Characterization of agrammatism in Tagalog: Evidence from narrative spontaneous speech

ABSTRACT

Background

Agrammatism in agglutinative languages exhibits preservation of verb morphology due to their rich morphological paradigms. Tagalog, an agglutinative language mainly spoken in the Philippines, remains uncharacterized in agrammatism yet holds potential for future research that can challenge existing theories and help advance novel ones.

Aims

The purpose of this study is to quantify and describe the characteristics of agrammatism in Tagalog in narrative spontaneous speech, to examine whether these patterns resemble those reported in other agglutinative languages, and to compare the results obtained to those emerging in literature for non-agglutinative languages.

Methods & Procedures

Narrative spontaneous speech was elicited from 10 individuals with non-fluent Broca’s aphasia and 12 matched controls. A series of parameters, both conventional and language-specific measures, were selected for analyses.

Outcomes & Results

Tagalog-speaking individuals with agrammatism exhibited slow and fragmented speech, with a strong preference for minor and simple constructions resulting to decrease in overall grammaticality and sentence complexity. Verb deficits were also found specifically in verbal predication, lexical diversity, and finiteness. There is a prevalent use of bare verb forms, consequently resulting to the decline of verb finiteness and inflections. Code-mixing is found to be extremely variable and statistically non-prominent between groups but patterns may be attributed to personal and sociolinguistic factors.

Conclusion

Patterns of agrammatism in Tagalog majorly mirror the trends reported in other agglutinative and non-agglutinative languages. Where Tagalog agrammatism departs from other agglutinative languages is the increased use of bare verbs as opposed to the obligatory inflected forms. Although still many questions are unresolved, Tagalog offers an interesting testing ground for variety of topics and issues in agrammatic aphasia.

KEYWORDS:

• Agrammatism
• aphasia
• Tagalog
• agglutinative language
• narrative speech
• verbs

Introduction

Typologically, agglutinative languages are distinguished by their rich word-formation process called “agglutination”, where multiple morphemes can be attached to a stem, each conveying distinctive grammatical information (Crystal, 2011). Despite being underrepresented due to language representation imbalance in aphasia research (Beveridge & Bak, 2011), numerous studies have investigated how agrammatism is manifested in speakers of morphologically rich languages. While it appears that most morphosyntactic deficits are shared across languages regardless of typology, agglutinative languages exhibit an intriguing characteristic divergent from others – the resistance of verb morphology to impairment (Druks, 2016). This is in contrast to languages with smaller morphological paradigms wherein verb inflections are more susceptible to omission (Grodzinsky, 1990). This peculiarity highlights the importance of archiving a broader range of languages to potentially uncover additional characteristics that can either bolster existing theories or give rise to entirely new ones, therefore improving our understanding of how agrammatism may manifest in different languages (Bates et al., 1991; Fedorenko et al., 2023; Paradis, 2001).

In this paper, we provide the first characterization of agrammatic narrative spontaneous speech in Tagalog, an agglutinative language spoken mainly in the Philippine archipelago. Natively spoken by communities in the capital city Manila and neighboring provinces (Javier & Or, 2022), Tagalog has been the most popular subject of language research in Philippine linguistics because of its predominance and intricate linguistic features (Pizarro-Guevara & Garcia, 2024). The agglutinative properties inherent in Tagalog are prominently evident in the formation of lexical words, dominantly in the verb conjugation system. What separates Tagalog from other agglutinative language (e.g., Turkish, Finnish, Japanese) is the use of bare verb stems, which are occasionally allowed in natural discourse across lifespan (Garcia & Kidd, 2022; Marzan, 2013).

In the next sections, we first outline a review of data of agrammatic characterization in narrative spontaneous speech from languages with similar morphological richness. Then, the grammatical features of Tagalog and the nature of its speakers will be described, which finally leads to the purpose of the current study.

Agrammatism in narrative spontaneous speech

Agrammatism is an acquired language disorder characterized by a core deficit in morphosyntactic production (Faroqi-Shah, 2023). Although there is heterogeneity in the population (Faroqi-Shah, 2023; Kolk & Van Grunsven, 1985; Miceli et al., 1989; Rochon et al., 2000; Sahraoui & Nespoulous, 2012), the core deficit of agrammatism is manifested through interrelated triad of symptoms: fragmented utterances and sentences, deficits in functional morphology, and scarcity of verbs (Faroqi-Shah, 2023; Menn & Obler, 1990; Saffran et al., 1989). These features have been observed not only in experimental conditions but also in more naturalistic contexts such as spontaneous speech. Assessment of spontaneous speech provides an ecologically valid approach of characterizing aphasic speech (Prins & Bastiaanse, 2004), evaluating strengths and weaknesses from diverse linguistic parameters, and mapping out the potential interplay among these measures (Webster & Howard, 2012). Many neurolinguistic investigations reported how agrammatism presents in the spontaneous speech of individuals with non-fluent Broca’s aphasia.

Agrammatic speech is described as slow, halting, and labored resulting in non-fluent productions typically consisting of 3–4 words per utterance (Menn & Obler, 1990), thus leading to reduced communicative content per unit of time (Chapey, 2012). With the shorter length of utterance, the complexity of sentence construction is significantly reduced. There is a notable preference for short, simpler, and overlearned linguistic structures with a canonical word order (Bastiaanse & Edwards, 2004; Menn & Obler, 1990), in which approximately half of the total utterances are considered grammatically correct (Hsu & Thompson, 2018; Rochon et al., 2000) which includes elliptical utterances (Kolk & Heeschen, 1990) and overused formulaic expressions (e.g., “I don’t know”; DeDe & Salis, 2020). Ungrammatical sentences, on the other hand, are judged to be erroneous due to errors of diverse origins, typically morphological (omission of functional morphemes, agreement errors), syntactic (word order, omission of obligatory argument), and/or lexical (paraphasias, semantic anomalies; Saffran et al., 1989). Additionally, the decline in structural complexity arises from the underuse of embeddings or subordination (Bastiaanse et al., 2002; Saffran et al., 1989).

Another hallmark feature of agrammatism is the omission of grammatical free and bound morphemes (Kean, 1977; Menn & Obler, 1990). Functional morphosyntactic devices, such as articles, connectives, and inflectional affixes are frequently dropped in agrammatic speech, while inflected forms are replaced with less marked ones (in English omitted; De Bleser & Luzzatti, 1994; Grodzinsky, 1990). Older accounts attributed this deficit to a more effortful lexical retrieval specific to function words than to content words (Bradley et al., 1980). Some usage-based accounts associate this particular vulnerability to the low-prominence and high-dependency of functional morphemes in relation to its surrounding linguistic elements (Boye et al., 2023), while others have observed some degree of preservation because of language-specific frequency patterns (e.g., case markers in German; noun articles in Italian; -ing in English; Bates et al., 1987; Faroqi-Shah & Thompson, 2004).

A well-studied phenomenon in agrammatism is the verb deficit. Cross-linguistically, the evidence regarding the scarcity of verbs varies. While many studies reported a reduction of verb use (Bastiaanse & Jonkers, 1998; Bird et al., 2000; Goodglass, 1993; Kim & Thompson, 2000; Menn & Obler, 1990), some others reported comparable verb proportions to those of healthy adults (Abuom & Bastiaanse, 2012; Anjarningsih et al., 2012; Lorch, 1990; Niemi et al., 1990). Nevertheless, verb proportions almost always exhibit some form of interaction with other parameters, commonly referred to as trade-off. An inverse relationship is normally observed, wherein a comparable proportion of verbs is often associated with a chance of compromised verb diversity, inflection, and/or finiteness (Bastiaanse & Jonkers, 1998). In the attempt to understand the susceptibility of verbs to impairment, potential variables have been examined, such as time reference (Bastiaanse et al., 2011; Bird et al., 2003), regularity (Faroqi-Shah, 2007), morphological complexity (Faroqi-Shah & Thompson, 2010), among others. Problems relating to argument structure appear to be the predominant account, with the internal argument of verbs influencing the use of processing resources during spontaneous speech. This may manifest as omission or substitution of inflectional markers and internal arguments, and/or substitution of verbs with less complex argument structure (Faroqi-Shah, 2023; Jonkers, 1999; Thompson, 2003).

Evidence of agrammatism in agglutinative languages

Similar to the features of agrammatism in non-agglutinative languages (e.g., English, German, Dutch, Italian, Spanish), speakers of agglutinative languages exhibit almost congruent patterns symptom-wise (see Appendix A for an overview). In summary, overall production is slow and laborious, composed of utterances that are simple, canonical, and lack complexity, in which a large proportion is considered ungrammatical. Some languages also report omission of grammatical morphemes, specifically free morphemes such as case markers in Korean, Basque, and Japanese (Halliwell, 2000; Laka & Korostola, 2001; Sasanuma, 1989), syntactic particles in Swahili, Indonesian, and Hungarian (Abuom & Bastiaanse, 2012; Anjarningsih et al., 2012; MacWhinney & Osmán-Sági, 1991), and closed-class words in Finnish (Helasvuo et al., 2001; Niemi et al., 1990).

