Disfluencies in spontaneous speech in persons with low-grade glioma before and after surgery

ABSTRACT

Impaired lexical retrieval is common in persons with low-grade glioma (LGG). Several studies have reported a discrepancy between subjective word-finding difficulties and results on formal tests. Analysis of spontaneous speech might be more sensitive to signs of word-finding difficulties, hence we aimed to explore disfluencies in a spontaneous-speech task performed by participants with presumed LGG before and after surgery. Further, we wanted to explore how the presence of disfluencies in spontaneous speech differed in the participants with and without objectively established lexical-retrieval impairment and with and without self-reported subjective experience of impaired language, speech and communication. Speech samples of 26 persons with presumed low-grade glioma were analysed with regard to disfluency features. The post-operative speech samples had a higher occurrence of fillers, implying more disfluent language production. The participants performed worse on two of the word fluency tests, and after surgery the number of participants who were assessed as having an impaired lexical retrieval had increased from 6 to 12. The number of participants who experienced a change in their language, speech or communication had increased from 9 to 12. Additional comparisons showed that those with impaired lexical retrieval had a higher proportion of false starts after surgery than those with normal lexical retrieval, and differences in articulation rate and speech rate, favouring those not having experienced any change in language, speech or communication. Taken together, the findings from this study strengthen the existing claim that temporal aspects of language and speech are important when assessing persons with gliomas.

KEYWORDS:

• Low-grade glioma
• disfluency
• pauses
• lexical impairment
• subjective language complaints

Introduction

Low-grade glioma (LGG) is a type of slow-growing brain tumour. It is often located in or near eloquent areas, that is, areas involved in language, cognitive, motor or sensory functions, and can therefore affect those functions (Duffau & Capelle, 2004). Most patients with LGG undergo surgery. Because of the tumour’s infiltrative nature, surgical resection may cause impairments in language and other cognitive abilities (Coburger et al., 2022). However, the vast majority experience only mild language impairments, mainly in the domain of lexical retrieval (e.g. Norrelgen et al., 2020; Papagno et al., 2012; Antonsson et al., 2018a). A person with lexical-retrieval difficulties needs more effort or time to find the desired word and as a result may have a disfluent language production. In spoken language, this may manifest itself in various ways, including as an increase in silent pauses, a higher presence of fillers such as ‘er’ or ‘hm’, and more self-interrupted sentences (where the person starts saying something but leaves the phrase incomplete). Such word-finding difficulties can cause frustrations which has been reported by Fama et al. (2021) who investigated how persons with post-stroke aphasia experience word-finding difficulties. There is to our knowledge no existing research regarding how persons with a brain tumour experience word-finding difficulties.

Both in clinical practice and in most studies examining language impairment in persons with gliomas, formal standardised tests are used to assess different aspects of language, including lexical retrieval (e.g. Antonsson et al., 2018b; Bello et al., 2007; Papagno et al., 2012). However, such tests sometimes fail to identify subtle language difficulties; several studies have reported a discrepancy between subjective complaints and objective test results (Antonsson, 2017; Brownsett et al., 2019; Gehring et al., 2015; Mooijman et al., 2021; Racine et al., 2015; Satoer et al., 2012). In a study investigating cognitive impairment in patients between 19 and 74 years of age, with gliomas in the left hemisphere, close to 60% reported having word-finding difficulties (Satoer et al., 2012) but all were assessed as having normal language or only a minimal discernible handicap according to the Aphasia Severity Rating Scale (Goodglass & Kaplan, 1983). Hence six in ten participants in that study had difficulties that would presumably not be apparent to a listener. This highlights a need to find other ways of examining subjective language complaints so as to better capture such subjectively experienced word-finding difficulties in language assessment.

Analysis of spontaneous speech, using qualitative or quantitative approaches, can be used to assess different linguistic levels, including fluency of speech. However, only a few studies have analysed spontaneous-speech production in brain-tumour patients. In one of them, Satoer et al. (2013) performed a linguistic analysis of spontaneous speech, finding that patients with gliomas near language-eloquent brain areas produced more incomplete sentences than a control group, both before and after surgery. The authors suggested that this could be due to a lexical problem. In a follow-up study including data from one year after surgery, they observed similar results but also a decreased lexical diversity and a shorter mean length of utterances compared with a control group (Satoer et al., 2018). Rofes et al. (2017) compared the outcomes of formal language assessment with those of a linguistic analysis based on a spontaneous-speech task in five patients with glioma. Their explicit aim was to test whether the spontaneous-speech task could replace language testing; they concluded that the different types of assessments complemented each other and detected comparable language impairments. The results from these previous studies are not clear cut, but they indicate that lexical deficits observed in tests are noticeable also in spontaneous speech. However, this needs to be investigated further.

Disfluencies are a common and natural part of the spoken language of all humans; they reflect the process of planning and producing language (MacGregor et al., 2010). Another type of disfluencies can also occur due to causes such as stuttering, but those types of disfluencies are not the focus in this study. Silent pauses, which are a common type of disfluency, are associated with lexical retrieval and sentence planning (Kircher et al., 2004), but they are also related to turn-taking management (Lundholm Fors, 2015) and may be used for rhetorical purposes (MacGregor et al., 2010). Fillers (filled pauses) are also common and seem to be employed for several reasons such as marking uncertainty and word choice, structuring discourse and managing turn-taking (Kosmala & Crible, 2021). Disfluencies also include fillers (filled pauses), repetitions, repairs and prolonged syllables and studies have shown that repetitions, restarts and repairs occur up to six times per hundred spoken words (Bortfeld et al., 2001; Tree, 1995). To listeners, an excessive occurrence of disfluencies not only signals that the speaker appears to be having problems finding the right word but may also affect their comprehension of the speaker’s message in the sense that a disfluency interrupts the ongoing processing that enables listeners to understand what is being said (MacGregor et al., 2010). A higher frequency of disfluencies has been reported in the speech of persons with different types of neurological conditions such as mild cognitive impairment and Alzheimer’s disease (Boschi et al., 2017; Szatloczki et al., 2015), Parkinson’s disease (Alvar et al., 2019) and stroke-induced aphasia (Angelopoulou et al., 2018). When it comes to spontaneous speech in glioma patients, the research is scarce and the analyses performed in previous work have mainly been based on transcripts and temporal aspects of language production have only been assessed perceptually, for example through the annotation of silent pauses and fillers. In the present study, by contrast, we highlight the temporal aspects by analysing different types of disfluencies in recordings of spoken-language production. Our primary aim was to explore disfluencies in a spontaneous-speech task performed before and after surgery by persons with presumed LGG. A secondary aim was to explore how the presence of disfluencies in spontaneous speech differed depending on whether participants had an objectively identified impairment to their lexical retrieval as well as on whether they reported a subjective experience of impaired language, speech and communication. The research questions were the following:

• Are there any differences with regard to disfluencies in spontaneous speech before surgery and at a three-month follow-up after surgery?

