Contributions to a neurophysiology of meaning: the interpretation of written messages could be an automatic stimulus-reaction mechanism before becoming conscious processing of information

Background. Even though the interpretation of natural language messages is generally conceived as the result of a conscious processing of the message content, the influence of unconscious factors is also well known. What is still insufficiently known is the way such factors work. We have tackled interpretation assuming it is a process, whose basic features are the same for the whole humankind, and employing a naturalistic approach (careful observation of phenomena in conditions the closest to “natural” ones, and precise description before and independently of data statistical analysis). Methodology. Our field research involved a random sample of 102 adults. We presented them with a complete real world-like case of written communication using unabridged message texts. We collected data (participants’ written reports on their interpretations) in controlled conditions through a specially designed questionnaire (closed and opened answers); then, we treated it through qualitative and quantitative methods. Principal Findings. We gathered some evidence that, in written message interpretation, between reading and the attribution of conscious meaning, an intermediate step could exist (we named it “disassembling”) which looks like an automatic reaction to the text words/expressions. Thus, the process of interpretation would be a discontinuous sequence of three steps having different natures: the initial “decoding” step (i.e., reading, which requires technical abilities), disassembling (the automatic reaction, an unconscious passage) and the final conscious attribution of meaning. If this is true, words and expressions would firstly function like physical stimuli, before being taken into account as symbols. Such hypothesis, once confirmed, could help explaining some links between the cultural (human communication) and the biological (stimulus-reaction mechanisms as the basis for meanings) dimension of humankind.


SECTION -The case: description and research's rationale
Introduction and rationale of the research. We examined, for our research, a series of real-world cases of interaction some of the authors had dealt with in their professional experience. The chosen cases were short enough to be easily handled and, at the same time, they were fully representative of the real world's complexity. The case to be created should have consisted of a realistic problem to challenge participants with; moreover, it should have been fully documented from start to end, consisting of written messages (e-mails) only and set inside an Italian corporation. We set up our case, we named it "The employee and the architect" (as a tribute to the protagonist characters) and we drew up the research protocol (see this Supporting Information, Section 3).
A complete description of the case can be found ahead in this present Section. In extreme synthesis, it goes on as an exchange of written messages (5 e-mails in total) between the employee and the architect; we have submitted these messages to the sample leading its members in a two-step work. In the first step, we have asked the participants to carefully read Messages #1, #2 and #3 in sequence; then, to interpret them and the situation they outline; finally, to report and display the "concrete elements" on which their interpretations were based. The rationale was: in vivo observation of the interpretation process, quali-quantitative analysis and formulation of a hypothesis.
In the second step, we have submitted to participants the last two messages asking them to read carefully the texts, to interpret them, then to solve a problem: the last two messages were two different versions of Message #4 and the problem was to indicate which of the two could have produced the final answer (fifth message). The rationale and its topics are put in a different sequence. Although XX expresses her satisfaction, no inspection has been carried out nor it has been requested any more.
10 21 196 197 INTRODUCTION 1. A case managed completely via e-mail, between an employee and a professional (the "architect"), has been set up. It concerns an interaction, related to a problem, inside an Italian corporation; the interaction lasts one month and a half. The problem developed and was completely solved through 5 transactions (5 messages were exchanged, chronologically labelled from #1 to #5). The employee starts the first transaction (Message #1) and concludes the interaction with the fifth one (Message #5).
2. During the action, the architect requests the opinion of a colleague of his; such request refers to a draft of the answer to Msg #3 spontaneously prepared by the architect (such draft is the first version of Msg #4, the "Hard"/H version).
The colleague studies the case and proposes an alternative Msg #4 (the "Softer"/S version); the advice is accepted by the architect, the "S" version is sent and produces the expected result, as the last reaction of the employee demonstrates (Msg #5).
NEUROPHYSIOLOGY of MEANING / Supporting Information 3. The used case is based on real cases which some of the authors had dealt with; it remains as close as possible to reality at the same time avoiding any reference or hint to the original real situations.
The QUESTIONNAIRE and its MANAGEMENT 4. Anonymity of respondents will be fully guaranteed during either the survey (questionnaire collection) or the analysis (data elaboration). No personal data will be asked; information that is necessary for statistical purposes (age, gender, education level and employment) will be requested as aggregated through pre-defined bins only.

