The Appendix of Comment according to Language into Act Theory A corpus-based research on Italian spoken data

The article deals with the détachement instances, an aspect of spoken language differing from the binary structure (Topic-Comment) considered to both semantically and informationally form the basic unit of spoken language. According to Language into Act Theory, détachment instances are considered specific information units called Appendix of Comment (APC), with a clear distinction from the Topic unit. The APC may be formally identified in the corpus through its distribution after the Comment and its prosodic performance via a suffix unit. The APC records a frequency value of 4.28% of reference units, which is significantly lower than that of the Topic (close to 20%). The morpho-syntactic fillings of the APC show a kind of “randomness”, that cannot truly be generalized, unlike Topic, since they are employed “in the moment”, as late adjunctions, echoes, repetitions, deictics, and formulas. APC doesn’t constitute a syntactic/semantic island, as Topic does, and its content is ultimately irrelevant, lacks an independent modality and is de-focused. The APC’s features confirm the specificity of its information function in order to textually integrate with the Comment, facilitating an agreement with the addressee, while the Topic serves as the field of application for the illocutionary force. The purpose of the article is to explain the distinction between the previous information functions by demonstrating the specific role of Appendix through examples derived from Italian corpus analysis.


Instances of détachement
This article deals with an aspect of spoken language that has been noted since it first began to be analysed in the literature and was perhaps first referred to by Vendryes in 1921 as détachement. It differs from the binary structure considered to both semantically and informationally form the basic reference unit of spoken language (variously referred to as the sentence, énoncé, utterance), and which has been developed over the course of the previous two centuries. Said structure consists of combinations referred to via various terminologies and according to different theoretical perspectives over time: point de depart-but du discours (Weill 1844), psychological subject-psychological predicate (Gabelenz 1891), thème-propos (Bally 1950), theme-rheme (Prague School, Sornicola & Svoboda 1991), given-new (Halliday 1976), topic-comment (Hockett 1958;Chafe 1994), prenoyau-noyau (Blanche-Benveniste 1997, topic-focus (Krifka 2012). However, tertium datur, that is instances of détachement, which go beyond this binary structure. The purpose of the article is to explain the clear functional distinction between instances occurring as first components in the binary structure, say Topic, and the detached ones, say Appendixes 1 . In the literature, different types of instances have been considered belong to the détachement domain, in the form of linguistic elements added to the righthand part of the utterance (in most cases at its conclusion) or after a syntactic VP, demonstrating non-canonical ordering. Indeed, they are usually considered to be outside of the syntactic configuration of the sentence or seemingly to infringe on the rules of word order of a given language 2 . They may take the form of postponed subjects for SVO languages, postverbal objects for SOV languages, lexical saturations of prior clitic pronouns, known as right-dislocation, sentence syntactic completions, adjunctions and corrections, repetitions and echoes, fillers, and formulas (see 3.1). However, there is no agreement on a shared identification of the domain. Language into Act Theory (L-AcT, Cresti 2000) is a framework developed by the LABLITA team for the analysis of spoken language, and within this framework we carried out corpus-driven research on the Italian section of the IPIC database consisting of 20,844 reference units (Panunzi & Gregori 2012), analysing the instance types mentioned above.
Within L-AcT, détachement instances are identified on the basis of their unusual distribution, their prosodic performance, morpho-syntactic and semantic features, justifying the assumption of a specific information function called the Appendix of Comment (APC). Specifically, the article will argue that APC cannot be considered a kind of postponed Topic, as if the latter's semantic content and function were essentially maintained, as often assumed in literature (Lambrecht 1981(Lambrecht , 1994Lombardi-Vallauri 2009Ortega-Santos 2016). We will explain that the difference does not depend only on their evident distribution (Topic before Comment vs. APC after it) and respective prosodic performance, via prefix unit for Topic and suffix unit for APC (see 2.2), but also for their broad difference in morpho-syntactic fulfilment and semantic features as corpus-based research makes emerge (see 3.1 and 4).
After a short summary of the L-AcT framework, in section 2 the paper will present the identification criteria for APCs and numerical data on their occurrences, along with descriptions of their prosodic performances and some distributional quantities. The typologies of the APC morpho-syntactic fillings are detailed in section 3 along with their respective percentages while section 4 deals with the semantic aspects of the APC that justify its functional definition and distinction from Topic.

Language into Act Theory
We will briefly summarise Language into Act Theory, referencing literature that deals with this approach (Cresti & Moneglia 2005, 2018aBossaglia et al. 2020). Within L-AcT, the primary reference unit for the analysis of speech has a pragmatic definition that is tied to a long tradition -that of the utterance -and corresponds to the accomplishment of an illocutionary act (Austin 1962). Even if also the phonetic and prosodic conclusion of the sentence fits with a long tradition (Karcevsky 1921;Crystall 1975), the identification of real utterances in spoken corpus through terminal prosodic breaks is an innovation introduced by L-AcT 3 .