• Is there a difference before and after surgery with regard to lexical retrieval or the number of persons who experience a change in language, speech and communication?

• Are there any differences in disfluency measures when dividing the participants into:

  1. those who had an objectively impaired lexical retrieval according to naming and word-fluency tests and those for whom no such impairment was detected, before and after surgery?

  2. those who had experienced a change in language, speech and communication and those who had not experienced such a change, before and after surgery?

Materials and methods

Participants

The study included 26 patients diagnosed with presumed LGG between November 2014 and September 2016 at the Sahlgrenska University Hospital, Gothenburg, Sweden. These patients were all included in a project investigating the impact on language and communication after surgical treatment for LGG. The inclusion criteria were age over 18 years, Swedish as native language, no severe developmental language or cognitive disorder, and a diagnosis of presumed LGG. No tests were done to screen for developmental language or cognitive disorder, but all participants were asked about previous history of language and communication difficulties or other medical history relating to communication. The diagnosis of presumed LGG was based on the patients’ MRI scans, physical examination and medical history. Initially, 32 persons met these criteria and participated in language testing. However, six of them were excluded, three for not participating in the three-month follow-up, two owing to technical recording errors and one for not talking about one of the topics requested. That participant instead talked about a well-known memory, which was repeated both before and after surgery in similar terms. The person in question had a congenital condition and a learning disability, which might explain why the instructions given were not followed. After surgery, histological examination showed that 5 of the final 26 participants had a tumour of grade III or IV (i.e. high-grade rather than low-grade glioma). However, since the relevant inclusion criterion was (pre-operatively) presumed LGG, all 26 participants were included in the analysis. The majority of the participants, 18 out of 26, had a tumour in the left hemisphere. Of these, 11 had a tumour in a language-eloquent area, i.e. inferior frontal, superior temporal, insula and inferior parietal areas in the left hemisphere, as categorised by a neurosurgeon in accordance with Chang et al. (2008). The remaining eight participants had a tumour in the right hemisphere. An overview of patient characteristics, broken down by tumour location, can be seen in Table 1.

Table 1. Demographic and tumour characteristics of all participants (total n = 26).

			Tumour data							Impaired lexical retrieval		Subjective change in language
	ID	Sex/Age/ Education (years)	Handedness	Location	Histology	Seizures	Previous treatment	Surgical treatment (EOR)	Post-op treatment	Pre-op	Post-op	Pre-op	Post-op
LH, language-eloquent area	4	M/60/13	R	Temporal, insula	OA II	No	No	2 PA	No	Yes	Yes	No	Yes
	6	F/46/16	R	Frontal	OA III	Yes	R, RxT, chemo	2 A	Low dose TMZ	Yes	Yes	No	Yes
	8	F/55/9	n/a	Frontal	A IV, GBM	Yes	Biopsy, RxT	3 A*	TMZ	Yes	Yes	Yes	Yes
	12	M/31/22	R	Frontal	A IV, GBM	Yes	No	2	RT+TMZ	No	Yes	No	Yes
	15	M/46/15	R	Multi-focal	A II	Yes	No	4	TMZ	Yes	No	No	No
	16	M/62/14	R	Frontal	OA III	Yes	R, chemo	3 A	SRS	Yes	Yes	Yes	Yes
	23	M/55/16	R	Temporal, insula	A IV, GBM	Yes	R, RxT	2 A*	TMZ	No	Yes	Yes	No
	24	M/31/16	L	Insula	A II	Yes	R	3 PA	Proton	No	Yes	Yes	Yes
	25	M/25/12	R	Fronto-temporal	O III	Yes	No	1	Proton+TMZ	Yes	Yes	Yes	Yes
	32	M/53/13	R	Frontal	O II	Yes	No	2	TMZ	No	No	No	No
	36	M/24/12	L	Frontal	A II	No	No	2 A	No	No	No	No	No
LH, non-language eloquent area	1	M/45/11	R	Frontal	OA III	Yes	No	2 A	RT+ TMZ	No	Yes	No	Yes
	2	F/37/19	R	Temporal	A II	No	No	2	No	No	No	No	No
	5	F/57/16	R	Frontal	OA II	Yes	No	2 A*	RT	No	Yes	No	Yes
	19	M/26/19	R	Temporal	A II	Yes	No	1	No	No	No	No	No
	27	M/56/15	R	Temporal	A II	Yes	No	2	No	No	No	Yes	Yes
	29	M/26/16	R	Temporal	Ganglio- glioma I	Yes	No	2	No	No	No	No	No
	34	M/67/17	R	Frontal	O II	No	No	1	No	No	No	No	No
Right hemisphere	7	M/43/20	R	Gyrus cinguli	OA III	Yes	No	2	RT+ TMZ	No	No	No	No
	11	F/56/12	R	Insula, frontal	O II	Yes	No	2	TMZ	No	No	No	No
	13	F/39/16	R	Frontal	O II	Yes	No	2	PCV	No	No	No	No
	17	F/34/15	R	Parietal	O II	Yes	R, RxT, chemo	1	No	No	No	Yes	Yes
	20	F/42/12	R	Frontal	A II	Yes	No	1	No	No	No	No	No
	26	M/44/17	R	Frontal, temporal, insula, thalamus	O II	Yes	No	3	TMZ	No	No	No	No
	28	F/62/11	R	Frontal	A II	Yes	R	3	RT	No	Yes	Yes	No
	35	M/43/11	L	Frontal	Ganglio- glioma I	Yes	No	1	No	No	Yes	Yes	Yes

LH = left hemisphere. Handedness: R, right; L, left. Histology: OA = oligoastrocytoma, A = astrocytoma, O = oligodendroglioma. Previous or post-operative treatment: R = resection, RT = radiotherapy. Surgical treatment: SRS = stereotactic radiosurgery, Chemo = chemotherapy, TMZ = temozolomide, PCV = procarbazine, CCNU and vincristine, Proton = proton-beam radiotherapy, EOR = extent of resection: (1) Gross total resection = 100%, (2) Sub-total resection = 90%–100%, (3) Partial resection < 90%, (4) Biopsy. A = awake surgery, PA = planned awake surgery. *Resection stopped because it was difficult for the patient to stay awake.