5.
For a better representation in the questionnaire, the case has been divided into two parts. In the first part (corresponding to the "Prologue" of the case description, see this Supporting Information, Section 2), the first 3 messages are gathered, in the same order they have been sent. These messages have been printed in sequence, in a single A4 page. The aim of this first part is to collect data about the interpretation process in general through a first set of questions. Such questions have been printed in another single A4 page (two opened questions, #1 and #2, the first sub-divided into three sub-questions).
6. In the second part (corresponding to the "Action" of the case description, see this SI, Section 2), the two versions of Msg #4 (the "Hard" and the "Softer" one) are presented, in separate A4 pages. They are submitted to participants in sequence (not simultaneously) and the remaining questions are printed in a last A4 page. At first (Questions #3 and #4) the participants' opinions are requested (separately) about the presumable effects of each version of Msg #4 on XX. In the end, after the transcription of the very brief Msg #5 (the employee's last reply), participants are requested (Final Question) to indicate which version (the "Hard" or the "Softer"), in their opinion, has produced the effect showed in Message #5. The aim of this second part is to collect data about the relationship between the interpretations of the alternative messages and the action (the choice) performed by participants.
7. All the questions (or sub-questions, if present) have been divided into two parts: in the first one, the interpretation of the respondent about one specific aspect is requested. In the second one, he/she is invited to "indicate the concrete elements (words, sentences, expressions etc…) on which your answer is based".

8.
A special attention has been dedicated to the wording of the questions.
Structural ambiguity of natural language implies the impossibility to formulate sentences that can be univocally interpreted by everybody, as the acknowledged Italian linguist De Mauro confirms 2 . Thus, any idea to pursue completely unambiguous formulations has been dropped. After the first careful formulation of the questions, two pilot-sessions will be set up for testing the questionnaire's suitability and gather indications about possible 2 The author (De Mauro, 1980) says that natural language is "equivocal" in etymological sense, from Latin aeque vocare (to name in the same way). That is: a same word can be used to refer to different things; different words can be used to indicate the same thing. corrections. In addition, ex-post specific controls will discard from quantitative analysis all the possibly remained ambiguous cases.
9. Same attention has been dedicated to possible statistical distortion effects. For example the YY's Colleague opinion on Msg #4/H (the original, "Hard" version) could influence respondents inducing some biases in their final choice; furthermore, there could be a possible precedence effect if the two versions of Msg #4 were submitted always in the same order. On these bases, the presentation of the two versions to the participants will be counterbalanced: all the participants will be informed that they are going to see, at first, the version spontaneously prepared by the "architect". The second (the "alternative" version) will be presented as suggested to him by one of his colleagues when asked for an advice. However, about one half of the sample will actually receive the two versions in that order (first Msg #4/H, then Msg #4/S); the remainder will receive them in the reverse order.
SURVEY and DATA COLLECTION: 10. All the conductors of the survey sessions (12 persons, in total) are members of the research group or in contact with it. Non-members will follow a brief training, led by one of the authors. All the conductors are committed to avoid expressing any comment about the message texts and concentrate on survey process conformity. Conductors have also to assure that the process is clear for the participants and that they understand the structure of the case and the questions. In order to minimize the speech necessities for the conductors, a title page has been prepared; it contains a presentation of the survey and the main context information (see this SI, Section 4). The conductors are due to invite participants to carefully read it. In the title page, the case will be presented as a real world case.

11
. Informed consent will be requested verbally, after the reading of the title page.
Written consent will not be collected for two reasons: the first is that it would imply the creation and management of a general database, paradoxically increasing, by its mere existence, the risks of accidental data diffusion. The second reason is that our data collection procedure (see also following points) anyway fully guarantees anonymity of participants. At the end of data collection, it will be impossible for everyone either to trace back participants starting from the filled questionnaires or to reconstruct the participants' list.
12. The 12 conductors will operate in a completely independent way and the participants will be enlisted by using their personal relationship network, extended until the third degree of separation. Enlisting requirements: adult condition (age>18 years), High-school degree at least. Exceptions about education level are accepted just for people whose literacy and life experience allow them to understand the case documentation without effort (see Note 1).
13. The conductors will collect questionnaires bereft of every personal indications (or even hints) related to participants. They will individually deliver the collected anonymous questionnaires to the authors' team and those documents will be randomly numbered and stored in a dedicated collection box. The

Title page
First of all, welcome and thank you for joining our research.
The e-mails on which this study is based will be submitted to you during the present session. They have been exchanged in a real working environment and they refer to an interaction that occurred in real life. They are presented in their original version; their text has not been modified to be used for this research. Of course, all the elements that specifically refer to persons, or to the real context, have been removed or appropriately altered for privacy reasons.
Your task consists in reading the messages, respecting their submission sequence. Please, read carefully and answer the questions intuitively, not analytically (although, not excluding some personal reflections, if necessary). Underline the text, take notes or look back at the message text, when deemed necessary, any time you need it.
All the questionnaires will be anonymous. We only ask you to give us general information about yourself, here below, for merely statistical purposes (data disaggregation).
[Questions followed on gender, age range, education level and employment (answers requested through pre-defined bins only).]

