Appunti Neurolinguistics

SUMMARY

Young controls and age-matched speakers showed VAS priming effects in transitive sentences but

not in unaccusative sentences.

PWA showed VAS priming effects in both transitive and unaccusative sentences.

GENERAL DISCUSSION

Speakers with agrammatic aphasia used different planning units from control speakers as a function

of the nature of GE processes required for sentence production.

28 of 76

  This suggests that the time courses of lexicalisation and functional structure generation processes

are affected differentially in agrammatic production, refining the impaired GE account.

Speakers with agrammatic aphasia did not plan sentences in a smaller linguistic unit compared with

control speakers. They engaged in a larger planning unit as the complexity of the task increased in

experiment 2 compared with experiment 1.

What goes wrong during passive sentence production in agrammatic aphasia ? : an eye-

tracking study. (Cho & Thompson, 2010)

13/10/21 - Lesson 6

LANGUAGE DEFICIT IN APHASIA

• Everyone agrees that there is reduced language function in aphasia, and that this has

consequences for communication

• Where the differences are: in characterising the nature of impairment:

1. Deficit in the linguistic representation

2. Adaptation to reduced cognitive resources

3. Differences in language processing.

TREATMENT APPROACHES

Where the differences are is whether you address the impairment (deficit in language functions) or

the consequences for communication (functionalist approach, life partification approaches).

Goal of the Study: to examine passive sentence production in agrammatic aphasia under conditions

of structural priming using eye-tracking while speaking. And to uncover whether different

production mechanisms underlie correct and incorrect production.

Lit Review: Passives in Aphasia

• Caplan & Hanna (1998); Faroqi-Shah & Thompson (2003); Menn et al. (1998), Weinrich, Boser,

McCall & Bishop (2001): problems to produce appropriate passive morphology (i.e., aux, -ed.

Preposition by).

• Caplan & Hanna (1998): 14 Broca’s aphasic speakers. Picture description of transitive events,

cued the agent or the theme. Verb was given.

• Errors: the bag give the clerk instead of the bag was given by the clerk.

• Faroqi-Shah and Thompson (2003): 50% errors for passive targets were role reversals (RRs), and

63% of these errors had errors in aux and/or prepositions

• HP: difficulty in producing passive sentences -> due to impaired grammatical morphology, not

impaired thematic role assignment.

• Bastiaanse and Edwards (2004): anagram task.

They arranged words in the order corresponding to a given picture (e.g., the child / is washed by

• / the mother).

PWA produced Reverse Role (RR) errors with passive targets.

• 29 of 76

  B&E: difficulty in mapping theta roles onto grammatical roles even passive morphology did not

• play a role in this task.

Methods note: this is a good production task for speakers who may also have dysarthria or

• articulation difficulties, but also non-native speakers when you want to focus on production and

not articulation.

• Caplan & Hanna (1998), Faroqi-Shah & Thompson (2003) methods: cued first noun to start with

making RR less likely.

If RRs in passives are by-products of impaired passive morphology, the number of RRs should

• decrease if you use an experimental paradigm that facilitates passive morphology.

SCORING

Like all scoring in production, you need criteria. When there are errors and omissions of function

words or grammatical morphology, it can be difficult to determine whether something is an active

with RR or a passive.

Bastiaanse et al. (2003): the boxer is punched by the bartender for the bartender is punched by the

boxer.

Faroqi-Shah & Thompson (2003): the bartender is punching the boxer for the bartender is punched

by the boxer.

Inconsistencies across studies is inevitable, as coring criteria have to be decided each time, based on

the study. This also highlights how much consideration you need to give to scoring and

interpretation with any production data.

So, the boxer is punching the bartender fo the bartender is punched by the boxer is scored as a RR

in some studies but not in others.

Off-line Production

There are limitations of off-line production data, because there is an ambiguity in interpreting

errors.

So the solution is that of introducing online measures as well: eye-tracking + speaking, or speech

onset latencies, provide additional measures to understand underlying processes in sentence

production in real-time

On-line Production

• In structurally guided production, before speech, speakers comprehend an event (“who is doing

what to whom”), and generate a corresponding message.

• You should be able to see this in passive production prior to speech.

* it’s in italics because this paper comes before the one of last week which explains.

Eye-movements During Speech

Speakers gaze at each referent in a visual display prior to mentioning it. Longer gaze durations prior

to speech reflect sensitive to factors such as frequency and codability.

30 of 76

  The first eye-tracking study while speaking in aphasia was made by Thompson, Dickey, Cho, Lee &

Griffin in 2007.

