The risk of metacommunication to manage interactional trouble in online chat counseling

The main handbook on online counseling used in the Netherlands recommends counselors to use metacommunication, i. e. to state what happens during the interaction, in case of interactional trouble. The aim of this study was to explore this interactional strategy by describing how counselors may use metacommunication to respond to interactional problems in online counseling of the “helpline” type (single chat sessions). Data were derived from a corpus of 197 sessions of the Dutch information chat service on alcohol and drugs and 348 sessions from a Dutch general chat help service. The theoretical and analytical framework of (digital) conversation analysis (CA) was used to analyze the data. The analyses show that metacommunication is used, but only rarely. Clients tend to log out from problematic sessions, often even before the counselor is able to identify interactional trouble at all. This means online counselors operate under the constant risk of clients’ disengagement. Second, we identified and described three metacommunicative practices counselors use to respond to problematic interaction: 1) Self-criticizing; the counselor is regretting (s)he was not able to help due to a lack of useful advice or to the boundaries of the institutional context. This strategy seems to elicit client acknowledgement followed by an aligned closing; 2) Accusing: the counselor is questioning the client’s advice recipiency. This strategy seems to elicit even more client resistance. 3) Explaining: the counselor explains the institutional tasks and responsibilities and/or suggests the client has unrealistic expectations. We conclude that although metacommunication can be useful to restore the relationship with the client when it is apologetic, it can also be counterproductive to deal with problematic interaction when the counselor questions the client’s willingness or ability to accept the advice. For this reason, our analysis warrants the advice to avoid metacommunication as a strategy in single session counseling.


Introduction
Online counseling is considered a valuable alternative to face-to-face treatment, and seems to be as effective as traditional forms of counseling (Fukkink/Hermanns 2009;King et al. 2006aKing et al. , 2006b;;Rochlen et al. 2004).However, a need remains for stronger empirical evidence to es-tablish efficacy and effectiveness and to better understand the unique mediating and facilitative variables of various types of online counseling processes (Richards/Viganó 2013).In the Netherlands, a growing number of services provide information and advice through text based chat sessions, accessed through websites.This form of counseling is characterized by it being online, anonymous, time-constrained, and usually without follow-up.The interactional context of chat counseling is likely to present clients and counselors with interactional challenges specific to this mediated counseling situation.For this project, we collaborated with the Dutch information chat service on alcohol and drugs (hosted by Trimbos) and a Dutch general chat help service (Korrelatie).These services have the purpose of providing the general public with accurate information and advice.The question we analyze in this article is how counselors manage interactional trouble in their chat session by using metacommunication as a strategy.We show that the use of metacommunication in actual chat counseling sessions creates a number of risks.For example, when the counselor questions the client's willingness or ability to accept the advice, this elicits more resistance rather than advice acknowledgement.This type of metacommunication is severely face-threatening and further distorts the counseling relationship.
In this paper we describe three metacommunicative practices counselors use to respond to problematic interaction: 1) Self-criticizing; the counselor is regretting (s)he was not able to help due to a lack of useful advice or to the boundaries of the institutional context; 2) Accusing: the counselor is questioning the client's advice recipiency; 3) Explaining: the counselor explains the institutional tasks and responsibilities and/or suggests the client has unrealistic expectations.Before we report about the analysis of these metacommunicative practices in more detail, we first give an overview of the literature regarding online counseling, providing and receiving advice, and metacommunication related to the stocks of interactional knowledge, and describe the data and methods that were used.

