Deferring the Decision Point: Treatment Assertions in Neurology Outpatient Consultations

ABSTRACT Recommendations can be implied by asserting some generalisation about a treatment’s benefit without overtly directing the patient to take it. Focusing on a collection of assertions in UK neurology consultations, this paper shows that these are overwhelmingly receipted as “merely” doing informing and argues that this is made possible by their ambiguous design: their relatively depersonalised formats convey that the neurologist is simply telling the patient what’s available, but the link made between the treatment and the patient’s condition implies that it will be of benefit. Thus, assertions, while stopping short of telling the patient what to do, are hearable as recommendation relevant. This delicates balance leaves it up to the patient to respond either to the implied or on-record action (recommending vs. informing). When treated as “merely” doing informing, assertions defer the decision point until the neurologist has done something more. Three main interactional functions of this are identified as follows: (i) indicating the existence of a solution to a concern, without making a decision relevant next; (ii) orienting to the patient’s right to choose; and (iii) making “cautious” recommendations.

As demonstrated by Stivers et al. (in press), recommendations can be implied "through the assertion of some generalisation about a treatment's benefit… without proferring a directive" (p. x). For example, in the UK neurology dataset, which forms the basis for the analysis presented in this paper, clinicians regularly initiated decision-making trajectories by asserting either that a particular named (type of) treatment could be helpful or, more generally, that suitable treatments existed for the patient's symptoms. Assertions thus "sit at the boundary between informationproviding statements and recommendations" (Stivers et al., in press), implying that the patient may benefit from starting treatment, but stopping short of an on-record recommendation that the patient ought to do so. This has implications for how patients might appropriately respond in next turn. As Stivers et al. (in press) have demonstrated, it is possible for patients to treat assertions as recommendations, responding by accepting or resisting the available option(s). However, it is also possible to treat the prior turn as simply doing informing, thereby responding only to the form of the turn and not to its implied action as a recommendation. As this paper will show, the latter was overwhelmingly the case for assertions in neurology, which were almost always followed by a minimal acknowledgement (e.g. "mm," "right," or a nod) or no response at all. Thus, it will be argued that these responses should not be assumed to be doing passive resistance, as they can be heard to do following full, on-record recommendations (Koenig, 2011;Stivers, 2005aStivers, , 2005bStivers, , 2007Stivers et al., in press). Rather, the analysis presented below shows how assertions-despite clearly being recommendation relevant-do not, in themselves, demand an immediate decision from the patient. By making it readily possible for the patient to treat them as "mere" informings, assertions can defer the decision point.
In this regard, assertions function similarly to patients' illness explanations as analysed by Gill and Maynard (2006). These authors show how illness explanations, by virtue of their design and sequential placement, can leave open what kind of response is relevant next (see also Stivers & Rossano, 2010). For instance, in Extract 1, the patient responds (line 4) to the doctor's question (lines 1-3) in a way that may imply that prior surgery (involving both a hysterectomy and bladder repair) caused her current experience of pain during intercourse. She also "adds a more overt, speculative explanation" (Gill & Maynard, 2006, p. 124) at lines 4-5, which focuses on the bladder repair specifically as a potential cause. In response (line 6), the doctor deals with the new information regarding how long she has been experiencing the pain. Although he implicitly notes the potential connection with the surgery ("ever since that surgery," line 8-9), he does not explicitly confirm or disconfirm whether the "bladder tie up" may be a cause of the patient's current pain.
Extract 1 (from Gill & Maynard, 2006, p. 124-see their Extract 9) By contrast, as Gill and Maynard (2006) show, patients can design their explanations as "frank questions that narrowly restrict doctors' response options, such that doctors are compelled to provide 'answers' by evaluating the explanations" (p. 126, italics in original). The distinction, then, is between those explanations that place significant interactional pressure on the doctor for a particular kind of response, and those that largely leave it to the doctor to decide how to respond; for example, with "a confirming or disconfirming evaluation" or, instead, by "treat[ing] the report as 'information' or 'data' and proceed[ing]… by simply nodding, or otherwise indicating receipt of the report" (p. 124). The present paper demonstrates how assertions function in a similar way to those explanations that place little interactional pressure on the recipient to address the implicit social action.
