Eloquent silences: A musical and lexical analysis of conversation between oncologists and their patients

Abstract

Objective:

Silences in doctor-patient communication can be “connectional” and communicative, in contrast to silences that indicate awkwardness or distraction. Musical and lexical analyses can identify and characterize connectional silences in consultations between oncologists and patients.

Methods:

Two medical students and a professor of voice screened all 1211 silences over 2 s in length from 124 oncology office visits. We developed a “strength of connection” taxonomy and examined ten connectional silences for lexical and musical features including pitch, volume, and speaker turn-taking rhythm.

Results:

We identified connectional silences with good reliability. Typical dialog rhythms surrounding connectional silences are characterized by relatively equal turn lengths and frequent short vocalizations. We found no pattern of volume and pitch variability around these silences. Connectional silences occurred in a wide variety of lexical contexts.

Conclusion:

Particular patterns of dialog rhythm mark connectional silences. Exploring structures of connectional silence extends our understanding of the audio-linguistic conditions that mark patient-clinician connection.

Practice implications:

Communicating with an awareness of pitch, rhythm, and silence – in addition to lexical content – can facilitate shared understanding and emotional connection.

1. Introduction

Silence is a communicative act, yet clinicians are rarely trained or sensitized to its importance and potential meanings. Silence in clinical settings has been studied since the 1970s in order to classify and understand its functions and etiologies [1–4]. Silence in clinical contexts is defined as an absence of verbal audio signal, lasting appreciably longer than the average time between speaking turns [5–7]. The meaning and function of silence is defined by context including ambient sounds, utterances before and after the silence, and visual cues such as facial expression and body position [3,5,7–10]. The literature from psychotherapy and linguistics suggests that silence within medical visits can be divided broadly into communicatory silences that are intentionally left silent and non-communicatory or interrupted communication, such as when a physician enters data on a computer [7,11].

Previous research on silence in patient-physician communication has identified a particular type of silence – which we call “connectional silence” – associated with emotional exchange [8,12]. Descriptions of these moments, presumed to build relationships and be therapeutic for patients, often contain elements of resonance, presence, and attentiveness [13]. These connectional silences may contribute to improved communication by facilitating shared understanding and emotions [14].

Silence in general – and connectional silences in particular – have been regarded either as intangible, part of the art of medicine, or as part of the lexical structure [15]. Yet, detailed attention to both the lexicon and musicality of language, including rhythm, pitch, volume, and tempo, may shed light on this phenomenon [16]. In this exploratory study, we build on previous work to examine the lexical and musical landscape around these silences in detail, and to identify common elements among connectional silences [17]. Given that this was an exploratory study, we did not know from the outset whether those elements would pertain to the words that preceded and followed the silence (the lexical elements) and/or the musical elements such as pitch and rhythm of the speakers’ voices. We sought to illuminate the co-occurrence of silence and connection to determine if there might be performative elements that could be taught to clinicians, thus enriching the patient-physician relationship [8,17,18].

2. Methods

We created a taxonomy of silence, and then analyzed both the lexical and musical contents for those silences that we identified as connectional. We used audio data from the observational stage of a multi-site randomized trial of an intervention to improve communication in oncology settings [19]. Audio-recorded encounters occurred between November 2011 and August 2012 in outpatient oncology clinics near Rochester, NY, and Sacramento, CA. Out of 53 oncologists contacted, 41 solid tumor oncologists in private and university practice settings each completed three audio-recordings of outpatient office visits with patients who had stage III or IV cancer (and an accompanying caregiver if available); these recordings were used in this secondary analysis (Table 1). Oncologists and study personnel at the time were unaware that conversational silences would be examined in a secondary analysis. Detailed recruitment, inclusion, exclusion, and human subject review board approval can be found in a previously published protocol [19].

Table 1.

(a) Patient demographics. (b) Physician demographics.

(a)
	N	%
All	124	100
Race
White	112	90
Other	12	10
Site
URMC	78	65
UCD	42	35
Patient Education
Some college or more	85	68
HS or less	39	33
Aggressive cancer
Non-aggressive	66	53
Aggressive	58	47
Patient Gender
Female	70	56
Male	54	44

(b)
	N	%
All	41	100
Physician gender
Male	29	71
Female	12	29
Physician Race
Asian	16	39
Black/AA	1	2
White	18	44
Other	1	2
Missing	5	12
Breast cancer physician
No	33	80
Yes	8	20
Physician age
Mean	Median	std
44.7	44	9.8

Initially, we set out to create a typology of moments where silence occurs in conversations in the context of advanced cancer. Analysis was conducted by a multidisciplinary team consisting of a medical student with a background in public health (JB), a medical student with a background in literature (RR), and a professor of voice (KC). All three analysts are trained musicians though RR is not currently practicing. Working from audio with parallel transcripts, we identified all silences greater than 2s in length.