2
Literature review

Online counseling
Compared to more traditional counseling environments, i. e. face-to-face or on the telephone, professionals working at chat services are confronted with challenges that are inherent to the online environment in which the interaction takes place.Various studies of online counseling interaction have examined the intricate relation between medium and counseling in recent years (for an overview, see Lamerichs/Stommel forthcoming).In this article, we present an analysis of professional chat interaction in counseling services that have the purpose of providing the general public with accurate information and advice.For reasons of space, we have singled out one study for detailed introduction as it studies professional chat interaction in the same counseling service, and because it was the indirect reason for the analysis we present in this article.The study compared the text-based chat and the telephone counseling of the Dutch information service for questions about alcohol and drugs.One of the findings was that closing a counseling session is more difficult in chat interaction than over the telephone (Stommel/Te Molder 2015).Both in chat and on the phone it is the client who initiates the interaction by seeking information or advice from the counselor, which implies that the client is also the one who closes the interaction.According to social norms, (s)he is also the one who has the epistemic rights to acknowledge whether sufficient information or advice has been given.Therefore, the participant norm is that it is the client who initiates the closing of a counseling session to make minimal imposition on the other speaker, in this case the counselor (Antaki 2002;Sacks 1992).In telephone counseling, clients overwhelmingly initiate closing by acknowledging the advice, but in chat counseling clients sometimes withhold a response to the advice or respond minimally ("okay").In those cases, counselors pose questions like "can you work with this?" and "have I answered your questions?" to receive a (more explicit) advice acknowledgement from the client and thus to terminate the session (Stommel/Te Molder 2015).This implies the relation between advice acknowledgement and (pre-)closing is not just a co-occurrence; rather, the participants treat advice acknowledgement as a precondition for closing.More specifically, when clients do not overtly acknowledge or appreciate the advice, counselors are confronted with the question how to close off the session.For this reason, Stommel and Te Molder (2015) suggest that the counselor's position in online chat counseling is interactionally more fragile than the client's, at least at this point.From a client perspective, one of the affordances of text chatting is that the advice can be more easily resisted.

Providing and receiving advice
Seeking help and providing advice have been discussed in the literature as social actions that are negotiated interactionally and that are socially accomplished (Butler et al. 2010 1988).The participants in their interaction orient to the overarching institutional goal of providing help in the form of information and/or advice.They mainly do so through their actions and the way they design these actions.In the course of these actions, the participants also jointly orient to their separate roles and responsibilities as help seeker and help provider.These separate roles implicate an asymmetrical counseling relationship.By offering advice, counselors display knowledge and expertise the client (advice recipient) is assumed not to have (Heritage/Sefi 1992).
When an advice is given the projected next turn is acknowledgement (preferred) or rejection (dispreferred).When advice is not acknowledged, it may be rejected explicitly, but when the advice is not acknowledged explicitly (enough) or soon (enough), this also implies advice resistance.In text-based chat sessions pauses between posts are much longer than the usual silences between spoken turns, and their nature is different: Users may be typing or repairing typed text (García/Jacobs 1999; Meredith/Stokoe 2014), they may be involved in multiple (chat) conversations and other activities at the same time (multitasking), or the Internet connection may be slow.However, the multi-interpretability of pauses seems to have different consequences for counselors and clients.For example, only a counselor may interpret a long pause as advice resistance.Counselors may then elaborate on their advice to create a new space (compare transition-relevant-place in spoken interaction) for the client to acknowledge the advice (Kinnel/Maynard 1996), or post pre-closing questions such as "have I answered your question?" or "can you work with this?" (Stommel/Te Molder 2015).Stommel and Te Molder (2015) found that counselors topicalize (apparent) advice resistance in some way very rarely.Surprisingly, the main handbook on online counseling used in the Netherlands (Schalken et al. 2013) recommends counselors to use metacommunication, i. e. to state what happens during the interaction, when clients show signs of advice resistance.In the literature on (psycho)therapy, metacommunication has also been discussed, primarily as a strategy to manage "ruptures" in the therapeutic alliance (cf.Binder/Strupp 1997;Elliot et al. 2011;Geller/Greenberg 2002;Lietaer 1993).1There seems to be a gap between counseling instructions and practice.

