caffarra, Kaan,

& van Heuven, 2017).

Importantly, the framework has no transformation rules. Canonical and non-canonical constructions are

separate units, meaning that passives are not derived from underlying structures in which the object

car

I can’t remember

N.

oh my god

fond of [N.]

blue

Adj.

Number

Interjection

dog

It’s alright

I don’t know

Oh dear

thank you

jog [someone’s] memory

[Adj.] [N.]

I don’t [V]

[SubjNP] [V] [ObjNP] [SubjNP] [aux] [VPP] (by NP)

walkINFL the dog

SubjNP doINFL VP

SubjNP beINFL injured [Number] [N.]

10

follows the verb. Rather, constructions are associated with another through similarities in form and/or

function. The English causative transitive (The man kills the lion) is related to transitives which encode

other meanings, such as mental experience (I saw a python) and possession (I have a black cat). Further,

one construction can also be a concrete instance of another, e.g. the formula I don’t know is an instance

of the intransitive construction as well as of the negation construction. Constructions can also have a

nested relationship. For example, the noun phrase construction is part of a great range of more complex

constructions.

4. The SLC in aphasia

The SLC describes a linguistic system that allows the generation of an infinite number of different sen-

tences, but often works with a limited possibility space as it relies on more concrete schemas and formu-

las. Abstract complex constructions have the greatest creative power but pose the highest demands to

lexical and syntactic networks as each morpheme needs to be accurately retrieved and integrated. Any

damage to lexical or grammatical systems therefore is more likely to impair the ability to process rare or

novel combinations and restrict the system to more concrete, lexicalized constructions (formulas). These

can be formulas and idioms, but also lexicalized derivations, such as the plural form of plural-dominant

nouns. In the framework, shoe as well as shoes would be considered lexicalized forms. Pluralization er-

rors are caused by retrieval of the incorrect, but more frequent, word form. Grammatical frame biases

are represented by partially lexicalized schemas with a lexicalized verb and slots for other constituents.

Such representations develop on the basis of distributions in language use, such as encountering a verb

primarily in a specific frame. For example, the bias of the verb admit towards complementizer sentences

(to which people with aphasia show sensitivity) can be represented as SubjNP admitINFL CP.

To demonstrate the framework further, we apply it to the passive that received much attention in apha-

sia research. In construction grammar and within the SLC, the full event passive (The lion is killed by the

man), which is commonly tested in aphasia batteries, is an abstract, complex construction. The passive is

used when the patient of the action is the topic and the agent is backgrounded. In Riches’ (2013) applica-

tion of Construction Grammar to developmental language disorder, he lists it as part of a larger construc-

tion network which includes other types of passives and “passive-like” constructions, such as short (trun-

cated) passives (The bread was eaten), adjectival passives (The vase is broken), get passives (Mary got

stung), the want something done construction (I want the roof fixed), the by-phrase (This picture is by

John) and the present perfect aspect (I have fixed the roof).

11

Instances of the passive can also be partially or fully lexicalized. They include formulaic passives such as

the die is cast as well as passive verb frames such as NP was born. The passive bias of a verb like injure

(Gahl et al., 2003) is explained by a passive schema (NP BE injured). By contrast, active sentences that

contain verbs with a passive bias require them to be inserted into more abstract frames (e.g. SubjNP V

ObjNP). The framework still assumes that passives are harder, and more likely to be affected by aphasia,

than actives. However, this is not because passives involve transformation from the canonical word or-

der. Rather, they are less frequent (Roland, Dick, & Elman, 2007) and learned later in life (Brooks &

Tomasello, 1999). However, these factors can vary from one linguistic community to another, and in

cultures where the passive is more frequent, processing of passives is less likely to be impaired in aphasia

(Jap et al., 2016). Individuals within a culture may also differ in the frequency with which they encounter

passives. For instance, since passives are considerably more frequent in writing than in speech, we could

expect that people who read more are better at understanding passives – a prediction which was con-

firmed by Street and Dąbrowska (2014).