While many similarities extend to both fragmented utterances and errors in functional morphology among languages, where agglutinative languages depart from the classical description of agrammatism is in the preservation of verbal morphology. This phenomenon is specifically highlighted in bilinguals who speak two typologically-distinct languages wherein verb inflection is better preserved in morphologically-complex language (Swahili) than in languages with simpler morphological paradigms (English; Abuom & Bastiaanse, 2012). This dissociation is explained due to the agglutinative languages’ heavy reliance on bound morphology in expressing grammatical functions (Druks, 2016), in contrast to less-inflected counterparts which predominantly rely on word order (Mahowald et al., 2023). Since many of these languages do not allow the use of bare verb stems, the functional morphology accompanying the verbs is given top priority during production as it holds high information value cues to grammatical relations (Slobin, 1991) and cannot be easily omitted (Boye & Bastiaanse, 2018). Moreover, drawing from child language development, these morphological rules are usually one of the initial linguistic skills mastered by typically developing children from morphologically-rich languages hence supporting the notion that what is acquired first will be lost last (Jakobson, 1941; Nedergaard et al., 2020). However, this notion remains contentious due to the existing counter-evidence regarding the parallels between agrammatic speech and child language patterns (Maviş, 2008).

This preservation, however, does not come without a cost. Many studies reported trade-offs between functional morphology and general verb usage. Similar to other languages, verb lexical richness, finiteness index, and inflectional complexity are compromised in agglutinative languages. While verb inflection might be preserved, agrammatic speech reveals a less diverse variety of verbs (Arslan et al., 2016; Sang, 2015), frequent substitution of inflectional morphemes resulting in errors in tense and agreement (Abuom & Bastiaanse, 2012; Halliwell, 2000; Laka & Korostola, 2001; Lartey et al., 2017; MacWhinney & Osmán-Sági, 1991), and the use of more complex morphological paradigms are mostly constrained (Halliwell, 2000; Nedergaard et al., 2020; Slobin, 1991). Highlighting on the last one, verb inflections are frequently substituted with less-marked acceptable word forms, restricting the use of verbs carrying several grammatical functions (Arslan et al., 2016; Halliwell, 2000; Helasvuo et al., 2001; Laka & Korostola, 2001; Niemi et al., 1990; Sasanuma, 1989).

While the omission of inflectional markers is generally observed in English agrammatism, this is not a viable option for most agglutinative languages, because stripping away bound morphemes results in non-existent and illegal word forms. Instead, morpheme substitution is the observed behavior in their speech (De Bleser & Luzzatti, 1994; Grodzinsky, 1990). Consequently, these substitutions lead to a diminished complexity of verbal markedness, resulting in the use of minimally-marked verb forms which are morphologically acceptable but often syntactically errorful (Anjarningsih et al., 2012; Halliwell, 2000; Helasvuo et al., 2001; Laka & Korostola, 2001; Niemi et al., 1990; Sang, 2015). In the only experimental study in Tagalog agrammatism (Bondoc et al., 2018) wherein they tested 5 native Tagalog-speaking individuals with agrammatism in their ability to produce and comprehend voice markings at sentence level, it was reported that all participants exhibited a high omission rate of case markers and evident use of bare verb roots during the production task. Although highly agglutinative, Tagalog allows the use of bare verbs in exceptional contexts (Garcia & Kidd, 2022). Typically developing children are also observed to be using these forms as they master the language’s conjugation paradigm (Marzan, 2013). Overall, these strategies may be viewed as a rational adaptation in agrammatism (Fedorenko et al., 2023) in an effort to convey the message in the simplest manner – by exploiting language-specific morphosyntactic devices that are readily available to prevent dealing with complexity (Menn & Obler, 1990; Paradis, 2001).

Relevant features of Tagalog grammar

Tagalog is an agglutinative Austronesian language from the Western Malayo-Polynesian branch (Schachter & Reid, 2011). It is mainly spoken as the native language of Metro Manila and its surrounding regions, but due to its national importance for basic education, local media, and commerce, approximately 70–90% of the population use Tagalog as either their first or second language (Adelaar & Himmelmann, 2004). English is designated as the second official language, which means that a huge percentage of the population is at least bilingual. Because of this, code-switching and mixing became dominant in everyday communication in populous cities. Many Tagalog speakers, especially those residing in Metro Manilla, are known to be avid code-switchers (Taglish) wherein they alternate their languages in a single course of utterance and modify words from one language to fit into the morphosyntactic rule of the other (see 1, English words “cramps” and “eat” embedded in Tagalog verbal morphology; Green & Abutalebi, 2013).

(1) Nagcracramps ang tiyan ko kapag nag-i-eat ako ng dairy

VCE.AGENT-MODE.INTENS-ASP.IMPERF.cramps SUBJ stomach my-1S whenever VCE.AGENT-MODE.DELIB-ASP.IMPERF.eat I-1S N.SUBJ dairy

“My stomach cramps whenever I eat dairy”

A basic clause in Tagalog is composed of two components: a subject and a predicate which can either be verbal (2) or non-verbal (3). The basic clause is the simplest, with the fewest components, focusing only on one proposition, and is used as the building block of more complicated constructions (compound and complex sentences). Generally, the most prevalent structure of a clause is predicate initial (Malicsi, 2013), which makes Tagalog a verb-first language while the rest of the arguments are free as word order does not distinguish the role assignment (Schachter & Reid, 2011). Because of its “pro-drop” nature (Kroeger, 1993), another unusual feature of Tagalog is the use of argument-dropping or zero anaphors (Park, 2020) which is similar to elliptical utterances. This is the usage of a gap in discourse which typically refers back to an antecedent that precedes the utterance somewhere in the discourse. This is commonly used in coordinated sentences (4), wherein the dropped subject or the predicate is referenced from the initial clause (Haspelmath, 2007).

(2) K-um-a-kanta ang babae

VCE.AGENT-ASP.IMPERF-MODE.CASUAL.sing SUBJ woman

“The woman is singing”

(3) Ma-lungkot ang bata

ADJ.sad SUBJ child

“The child is sad”

(4) Nag-bukas si Tess ng regalo tapos si Maria din

VCE.AGENT-ASP.PERF-MODE-DELIB.open DET Tess N.SUBJ gift and then DET Maria too

“Tess opened the gift and then Maria too”

Word formation in Tagalog is built from three basic morphological classes (Nolasco, 2005, 2011). First, there are roots, which are bare forms capable of taking in affixes. Secondly, affixes are bound morphemes that are embedded within root words and carry grammatical information that can alter both the semantics and syntax of the formed word. Lastly, particles, contrary to roots, are free morphemes that can occur independently even without an affix attached to it. Although particles constitute complete words, they are unable to stand alone and must be used to accompany lexical words. When used in combination with other words, particles provide additional semantic information to the context (5).

(5) Ma-galing na daw

ADJ.heal PART-already PART-reportedly

“They say he/she is already recovered/healed”

The agglutinative properties of Tagalog are mainly observed in its intricate verbal morphology. What makes the language’s morphological system complex is the nesting (or agglutination) of grammatical features namely voice, aspect, and mode (6; Cena, 2014). A fully realized verb can originate from any bare root form. Once this root is assigned with a voice marker, an infinitive verb is formed. Depending on the intended manner of the verb, one or more mode markers can be added, which comes in numerous hybrid forms with voice (Schachter & Otanes, 1972). The infinitive marking later assigns which inflectional paradigm for aspect the verb will follow. By rule, a Tagalog verb can only be considered finite once it is inflected with aspect (Wurmbrand, 2013). Although the verb formation appears to be very complex, the use of bare verb roots is sometimes allowed. This leeway is evidently used by developing children, as well as by adults during child-directed speech, for imperative and exhortative purposes (7; Garcia & Kidd, 2022; Marzan, 2013).

(6) I-pa-p(a)-ag-bili

VCE.APPL-ASP.CONTEMP-MODE.CAUSE-MODE-DELIB.buy

“Will be caused to be bought”

(7) Tara kain tayo Tara k-um-ain tayo

come BARE.eat us vs. come VCE.AGENT-ASP.INF.eat us

“Come, let us eat” “Come, let us eat”

The current study

The aim of the present study is two-fold. First, drawing from the reviewed general manifestations of agrammatism, we seek to quantify and describe the characteristics of spoken narrative production of Tagalog-speaking individuals with non-fluent Broca’s aphasia, comparing them with neurotypical controls. Second, our investigation extends to examining whether the agrammatic patterns observed in Tagalog resemble those found in other agglutinative languages.