Message #1 (description)
A female line-worker (the "employee", named "XX") writes a 67 word e-mail to the Project Account (the "architect") about the installation of the heating plant in her office. She requires an inspection, claiming about "flaws" in the present state of the works. Flaws are no better detailed. In her request, she declares that she is also speaking in the name of some colleagues and she uses the expression: "we would be pleased if, at least once, someone of our Corporation would come here and control…".

Message #2 (description)
The Project Account (a male professional, the "architect", named "YY") answers to XX. In his brief message (48 words) he declares regularity in the Project progress, and ends with: "at the moment, the progress substantially complies with the chronogram".

Block #2
I want to premise that, for the sake of a wise management of the work process, intended to optimize the utilization of our Corporation resources (exactly, in order to avoid wasting public money): -Before Project start, I asked the Director of your structure (B wing of the building), Dr. KK, to put a specific person in charge of controlling the work's progress; -As far as I am concerned, the indicated person is, and will remain, Dr. ZZ; -Dr. ZZ carefully planned the project development steps with us; -Each office, situated in the B wing of the building, has been already supplied with heating systems (hardware), fully complying with the timetable agreed with Mrs. ZZ; -The heating plant is now working, even though in provisional mode.
I do recommend you to send any communication, concerning the mentioned Project, to the specific person in charge of controlling, in order to avoid (as already happened) message exchange with personnel that is not directly and formally involved within the process.

Block #3
However, I inform you that, at the moment, the works under discussion have been suspended, in order to enable the provisioning of the plant-control software. It will manage automatically the heating system in the offices, including yours, regulating the warm air diffusion (in order, as said above, to reduce any waste of money).
As soon as the software will be installed by the contractor, the works will come to end. By the way, in this phase they should not affect the rooms situated in the B wing of the building at all, but only the thermo-station.
All quantitative and qualitative controls, requested by the CHK form [formal inspection document], will be carried out after the end of the works and just before their compliance to fixed quality standards will be attested, as prescribed by the current rules.

Block #4
This said, I have found your objections very interesting. For this reason, once the real existence of the problems you have marked will be assessed, I will certainly solve them as a part of my duty.

Block #2
I remember your last message, which I have already answered, and now I really thank you for this new one. In fact, we do believe that the attention of our colleagues, on field operating with structures and plants we provide, is fundamental to complete our tasks at best.

Block #3
In order to optimize our contribution, I have been since the beginning asking for a unique person in charge of controlling the works, accounted for your office's building. This person is Doctor ZZ (I might have already mentioned her in my previous answer even though, at present time, I am not certain about this). Her duty is to collect all the observations expressed by the staff about the work in progress, then to send it directly to my office. I think you already know her and she is going to receive a copy of the present message. I thought this would make communication easier.

Block #4
Concerning your request, you can be certain that, so far, our Project has been developed by following all the technical and formal standards prescribed by the current rules. In addition, I inform you that the works are not yet concluded and final checks (along with possible inspections) are about to be carefully planned. Please, inform your colleagues about the existence of a person in charge of control and do not hesitate to contact her in the case of further observations or possible problems. As I said, she will return your indications to us; this way, I assure you they will not be ignored.

* Please, read Message #4 and answer to the following questions:
In your opinion, what effect will this version produce on XX?
Could you indicate the concrete elements (words, sentences, expressions etc…) on which your answer is based?

* Please, read alternative Message #and answer to the following questions:
In your opinion, what effect will the alternative version produce on XX?
Could you indicate the concrete elements (words, sentences, expressions etc…) on which your answer is based?

