“I think” in political speech

Simon-Vandenbergen (2000: 61) concluded her study of I think in political discourse by noting the importance of further study of its prosodic realisation. Consequently, I investigate the prosodic realisations of I think in political debates. At the same time, I examine the lexico-grammatical form of the construction, and its surrounding co-text. My exploration confirmed that I think is frequent in political speech, and revealed that it projected four types of meanings. Three of the meanings occurred irrespective of the intonational choices, though prosody influenced the likelihood of the occurrence of a particular meaning. There was a greater likelihood of the speaker expressing a tentative statement if think was prominent/tonic. Intonational prominence on I explicitly warranted the source of the evaluation. When the construction did not contain an intonational prominence it tended to signal commitment to a proposition, or if followed by a filled pause or rhythmic disjunction a hesitation marker.

1 Outline I think and related constructions such as I guess, I believe and I know have been extensively studied from a number of angles.These are from a semantic viewpoint (e.g.Urmson, 1952;Thompson and Mulac, 1991a;Simon-Vandenbergen, 2000), according to their syntax (e.g.Thompson and Mulac, 1991b;Quirk, Greenbaum, Leech and Svartik, 1985;Halliday and Matthiessen, 2014), according to their use (e.g.Boye and Harder, 2007;Kearns, 2007), and from a historical perspective (e.g.Hooper, 1975;Brinton, 2008).Yet, none of the studies men- tioned above have considered the role of prosody.Studies, which have incorporated prosody, such as Dehé and Wichmann (2010 a and b), Dehé (2014) and Kaltenböck (2009) are much rarer.In this article I will illustrate the contribution of prosody in disambiguating the meaning of individual I think tokens in televised political debate.
In section 2, I critically review previous studies of I think and related constructions in order to situate my study in the wider literature.Then in section 3, I set out the intonation framework used in this chapter, and relate it to previous prosodic investigations of I think especially those situated in the field of politics.In section 4, I describe the political corpus and illustrate how the I think tokens were coded intonationally.Section 5 discusses the results, while in section 6, I summarise the findings.

2
I think as complex discourse marker Urmson (1952: 495), one of the earliest studies of I think, argued that a class of mental processes, which he dubbed parenthetical, and which are used in the first person in indicative mood in the present simple tense, "functioned as signals to guide the hearer to a proper appreciation of the statement in its context".These mental processes were not part of the truth conditions of the statement, but rather signalled its reliability.In a similar manner, Halliday and Matthiessen (2014: 693) label the construction I think an incongruent (metaphorical) modalization with an explicit subjective source.For them the congruent means of signalling the speakers' view of the probability of a statement being true is to signal it within the clause through the use of an interpersonal adjunct such as probably or a modal verb such as will.They classify I think as realising a median level of probability, thus examples (1a) to (1d) are ideationally synonymous.1 (1a) I think Mary was at the party.(Subjective-explicit) (1b) Mary will have been at the party.(Subjective-implicit) (1c) Mary was probably at the party.(Objective-implicit) (1d) It is likely that Mary was at the party.(Objective-explicit) 1 Similar views are expressed by Aijmer (1997) and Van Bogaert (2010: 403), though these scholars argue that speakers may at times use I think to realize a stronger sense of commitment than can be glossed by probably, e.g.example (26) this paper.Yet, ( 1a) is open to two possible readings.The first where the construction I think is an epistemic comment clause; the second where I think is a projecting mental clause which projects the second relational clause.Thompson andMulac (1991a/1991b) argue that the presence of the complementizer that cues the hearers' analysis of constructions such as I think.Its absence signals that the mental projecting clause is to be reanalysed as an epistemic comment clause.Kearns (2007: 483) rejects this argument and states that I think can only be identified as a projecting mental process if the mental verb plus subject determines the form of the corresponding tag question, contrast examples (2a) and (2b).Van Bogaert (2010) reports two further tests for distinguishing mental projecting clauses from epistemic comment clauses.These are transparency to negation and transparency to factive sentence adverbials, e.g. ( 2c) to (2f).However, without recourse to contextual information it is hard to see how any such tests can lead to a definitive claim of the status of the I think construction in ( 1a).Yet while it seems clear that (2a), ( 2c) and (2e) are the more natural examples it is possible to create contexts where the opposite is the case.2(2a) I think (that) she was at the party, wasn't she? = Epistemic comment clause (2b) I think (that) she was at the party, don't I? = Mental projecting clause (2c) I don't think she was at the party = I think she wasn't at the party.(2d) I don't think she was at the party = It's not that I think she was at the party.
I know she was.(2e) Unfortunately I think she was at the party = I think unfortunately she was at the party.(2f) Unfortunately I think she was at the party = It is unfortunate that I think she was at the party.I wish I didn't think so! Boye and Harder (2007) point out the hybrid nature of I think and related constructions which they label "complement-taking predicates" (CTPs),3 They along with Kearns (2007) and Van Bogaert (2010) argue that a functional reanal- ysis of I think as an epistemic comment clause does not necessarily entail a structural reanalysis.I think needs to be described not only in terms of its structure, but also in terms of its use.Boye and Harder (2007: 590)  Boye and Harder (ibid:591) suggest that I think has "only just reached position C".This results in some tokens of I think at stage A remaining as mental projecting clauses, or in their original terms main clauses with primary status, e.g.I think she is wonderful don't I?.Others at stage B are hybrid in the sense that structurally the I think construction remains a mental projecting clause but has secondary status, e.g. the appropriate response to I think that she is wonderful is yes she is/no she isn't and not do you.At stage C I think has secondary status and modifies the proposition she is wonderful, e.g.I think she is wonderful isn't she?Unlike Thompson and Mulac (1991a/b) the authors, discussed above, do not claim that there is a single structural marker, such as the complementizer that, which distinguishes between the mental projecting I think clause and the epistemic comment clause I think.Rather they argue that the probes, see examples (2a) to (2f), they use to determine the usage and structural status of I think are context dependent.Yet, despite the fact that speakers produce prosodically appropriate utterances and these prosodic choices project the communicative status of utterances and the informational status of lexical elements, none of the authors, reviewed above, has examined prosody while attempting to disambiguate the various meaning signalled by the use of I think in context.