However, even in English it is possible for some individuals to have intact passive processing, and at the

same time, impaired active processing (Druks & Marshall, 1995; Zimmerer et al., 2014). The SLC frame-

work is compatible with these cases because it sees actives and passives as independent constructions

rather than the passive being derived from canonical actives. It is not clear how large this group of

speakers is, nor is their impairment of actives sufficiently understood. One possible explanation for the

latter is an impairment of sequence order processing, which would affect the capacity to interpret the

order of constituents in a sentence. This would cause greater difficulties for processing of English actives,

in which order determines semantic roles (e.g. The lion kills the man vs. The man kills the lion), than of

the morphologically marked passive (The man is killed by the lion vs. By the lion the man is killed). Zim-

merer and Varley (2015) demonstrated that WR, whose comprehension of actives was severely impaired

but who successfully interpreted passives, showed no sensitivity to the order of visual stimulus sequenc-

es in an artificial grammar learning task.

The construction by NP is nested within the full passive. It serves prototypically as an agentive marker

and is also used in other constructions in addition to the passive, for instance in action nominalizations

(e.g. a strike by lorry drivers) or to mark the author of a text of work of art (The Black Cloud by Fred

Hoyle). WR interpreted the by-phrase as an agentive marker even when it occurred within an active sen-

tence and encoded spatial information. For example, when presented with the sentence The man by the

woman shot the rabbit, he assigned agency to the woman (Zimmerer et al., 2014).

12

5. Future research: aphasic profiles, neurological networks, processing theories and clinical implica-

tions

We are at early stages of applying constructionist, usage-based frameworks to aphasia. The reviewed

research is very mixed with regards to aphasia types, methods and aspects of processing. One major task

is to understand whether aphasic symptoms can be mapped to impairments of specific parts of the con-

structicon, and how distinction among the dimensions of complexity and abstractness can help in map-

ping out aphasic language. Because individual words are also represented as part of some complex con-

structions, grammatical and lexical behaviour is hard to disentangle. This is reflected by observations

that neither lexical nor grammatical impairment occur in isolation in aphasia (Bates & Goodman, 1997;

Dick et al., 2001).

Impairment can affect the ability to combine constructions (superimposition, juxtaposition). Data from

language acquisition suggest that two-year-olds first produce utterances that require either no or one

combination, and only with increasing age and experience become able to combine more constructions

at a time (Dąbrowska, 2014; Dąbrowska & Lieven, 2005). Such limitation may also affect people with

aphasia. It is possible that the combinatorial operations become entirely unavailable, or that processing

constraints limit the number of combinations that can be carried out for a given utterance. In principle,

accounts of reduced processing capacity (Caplan, Waters, DeDe, Michaud, & Reddy, 2007; Kolk &

Heeschen, 1990) are applicable to Construction Grammar, but they require some modification. Memory

demands of a given utterance cannot be determined just by analyzing its formal syntactic structure. One

needs to take into account whether constructions are frequent and/or idiomatic.

There is only a small amount of work that tests the SLC against neurocognitive data. Influential neurolog-

ical models were designed around the distinction between words and rules and are difficult to apply to

Construction Grammar. For example, Ullman (2001, 2004, 2016) describes a model of declarative and

procedural memory in language processing. He associates declarative memory with word knowledge

(form-meaning pairing), and identifies regions in the medial temporal lobe and surrounding structures as

central to its processing. Procedural memory on the other hand is associated with grammatical rules and

regions in the frontal lobe, basal-ganglia, parietal lobe and cerebellum. This strict separation between

meaning and abstract grammatical knowledge clashes with the notion in Construction Grammar that

even fully abstract constructions are form-meaning pairings and therefore have a declarative element.

The only clearly procedural operations in the framework are combinatorial, i.e., superimposition and

juxtaposition.