Metacommunication and the stocks of interactional knowledge
Metacommunication can be defined as what we do when we reflexively talk about talk for some pragmatic purpose (Verschueren 1999).Synonyms used for metacommunication include metatalk (Schiffrin 1980) and metadiscourse (Craig 2005).The role of metacommunication in our day to day interactions depends on social and cultural aspects and may be related to specific settings: psychotherapy is pre-eminently characterized by metatalk (Schiffrin 1980).It has been claimed more generally that there is currently a pervasive interest in metadiscourse in social interaction due to our cultural preoccupation with communication (Craig 2005).The fact that recent counseling protocols like Schalken et al. (2013) include a paragraph on metacommunication (Van Gemert/Van Hoogenhuyze 2013) as a counseling strategy may be considered a result of this.
Metacommunication comprises a large set of communication practices that are employed for a variety of functions and purposes in social interaction.For example, it may be used to structure the talk at hand ("at the beginning you said..."), explicate an interactional act ("I would advise you to..."), organize participants' availability ("are you still there?"),or evaluate aspects of the preceding interaction ("If you don't respond to my questions I cannot help you") (cf. Schriffrin 1980;Van Nijnatten 2006).Additionally, metadiscursive linguistic strategies contribute to relationship and identity building.However, Gordon/Luke (2016: 29) pointed out that the relational effect of the use of metadiscourse is not straightforward (cf.Tannen 1994).For instance, talking about how counselors-in-training want to be or are addressed by their clients can enhance their (developing) professional identity but also challenge it.
The relational function of metacommunication is relevant to our analysis of interaction between counselors and clients as this relationship is asymmetrical and characterized by institutional goals such as advice giving.
Another function which is particularly relevant for our study is the use of metacommunication to overcome (communication) problems.Very basically, this function can be observed in utterances like "we should talk" (Craig 2005; see also Buttny 2010).Such utterances acknowledge there is a problem and propose talk as a solution or way to deal with the trouble.In this example, the counselor steps back from the content and discusses the interaction itself.In the first post, the counselor describes the preceding interaction in terms of interactional activities; the client repeatedly asked something while an answer to the counselor's question was interactionally relevant and withheld by the client.In her second post, the counselor enquires if the client does not like the counselor's questions, which is formatted as a declarative question (the question mark renders it a question rather than a "noticing") projecting confirmation.Thereby, the counselor questions the client's counseling recipiency.Peräkylä/Vehviläinen (2003) propose that assumptions of useful interactional strategies (such as metacommunication) may be seen as constituting a Stock of Interactional Knowledge (SIK).SIKs can be found in professional texts, in training manuals, and in written and spoken instructions delivered in the context of professional training and supervision.Peräkylä/Vehviläinen (2003) propose that Conversation Analysis (CA) may (i) falsify or correct assumptions that are part of a SIK, (ii) provide a more detailed picture of practices that are described in a SIK, (iii) add a new dimension to the understanding of practices described by a SIK, or (iv) provide the description of practices not provided by a very abstract or general SIK.By making individual solutions to general interactional problems (resulting from general institutional goals) more explicit, conversation-analytic research can contribute to a better understanding of these SIKs, which in turn could lead to an improvement in both the day-to-day practice of online counseling and the teaching of specific conversational techniques to prospective online counselors.
This study uses (digital) conversation analysis (Giles et al. 2015 Internet, 58 % used the Internet for health purposes.For this project, we collaborated with the Dutch information chat service on alcohol and drugs (hosted by Trimbos) and a Dutch general chat help service (Korrelatie).These services have the purpose of providing the general public with accurate information and advice.
The Trimbos Institute reported that 1.585 serious chat sessions took place on their online platforms in 2015, which is an increase of 18 % compared to 2014 and 30 % compared to 2013.
The average time of a chat session was 18 minutes; the protocol recommends a maximum duration of 45 minutes.On the home pages of their websites (www.alcoholinfo.nland www.drugsinfo.nl)visitors are invited to call, chat or email.The chat service can be reached via the websites on week days between 13PM and 17PM.The chat window is one on one, counselors are randomly assigned to new visitors, and no explicit statements about anonymity are mentioned.
Korrelatie reported that, in 2015, 17,000 people reached out to Korrelatie for anonymous help.42 % of these contacts were via chat, which means 7.140 chat sessions took place on their online platform.Korrelatie also reported an increase of client contacts via chat compared to e-mail and telephone from 21 % in 2013 to 34 % in 2014 and 42 % in 2015.The guidelines for a Korrelatie chat session indicate a maximum duration of 30 minutes.However, online counselors can decide whether or not to adhere to this time constraint.The homepage of the website (www.korrelatie.nl)presents a direct link to the chat service which is available on week days between 9AM and 17:30PM.When visitors start a chat session they are first contacted by the "front desk" who provides a link to a page on the organization's rules and privacy statement.When an online counselor is available (s)he will be added to the chat session and the "front desk" leaves the session.It is possible for visitors to indicate a preference for a specific counselor.
Online chat counseling is usually demand-driven; the client seeks information or advice.
Online counselors working at the chat services of Trimbos and Korrelatie are trained for these tasks by their employer.The training from both institutions is based on the handbook from Schalken et al. (2013).Most counselors have a background in social work or psychology.The websites of these organizations state information provision and advice giving as their main tasks.It is therefore not surprising that both clients and counselors usually display an orientation towards requesting information or advice as the main interactional activity.
The study was approved by the ethics committee of the Faculty of Arts, Radboud University, under the conditions that the chats did not include information that could lead to the actual participants and that all researchers had signed a vow of secrecy.