Methodology

Overall, our methodology are in accordance to the identified best practices in the assessment of narrative spontaneous in agrammatic characterization. This later allows replicability in future research and support clinical implementation (Linnik et al., 2016; Stark et al., 2023).

Participants

Ten individuals with non-fluent aphasia (non-fluent IWA) from Greater Metro Manila participated in this study. They are all primary speakers of Tagalog, and English is their second language as indicated by the Language Experience and Proficiency Questionnaire (LEAP-Q, Kaushanskaya et al., 2020). All were diagnosed with non-fluent Broca’s aphasia secondary to a first symptomatic stroke by registered speech-language pathologists (SLP). Since no standardized aphasia test is available for Tagalog, ensuring correct classification of non-fluent IWA into the non-fluent category involved scoring all samples using the spontaneous speech subtest scoring matrix (fluency, grammatical competence, and paraphasias) of the Western Aphasia Battery – Revised (WAB-R; Kertesz, 2006). This was conducted by the first author (a trained SLP) and reached an agreement with two other SLPs. Additionally, although most of the non-fluent IWA had co-existing motor speech disorders such as apraxia of speech (AOS) and dysarthria, none were severe enough to warrant exclusion (MacWhinney & Osmán-Sági, 1991). Time post-onset of stroke was at least six months, and none had co-existing comorbidities associated with cognitive problems (e.g., progressive neurodegenerative disorders; MacWhinney et al., 2011). The control group comprised 12 non-brain damaged speakers (NBD) who were age-, gender- and education-matched. Summarized in Table 1 is the demographic information of both groups as per DESCRIBE guidelines (Wallace et al., 2023).

Table 1. Demographic Information of non-fluent IWA (NF) and NBD Control (HC) Participants .

Participant	Age	Gender	Occupation	YOE	Handedness	TPO	Etiology	SLP Dx	MSD
NF001	42	Male	Massage therapist	13	Right	21	Left CVA – Ischemic	Broca’s	AOS & Dysarthria
NF002	34	Male	Public Accountant	14	Right	10	Left CVA – Unspecified	Broca’s	AOS
NF003	58	Male	Telecom Engineer	15	Right	36	Left CVA – Hemorrhagic	Broca’s	AOS & Dysarthria
NF004	52	Male	Company Driver	12	Right	6	Left CVA – Hemorrhagic	Broca’s	AOS & Dysarthria
NF005	40	Female	Call Center Agent	14	Right	16	Left CVA – Ischemic	Broca’s	None
NF006	49	Female	Sales Manager	14	Right	49	Left CVA – Ischemic	Broca’s	AOS
NF007	43	Male	Marine Surveyor	15	Right	44	Left CVA – Ischemic	Broca’s	AOS
NF008	58	Male	Production Manager	14	Right	16	Left CVA – Unspecified	Broca’s	AOS
NF009	52	Female	Medical Radiologist	22	Right	8	Left CVA – Ischemic	Broca’s	AOS
NF010	45	Female	Supermarket Manager	12	Right	8	Left CVA – Hemorrhagic	Broca’s	AOS
HC001	50	Male	Athletic Director	18	Right
HC002	56	Male	Real Estate Director	16	Right
HC003	51	Male	Data Collection Analyst	17	Right
HC004	42	Male	Flight Attendant	14	Right
HC005	55	Male	Sales Manager	16	Right
HC006	43	Female	Dentist	18	Right
HC007	52	Female	House Helper	11	Right
HC008	63	Male	Retired Engineer	15	Right
HC009	57	Female	Physician	20	Right
HC010	61	Female	Government Employee	16	Right
HC011	51	Female	Businesswoman	14	Right
HC012	40	Female	Housewife	12	Right

Note: YOE = Years of Education, in years; TPO = Time Post-Onset, in months; SLP Dx = SLP Diagnosis; MSD = Motor Speech Disorder; AOS = Apraxia of Speech

Materials & procedures

For eliciting maximum productive speech from the participants, we developed a Tagalog narrative speech elicitation protocol that incorporates multiple elicitation paradigms. The protocol is divided into three subsections. First, a personal interview based on AphasiaBank guidelines (MacWhinney et al., 2011) wherein participants were asked to talk about their stroke (non-fluent IWA) or illness/injury (NBD), their recovery story, and a memorable event (Kleinman et al., 1978; Olness & Ulatowska, 2011). Secondly, narrative elicitation was elicited through three picture description tasks: (a-b) two black-and-white picture panels, namely “Broken Window story” and “Refused Umbrella story” (MacWhinney et al., 2011; Menn & Obler, 1990, p. 36; Richardson & Dalton, 2016) and (c) a colored-picture panel called “Little Birds” from the Tagalog version of the Multilingual Assessment Instrument for Narratives (MAIN-Tagalog; Amora et al., 2020). Thirdly, a story-retelling task from MAIN-Tagalog called “Cat” was presented.

The protocol was designed with Tagalog instructions for the SLP who assisted with data collection to follow, aiming not only to establish a naturalistic elicitation environment but also to encourage participants to predominantly use Tagalog in their spontaneous speech. The protocol also included the sequence of the tasks, step-by-step instructions, scripts for permitted neutral cues, and a few troubleshooting reminders.

All elicitation sessions were conducted in a quiet setting and were audio-recorded. For each task, participants were prompted to produce as much speech as possible by providing encouragement, neutral cues, and probing questions as necessary. Two SLPs with linguistic training backgrounds (including the first author) transcribed all speech samples orthographically. Transcriptions were segmented primarily based on the minimum syntactic unit (clauses), with lesser weight given to prosodic features and pauses (Boxum et al., 2013), to determine utterance boundaries that convey an understandable meaning (Saffran et al., 1989; Wagenaar et al., 1975).

Analysis

While a minimum word count of 200 words is recommended for analysis in agglutinative languages (Abuom & Bastiaanse, 2012), only two non-fluent IWA met this requirement. Consequently, the authors opted to include all non-fluent IWA, who successfully completed the entire elicitation protocol, irrespective of their word counts. Following the standard in narrative speech studies, a sum of 300 words were extracted from different elicitation paradigms (based on the proportions of non-fluent IWA sample) for the analysis of the NBD samples (Prins & Bastiaanse, 2004; Vermeulen et al., 1989).

In order to characterize the nature of spontaneous speech of the sample, the following parameters were measured: rate of speech, mean length of utterance in words (MLUw) and in morphemes (MLUm), sentence types, number of embeddings, grammaticality, finiteness index, total number of verbs, proportion of modal and lexical verbs, and verb lexical diversity. Additionally, other parameters were introduced as they were deemed unique for Tagalog: average number of verbal and non-verbal predicate types per utterance (Malicsi, 2013), morphological density measuring the proportion of morphological classes (Nolasco, 2011), and percentage of code-mixed words (Goral et al., 2019). For full description of each parameter, please refer to Appendix B.

For statistical analysis, due to the small sample size, non-parametric Mann-Whitney U tests were utilized to assess the significance of distinctions (alpha level = 0.05, two-tailed in all comparisons) between the non-fluent IWA and the NBD group. These tests were conducted for each variable separately. For effect sizes, we report the rank biserial correlation (r) ranging from −1 to +1. Please refer to Appendix C for all the raw scores for each parameter. For inter-rater reliability measures (IRR), together with the first author, two other native Tagalog SLPs with at least 5 years of experience in aphasia management took part in this part of the study. They underwent a training session that included familiarization with the analysis scheme, practice exercises, and discussions to clarify and resolve any ambiguities. Each rater provided their ratings independently. Krippendorff’s Alpha (Krippendorff, 2019) was employed to assess the interrater reliability of the analysis protocol and the data set was processed using a web-based K-Alpha calculator (Marzi et al., 2024). Statistics yielded a Krippendorff’s Alpha coefficient of 0.983 with 95% confidence interval [0.960 and 0.994] which indicates satisfactory level of agreement and a reliable rating.

Results

Speech production parameters

The results revealed lower scores in all speech fluency measures for the non-fluent IWA compared to NBD speakers (see Table 2). They had a significantly slower rate of speech (WPM), U = 1.00, p < .001, r = 0.98, and shorter utterance length, both in words (MLUw), U = 0.00, p < .001, r = 1.0, and in morphemes (MLUm), U = 0.00, p < .001, r = 1.0. Contrary to these findings, non-fluent IWA produced more utterances during the speech sample elicitation compared to the NBD, U = 5.00, p < .001, r = 0.92.