13

In their review of neurophysiological evidence, Pulvermüller, Cappelle and Shtyrov (2013) conclude that

there is evidence for lexicalized schemas, but caution against seeing a continuum between words and

these more complex representations. They argue that because mismatch negativity is reduced by lexical

violations (pseudowords), but increased by grammatical violations (see also Pulvermüller & Shtyrov,

2006), lexical and grammatical processing are fundamentally different. In contrast, Dam and Desai (2016)

propose that abstract constructions, just like words, may be part of an embodied linguistic network. In

their fMRI study they found that (di-)transitive constructions, which are associated with caused motion

(e.g. You threw the ball to her) involve regions associated with reaching and grasping even when the verb

is abstract (e.g. He communicated the news to you). Blank, Balewski, Mahowald and Fedorenko (2016)

found upregulation of the whole perisylvian region in response to grammatical complexity, suggesting a

blurring of distinctions between lexicon and grammar.

Clinical implications could be substantial. One crucial variable in this review has been usage-frequency,

which can be measured quickly using computerized and corpus-based tools. It is possible to determine

the degree to which an individual relies on frequent words or strong collocates. Impairment may be

traceable as a deviation from normative samples. The Frequency in Language Analysis Tool (Zimmerer,

Wibrow, & Coleman, 2017) was built for that purpose and can distinguish not only aphasic individuals,

but also people with Alzheimer’s disease from neurotypical speakers (Bruns et al., 2018; Zimmerer et al.,

2018; Zimmerer, Wibrow, & Varley, 2016). Speakers in clinical groups used more frequent and strongly

collocated language forms. Tools like these can assist clinicians in making more accurate, and in case of

neurodegenerative diseases, earlier, diagnoses. They may also be used to track change as the result of

intervention or disease progression.

One strand of intervention aims to compensate for language impairment by adjusting the behaviour of

non-aphasic communication partners in a way that it becomes easier for the person with aphasia (e.g.

Simmons-Mackie, 2018). Communication guidelines typically suggest use of simpler sentences. While

this is typically interpreted as a call for simpler syntactic structures, the evidence suggests that frequency

effects need to be considered. Common utterances, such as formulas or sentences in which the verb

occurs in the preferred grammatical frame are easier for a person with aphasia. The effectiveness of

communicating using “common language” has not been tested, but it is possible that future communica-

tion guidelines will find it relevant.

Intervention can utilize the concept of networked constructions. In the case in which the individual is

unable to apply abstract constructions, one could assess which concrete constructions, such as formulas

and verb-specific frames, are available, and use them as a basis to build up more high-level representa-

14

tions. Another approach is to work on a target construction by training related, but simpler and perhaps

more intact constructions, such as the adjectival passive (The vase is broken) and the present perfect (I

have fixed the roof) aspect in order to support the event passive (Riches, 2013). Further, one would pre-

dict that training would benefit from practice items which place the target construction within a com-

municative context in which it is commonly used. With regards to word-retrieval therapy, the framework

suggests supplementing lexical intervention with frames within which the target word occurs frequently.

Construction Grammar and the SLC can help explain aphasic data, and in turn research on aphasia can

also be used to test and inform the theory. Construction Grammar motivates new research questions.

While in generative frameworks there is much focus on how stored representations (words, morphemes)

are combined, here, one critical question is what the stored units are, especially at the multiword level.

Which utterances rely more on abstract schemas, and which on concrete constructions? As demonstrat-

ed in this article, an impaired linguistic system with reduced creativity can provide evidence for what

counts as a construction, and how constructions may be associated or dissociated from another.

References

Arnon, I., McCauley, S. M., & Christiansen, M. H. (2017). Digging up the building blocks of language: Age-

of-acquisition effects for multiword phrases. Journal of Memory and Language, 92, 265–280.

http://doi.org/10.1016/j.jml.2016.07.004

Avrutin, S. (2000). Comprehension of Discourse-Linked and Non-Discourse-Linked Questions by Children

and Broca’s Aphasics. Language and the Brain, 295–313. http://doi.org/10.1016/B978-012304260-

6/50017-7

Bates, E., & Goodman, J. (1997). On the inseperability of grammar and the lexicon Evidence from

acquisition, aphasia, and real-time processing. Language and Cognitive Processes, 12, 507–584.