Note: The participants are grouped by tumour location: language-eloquent areas of the left hemisphere, non-language eloquent areas of the left hemisphere or right hemisphere. Eloquent areas are defined in accordance with Chang et al. (2008).

Data collection

All data collection was carried out by the first author. The spontaneous-speech task was performed as part of a comprehensive language assessment that also included formal aphasia assessments including tests of lexical retrieval, high-level language ability and assessment of aphasia severity as well as a writing task. The spontaneous-speech task was the first task administered after participants were asked questions about their background for purposes of establishing their medical history and after they had the test procedure described to them. Results from those formal language tests and analyses of that writing task have been previously reported in Antonsson et al. (2018a; 2018b; 2018c).

Spontaneous-speech task

The speech material used was elicited from participants by asking them to talk a little about a familiar topic such as a hobby, their family or their work. At the pre-operative assessment, each participant could choose freely which of the topics suggested to talk about. However, at the post-operative three-month follow-up, the examiner steered participants towards talking about a topic other than that chosen at the pre-operative assessment. The rationale for using another topic was to avoid repetition of similar information at the post-operative follow-up due to the relatively short time span between the two assessments. The task was recorded using a video camera, and audio files of each speech sample were extracted from the recordings. The recordings varied greatly in length (pre-operative range: 2 min 33 sec to 8 min 45 sec, post-operative range: 1 min 35 sec to 10 min 33 sec). For this reason, only the first two minutes of each sample were transcribed. In the few cases where a participant spoke for less than two minutes, their recording was transcribed in its entirety. The cut-off of two minutes was chosen to ensure that each sample would be focused on a single topic; many of the participants who spoke for several minutes tended to drift to other, related topics.

Data preparation

The recordings were transcribed orthographically by a certified speech-language pathologist with experience in transcribing the speech of persons with a neurological impairment. The transcribed samples had a median number of words of 301.5 at the pre-operative assessment and 268.5 at the pre-operative assessment (pre-operative range: 167–372 words, post-operative range: 136–459 words. The transcriber was instructed to annotate any fillers, false starts (self-interrupted words), self-interrupted sentences, paraphasias and neologisms, and repetitions. Next, the speech was segmented into sentences. Segmentation was based on clauses, defined as having to include one finite verb; a sentence was defined as consisting of one or several clauses making up a unit of meaning. In addition, segmentation was also based on the speakers’ prosodic markers indicating a sentence break; for example, a falling intonation pattern could be indicative of the end of an utterance and hence mark a sentence break. While there are various manuals and guidelines available for segmentation of speech into sentences or utterances, we ultimately chose to define our own guidelines to take into account both syntax and prosody. All questions asked and comments made by the examiner were also annotated and later excluded from all analyses. At a later stage, the transcriptions were checked and corrected as needed by the first author, who is also a certified speech-language pathologist with experience in transcribing the speech of persons with a neurological impairment. The process of improving the transcriptions was an iterative one, where changes were made continuously throughout the analysis process. Each transcription was checked against the audio file at least five times. The transcriptions were aligned with the audio recordings using Webmaus (Kisler et al., 2017), and subsequent manual corrections were made using Praat (Boersma, 2002).

Analysis of disfluencies in spontaneous speech

To analyse disfluencies in spontaneous speech, various measures reflecting different aspects of speech fluency were used. The first step involved establishing, for each recording, the total recording time (excluding the examiner’s utterances), the total articulation time (total recording time excluding silent pauses and fillers) and the total number of words (including paraphasias but excluding fillers). Then the speech rate (in words per minute) was calculated as the total number of words divided by the total recording time, to give an overall indication of how fast participants spoke. The articulation rate (also in words per minutes) was calculated as the total number of words divided by the total articulation time, to reflect how fast the words were articulated. Further, the total articulation time was divided by the total recording time to obtain the articulation ratio, to reflect how much participants spoke in relation to their pause time.

Temporal aspects of the data were measured in Praat, using a TextGrid aligned with the audio signal. In a second step, a Praat script was used to identify all silent intervals (equivalent to empty stretches in the TextGrid) longer than 120 milliseconds in recordings that did not include speech or any other sounds produced by participants, such as coughing or laughing; these are called ‘silent pauses’. The cut-off of 120 milliseconds was used since it has been identified as the threshold for when a listener can perceive an acoustic silence in speech (Heldner, 2011). The silent pauses were measured and durational data was extracted, also through a Praat script. ‘Fillers’ were defined as sounds that indicate planning or hesitation but lack lexical content, such as ‘uh’ and ‘um’. To analyse these, we used a list of all fillers that had been tagged as such in the manual transcription, and from this list we extracted durational features from the TextGrid through the use of a Praat script. For both silent pauses and fillers, the proportion of the total recording time occupied by them (silent-pause ratio and filler ratio), their mean length in seconds (silent-pause length and filler length) and their frequency in terms of their number relative to the total number of words (silent pauses/words and fillers/words) were calculated. Third, a number of additional measures reflecting specific types of disfluencies – false starts, self-interrupted sentences, paraphasias and neologisms, and word repetitions – were used, reflecting the proportion of each type relative to the total number of words (or sentences in the case of self-interrupted sentences). A false start is where a speaker starts saying a word but does not complete it. A self-interrupted sentence is a similar phenomenon: a speaker begins a sentence but does not complete it. Paraphasias in this context refer to phonological paraphasias only, since it is difficult to consistently detect semantic paraphasias in a free task such as the one used here. Neologisms were merged with paraphasias into a single category since they both reflect some kind of word-level error. If several disfluencies occurred in a row, they were analysed as separate instances, the only exception from this rule being that repetitions where two or more words were repeated after each other were counted as a single occurrence. An overview of the measures used, and some examples, can be found in Table 2.

Table 2. Overview of variables used.