Table 2. Speech Fluency Parameters.

	Total Number of Words	Speech Rate (WPM)	MLU in words (MLUw)	MLU in morphemes (MLUm)	Number of utterances
NF001	74	15.89	1.61	1.72	46
NF002	146	18.04	2.81	3.08	52
NF003	170	45.64	2.70	3.05	63
NF004	121	31.40	3.46	4.57	35
NF005	159	25.60	2.21	2.49	72
NF006	255	62.68	4.40	4.71	58
NF007	131	31.80	2.26	2.72	58
NF008	137	37.14	3.19	3.81	43
NF009	159	28.14	3.12	4.20	51
NF010	233	21.18	3.76	4.73	62
Non-fluent IWA Mean	158.5	31.75	2.95	3.51	54
Std. Dev.	52.6	14.07	0.82	1.05	10.85
NBD Mean	300	90.38	9.16	12.68	34.25
NBD Range		56.66–136.96	7.50–11.77	9.73–15.73	26–42
Std. Dev.	0.00	20.46	1.46	2.11	5.43

With more utterances, non-fluent IWA produced more sentences in total than the NBD group, U = 0.00, p < .001, r = 1.0. With a closer look, they produced simpler and shorter sentences as seen in the significant overuse of minor, U = 0.00, p < .001, r = 1.0, and simple sentences, U = 4.50, p < .001, r = 0.93. For more complex constructions, no difference in the production of compound sentences was found, U = 40.50, p = 0.203, r = 0.33, but non-fluent IWA produced significantly lower number of complex sentences than NBD, U = 1.50, p < .001, r = 0.98, (see Table 3).

Table 3. Sentence grammaticality and complexity parameters.

	Total Number of Sentences	Sentence Types				Percentage of Grammaticality	Number of Embeddings	Average # of Predicate Types per Utterance
		Minor	Simple	Compound	Complex			Non-Verbal Predicates	Verbal Predicates
NF001	46	38	7	0	1	30.43%	1	0.22	0.54
NF002	52	32	13	0	7	13.46%	7	0.10	0.50
NF003	62	34	26	2	0	24.19%	0	0.14	0.60
NF004	34	10	21	1	2	35.29%	2	0.00	1.00
NF005	72	50	15	0	7	34.72%	7	0.39	0.51
NF006	50	15	16	8	11	26.00%	11	0.35	0.57
NF007	58	38	15	0	5	29.31%	6	0.26	0.66
NF008	43	23	14	0	6	30.23%	7	0.33	0.74
NF009	42	18	15	5	4	38.10%	4	0.12	0.90
NF010	45	14	15	10	6	35.56%	6	0.19	0.86
Non-fluent IWA Mean	50.4	27.2	15.7	2.6	4.9	29.73%	5.1	0.21	0.69
Std. Dev.	11.08	13.10	4.97	3.75	3.28	7.23	3.35	0.12	0.18
NBD Mean	22.25	0.08	4.50	3.17	14.50	93.13%	24.25	0.27	1.43
NBD Range	11–33	1	0–14	1–8	9–19	86.21–100	15–29	0.11–0.39	1.20–1.77
Std. Dev.	6.60	0.29	4.19	2.21	3.12	4.22	3.84	0.11	0.16

Meanwhile, the grammaticality of sentences and the number of embeddings were significantly lower for non-fluent IWA. None of the non-fluent IWA reached 40% grammaticality in sentences, U = 0.00, p < .001, r = 1.0, a result of a large number of predicate omissions (see (8) below), argument droppings (9), frequent omission of grammatical affixes and particles (10), leading to ungrammatical constructions. Moreover, due to a strong preference for simpler constructions, the number of embeddings was considerably lower in non-fluent IWA as well, U = 0.00, p < .001, r = 1.0.

(8) Missing predicate/verb

Bata ball

child ball

“child ball”

(9) Missing subject/agent

lalaro ng bola

VCE.AGENT-ASP.CONT.play N.SUBJ ball

“will play ball”

(10) Missing determiners/case markers

lulundag paruparo, nahulog cat

VCE.AGENT-ASP.CONT.jump butterfly VCE.AGENT-ASP.PERF.fall cat

“butterfly will jump, cat fell”

Non-fluent IWA produced significantly fewer predicates than NBD, U = 27.00, p = 0.032, r = 0.55. Further examination of average number of each predicate types per utterance revealed that this difference is driven by the significantly lower use of verbal predicates per utterance, U = 0.00, p < .001, r = 0.79, whereas the average number of non-verbal predicates per utterance was similar in both groups, U = 40.00, p = 0.198, r = 0.33.

Functional morphology parameters

As mentioned above, the large proportion of ungrammatical sentences in non-fluent IWA was often caused by the frequent omission of grammatical affixes and particles. As shown in Table 4, the use of root words was noticeably higher in non-fluent IWA compared to NBD, U = 5.00, p < .001, r = 0.92, while the pattern was reversed for affixes, U = 28.00, p = 0.036, r = 0.53, and particles, U = 12.00, p < .001, r = 0.80, which non-fluent IWA used significantly fewer than NBD participants.

Table 4. Morphological class parameters.

		Proportion of Morphological Classes
	Total Number of Morphemes	Roots	Affixes	Particles
NF001	79	79.75%	5.06%	15.19%
NF002	156	83.97%	7.69%	8.33%
NF003	189	66.67%	11.11%	22.22%
NF004	161	52.17%	23.60%	24.22%
NF005	185	66.49%	11.89%	21.62%
NF006	273	58.24%	4.76%	37%
NF007	157	72.61%	16.56%	10.83%
NF008	167	58.68%	16.77%	24.55%
NF009	215	47.91%	24.19%	27.91%
NF010	252	66.67%	23.81%	9.52%
Non-fluent IWA Mean	183.4	65.3%	14.5%	20.1%
Std. Dev.	54.65	11.47	7.61	9.13
NBD Mean	429.67	46.1%	21.6%	32.3%
NBD Range	367–544	36.64–53.13	18.8–26.57	27.62–36.67
Std. Dev.	46.47	4.36	12.66	4.78

Verb-related parameters

Based on the proportion of the analyzed sample, non-fluent IWA significantly used a higher proportion of verbs than the control speakers, U = 16.00, p = 0.004, r = 0.73. Upon closer inspection of verb types, even with lower word counts, the overuse of both modal, U = 17.00, p = 0.005, r = 0.72, and lexical verbs, U = 23.00, p = 0.015, r = 0.62, was apparent in non-fluent IWA’s speech (see Table 5).

Table 5. Verb-Related Parameters.

	Total Verb Percentage	Proportion of Verb Types		Percentage of Bare Verb Use	Finiteness Index	Verb Diversity	Noun – Verb Ratio
		Modal Verbs	Lexical Verbs
NF001	33.78%	14.86%	18.92%	71.43%	0.21	0.71	0.72
NF002	17.81%	0.68%	17.12%	56.00%	0.14	0.76	3.73
NF003	23.53%	5.29%	18.24%	54.84%	0.45	0.71	1.65
NF004	30.58%	2.48%	28.10%	20.59%	0.53	0.74	1.05
NF005	22.01%	5.66%	16.35%	69.23%	0.21	0.69	1.17
NF006	13.33%	5.10%	8.24%	66.67%	0.25	0.52	1.50
NF007	29.77%	5.34%	24.43%	50.00%	0.36	0.84	1.33
NF008	24.82%	6.57%	18.25%	24.00%	0.63	0.68	0.74
NF009	29.56%	5.66%	23.90%	21.05%	0.56	0.76	0.62
NF010	23.61%	3.86%	19.74%	26.09%	0.69	0.67	0.80
Non-fluent IWA Mean	24.88%	5.57%	19.31%	46.00%	0.40	0.71	1.33
Std. Dev.	6.28	3.72	5.41	21	0.20	0.08	0.91
NBD Mean	17.92%	2.39%	15.55%	2.30%	0.97	0.81	1.14
NBD Range	16.33–21.33	1.67–3.33	14–19	0–17.5	0.09–1.0	0.75–0.86	0.96–1.42
Std. Dev.	1.86	0.44	1.806	2.71 (4.7)	0.04	0.04	0.14

The finiteness index was significantly lower in non-fluent IWA, U = 0.00, p < .001, r = 1.0. This significantly low usage of finite verbs was attributed to the high percentage of bare verbs, U = 0.00, p < .001, r = 1.0. Additionally, non-fluent IWA had lower verb diversity, U = 15.00, p = 0.001, r = 0.75. Although with great variability amongst individual participants, there was no significant between-group difference in noun-verb ratio, U = 55.00, p = 0.767, r = 0.08.