Berndt, R. S., & Caramazza, A. (1999). How “regular” is sentence comprehension in Broca’s aphasia? It

depends on how you select the patients. Brain and Language, 67(3), 242–247. Retrieved from

http://www.sciencedirect.com/science/article/pii/S0093934X99921302

Berndt, R. S., Mitchum, C. C., & Haendiges, A. N. (1996). Comprehension of reversible sentences in

“agrammatism”: a meta-analysis. Cognition, 58(3), 289–308. Retrieved from

http://www.sciencedirect.com/science/article/pii/0010027795006826

15

Biedermann, B., Lorenz, A., Beyersmann, E., & Nickels, L. (2013). The influence of plural dominance in

aphasic word production. Aphasiology, 26(8), 985–1004.

Blank, I., Balewski, Z., Mahowald, K., & Fedorenko, E. (2016). Syntactic processing is distributed across

the language system. NeuroImage, 127, 307–323.

http://doi.org/10.1016/j.neuroimage.2015.11.069

Brooks, P. J., & Tomasello, M. (1999). Learning the English passive construction. In B. Fox, D. Juracsky, &

L. Moichaelis (Eds.), Cognition and Function in Language. Stanford, CA: CSLI Publication.

Bruns, C., Varley, R. A., Zimmerer, V. C., Carragher, M., Brekelmans, G., & Beeke, S. (2018). “I don’t

know”: a usage-based approach to familiar collocations in non-fluent aphasia. Aphasiology, 33(2),

140–162. http://doi.org/10.1080/02687038.2018.1535692

Caplan, D., Waters, G. S., DeDe, G., Michaud, J., & Reddy, A. (2007). A study of syntactic processing in

aphasia I: Behavioral (psycholinguistic) aspects. Brain and Language, 101(2), 103–150. Retrieved

from http://www.sciencedirect.com/science/article/pii/S0093934X06002409

Caramazza, A., Capasso, R., Capitani, E., & Miceli, G. (2005). Patterns of comprehension performance in

agrammatic Broca’s aphasia: A test of the Trace Deletion Hypothesis. Brain and Language, 94(1),

43–53. Retrieved from http://www.sciencedirect.com/science/article/pii/S0093934X04002986

Caramazza, A., Capitani, E., Rey, A., & Berndt, R. S. (2001). Agrammatic Broca’s Aphasia Is Not Associated

with a Single Pattern of Comprehension Performance. Brain and Language, 76(2), 158–184.

http://doi.org/doi:10.1006/brln.1999.2275

Caramazza, A., & Zurif, E. B. (1976). Dissociation of Algorithmic and Heuristic Processes in Language

Comprehension: Evidence from Aphasia. Brain and Language, 3, 572–582.

http://doi.org/https://doi.org/10.1016/0093-934X(76)90048-1

Chafe, W. L. (1968). Idiomaticity as an anomaly in the Chomskyan paradigm. Foundations of Language,

4(2), 109–127.

Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.

Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris.

Code, C. (1982). Neurolinguistic analysis of recurrent utterance in aphasia. Cortex, 18, 141–152.

http://doi.org/10.1016/S0010-9452(82)80025-7

16

Code, C. (1989). Speech automatisms and recurring utterances. In C. Code (Ed.), The Characteristics of

Aphasia (pp. 155–163). Hove and London: Laurence Erlbaum.

Code, C. (2013). Letter to the editor and author’s response: Did Leborgne have one or two speech

automatisms? Journal of the History of the Neurosciences, 22, 319–320.