The example (1) shows a dialogic turn which is composed of four utterances, each concluded by prosodic terminal breaks and marked by double slashes and a superscript tag (COM). Transcripts follow the L-AcT format, where prosodic units are marked at their prosodic boundary with their information function tag, and the boundaries are classified as either terminal (//) or non-terminal (/) (Moneglia & Raso 2014). The illocutionary force of each pragmatic unit is reported in the sub-tier denoted "%ill:", according to the tag set in Cresti (2020). The IPIC file ID is found within the square brackets.
(1) Moreover, each utterance corresponds to an information pattern -following L-AcT terminology -which is performed by a prosodic pattern, with a direct correlation between the information units and prosodic units. Each prosodic unit within the prosodic pattern is signalled by a non-terminal prosodic break and contains a perceivable f0 movement. In L-AcT, the core of the information pattern is formed by a Comment information unit which accomplishes the illocutionary force of the utterance allowing its pragmatic interpretation. The information pattern depends on the occurrence of the Comment, which is necessary and sufficient for performing the utterance. For instance, (2) is a turn corresponding to a sequence of simple information patterns, each composed of an only Comment unit accomplishing a specific illocution.
(2) *MAR: chiamo // COM il telefono staccato // COM il cellulare staccato // COM 'I call. The corpus-driven research shows that terminal prosodic breaks can also identify a second reference unit still accomplishing a pragmatic value but going beyond the utterance: the stanza. It is composed of at least two Bound Comments (COB), or sub-patterns of information supporting a COB (Cresti 2000(Cresti , 2010Panunzi & Saccone 2018;Cresti & Moneglia 2020;Saccone 2020Saccone , 2021. The stanza accomplishes specific pragmatic activities such as descriptions, narrations, explanations and instructions. It is performed via a sequence of homogeneous, "weak" illocutionary acts, falling outside of any previous program, through an adjunction process that follows the flow of thought (Chafe 1994 The lowercase letter r in subscript, preceded by a dash, is added at the end of an information function tag, for instance COM-r, when it represents a unit of information that is part of a reported speech. 7 The superscript tag SCA marks a chunk of speech that does not directly develop an information function but still participates in its composition. For instance: all'incirca / SCA otto settimane // COM . This happens in cases where the prosody simply divides into parts -that is, scans -a textual information unit that is too long to be performed as one prosodic unit. Italian IPIC records approximately 90% of utterances (18,810) versus 10% of stanzas (2,034).

Appendix of Comment
Within the set of L-AcT information functions the Appendix of Comment unit belongs to the textual information units, as does the Comment, since they participate in the semantic content of the utterance. However, with respect to the information structure, the APC must be distinguished from the Comment in that it is not responsible for carrying the illocutionary force; though not usually considered in other theoretical frameworks, this distinctive feature is fundamental to our approach. Using the L-AcT methodology, the APC is identified in the corpus through a set of prosodic, distributional, morpho-syntactic and semantic features, allowing to support its functional definition. For the information function of the APC, we must take into account a tradition of studies that have dealt with spoken chunks appearing to be delayed, fetched in memory or later added on, or to simply occur in the right periphery of an utterance. These were often analysed and explained in isolation from one another and thus ultimately lead to differing definitions. As we anticipated, there is no agreement on a shared identification of the domain.
We would like to anticipate our functional definition of APC which we will better explain in the course of the article. In keeping with the semantic perspective that has been termed after-thought explanation, L-AcT agrees that the speaker adds later content to the information pattern via the Appendix. Specifically, Appendices are considered to add unimportant linguistic contents to what has already been expressed with the Comment. In fact, the APC's contents are redundant, even trivial, although it must be emphasised that they are made on purpose by the speaker toward the addressee with the intention of expressing agreement with him. Even if APC's textual integration adds little or no content, within the relational exchange it counts as a conventional expression of agreement with the addressee.

2.2
Prosodic performance of the APC According to the L-AcT framework, an Appendix unit must be realized through a dedicated prosodic unit known as the suffix (Cresti & Firenzuoli 2002Cresti 2012a. The f0 movement of suffix is seen to be flat or gradually descending and doesn't present a nucleus. The movement is simple in terms of its composition, since it does not record any preparation, in contrast to the prefix unit performing a Topic information unit and the root prosodic unit performing the Comment. Its recorded hertz values are lower than those of the Comment and there is found to be a jump followed by a considerable lowering in intensity. The phonetic execution is often inaccurate unlike that of the Topic. Overall, we can say that the suffix unit does not realize a prosodic prominence, but rather follows a defocalisation strategy.
Within a comparison with Topic, several corpus-based works have shown that a specific prosodic unit -prefix -is dedicated to the performance of the Topic The prefix unit is characterized by the presence of a nucleus and in the whole represents a prosodic prominence 9 .
The suffix unit performing the APC -necessarily occurring after the root unit of Comment -is preceded by a non-terminal break, and in approximately 82% of cases concludes the reference unit in which it participates.
In example (10), some men are playing poker, but they are total novices. Player CIC explains that the direction of dealing cards in this game is the opposite of the one started.