I think in political speech
Studies such as Aijmer (1997) report that I think is typical of informal conversation where it is mostly used to signal uncertainty or tentativeness.Yet, Simon-Vanderbergen (2000: 47) found that I think appears to be more than twice as common in political interviews than in informal conversation.She (1997 and2000) found, that in interviews, politicians frequently employ I think constructions not only to project hedging or a lack of commitment to their utterances, see Jucker (1986), but also to project their commitment to the truth value of their statements.She considered this unsurprising in that political interviews are specifically focused around the opinions of the political interlocutors.Fetzer (2011 and2014) noted that in political discourse the co-occurrence of I think with different pragmatic markers determines its communicative function as either strengthening or weakening the illocutionary force of the argument.
In informal conversation the BNC reports that 3 out of every 4 instances of I think occur in clause initial position while in her corpus of radio political interviews initial I Think occurred more than 9 out of every 10 times (Simon-Vandenbergen 2000: 48).This indicates that clause initial position is the unmarked location for I think especially in political speech.In the data studied here initial I think similarly made up the overwhelming majority of the I think tokens found, see section 3. Halliday and Matthiessen (2014: 109) have noted that if speakers wish to project their attitude towards a proposition they typically do it as part of the Theme.Therefore, I think tokens in initial position are more informationally salient than those found elsewhere.As a result, in this paper I will examine only instances of I Think found in initial position.