Conklin, K., & Schmitt, N. (2012). The Processing of Formulaic Language. Annual Review of Applied

Linguistics, 32, 45–61. http://doi.org/10.1017/S0267190512000074

Croft, W. (2001). Radical construction grammar: Syntactic theory in typological perspective. Oxford:

Oxford University Press.

Dąbrowska, E. (2014). Recycling utterances: A speaker’s guide to sentence processing. Cognitive

Linguistics, 25, 617–653.

Dąbrowska, E. (2015). Individual differences in grammatical knowledge. In Handbook of Cognitive

Linguistics (pp. 649–667). Berlin: De Gruyter Mouton.

Dąbrowska, E., & Lieven, E. (2005). Towards a lexically specific grammar of children’s question

constructions. Cognitive Linguistics, 16(3), 437–474.

http://doi.org/https://doi.org/10.1515/cogl.2005.16.3.437

Dam, W. O., & Desai, R. H. (2016). The semantics of syntax: The grounding of transitive and intransitive

constructions. Journal of Cognitive Neuroscience, 28(5), 693–709.

DeDe, G. (2013a). Effects of verb bias and syntactic ambiguity on reading in people with aphasia.

Aphasiology, 27(12), 1408–1425. http://doi.org/10.1080/02687038.2013.843151

DeDe, G. (2013b). Verb transitivity bias affects on-line sentence reading in people with aphasia.

Aphasiology, 27(3), 326–343. http://doi.org/10.1080/02687038.2012.725243

Dick, F., Bates, E., Wulfeck, B., Utman, J. A., Dronkers, N., & Gernsbacher, M. A. (2001). Language deficits,

localization, and grammar: evidence for a distributive model of language breakdown in aphasic

patients and neurologically intact individuals. Psychological Review, 108(4), 759–88. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/11699116

Drai, D., & Grodzinsky, Y. (2006). A new empirical angle on the variability debate: Quantitative

neurosyntactic analyses of a large data set from Brocaâ€TMs Aphasia. Brain and Language, 96(2),

117–128. Retrieved from http://www.sciencedirect.com/science/article/pii/S0093934X05001161

17

Druks, J., & Marshall, J. C. (1995). When passives are easier than actives: two case studies of aphasic

comprehension. Cognition, 55, 311–331. http://doi.org/https://doi.org/10.1016/0010-

0277(94)00651-Z

Ellis, N., Römer, U., & O’Donnell, M. B. (2016). Usage-based Approaches to Language Acquisition and

Processing: Cognitive and Corpus Investigations of Construction Grammar. Malden, MA: Wiley-

Blackwell.

Erman, B., & Warren, B. (2000). The idiom principle and the open choice principle. Text - Interdisciplinary

Journal for the Study of Discourse, 20(1), 29–62.

Fillmore, C. J., Kay, P., Michaelis, L. A., & Sag, I. A. (2003). Construction Grammar. Stanford, CA: CSLI

Publications.

Fillmore, C. J., Kay, P., & O’Connor, M. C. (1988). Regularity and idiomaticity in grammatical

constructions: The case of let alone. Language, 64(3), 501–538.

Friedmann, N., & Grodzinsky, Y. (1997). Tense and Agreement in Agrammatic Production: Pruning the

Syntactic Tree. Brain and Language, 56(3), 397–425. http://doi.org/10.1006/brln.1997.1795

Gahl, S. (2002). Lexical biases in aphasic sentence comprehension: An experimental and corpus linguistic

study. Aphasiology, 16(12), 1173–1198. http://doi.org/10.1080/02687030244000428

Gahl, S., Menn, L., Ramsberger, G., Juracsky, D., Elder, E., Rewega, M., & Audrey, L. H. (2003). Syntactic

frame and verb bias in aphasia: Plausibility judgments of undergoer-subject sentences. Brain and

Cognition, 53, 223–228. http://doi.org/https://doi.org/10.1016/S0278-2626(03)00114-3

Goldberg, A. (2019). Explain Me This. Princeton, New Jersey: Princeton University Press.