(10) *CIC: è così / COM il giro // APC 'it is like this, the direction' %ill: instruction [ifamcv14,4] Although less frequent, APCs may also be followed by other information unitsspecifically, the Parenthesis 1.4% and the Phatic 3.8% -which, in turn, end the utterance; in this case, the APC is concluded by another non-terminal break. We can see this in the f0 tracks from example (11), where a mother gives an order to her child to go and see something after making the same request several times.
(11) *DAN: vai / COM a vedere / APC eh // PHA 'go to see it then' %ill: order [ifamcv15] When the APC concludes the utterance or the stanza, it is performed through the same type of suffix unit. However, an APC may also occur within a stanza, following a Bound Comment and in turn be followed by other information units, often taking in the case the form of a short and ascending tail. Let's look at the f0 tracks in example (12) which corresponds to a stanza of narration 11 . This is composed of a first COB, followed by a short APC, followed by a new information pattern which is linked to the first one by a Parenthesis, followed by a Discourse Connector, and then finally a Comment.
(12) *ELA: io però / TOP me la ricordo / COB la tu' mamma / APC che quindi doveva avere una certa età / PAR ma / DCT fresca // COM 'but I remember her, your mother, that she then had to have already a certain age, but, fresh'  The first COB (but I remember) is followed by the APC (your mother), performed as a short ascending tail. Thus, in our opinion even for the suffix unit we can assume the existence of at least two variants: the most common records a flat or descending movement at the end of the utterance, while the second variant corresponding to a short ascending movement, necessarily occurs within a stanza. However, we lack sufficient research on this aspect.

Distribution of the APC
The APC units that have been identified in the IPIC database 12 , present the following values: of the total reference units (20,844: 18,810 utterances, 2,034 stanzas) the number of APC units is 893, corresponding to an overall frequency the 4.28%. Their occurrence records differing proportions regarding utterances and stanzas: the number of utterances with APCs in them is 780 while for stanzas it's 113, however their respective frequency is 3.52% and 5.55%. This implies that APCs are more commonly used in the composition of stanzas than in that of utterances. APCs may also occur more times, mostly twice, but while utterances with two APCs amounted to 3.8% stanzas with two APCs amounted to 9.7%. Thus, data on the APC, as well as the Topic, the Parenthesis and DCT information units, confirms the presence of a richer information structure in stanzas (Saccone 2021).
A preliminary question about the distribution of APC units concerns the identification of the reference unit to which the APC belongs. The attribution of a particular chunk to the information pattern of a reference unit, in terms of which the Appendix must be evaluated, is not easy. If only the transcription of speech is considered, attribution to a specific information pattern may be uncertain. Listening, even, without appropriate consideration given to the terminal or nonterminal values of prosodic breaks makes it impossible to decide which group of words should be taken together. Consider the following example: (13) *DNA: sicché / DCT io andai alla laurea // COM 'so, I went to my dissertation' %ill: narration If we consider the simple transcription of the first stretch of the acoustic sequence within the third turn, il giorno dopo mi richiamarono a lavorare ('the day after they called me once more for work'), it could be interpreted as only one utterance accomplishing an assertive illocution. This corresponds to a simple information pattern composed of a Comment unit. If, however, the sequence has its terminal 12 IPIC is a database of spoken corpora with text/sound alignment and tagging per information unit types according to the L-AcT framework. IPIC archives a set of spoken resources: the informal section of C-ORAL-ROM Italian (20,844 reference units); the Brazilian Portuguese mini-corpus from C-ORAL-BRAZIL (5,511 ref. un.); the Spanish mini-corpus from C-Or-DiAL (5,765 ref. un.); the American English mini-corpus from the Santa Barbara Corpus by LEEL (3,452 ref. un.). The annotation of information unit types in the Italian IPIC has been achieved by a single expert annotator and has been revised by the LABLITA team on the basis of both perceptual evidence and instrumental control. No interrater agreement has been calculated on this annotation. All examples presented in the article have been verified by the author. The database is freely available online at http://www.lablita.it/app/dbipic/ breaks identified it is observed that it corresponds to two utterances (il giorno dopo / TOP mi richiamarono // COM ) and (a lavorare // COM ). Perceptually we can recognise two utterances that are separated by a terminal break. Each utterance accomplishes an independent illocution (assertion, contrast), interpretable in isolation. The fact that the speaker DNA has been compelled to go back to work the day after his dissertation is considered exceptional, indicating a sense of contrast. However, the dialogue could have occurred in different circumstances. As shown in the artificial example (13a), if the content (for work) had already been mentioned by the speaker, then this later instance of it could simply have been added in a repetitive manner within the first utterance, forming an Appendix after the Comment and as a sort of loose confirmation toward the addressee.
(13a) *DNA: sai / CNT dopo la tesi / TOP sono dovuto tornare subito a lavorare // COM il giorno dopo / TOP mi richiamarono / COM a lavorare // APC 'you know, after my dissertation, I had to come back immediately to work. The next day, they called me back, to work' %ill:

[1] narration; [2] narration
In conclusion, the occurrence of the APC can only be properly recognized within the reference unit in which it occurs, after the Comment that is the core of the information pattern in which the APC participates and performed prosodically through a suffix unit. According to the IPIC distribution data, approximately 82% of APCs not only follow the Comment or the Bound Comment but also end the utterance or stanza in which they occur.