Although treatment recommendations have long been of special interest to those investigating communication in medical care, there has been little work exploring whether particular turn designs may be employed, systematically, in particular sequential environments. A notable exception is a study of psychiatric consultations in Japan, which compares two formats for treatment proposals: an inclusive "we" form (translated as "let's do x" or "how about x") and declarative evaluations (such as, "it might be better to x") (Kushida & Yamakawa, 2015). The authors demonstrate that the former turn design is used when decision-making proper is relevant next; the latter, by contrast, "is used to propose a treatment cautiously when the sequential environment is not yet ready for decision-making" (p. 522). Kushida and Yamakawa's focal turn designs all fall within the broader action type of proposals. Their declarative evaluations are not, then, direct translations of any of the assertion formats identified in the present study. However, they appear to be functioning in similar ways. As Kushida and Yamakawa argue, although these turns are recognisable as proposals, they do not create an on-record decisionmaking moment, partly due to the grammatical form, whichlike the assertions analysed here-can be treated simply as information. It is thus left to the patient whether to respond with a news receipt or to orient to the "less official" action of proposing (p. 532).
Following a similar line of analytic argument, this paper will show how assertions are poised between doing "simple" informing and doing recommending. Exploring their function, the paper will argue that the "off-record" nature of assertions makes them well suited to performing three types of interactional work: (i) indicating the existence of a solution to a patient's concern, without making a decision relevant next; (ii) orienting to the patient's right to choose; and (iii) making "cautious" recommendations.

Data and method
The analysis reported here made use of an existing dataset, collected as part of a project aimed at identifying how clinicians offer patients choice, and the interactional consequences thereof (Reuber, Toerien, Shaw, & Duncan, 2015). The original project was funded by the United Kingdom's National Institute for Health Research. The main dataset consists of recordings of 224 consultations collected in two major clinical neuroscience centres in the UK, between February and September 2012. Participants could choose whether to be audio or videorecorded. Approvals were obtained from the appropriate UK National Health Service Research Ethics Committee and the participating hospitals' Research and Development departments.
For the collaborative project reported here, 50 physicianinitiated recommendations were identified that met the codebook criteria. Each of these was coded for the features described in Stivers et al. (in press), including the five main types of social action performed through the recommending turns: pronouncements, suggestions, proposals, offers, and assertions. This produced an unexpected finding: almost half of the neurology cases (48%) were coded as assertions. This is more than three times the number found in the UK psychiatry consultations (15%) and three times the number in the UK and US primary care consultations combined (16%). Assertions were more common in the UK primary care consultations (16%) than in those from the US (5%), but clearly the neurology effect goes beyond the UK-US difference.
Subsequent qualitative analysis thus focused in detail on the neurology assertions specifically, using the tools and perspective of conversation analysis (CA) to examine how these functioned in the recorded clinical interactions between neurologists and patients. This involved searching for patterns in the design of the assertions, patients' responses to them, and the sequential placement of the assertions within the wider interaction. For introductions to conversation analysis, see Drew (2005), Sidnell and Stivers (2013.

Introducing neurology assertions and their responses
In the neurology dataset, clinicians initiated treatment decision-making trajectories by asserting either that a particular named (type of) treatment could be helpful (as in Extract 2) or, more generally, that suitable treatments existed for the patient's symptoms (as in Extract 3). Each of these implies a recommendation by making a link between the treatment and the patient's complaint(s). In Extract 2, this is done through mentioning a diagnostic category ("an acute relapse") that was applied to the patient earlier in the consultation. The definite article ("the steroids" rather than "steroids") also implicitly links the assertion to an earlier discussion about whether the patient had already been prescribed these (he had not). In Extract 3, the link is made through use of the definite article to index the patient's earlier report of symptoms: "the burning sensations… the… painful and tight sensations". Thus, these turns-in common with all the assertions in our collection-are not neutral with respect to patients' concerns; all are tilted, at least somewhat, in favour of treatment as a means of addressing these. Given the generally agreed right for clinicians to "know best" about treatment, this tilt carries particular institutional weight. As Stivers et al. (in press) put it: assertions "can carry the force of a recommendation because they leverage the epistemic authority of the physician into the deontic force of a recommendation through a stepwise process of inference" (p. x).
Nonetheless, assertions stop short of an on-record recommendation that the patient start a particular treatment. They are, then, deontically mitigated in comparison to the other formats shown in Stivers et al. (in press). In Extracts 2 and 3, this is evident in the work done to avoid overtly personalising the information provided. It is presented as a general fact that "the steroids" can be helpful for an acute relapse (Extract 2) and that "medications" exist for treating certain symptoms of multiple sclerosis (Extract 3). Neither neurologist refers explicitly to his own view on the treatment ("the steroids can be helpful" rather than "I think steroids would be helpful," and "there are medications for treating" rather than "I would suggest taking medication"), and neither neurologist refers to the patient personally ("an acute relapse" and "the burning sensations" rather than "your symptoms"). Assertions are poised, then, somewhere between informing-"merely" providing the patient with information about available treatment-and a recommendation that this patient ought to take this treatment.