Supplemental parameters

No significant difference in the number of instances of code-mixing (Taglish) was found between groups, U = 38.00, p = 0.159, r = 0.37 (see Table 6). Individual data were greatly varied and could be associated with specific participants’ professional background and language experiences. Qualitatively, most of the code-mixed English words are common nouns (e.g., ball, butterfly, bird, cat) and some verbs, in either inflected (e.g., na-save, re-rain, nagra-run) or uninflected bare forms (e.g., catch, exercise, kick).

Table 6. Code-mixing parameters.

	Number of Code-Mixed Words	Total Number of Words	% Code-Mixing
NF001	10	74	13.51%
NF002	93	146	63.70%
NF003	66	170	38.82%
NF004	5	121	4.13%
NF005	25	159	15.72%
NF006	21	255	8.24%
NF007	35	131	26.72%
NF008	18	137	13.14%
NF009	11	159	6.92%
NF010	10	233	4.29%
Non-fluent IWA Mean	29.4	158.5	19.52%
Std. Dev.	28.5	52.6	18.9
NBD Mean	26.08	300	8.69%
NBD Range	4–55	300	1.33–18.33
Std. Dev.	15.76	0.00	5.25

Discussion

In an effort to improve to our understanding of the impact of agrammatism across diverse language typologies, this study generated findings from the first-ever characterization of spontaneous speech in Tagalog-speaking individuals with non-fluent aphasia. We selected a range of linguistic parameters which were sub-grouped based on agrammatism’s triad of symptoms (Faroqi-Shah, 2023), encompassing both conventional and language-specific measures distinctive to Tagalog. Below, the findings will be critically evaluated for congruence with existing literature on aphasia, more specifically in agglutinative languages. Our results highlight the relevance of characterization of agrammatism in less explored languages, emphasizing their crucial role in understanding the manifestation of agrammatism from a cross-linguistic standpoint.

Fragmented utterances and sentences

As expected, non-fluent IWA’s many dysfluencies in their speech production result in slower speech rate. However, these low fluency rates cannot be solely accounted by either motor speech planning or language deficit as this may be due to interplay these factors (Faroqi-Shah, 2023). Unfortunately, our current data cannot account for these potential interactions, hence this issue may be carried over for follow-up research in the future as non-fluency is considered to be a strong predictor of morphosyntactic ability in aphasic speakers (MacWhinney et al., 2011) and is regarded as “stalling for time” during linguistic planning (DeDe & Salis, 2020). The pattern of extended discourse with oversimplified and fragmented syntax, which can be seen as a compensatory strategy (Sahraoui & Nespoulous, 2012), was observed in our findings as non-fluent IWA produced a higher number of shorter utterances compared to NBD. Although this may be the case, note that because of the 300-word cap, the analyzable portions from the NBD may not entirely express the same content as the non-fluent IWA’s samples.

At sentence level, non-fluent IWA displayed significantly lower scores in multiple indices such as grammaticality, predication, and subordination. As a rational behavior to flexibly adapt their language, non-fluent IWA produced significantly higher numbers of sentences which are mostly minor and simple constructions, preserving the most informative parts (Fedorenko et al., 2023). This increased number of utterances and sentences observed in non-fluent IWA is primarily attributed to the lack of complexity in sentence construction, contrary to the performance of the NBD group. One might intuitively assume that if a person produces more sentences, the number of predicates should be at least equivalent. Both groups exhibited a mirroring pattern wherein they both utilized a higher frequency of verbal predicates compared to non-verbal predicates. However, upon closer comparison against NBD speakers’ performance, non-fluent IWA’s use of verbal predicates were extremely reduced while non-verbal predication was preserved. Among other factors such as argument omission and substitution/omission of functional morphemes, the reduction in verbal predicates actually contributed to the low proportion of grammatical utterances. Moreover, due to the preference for syntactically simpler constructions, the number of sentence embeddings, which is a key element to the formation of complex sentences, was considerably lower when compared to the controls.

Deficiencies in functional morphology

A trend aligns with Menn and Obler’s (1990) summary of the limited use of morphosyntactic devices in any particular language, either through omission or substitution, while content words remain. In our case, root words functioning as content words were profusely used. Meanwhile, grammatical morphemes such as affixes and particles were often omitted or extremely underused. As a consequence, these patterns influenced both speech productivity and grammaticality negatively. Notably, performance among non-fluent IWA varied, specifically with the use of affixes in some non-fluent IWA’s (NF004, NF009, & NF010) being at par with NBD speakers, highlighting the heterogeneity of the agrammatic population (Faroqi-Shah, 2023; Kolk & Van Grunsven, 1985; Miceli et al., 1989; Rochon et al., 2000). Since our data only provided a surface analysis with regards to the production of morphological devices, a closer look of specific preferences in the use of affixes and particles (bound vs free morphemes) may be warranted in future work.

Verb trade-offs

Although the ratios between nouns and verbs are variable, regardless of types, verbs were apparently overused even with shorter speech which is contrasting to previous literature reporting consistent verb omissions (Bird et al., 2000; Goodglass, 1993; Kim & Thompson, 2000). Although unexpected, this overuse is not surprising as the functional elements were dropped, preserving elements with high prominence value such as verbs (Boye et al., 2023). However, trade-off patterns were yet again observed in verb indices such as low scores for finiteness and lexical diversity which were construed as a sign of compensation (Anjarningsih et al., 2012; Bastiaanse & Jonkers, 1998).

One major finding highlighted in this paper is the increased use of bare verbs, similar to patterns found in languages that allow bare verb use such as in English (Grodzinsky, 1984, 1990). This is where Tagalog stands out among other agglutinative languages studied to date in the agrammatic literature. Unlike others languages (Abuom & Bastiaanse, 2012; Arslan et al., 2016; MacWhinney & Osmán-Sági, 1991; Nedergaard et al., 2020; Niemi et al., 1990; Sasanuma, 1989) where a bare form cannot simply exist on its own, irrespective of its complex verb conjugation system, Tagalog uniquely allows its speakers to use of bare verbs in exceptional contexts such as in imperative and exhortative functions (Garcia & Kidd, 2022; Marzan, 2013). This normally happens when the affixes attached to the verb play a less critical role and are dispensable, such as in infinitive forms. However, the non-fluent IWA’s exploitation of bare verb forms was not only limited to infinitive forms, but also to finite verbs in obligatory contexts (11a-b). This is again in parallel with child connected speech (Marzan et al., 2014) and experimental data on Tagalog agrammatism (Bondoc et al., 2018), wherein omission of required inflections were more common than affix substitution. While we acknowledge that there is a substantial variability in proportions, the overuse of bare verb forms is consistently observed across all non-fluent IWA, thereby indicating its status as a pervasive symptom of agrammatism in Tagalog.

(11a) Non-fluent IWA’s utterance

Aso kagat ang buntot

dog BARE.bite SUBJ tail

“dog bite the tail”

(11b) Reconstructed utterance

Yung aso, kinagat ang buntot

DET dog VCE.OBJECT-ASP.PERF.bite SUBJ tail

“the tail was bitten by the dog”

Ultimately, the fragmented shape and grammatically-impoverished character of agrammatic speech hinges on verb-related deficits (Berndt et al., 1997; Edwards, 1998), which aligns closely with our data. These problems may be caused by the grammatical information hosted by verbs (Boye et al., 2023; Druks, 2016; Slobin, 1991) and the complexity of predicate-argument structure (Bastiaanse & Jonkers, 1998; Malyutina et al., 2016; Miceli et al., 1984; Thompson, 2003) that potentially creates a processing bottleneck during speech production (Faroqi-Shah, 2023). As to how exactly Tagalog verbs in non-fluent IWA behave is beyond the scope of this paper. Nevertheless, this particular issue warrants more focused attention in future research, specifically investigating Tagalog verbs’ intricacies, both in narrative speech and controlled conditions – from its unique grammatical features, argument structure, and its morphological richness.

Code-mixing in Tagalog

As Tagalog speakers are known to be avid code-switchers, a noteworthy finding in our narrative speech analysis was the comparable degree of code-mixing between groups. Although extremely varied, both non-fluent IWA and NBD speakers’ percentage of code-mixing may have been somehow influenced by sociolinguistic factors such as occupational background and language experience of each individual (Goral et al., 2019). Participants with the highest degree of code-mixing from either group had a history of working overseas or from a workforce that uses the mix English-Tagalog as the medium of communication, which may explain as to why these particular participants code-mixed more than others. Again, this pattern points towards exploitation of not only by language-specific structures, but also language habits to aid communicative success. These results, however, should be interpreted with caution, as no potential difference can be drawn due to the absence of controls for variables number of languages spoken, language proficiency, and language experiences among our recruited participants. Although our data were inconclusive, as Tagalog provides an interesting platform to examine the intricacies of bi/multilingualism and code-mixing in aphasia, this specific topic of interest could be given more attention in a more controlled study in terms of participant recruitment and testing conditions.