(14) *CLA: e allora / DCT voti in una certa maniera / CMM ti garantisco qua / CMM ti garantisco là / CMM eccetera // APC 'and then, you vote in a certain way, they guarantee you here, they guarantee you there, and so on' %ill: exemplifying list [ifammn02,128] Nonetheless, the APC may be also followed by one or more information units, specifically by a Phatic (3.8%) or a Parenthesis (1.4%): (15) *AGO: tu frequenti anche poco / COM te / APC mi sembra / PAR vero // PHA 'you frequent even a little, you, it seems to me, don't you' %ill: assertion of evidence [ipuddl03,116] (16) *GIA: loro / TOP con tutti / COM provano / APC capito // PHA 'they, with everyone, try, do you understand?' %ill: contrast [ifamdl16,27] As we have already seen in (12), the APC can also follow a Bound Comment within a stanza. More rarely the APC can occur inside a CMM within some illocutionary patterns.
(19) *MAX: eh / INP possono essere / SCA sottoscritti anche a due nomi / CMM questi prodotti / APC o è un nome solo ? CMM 'eh, can these products also be subscribed to two names, or is it just one name?' %ill: alternative question [ipubdl4,374] As mentioned, reference units take the form of utterances and stanzas and only chunks occurring after the Comment of an utterance or a Bound Comment in a stanza are able to develop an Appendix function. Naturally, they must also be performed through a suffix prosodic form. On the whole, the ability to formally identify information units in corpora allows us to also quantify their relative frequencies. Part of our objective is to understand the difference in frequency between the Topic (approximately 20%), and the APC (4.28%), which constitutes a further characteristic in their distinction.

Morpho-syntactic fillings
There is much corpus-based research into the Topic conducted on Romance speech corpora according to the L-AcT methodology that detail its morphosyntactic fulfilments (Signorini 2005;Mittmann-Malvessi 2012;Cavalcante 2015;Cresti & Moneglia 2018b). Within L-AcT, Topic is considered the field of application of the illocutionary force and must meet some semantic conditions, since it must ensure a 'stable' reference on which the illocution can operate. Topic is conceived as a syntactic and semantic island that must be interpretable, let's say, for itself by the addressee, excluding, for instance, negative pronouns or modal deontic expressions. These would not be adequate to ensure the cognitive frame to perform the expected illocutionary application (Cresti & Moneglia 2018b However, nearly 10% of Topic fillings are adverbial phrases (in my opinion, probably) that direct the illocution of Comment towards the speaker himself and are considered modal Topics. Let see (22) and (23): Both the referential and modal morpho-syntactic choices are in any case consistent with the role of field of application for Comment. Until now there have been few corpus-based studies on the systematic morpho-syntactic fillings of APC according to the L-AcT methodology (Cresti & Firenzuoli 2002;Cresti 2012a) 13 . However, morpho-syntactic correspondences for the APC already show it to be extremely varied and the work currently being conducted on our corpus confirms this observation.
We can see this variation in classifications taken from corpus analysis, found below, which report many morpho-syntactic groupings for the APC fillings. We have assembled them to look for comparable syntactic behaviours and semantic strategies 14 .
-Clitic saturation a. Saturation of clitics in the Comment through phrases or completive clauses (subjective, objective, indirect interrogative) (8.25%) The broad morpho-syntactic variation of the APC fillings cannot be made into any form of consistent generalization. We tried to identify 5 different morphosyntactic and semantic filling strategies (saturation, morpho-syntactic completion and addition, echolalia, fillers, correction), collecting them into 9 sub-groups. The completion group collected under label d., recording 35% of occurrences, may be considered a core type but in effect it corresponds to at least four sub-groups (casual PP, circumstantial PP, Adv, subordinate clauses). We grouped these together merely for convenience and the overall classification could have been elaborated further. Each element seems to have been assigned at the time by the speaker via momentary opportunities, and proceeds with a lexical saturation of clitics, deictic pronouns, empty adverbs, echolalic strategies, and so on. They seem to follow a local stimulus or "inspiration", and we can almost speak of a morpho-syntactic "randomness" in the APCs.
This element of chance accords precisely with the Appendix's function, whose semantic content is "not relevant" and is used at the time with precautionary purpose to reassure the addressee. The degree of randomness yielded by the APCs is a non-issue since their response to the addressee does not involve much or any decoding due to their "triviality".
The difference in morpho-syntactic fillings for the Topic and the APC comes from corpus data evidence. However, not only are the morpho-syntactic Topic and APC fillings different, but we want also to highlight that more than 40% of those found in APCs -morpho-syntactic completions and additions to partial sentence configurations, relative clauses and cleft clauses, echolalic strategies, lexical correction, dicendi gloss, conclusive phraseology -cannot be used as Topics for evident syntactic and semantic reasons.