They studies sentence production with transitive and ditransitive verbs. They found that PWA’s eye

movements were similar to controls for correctly produced sentences, but not for incorrect ones.

Specific Goals of the Study

We have two main questions:

1. Can impairments in passive sentence structure occur independently of impairments in passive

morphology?

2. Can online measures provide an insight into the different types of errors observed in offline

sentence production in speakers with aphasia?

PARTICIPANTS

There were 9 participants with stroke-induced aphasia (PWA), 12 age and educated matched control

+ 16 healthy young adults.

They were all native speakers of English. The only exclusion criteria decided was a prior

neurological or speech-language deficit.

Language Profile

Participants had mild to moderate non-fluent phase, relatively preserved auditory comprehension,

difficulty with non-canonical sentences compared to canonical sentences + -ed form was impaired

whereas the -ing form was relatively preserved.

MATERIALS

• 20 active and 20 passive prime sentences with accompanying pictures

• 40 target pictures of transitive reversible events

• Prime sentences were presented in written and auditory format

• Same verb in the prime and the target, different nouns.

• Verbs were regular.

It is always useful to carefully read well designed studies (that worked), for information about

procedures, details about location of stimuli, font, variables to control for etc.

Norming

It is also important to figure out whether your materials work or not. So you go through norming,

where you usually need around 10 participants who test your materials.

In this study, pictures were normed by asking native speakers to describe them. You want pictures

that yield target descriptions (transitive) and some proportion of passives.

PROCEDURE

First there was a familiarisation process for noun and verbs.

31 of 76

  The experiment was composed of 20 trials each. In the priming trial, participants viewed and

listened to the prime sentence and were asked to repeat it out lou.d

On the target trial, participants were asked to describe the target picture using the same structure as

the prime.

So, instructions were explicit (not to be confused with explicit training of the derivation of a

passive).

DATA ANALYSIS: scoring for structure

The sentence was scored as active if the structure was: DP agent, aux V-ing, DP theme

The sentence was scored as passive if the structure was: DP theme, aux V, PP-by (required aux -ed

and by).

Semantic substitutions were considered ok, as well as different DP.

Phonemic paraphasia was ok if 50% of target phonemes were correct.

Scoring for errors

The error was scored as grammatical morphology error if there were correct thematic roles but one

or more errors in morphology.

The RR error: two participants required and morphology indicated wrong role.

Sentence type error: active produced in the passive priming condition or viceversa.

Incorrect lexical items: non-semantically related lexical substitutions, verbs with different AS,

perseverations for the prime.

Other: everything else, or fragments.

Speech onset times

They considered only correct responses. They were studied for N1, V, N2. The time frame was from

onset of the picture to onset of each word. Speech onset was determined manually using Praat.

Eye gaze durations

Only correct responses were considered. The area of interest was decided around an event

participant. Fixation needed to last at least 100 ms. The gaze duration for each participant was

determined by the sum of all fixations in the area of interest.

Gaze durations aligned with onset of N1, onset of V and onset of N2.

RESULTS: sentence type 32 of 76

  Speech Onset

PWA took longer to produce passives with RR (mean latency = 6907 ms) than active-for-passive

descriptions (mean latency = 4219).

PWA took longer to produce actives-for-passives than to produce correct passives (mean latency =

5773).

Gaze duration to N1 in RRs, actives for passives and correct passives by region

Passives with RRs: longer looking times to N1 compared to correct passives both in the PRE N1

region and the N1-V region.

When producing actives-for-passives this effect only showed up in the N1-V region, not the PRE

N1 region.

[Table: gaze duration to N1 in RRs, and correct passives by region]

[Table: gaze duration to N1 in actives-for-passives and correct passives by region]

33 of 76

  SUMMARY AND DISCUSSION

Impairments in passive sentence structure can occur regardless of impairments in passive

morphology.

Under conditions of structural priming PWA were able to produce passive sentences with relatively

preserved passive morphology. Still they produced high RR error rates.

If the difficulty with passives were a consequence of difficulties with passive morphology we

wouldn’t expect RR errors.

So, RR errors in aphasia are not a consequence of difficulty accessing and producing passive

morphology.

Online measures showed different patterns in two common types of errors (RRs and actives-for-

passives).

Speech onset latencies were longer for RRs than for correct passives. PWA accessed the lemma but

mapped roles incorrectly. Speech onset times in actives-for-passives were shorter.

Gaze duration data showed that both RRs and actives-for-passives had longer gaze durations to N1

during speech compared to correct passives. Before speech, only RR sentences had longer gaze

durations to N1.

Producing complex sentences is computationally costly. Different aspect