Goldberg, A. E. (2003). Constructions: a new theoretical approach to language. TRENDS in Cognitive

Sciences, 7(5), 219–224. http://doi.org/10.1016/S1364-6613(03)00080-9

Goldberg, A. E. (2006). Constructions at work: The nature of generalization in language. New York:

Oxford University Press.

Goodglass, H., Christiansen, J. A., & Gallagher, R. E. (1994). Syntactic constructions used by agrammatic

speakers: Comparison with conduction aphasics and normals. Neuropsychology, 8(4), 598–613.

Grodzinsky, Y. (2000). The Neurology of Syntax: Language Use Without Broca’s Area. Behavioral and

Brain Sciences, 23, 1–71. http://doi.org/10.1017/S0140525X00002399

18

Hatchard, R. (2015). A construction-based approach to spoken language in aphasia. University of

Sheffield. Retrieved from etheses.whiterose.ac.uk/10385

Hatchard, R., & Lieven, E. (n.d.). Inflection of nouns for grammatical number in spoken narratives by

people with aphasia: A challenge for rule-based models. Language and Cognition.

Hopper, P., & Thompson, S. A. (1980). Transivity in grammar and discourse. Language, 56(251–299).

Jackson, H. (1874). On the nature of duality in the brain. Brain, 38, 80–86.

Jap, B. A., Martinez-Ferreiro, S., & Bastiaanse, R. (2016). The effect of syntactic frequency on sentence

comprehension in standard Indonesian Broca’s aphasia. Aphasiology, 30(11), 1325–1340.

http://doi.org/10.1080/02687038.2016.1148902

Kolk, H., & Heeschen, C. (1990). Adaptation symptoms and impairment symptoms in Broca’s aphasia.

Aphasiology, 4(3), 221–231. http://doi.org/10.1080/02687039008249075

Lakoff, G. (1977). Linguistic gestalts. Chicago Linguistic Society, 13, 236–287.

Marcus, G. F. (1995). Children’s overregularization of English plurals: a quantitative analysis. Journal of

Child Language, 22(02), 447–459. http://doi.org/10.1017/S0305000900009879

Mauner, G., Fromkin, V. A., & Cornell, T. L. (1993). Comprehension and acceptability judgments in

agrammatism: Disruptuions in the syntax of referencial dependency. Brain and Language, 45, 340–

370.

Pulvermüller, F., Cappelle, B., & Shtyrov, Y. (2013). Brain basis of meaning, words, construction and

grammar. In T. Hoffmann & G. Trousdale (Eds.), The Oxford Handbook of Construction Grammar

(pp. 397–416). New York, NY: Oxford University Press.

Pulvermüller, F., & Shtyrov, Y. (2006). Language outside the focus of attention: The mismatch negativity

as a tool for studying higher cognitive processes. Progress in Neurobiology, 79(1), 49–71.

http://doi.org/10.1016/j.pneurobio.2006.04.004

Rice, S. A. (1987). Towards a cognitive model of transitivity. University of California, San Diego.

Riches, N. G. (2013). Treating the passive in children with specific language impairment: A usage-based

approach. Child Language Teaching and Therapy, 0265659012466667-.

http://doi.org/10.1177/0265659012466667

Roland, D., Dick, F., & Elman, J. L. (2007). Frequency of basic English grammatical structures: A corpus

19

analysis. Journal of Memory and Language, 57, 348–379.

Sidtis, D. V. L., & Yang, S. (2017). Formulaic language performance in left- and right-hemisphere damaged

patients: structured testing. Aphasiology, 31(1), 82–99.

http://doi.org/10.1080/02687038.2016.1157136

Simmons-Mackie, N. (2018). Communication partner training in aphasia: reflections on communication

accommodation theory. Aphasiology, 32(10), 1215–1224.

http://doi.org/10.1080/02687038.2018.1428282

Siyanova-Chanturia, A. (2015). On the “holistic” nature of formulaic language. Corpus Linguistics and

Linguistic Theory, 11(2), 285–301.