Silence identification was completed by JB and RR by scanning the waveform in detail for areas of low intensity to ensure that silences as short as one second were conspicuous. Length was measured using an Atlas Ti waveform display by zooming to a scale where one second stretches almost two inches across the screen, highlighting the section of interest, and listening to see if there was any low volume vocal activity. The endpoints of the silence were then adjusted manually to the nearest 0.1 s. Sensitivity was achieved by listening to the entire recording in case there were times with high intensity waveform being created by nonverbal sources such as exam room tables or a ringing cell phone. Ten complete visits were coded by both JB and RR to check accuracy in silence identification and length measurement. A 6% disagreement on borderline-length silences (1.9–2.1 s) was resolved through consensus and we refined the criteria and technique for silence length measurement.

Using a modified grounded approach sensitized by the empirical literature on conversational silences and in consultation with a linguist, we developed descriptive codes [20,21]. Through an iterative process, we eventually defined 10 broad dimensions – 5 lexical and 5 musical – within which a priori and emergent codes were developed to describe distinct patterns of communication 30 s before and after each silence [22,23]. We used musical notation to capture overlapping verbal and non-verbal sounds and to compare speakers’ pitch, tempo, and volume even when they overlapped. Our 10 dimensions allowed for axial coding of lexical and musical characteristics around each silence (See Tables 2a and 2b for a list and definitions of these dimensions).

Table 2a.

Lexical dimensions and codes within each dimension.

Regional conversation context

“Bad news” discussion during which physician uses negative adjectives or patient expresses disappointment

Signposts of what the future might bring

Biomedical talk concerned with markers of disease progression/remission

Review of scans/labs

Symptoms discussion using medical words as in a review of systems

Psychosocial experience of the patient

Small talk about weather, travel, or other unrelated talk

Specific prompt

Computer/paperwork task

Patient/caregiver expresses preference

Patient/caregiver expresses concern/emotional cue

Doctor acknowledges/expresses concern/empathy

Patient or caregiver acknowledges/accepts/understands before silence

Question answered before silence

Question asked before silence

Small talk filler

Unknown or internal though process

Utterance before silence

Incomplete

Complete

Subvocal proceeds silence

Subvocal ends silence

Speaker turn order across silence

Caregiver-caregiver

Caregiver-doctor

Caregiver-patient

Doctor-caregiver

Doctor-doctor

Doctor-patient

Patient-caregiver

Patient-doctor

Patient-patient

Conversational outcome

Concern

Doctor acknowledges/expresses concern/empathy

Subject changed in question form

Subject changed in statement form

Patient/caregiver acknowledges/accepts/understands

Question answered

Question asked

Patient/caregiver expresses preference/value

Table 2b.

Musical dimensions and parameters measured within each dimension.

Dialog rhythm visualization (see Fig. 1)
Pitch variability within each speaker’s turn
- Notated on 12 tone staff
- Recorded as difference in half steps between lowest and highest note in each turn.
Volume variability within each speaker’s turn
- Notated with standard musical notation from pp to ff
Volume variability across the silence
- Difference between last word before silence and first word after silence
Relative stressed syllable speed
- Notated from quarter to thirty-second divisions, shows relative speed between speakers.

As coding progressed, the team noted one kind of silence that marked emotionally intense moments. We sought descriptions of similar phenomena in the medical literature and developed a “strength of connection” taxonomy with subcodes of “connectional,” “invitational,” “neutral,” and “disengaged” silences (Table 3) [12,24–26]. We defined a connectional silence as one in which there was an implicit or explicit patient emotional cue, doctor recognition of that emotion, and an indication of emotional resonance between doctor and patient. We defined connectional silences narrowly; for example, in one case, the patient recognized a doctor’s emotion, but not vice versa, and the segment was not coded as connectional because it lacked reciprocity. Invitational silences were those that opened up a conversation, whereas disengaged silences were activities that distanced patient and doctor, such as entering data on a computer. In order to validate the “strength of connection” taxonomy, 10 conversation segments with one silence each were reviewed by 3 expert physicians who were not study participants and not involved in the coding process. The strength of connection was identified reliably by this panel with an 87% overall agreement (Cohen’s kappa = 0.72) [27].