Let see for instance the experiment of the anteposition in Topic of some chunks conceived as an APC.
(28) *ZIA: in terra / COM stavano // APC 'on earth, they stood' %ill: narration [ifamn08,408] *(28a) *ZIA: stavano / TOP in terra // COM 'they stood, on earth' (35) *NIC: adesso / PHA guarda questa / COM che ho fatto // APC 'now, look at this (photo), which I made' %ill: presentation [ifamdl17,12] *(35a) *NIC: adesso / PHA che ho fatto / TOP guarda questa // COM 'now, which I made, look at this (photo)' (37) *NAN: ma quando l'aveva fatto / TOP l'era splendido / COM splendido // APC 'but when he did, it was splendid, splendid' %ill: assertion of evidence [ifammn21,115] *(37a) *NAN: ma quando l'aveva fatto / TOP splendido / TOP l'era splendido // COM 'but when he did, splendid, it was splendid' (42) *CLA: e allora / DCT voti in una certa maniera / CMM ti garantisco qua / CMM ti garantisco là / CMM eccetera // APC 'and then, you vote in a certain way, they guarantee you here, they guarantee you there, and so on' %ill: exemplifying list [ifammn02,128] *(42a) *CLA: e allora / DCT voti in una certa maniera / CMM eccetera / TOP ti garantisco qua / CMM ti garantisco là // CMM 'and then, you vote in a certain way, and so on, they guarantee you here, they guarantee you there' (47) *FRA: è / SCA perché reagisce a stimoli / COM cose // APC 'it depends on the fact it reacts to stimuli, things' %ill: admission [ifamcv06,115] *(47a) *FRA: cose / TOP è / SCA perché reagisce a stimoli // COM 'things, it depends on the fact it reacts to stimuli' The artificial anteposition in Topic of the linguistic stuff, conceived as an APC, produces unacceptable utterances corresponding to the series of examples preceded by an asterisk and marked with the letter a. The Topic function cannot be filled by a verbal form in the past time, by a relative clause before its head, an echo before its occurrence, a correction before the term to correct, a conclusive formula, that on the contrary are common fillings of APC. In conclusion, even without considering the distributive and prosodic features characterizing the Topic and the APC, they are distinct for a large part of their usual morphosyntactic fillings.

The semantic change
The morpho-syntactic characteristics of the APC and its randomness have been verified via objective data, but also its semantic counterparts should be taken into consideration. Generally speaking, expressions occurring in right or left distributions, and necessarily realized through prefix or suffix units, lead to specific interpretations depending on their different functional roles. Sometimes a typical case has been used to try and demonstrate the equivalence between the Topic and the APC functions: one in which the Topic and the APC are both present in the information pattern of an utterance and are filled by the same expression (usually a NP). However, if all aspects are really considered, the case validates the difference between the function of Topic and that of the APC, rather to confirm their equivalence. Let us look at the following example: (48) *CAR: ma come si chiamava l'autore?
'but what was the author's name?' %ill: wh question *GIU: eh? %ill: repetition ask *CAR: com'era l'autore? 'how was the author?' %ill: wh question (insistence attitude) *GIU: l'autore / TOP 'un me lo ricordo / COM chi era / SCA l'autore / APC vedi // PHA 'the author, I can't remember who was, the author' %ill: admission [ifamcv19,123] By the first occurrence of the author in the Topic, the speaker GIU takes up the nominal phrase from the previous question asked by CAR and gives her answer signalling to the latter that after an initial misunderstanding her question has been accepted. Thus, it serves as the field of application for the assumption of ignorance by GIU, reactivating the discourse topic. GIU, however, also adds a scanned echolalic APC afterwards. You might think that it is the repetition of the Topic, but it represents only the repetition of its empty content and not of its function. We would underline the differing prosodic performances for the same expression: the Topic is characterized by a prosodic prominence, realized through a prefix unit of the second type (raising), while the f0 lowering in the suffix unit produces a defocusing effect in the APC. The second occurrence of the author is devoid of meaning and can no longer serve to activate the discourse. In conclusion, the two occurrences of the same expression develop distinct information functions that are difficult to confuse. In addition to the examples derived from the corpora, we propose experiments that we carried out on the artificial commutation of the order of information units from their original conception and distribution such as the APC, to the Topic position. Although the substitution may have been adequate from a syntactic point of view (ignoring their prosodic unacceptability) the results clearly demonstrate how inappropriate the replacements are from an information perspective. The example goes: two girls are chatting and among many arguments SAB asks IDA about presents for the latter's birthday. IDA gives a generic answer where she recognizes that she doesn't remember any more of her gifts and the information is smoothed by the repetition of SAB's words in the APC.
(49) *SAB: poi che cosa ti hanno regalato? COM […] 'and then what they gave you?' %ill: wh question *IDA: non mi ricordo più / COM che cosa mi hanno regalato // APC 'I cannot remember, what they gave me' %ill: answer [ifamdl18,121] If we commutate the order of the APC content, placing it in the Topic position as in (49a), although the answer would be acceptable from a syntactic point of view, it creates a 'sharp' attitude with the meaning "why do you care about my business?". The anteposition in the Topic of the completive clause (what they gave me) conveys its full meaning, since it behaves as a syntactic/semantic island with an alethic modality (see 4.2) instead of being an echolalic APC following its original conception. Therefore, within the spoken exchange the clause changes its semantic interpretation with an unpleasant effect.