Message #5 (full text)
Thank you very much for your interest and for the information. That was very kind of you and your answer was exhaustive.
Best regards XX

Consider that Message #5 was the final reaction of XX and answer the following questions:
In

SECTION 5 -Case structure and communication critical points
Focusing on the communication aspects of our case, we can synthesize its structure as in Table S1, which accounts also for the critical points of the interaction between the employee and the architect. Such scheme can be translated in plain language as it follows: apparently, the employee (working for the architect's same corporation but belonging to a different branch, with no executive commission) was complaining, through Message #1, about the quality of the heating plant installation. However, some lacks of matter (for example the claimed "flaws" were not specified) suggest to figure  we can detect in it five main content blocks (they are marked through specific subheadings along the message text). Msg #4/S (the "Softer" version, suggested to YY by one colleague of his) maintains the same content while its written form is reviewed and its sequence modified. In practice, the "alternative message" #4/S presents the same content blocks of Msg #4/H in a different order and under a new written form. We have synthesized a comparison of the two structures in Table S2.
The substantial difference between the "Hard" and the "Softer" versions of Message #4 is founded on the diverse approach to the arising conflict: while the "Hard" spontaneous reaction of YY approached it through a direct confrontation, the alternative "Softer" version maintains the same information content but approaches the relation with XX in terms of welcome and acknowledgement. Our work was aimed to explore the process of message interpretation, sharing the general assumption that the communication process is universally uniform. We mean that human communication, although its expressions appear extremely variable, must however stem from a unique base of fundamental factors and processes. Something like a limb in a heterogeneous sample of humans: its aspect looks very different in function of sex, age, size, health and so on; nonetheless, it is based on a unique anatomical and functional scheme. For this, the sample's representativeness with respect to the Italian people was not critical. Thus, we decided to increase, as much as possible, the amount of participants while easing the sampling process (see research protocol, in this Supporting Information, Section 3, point 12).
We recruited 102 participants in our sample, whose characteristics are displayed in manuscript Tables 1-3. The total sample composition (manuscript Table 1) shows an exceeding rate of women vs. men and of Graduates/Post-graduates vs. High-school degree granted members (columns "Education", "Gr" bin vs. "Dg" bin; people granted with Elementary degree are inessential, only 4 out of 102). We also highlight the high rate of students and unemployed vs. employed members (columns "Employment", "E" and "F" bins vs. others). For these reasons, even if sample statistical analysis is less relevant in our work, we have drawn more balanced sub-samples from the total sample.  In this section we present in detail an assessment about the amount of the collected materials ("how much" the respondents have written in their answers, the "physical amount" of the answers).
Starting data analysis, we firstly transcribed into a .xls file the filled questionnaires; such file turned into 1 tab containing 8 data-sheets, one for each main question or data source (information for disaggregating data, Questions #1-a, #1-b, #1-c,  Section 5 and Table S1). Even if some participants refer to these passages in their answers, none stresses them as particularly critical and almost none labels them as "threat" or "personal attack". Finally, while examining the answers to Questions #3 and #4 and to the Final Question, we found that about one fourth of the sample (mean for the three questions 26.5%, range 16% -36%) overtly stated, at least once, the impossibility to analytically answer to the second part of the questions (which requested to point out the "concrete elements" that induced the answer to the first part). These respondents describe their answers to the first part of the questions as the result of "a general impression", "a sensation/a perception"; in other cases they present such answers as "an opinion drawn from the whole message" or In order to check the existence of possible imbalances in the collected data, we explored the distribution of the answer texts with respect, by one hand, to the questionnaire questions/sub-questions and, by the other hand, to the respondents. We quantified these texts through the amount of words and characters contained in the filled questionnaires. We remind that each question/sub-question was divided into two items; when we refer to "totals", we mean that the presented data are the result of summing values related to the "strict" answer (first item, i.e. first part of the question) and values related to the indicated "concrete elements" (second item, i.e. second part of the question).
a -Text amount distribution with respect to items. The results of this first analysis are displayed in Table S3 and Fig. S1. Table S3 shows totals and some statistical indexes with regards to the distribution of the answer text amounts on questions/sub-questions.
Data referred to all the answers (left part) are compared with those excluding Question #2 (right part). The reason of such exclusion: answering was under condition and Question #2 was answered by only a part of the sample. In order to investigate the distribution shape, we drew the histogram of Fig. S1, which displays the percent distribution of text amounts (in terms of words and characters, Question #2 excluded) with respect to the questionnaire items. It shows evident lower levels for Questions #1-b and #1-c (whose minimum, all the same, is around 7%); the rest of the values seesaws between 9% and 11% (the general percent mean, per item, is 100:11=9.1%, see Table   S3, right part, "% Gen. means per item" row). About this, we must consider that several respondents answered in short to subquestions #1-b and #1-c, just indicating some references to the previous sub-question (#1-a, indeed having the highest values). Thus we prefer to use, for comparing different items, values referring to the percent mean of the three sub-questions of Question #1, that is 8.3% both for words and for characters (SI = spaces included). On the whole, we have a range oscillating between 8.3% and 11.1% (for words) or 11.3% (for characters). No meaningful difference is recordable; the distribution of the answer texts with respect to the questionnaire items has an almost rectangular shape and can be assessed as satisfactorily uniform. Actually, no question at all has been neglected by respondents.
b -Sample distribution with respect to the text amounts. The results of this analysis are displayed in Table S4 and Fig. S2 and S3. Table S4 shows totals and some statistical indexes referred to the amounts of text (in terms of words and characters, Question #2 excluded) provided by respondents through their answers. Data are displayed separating values referred to the first item of the questions ("strict" answer) from those referred to the second one ("concrete elements"). In order to investigate the distribution shape, we drew two histograms, in which participants have been grouped in   In addition, Table S8 and Table S9 contain the source data for the manuscript Fig.7 and Fig.8, which display, for the sub-samples "Age" and "Employment", the same kind of histograms of Fig. S8-S9-S10-S11.