Table 3.

Strength of connection definitions.

Definition	Inclusion	Exclusion	Example
Connectional silence: a moment of shared emotion and meaning.	Requires implicit or explicit patient emotion, doctor recognition of that emotion, and an indication of emotional resonance between doctor and patient. The silence must occur within or immediately adjacent to this sequence of emotion-recognition-resonance.	No doctor recognition of patient’s emotion. Patient recognition of physician’s emotional cue. Silence occurs immediately after subject change.	P 503: I feel like I’m doing something, I feel like I’m contributing. D 412: Well, you know what, [Name]? Do it but don’t do it to your own detriment. P: Yeah. Okay. {Silence 3.8 s} I got a lot of that stuff going. D: It’s rare of me to tell somebody point blank you’ve got to stop. However, I will say you have my permission to set limits.
Invitational Silence: a moment where the doctor invites the patient to express an emotion or the patient invites the doctor to recognize an emotional cue.	Patient emotional expressed or doctor leaves the door open for patient to express some emotion.	Invitation results in expression or recognition of emotion.	D481: Have you talked to your family at all any more about the disease? P106: Uhm [silence 2.5 s] D: Not really P: Well, I talked to my boyfriend. I told him about it. And he’s understanding. I haven’t told my mom about the two-year kind of average thing yet. Um, but. . I joke with her. You know, how can I diet like this. She said don’t eat sugar. It hates cancer. D: Right, right. P: I go I’m gonna die anyways, mom, let me eat it. [laughs] D: No. I know it’s not easy to have those conversations. P: Yeah.
Neutral Silence: a pause in conversation that does not affect the emotion of the conversation. Often without apparent etiology, conversation sounds no different if the pause is removed from that moment of conversation.	Silence has no apparent effect on the conversation. Silence can be removed and conversation sounds the same.	Fits better into one of the other categories. Presence of emotional content.	P 504: So, what we need to do is then continuously or periodically do the scan to see if it’s there and if it’s starting to grow. And then attack it when it does. D400: Um hm. [Silence 4.1 s] D: Does that make sense to you? P: It does. It does. I mean, I wish there was some way to completely eradicate everything that’s there and get that one rebellious gorilla cell from being able to start nesting again. But, you know, again, hopefully with the scans that we do we’ll be able to find it quickly and address it and attack it quickly.
Disengaged Silence: A silence that seems to embody distraction, specifically when the doctor does not seem engaged in communication with the patient.	Occurs before the natural end of a conversational topic. Results in a loss of coherence. Divides two topics without any transition statements.	Physical exam activities in close proximity. Conversation sounds normal if recording played without silence (didn’t miss a beat).	D412: For the clot pain, Oxycodone is fine and sometimes ibuprofen, you know, Advil … P 503: Okay. D:.. Is actually better for clot pain. So ifyou really had somebody brushes it or it’s hurting and not going away try some ibuprofen. It’s good, too. P: Okay. [Silence 21.6 s, sounds of typing] D: How did the drive feel today? Was it okay? C: Pretty good. P: Not so bad. Not bad at all. D: How much snow where you are? P: We’ve got quite a bit.

The 124 recordings were divided among three coders (JB, KC, and RR) after they reached an inter-rater reliability (Cohen’s kappa) of greater than 0.60; kappa reached 0.64 after coding the first 67 silences [27–29]. All silences coded as “connectional” or “invitational” were reviewed a second time by at least two coders and disagreements were resolved by consensus. We coded all “connectional” silences, including the 30 s before and after each silence, along all ten dimensions (Table 2).

Due to challenges examining speech tempo (see Alexander et al. [17]), analysis of the micro timings of particular syllables was abandoned in favor of larger scale rhythmic analysis. We attempted to characterize relative between-speaker tempo through standard western musical subdivision but only relative within-speaker tempos could be visualized (Fig. 2). In a few small sections where we calculated syllables per second, utterance length appeared to be a non-linear inverse predictor of syllable speed, though the small sample size did not permit further analysis.

Fig. 2.