Variable	Definition	Examples
Temporal variables:
Speech rate (words/minute)	All words* including paraphasias/total recording time**
Articulation rate (words/minute)	All words* including paraphasias/total articulation time
Articulation ratio (%)	Articulation time/total recording time**
Silent-pause ratio (%)	Total length of silent pauses above 120 ms/total recording time**
Silent-pause length (seconds)	Mean length of silent pauses above 120 ms
Silent pauses/words (n)	Number of silent pauses above 120 ms/total number of words
Filler ratio (%)	Total filler time/total recording time**	‘uh’ or ‘um’
Filler length (seconds)	Mean length of fillers
Fillers/words (n)	Number of fillers/total number of words*
False starts (%)	Number of false starts/total number of words*	‘and each time it took well about ten s- or um about five seconds’
Linguistic variables:
Self-interrupted sentences (%)	Number of self-interrupted sentences/total number of sentences	‘but I still think it is… when I do get out it’s nice’
Paraphasias and neologisms (%)	Number of paraphasias and neologisms/total number of words*	Paraphasia: ‘flowder’ instead of ‘flower’ Neologism: ‘bamber’ instead of ‘band-aid’
Repetitions (%)	Number of whole word repetitions/total number of words*	‘and then we will travel travel back by train’

* All words encompass all articulated words including paraphasias, neologisms, self-interrupted sentences, and repetitions. Fillers and false starts were not regarded as words and hence not included. **The examiner’s comments are excluded from the total recording time, which is used to calculate the speech rate, articulation percentage, silent-pause ratio and filler ratio.

Assessment of lexical-retrieval ability and screening procedure for subjective complaints

To assess lexical retrieval, tests of naming and word fluency were used. The Boston Naming Test (BNT) is a test of confrontation naming where participants are asked to name pictures of objects. In the present study, the BNT was presented on a computer screen (digitalisation of picture material with the permission of the copyright holder: Kaplan et al. (2001)). The test was scored and compared with Swedish norms in accordance with Tallberg (2005). Word-fluency tests measure a person’s ability to generate words in a particular category within a limited time. The tests included here were letter-fluency tests designed to yield words beginning with certain letters (F, A and S) and semantic-fluency tests designed to yield words belonging to certain categories (animals and verbs). They were administered and scored in accordance with Tallberg et al. (2008). A person was deemed to have impaired lexical retrieval if he or she scored more than two standard deviations below the norm on at least one of the tests (the BNT or a word-fluency test).

When it comes to subjective i.e. self-reported complaints, all of the participants were asked before their surgery whether they had experienced any change in their language, speech or communication that could be related to the tumour. They were given examples of different types of symptoms such as difficulties in word-finding, articulation, language comprehension, listening to several speakers at once, or reading or writing. At the three-month follow-up after surgery, they were asked whether they had experienced any such change after their surgery compared with the pre-operative situation. The participants’ responses to these questions were divided into two categories: ‘yes’ (those who clearly had experienced some kind of change or impairment) and ‘no’ (those who had not experienced any change). Participants who were uncertain were asked to elaborate upon any changes that they had experienced; if they remained unsure whether they had experienced any change, their response was scored as ‘no’.

An overview of the individual participants’ categorisation in terms of objectively established normal or impaired lexical-retrieval ability and in terms of subjective experience of changes in language, speech and communication is given in Table 1. To illustrate their categorisation before and after surgery as well as the overlap between the two different kinds of categorisation, Figure 1 visualises the numbers of participants who had experienced a change in their language, speech or communication and/or were found by formal tests to have a normal or impaired lexical-retrieval ability, before and after surgery.

Figure 1. Sankey chart illustrating the numbers of participants who experienced a change in their language, speech or communication and/or had an objectively established normal or impaired lexical-retrieval ability before and after their surgery.

Statistical analysis

All data were visually inspected and analysed using the Shapiro – Wilk test of normality to determine whether parametric or non-parametric tests should be used for the statistical comparisons. The comparisons between pre-operative and post-operative performance were analysed using either a paired t-test or the Wilcoxon signed-rank test, depending on the data distribution. McNemar’s test was used to compare the number of participants who had an impaired lexical retrieval before and after surgery. The same method was used to compare the number of participants who experienced a change in language, speech or communication before and after surgery.

The comparisons between groups performed to answer the third research questions were analysed using the Mann – Whitney U test for independent samples. As multiple comparisons were made, a Bonferroni correction was performed; the level of statistical significance was set to p < 0.004 for the comparisons for the whole group before and after surgery, p < 0.008 for the comparison of tests of lexical retrieval and subjectively experienced change in language, speech or communication, and p < 0.004 for the comparisons between groups based on objectively established lexical-retrieval ability. For the comparisons between groups based on subjective experience of changes in language, speech or communication – the Bonferroni-corrected p-level was set to p < 0.003. Significant results are reported at the Bonferroni-corrected level and at p < 0.05. IBM SPSS Statistics version 23 was used for computation.

Ethical considerations

The present study is covered by ethical approval (reference number: 625–14) given by the Regional Ethical Review Board of Gothenburg. The participants were informed that they could withdraw their participation at any time. All data were coded and anonymised.

Results

Disfluencies in spontaneous speech before and after surgery

Table 3 presents an overview of the participants’ disfluency-measure scores before and after surgery. The only difference that was significant at the Bonferroni-corrected level concerned the proportion of fillers, which were more frequent after surgery: before surgery the median value was 0.03 fillers/word, but after surgery it was 0.05 fillers/word (Z = −3.22, p = 0.001). In addition, changes at the p-level of .05 were observed for articulation percentage, silent-pause length and silent pauses/word. All of these changes point in the direction of more disfluent performance after surgery: the proportion of speech in each speech sample was lower after surgery, while silent pauses were longer on average and accounted for a larger percentage of the total recording time after surgery.

Table 3. Comparison of disfluencies in spontaneous speech before and after surgery (n = 26).

	Before surgery		Follow-up 3 months after surgery
	M	Mdn	M	Mdn
	(SD)	Min – Max	(SD)	Min – Max	Sig.
Speech rate (words/minute)	187	185	175.6	172	t = 1.81
	(30)	(125–260)	(42.9)	(97–277)	p = 0.083
Articulation rate (words/minute)	239	234	236	239	t = .55
	(30)	(187–313)	(35)	(169–315)	p = 0.590
Articulation ratio (%)	78.2	78.7	73.6	75.6	t = 2.68
	(7.5)	(64–89)	(9.6)	(52–90)	p = 0.013*
Silent-pause ratio (%)	13.0	13.5	15.4	15.0	Z = −2.04
	(4.7)	(4.4–22.3)	(5.7)	(4.2–27.1)	p = .052
Silent-pause length (seconds)	0.56	0.56	0.63	0.59	Z = −2.27
	(0.13)	(0.34–0.79)	(0.17)	(0.24–0.96)	p = .023*
Silent pauses/words (n)	0.10	0.09	0.11	0.11	t = −2.53
	(0.04)	(0.04–0.18)	(0.05)	(0.03–0.27)	p = .018*
Filler ratio (%)	4.9	3.2	6.0	5.0	Z = −1.33
	(3.9)	(0.3–14.2)	(4.8)	(0.4–22.2)	p = .182
Filler length (seconds)	0.42	0.41	0.39	0.36	Z = −1.05
	(0.11)	(0.20–0.66)	(0.12)	(0.23–0.71)	p = .292
Fillers/words (n)	0.05	0.03	0.07	0.05	Z = −3.22
	(0.03)	(0.003–0.14)	(0.06)	(0.007–0.28)	p = .001**
False starts (%)	0.2	1.3	0.5	0.7	Z = −.39
	(0.2)	(0–4.9)	(0.8)	(0–8.0)	p = .696
Self-interrupted sentences (%)	10.2	10.1	10.0	7.8	Z = −.80
	(7.4)	(0–29.6)	(10.4)	(0–40.0)	p = .424
Paraphasias and neologisms (%)	0.2	0	0.4	0	Z = −1.45
	(0.2)	(0–0.7)	(0.8)	(0–3.3)	p = .148
Repetitions (%)	0.4	0.3	0.5	0.4	Z = −.99
	(0.5)	(0–2.0)	(0.6)	(0–2.0)	p = .322