Although, as Stivers et al. (in press) have demonstrated, patients can and do sometimes treat assertions as recommendations, it is also possible to treat them as simply doing informing. The latter was overwhelmingly the case for assertions in neurology, which were almost always followed by a minimal acknowledgement (e.g. "mm," "right," or a nod) or no response at all. Extracts 4-6 provide illustrations, with assertions in boldface and and arrowed lines showing their responses. The patients in 4-6 treat the prior turns as nothing more than informing, to be-at most-simply acknowledged (see Gardner, 1997Gardner, , 2001Gardner, , 2007. They neither actively accept nor actively resist the implied recommendation for the medications just introduced. Crucially, such minimal responses occurred routinely after neurology assertions, regardless of whether the patient went on to accept or resist a treatment option.
The rest of the analysis presented here explores the function of the neurology assertions, arguing that their "off-record" nature makes them well suited to performing three types of interactional work, each of which is discussed below.

Indicating the existence of a solution to a patient's concern
Of the 24 assertions in the neurology data, five occurred in immediate response to a patient's reported concern or ongoing trouble, and an additional case responded-at some remove-to the patient's request for a medication review as part of her reason for the visit, this being predicated on her concern about possible side effects. 1 Cases included in this section total 25% of the neurology assertions. Extract 7a shows an example from a multiple sclerosis (MS) review. As part of the history-taking (not all data shown), the patient has revealed that she had a relapse the previous year, keeping her off work for months. She discloses several ways in which she has been struggling since then. The extract begins at the point where the patient is summarising the pressure she is under (lines 1-2 and 4-5 and 8).
Confirming that this will be affecting her condition (lines 7, 9), the neurologist explains about the relationship between MS and stress. Although it is difficult to hear what the neurologist says, the patient's response reiterates the negative impact of being busy (lines 20-22). Through a series of questions just like the preliminaries shown in Barnes (in press), the neurologist works towards some possible solutions, centred on obtaining support at work (lines 24, 26, 30). Having already tried the courses of action implicitly recom- Although we excluded cases where patients overtly initiated a treatment decision-making sequence (see Stivers & Barnes, in press), we included cases where neurologists considered treatment in response to patients' reported concerns about symptoms, since proposing treatment as a solution to patients' troubles is core to the business of medical consultations. mended through these questions (see also Shaw, Potter, & Hepburn, 2015), the patient concludes that there is little to be done since the job itself cannot be changed (lines 38, 40-42, 45). This leaves an on-going problem, construed as not solvable through the courses of action introduced by the neurologist. It is in this context that he produces a turn containing two assertions: "there are medications that can help with the sensory symptoms" (lines 46-47, 49), and "there are medications that can help with the fatigue" (lines 49, 51).
Extract 7a (UK050101401-2; Multiple Sclerosis) These assertions are poised between doing informing and doing recommending, following a very similar format to that in Extract 3. This entails the impersonally formulated announcement that "there are medications," followed by the construction of these as potentially helpful for the patient's reported symptoms. The assertions in Extract 7a respond directly to the patient's unresolved troubles at work, indicating that a (potential) solution is-contrary to her expectations -available. They are thus hearable as recommendation relevant, without overtly directing the patient to take the proferred treatment. The patient treats this as information only, producing minimal acknowledgements at lines 48, 50, and 53. In next turn, the neurologist resumes the on-going activitythe MS review-with further questioning (shown in Extract 7b, lines 54, 56, 58). Only once this is complete, does he return to the matter of treatment, producing a mitigated recommendation in the form of a suggestion (boldface lines 70-71, 73), which the patient accepts at line 75. Through further discussion (not shown), they ultimately settle on pregabalin, following the patient's announcement that she was treated (effectively) with this in the past. The neurologist describes this as a "newer version" of gabapentin, thus making of this a decision about which form of the drug to take. The patient readily accepts the recommendation to resume her previous dose.