SECTION 12 -The "block preference" analysis
The second indicator we have used (block preference indicator), was built starting from the consideration (this SI, Sections 4 and 5) that the "Hard" (H) and the "Softer" (S) version of Message #4 contain the same content blocks (it was an overt decision of YY's "colleague"), differing for the order of presentation and for linguistic form. Each block is identified as concerning a given content (see this SI, Section 5 and Table S2). Then, we investigated about possible differences regarding the attention paid by "H" and "S" choosers to different blocks, while answering to Questions #3 and #4 (predictions of the messages' effects on XX). Our goal was to explore finer characteristics in the choice process. Specifically, we intended to verify if the different choices ("H" or "S") were linked to differences in focusing on the blocks or in detecting diverse characteristics inside same blocks. In the first case the different contents, ascribable to the different blocks, would lead the process; in the second case, other factors would play a critical role.
To build the block preference indicator we, at first, examined the answers to Questions #3 and #4 and highlighted all the direct references to Message "H" and Message "S" texts (i.e. sentences in quotation marks or undoubtedly referring to clearly identifiable passages). Then, we associated them to the text blocks. Results from this part of the analysis are displayed in Tables S10-S13 5 ; they contain clear indications about the 5 Tables S10 and S11 display data with regards to the amount of references to each block expressed by participants. In Table S10, totals for each block and each evaluated message (as well as general totals) can be higher than the people amount, given that each person can express more than one references. Tables S12 and S13 display data with regards to the amount of participants that referred to each block. In Table S12, totals for each block and each evaluated message must be inferior to the participants' amount; however, the general totals can be higher, given that each person could refer to more than one block. message blocks which the attention of participants has fallen upon. We will base our analysis on Table S12 data; blocks are displayed along with the texts of the "Hard" (H) and the "Softer" (S) versions of Message #4; a comparison among them is presented in Table S2. inside the organisation. Both "H" and "S" choosers give Block #2 a prevalent attention, when they read it in the "Hard" (H) message. However, when they read it in the "Softer" (S) version, we see that "H" choosers maintain their preference (with a little shift towards Block #3, containing specific information) while "S" choosers pay the minimum of 6 We remind that Message "S" maintained the same content of Message "H", and that content was divided into analogous text blocks, but varying their sequence (besides their written form). For reliable comparing, it has been necessary to give each "S" block a "converted number", that is the same of the correspondent block in Message "H" (see this SI, Section 5, and Table S2, extreme right column). From now on, until express notice, all the numeric references to "S" blocks must be intended as converted numbers. information is what it matters and they pay little attention to relational aspects. Such situation reminds the differences between "H" and "S" choosers' behaviours highlighted by coherence indicator analysis (specifically, the sample distribution with respect to coherence level).
We successively noted that a minority of "S" choosers, while evaluating the "Hard" (H) message, focused on Block #4 (the relational acceptance passage) and rated it, overwhelmingly, negative (4+15=19, see Table S12, Block #4 row, column "S" choosers/"H" evaluation). Some of them, for example, justify their evaluation interpreting that YY overtly declares that he does not trust XX, given that he says he reserves himself to check for the real existence of the problem, before intervening 7 . They 7 We observe that, as widely discussed in the manuscript (specially in the Discussion section), the question is not linked to the information per se, nor it regards YY's right to control. The question is "the fact that" YY decided to overtly declare, in a certain point of his message and under a certain form, his doubt and his intentions.
the second option, we got the impression that such formulation could result weak and that the observed processes cannot be restrained to such dichotomy. Then, how can we explain our observations? The picture can be synthesized as it follows:  When predicting Message "H" effects, both "H" and "S" choosers mainly focus on the same block but they are attracted by different characteristics: "H" choosers by its information content; "S" choosers by its relational impact.
 When predicting Message "S" effects, "H" and "S" choosers mainly focus on different blocks. However, their answers show that such behaviour is linked to the attraction they feel towards the same characteristics that stimulated them in the previous case: "H" choosers insist on privileging information content (and Blocks #2 and #3, that concentrate the information); "S" choosers shift towards new blocks that make evident the relational care of YY with regards to XX (Blocks #3 and #4, converted numbers, see Note 6).
One last aspect to be cleared: the second point contains, besides the specific divergence in focusing, a new example of the first case, i.e. the same focusing joined to attention paid to different characteristics. Actually, both "H and "S" choosers focus also on Block #3 (converted number) of the "Softer" (S) message, that is labelled as "Information" in Table S2. However, even though that block undoubtedly contains information, the two versions present it in different ways. Confronting the texts, we can easily verify that the "H" version bears just technical and formal contents while the "S" version pays attention to present the information as a "service" for the colleagues.  This histogram shows that the word and character percent amounts resulting from the respondents' answers vary, with respect to items, from 6.6% to 11.9% (words) and from 7.0% to 11.6% (characters, spaces included). The range reduces to 8.3%-11.1% (words) and 8.3%-11.3% (characters SI) if the three sub-questions of Question #1 are grouped together and their mean is considered (see text for details). The amounts appear to be distributed in an almost rectangular distribution (i.e. in a satisfactorily uniform shape) across the questions of the questionnaire. On the whole, no item seems to be definitely privileged, or neglected, by participants. With respect to questions, the respondents' indications about the focused components (see text for definition) present a rectangular-like percent distribution (differences in a range around 5%, from 12% to 17% about, source data from Table 8 The respondents' indications have been grouped in bins by type. The presented percent distribution (source data from manuscript Table 8, "%" row) has been built through ranking the first six types (from "Symbols" to "Whole") by their increasing size. The remaining three types (Information content, Other components and Grammar notations) have been added ranking them by decreasing percent values. The highest frequencies correspond to middle-sized "chunks" of the messages.