Data and analytic procedure
The data were derived from a corpus of 197 sessions of the Trimbos Institute and 348 sessions from Korrelatie.These were single sessions on a broad range of topics.We selected problematic chat sessions with metacommunication from this data set following three steps.First, we excluded chat sessions that lacked interaction between a client and a counselor.This often occurred at the general chat help service when clients were told by the "front desk" to wait for a counselor to become available.Often these clients logged out before the session started.Second, we selected sessions with "unsmooth"/problematic interaction.Inclusion criteria in this step were: client's non-responses, long pauses, overt resistance, lack of advice acknowledgement, and unilateral closing.We also used our intuition by asking ourselves the question: does the counselor have to do more work than usually for the interaction to run smoothly?All sessions were screened for problematic interaction by three people (the first author and two undergraduate students).When there was doubt or disagreement, the first author read the chat session again and decided whether it had to be included or not.These first two steps resulted in a collection of 49 sessions of the Trimbos Institute (25 % of the total Trimbos collection) and 57 sessions from the general chat help service (16 % of the total Korrelatie collection).
Of the chat sessions that we identified as problematic, in 30 sessions of the Trimbos Institute (61 %) and 30 sessions of Korrelatie (53 %), the client disengaged from the interaction.Because these sessions did not include metacommunication, we excluded them from our data set.However, disengagement may provide an explanation for the rare use of metacommunication in our data.Therefore, we will discuss client disengagement without metacommunication in some detail here.Previous studies of chat closings (Markman 2009;Raclaw 2008) analyzed ways of closing chats with more archetypical sequencing and with automated triggers.Closings tended to be designed in alignment, being initiated with a preclosing which received an aligned response.In contrast, in our data some closings are disaligning actions, hence they are "premature".Premature closing occurs when a different type of action (e.g. an answer to a question) is relevant but absent.In lieu of the relevant action, the client produces an automated trigger ("client has left the chat") or remains silent upon which the counselor produces an automated trigger ("chat is ended").Excerpt 1 shows an example of the latter variant.
The client's reason for starting the session was that her boyfriend had a bad trip and she did not know what to do.One of her first posts was help me, which gave the session an acute nature.Excerpt 1 begins with the counselor's question about the situation, which is followed by advice in line 3.

Excerpt 1
Alcohol-and drugs service (Trimbos), chat T38 The question whether the client's boyfriend is freaking out does not receive an answer for 0.36 seconds.Then, the counselor seems to post general advice on what to do in case of a bad trip into the chat session -presumably copied from a website or local information document.The client's nee ('no') in line 7 responds to the question in line 1.This becomes evident by the time stamps in lines 3 and 7 which are only 2 seconds apart, indicating simultaneous typing.The counselor, however, does not take this up and rephrases her advice in a colloquial format (lines 9-10).Then, for almost a minute, the client neither responds to, nor acknowledges the advice.This is problematic from the perspective of the counselor; advice acknowledgement is usually pursued in online counseling of the "helpline" type (single chat sessions) (Stommel/Te Molder 2015).In this case, the counselor seems to look for where she might have missed something and initiates repair (Meredith/Stokoe 2014; Schegloff 1992) of the no response (line 7).However, the client then disengages, having received advice, but not having acknowledged, accepted or, for that matter, rejected it.Hence, she leaves the session "unclosed".The counselor arguably has tried to "restore" the session, but failed.
Cases like Excerpt 1 could be easily explained by many other things than the quality of the counseling.Clients could be multitasking, be called away, have internet connection problems, etc.However, overall there seems to be a pattern that problematic interaction increases chances that a client logs out prematurely.
For our analysis, with an interest in understanding how counselors use metacommunication to respond to interactional trouble, this meant that after excluding all sessions where the client disengaged from the session without any explicit warning and no metacommunicative strategies were used, we were left with 46 sessions.The third step was to identify those sessions in which counselors used metacommunication, i. e. the counselor steps out of the content and discusses the course of the interaction (Van Gemert/Van Hoogenhuyze 2013).This resulted in a collection of 15 sessions from the chat service on alcohol and drugs and 14 sessions from the general chat help service.The 17 sessions with interactional trouble but no metacommunication that were excluded in this step involved hoaxing, transgressive questions (e. g. sexual requests), language barriers, or breaking of organizational rules, or the counselor simply "ignoring" trouble (e. g. using a pre-closing question).