Note: *significant at p-level .05. **significant at the Bonferroni-corrected level of p < 0.004.

Performance on tests of lexical retrieval and presence of subjectively experienced change in language, speech or communication before and after surgery

Table 4 presents the participants’ performance on tests of lexical retrieval and the presence subjectively experienced change in language, speech or communication. The participants performed worse on two of the word fluency tests, letter fluency and the semantic category Animals. Only the category Animals survived a Bonferroni correction.

Table 4. Comparison of tests of lexical retrieval and subjectively experienced change in language, speech or communication, before and after surgery (n = 26).

	Before surgery		Follow-up 3 months after surgery
	M	Mdn	M	Mdn
	(SD)	Min – Max	(SD)	Min – Max	Sig.
Tests of lexical retrieval:
BNT	51.3	53	49.3	52	Z = −.945
	(5.1)	(37–58)	(10.2)	(14–58)	p = .344
FAS (letter fluency)	40.7	42.5	35.3	36	Z = −2.453
	(14.2)	(14–71)	(17.8)	(6–64)	p = .014*
Animals	21.1	20	15.7	15	Z = −2–762
	(5.9)	(11–32)	(8.9)	(0–32)	p = .006**
Verbs	17.8	19.5	18.8	17.5	Z = −.864
	(6.3)	(6–30)	(7.7)	(4–31)	p = .388
Impaired lexical retrieval (n):	6 (23.1%)		12 (46.2%)		p = 0.070
Experienced change in language, speech or communication (n):	9 (34.6%)		12 (46.2%)		p = 0.453

Note: *significant at p-level .05. **significant at the Bonferroni-corrected level of p < 0.008.

There was an increase from 6 to 12 participants who were assessed as having an impaired lexical retrieval after surgery, but this change was not significant according to the exact McNemar’s test. Nor was the increase in the number of persons who experienced a change in their language, speech or communication, which increased from 9 to 12 participants after surgery.

Comparison of disfluency measures between participants who had an objectively established lexical-retrieval impairment and those who had normal lexical retrieval, before and after surgery

Before surgery, no measure of disfluency discriminated those with an impaired lexical-retrieval ability from those with a normal lexical-retrieval ability; see Table 5 for an overview. After surgery, the participants with an impaired lexical retrieval (n = 12) had a higher proportion of false starts (normal: Mdn = 0.3%; impaired: Mdn = 2.7%; U = 35, p = 0.005) than those with normal lexical retrieval (n = 14). This was the only difference seen at the Bonferroni-corrected level, but at the uncorrected level differences were detected in three additional disfluency measures: the participants with impaired lexical retrieval had a slower speech rate, a slower articulation rate and a longer silent-pause length than those with normal lexical retrieval.

Table 5. Comparison of disfluency measures as between participants who had lexical-retrieval difficulties and those who did not, before and after surgery.

	Before surgery					Follow-up 3 months after surgery
	Normal lexical retrieval		Impaired lexical retrieval			Normal lexical retrieval		Impaired lexical retrieval
	n = 20		n = 6			n = 14		n = 12
	M	Mdn	M	Mdn		M	Mdn	M	Mdn
	(SD)	min – max	(SD)	min – max	Sig.	(SD)	min – max	(SD)	min – max	Sig.
Speech rate (words/min)	190	187	177	181	U = 47	194	183	154	160	U = 35
	(28)	(137–260)	(37)	(125–222)	p = .457	(38)	(139–277)	(39)	(97–228)	p = .011*
Articulation rate (words/min)	242	235	228	228	U = 43	252	249	218	210	U = 38
	(31)	(187–313)	(22)	(197–259)	p = .324	(30)	(196–315)	(32)	(169–284	p = .017*
Articulation	78.6	78.7	76.8	79.5	U = 55	76.9	77.2	69.9	70.6	U = 56
ratio (%)	(7.0)	(65–89)	(9.6)	(64–86)	p = .790	(8.3)	(62–90)	(10.0)	(52–81)	p = .160
Silent-pause	12.9	13.5	13.5	11.6	U = 57	14.1	13.6	16.8	15.4	U = 59
ratio (%)	(4.3)	(4.4–22.3)	(6.3)	(7.6–20.9)	p = .882	(6.5)	(4.2–27.1)	(4.4)	(11.1–23.2)	p = .212
Silent-pause	0.53	0.52	0.64	0.63	U = 29	0.55	0.56	0.71	0.67	U = 31
length (sec)	(0.14)	(0.34–0.79)	(0.08)	(0.56–0.74)	p = .062	(0.15)	(0.24–0.92)	(0.16)	(0.48–0.97)	p = .005*
Silent pauses/words (n)	0.10	0.10	0.10	0.07	U = 53	0.10	0.10	0.14	0.13	U = 52
	(0.03)	(0.03–0.17)	(0.05)	(0.06–0.18)	p = .700	(0.04)	(0.03–0.15)	(0.06)	(0.05–0.27)	p = .106
Filler ratio (%)	4.9	3.7	4.9	2.2	U = 48	5.5	5.0	6.5	5.5	U = 82
	(3.5)	(1.0–14.2)	(5.7)	(0.3–12.6)	p = .494	(3.6)	(0.2–15.1)	(6.1)	(0.4–22.2)	p = .940
Filler length (sec)	0.43	0.42	0.40	0.35	U = 45	0.41	0.38	0.37	0.33	U = 70
	(0.10)	(0.20–0.62)	(0.15)	(0.26–0.66)	p = .387	(0.13)	(0.27–0.71)	(0.12)	(0.23–0.58)	p = .494
Fillers/words (n)	0.04	0.03	0.05	0.04	U = 52	0.05	0.04	0.10	0.06	U = 62
	(0.03)	(0.01–0.14)	(0.05)	(0.003–0.13)	p = .656	(0.02)	(0.02–0.09)	(0.08)	(0.007–0.28)	p = .274
False starts	1.3	1.2	1.9	2.0	U = 35.5	0.5	0.3	2.9	2.7	U = 27.5
(%)	(1.0)	(0–4.9)	(1.1)	(0.7–3.7)	p = .139	(0.7)	(0–2.5)	(2.7)	(0–8.0)	p = .003**
Self-interrupted sentences (%)	9.6	10.1	12.5	10.1	U = 50	7.0	7.6	13.3	9.3	U = 67.5
	(6.8)	(0–24)	(9.6)	(3.6–29.6)	p = .573	(5.2)	(0–17)	(13.7)	(0–40.0)	p = .403
Paraphasias and neologisms (%)	0.2	0.1	0.2	0	U = 57	0.2	0	0.7	0.4	U = 56
	(0.2)	(0–0.7)	(0.3)	(0–0.7)	p = .882	(0.5)	(0–1.4)	(0.9)	(0–3.3)	p = .160
Repetitions	0.5	0.3	0.3	0.2	U = 46.5	0.5	0.3	0.6	0.4	U = 80.5
(%)	(0.5)	(0–2.0)	(0.3)	(0–1.0)	p = .421	(0.6)	(0–2.0)	(0.6)	(0–2.0)	p = .860