#1
Form of address Form of address S "1"  S "1" converted #2 Re-addressing XX Relational acceptance S "2"  S "4" converted #3 Information Re-addressing XX S "3"  S "2" converted #4 Relational acceptance Information S "4"  S "3" converted #5 Form of saluting Form of saluting S "5"  S "5" converted     Predictions about the effects that the "Hard" (H) or the "Softer" (S) version of Message #4 could have on the receiver (the "employee" XX) are independently expressed, by each member of the sample, through answering to Questions #3 and #4. Answers are classified through the dummy variable "Expected effects" (possible values "+", if respondents point out that the message will solve the XX-YY conflict, or "-" in the opposite case). The H-/S-2 8.3% 7 9.4% 9 9.2% 1 6.7% 5 12.2% 6 10.7% 1 5.6% 5 11.6% 6 9.8%  In this table the combined predictions (see Table S5) about the effects that the "Hard" (H) or the "Softer" (S) version of Message #4 could have on the receiver (the "employee" XX) are crossed with the final choices of the respondents (all the variables are independent). Data shows the association (for the total sample and the two control subsamples) between the "H-/S+" combination and the S-version (the "Softer" one) as final choice. In addition, some correlations between the two variables is underlined by Chisquared test (p=0.001, total sample; p=0.035, sub-sample "AGE"; p=0.009, sub-sample "EMPLOYMENT") and Fisher's Exact test (p=0.002, total sample; p=0.027, sub-sample "AGE"; p=0.008, sub-sample "EMPLOYMENT").     The sequence of the blocks belonging to Message "H" is the original one (as it appears in the actual message); the sequence belonging to Message "S" is converted (see SI, Section 12 and Note 6, for details). [Legend: +/-= type of predicted effect (resolution or escalation of the conflict) of the "Hard" (H) and the "Softer" (S) version of Msg #4 on XX.] The   The sequence of the blocks belonging to Message "H" is the original one (as it appears in the actual message); the sequence belonging to Message "S" is converted (see SI, Section 12 and Note 6, for details). [Legend: +/-= type of predicted effect (resolution or escalation of the conflict) of the "Hard" (H) and the "Softer" (S) version of Msg #4 on XX.] The