3
The intonational framework Prior to describing the intonational realisation of I think, it is first necessary to briefly sketch the intonational framework, British School intonation, used to transcribe and categorise the politicians' speech.Speech is articulated as a series of tone units which themselves consist of a mandatory tonic syllable or nucleus and optional prominent and non-prominent syllables.The tonic syllable is identified by being the most prominent syllable within the tone unit, and by being the locus of the major tone movement in the tone unit.
There are five primary tone movements fall (\), rise (/), fall-rise (\/), level (-) and rise-fall (/\).There may be optional prominent syllables prior to the Tonic.The initial prominence is known as the onset syllable and it represents the beginning of the head which continues until the tonic.The onset syllable may be pitched as high, mid or low relative to the prior onset.contain non-prominent syllables.If present non-prominent syllables prior to the onset are known as the prehead.Syllables after the tonic are known as the tail.The internal structure of a tone unit is set out in example (4), with optional elements in italics.Prominence, both tonic and pretonic, usually involves a degree of pitch change and often involves a local maximum or minimum pitch height, occurs on a rhythmic beat, and is usually accompanied by increased duration and loudness.For further information see Cruttenden (1997: 13), Crystal (1969: 207-210), Brazil (1997:14), Halliday and Greaves (2008: 211) and Ladd (2008: 48).The intonation transcription conventions are found in the Appendix.
(4) with SOME of the ap PAL ing things onset Prehead Head Tonic Tail A clause initial I think construction can be articulated as follows: (5) a. as an independent tone unit; (i) with I tonic and think in the tail; (ii) with think tonic and I as prehead.b. as part of the prehead of a tone unit.c. as part of the head of the tone unit; (i) with think as onset and I as prehead; (ii) with I as onset and think as a non-prominent syllable in the head.Kaltenböck (2007 and2009) A speaker can produce the I think construction (a) in an independent tone unit, (b) in the prehead or (c) in the head.If the construction is in an independent tone unit or in the head the speaker make either element prominent.Finally the speaker may reduce the intonational prominence of I think by producing it in the prehead, as a stream of unaccented syllables; see example (8) below, where the token was pronounced as [aɪθɪŋʔ].Such unaccented syllables may or may not be rhythmically integrated with the remainder of the tone unit.In all cases the tone unit containing the I think construction may be realised with any of the five primary tones.The options are detailed in Figure 1.Now, that we have seen the forms which articulations of I think may realise, it is time to explore the communicative significance of possible selections made by speakers.Numerous scholars e.g.Chafe (1994), Cruttenden (1997) and Halliday and Greaves (2008) have noted that speakers segment speech into units containing only one single idea, or piece of information, which is expressed in what are variously known as intonation groups, intonation units or tone units.If a speaker separates the I think construction off from the surrounding discourse by placing it into its own tone unit he/she focuses attention on it as a piece of information worthy of consideration in its own right.While there has been a long tradition of classifying the spoken information structure of West-Germanic languages, such as English, as a binary contrast signalled by the presence of a prominent syllable, e.g.(Bolinger, 1972;Pierrehumbert and Hirschberg. 1990) etc., this is not the entire story.Speakers may make syllables prominent in order to have an independent onset and not because they wish to signal informational salience, see (Brazil, 1997).Speakers' tone choices4 project the content of a tone unit as either a piece of major information which represents an act of telling or as a piece of incomplete or minor information (Brazil, 1997;Gussenhoven, 2004;Tench, 1996 etc.) Speakers signal major information through the selection of end falling tone, while end-rising tones that precede falling tones signal incomplete information.Minor information is signalled by end-rising tones that follow falling tones.In addition speakers have the option of withdrawing from the immediate communicative context by producing a tone unit with level tone (Brazil, 1997;Tench, 1997).
o'grady 10. 1163/18773109-00901006 | International Review of Pragmatics (2017) 1-35 Speakers may pitch the initial prominent syllable (onset) of their tone unit as either high, mid or low key relative to the height of the onset in the previous tone unit (Brazil, 1997).5A high key signals that the following stretch of speech contains information, which the speaker projects, the hearer will find contrary to the previously created expectations.The selection of mid key carries no such contrastive overtones.A high key immediately following a low pitch and an extended pause signals a shift in topic (Wichmann, 2000), while a high key not immediately preceded by a low pitch or a pause signals an unexpected transition within a continuing topic (O'Grady, 2013: 130).
To summarise when producing initial I think speakers may choose to place it in its own tone unit.If they do, either word may be tonic, and it may have a high, mid or low onset.Speakers may reduce the intonational prominence of I think by producing it in the onset.The onset may be pitched as high, mid or low.Alternatively speakers may further reduce the intonational prominence of I think by deaccenting the construction and placing it in the prehead.

3.1
Previous prosodic studies of initial 'I think' Kaltenböck (2009: 60) compared the prosodic realisation of I think and I think that.He found that there appeared to be, "no fundamental difference in usage" between the two constructions in initial position.The construction I think, regardless of whether or not it was followed by the that complementizer, "generally has secondary status as a qualifier of the proposition in the following clause rather than being the main assertion itself", (:67).He rejected the predictions made by cognitive-functional accounts, such as Langacker (1991: 436-438), that intonational prominence is the key to disambiguate between mental projecting clause I think status and epistemic clause status I think.Kaltenböck (2007: 6) claimed that the placement of initial I think, in the absence of the complementizer that, into the separate tone unit indicated epistemic clause status.However, in light of his 2009 finding that that tends to be rhythmically integrated or chunked with I think rather than with the following material he would presumably consider that the placement of I think that into an independent tone unit similarly indicates its epistemic clause status.Intuitively this is appealing, as clause initial adjuncts are frequently found within independent tone units (Cruttenden, 1997: 69).Pragmatics (2017) 1-35 | 10.1163/18773109-00901006In their careful corpus investigation of clause initial I think (that)6 Dehé and Wichmann (2010a) examine which element the I or think is prominent and conclude that the most important feature disambiguating the uses of I think is prominence.A prominent I whether in the head or tonic indicates a mental projecting clause while a prominent think indicates an epistemic comment clause.Non-prominence indicates a discourse marker.This suggestion has the benefit of being in accord with the intonational literature, which notes that prominence reflects the intonational salience of a lexical item in context.Furthermore it accords well with Boye and Harder's test of addressability.For example, ( 6) with prominent syllables underlined, can potentially be responded to by the question Do you with tonic prominence on the you.