Analytical approach
We used the theoretical and analytical framework of (digital) conversation analysis (CA) to examine the use of metacommunication in response to interactional trouble in online counseling (Giles et al. 2015;Hutchby/Wooffitt 2008).Traditionally CA is a data-driven research approach but in recent years researchers have been looking for ways to use CA also as an ap-Margot Jager and Wyke Stommel: The risk of metacommunication to manage interactional trouble in online chat counseling ISSN 1615-3014 199 proach to answer specific questions derived from theory and practice (Antaki 2011).The analysis reported in this article is the result of an iterative analytical process involving data, theory, and the SIK on metacommunication described in the handbook on online counseling (Van Gemert/Van Hoogenhuyze 2013).

Analysis
We observed three recurrent metacommunicative practices. 2The first practice involves the counselor regretting not being able to or not having been able to help.The second practice involves the counselor questioning the client's counseling recipiency.The third practice is different in the sense that it is the client who first initiates metacommunication, usually by accusing the counselor of not providing sufficient information or advice.The counselor, in response, uses metacommunication to explain the institutional boundaries, thereby implicitly accusing the client of having unrealistic expectations.

Regretting inability to help
The first practice we identified was the counselor regretting not being able to, or not having been able to help.This strategy usually occurred towards the end of a chat session.Excerpt 2 shows an example of the counselor accounting for her own lack of counseling competence.
The client is a 36 years old woman who has problems with her husband; she wants to separate but cannot afford it.Her initial question was how she could live with him in one house, while he is verbally aggressive and unfaithful to the arrangements they made.The counselor has advised her to leave the house if it is impossible to live with him, but the client has not accepted this advice, which brings the counselor to ask: could you formulate your question to me once more, [name client]?After some more posts back and forth the client writes: now that I am typing it I see it is a bit of a difficult question.And I realize I am the victim here.
The interaction continues with posts about possible explanations of the husband's aggression and treatment he received.Excerpt 2 begins 44 minutes after the start of the chat session, which is already beyond the time limit.

Excerpt 2
General help service (Korrelatie), chat K414 In line 1 the counselor initiates closing of the chat session due to time constraints (referring to the maximum of 30 minutes for a session as announced on the website).In this same post, she also evaluates the session using metacommunication, which turns out to be a rather selfdeprecating assessment, as she writes Ik heb het gevoel dat ik je niet echt heb kunnen helpen Margot Jager and Wyke Stommel: The risk of metacommunication to manage interactional trouble in online chat counseling ISSN 1615-3014 201 ('I have the feeling that I haven't really been able to help you').Thereby, she implicitly devaluates her own counseling competence.In a next sentence, she seems to shift from discussing her counseling competence towards shared "blame" (we hebben elkaar niet steeds begrepen ('we haven't always understood each other') of an apparently unsuccessful chat session.
After another post of the counselor that is a response to an earlier suggestion of the client that her husband might suffer from PTSS, the client states: ik heb er wel wat aan gehad ('it did help me though') (line 8).This is a direct response to the counselor's metacommunicative post in lines 1-3.By using wel ('though') the client seems to deny not just the counselor's competence but also the suggested lack of mutual understanding.She explains this by positively evaluating the "talking" (line 10), which implies that her advice seeking question at the start of the session has been replaced with the request to talk about it, which is exactly what she got from the counselor.In line 12, the client further accounts for her positive evaluation of the chat session by complimenting the chat service in general using the plural pronoun jullie (English pl.'you').In this example, the counselor's strategy of self-deprecating leads to several posts of the client denying the counselor was not able to help.In this case the selfdeprecating strategy of the counselor clears the floor for an aligned closing of the session.The participants use nine more posts (lines 16-34) to collaboratively close the session.

Questioning the client's advice recipiency
The second practice we identified was the counselor questioning the client's counseling recipiency.Excerpt 3 shows an example.The client presented himself as a lecturer whose student has been hospitalized at one of the institutions for addiction care which man the chat service.He wants to write the student a postcard and asks for the room number, but the counselor has no access to this kind of information.The client has just given the name of the student when Excerpt 3 starts.