Musical notation of Example #500.

We found dialog rhythm [30], a measure of how frequently there is a change of speaker and how long each speaker speaks, to be the most informative measure of rhythm in these conversations. Using Atlas Ti, we notated and measured dialog rhythm by coloring the waveform of the conversation segments with colors representing individual speakers (Fig. 1). We then compared the dialog rhythm visualizations across all ten connectional silences to look for patterns [31]. The patterns of connectional silences were then examined for common features.

Fig. 1.

Visualization of speaker turns surrounding connectional silences to show dialog rhythm and utterance density.

Next, voice pitches were coded by KC using musical notation to approximate the semitones of speech to a 12-tone scale. KC has extensive experience transcribing speech – in a number of languages – into musical notation. Volume was coded with standard musical dynamic markings ranging from pianissimo (very quiet) to fortissimo (very loud), each assigned a numerical value to create a discreet ordinal variable for comparison of variance (pianissimo = 1 to fortissimo = 6). All three analysts reached consensus on phraseology by listening for borders delineated by breath divisions, large pitch changes, and rests. Once each silence was fully annotated, notated, and visualized, we measured pitch and volume variability by transforming half tones and volume measurement to ordinal integers, and tempo difference between speakers using a trichotomous distinction (slower, the same, faster)(Fig. 2).

3. Results

Characteristics of the 124 patients and 41 oncologists in the study are described in Table 1. There were 1211 silences that exceeded 2 s in duration; 440 were disengaged, 700 neutral, 61 invitational, and 10 connectional (Fig. 3). The 10 connectional silences occurred in 5 discrete conversations (see examples in Table 5). All connectional silences were located in the latter two-thirds of the conversation. Connectional silences were on average 3.6 s long whereas all other silences had a mean of 5.79 s (Table 4a). Within the 3 conversations that contained more than one example, the silences were often grouped together within one minute of each other.

Fig. 3.

Silence distribution by “Strength of Connection”.

Table 5.

Lexical content surrounding connectional silences.