We also see frequent cases in which a certain expression -mostly adverbs, conjunctions, connectives, i.e. polyfunctional morphemes (Debaisieux 2013)changes its meaning if it is employed as a Topic or APC. Morphemes such as allora (then) and però (but) can be employed as an APC to end an utterance, but they may also occur in the opening of an utterance as a Topic: the two functional distributions imply a semantic change. Let see: When the morpheme allora develops a Topic function, it has a referential meaning of past time (temporal adverb). Also, però, as a Topic, assumes an adverbial modal value "taking into account all the previous things". Conversely, when both morphemes appear after a Comment with an APC function, as in (52) and (40), from a semantic point of view they behave as connectives showing a weak concessive meaning.
(52) *MAX: chiudo tutto / COM allora // APC 'I block everything, then' %ill: request of confirmation [ifamcv17,203] (40) *DAN: mangiala tutta un boccone / COM però // APC 'eat it all in one bite, then' %ill: order [ifamcv15,165] The two morphemes, allora and però, are even interchangeable, given their shared connective role and generic value of agreement. We translated them into English with the same morpheme: then 15 . This is acknowledged in the literature: see Cimmino's research (forthcoming) on the ma and però morphemes which could be considered "equivalent" from a syntactic and semantic point of view. Conversely, the corpus-driven research carried out on IPIC confirms on the one hand their semantic asymmetry, as proposed for instance also by Beeching & Detges (2014) 16 , and on the other their different discourse functions starting from their information features. Specifically, in the right periphery distribution però -the author says -can play a modulating function at a pragmatic level, offering an intersubjective interpretation 17 .
We would also like to cite the case of anyway as analysed by Haselow (2015) in research into the processes of grammaticalization. He underlines how the expression passes from a clause-internal adverbial into a discourse-grammatical marker with a predominantly global connective function in the left periphery of an utterance. However, Haselow specifically emphasizes that when anyway occurs in the right periphery it develops a local connective function where it also exhibits a modifying effect at the illocutionary level.
In conclusion, if an expression is isolated, properly performed by prosody, its interpretation changes depending on its occurrence in the so called left or right periphery (Ziv 1994; Averintseva-Klisch 2008) but according to L-AcT, it depends on the fact that they develop two different information functions.

The absence of modality in APC
As we anticipated the information function of the Topic corresponds to a syntactic and semantic island characterized by semantic conditions. This restriction comes from the necessity of the Topic to provide a sure and stable reference for the addressee. According to L-AcT, the latter is called to make an explicit commitment on the illocution of the Comment since he must proceed to the pragmatic appreciation of the utterance. Comment's content in the case of directive illocutions could be a deontic modality, or in the case of expressive forces, for instance an epistemic modality. However, the Topic content must present a sure semantic basis for the Comment, thus it must be also conceived within a sure and independent modality, which is to say an alethic modality, and corpus data confirm this theoretical request. The independent modal characterization of the Topic must be emphasized since it is generally ignored by 15 Even if the example were a story that takes place in the past tense 'I blocked everything, then', the morpheme occurring at the end of the utterance and performed by a suffix unit would continue to express a concessive value. 16 The authors support the hypothesis of a functional asymmetry claiming that the same connectives when they occur in the left periphery are characterized by subjective functions, while in the right periphery they are dedicated to intersubjective functions. 17 However, when però, prosodically isolated, occurs in the end of an utterance, occasionally it may also convey a counter-expectative value expressing an epistemic modality. In this case it develops a Parenthesis function that is confirmed by its prosodic performance. For prosodic cues distinguishing APC and Parenthesis information units see Saccone (2021). the literature, which lacks the illocutionary conception of the information structure 18 .
Conversely, such restrictions have not been found in the case of the APC, which demands only that its semantics are easy to decode, as with repetitions, formulas and deictics. We can say that because of its irrelevance the APC's content doesn't constitute an independent syntactic and semantic island, as Topic does, and is devoid of an independent modality.
However, let see also the case when the content of the APC, beyond being irrelevant, may correspond to a partial syntactic completion of the Comment, that it could be considered an adequate semantic content to express a modality own. We assume that even when the APC content constitutes a syntactic continuity with the Comment, it doesn't require an independent modality since it only fulfils a sort of "grammatical responsibility". Let's look at various cases of completion: In example (53), the completion in the APC concerns a NP developing a role of postponed subject (those records), already cited and at one point omitted out of the sentence with respect to the question accomplished by the Comment. In example (54), the verb in the APC seems to saturate the predicate of the previous nominal sentence, but the location through the PP (on the ground) would have been a sufficient description to the addressee. In example (32) the partitive specification (of spells) in the APC is the repetition of already told information. Finally, in example (35) the NP object within the instruction accomplished by the Comment is accompanied by a useless relative clause (that I made) since the picture has obviously been made by the speaker. In all the previous examples the content in the Comment, accomplishing a specific illocution, would be sufficient to answer or explain the situation to the addressee. Even if APCs are removed, they do not change the interpretability of the utterance. Moreover, those completions in APC behave as supplementary syntactic pieces, like patches later added to the whole, that is to a predefined syntactic configuration. They are not syntactic/semantic islands responding to an independent conception rather following a syntactic trajectory. A syntactic configuration must be conceived in its entirety within the same modality (Cresti 2014), but if the APC content is the completion of a partial syntactic structure predefined in the Comment, it cannot meet another modality.