Extract 7b (UK050101401-2; Multiple Sclerosis) The trajectory shown in Extracts 7a-b was common across all six of the responsive assertions: patients treated these as information, responding minimally in five, and with a (delayed) positive assessment of the information in the sixth; only when the neurologists went on to make an explicit recommendation (which occurred in all six), did the patients engage overtly in decision-making, treating the prior turn as making acceptance of a treatment relevant. This usually occurred at some remove from the assertion, depending on what activity was underway. For example, when they were partway through history-taking, the explicit recommendation was deferred until that was complete. Thus, these assertion turns handled the interactional relevance of responding to a patient's concern with a treatment solution, while at the same time avoiding setting up-immediately next -the activity of treatment decision-making. This was achieved by using a turn design that placed very little pressure on the patient for an immediate acceptance (see Gill & Maynard, 2006); minimal acknowledgements were routinely produced and routinely treated as sufficient for the neurologist to return to the ongoing activity. Assertions can function, then, to delay the decision point in the service of first completing another activity. [Oh(h)(h)a->two or three cups of tea a day< 60 I don't (drink)/(take) coffee, ((8 lines omitted in which the neurologist asks about the patient's intake of 'fizzy drinks' and explains that these "can sometimes create symptoms")) 69 Orienting to the patient's right to choose A further six of the 24 neurology assertions (25%) were used to introduce a list of treatment options and/or to introduce one or more of the options. In an additional eight (33%) cases, the neurologist initiated a decision about a single treatment but made it clear that not treating was an option, giving the patient a potential yes/no choice. This section considers both a) multiple options and b) yes/no decisions, showing how assertions can allow neurologists to defer the decision point in the service of orienting to the patient's right to choose. Cases included in this section total 58% (n = 14) of the neurology assertions.

Multiple option-listing
Extract 8 shows a case where assertions function as part of a wider trajectory known as "option-listing" (Toerien, Shaw, & Reuber, 2013). Option-listing typically includes both an announcement that there are options and a subsequent listing thereof (often with information about their pros and cons). In Extract 8, the neurologist self-repairs from naming a particular tablet (lines 2-3) to setting up a list of options (line 3), thereby announcing that there is a choice. He goes on to list the options (bolded lines 4-12). Both the announcement of a forthcoming list and each of the options are introduced using the format, "there is/are x," where x is a generic reference to treatment (line 3), a specific treatment name (lines 4 and 8), or a reference to a drug type (line 6).
Extract 8 (UK050207201; Diagnosis not yet confirmed (awaiting scans) but patient experiencing symptoms described by neurologist as "neuropathic pain") In setting up a list, the neurologist projects that there is a decision to be made, but that the decision point is to be deferred until all options have been described. As shown in the previous section, the assertion format readily achieves this deferral by virtue of being treatable as "just informing". Patients can thus respond minimally to the initiating assertion-for example, the nod in Extract 8, line 5-without being seen to be resistant to treatment. Furthermore, the relative neutrality of the assertion format allows the neurologist to list each treatment in a way that, at least to some extent, poses each as a viable option. For instance, in Extract 8, each option is framed impersonally ("there is/are"), thereby holding back from an overt recommendation for one over another. At the same time, the neurologist makes a link between the options and the Patient's complaints, highlighting the potential benefit. He also positions the options as clinically endorsed-through the institutional "we" (line 3) and the claim that these tablets are in common use for pain (lines 1-3). Thus, while efforts are made to produce all the options as legitimate, there is a slant in favour of treatment of some kind. Nevertheless, as long as the listing is on-going, the patient can produce minimal responses (lines 5, 9, 14) without being heard to be resistant-precisely because the construction of a list defers the relevance of the decision itself.
How the decision point was constructed varied considerably across our option-listing cases. Notably, these ranged with respect to how "open" the decision was (Reuber et al., 2015;Toerien, Shaw, Duncan, & Reuber, 2011). In some, the neurologist positioned the decision as lying entirely in the patient's domain. For example, following extensive information provision regarding migraine treatment options (data not shown), the neurologist in Extract 9 explicitly seeks the patient's preference (shaded turn, lines 9-10).

Extract 9 (UK060601601; Migraine)
Extract 10 also shows a turn in which the neurologist seeks the patient's view on treatment (see line 6). In both cases, the assertion turn itself occurred at some remove from the decision point, with extensive information provision about the treatment options occurring in between (data not shown). They differ, however, in the extent to which the neurologist is willing to provide his view. In Extract 9, the neurologist seeks the patient's preference (lines 9-10) after explicitly excluding himself from the decision (see boldface lines 3, 6). By contrast, in Extract 10, the neurologist seeks the patient's view after voicing his own: "I think we should consider… disease modifying therapies" (boldface lines 5-6). Although cautiously worded, this is a clear bid for treatment, backed up with the treatment-relevant diagnostic announcement that the patient appears to "have had a further relapse" (line 4).  It is not the case, then, that assertions-and the information that typically follows them-avoid conveying anything of the neurologist's stance towards the options. The impersonal framing notwithstanding, this stance can become evident both through the design of the list and the way in which the patient is (and, sometimes, is not) given a chance to express his/her view. Rather, the analysis shows how the delicate balancing act that assertions perform-poised between informing and recommending-can be used to defer decision-making. That deferral is done here in the service of offering choice (however tilted in a particular direction that choice might be). By asserting both that there is more than one option, and then using assertions to list the options, neurologists can deliver extended information as a basis for decision-making, while holding back both from an overt recommendation and from placing interactional pressure on the patient for an "active" response to each option. This, then, sets up the possibility for the patient, subsequently, to choose from those options that have been introduced.