International Review of
(6) I think they should be converted into the homes that people need for young families like yours?
Furthermore, the context of the utterance, a challenge to the speaker to state what he would do with the numerous unused commercial buildings in British cities, does not call for a tentative response.In other words, in example (6) I think is not commutable with probably.This view entails that what appears to be a structurally hybrid construction is not, once intonation is taken into account.

Data and method
The corpus studied here is the series of three televised leaders' debates held prior to the UK general election in 2010.The debates were between Gordon Brown, the then Labour Prime Minister, David Cameron the leader of the main opposition Conservative party and Nick Clegg the leader of the smaller Liberal Democrat party.In total the 3 leaders produced 50,236 words across the three debates.7 Table 1  sentence initial position, or immediately following a conjunction which linked two paratactic clause nexuses were classified as initial.Four instances of initial I think in debate 1 had not been transcribed in the orthographic transcription but are counted here.Overall, 82.2% of the I think tokens produced by the 3 politicians occurred in initial position.This number falls between the figure of 93% and 74% initial I thinks reported for political interviews and casual conversation by Simon-Vandenbergen (2000: 48).This suggests that the televised political debates are a mix of political interview and unscripted conversation.
Table 2, however, shows that the overall use of initial I think in the debates far exceeds that previously reported for conversation and indeed for two of the speakers exceeded the frequency of initial I think tokens found in political interviews.One speaker, Gordon Brown, produced far fewer tokens of initial I think, and I will suggest a possible explanation for this in section 6. Fetzer (2014: 78) reports that in the speeches and political interviews in her corpus that there was an increase in the frequency of use of I think in the data collected from the period 1997-2003 compared to that collected in 1990.The data reported in is becoming more frequent.Fetzer (2014) argues that the increased use of I think provides support for Fairclough's (1992:204) claim that UK institutional discourse is being increasingly conversationalized.Yet, Simon-Vandenbergen's finding that I think was more common in political interviews than conversations and that the use of I think in political interviews realised a distinct communicative purpose may indicate the opposite.This is an issue I will return to in section 5 once I discussed the lack of internal fixation of the construction, (Van Bogaert, 2010:410), the genre structure of the televised political debates, and the intonational realisation of the construction found in the corpus.
In the interests of ensuring that the data captured all possible uses of initial I think I included variant forms of the construction as detailed below.Van Bogaert (2010) has convincingly argued for the existence of an I think construction with variable wording.Perhaps more controversially I have followed Fetzer (2011:262) in including we think as a variant of I think because leaders of political parties frequently speak on behalf of their parties.Table 3 details the results.
It is clear that as expected the wording I think is the most frequent.It occurs around 85 % of the time.This proved to be the case for all 3 speakers with over 80% of each leader's initial I think being worded as I think.When examining the communicative function of I think the most frequent wording will be considered to be the paradigmatic example of the construction.Accordingly the meaning of the function of variant forms will be analysed as alternations of the paradigmatic construction.
The televised political debates were highly structured events, see O'Grady (2014) for a fuller description.Each debate started and concluded with the 3  politicians making a statement to camera.In between they addressed 8 topics, first by addressing the question itself and then once all the leaders had spoken by replying to the other leaders' answers.If the moderator saw fit there was a further period of additional free debate where the politicians more explicitly addressed their peers' arguments.The politicians concluded each debate by producing their closing remarks.
Table 4 provides details of the percentage of speaking time across the three debates measured in tone units and broken down into stages in terms of the debates' Generic Structure Potential, (Hasan, 1996:53-58;Halliday and Hasan, 1989:63-66).I coded the corpus into tone/information units rather than time in order to quantify the information produced by each speaker at each stage of the debate.