In this regard, it is interesting to note that new semantic additions are forbidden in completions made by the APC, showing their dependent conception on Comment. Let see the case of the insertion of new clitics in the APC, anaphorically relating to the phrase in the Comment, that result being ungrammatical in Italian. We can verify it in the comparison of the artificial example (55), which is acceptable in Italian, with its version (55a), where a new clitic is added, and which is not.
(55) *ABC: che cosa ti hanno rubato ? COM 'what have they stolen from you?' %ill: wh-question *DEF: la bicicletta / COM mi hanno rubato // APC 'the bicycle, they have stolen from me ' %ill: answer The second turn of example (55) records a Comment realizing an answer, through only a NP (the bicycle) performed by a root prosodic unit, that is followed by an APC (they have stolen from me) performed by a suffix unit. The morpho-syntactic filling of the APC is the syntactic completion of a possible sentence (they have stolen the bicycle from me), through the partial echo of the preceding question. On the contrary, version (55a), recording the insertion of the clitic la -and the consequent necessary agreement in Italian with the feminine form of the past participle -, is ungrammatical in Italian.
*(55a) *ABC: la bicicletta / COM me l'hanno rubata // APC 'the bicycle, they have stolen it from me' %ill: answer The insertion of the clitic la (they have stolen it from me) would transforms the APC's content from a simply echolalic completion in an entirely new autonomous sentence which is unacceptable until when it is realized by a suffix prosodic unit. The occurrence of the clitic la would be acceptable only if a second independent utterance is performed, as in (55b): (55b) *ABC: la bicicletta // COM me l'hanno rubata // COM 'the bicycle. They have stolen it from me (unfortunately)' %ill: [1] answer; [2] expression of regret Evidently, in example (55b) each utterance has its own illocution -an answer and an expression of regret -, is properly performed by a dedicated root prosodic unit and each of them is an independent syntactic/semantic island. In the second utterance the clitic la is part of a new sentence within a Comment, thus it is focused, and the clitic can develop an anaphoric reference to the previous utterance. Corpus-driven research, indeed, showed that clitics may refer anaphorically to semantic contents in previous utterances, not only within a certain distance, but also if they are focalized (that is, if they develop a Topic or a Comment function). We didn't find instances of clitics dating anaphorically to earlier content developing information functions of either the Parenthesis or the APC, which appears to be a significant finding offered by corpus-based research (Cresti 2009).
Generally speaking, a completion in APC can only follow a predefined syntactic program, confirming its lack of independent semantic conception and modality.
Phenomena like APC syntactic completions have been acknowledged in literature when completion is performed by the addressee and are known as "anticipatory completion". These utterances are syntactically suspended by the speaker and completed by the addressee within an "ongoing trajectory", as an achievement of a shared perspective (Hayashi 2003) or the syntax of a sentence in progress (Lerner 1993(Lerner , 2004.
Comparing the APC with anticipatory completions by the addressee, we notice that this latter phenomenon is motivated mostly by the lexical incertitude of the speaker, who signals his difficulty prosodically, or through a veiled request of comprehension and confirmation. Conversely, in the case of the APC, the completion is performed by a single speaker, who through the addition of the "trivial" delayed content fits a predefined syntactic model. In some sense the speaker seems to take the place of or mimic the addressee, playing to the latter's expectations. In this way the speaker signals a sort of agreement with the addressee on the content being discussed.
In conclusion, APC's completions are devoid of an original point of view and conceived within the same modality of Comment, unlike the Topic, which must present an alethic modality independent from that of the Comment. The modality of the APC is somehow "inherited" from the Comment, or better yet is inactive 19 .

The APC's semantic weakness
As mentioned, the nature and the function of those entities that we have referred to as détachement instances have been the subject of broad discussion since the 1990s, according to which APCs are often defined as exceptional cases of Topic: anti-Topic (Lambrecht 1981), Topic Continuity in Discourse (Givon 1984), postposed Topic (Lombardi-Vallauri 2009, right edge Topic (Ortega-Santos 2016). Terminology is significant as reveals a conception that directly compares the détachement elements with those occurring in the left periphery of the utterance, participating in the Topic-Comment binary structure 20 .
In our opinion the conception of the APC in terms of a variant of Topic derives from the lack of a proper identification practice in the corpus of those instances that are in effect added later 21 . Thus, the specificity of the APC's function has not been deeply investigated, ignoring its semantic irrelevance and in parallel the aim at agreement with the addressee.