Yes/no choice Assertions were also used to set up the possibility of a yes/no choice for the patient, as shown in Extract 11. The extract begins partway through a wider decision-making activity. Having received overt acceptance of a recommendation against diseasemodifying therapy (data not shown), the neurologist marks a shift to a second issue: her "taking things easy" (lines 1-2). The patient's difficulty with fatigue was implicitly introduced near the start of the consultation, through her claim to be feeling better after reducing her activities (not shown). At lines 1-2, the neurologist picks up on this, seeking confirmation that she is "not working at the moment." The patient responds with an account of her difficulties at work and her decision to take time off (not shown). The neurologist pursues this through a series of questions, which address the symptom of fatigue. The last of these-like those analysed in Barnes (in press)-checks whether the patient has already received treatment for the problems she has described (lines 26-27). When her response (lines 28, 30) clears the way for a recommendation on the grounds that she has not received treatment, the neurologist produces the focal assertion turn. This appears to be tied back to his prior question through the use of "because" (line 31), making this an account for the prior question, although this is not clearly audible.
The assertion is framed using the now familiar formulation, "there are x," where x is a generic reference to "certain drugs" (line 31). This turn, once again, is poised between an informing and a recommendation. The impersonal framing presents this as information about the existence of treatment rather than as a recommendation that this patient ought to take the treatment. At the same time, the claim that this treatment "can help" (lines 31-32) the patient's reported symptom-which has just been described as getting in the way of her work-tilts this in favour of treatment. Responding to the assertion, the patient does no more than receipt the information (line 33), this time with a continuer, indicating her understanding that more is to come (Schegloff, 1981). The neurologist continues with more "informing," explaining both the nature of fatigue in MS and how its treatment can help improve the patient's symptoms (lines 34-44). This information is clearly tilted in favour of treatment-evident especially in the increment, which upgrades the neurologist's claim from "Can improve these symptoms" to "Can improve these symptoms. Quite significantly" (line 42 and 44). Throughout, however, the patient produces either continuers or minimal receipts (bolded lines 33, 36, 38, 40, 43, 45), thereby not treating the decision itself as relevant yet. Extract 11 (UK50100201; Multiple Sclerosis) 01 Neu: U::m (0.6) no:w in terms of (0.4) taking things: easy, 02 you're not working at the moment, ((23 lines omitted during which they establish the kinds of difficulties the patient was eperiencing)) 26 Neu: And-(0.2) have you been given any medication 27 to help with the fati::gue. A decision point is produced in an on-record fashion only at lines 47-48, where the neurologist reissues the assertion, still framed with the impersonal "there are x" format and done with the generic reference form ("a few medications"). This second version includes more endorsement through the more personalised part of the turn: "that we can try to try and achieve that" (compare with "that can help with fatigue," lines 31-32). However, the neurologist quickly adds a question to the patient (lines 47-48). This pulls back from the upgraded deontic force, making relevant the patient's announcement of whether she wants to try the treatment. Like the decision points shown in Extracts 10 and 11, the turn marked as →II in Extract 12 explicitly orients to the patient's right to choose. In response, the patient gives the go-ahead (Schegloff, 2007) for the neurologist to recommend treatment more fully. This he does by providing further recommendation-relevant information, now naming a specific drug (line 52), and explaining how to take it (lines 55-59), its potential side effects (lines 83-84), and its likelihood of working (lines 86-90). This assumes the patient will start treatment. In that sense, lines 51-90 may be seen as a recommendation for amantadine, specifically. However, the neurologist also goes one step further, producing a personal endorsement of the drug (line 92) and an account for recommending it (lines 95-96, 98). Only then does the patient produce an explicit acceptance: "Okay" (line 93) and "Okay that's fine" (lines 99, 101), despite having given the go-ahead at line 49.