To illustrate in the first debate, Gordon Brown produced more information in the Free debate stage than he did in the other stages.I similarly recorded which stage all initial I thinks occurred in.While the politicians were unaware of the exact wording of the 8 questions they were forewarned as to the theme of the debates.Furthermore, topics were repeated from week to week, Thus, we can predict that the introductory statements, concluding remarks and question responses were likely to have been more scripted than the other stages.We can also predict that as arguments are recycled in the later debates they will be more scripted than the earlier ones.In the following paragraphs, I will examine whether the distribution and prosodic realisation of I thinks in the corpus sheds some light on whether I think is a feature of less scripted speech or whether it occurs equally or more frequently in more scripted speech.The opening statements amounted to 3.45 % of the total number of tone units, see Table 4, but only 0.8% of the I think constructions occurred in the pre-planned introductory statements.The concluding remarks formed 4.3 % of the speakers' contributions but contained 3.5 % of the I think construction.The majority of the speakers' contributions to the debates were found within the Question Response, Response to other leaders and Free Debate stages which amounted to 29.6 %, 28.4% and 34.3 % of the tone units respectively.The percentage of I think constructions in these three stages was 34.9%, 35.3 % and 26.1 % respectively.If the distribution of I thinks had been unaffected by the generic structure of the debates, we would expect to find roughly an equivalent proportion of I thinks to tone unit within each stage.However, as comparison of the pie charts in Figure 3 illustrates this was not exactly the case.The proportion of I thinks, indicated on the left, in the most scripted stages-the introductory statement and the concluding remarks-was lower than the respective proportion of tone units, indicated on the right.Conversely in the least scripted stage the proportion of I thinks was greater than the proportion of tone units.
A chi square test was conducted and found a significant difference (p = 0.02429) between the numbers of I thinks and tone units found proportionally within each stage.This evidence provides some support for the claim that I think is a feature of spontaneous rather than scripted speech e.g.(Aijmer 1997).Further the decline in the frequency of I thinks across the three debates suggests that the recycling of topics across the debates resulted in the later debates being more prepared and practiced than the earlier ones.The most common usage of the I think type for all speakers is its realisation within the prehead, with 135 instances of the construction found in that position.There were two different patterns illustrated in the data: for the first see ( 7) where it is integrated rhythmically into the tone unit; for the second, see (8) where it is not.The latter pattern was rare with only seven of the I think preheads showing rhythmic disjunction.
(7) | i think the nhS is a WONderful WONderful \THING l| (DC1-2820) [QR-7]9 In example (7) there are four prominent syllables with the head beginning at the first prominent syllable S. The lack of prominence on I think projects that the construction is projected as informationally non-salient or recoverable.In (8) the construction is not fully integrated into the tone unit.The speaker is clearly searching for the appropriate words.In ( 9), Cameron resets the pitch to mid by pitching the onset syllable think at 192.9hz, a level equivalent to the immediately prior mid onset.In (10) he selects a high onset reset by resetting the onset syllable to 236.9hz, a level perceptually higher than the immediately preceding onset.As the previous tone unit was itself completed by a fall to low the following high onset projects the introduction of a new topic.One of the elements within the construction I think (that) was made tonic by the speakers on 46 occasions.The speakers selected a mid or high tonic.
Mid was the more usual choice being chosen on 29 occasions.The speakers tended to place the tonic syllable on think, which was made tonic 29 times.The speakers selected falling tone on 35 occasions, 26 of which coincided with think as tonic.Out of the 12 occasions that I was tonic, end falling tone was selected on 8 occasions.( 11) illustrates a tone unit with falling tone and a high onset on the tonic syllable think, while (12) illustrates one with rising tone and a mid onset of the tonic syllable I.
International Review of Pragmatics ( 2017 Finally, there were 11 instances of the I do/don't think construction which did not fit into any of the above categories.In all the examples do or don't were either prominent or tonic which meant I was placed in the prehead and think in either the head as a non-prominent syllable or in the tail.Examples ( 13) and ( 14) illustrate: The most common occurrence of I think, as noted previously, was in the prehead which accounted for over half of the realisations of the construction as shown in Figure 12.This is in line with previous findings e.g.Dehé (2014: 188).Dehé and Wichmann (2010a) note that I think occurring in the prehead functions as a discourse marker.It is noticeable however, that the prehead use of I think declines across the debates from 63 in debate 1 to 53 in debate 2 and to 24 in debate 3. Conversely there was no decline in the number of I thinks found either in the head or in an independent tone unit across the debates.This provides some evidence that the political leaders treated the second and third debates less as conversations and more as political speeches.
In order to examine the communicative functions of initial I think in the prehead, head and tonic I manually examined all 254 tokens in context by (a) considering the surrounding co-text, (b) the presence of lexical items signalling obligation, necessity, desirability tentativeness and uncertainty, (c) the presence of non-junctural pauses, hesitation markers and filled pauses, and (d) whether or not the I think was commutable by modal expressions such as probable, see Boye andHarder (2007), Fetzer (2014), Kearns (2007), Kaltenböck (2009), Simon-Vandenbergen (2000), andVan Bogaert (2010) for further information on methods used to disambiguate different uses of I think.