Specifically, APC cannot be a device that exist to reactivate a discourse topic, as assumed in literature (Givon 1984), because this would not allow us to 19 It is no coincidence that in the jumble of random expressions fulfilling the APC we do not find clear forms of modal evaluation, that would presuppose an autonomous attitude. However, it must be recognized that there is a difficulty in prosodically distinguishing cases of APC from Parenthesis occurring at the end of the utterance. Only a systematic corpus-based research on the latter information unit could give greater certainty to the exclusion of ambiguous cases. 20 See the contributions made by Lombardi-Vallauri (2007, who explicitly compares the notion of the postposed Topic with that of the APC as defined within L-AcT, even if, in the end, he considers the dètachement instances to still be a kind of Topic. 21 As anticipated, within L-AcT, the Topic and the Appendix information units are clearly distinct in their distributions relative to the illocutionary accomplishment (before the Comment, after the Comment), and through their prosodic performances (via prefix and suffix units).
distinguish the actual function, that in effect is carried out by the real Topic, and the scarce semantic adjunction or even empty repetition developed by the Appendix. See for instance example (53), where the different functions of a same expression (discourse topic reactivation and echolalic repetition) are highlighted.
Moreover, the conception of the APC as a discourse reactivation device should presuppose that is addressed to the continuation of the next utterances within the speech, while given that it serves as an inert integration or completion of the preceding Comment, it is rather characterized by a backward involvement. Interactionism too has defined détachment instances in term of retroactive operation, that are locally limited and aimed at modification of the foregoing utterance (Couper-Kulen & Ono 2007).
The semantic irrelevance of APC content is confirmed by corpus data, the analysis of which shows a kind of semantic inertia on the part of the APC. This could already be inferred from the lack of an independent modality; however other semantic features lead to this conclusion too. We will briefly mention them: defocusing and explicitness.
As anticipated for examples (49) and (49a), while the Topic signalled through a prosodic prominence conveys a semantic focus, the Appendix does not prosodically highlight its content, but rather follows a de-focalisation strategy. Evidence that the content of the Topic is focalized, while that of the APC is not, depends not only on its peculiar prosodic performance but also on the ability to be the semantic antecedent for clitics in subsequent utterances. Following the comparison between examples (55), (55a) and (55b), we have already shown that the content of APC is defocused, since it doesn't allow the insertion of an active anaphoric pronoun.
Additionally, with the Topic, the speaker proposes content (both old and new), which is informationally salient and verifiable (alethic). However, the relevant aspect of the Topic's content is proposed implicitly to the addressee, who is called upon to pragmatically interpret the illocution of the Comment and respond appropriately. In order for this to succeed, the Topic's relevance is, to some degree, hidden and meaning is taken for granted or implicit (Lombardi-Vallauri 2019; Lombardi-Vallauri & Masia 2020). In our perspective this follows from the fact that the addressee must concentrate on the Comment. Conversely, the "random" expressions yielded by the APC do not necessitate any interpretation or "decoding" by the addressee, who, as we have seen in the case of completion, may have even said the expressions himself. The addressee knows that what is added later -after the illocutionary accomplishment of the Comment, defocused and devoid of independent modality -is given as a form of agreement to him. The trivial content in the APC doesn't contain any implicit semantic load because this would be counterproductive. The speaker chooses a strategy of inert semantic adjunction since the addressee has already received the core illocutionary information and no longer has to interpret it. Rather he has only to comprehend the conciliatory attitude of the speaker. Thus, the content of the APC is, in one sense, irrelevant, but, on the other hand, also exhibited purposefully. From a semantic point of view the Topic and the Appendix have opposite contents: that of the Topic is relevant and implicit and that of the APC irrelevant and explicit.
In conclusion, the Appendix is not an independent syntactic/semantic island, is devoid of independent modality, defocused, trivially explicit, an inadequate antecedent for subsequent utterances and a mild retroactive operator. Thus, it cannot perform the role of any kind of Topic, even postponed, and must in fact be distinguished from the latter.

Conclusion
The function of the APC is to textually integrate with the Comment, in order to facilitate a concordance or search for agreement with the addressee. The APC unit may be formally identified in the corpus through its placement after the Comment and its prosodic performance via a suffix unit. The APC records a frequency value of 4.28% of reference units, which is significantly lower than that of the Topic (close to 20%). The morpho-syntactic fillings of the APC show major variationwe could say a morpho-syntactic "randomness" -and cannot truly be generalized since these expressions are employed "in the moment", in the form of lexical saturations, late adjunctions, syntactic completions, echoes, repetitions, deictics, and formulas. The APC's content doesn't constitute a syntactic/semantic island, as Topic does. It is ultimately irrelevant, lacks an independent modality, is de-focused and behaves as a mild retroactive operator. However, given that the APC is produced purposefully to reassure the addressee or agree with them, its content is also explicit, unlike that of the Topic.
The APC's features confirm the specificity of its information function, allowing it to be clearly distinguished from that of the Topic.