An account of agrammatism in Tagalog

Irrespective of the heterogeneity in the population, at the group level, agrammatism in Tagalog resembled not only the patterns found in other agglutinative languages but also universally. However, where Tagalog departed from this pattern relates to Tagalog-specific characteristics. In summary, Tagalog-speaking individuals with agrammatism had a slow rate of speech and reduced utterance length, leading to a preference for simpler yet often ungrammatical constructions. Sentence complexity was constrained by a limited use of embeddings. Despite a higher proportion of verb usage, deficits in verb-related indices such as verbal predication, lexical diversity, and finiteness were evident. Finiteness in particular was hampered by the high proportion of bare verb use. When forms like bare forms naturally exist in a specific language, individuals with agrammatism are predicted to overuse these alternatives to navigate through complexities (Paradis, 2001).

This overuse of bare verb forms is not a new finding in Tagalog agrammatism, as it has also been reported in an earlier experimental study (Bondoc et al., 2018). Our non-fluent IWA group, who were constrained by limited linguistic capacity, naturally gravitated towards relying on this readily exploitable language feature during discourse. Evidence from languages with complex verb paradigms implies that verb morphology is preserved in agrammatism due to its heavy reliance on bound morphology on expressing grammatical meanings (Druks, 2016; MacWhinney, 1987; Slobin, 1991). While this also holds true for Tagalog due to its complex verbal morphology, in order to get the message across efficiently, adapting the production using language-dependent features must be capitalized, hence resulting to agrammatic speech (Fedorenko et al., 2023; Goral et al., 2019; Grodzinsky, 1990; Paradis, 2001). Moreover, because of the bilingual characteristic of most Tagalog-speakers, bilingualism enables its speakers to not only exploit one but two languages (English) that both allow the use of bare verb forms, as an accommodation to the costly demands of communication.

Conclusions of theoretical and clinical relevance

Tagalog offers a suitable research ground for theory testing in both narrative speech and experimental settings. Regardless of its many idiosyncrasies, Tagalog exhibits morphosyntactic deficit patterns similarly observed in multiple documented languages for agrammatism, both in agglutinative and non-agglutinative typologies. However, our findings challenge some assumptions about how agrammatism manifests in agglutinative languages. The widely used notion that agrammatism solely impairs syntax and not morphology (Anjarningsih et al., 2012; Nedergaard et al., 2020; Niemi et al., 1990; Sasanuma, 1989) is contested. Tagalog’s morphology and syntax are tightly linked and bare verbs result in legal lexical word forms in Tagalog. In our study, the omission of verbal affixes cascaded in to commission of wide range of morphosyntactic errors in non-fluent IWA’s narrative spontaneous speech. Due to its complex verbal system and allowance of bare verb use, many of these verbal features can advantageous for future experimentation specifically in terms of verb usage, such as verbal focus and thematic role assignment (Garcia & Kidd, 2020), realization of time reference (Bastiaanse et al., 2011), and verb argument structures (Faroqi-Shah, 2023; Thompson, 2003). With these future research directions, we can investigate what precisely motivates the heightened preference for bare forms over inflected forms in Tagalog agrammatism. Moreover, the individual variability in non-fluent IWA group in many parameters opens an opportunity for further characterization of compensatory strategies in agrammatism (Goral et al., 2019; Sahraoui & Nespoulous, 2012).

Clinically, this research offers an ecologically-valid method of analyzing aphasic speech, integrating diverse parameters that address language areas relevant for assessment and rehabilitation of individuals with aphasia (Dipper et al., 2021; Linnik et al., 2016; Stark et al., 2023). Although other parameters may be incorporated in protocol in future research protocols, the existing procedures could be applied by clinicians in their respective practices to determine whether persons with non-fluent Broca’s aphasia align with the Tagalog agrammatic profile described in this paper. Nonetheless, further initiatives are essential to establish reliability among Tagalog-speaking clinicians, ensuring a standardized quality in the clinical application of the protocol.

Acknowledgments

We would like to express our sincerest gratitude to all Filipino speech-language pathologists who extended their assistance in the recruitment process, data collection, transcription and analyses, and expert consultations. The authors are also thankful for all the participants and their families who devoted their time participating in this study. Mabuhay kayong lahat!

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors have no funding to report.

Appendix A

Agrammatic narrative spontaneous speech studies from agglutinative and polysynthetic languages

Table A1 Summary of agrammatic narrative speech studies in agglutinative and polysynthetic languages

Study	Language (Family)	Methods	Sentence Production	Grammatical Morphology	Verbs
Slobin (1991)	Turkish (Turkic)	N = 7 with Broca’s aphasia N = 10 controls Tasks: picture description and personal narratives	Use of canonical word order	Verb suffixes were used appropriately Case system is preserved Low diversity of verb inflection Correct order of verb morphemes	Limited set of verb forms
Arslan et al. (2016)	Turkish (Turkic)	N = 10 with Broca’s aphasia N = 10 controls Tasks: personal interview and picture description	Shorter MLU Lower percentage of correct sentences Fewer embeddings with non-finite verbs Normal embeddings with finite verbs	Verb inflection relatively intact Infinitive and narrative markers are produced slightly higher than normal	Normal frequency of lexical verbs Reduced verb diversity Lesser verb per utterance
Abuom & Bastiaanse (2012)	Swahili (Atlantic-Congo)	N = 6 with non-fluent aphasia Tasks: picture description and personal interview	Shorter MLU Lower proportion of grammatical sentences Low use of complex sentences	Verb inflection for past is more impaired Verb inflection is more preserved in Swahili than in English Frequent omission of free morphemes Substitution of inflections leading to tense/agreement errors	Verb tokens and diversity is similar to healthy controls
Sang (2015)	Swahili (Atlantic-Congo)	N = 6 with non-fluent aphasia N = 6 controls Tasks: picture description and personal narratives	Shorter MLU Simple sentences	Mostly errorless in verb inflection, except one subject who used bare verb stems	Preference to verbs with one internal argument Lower verb TTR
Lartey (2017)	Akan (Atlantic Congo)	N = 5 with aphasia N = 5 controls Tasks: personal interview	More grammatical errors Fewer embedded sentences Shorter MLU	Past verb inflection is more impaired	Normal verb diversity
Niemi et al. (1990)	Finnish (Uralic)	N = 2 with Broca’s aphasia N = 2 controls Tasks: personal interview and picture description	Shorter MLU Use of simple and short phrases and sentences Reduced grammaticality of syntactic structures Reliance on canonical word order	Few morphological errors on noun case markings Verb affixation used appropriately Verb affixes were replaced by less marked ones Omission of closed-class words	Verb frequency is similar to controls Verb TTR is within normal limits
Helasvuo et al. (2001)	Finnish (Uralic)	N = 4 with Broca’s aphasia Tasks: conversation and picture description	Word order patterns are often peculiar	Appropriate use of case endings Tendency to use uninflected noun phrase on some obligatory contexts Frequent use of unmarked 3^rd person verb form – typically in less frequent and morphologically complex verbs Auxiliaries in periphrastic forms are omitted	No data available
Sasanuma (1989)	Japanese (Japonic)	N = 2 agrammatic patients N = 2 controls Tasks: personal narratives, picture description, and story-telling	Short phrase length Simplified grammar	Omission of grammatical morphemes in obligatory contexts Verb inflection forms were preserved almost perfectly	Omission of copula, auxiliary verbs, and some main verbs
Halliwell (2000)	Korean (Koreanic)	N = 2 with agrammatic aphasia N = 2 controls Tasks: personal interview, storytelling, and picture description	Shorter MLU Preference for simple syntactic constructions and word order High proportion of single element utterances	Substitution of verb inflections resulting to tense errors Omission of nominal case markers Omission of grammatical elements Verbs are minimally inflected (non-reliance to zero morpheme)	Lexical choice of verbs is at par with controls More verbs are omitted than nouns
Nedergaard et al. (2020)	WestGreenlandic (Eskimo-Aleut)	N = 5 with expressive aphasia N = 5 controls Tasks: personal narrative, picture narrative, and picture description	Shorter MLU Production of less complex syntactic forms Normal speech rate	No difference in morphemes per word and per utterance No significant morphological impairment	Lower proportion of subordinate, valence-changing, and transitive verbs
Anjarningsih et al. (2012)	Indonesian(Austronesian)	N = 6 with Broca’s aphasia N = 15 controls Tasks: semi-standardized interview	Shorter MLU Simpler sentence structures Higher proportion of ungrammatical sentences (argument or predicate omission) Normal production of passive sentences	Fewer syntactic particles Overuse of inflectional affixes over derivational Low percentage of reduplication	Normal-to-Above Normal verb production
Chiang & Razak (2005)	Malay (Austronesian)	N = 1 with agrammatic aphasia Tasks: conversation covering different settings	Short and incomplete utterances Mostly noun phrase & negative clauses	No data available	Higher proportion of nouns over verbs
Laka & Korostola (2001)	Basque (Isolate)	N = 8 Broca’s aphasia Tasks: unspecified type of interview	Ungrammatical constructions	Nominal case marking and determiner errors Substitution errors on verb inflections Verbs substituted to nominalized form which lacks inflectional morphology Agreement errors leading to use of null verbal paradigms Creation of impossible auxiliaries (combining morphemes from conflicting paradigms)	Low frequency of auxiliary verbs Low frequency of inflected verb forms Errors in selecting intransitive forms
MacWhinney & OsmÁn-SÁgi (1991)	Hungarian (Uralic)	N = 9 with Broca’s aphasia N = 15 controls Tasks: picture description	High rate of argument omission, mostly subjects and indirect objects	High level of article and case marking omission Mostly substitution errors of functional morphemes Preference for unmarked nominals No clear evidence of unmarked verb substitution	No data available