Excerpt 3
Alcohol-and drugs service (Trimbos), chat T68 The counselor rejects the client's request for information and provides a reason for it (lines 1-2).When this rejection does not receive a reply, the counselor proposes a further action, making relevant a response.However, a response does not follow for over two minutes (line 6).Then, the counselor asks whether the client is still engaged with the session.After another pause of more than a minute, in lines 9-10 the counselor uses metacommunication to "accuse" the client of not fulfilling his interactional role: ik probeer je te helpen maar als ik geen reactie krijg dan gaat dat moeilijk ('I am trying to help you but if I don't get response that's difficult').This post also does not receive a client response and after 51 seconds the client disengages from the chat session by logging out.In lines 1-3, the counselor refers the daughter and her mother to an organization related to the Trimbos Institute where they can participate in group sessions from an addiction care institution.This is an external service but not a home service contrary to what the client desired.
In lines 5-6, when the client has not acknowledged the advice, the counselor poses a question that elicits an advice acknowledgement as it heavily prefers yes as an answer.This format is similar to the pre-closing questions (for example: "Have I answered your question like this?") described by Stommel/Te Molder (2015).Alternatively, the counselor could have responded to the advice resistance here using metacommunication.In this case, an answer like "yes, thank you for the advice" would have initiated the closing.However, instead, in line 8 the client claims to understand the counselor has 'to say all this' (Ja ik begrijp dat u dit allemaal moet zeggen), which disqualifies the counselor as a competent advice provider.She continues to resist the advice (even if she does not seem to refer to the advice to take part in a session).
Next, in lines 12-16, the counselor attempts to reinstall her advice competence by denying she has to say things and making explicit her institutional goal: ik probeer jullie te helpen ('I'm trying to help you').She goes on giving advice in directive formats using deontic modalities and imperatives.Note the client does not post a response now and non-response in this case bolsters the earlier advice resistance.In a further post (lines 18-20), the counselor accounts for her forceful advice in lines 12-16: ik weet dat dat makkeliker gezegd dan gedaan is, maar ('I know it is easier said than done, but').Again, there is no uptake from the client.Then the counselor posts pre-closing questions of the type we saw earlier, initiating closing.The client thanks the counselor which is "doing advice recipiency", but also denies she has been helped, hence declaring the institutional goal has not been achieved.The counselor now uses metacommunication by "regretting" unsatisfying advice but also by questioning the client's advice recipiency and defending her own advice competence.This issue is not resolved and the session is not closed collaboratively (the counselor's post in lines 29-30 is the last one in this chat session).

Explaining institutional tasks and responsibilities
The third practice we identified was the counselor using metacommunication to explain the institution's tasks and responsibilities, thereby implicitly accusing the client of having unrealistic expectations.This practice occurs in response to allegations or complaints of the client, and is typically used in combination with other meta-communicative practices.
Excerpt 5 shows an example where the counselor refers to the organization's restricted tasks and goals as an account for not having been able to help in response to a complaint of the client.In this session the client is a 19-year-old man who is wondering whether he might have an alcohol problem.Already in his first post he expresses his doubts (I am not sure but...) and when the counselor asks why he thinks he might have an alcohol problem he refers to friends who addressed his drinking behavior.Before the start of this excerpt the session has been running for 35 minutes already (the protocol suggests a maximum duration of 45 minutes), and the counselor eventually provides advice by suggesting that the client may want to look at a website of the addiction care institution in his neighborhood and make an appointment there.This is where Excerpt 5 starts.

Excerpt 5
Alcohol-and drugs service (Trimbos), chat T71 The counselor's advice in lines 1-2 is, after a short pause, followed by a pre-closing question eliciting advice acknowledgement (Stommel/Te Molder 2015).This elicits the client to accuse the counselor of wanting to cut off the chat (lines 6-7).The counselor starts her metacommunicative response to the client's accusation by stating the main task of the chat service (referral), which is followed by an apology: De chat is een doorverwijsplek, excuses als je je afgekapt voelt dat is niet de bedoeling ('The chat is a place of referral, apologies if you feel cut off that is not the intention').However, this apology is downgraded by transferring responsibility for the offense to the client; the counselor apologizes for the client's offended feelings and states that cutting off was not her intention.Next, the counselor provides an account (lines 12-13) explaining that it was his lack of response to her advice which lead to the preclosing move.This way, the counselor not only avoids taking responsibility for the client's feelings of offense, but even suggests that the client is the one who caused interactional trouble by not complying to the advisory nature of this particular institutional context.The client provides an account for his lack of recipiency (line 15), thereby treating the accusatory metacommunicative remarks of the counselor as legitimate.After another 2.15 minute pause the counselor makes another (accusatory) metacommunicative move by stating that she has good intentions, ik zou je graag verder helpen ('would love to help'), which she cannot fulfill due to the client's lack of responsiveness to her ideas.The client does not respond and the counse-  -20).Now the client responds and acknowledges the advice the counselor gave earlier in the session (talking about it with his therapist).After this the counselor initiates closing which is collaboratively achieved (data not shown).
Excerpt 6 shows another example of a counselor explaining the institution's tasks to respond to a complaint from a client.Here, the client is less compliant than the one in Excerpt 5 and starts a negotiation of what the counselor has to offer.In this session the client is a 41-yearold woman who, already in her first post, provides a lot of information about her life and the problems she encounters.The main topics she addresses are her (disappointing) therapeutic history and her potential new job.Before the start of this excerpt the session has been running for over 30 minutes already, and the counselor eventually provides advice by suggesting that the client may want to look for some sort of coaching to help her in her new job instead of going for therapy.This leads to several posts in which the client rejects the advice and evaluates it as not new.This is where Excerpt 6 starts.