Note: *significant at p-level .05. **significant at the Bonferroni-corrected level of p < 0.004.

Comparison of disfluency measures between participants with a subjectively experienced change in language, speech or communication and those who had not experienced such a change, before and after surgery

At the Bonferroni-corrected level, no measure of disfluency differed as between those with and without a subjectively experienced change in language, speech or communication, either before or after surgery; see Table 6 for an overview. At the uncorrected level, differences were seen in the speech rate and articulation rate after surgery: the participants who had experienced a change in communication and language (n = 12) had a slower speech rate and a slower articulation rate than those who had not experienced such a change (n = 14).

Table 6. Comparison of disfluency measures as between participants who had experienced a change in speech, language or communication and those who had not experienced such a change, before and after surgery.

	Before surgery					Follow-up 3 months after surgery
	No change		Change			No change		Change
	n = 17		n = 9			n = 14		n = 12
	M	Mdn	M	Mdn		M	Mdn	M	Mdn
	(SD)	min – max	(SD)	min – max	Sig.	(SD)	min – max	(SD)	min – max	Sig.
Speech rate	190	188	182	183	U = 67	193	183	155	159	U = 33
(words/min)	(32)	(137–260)	(29)	(125–216)	p = .634	(36)	(139–277)	(43)	(97–251)	p = .008*
Articulation rate	245	235	228	228	U = 52	251	249	219	214	U = 37
(words/min)	(31)	(187–313)	(23)	(197–266)	p = .200	(31)	(196–315)	(32)	(169–282	p = .015*
Articulation ratio (%)	77.3	78.5	79.7	79.3	U = 64	76.7	77.8	70.1	70.0	U = 52
	(7.7)	(65–89)	(7.3)	(64–87)	p = .525	(7.6)	(62–90)	(10.7)	(52–89)	p = .106
Silent-pause ratio (%)	13.1	12.8	12.8	13.7	U = 75	14.1	13.0	16.9	17.1	U = 54
	(5.0)	(4.4–22.2)	(4.5)	(6.2–14.7)	p = .958	(6.0)	(4.2–27.1)	(5.4)	(5.4–23.2)	p = .131
Silent-pause length (sec)	0.53	0.54	0.61	0.64	U = 50	0.58	0.58	0.68	0.67	U = 47
	(0.11)	(0.34–0.69)	(0.15)	(0.35–0.79)	p = .164	(0.11)	(0.44–0.92)	(0.21)	(0.24–0.96)	p = .060
Silent pauses/words (n)	0.10	0.10	0.09	0.08	U = 63	0.10	0.10	0.14	0.13	U = 46
	(0.04)	(0.04–0.18)	(0.03)	(0.07–0.15)	p = .491	(0.04)	(0.03–0.15)	(0.06)	(0.05–0.27)	p = .053
Filler ratio (%)	5.5	4.2	3.9	3.0	U = 54	5.2	4.1	6.9	6.3	U = 73
	(3.9)	(1.0–14.2)	(4.0)	(0.2–12.6)	p = .241	(3.8)	(0.9–15.1)	(5.9)	(0.4–22.2)	p = .596
Filler length (sec)	0.44	0.42	0.38	0.40	U = 51	0.42	0.38	0.36	0.33	U = 59
	(0.12)	(0.20–0.66)	(0.08)	(0.24–0.49)	p = .181	(0.13)	(0.27–0.71)	(0.10)	(0.23–0.56)	p = 0.212
Fillers/words (n)	0.05	0.04	0.04	0.33	U = 59	0.05	0.05	0.10	0.06	U = 58
	(0.03)	(0.01–0.14)	(0.04)	(0.003–0.13)	p = .367	(0.03)	(0.01–0.09)	(0.08)	(0.008–0.28)	p = .193
False starts (%)	1.3	0.8	1.7	1.3	U = 67.5	1.0	0.5	2.3	1.4	U = 53.5
	(1.0)	(0–3.4)	(1.6)	(0–4.9)	p = .634	(1.6)	(0–6.2)	(2.6)	(0–8.0)	p = .118
Self-interrupted sentences (%)	12.0	12.0	6.9	5.3	U = 43.5	8.6	7.6	11.5	8.3	U = 77.5
	(8.0)	(0–3.4)	(4.9)	(0–15.8)	p = .075	(8.4)	(0–32)	(12.5)	(0–40.0)	p = .742
Paraphasias and neologisms (%)	0.3	0.3	0.08	0	U = 44.5	0.3	0	0.6	0.2	U = 68
	(0.3)	(0–0.7)	(0.1)	(0–0.4)	p = .085	(0.5)	(0–1.4)	(0.9)	(0–3.3)	p = .432
Repetitions (%)	0.3	0.3	0.5	0.3	U = 63	0.5	0.3	0.6	0.4	U = 71
	(0.4)	(0–1.0)	(0.6)	(0–2.0)	p = .491	(0.6)	(0–2.0)	(0.7)	(0–2.0)	p = .527
Lexical-retrieval measures:
BNT	52.6	54	48.8	50	U = 39	52.7	53.5	45.3	51.5	U = 54.5
	(4.4)	(42–58)	(5.5)	(37–54)	p = .045*	(4.5)	(45–58)	(13.5)	(14–57)	p = .131
FAS	45.8	48	31.3	29	U = 32	47.4	47	21.2	18	U = 12.5
	(13.6)	(20–71)	(10.3)	(14–45)	p = .016*	(11.9)	(26–64)	(12.1)	(6–51)	p < .001**
Animals	22.7	22	18.0	18	U = 39	20.9	20	9.5	8.5	U = 18.5
	(5.6)	(13–32)	(5.3)	(11–27)	p = .045*	(6.6)	(10–32)	(7.1)	(0–25)	p < .001**
Verbs	20.0	20	12.7	12	U = 33.5	22.9	24.5	14.0	13.5	U = 26
	(5.6)	(6–30)	(5.5)	(9–24)	p = .018*	(6.3)	(11–31)	(6.4)	(4–24)	p = .002**