Siyanova-Chanturia, A., Conklin, K., Caffarra, S., Kaan, E., & van Heuven, W. J. B. (2017). Representation

and processing of multi-word expressions in the brain. Brain and Language, 175, 111–122.

http://doi.org/10.1016/j.bandl.2017.10.004

Street, J. A., & Dąbrowska, E. (2014). Lexically specific knowledge and individual difference in adult native

speakers’ processing of the English passive. Applied Psycholinguistics, 35(1), 97–118.

Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition.

Cambridge, MA: Harvard University Press.

Ullman, M. T. (2001). A neurocognitive perspective on language: the declarative/procedural model.

Nature Reviews. Neuroscience, 2(10), 717–26. http://doi.org/10.1038/35094573

Ullman, M. T. (2004). Contributions of memory circuits to language: the declarative/procedural model.

Cognition, 92, 231–270.

Ullman, M. T. (2016). The Declarative/Procedural Model: A Neurobiological Model of Language Learning,

Knowledge, and Use. http://doi.org/10.1016/B978-0-12-407794-2.00076-6

van Lancker-Sidtis, D., & Rallon, G. (2004). Tracking the incidence of formulaic expressions in everyday

speech: methods for classification and verification. Language & Communication, 24, 207–240.

Van Lancker Sidtis, D. (2012). Formulaic Language and Language Disorders. Annual Review of Applied

Linguistics, 32, 62–80. http://doi.org/10.1017/S0267190512000104

Van Lancker Sidtis, D., & Postman, W. A. (2006). Formulaic expressions in spontaneous speech of left-

and right-hemisphere-damaged subjects. Aphasiology, 20(5), 411–426.

20

http://doi.org/10.1080/02687030500538148

Van Lancker Sidtis, D., & Yang, S. (2016). Formulaic language performance in left- and right-hemisphere

damaged patients: structured testing. Aphasiology, 31(March), 1–18.

http://doi.org/10.1080/02687038.2016.1157136

Varley, R., & Zimmerer, V. (2018). Language impairments in acquired aphasia: Features and frameworks.

In A. Bar-On & D. Ravid (Eds.), Handbook of communication disorders: Theoretical, empirical and

applied linguistic perspectives (pp. 881–897). Boston/Berlin: Walter de Gruyter.

Wray, A. (2012). What do we (think we) know about formulaic language? An evaluation of the current

state of play. Annual Review of Applied Linguistics, 32, 231–254.

http://doi.org/10.1017/S026719051200013X

Zimmerer, V. C., Dąbrowska, E., Romanowski, C. A. J., Blank, C., & Varley, R. A. (2014). Preservation of

passive constructions in a patient with primary progressive aphasia. Cortex, 50, 7–18.

http://doi.org/https://doi.org/10.1016/j.cortex.2013.09.007

Zimmerer, V. C., Newman, L., Thomson, R., Coleman, M. J., & Varley, R. A. (2018). Automated analysis of

language production in aphasia and right hemisphere damage: Frequency and collocation strength.

Aphasiology, 32(11), 1267–1283. http://doi.org/10.1080/02687038.2018.1497138

Zimmerer, V. C., & Varley, R. A. (2015). A case of “order insensitivity”? Natural and artificial language

processing in a man with primary progressive aphasia. Cortex, 69.

http://doi.org/10.1016/j.cortex.2015.05.006

Zimmerer, V. C., Wibrow, M., & Varley, R. A. (2016). Formulaic language in people with probable

Alzheimer’s Disease: a frequency-based approach. Journal of Alzheimer’s Disease, 53, 1145–1160.

http://doi.org/10.3233/JAD-160099

Zurif, E. B., Caramazza, A., & Myerson, R. (1972). Grammatical judgments of agrammatic aphasics.

Neuropsychologia, 10, 405–417.