Patient 500, Doctor 412	Patient 503, Doctor 412
{12:38} C: He says all that means is that the cancer tumor is not growing, we’re not curing, we’re not in remission. We understand all that. So it means it’s not growing, that’s good news for that part. D: It is good news. C: In a little dose. D: Yes. It is. It is good news. But ultimately the blood work can be deceptive so we only believe scans. Blood work, we always like for it to come down but sometimes - C: Usually it’s a little bit of a guide? D: Yes, exactly, but the key is the scan. C: That’s all right. We have a little ray of hope in all of it. D: Yeah, I know. C: It’s tough. D: What you feel is really hard. It’s really hard. {Connectional SILENCE 2.8 seconds} C: I just wish he had a better quality of life D: I know I know	{20:59} P: Okay. We’ve agreed a long time ago but, you know, if it gets to that point where it’s not working it’s… just keep me comfortable. D: It’s not worth it. P: Yeah. D: And sometimes when we do that patients actually have a great quality of life and can live a while. Because we stopped at a point where the body was doing okay. If we had done more chemo it may do more harm and actually get strength and get to do a lot of stuff. P: Right. D: And what I tell patients is that when the time comes, don’t feel like it’s a decision that you can never turn back on. P: Right. D: Some patients say well, let me try one or two rounds of the second chemo and if it’s not good, stop. Some patients say I want to go home, do X, Y and Z and then they’re here, let’s do that chemo, you know?
{Connectional SILENCE 2.6 second} D: Do you want to stop altogether? With the chemo? I feel- P: No, no. We’re going fine. D: Okay. You want to keep going, okay. C: No, we want to keep going. D: Cause certainly that’s always an option. And as long as I feel you can tolerate it, even at a lower dose or a changed schedule I will always do that. P: Right, right. D: But if ever you reach a point, you know, and if I feel you’ve reached the point I’ll tell you you know myself as well, that you wanted to stop with the chemotherapy and just have a quality of life and let this take its course, that’s reasonable. C: No. D: We’ll do our best to get you through. Okay. You are doing a good job. I know it’s hard. {14:06}	P: Right. D: But I think that is a courageous decision you are saying and I’m … I would support that, yeah. But I think that we need the brain, we need to, because, you know, that could get out of hand. So I think the steroids we’ll start and I’m glad we’ll go up on the Prilosec. P: Right. D: Because the steroids can make heartburn worse, too. P: Right. D: And then Dr. KK will do the radiation. P: Okay. D: {22:15} You do want the liver biopsy? You want to know? P: Yes. D: Okay. I think that’s fair. I want to know. But if you didn’t want to go through it… P: No, I want to know. D: Yeah. P: I want to know. And that way if there’s a chance that we- D: That it’s something completely different that we haven’t addressed. P: Yup. And that way, too, we’ll know if it is the same we know it’s not working. D: Yeah. P: Yeah. [Connectional Silence 6.3 seconds] D: So you get off from chemo. P: [Chuckles] D: No three days of fun for us P: Oh, boy. D: We’ll put in for Dr. KK to see you on Friday for the biopsy. And I’ll see you in two weeks, you know, wherever you are. You’ll be either towards the end of your radiation. P: Right. {23:05}
Patient 514, Doctor 408	Patient 520, Doctor 420
D: {33:15} I think the Taxol is the best chemo for you to treat your cancer. So, if he needs to do something for the eye well do that. We can have both. He’s not saying to stop the Taxol. P: Oh, no, no. He’s for it. I said, I don’t see you til Friday, I’m gonna have a treatment on Monday. He said that’s fine, no, I want you do that, he said. D: Good. And I am very pleased with the results. So, I hope the things we talked about, you know, I don’t know if you had more questions about what happens next or… ? P: No, for some reason I guess I just in my head was gonna be on Taxol for the rest of my life and everything was gonna be hunky dory and. …	C: Like if he does two gigs in a day he’s exhausted, more so than before. D: So what does that make you think? P: I like that feeling. I’d rather be exhausted. D: That’s your choice. P: Yeah. D: And it may bug you because you want to see him be not exhausted. But you have a choice to say only one. P: Yeah. Yeah. Well. D: I think you need to feel free to set limits {Connectional Silence 2.1 seconds @10:40} cause you’re probably not gonna hurt anybody else’s feeling by saying this is all I can do.
Silence 6.04 seconds @ 34:03 you know. I knew early on, I mean you told me early on it’s not like and then this will be the rest of my life. Something, you know D: Something. P: might go down, and so .. {Connectional Silence 2.9 seconds} D: That can be a very hard thing to think about. That here we found something that’s helping but you can’t stay on you can’t stay on it for the rest of your life. The reason for that is because neuropathy can get very serious, even to the point that you can’t feel or balance or things like that. P: Well, I have problems with balance, too unless I stand up like you had me stand there a minute and get my wits about me. D: Right. P: I can’t just jump up. I’ve learned. {34:59}	P Yeah. I know it. I gotta take- C: He hates doing it, though. P: I hate doing that. D: It sounds to me like you’re grieving your normal life and it’s not something you want to give up yet. P: A huh. D: Kind of normal. P: That makes sense. C: He did do, what, 17 gigs in 11 days? D: Oy. C: Over the St. Patrick’s Day. {Silence 2.72 seconds @11:09} P: I feel like I’m doing something, I feel like I’m contributing. D: Well, you know what, [Name]? Do it but don’t do it to your own detriment. P: Yeah. Okay. {Silence 3.78 @ 11:21} I got a lot of that stuff going. D: It’s rare of me to tell somebody point blank you’ve got to stop. However, I will say you have my permission to set limits. P: Okay. {Connectional Silence 6.4 seconds} C: [Chuckles] He can’t stand the thought of it. I can tell by his laugh . | [laughs] D: I know he can’t stand the thought of it. P: No, that’s okay. I… I’ll get used to it. D: I don’t think you’re a danger to yourself or others. P: Oh, that’s good. D: Just, you know, you might have to prioritize. P: Yeah. D: Okay. What else? {12:05}

Table 4a.

Silence length and locations.

	Mean length (sec)	Median	SD	Location mean	Location range
Connectional (n=10)	3.74	3.21	1.37	61%	35–94%
All other silences (n = 1201)	5.79	3.51	7.91	51%	1–99%

The dialog rhythm around connectional silences is notable for a high density of vocal contributions by multiple speakers, often short continuers rather than full speaker turns. The connectional silences were surrounded by an average of 17.2 speaker alternations per minute of conversation (Fig. 1). In addition to this density of participation, the pattern of voices around these silences exhibited many short utterances and frequent overlapping vocalizations. For example, both doctor 412 and doctor 408 mirror their patients’ utterance length and volume following each connectional silence (Fig. 1). The other musical dimensions did not reveal a consistent pattern across these ten silences. For example, there was greater pitch variability within doctors’ utterances around only 3 of the 10 examples. There was volume variability of at least 2 volume levels in at least one of the speakers before a majority of these connectional silences (Table 4b). Our tempo measurements revealed that the physician often spoke more slowly than the patient and further decreased their speaking rate in close proximity to the connectional silence (Fig. 2).