Note: *significant at p-level .05. **significant at the Bonferroni-corrected level of p < 0.003.

Scores on tests of lexical retrieval were also included in this comparison, to see whether performance on those tests differed between the groups. It was found that, after surgery, the group having experienced a change in language, speech or communication performed worse on all word-fluency tests than the group not having experienced any such change. Differences in scores on tests of lexical retrieval were also seen before surgery, but they did not survive Bonferroni correction.

Discussion

The primary aim of the present study concerned how different aspects of disfluencies in spontaneous speech are affected by tumour surgery in participants with presumed LGG. While several measures suggested that the participants’ speech was more disfluent after surgery, only a single difference – an increased frequency for fillers after surgery – survived significance-level correction. We also examined the performance on tests of lexical retrieval, where participants performed worse on two of the word fluency tests after surgery, but not on BNT, indicating that lexical access under time pressure was more affected than confrontational naming. The number of persons who were assessed as having an impaired lexical retrieval increased after surgery, and there was also an increase in the number of persons who experienced a change in their language, speech or communication.

We also explored whether those participants who had been objectively found to have an impaired lexical-retrieval ability or who had subjectively experienced a change in language, speech or communication differed in terms of disfluencies in spontaneous speech from those with normal lexical retrieval and those not having experienced such a change, respectively. Overall, the breakdown based on objectively identified deficits in lexical retrieval yielded more differences between the groups than the breakdown based on subjective experienced changes. Since the sample size was small, it is not particularly surprising that several differences did not attain the Bonferroni-corrected level of significance. The discussion in the following sections will deal both with those significant results that passed the Bonferroni test and those that did not.

Reasons for the increase in disfluencies in spontaneous speech seen after surgery

An increased occurrence of pauses and fillers as well as longer silent pauses, as seen in the present study, can indicate that more time or effort is needed to plan and produce spoken language after surgery than before. This could be related to impaired lexical retrieval, which is a language deficit commonly seen in persons with glioma both before and after surgery (Papagno et al., 2012; Racine et al., 2015; Santini et al., 2012; Satoer et al., 2012; Tucha et al., 2000). Our findings differ from those of Satoer et al. (2013, 2018) in that those authors did not see a decline in any of the measures they analysed as between a pre-operative assessment and a follow-up three months post-operatively. It could be that the measures used in the present study, focusing on temporal aspects of spoken language, are more sensitive to tumour-related changes, but the differences in findings could also be related to differences in participant characteristics. While the durational changes in silent pauses are small, it is worth noting that even a 0.1 s change in pause length may influence the perceived communicative function of a pause in conversation (Roberts & Francis, 2013). Temporal aspects of language have proved sensitive to change in both language production and language processing in persons with glioma (Mooijman et al., 2021; Moritz-Gasser et al., 2012). Moritz-Gasser et al. (2012) found that lexical-access speed (i.e. response times on a naming task) was associated with the likelihood that a participant would be able to return to work after surgery, even after other cognitive factors were controlled for. For this reason, and because fluency is an important aspect of efficient spoken language, those authors emphasised the need to include temporal measures of language production in assessment. Corresponding findings have also been made with regard to written communication. In a study investigating the writing process in participants with glioma (partly the same individuals as in the present study), there was an increase in pauses before words after surgery (Antonsson et al., 2018c). In the same study, it was also noted that performance on tests of oral lexical retrieval (BNT and word fluency) had a strong positive relationship with writing fluency both before and after surgery. Taken together, these findings support a lexical deficit. However, word-finding difficulties may not solely be the result of impaired lexical retrieval. Mooijman et al. (2021) suggest a more general cognitive decline, based on their observation that word-finding difficulties were related not only to the speed of language processing but also to processing speed in a non-verbal task.

Although not the focus in this study, disfluencies in speech can also occur due to causes such as stuttering (developmental or neurogenic) or other fluency disorders. We did not screen for this but asked all participants of their history of language, speech, and communication disorders and if they had observed any changes in these areas. To our knowledge none of the participants had neither, but future studies should examine this more.

The role of objectively established lexical-retrieval impairment and of subjective complaints

To explore how the presence or absence of objectively established lexical-retrieval impairment and of subjective experiences of language change relate to the presence of disfluencies in spontaneous speech, we divided the participants into groups and compared their speech before and after surgery. As illustrated in Figure 1, there is an overlap between having impaired lexical retrieval and experiencing subjective language deficits, and vice versa, but in some cases there was a discrepancy between those two categorisations. It is in fact well known that subjective and objective measures of language impairment can differ, and several other studies have reported similar discrepancies between subjective complaints and test results (e.g. Antonsson, 2017; Brownsett et al., 2019; Gehring et al., 2015; Mooijman et al., 2021; Racine et al., 2015; Satoer et al., 2012).

Surprisingly, we did not see any significant differences before surgery when comparing disfluencies across the two categories with regard to objectively established lexical-retrieval impairment. A possible reason for this might be that the impairments seen before surgery are predominantly mild (indeed, few participants reported any language complaints) (Antonsson et al., 2018a). After surgery, by contrast, several aspects – including speech rate, silent-pause length and the proportion of false starts – differentiated the participants with impaired lexical retrieval from those with normal lexical retrieval. This might be because more participants had impaired lexical retrieval after surgery, and perhaps also because the retrieval problems were generally more pronounced after surgery. These results support previous findings that a lexical deficit affects language production in persons with glioma, and they also shed some light on the functional consequences that impaired lexical retrieval exerts on spoken-language production.

