A person recovering from stroke may speak slowly, pause frequently, struggle to pronounce words, or have difficulty finding the right word.
From the outside, these problems can look similar.
Clinically, however, they may arise from very different disorders:
aphasia, dysarthria, and apraxia of speech.
Understanding the difference matters because the metrics used to track improvement should also be different.
A speech-production measure such as articulation rate can be highly informative for one disorder while providing only limited information about another.
This distinction is particularly important as digital rehabilitation platforms begin using automated speech analysis to monitor communication longitudinally.
Dysarthria Is Primarily a Speech-Execution Disorder
Dysarthria is a group of neurogenic motor speech disorders caused by impaired control of the muscles involved in speech.
ASHA describes dysarthria as involving abnormalities in the strength, speed, range, steadiness, tone, or accuracy of the movements needed for respiration, phonation, resonance, articulation, and prosody. It can result from neurological conditions including stroke, traumatic brain injury, cerebral palsy, Parkinson disease, multiple sclerosis, ALS, and others.
The person's language may be intact.
They may know exactly what they want to say and understand the conversation perfectly, but the neuromuscular system responsible for producing intelligible speech does not execute the movement normally.
Depending on the type of dysarthria, clinicians may hear:
Slurred or imprecise articulation
Abnormally slow or fast speech
Reduced loudness
Abnormal voice quality
Reduced or excessive pitch variation
Irregular rhythm
Altered resonance
Reduced intelligibility
Because dysarthria involves the execution of speech movements, acoustic and temporal measures can provide useful longitudinal information.
What Should Be Tracked in Dysarthria?
Potential measures include:
Articulation rate
Overall speaking rate
Pause duration and frequency
Speech intelligibility
Sound-production accuracy
Loudness
Pitch range and variability
Voice quality
Respiratory–phonatory coordination
Performance deterioration during prolonged speaking
ASHA specifically recommends evaluating speech across phonation, articulation, resonance, respiration, and prosody and considering intelligibility, naturalness, communicative efficiency, and participation—not merely one acoustic measurement.
That last point matters.
A person's loudness might improve without their everyday communication becoming substantially easier. Therefore, objective acoustic measurements should be combined with functional outcomes.
Apraxia of Speech Is Different
Acquired apraxia of speech, or AOS, is primarily a disorder of motor planning and programming rather than muscle weakness.
A person may have adequate strength in the speech musculature but have difficulty planning and sequencing the precise movements necessary for speech.
Common characteristics include:
Sound distortions
Distorted substitutions
Difficulty initiating speech
Groping for articulatory positions
Slowed speech
Syllable segmentation
Abnormal stress
Repeated attempts at self-correction
Increasing difficulty as words or utterances become more complex
ASHA notes that sound accuracy often decreases as syllable or sentence complexity increases. Automatic or highly familiar speech may sometimes be easier than novel, generative speech, although this is not an absolute diagnostic rule.
Repetition Consistency Can Be Informative—but Requires Caution
A frequently taught distinction is:
Dysarthria → relatively consistent errors
Apraxia → relatively inconsistent errors
That is useful as a clinical heuristic, but the science is more nuanced.
Research examining repeated multisyllabic words has found that the nature and consistency of AOS errors depend partly on what unit is analyzed and that inconsistency alone should not be used as the definitive diagnostic feature of apraxia.
Therefore, a digital system should not diagnose apraxia simply because repeated recordings contain different errors.
Instead, it could characterize patterns for clinician review.
What Could Longitudinal Tracking Measure in Apraxia?
Useful candidate measures include:
Speech initiation latency
Articulation rate
Inter-syllable timing
Within-word pauses
Sound distortions
Self-correction attempts
False starts
Repeated syllables
Prosodic abnormalities
Error consistency across repeated productions
Change in accuracy as syllable length increases
For example, a patient might repeat:
“cat”
accurately several times but struggle increasingly with:
“categorical”
or longer phrases.
Monitoring performance systematically across increasing phonetic and syllabic complexity could reveal patterns that a simple correct/incorrect score would miss.
Aphasia Requires a Fundamentally Different Framework
Aphasia is primarily an acquired language disorder.
It typically results from damage to the language-dominant hemisphere, most commonly after stroke, and can affect:
Speaking
Understanding spoken language
Reading
Writing
The extent of impairment differs substantially among individuals.
This distinction is critical:
A person can produce speech perfectly from a motor standpoint and still have severe aphasia.
For example, someone may speak rapidly with normal articulation, loudness, and pitch.
Yet the sentences may contain incorrect words, neologisms, semantic errors, or very little meaningful information.
If a monitoring system evaluated only:
articulation rate,
loudness,
pitch,
voice onset, and
pause duration,
that person's speech might look relatively normal.
Their language impairment would be missed.
Pause Time Shows Why Context Matters
Pause duration is a good example of why one metric cannot mean the same thing for every communication disorder.
In nonfluent aphasia, prolonged pauses may reflect difficulty retrieving words.
Research on connected speech shows that nonfluent post-stroke aphasia is often associated with lower speech quantity, slower rate, increased pausing, and reduced grammatical complexity.
Therefore, decreasing word-finding pauses may be a useful recovery signal for some patients.
But consider a person with fluent aphasia.
The person might produce a continuous stream of speech with very few pauses.
A pause metric could therefore look excellent.
Yet the speech might contain:
Semantic paraphasias
Neologisms
Empty language
Missing critical information
Poor coherence
In this situation, less pausing does not necessarily mean better communication.