Excerpt 6
General help service (Korrelatie), chat K163 After this rejection of her advice in lines 1-5, the counselor uses metacommunication to accuse the client of not giving her space to help, thereby implicitly blaming her for not being a good advice recipient, i. e. if you ask for advice you should allow the counselor the opportunity to give advice.In the second part of the accusatory post (lines 7-8), the counselor suggests that the client has unrealistic expectations of what the organization has to offer: misschien verwacht je meer van deze organisatie dan we je kunnen bieden ('maybe you expect more of this organization than we are able to offer').Thereby, she does not treat the client's accusation as directed to her personally but as inherent to the nature of the service.The client, in response, does not defend her advice recipiency, but initiates a discussion of the tasks and responsibilities of the counselor (as a representative of the organization).This way, she invites the counselor to explain herself, which is indeed what happens in lines 16-18.After a general description of the way things work, the counselor again accuses the client of not giving her enough space, and even to be judgmental on the things she says.After a short pause (line 21) the counselor downgrades her prior accusation by explicitly stating that she does not blame the client for the interactional trouble.However, the second part of her post is a butprefaced implicit accusation of the client making it difficult for her to fulfill her institutional role as help provider.In line 25, the client then starts to explain herself, thereby treating the counselor's accusation as legitimate.The session goes on with several posts in which client and counselor further discuss what it is the client is looking for.The session is getting back on track and eventually has a positive collaborative closing.
The sessions in which the metacommunication stretches to a discussion of the nature of the particular service also always include accusatory posts with regard to advice recipiency.Overall, this strategy seems less disturbing for the counseling relationship, as it moves the discussion away from personal accusations.