Table 4b.

Musical characteristics of connectional silences.

	Mean	Median	SD	Min	Max
Dialog rhythm density	17.2	17.5	5.02	8	25
Doctor pitch variability (semitones)	12.2	14	3.99	6	17
Patient pitch variability (semitones)	10.0	8.5	4.59	5	16
Doctor volume variability	2	2	0.47	1	3
Patient volume variability	1.7	1.5	1.49	0	5
Volume difference across connectional silence	2	2	3	0	4

Invitational, neutral, and disengaged silences during these conversations often delineated boundaries between thoughts, phrases, and tasks. Connectional silences, however, did not mark the end of a thought (see both examples P514 and P520 below); rather, the silence became part of a larger phrase, rather than marking an immediate end to the preceding phrase. Based on the participants’ speech before and after the silence, connectional silences did not interrupt the dialogue even when they occurred within the middle of a lexical phrase.

Connectional silences occurred in parts of conversations regarding biomedical treatment, future expectations, signposting about what to expect in the future, and discussion of the psychosocial experience. For example, patient 514 talks about her future as she realizes that her current health will not last. Here and elsewhere, the utterances before connectional silences were sometimes incomplete, trailing off in the middle of a thought.

Patient 514: No, for some reason I guess I just in my head was gonna be on Taxolfor the rest of my life and everything was gonna be hunky dory and …

{connectional silence 2.9 s}

Doctor 408: [inhales, as if to respond]

Patient 514: you know. I knew early on, I mean you told me early on it’s not like and then this will be the rest of my life. Something, you know, might go down.

Doctor 408: Something. That can be a very hard thing to think about. That here we found something that’s helping but you can’t stay on it for the rest of your life.

The utterance prior to a connectional silence was often a physician expression of concern or empathy, or a patient or caregiver acknowledging a harsh new reality. For example, in the discussion with patient 520, doctor 408 expresses concern that the patient continues to work so hard and offers to support the patient in setting limits.

Doctor 408: It’s rare of me to tell somebody point blank you’ve got to stop. However, I will say you have my permission to set limits.

Patient 520: Okay.

{Connectional Silence 6.44 s}

Patient’s wife: [Chuckles] He can’t stand the thought of it. I can tell by his laugh [laughs]

Doctor: I know he can’t stand the thought of it.

Patient: No, that’s okay. I… I’ll get used to it.

After connectional silences the lexical content was variable, such as a physician statement of concern, patient question, patient expression of concern, or continuation of the same phrase that came before the silence (see Table 5). Interestingly, the topic was not changed directly following a connectional silence. There was often a subtle reorienting of the lexical content, such as with patient 116 in which the physician doesn’t name the patient’s emotion (fear) but instead addresses the source of that fear (pain/physical suffering). We could not identify any consistent pattern of linguistic turn-taking after the silence; silences might be terminated by any party.

4. Discussion and conclusions

4.1. Discussion

In this study of connectional silences, we observed that patterns of vocalization around each occurrence created a unique rhythm characterized by bi- or tri-lateral participation with rapid alternation between speakers. This high frequency turn-taking reflects a high degree of engagement among speaking parties suggesting that connectional silence itself is a manifestation of interpersonal coordination. Alternatively, the rhythm surrounding these potentially empathetic moments might act to reset and break up the intense emotion and get back toward longer phrases and territory that is more emotionally neutral.

Silence in these examples requires that two or three people, all talking and contributing to a conversation, suddenly ‘agree’ to be silent simultaneously – a musical agreement that we interpreted as a manifestation of interpersonal synchrony. We assume that communication is intersubjective and iterative – meaning that it is a result of mutual influence and there are elements in which conversation partners complete each others’ thoughts, speak each others’ emotions and facilitate each others’ cognitive processing [32,33]. Our findings suggest that among other things, this coordination may be related to cognitive mirroring and entrainment, and may reflect interpersonal engagement and presence. Briefly, these functions refer to our mind’s ability to model another person’s actions with or without actually performing them physically [6,34–38]. During a connectional silence, speakers may be entrained physically, and the mirroring of each other’s vocal activity before, during, and after each silence may be an indication of a multi-level synchronization that facilitates this emotional exchange [39,40].