5.1
Prehead There were four uses of I think found in prehead position, two of which have been previously described by Simon-Vandenbergen (2000) in her investigation of I think in political interviews.The first use indicates the strength of the speakers' commitment to the proposition.It is signalled by the co-presence of clauses containing boosting lexis.47.9% of the occurrences of initial I think in the prehead co-occurred with the presence of a clause containing boosting lexis such as the intensifying adverb really and the adjective big in example (15a)-see also examples ( 7) and ( 9) above.Cameron projects his strong commitment to the desirability of increased rail transport in example (15a) while employing I think to allow him the option of distancing himself from his assessment of the necessity of increased rail travel.It is noteworthy that in (15a) as well as (15b) and (15c) that the speaker not only makes the intensifying lexis intonationally salient projecting it as informationally new, but also that he produces a falling tone signalling that the proposition is an act of telling.
Nick Clegg in (15b) selects the adjective important to project his commitment to creative teaching.The I think construction as in (15b) does not signal an International Review of Pragmatics ( 2017 The second main pattern which was found on 25 % of occasions involved the co-presence of a clause signalling obligation, desirability or necessity e.g.The co-presence of the clause expressing obligation or necessity projects that the speaker is qualifying a proposal by signalling his evaluation of its desirability or advisability (Simon-Vanderbergen, 2000:53).The lexical item projecting the obligation, desirability or necessity of the proposed action was usually intonationally prominent but on occasions speakers downplayed its salience.For instance, in (16c) the modal should is non-prominent, projecting Cameron's view that the necessity of high-speed rail is presupposed.As in examples (15a-15c) the modal probably cannot be commuted for I think.Unlike examples (15a-c) and (16a-c) these examples project genuine uncertainty or tentativeness as can be seen from the fact that the modal probably is commutable with I think.For instance Brown could have rephrased his utterance by saying that It is probable Nick agrees with me.Clegg is unsure of where a previous questioner was from, and Cameron of whether or not the public is beginning to get angry.It is noticeable that each example contains a final end-rising tone which projects that the speaker is signalling incomplete information, which he or another speaker will subsequently clarify.
The final pattern, identified on 5.7 % of occasions, was where the speaker produced I think as he was struggling to put his message together on the fly.This pattern was most prevalent in the speech of Nick Clegg who accounted for 87.5 % of the examples, and is exemplified by (18)-see also ( 8 To conclude the discussion of initial I think in the prehead does not realise a single communicative function.Rather the combination of lexis and prosody in context creates redundant information (see Fetzer, 2011) which allows the hearer to unpack the speakers' intended meaning.

Head
The second most common occurrence of I think located in the political debates was in the head where it accounted for 25.1 % of the realizations of the construction.There were four different prosodic articulations of I think: a mid onset prominence on think, a high onset prominence on think, a mid onset prominence on I and a high onset prominence on I. Examination of the corpus revealed that 57.8% of high onset thinks and 80% of high onset Is were in phonological paragraph initial position.Speakers signal the start of a new discourse topic through the presence of a high onset immediately following a drop to low pitch in the previous tone unit, see O'Grady (2013), Tench (1996) and Wichmann (2000)  Three of the four types of meaning identified above were found when the construction was in the head.On 38.6% of occasions the speaker signalled strong definite commitment to the proposition, on 25 % of occasions their stance towards a proposal and on the remaining 36.4% of occasions their tentativeness.Thus, the selection of a prominence on think seems to result in a usage which retains more of its propositional value as articulating a median explicit subjective probability.It is not possible to commute the modal probable or the expression it is probable without altering speaker meaning.
Examples (20-22) illustrate.In (20) Clegg overtly proclaims his evaluation that the provision of tax breaks for the rich is wrong.By selecting think as prominent rather than I he emphasises the strength of his assessment.Cameron in ( 21) projects his evaluation that a change in policy is required in order to reduce the deficit.The prominence on think and the lack of prominence on need project a context where the high obligation to change policy is presented as a matter of common understanding, but his own assessment of the strength of his commitment is emphasised.Yet, while emphasising their opinion the politicians have introduced a degree of uncertainty and deniability into their claims.In example ( 22) Cameron projects his claim that a person was climate minister on two occasions as tentative.
The most common meaning realized by a prominent I affirmed the speaker's evaluation of the necessity and desirability of an action, and occurred on 66.6 % of occasions.The speaker contrasts his opinion with that of his political rivals e.g.Cameron projects that that it is his (and his party's) opinion that the levy is necessary.The lack of prominence on need projects a context where the high obligation to impose the levy is projected as presupposed common sense.The intonational prominence on I explicitly warranted the source of the evaluation in order, Cameron hopes, to add to its bona fides.The other two strands of meaning are present with strong commitment realized on 20% of occasions, and tentativeness on 13.4% of occasions.In (23b) Clegg projects the uniqueness of his assessment of the importance of his party's innovative immigration policy, and by so doing implicitly criticizes his rivals for their lack of vision.He thinks it is important not they.Clegg's selection of prominence on I in (23c) projects that he and he alone is reasonable in his assessment of the prohibitive cost of replacing the trident nuclear system.To sum up prominence on I projects a distance between the speaker and his rivals by explicitly warranting the source of the evaluation or proposition.