What Should Be Measured in Fluent Aphasia?
For fluent aphasia, discourse informativeness becomes particularly important.
Potential longitudinal metrics include:
Correct Information Units (CIUs)
%CIUs
CIUs per minute
Main concepts communicated
Semantic paraphasias
Phonological paraphasias
Neologisms
Lexical diversity
Relevant versus empty words
Sentence accuracy
Topic maintenance
Coherence
Auditory comprehension
Functional conversational success
CIU analysis is especially useful because it asks not simply how many words were spoken, but how much accurate, relevant information was successfully communicated.
Research has found measures such as total CIUs, lexical diversity, main concepts, and word-finding behaviors useful for analyzing aphasic discourse.
A 2024 review of aphasic discourse assessment also emphasized that meaningful analysis can occur at several levels—including linguistic accuracy, semantic content, discourse structure, coherence, and pragmatic communication.
One Patient May Have More Than One Disorder
Another major complication is that these disorders can coexist.
A stroke survivor may have:
Aphasia + apraxia of speech
or:
Aphasia + dysarthria
or even multiple communication impairments simultaneously.
ASHA specifically recommends including a motor-speech examination as part of aphasia assessment when appropriate because dysarthria and apraxia can coexist with aphasia.
Therefore, a digital communication platform should not force every speech abnormality into one category.
Instead, it can maintain separate but related measurement domains.
A Better Model for Longitudinal Tracking
For a platform such as AgainAbility, this suggests three distinct layers.
1. Motor Speech Layer
Useful for dysarthria and apraxia:
Speech onset
Articulation rate
Pause timing
Pitch variability
Loudness variability
Prosody
Articulatory accuracy
Repeated-word consistency
Complexity effects
Intelligibility
2. Language Layer
Primarily for aphasia:
Naming accuracy
Response latency
Cue dependence
Semantic errors
Phonological errors
Listening comprehension
Reading comprehension
Writing
Lexical diversity
3. Functional Communication Layer
Useful across disorders:
Correct Information Units
Information per minute
Main concepts
Conversation success
Ability to communicate needs
Real-life role-play performance
Communication partner support required
This final layer is particularly important because impairment scores do not always tell us whether someone's everyday communication has improved.
A 2025 review of functional communication assessment in aphasia concluded that meaningful evaluation should consider communication that is contextual, multimodal, and interactive, reflecting real-world communication rather than only laboratory-style language tasks.
The Same Metric Can Mean Different Things
Consider a simple dashboard:
MetricPatient APatient BSpeaking rate↑↑Pause time↓↓Loudnessstablestable
At first glance, both patients appear to be improving.
But suppose:
Patient A has nonfluent aphasia.
Their speaking rate increases, pauses decrease, naming improves, and CIUs increase.
That pattern may represent meaningful improvement.
Now suppose:
Patient B has fluent aphasia.
Their speaking rate increases and pauses decrease—but semantic errors increase and %CIUs falls.
They are producing more speech but less useful information.
The acoustic metrics improved.
The communication outcome did not.
This illustrates why digital rehabilitation should avoid interpreting individual speech metrics outside the patient's communication profile.
From Universal Metrics to Personalized Metrics
The future of digital speech rehabilitation may therefore not be a single universal “speech score.”
A more clinically meaningful approach is to determine:
What type of communication impairment does this person have, and which metrics are relevant to that impairment?
For someone with dysarthria, loudness and articulation may be central.
For someone with apraxia of speech, articulatory planning, complexity effects, and repeated-production patterns may matter more.
For someone with nonfluent aphasia, naming latency, pauses, grammatical output, and informativeness may be important.
For someone with fluent aphasia, measuring pause duration alone may contribute little unless it is accompanied by semantic accuracy, comprehension, and discourse informativeness.
Final Thoughts
Aphasia, dysarthria, and apraxia of speech can all make spoken communication difficult, but they represent fundamentally different neurological problems.
Dysarthria primarily affects speech execution.
Apraxia of speech primarily affects speech motor planning and programming.
Aphasia primarily affects language processing.
Their symptoms can overlap, and individuals may have more than one disorder.
That is why longitudinal monitoring should not simply collect every available speech metric and assume that improvement in each metric represents recovery.
The more meaningful approach is:
Measure the right variable for the right disorder—and interpret it in the context of functional communication.
For digital rehabilitation, that distinction may be one of the keys to turning large amounts of speech data into information that is genuinely useful to patients and clinicians.
References
American Speech-Language-Hearing Association. Dysarthria in Adults. ASHA Practice Portal.
American Speech-Language-Hearing Association. Acquired Apraxia of Speech. ASHA Practice Portal.
American Speech-Language-Hearing Association. Aphasia. ASHA Practice Portal.
National Institute on Deafness and Other Communication Disorders. Aphasia.
Bislick L, McNeil MR, Spencer KA, Yorkston K, Kendall DL. The Nature of Error Consistency in Individuals With Acquired Apraxia of Speech and Aphasia. American Journal of Speech-Language Pathology. 2017.
Stark BC, et al. A scoping review of transcription-less practices for analysis of aphasic discourse and implications for future research. International Journal of Language & Communication Disorders. 2024.
Hammond, et al. Assessing Functional Communication in Persons With Aphasia: A Scoping Review of Formal and Informal Measures. International Journal of Language & Communication Disorders. 2025.