This was consistent across the neurology assertionsexplicit, immediate acceptance was never done in next turn (i.e. line 33 in Extract 11), even when the patient went on to accept treatment (apparently quite readily) later in the decision-making trajectory. Extract 11 illustrates this, showing that the minimal responses across lines 33-45 do not appear to be related to forthcoming resistance to treatment. In some cases, patients did actively resist treatment following the neurologist's orientation to their right to choose. In these, the same pattern shown in Extract 11 was also typically evident: an assertion containing a generic reference to treatment (see →I), minimally receipted; a subsequent orientation to the patient's right to choose whether to undergo treatment (see →II); a go-ahead or blocking response, followed by a recommendation tailored to that response (see lines 51-98). In other words, where the patient's response blocked the option to treat, the subsequent recommendation was typically against treatment; only on two occasions did the neurologist pursue agreement to treat in the face of such resistance.
In this respect, these assertions function similarly to those used in option-listing. Poised somewhere between an informing and a recommendation, the assertions allow for the construction of a wider decision-making trajectory by placing very little interactional pressure on the patient for an immediate acceptance of the just-mentioned treatment. Like in option-listing, this defers the decision point proper until two things have been done: the delivery of recommendationrelevant information and the creation of a slot for the patient, explicitly, to position him/herself for or against treatment. Again, the information delivered by the neurologist is not neutral with respect to that decision. But the trajectory allows the neurologist both to construct her/himself as knowing best about the available treatments and their potential benefit and to construct the patient as knowing best about her/his treatment preferences. And again, as in option-listing, the assertion can be done to help facilitate-for yes/no treatment decisions-a degree of patient choice.

"Cautious" recommendations
This final analytic section examines four cases (15%) that are notably different to the rest in their turn design. In each, the neurologist refers to a single, named treatment as a standalone option, as shown in boldface in Extracts 12-15. In the rest of the neurology assertions, treatments were only named when they were part of a wider multi-option list. Apart from the four below, all the assertions that referred only to yes/no decisions used generic references (e.g. "medication(s)" in Extracts 4, 5, 6 and 7a; "drugs" in Extract 11). The neurologist is arguably more clearly hearable as doing recommending when naming a treatment since generic references do not make clear what, precisely, the patient might accept/resist in next turn. Nevertheless, in three of the four cases shown above (Extracts 12-14), the patients responded minimally or not at all to the assertion, even though all three went on to accept the treatment. In other words, despite the different treatment formulation, these assertions seem-like those in the previous sections -to be treated, typically, as simply doing informing. Again, then, the assertion seems to defer the decision point proper. This can be seen in Extract 16, which shows the wider decision-making trajectory for the case shown in Extract 12. In common with all the cases in this final sub-collection, the decision is only reached after a full, on-record recommendation is made (lines 6-8). Thus, the decision sequence includes the assertion (lines 1-2), which is not receipted vocally (we only have an audio recording so cannot be sure if anything occurred nonvocally), and a recommendation (lines 6-8), which is accepted (line 9).
Extract 16 (UK050208202; Multiple Sclerosis)a continuation of Extract 12 The final case (Extract 17) is atypical in that the patient responds with active resistance. He does so on the grounds that he believes that he is already taking a beta-blocker (or something similar). This is perhaps due to the neurologist's description of it being used for high blood pressure (lines 3-4), a condition for which the patient has already said he is being treated (data not shown). Thus, despite the general construction of the assertion turn ("generally the treatment for that is…", lines 3-5) and the provision of additional information about the typical use of such a tablet (line 3), the patient treats the assertion as conveying a specific recommendation for himself. This is evident in the way his resistance is personalised and specific ("I'm taking simvastatin," lines 6-7). This case thus supports the finding, in primary care, that assertions may be treated as doing recommending (Stivers et al., in press).

Extract 17 (UK050207101; Postural tremors)-a continuation of Extract 15
The neurologist's pursuit of agreement by the patient to take a beta-blocker-across lines 9-19-ratifies the patient's understanding of the assertion as a recommendation. This is evident in the neurologist's treatment of the patient's turn as resistant. First, he counters the patient's resistance by correcting his misclassification of simvastatin as a beta-blocker (lines 9-11), which the patient accepts (line 12). Second, in correcting his own claim that the patient is not already taking a beta-blocker ("well you're on atenolol," line 11), the neurologist produces another assertion, this time naming a specific beta-blocker ("propranolol," line 16) and providing an account for why it might be more helpful than the one the patient is taking (lines 16-19). In so doing, he justifies the treatment on the grounds of its likely benefit to the patient. Finally, the neurologist produces an explicit recommendation in the form of a proposal (lines 21,22), which the patient accepts. The neurologist's response to the patient's resistant turn at lines 6-7 thus exposes the implicit recommendation produced through the assertion at lines 1-4, ratcheting up the interactional pressure for acceptance. The patient's response shifts from nonvocal acknowledgment (line 20) to more explicit acceptance (line 23).