The literature on mirroring and communication in other contexts may be relevant here. For example, speakers mirror each other’s dialog rhythms in settings of successful negotiation and courting; more balanced speaking turns and more frequent interjections indicate greater engagement [30,35–38,40–42]. The same may be true of clinicians, in that this particular type of silence as defined by its dialog rhythm may mark mirroring, entrainment and interpersonal synchrony.

In this study, we have suggested a new way to examine silences empirically and potentially to identify those that co-occur with emotionally salient moments in cancer care. To describe the occurrence, placement, lexical, and musical aspects of silence, we have employed visualizations of the dialog rhythm surrounding connectional moments. Our examples suggest that there is not always increased variability of tempo and volume associated with emotional dialog as others had reported previously [17,43]. Our use of audio data and its transformation into musical notation and rhythm visualization enriched our analyses and provided insights that may eventually supercede the lexical content contained on medical transcripts.

There are several limitations to our study. In this exploratory study, we sought only to characterize connectional silences, realizing that there may have been other connectional moments in which there was no silence. Our data permitted us access to audio channels of communication but we could not capture other nonverbal behavior such as facial expression, body position, and gaze direction. Each of these provides additional context that is not captured through audio data alone. We included only oncologists and their patients with advanced cancer, who spoke fluent English, and who lived in two discrete regions of the USA. While cultural norms might affect patterns of dialog rhythm and silence, we did not have detailed ethnolinguistic data available [44]. Pitch and volume measurement relied on human raters due to background noise that inhibited digital quantification in these real life recordings.

4.2. Conclusion

As patients, caregivers, and clinicians communicate, they adapt in real time to rhythmic patterns of vocalizing and silence of the other party in addition to lexical content. We chose silences co-occurring with poignant moments of connection to examine one way in which communication facilitates patient-centered care–through the use of silence. Silences are bounded by both lexical and musical elements; here we have elucidated rhythmic elements of speech which may be crucial to the communication of emotion and the formation of patient-clinician relationships [17,45–48]. In particular, dialog rhythm mirroring was a marker of this particular kind of silence [3]. Given that connectional silences are relatively infrequent, new developments in computer-automated analysis of dialog rhythm patterns may facilitate investigations using larger data sets and thus reliably identify and characterize connectional moments. Perhaps measures like speaker alternations per minute and dialog rhythm around silences could shed light onto how physician engagement relates to silence in conversation.

The changes in dialog rhythm surrounding connectional silences have implications for those seeking to measure and practice shared presence, empathy, and engagement [13,49]. Characterizing clinicians’ speech patterns provides an additional key to understanding how clinicians might enhance the patient experience and communication outcomes [34]. We have demonstrated that connectional silences are empirically identifiable through their paralinguistic features, especially dialog rhythm. Future research should explore whether specific types of silence have a corresponding audio “fingerprint” that can reliably distinguish connectional from other types of silences.

If those present must collaborate to have silence, dialog rhythm mirroring is likely both a marker of and a vehicle for empathy and shared connection [50]. Music performers increase emotional resonance through the use of silence, perhaps by creating a connection as both performer and listener can mirror each other, quiet and still [16,38]. Similarly, this one type of silence in the exam room, in the context of engaged dialog rhythm, might be a moment that has the potential to enhance recognition of emotion and provide space that allows a clinician to address suffering [13].

4.3. Practice implications

Both lexical and musical elements of communication are complex, and a successful symbiosis such as occurs during empathy needs both carefully chosen words and a carefully chosen way to deliver them in terms of pitch, rhythm, and silence [48,51]. To achieve overall coherence and to cultivate the presence of mind necessary to perform this musical “work” requires self-awareness and deliberate practice. Communication training in medicine has largely focused on teaching physicians what to say. Our research suggests that this training would benefit from an additional focus on how to say it and when to say nothing in order to be in-tune with our patients. A coaching model used by musicians may prove to be an effective model for clinicians [52–54]. Understanding silences that mark patient-centered communication could be valuable in both teaching and evaluation of effective and complete communication.