'I think' as tonic
The third most common occurrence of I think was as tonic in a separate tone unit.This accounted for 16.1 % of the realization of the construction.The separation of I think from the reminder of the proposition projected that the speakers and their assessment of the proposition realises a single piece of information separate from the following proposition.Four prosodic patterns were identified.The high tonic on think and I accounted for 34.1 % and 7.3 % of tonic I thinks and functioned chiefly to project unexpected contrasts, but not to signal the introduction of new discourse topics.Three of the four strands of meaning described above were found.The most common meaning realized by a tonic think was to project the tentativeness of the likelihood of a proposition occurring.This meaning was found on 43.4% of occasions.The second most common meaning was a commitment to the expressed proposition which occurred on 36.6 % of occasions.The speaker's evaluation of the advisability/desirability of a proposal was the final meaning identified and it occurred on the remaining 23.3% of occasions.For tonic I the chief meaning expressed is a tentative commitment to a proposition, and an evaluation of the advisability/desirability of a proposal (45.4% each) with the expression of a strong commitment to a proposition making up the remainder of the cases.
By presenting the I think construction as a separate information unit and one which was mostly projected through the co-selection of an end-falling tone as major information, the speakers projected that the construction retains much of its propositional value.For instance, in (24) the hearer, prior to noting, Nick Clegg's claim of agreement must first assess the strength of his claim which is presented as being Clegg's opinion and therefore notionally disputable.The remainder of the turn initial utterance is presented in example (25).(25) Well I think everyone is agreed that if we were to do this again which is Stuart's question we'd need to make sure that we've got the right equipment the right resources.
It is clear that while Clegg presents his proposition as notionally disputable the presence of the evaluative adjective right on two occasions and the deontic modal need functions to contract the room for dispute.In (26) conversely the strength of Clegg's claim is increased, or so the speaker hopes, by being warranted through his standing as a politician. (

'I think' as other
The other category accounted for the remaining 11 initial I thinks.On 6 occasions don't was prominent or tonic with do prominent or tonic the remaining 5 times.By projecting the polarity contrast between do and don't as informationally new the speaker emphasises the significance of his polarity choice.
In (27) Cameron emphasises the strength of his positive commitment to his proposition that there has been too much immigration to the UK under the Labour government by emphasising the positive polarity of his assertion.He further emphasises his commitment to the truth of his assertion by his coselection of rise-falling tone, and by the fact that the construction is placed into its own information unit.In (28) he projects his assessment of the negative impact of a hung parliament, while simultaneously allowing himself to mitigate the force of his assertion by leaving himself room to argue that the truth value of his assertion was merely probable.To conclude, some evidence has been presented supporting the claim that I think, (Aijmer, 1997) is more typically found in less scripted rather than more scripted speech.Yet, support has also been found extending the view expressed by Simon-Vandenbergen (2000) that the meanings realised by initial I think in political interviews are also found in televised pre-electoral debate and distinct from those found in conversation.Unsurprisingly no evidence was found for a one-to-one relationship between the three grammatical structures, main clause, comment clause and discourse marker, identified by Dehé and Wichmann (2010) and the 4 meanings realised by the use of initial I think in the leaders' debate.
The presence of co-occurring lexis and prosodic choice indicated which of the four meanings occurred regardless of whether I think was non-prominent in the prehead, prominent in the head or produced in a separate tone unit.
It was noted, above, that one speaker Gordon Brown produced far fewer I thinks in initial position than did his two rivals.A possible reason for this was that he, as Prime Minister, had less scope to indicate the desirability or necessity of future actions.I conducted a manual search of the corpus and found that Brown's use of deontic modal verbs, when discussing the future, was far less frequent than the other two speakers.Simply put his discourse centred on what his government was doing, while the other speakers focused on what should/ought to/needed to be done once they were in power.
Initial I think realised four types of meaning.The first of which is that in conjunction with boosting lexis it projected the speaker's commitment to a proposition by allowing him to seem confident and assured.Simultaneously the I think construction created, at least notionally, room for a hearer to dispute the speaker's opinion.The second meaning occurred in the absence of boosting lexis where the I think construction signalled uncertainty or tentativeness.The third meaning occurred in the co-presence of modal verbs and evaluative lexis.The leaders used the construction to signal their evaluation of proposals and to signal what in their opinion was desirable or advisable.However, as example (28) indicated the meanings created by initial I think were defeasible by the surrounding co-text.The fourth meaning was indicated by the co-occurrence of I think with hesitation markers and filled pauses.In these examples the initial I think signalled that it functioned as a pragmatic placeholder designed to provide the speaker with more planning time to assemble his message.
While the unmarked position for I think was in the prehead, speakers had the option of making either I or think prominent.By choosing to make think prominent the leaders projected it as informationally salient.The construction with think as onset realised three potential meanings depending on the co-text.By selecting I as onset the speakers personalised their commitment and distanced themselves from their political rivals.By placing I think into a separate unit, the leaders treated it as a piece of independent information, backgrounded or fore- grounded by the co-selection of tone.Without consideration of the strength of the claim or the identity of the author of the claim the hearers were not in a position to judge the validity of the proposition or evaluate the necessity of a proposal.
Three of the four types of meaning: commitment to a proposition, evaluation of the advisability/desirability of a proposal and tentativeness were present regardless of how the construction was intoned.However, there was greater likelihood of the speaker expressing a tentative statement if think was prominent or tonic.The speakers tended to make I prominent or tonic when projecting their individual evaluation of the desirability/advisability of a proposal.To conclude the intonational realisation of the I think construction did not by itself lead to the creation of new independent meaning but rather it signalled how the speaker intended the hearer to understand his commitment to a proposition or his evaluation of a proposal.
combine a functional and structural reanalysis of CTPs to produce the model illustrated in example (3).The bold terms represent grammatical terms and the non-bold terms represent usage terms.The illustrative clause I think she is wonderful and the glosses are my made up examples.(3) A: Lexical and Primary CTP [I think] = [Mental projecting clause and primary status].[She is wonderful] = [Projected clause].Usage reanalysis B: Lexical and secondary CTP [I think] = [Mental projecting clause and secondary status].[She is wonderful] = [Projected clause and primary status].Structural reanalysis: grammaticalization of CTP C: Grammatical and secondary CTP [I think] = [Epistemic comment clause and secondary status].[She is wonderful] = [Main clause].