Putting the resistance to one side, we see the same two-step pattern shown in Extract 16: an assertion in which a specific treatment or type of treatment is named (see →I), followed by an overt recommendation (see →II). This occurred in all four cases.
So what's going on in these four? These assertions function, it can be argued, as highly mitigated, or deontically "cautious," recommendations. In that respect, they resemble the yes/no cases shown in the previous section. However, whereas the cautiousness in those cases opened up space for an explicit orientation to the patient's right to choose, in these cases, it opens up space for handling a variety of difficulties with respect to the decision. In Extract 16, the assertion can be heard as persuasive, just as the subsequent question (lines 3-4) may be heard as implying a need for treatment. Thus, before acceptance becomes overtly relevant from the patient, the neurologist has made a case for the treatment. This seems to be predicated on the possibility that this patient may resist steroids. Earlier in the consultation, it became clear that the MS nurse had already recommended steroids, but that the patient has not taken any. Thus, the cautiousness here appears to be handling this indicator of reluctance.
Reasons for a cautious approach to recommending were evident in all four cases. Just prior to Extract 17, the neurologist reports a degree of uncertainty about the diagnosis, to be resolved with further testing. Thus, the generalised assertion ("generally the treatment for that is…", lines 1-2) serves to indicate a treatment option that is uncertain, dependent on the outcome of the tests. In the other two cases, the patients had already tried treatments that had not been effective. Thus, the assertions introduce alternatives against this backdrop of treatment failure. In all cases, then, there is reason for a deontically cautious approach to the action of recommending a particular drug. Again, the assertion's delicate balance between informing and recommending helps to minimise the pressure on the patient for an immediate acceptance, deferring the decision point until further treatment-relevant information provision (and sometimes information gathering) has been done. Unlike in the prior cases, however, the goal here seems to be the "step-wise" building of a case for treatment, rather than the construction of a choice.

Discussion
This paper has shown that assertions, at least in the UK neurology collection, are overwhelmingly receipted as doing nothing more than information provision. This is made possible by their ambiguous design: on the one hand, their relatively depersonalised formats convey that the neurologist is "merely" informing the patient about what's available; on the other, the link made between the available treatment and the patient's condition implies that it will benefit the patient. Thus, assertions, while they stop short of explicitly telling the patient what s/he ought to do, are hearable as recommendation relevant. This delicate balance leaves it up to the patient whether to orient, in response, to the implied action (recommending) or the on-record action (informing). In that regard, assertions function in a similar way to the illness explanations (Gill & Maynard, 2006) and the declarative evaluations (Kushida & Yamakawa, 2015) discussed in the introduction, above.
When treated by the patient as doing "mere" informing, assertions defer the decision point until something more has been done by the neurologist. The analysis presented here has demonstrated three main interactional functions served by this delay. First, in responsive position, they can be used to address patient concerns-indicating that a solution does exist -without actively making a decision relevant next. In this way, they can avoid derailing another on-going activity (e.g. history-taking). This parallels Gill and Maynard's (2006) analysis, which showed that, by avoiding direct questions to the doctor, patients can get candidate explanations on the table without disrupting the doctor's information-gathering activity. Doctors, reciprocally, usually choose not to topicalise the explanations during that activity, allowing them to "stay on course" with respect to the overall organisation of the visit (p. 117).
Second, this paper has shown that assertions can be used to construct both multi-option and yes/no choices for the patient. For the former, assertions may be used to announce that there are multiple options and to introduce one or more of the options from which the patient might choose. Similarly, for yes/no decisions, they can defer the decision point until two things have been done: the delivery of recommendation-relevant information and the creation of a slot for the patient, explicitly, to position him/herself for or against treatment. Typically, the neurologist then went on to align with the patient's view in how s/he designed the recommendation proper. However, it should be emphasised that, across this second subset, the extent to which the alternatives were equally endorsed by the neurologist varied considerably. These assertions were not, then, typically used to offer entirely "open" decisions for patients; they did all, however, facilitate a clear orientation by the neurologists to the patients' right to choose.
Third, this paper showed that assertions can be used to construct a "step-wise" approach to the recommendation, where the assertion is followed by a more explicit recommending turn. This seems to handle various reasons for caution, including diagnostic uncertainty, the likelihood of patient resistance, and uncertain efficacy of the proffered drug. These cases are akin to the declarative evaluations discussed by Kushida and Yamakawa (2015), who showed that the decision proper may be delayed if the proposal either has an obscure relationship to the patient's stated concern or is based on a diagnosis that differs from the patient's understanding of their condition.