Appendix B

Tagalog Spontaneous Speech Analysis – Description of Parameters

Utterance Segmentation

Utterance segments are primarily parsed based on the minimum syntactic unit, which is clauses. Prosodic features, such as intonation and pauses, serve as secondary criteria (Boxum et al., 2013). A simple independent clause is considered as one utterance. In cases where there are dependent clauses, they shall be combined with the main independent clause and counted as one utterance.

Example utterance chunk:

“Tumapos ang bola papunta sa bintana. So after nangyari yun, tumayo yung may-ari ng bahay at nakita niya ang bata na naglalaro ng bola.”

Table

Utterance 1	Tumapon ang bola sa bintana. independent clause “The ball flew to the window.”
Utterance 2	So after nangyari yun, tumayo yung may-ari ng bahay dependent clause independent clause “So after that happened, the owner of the house stood-up”
Utterance 3	at nakita niya ang bata na naglalaro ng bola. independent clause dependent clause “and he saw the child who is playing ball”

Rate of speech

Measured using words per minute (WPM), the total number of all identifiable words (including repetitions, self-corrections, hesitations, and the like) produced by the participants was divided by the total number of minutes of the participants’ speech (Saffran et al., 1989).

Mean Length of Utterance

Considered as the most useful and meaningful measure of syntactic complexity in both children (Templin, 1957) and adults with aphasia (Vermeulen et al., 1989; Wagenaar et al., 1975), mean length of utterance (MLU) was measured in two-ways: (a) MLU in words (MLUw) and (b) MLU in morphemes (MLUm). The latter being added to account for typological differences given that Tagalog is highly agglutinative in nature (Allen & Dench, 2015). Short responses such as “yes/no”, short direct responses to questions, repetitions to the elicitors’ speech, neologisms or stereotypic utterances, and repaired utterances were excluded from the count to avoid skewing of MLU values (Saffran et al., 1989; Vermeulen et al., 1989). To yield these measures, the total number of words and morphemes were divided by the total number of participants’ analyzable utterances.

Sentence types

To fully describe the participants’ performance, different sentence types that are both universal and unique to Tagalog’s structures were determined:

• Minor sentences – this type categorizes phrases/sentences that are classified as mazes (short expressions that are non-propositional) (DeDe & Salis, 2020), elliptical responses (Kolk & Heeschen, 1990) which are typically discourse-linked (Anjarningsih et al., 2012; Garcia & Kidd, 2022), and sentence fragments with either predicate or argument omissions. However, mazes will be excluded from the final analysis.

  Table

  Utterance 1

  Pusa aso buntot “Cat dog tail”

  Minor sentence

• Simple sentences – This type of sentence is considered to be the “basic structure” that is later used to form more complex constructions. It is essentially formed with a subject and a predicate which can either be verbal or non-verbal in nature (Malicsi, 2013). Utterances with a subject and a predicate, even though ungrammatical due to function word omissions (for example), is still considered a simple sentence.

  Table

  Utterance 1

  Pinagalitan siya ng nanay niya “He was scolded by his mother”

  Simple sentence

• Compound sentences – this category are for combinations of two independent clauses, and are must be clearly linked by a coordinating conjunction.

  Table

  Utterance 1	Naexcite yung pusa “The cat got excited”	Compound sentence
  Utterance 2	At hinabol niya yung paruparo. “and it ran after the butterfly”

• Complex sentences – a group of sentences that are formed by combinations of one independent clause and one or more dependent clauses (embeddings), which may or may not be glued together by a subordinating conjunction. Complex sentences may also be derived from multiple number of utterances with both independent and dependent clauses, as long as these utterances are linked by coordinating conjunctions.

  Table

  Utterance 1	So after nangyari yun, tumayo ang may-ari ng bahay “So after that happened, the owner of the house stood-up”	Complex sentence
  Utterance 2	at nakita niya ang bata na naglalaro ng bola. “and he saw the child who is playing ball.”

Number of Embeddings

A complex sentence can have one or more embedded/subordinate clauses in it such as complement, relative, adverbial, and report-type clauses. Aside from shorter MLU, non-fluent IWA are shown to have reduced use of embeddings or subordination (Prins & Bastiaanse, 2004). The total number of embeddings across the whole sample were counted (Arslan et al., 2016; Nedergaard et al., 2020; Saffran et al., 1989).

Table

Utterance 1	Sabi ko sa mga kasama ko, “uwi na tayo” “I told them, ‘let’s go home’”	Complex Sentence	1 embedding
Utterance 2	Kasi hindi ako makahinga “because I can’t breathe”
Utterance 3	“Magnenebulizer ako sa bahay”, sabi kong ganyan. “‘I will use the nebulizer at home’, I said.”	Complex Sentence	1 embedding
TOTAL EMBEDDINGS			2 embeddings

Grammaticality

This measures for the proportion of well-formed sentences (Saffran et al., 1989) that were judged to be appropriate, propositional, and discourse-licensed (Anjarningsih et al., 2012) in cases of argument dropping or predicate omission (a typical case for compound sentences). Ungrammatical utterances include sentences without a finite verb, omission of function words such as case markers, linkers, and affixes (Bondoc et al., 2018), and constructions with other remarkable errors may it be morphological, syntactic, and/or lexical in nature (Arslan et al., 2016).

Table

	Utterance	Sentence Type	Nature of Error	Grammaticality
Utterance 1	Umalis yung ibon “The bird went away”	Compound	Subject dropping in 2^nd clause but it is discourse-licensed	1
Utterance 2	Tapos lumipad. “and then it flew”
Utterance 3	Yung aso kinagat “The dog bit”	Simple	Missing argument	0
Utterance 4	Aso umalis para hanap pusa “Dog went away to seek cat”	Complex	Verb affix omission; Case marker omission; Missing argument	0
Utterance 5	At hinabol ng aso “and the dog chased”
TOTAL GRAMMATICALITY				1/3 or 33%

Predicate Types

As mentioned above, the simplest structure in Tagalog must contain a predicate. However, unlike other languages which typically rely on verbs alone for predication, Tagalog can either use a verbal or a non-verbal type of predicate (Malicsi, 2013). The average number of both predicate types per utterance were calculated (Anjarningsih et al., 2012).

Proportion of Morphological Classes

The degree of how functional morphology (such as free and bound morphemes) is used was computed, as non-fluent IWA are observed to prevalently omit these grammatical elements (Bates et al., 1987; Boye & Bastiaanse, 2018; Menn & Obler, 1990). Tagalog morphology is built upon three morphological classes namely roots, affixes, and particles (Nolasco, 2011). Roots (or stems) are any bare form capable of taking in one or more affixes (nominal, verbal, adjectival) to create more complex word forms (also includes numerals and pronouns). These forms can be combined with free morphemes called particles (including determiners, linkers/connectives, and enclitic particles) to add another layer of syntactic or semantic meaning. The number of occurrences per class in relation to the total number of morphemes were computed to yield the proportion of roots, affixes, and particles in each sample.

Table

Utterance	Root	Affix	Particles
Naramdaman ko yung pananakit ng aking katawan “I felt the aching of my body”	5	4	3
	ramdam, ko, sakit, akin, katawan	na-, -an (naramdaman) pa-, CVr (pananakit)	yung, ng, -ng (aking)
MORPHEME TOTAL: 12	5/12 or 42%	4/12 or 33%	3/12 or 25%

Number and Proportion of Verbs

Verbs are divided into two major types: modals and lexical verbs. Modals are a special group of verbs that cannot be inflected (e.g. “gusto” as in want, “kailangan” as in need). This includes the existential verbs (“mayroon” as in have). Lexical verbs, on the other hand, are the types that are capable of taking in inflectional markers. The total number of all verbs, alongside the proportion of modal and lexical verbs, were computed.