As regards the breakdown by whether participants had experienced any change in their language, speech or communication, the only measures found to differ were the speech rate and the articulation rate, and only after surgery; both of those rates were lower on average for the participants who had experienced a change. When asking the participants whether they had experienced any change in their language, speech or communication, we did not ask to what degree this affected their quality of life and everyday participation. This is of course important when interpreting to what extent their problems affect their life. However, in a follow-up study, recently published by Åke et al. (2022) we interviewed 12 of the same participants. They expressed changed communication that affected word finding, motor speech and comprehension, and that communication required a greater effort. Even though most participants did not feel that it affected their everyday life, they expressed that it was not like before, and many expressed feeling frustrated with their communication.

Although, as noted, there is some overlap between the two categorisations, it might seem reasonable that the difference between participants broken down on the basis of an objective measure (in this case BNT or word fluency) should be more pronounced than that between participants broken down on the basis of a subjective experience. It would obviously have been preferable if we had been able to specify the ways in which participants had experienced change. It should be noted that, as we have described in previous publications relating to the same individuals (Antonsson, 2017; Antonsson et al. 2018a), the most commonly reported type of change in language, speech or communication was word-finding difficulties. However, since some participants were fairly vague in their answers or had experienced other kinds of changes in their language, speech or communication, we chose not to make a further breakdown but instead to look at the whole group who had reported some kind of change.

The discrepancy not infrequently found between subjective and objective measures is often highlighted as a problem in the assessment of language-related and cognitive complaints in persons with glioma. Attempts have been made to develop specific test instruments or batteries of tests for this patient group to better capture their language impairments (Papagno et al., 2012; Satoer et al., 2019), but there is still no consensus on what language tests should be used (Rofes et al., 2017). It could be that this discrepancy between subjective complaints and test results is unavoidable, and perhaps we should rely more on the subjective experience rather than trying to adapt or develop tests that can capture subjective complaints. When it comes to cognitive impairments seen in the ageing population, mild cognitive complaints not accompanied by any objectively observable cognitive decline are referred to as ‘subjective cognitive impairment’ (SCI) (Mendonça et al., 2016). It might be appropriate to adopt similar terminology for neurological diseases where subjective language complaints are not reflected in objective test results.

Methodological considerations and future directions

Several of the methodological choices made in the present study may have affected its results. Some of them we have already mentioned, for example that we merged all types of experienced change in language, speech or communication into a single category. Another is the choice of task for the elicitation of spontaneous speech. The task we used has both drawbacks and benefits. A more structured task, such as a picture elicitation, has the benefit of prompting similar responses, with a comparable structure, from all participants. However, a freer task such as ours might have greater ecological validity in that it permits participants to produce a more personal narrative and probably places higher demands on their ability to plan as well as execute spoken language, given that neither overall content nor specific words are prompted. Some studies of spontaneous speech (e.g. Rofes et al., 2018) include both a free task and a more structured task, such as a picture-elicitation task or a task involving the retelling of a story. Since the spontaneous-speech task used in the present study was administered as part of an extensive language-test battery including a comprehensive aphasia test, we did not deem it feasible to include additional tasks. However, we recommend that future research, if possible, should use different kinds of tasks to enable comparison between them. It would also be preferable that one of the elicitation tasks included in future studies to be similar before and after surgery to control for linguistic variables such as word frequency. One might in future studies also explore other types of linguistic analyses suitable for discourse data such as measures of coherence and cohesion, which has previously been explored in persons with post-stroke aphasia (Whitworth et al., 2015) and in persons with early cognitive impairments (Kim et al., 2019; Mazzon et al., 2019).

Another methodological reflection is that most analyses of disfluencies are quite cumbersome, and would not be suited for a clinical setting (silent pauses is the type of disfluency that is most suitable for automatic analysis at present, since it can be done more or less automatically, depending on the level of precision desirable). For English the development of automatic transcription has come further, but in small languages such as Swedish, there is a lack of tools that are suited for a clinical use. With technical advances, hopefully the development of automatic transcription tools and tools for automatic analyses of speech, language and disfluencies, are not too far into the future.

Our study included participants with right hemisphere impairment, who may have poor insight in their communicative or cognitive impairment. Since we only asked the participants themselves if they experienced any change in their language, speech or communication, we can not rule out that others such as family, friends or co-workers have noticed differences in their communication. Hopefully, we can learn more about this in future studies, since we are currently collecting data on a questionnaire on communicative impairment with two forms, one directed to the patient and one directed to a significant other.

A further limitation of the present study is that it does not include a control group, which would have allowed us to investigate whether the participants with gliomas differed with regard to disfluencies from persons without any neurological conditions. Previous studies in this field suggest that participants with glioma differ from controls on several aspects of spontaneous speech (Rofes et al., 2018; Satoer et al., 2013, 2018). However, the principal aims of the present study related to the identification of differences within the group of persons with gliomas.

It should also be acknowledged that the present study does not control for the influence of differences between the groups compared before and after surgery other than the presence or absence of lexical-retrieval impairment and subjective experiences of changes in language, speech and communication. In fact, a wide range of factors may influence how fluently a person speaks. Those factors may be both language-related and speech-related. In fact, in the comparisons based on objectively identified deficits in lexical retrieval and on subjectively experienced change, differences after surgery were seen not only in the speech rate, but also in the articulation rate. This suggests that the motor execution of language – that is, speech – is an important factor to be included in assessments of disfluencies in spontaneous speech. Another possible explanation already discussed is the presence of a more general cognitive decline post-surgery. Hence, it would have added to the value of the present study to include cognitive tests assessing processing speed and working memory, as well as assessments of psycho-emotional well-being and presence of fatigue.

Conclusion

The present study shows that there was an increase in disfluencies in the spontaneous speech of persons with presumed low-grade glioma after surgery. That increase might be related to impaired lexical retrieval. Further, more differences were found when the participants were broken down based on objectively identified deficits in lexical retrieval than when they were broken down based on subjectively experienced change. Taken together, the findings from this study further strengthen the previous claim that temporal aspects of language and speech production are important in the assessment of persons with gliomas as well as in the assessment of persons with subtle language difficulties in general.

Disclosure statement

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

Additional information

Funding

This work was supported by grants from the Swedish State under the ALF agreement between the central government and the county councils under Grant [No. 76510].