Although the sample size is too small to be confident about the generalisability of the patterns identified here, it is intriguing to consider the possibility that whether or not a treatment is named in an assertion may be critical to whether the neurologist is hearably initiating a recommendation trajectory or an opportunity for patient choice. All the initiating assertions (i.e. excluding those that were responsive to patient concerns) showed the same correlation between reference form and subsequent action: when a specific treatment or type of treatment was named (see →I in Extracts 16 and 17), then the neurologist went on to produce an overt recommendation (see →II); when treatment was referred to generically (see →I in Extract 11), the neurologist went on to orient to the patient's right to choose (see →II). This implies that assertions-described here as poised between recommending and informing-may lie along a continuum, ranging from most like a recommendation to most like "mere" information provision. The analysis presented here suggests that the reference form can push the assertion more towards one or the other end of the continuum, but it seems likely that other features of the turn design will play a role too. This warrants further research.
In all the neurology cases, assertions played both an epistemic (see Heritage, 2012aHeritage, , 2012b and deontic (see Stevanovic & Peräkylä, 2012) role: they enabled the delivery of treatmentrelevant information while placing minimal interactional pressure on the patient for an active response. In that regard, they are deontically mitigated relative to the other action forms discussed in Stivers et al. (in press). Nevertheless, depending on how the information is delivered, they can be done in the service of seeking-in various ways-to persuade the patient of the benefits of treatment, in advance of the recommendation proper. Comparable persuasive moves have been demonstrated by previous studies of how treatment recommendations may be "pre-figured." For example, Silverman (1981) showed how, despite an apparent emphasis on parents' preferences, paediatric cardiology consultations for children with Down's syndrome almost always ended in a decision that accorded with clinic policy (not to treat). He showed how this was subtly achieved across the course of the consultation, through a series of moves that laid the groundwork for non-intervention (see also . Similarly, in their analysis of a "pressured decision" in psychiatry, Quirk, Chaplin, Lelliott, and Seale (2011, p. 99) showed how the psychiatrist pushed for an acceptance of medication through a series of preliminary questions. It is notable that these questions were preceded by an informing-"You know olanzapine isn't the only drug?" (line 1, p. 100)-which, although not done as an assertion, is comparable in that it introduces the possibility of another treatment without expressly recommending it. A series of papers investigating surgeons' recommendations for and against surgery likewise showed that surgeons may work to "'get the patient on board'" with their preferred course of action before delivering the recommendation proper Raymond, 2011, p. 1036, andsee;.
A similar argument could be made for some of the cases included in our sample-most clearly those in which a specific treatment is named in the assertion turn. Like Barnes (in press), this paper thus contributes to an understanding of the work clinicians may do-in advance of the recommendation proper-to try to secure acceptance or reduce resistance. However, previous studies have seldom considered the ways in which explicit moments of choice may be generated for patients. In the neurology data, the majority of assertions were used to initiate a decision-making trajectory that included just such a moment. Although information provision prior to a moment of choice is rarely (if ever) neutral-and thus can also be hearably persuasive-information is a prerequisite for valid decision-making. Thus, the informing done through assertions can also be understood as a means of providing patients with (some of) the epistemic resources needed to make a choice. Without such resources, patients demonstrably struggle to choose, as we have shown elsewhere (Reuber et al., 2015). Assertions can, then, handle an interactional dilemma: how to reduce the epistemic gradient between clinician and patient (Heritage, 2012a(Heritage, , 2012b while avoiding directly telling the patient what to do. This study is clearly limited both with respect to sample size and the focus only on neurology assertions. Further research should test these findings in larger datasets and across other settings. Nevertheless, the analysis presented here suggests two likely reasons for the large number of assertions in neurology compared with primary care. First, neurologists may be more likely to be prescribing treatments with which patients could not be assumed to be familiar (unlike, say, antibiotics in primary care in the UK and US). Thus, information provision as a basis for decision-making may be more commonly deemed to be necessary in neurology. Second, there is a strong emphasis on patient choice in the policy literature for neurology within the National Health Service in the UK (see Department of Health, 2000, 2005. Given that assertions were most commonly part of decision-making trajectories oriented to patient choice, this emphasis in UK neurology may partly account for their prevalence in our sample. Why this should not, similarly, be the case in UK psychiatry is not yet known. However, given the finding that assertions were generally treated as recommendations in the primary care datasets, it seems likely that they are not being used (or not often being used) to generate a slot for patient choice in primary care. Assertions may, then, be one of the ways in which neurologists are seeking to enact the call to enable patients with chronic conditions to "become key decision-makers in the treatment process" (Department of Health, 2000, p. 5).