Figure 2
Figure2shows the number of I thinks found within each of the stages set out earlier in Table4.See Appendix 2 for a breakdown by individual speaker.
figure 3 A comparison of the % of I Thinks and Tone Units produced in each stage Table 5 details the leaders' intonational realisation of the 254 initial I think tokens in the debates.I used Praat (Boersma and Weenink, 2013) to, visualise the pitch curve and waveform, and to produce the spectrographs reproduced below.The most common usage of the I think type for all speakers is its realisation within the prehead, with 135 instances of the construction found in that position.There were two different patterns illustrated in the data: for the first see (7) where it is integrated rhythmically into the tone unit; for the second, see (8) where it is not.The latter pattern was rare with only seven of the I think preheads showing rhythmic disjunction.

(
13) | but i DON'T think that /MEANS | you should hSTOP someone … VIsiting our \COUNtry | (GB2-1479) [RL-4] (14) | But i /\DO think | it's /GOT | OUT of \/conTROL| (DC1-401) [FD-1]Now that I have illustrated the form of the prosodic realisation of I think found in the televised political debates, I will in the next section investigate the communicative effects of the I think construction in the debates.
figure 12 The prosodic realisation of I Think assessment of median probability, but rather functions to allow Clegg to signal his commitment while simultaneously keeping the option open of distancing himself from his claim.In (15c), Gordon Brown selects the intensifying adverb in addition to the adjective to achieve a similar communicative function.In all examples (15a-15c) the speakers could not have substituted the modal adjunct probably for I think.Instead the construction signals a meaning akin to the expression in my opinion.(15a) | i think it would be a REAlly BIG step \FORwardl | (DC2-981) [QR-3] (15b) | i think hcreaTIVity is imPORtant in the \CLASSroom | (NC1-1628)[R/OL-4] (15c) | and i think that's inCREDibly \imPORtantl | (GB2-908)

(
28) | i DON'T think a hung \PARliament | will be lGOOD for l\BRItainl | /18773109-00901006 | International Review of Pragmatics (2017) 1-35 details the number of I think tokens produced by the politicians and indicates whether they occurred in initial or non-initial position.The official transcripts were examined in order to see if a particular instance of I think was initial.8Instances of I think which were found either in The distribution of I think produced by the politicians o'grady 10.1163/18773109-00901006 | International Review of Pragmatics (2017) 1-35table 1

table 3
Table 2 provides some indirect evidence that the use of I think The variants of I think found in the debates International Review ofPragmatics (2017) 1-35 | 10.1163/18773109-00901006

table 4
The percentage of tone units in the stages of the debates for further details.Example (19) illustrates David Cameron signalling an overt contrast between what he claims the existing government has done and what needs to be done.