Percentage of Bare Verbs

Since it is allowed in the language and is observed to be used by typically developing children and adults in natural discourse (Garcia & Kidd, 2022; Marzan, 2013), as well as agrammatic speakers during experimental conditions, the number and percentage bare verb roots (or bare infinitives) used in each sample were also counted. Whether a root word is used as a bare verb is determined based on the overall context of the utterance (topic of the conversation, utterances surrounding the root word in question, word order, and/or lexical/grammatical elements preceding/following the target) and the assessment of a native speaker.

Finiteness Index

This is to measure the proportion of verbs inflected against all possible contexts (Boxum et al., 2013). The total number of correctly inflected finite verbs is divided to the total number of possible contexts where finite verbs are obligatory (Bastiaanse et al., 2002b; Saffran et al., 1989). In Tagalog, a key distinguishing factor in classifying finite versus non-finite verbs is the presence of aspectual marker, meaning if the verb is only inflected with a voice marker, it is considered an infinitive (Wurmbrand, 2013).

Table

Utterance	Verb	Inflected for Aspect?	Obligatory Context
Nagugutom na ang mga ibon “The birds are getting hungry”	nagugutom	1	1
Hanap yung nanay ng pagkain “The mother look for food”	hanap	0 (bare verb)	1 (humahanap/ naghahanap)
Uwi na tayo, sabi ng ibon “Let’s go home, the bird said”	uwi	0 (bare)	0 (Infinitive – umuwi)
	sabi	0 (bare)	1 (sinabi)
Kailangan kong maglaba “I need to do laundry”	maglaba	0 (infinitive)	0 (Infinitive)
Aakyatan ibon ng pusa “Climb on bird by cat”	aakyatan	0 (incorrectly inflected)	1 (aakyat or aakyatin)
TOTAL FINITENESS INDEX		1/4 or 0.25

Verb Diversity

Traditionally, the overall lexical richness of narrative speech is measured using the type-token ratio (TTR) across all word classes (Templin, 1957; Vermeulen et al., 1989). For the purpose of cross-linguistic comparison in agrammatic patterns, we specifically computed for the verb diversity alone. To the best of our knowledge, there is no standard approach in computing the TTR in Philippine-type languages. In Tagalog, verb lemmas are derived not from the roots but from the infinitival forms. Therefore, the infinitival forms will serve as the basis for counting verb types. All potential inflected forms of the base infinitive will be categorized under the same type. Bare forms will be counted in the same lemma of an existing infinitive form with the same base root, else it will have its own type if there is no inflected form available. The total number of unique verb types is then divided by the total number of lexical verbs or verb tokens in the sample.

Table

akyat – aakyat – naalala – maalala – inaalala – sabi – umuwi – umuuwi – nakauwi – natamaan - tinamaan
Type (Infinitive form)	Tokens
umakyat	2 = aakyat, akyat (bare)
maalala	2 = maalala, naalala
alalahanin	1 = inaalala
umuwi	2 = umuwi, umuuwi
makauwi	1 = nakauwi
sabi	1 = sabi (bare)
matamaan	1 = natamaan
tamaan	1 = tinamaan
TOTAL TTR	8 types/11 tokens or 0.727

Noun-Verb Ratio

In the context of Indo-European languages, individuals with agrammatism use fewer verbs than normal, and in turn, overuse nouns as compensation (Zingeser & Berndt, 1990). In this measure, the proportion of nouns against verbs was computed by dividing the total number of nouns over verbs.

Percentage of Code-Mixed Words

Language mixing and switching in non-fluent IWA is suggested to be a strategy to maximize communication in bilinguals (Goral et al., 2019; Lerman et al., 2019). Since Tagalog-speakers from Metro Manila are typically Tagalog-English bilingual in nature (Bautista, 2004), the percentage of total number language-mixed words (English) divided by the total number of words was calculated. This includes all nouns, verb roots used in Tagalog verb conjugation, and others. Only proper nouns in English were not counted (e.g. brand names such as “Burger-King”).

Appendix C

Raw data of both non-fluent IWA (NF) and NBD (HC) groups

Table C1 Raw data for sentence production parameters

Participants	# Words	WPM	MLUw	MLUm	# Utterances	# Sentences	# Minor	# Simple	# Compound	# Complex	# Embed.	# Gram.
NF001	74	15.89	1.61	1.72	46	46	38	7	0	1	1	14
NF002	146	18.04	2.81	3.08	52	52	32	13	0	7	7	7
NF003	170	45.64	2.70	3.05	63	62	34	26	2	0	0	15
NF004	121	31.40	3.46	4.57	35	34	10	21	1	2	2	12
NF005	159	25.60	2.21	2.49	72	72	50	15	0	7	7	25
NF006	255	62.68	4.40	4.71	58	50	15	16	8	11	11	13
NF007	131	31.80	2.26	2.72	58	58	38	15	0	5	6	17
NF008	137	37.14	3.19	3.81	43	43	23	14	0	6	7	13
NF009	159	28.14	3.12	4.20	51	42	18	15	5	4	4	16
NF010	233	21.18	3.76	4.73	62	45	14	15	10	6	6	16
HC001	300	81.24	11.77	15.73	26	13	0	0	1	12	21	12
HC002	300	86.73	10.97	15.38	29	11	0	1	1	9	25	10
HC003	300	93.13	8.51	11.54	35	19	1	0	4	14	27	18
HC004	300	93.83	9.48	13.17	31	19	0	2	4	13	22	17
HC005	300	97.70	8.89	13.09	35	20	0	3	5	12	25	19
HC006	300	71.31	7.92	11.69	39	29	0	6	4	19	29	25
HC007	300	71.86	7.50	9.95	42	30	0	9	5	16	27	29
HC008	300	56.66	9.24	13.76	33	24	0	4	2	18	26	23
HC009	300	96.13	9.27	12.46	33	26	0	8	1	17	28	25
HC010	300	88.98	7.42	9.73	41	33	0	14	2	17	22	31
HC011	300	110.06	11.07	15.33	27	22	0	5	1	16	24	19
HC012	300	136.96	7.88	10.38	40	21	0	2	8	11	15	21

Table C2 Raw data for functional morphology and verb-related parameters

Participants	# Preds.	# Non-verb Pred.	# Verb Pred.	# Verbs	# Modal	# Lexical	# Bare	# Finite	# Oblig Con	Verb Types	Verb Tokens	# Noun	# Root	# Affix	# Particle	# English
NF001	35	10	25	25	11	14	10	3	14	10	14	18	63	4	12	10
NF002	31	5	26	26	1	25	14	3	21	19	25	97	131	12	13	93
NF003	47	9	38	40	9	31	17	13	29	22	31	66	126	21	42	66
NF004	35	0	35	37	3	34	7	16	30	25	34	39	84	38	39	5
NF005	65	28	37	35	9	26	18	5	24	18	26	41	123	22	40	25
NF006	53	20	33	34	13	21	14	5	20	11	21	51	159	13	101	21
NF007	53	15	38	39	7	32	16	11	31	27	32	52	114	26	17	35
NF008	46	14	32	34	9	25	6	15	24	17	25	25	98	28	41	18
NF009	52	6	46	47	9	38	8	20	36	29	38	29	103	52	60	11
NF010	65	12	53	55	9	46	12	29	42	31	46	44	168	60	24	10
HC001	49	3	46	50	8	42	0	37	37	34	42	59	182	77	150	19
HC002	53	11	42	50	8	42	1	33	34	34	42	71	193	94	159	21
HC003	58	11	47	51	7	44	1	38	38	36	44	49	166	81	120	51
HC004	54	11	43	49	5	44	0	33	33	38	44	63	189	98	123	39
HC005	60	13	47	51	7	44	1	34	35	37	44	55	201	114	133	4
HC006	67	12	55	61	8	53	0	44	44	43	53	69	197	123	143	11
HC007	69	10	59	64	8	56	1	47	47	45	56	65	210	94	116	28
HC008	58	13	45	51	7	44	1	35	37	37	44	61	203	103	248	22
HC009	60	13	47	50	6	44	1	40	41	37	44	63	200	89	122	55
HC010	61	5	56	63	6	57	10	39	46	43	57	62	212	75	112	16
HC011	51	4	47	50	6	44	1	37	38	33	44	57	198	84	132	14
HC012	54	6	48	55	10	45	1	40	40	35	45	57	210	80	125	33