Discussion
In this article we have described how counselors use metacommunication to respond to interactional problems in online counseling of the "helpline" type (single chat sessions).We found that metacommunication is used, but only rarely.Clients tend to log out from problematic sessions, often even before the counselor is able to identify interactional trouble at all.This means online counselors operate under the constant risk of clients' disengagement.Second, we identified and described three metacommunicative practices counselors use to respond to problematic interaction: 1) Self-criticizing; the counselor is regretting (s)he was not being able to help due to a lack of useful advice or to the boundaries of the institutional context.This strategy seems to elicit client acknowledgement followed by an aligned closing; 2) Accusing; the counselor is questioning the client's advice recipiency.This strategy seems to elicit even more client resistance.3) Explaining; the counselor explains the institutional tasks and responsibilities, thereby implicitly accusing the client of having unrealistic expectations.Hence, each of these metacommunicative strategies contributes to the relationship between client and counselor, either restoring it or harming it further (cf.Gordon/Luke 2016).
The three strategies are sometimes used in combination in one post, or step by step during the session.
The scarcity of metacommunication as a strategy to respond to interactional problems is surprising given the ritualized nature in which we use metadiscourse to restore (communication) problems (Craig 2005).It might be related to, or explained by, a number of aspects relevant to the medium and the setting.First, the online text-based medium creates the possibility or risk of disengagement.The client may be gone before the counselor has (taken) the chance to initiate a discussion on the course of the interaction.It seems that the affordance of logging out puts the client in a structurally more powerful position than the counselor.Second, counselors may be reluctant to use metacommunication because they are uncertain whether the client really is resistant towards the advice.After all, they have only typed words.Aspects that may help to detect resistance are absent in chat interaction; there are no voice and intonation, no facial expressions and body movements available to base their judgment on.Acting upon advice resistance seems delicate, probably partly because of its potentially accusatory nature.Therefore, counselors may be inclined to interpret silences and minimal acknowledgements in response to advice as related to the medium of chat rather than as resistant behavior.Third, acting upon advice resistance by topicalizing it (metacommunication) can harm the counseling relationship.Research on counseling and therapy has found that the client-counselor alliance is a major factor determining effectiveness of care (Duncan et al. 2010).In timeconstrained, anonymous, text-based online counseling such an alliance may be more difficult to establish, more easily ruptured, and harder to resolve.One reason for the lack of metacommunication to respond to interactional trouble in online counseling may be that it is highly face-threatening and therefore (unconsciously) avoided by counselors.Effective metacommunication presumes some level of established positive therapeutic alliance (Kiesler 1996).Research suggesting that metacommunication may be an effective strategy to manage "ruptures" in the therapeutic alliance is all based on longer counseling trajectories (cf.Binder/Strupp 1997;Elliot et al. 2011;Geller/Greenberg 2002;Lietaer 1993).Online counselors only have a single session to build an alliance and to solve alliance ruptures when these occur.When initiating metacommunication, counselors need to figure out how to topicalize the (disrupted) interaction, while at the same time avoiding an irrevocable alliance rupture.Viewing the identified practices as metacommunication, we conclude that metacommunication can be useful to restore the relationship with the client when it is apologetic.A minor drawback of this strategy may be that it reduces the asymmetry between the counselor and the client.This may even undermine the counselor's identity as an expert and therefore a competent and Margot Jager and Wyke Stommel: The risk of metacommunication to manage interactional trouble in online chat counseling ISSN 1615-3014 209 reliable advice giver.By using apologetic metacommunication the counselor's identity is reduced to someone who is at the most able to offer a listening ear.
The analysis shows that metacommunication can also be counterproductive to deal with problematic interaction.On this basis, it seems fair to question the SIK of metacommunication in single session online counseling.More precisely, our analysis corrects the conviction that metacommunication is an effective strategy in online counseling.It can be, but it frequently is not.This happens when the counselor questions the client's willingness or ability to accept the advice, which is a reproach.This type of metacommunication is severely face-threatening and elicits more resistance rather than advice acknowledgement.Hence, it further distorts the counseling relationship.An addition related to the risk of metacommunication seems legitimate.Metacommunication may be perceived as accusatory, which could disturb the fragile counseling relationship or it may undermine the counselor's expertise and counseling competence.For this reason, our analysis warrants the advice to avoid metacommunication as a strategy in single session counseling.
Excerpt 4 shows another example.The client is a 17-year old girl who explains that she has found out that her father is addicted to smoking marihuana.He is very aggressive and denies everything.Her question is: what should I do?.The counselor immediately posts extensive advice aimed at the girl: she should seek help to discuss how to deal with this situation, and her mother should be the one who solves things with her husband (the client's father).The client rejects this advice with: I don't need help, he needs help I would think!!After another advice-resistance sequence focusing on the mother's role, the client poses a concrete, practical question, suggesting what the family needs is an external service to take care of the problems with her father.The counselor only loosely responds to the client's question, and posts another piece of advice.This is where Excerpt 4 starts.for a long time I think that your mother should impose consequences.from now on no smoking anymore or else he leaves the house.and in case of aggression call the police for a change.you have to make clear to your father that it is really not accepted anymore' voeren vind ik dat het vanuit je moeder ook eens 14 consequemties moet hebben.vanaf nu niet meer blowen of anders gaat hij 15 het huis uit.en bij agressie een keer de politie inschakelen.je moet je 16 vader duidelijk maken dat het echt niet meer gepikt gaat worden 'I don't have to say things, I'm trying to help you [pl].if you [pl] are al- En wat heeft het formuleren van doelen voor zin.Dat heb ik al miljoen keer 2 gedaan.weetje Co ik begin in dit gesprek alleen maar bozer te worden.'Andwhat is the use of formulating goals.I've already done that a million times.you know Co I start to get even more angry in this conversation.We try to support in finding solutions here.Sometimes it helps to give advice and sometimes it helps to examine why certain things don't work out.I have the feeling that you've already had so many negative experiences that I get little space to think with you and am judged on many things that I say.