Experimenting on Factivity. An investigation into the prosody of factive sentences
When we speak, we often mean more than what we say, that is, more than what our utterances literally mean. In these cases, there is a surplus of content which is not expressed explicitly but is implied and, therefore, is part of the conveyed message. An example is when a speaker uses a so-called factive verb, that is, a verb such as know, discover, etc., followed by a that-clause, as in (1):
(1) Jim knows that Sue lives in Rome
A sentence such as (1) conveys implicitly that the content of the that-clause is true: Sue lives in Rome. However, this implication has been recently debated since factive verbs can also be used in contexts where the content of the that-clause is not true, as in (2):
(2) Everyone knew that stress caused ulcers, before two doctors proved that ulcers are caused by bacterial infection (Hazlett 2010)
Interestingly, results from our previous studies suggest that factivity (i.e.whether the content of the that-clause is assumed to be true or not) is audible in the way the speaker pronounces the sentence. In the project we want to confirm this preliminary finding with empirical methods. Following a two-year-long time plan, we intend to record a collection of spoken examples comparable to (2) in the languages Italian and English, analyze them acoustically, and use them in a perception test with human listeners. By providing new experimental data, the project will deepen our understanding of factivity and of how implied meaning is conveyed in communicative contexts.
(1) Jim knows that Sue lives in Rome
A sentence such as (1) conveys implicitly that the content of the that-clause is true: Sue lives in Rome. However, this implication has been recently debated since factive verbs can also be used in contexts where the content of the that-clause is not true, as in (2):
(2) Everyone knew that stress caused ulcers, before two doctors proved that ulcers are caused by bacterial infection (Hazlett 2010)
Interestingly, results from our previous studies suggest that factivity (i.e.whether the content of the that-clause is assumed to be true or not) is audible in the way the speaker pronounces the sentence. In the project we want to confirm this preliminary finding with empirical methods. Following a two-year-long time plan, we intend to record a collection of spoken examples comparable to (2) in the languages Italian and English, analyze them acoustically, and use them in a perception test with human listeners. By providing new experimental data, the project will deepen our understanding of factivity and of how implied meaning is conveyed in communicative contexts.
Final report
Purpose and development
The purpose of our project is to contribute to a broader understanding of the linguistic class of factive predicates, such as 'know', 'remember', 'discover', 'realize', etc., which typically establish a relation of implication to the truth of the proposition expressed by their complement clause (p). This relation of implication is called "factivity". To illustrate, a sentence such as (1) asserts that John holds a mental state of knowledge and implies that the object of John’s knowledge – the proposition 'Mary lives in Italy' – is true.
(1) John knows that Mary lives in Italy.
Several works by linguists and philosophers have highlighted the occurrence of factive predicates in non-factive contexts, that is, in contexts where p is false, as illustrated in (2):
(2) Everyone knew that stress caused ulcers, before two Australian doctors in the early 80s proved that ulcers are actually caused by bacterial infection. (ex. (1) in Hazlett 2010: 501)
For Italian, previous investigations have demonstrated that the verb 'sapere' ('know) is polysemous between a factive and a non-factive sense, supporting the idea that this word, when used in mental states ascriptions, has undergone a process of semantic change that has led to the generalization of its meaning into 'believe' (Colonna Dahlman & van de Weijer 2019, 2022). Based upon empirical evidence provided by several scholars, we assume a similar ongoing process of generalization for English 'know'.
The non-factive use of factive predicates is a phenomenon that is typical for spoken language, thus our research project has focused on the acoustic realization (notably the prosody) of factive predicates in factive and non-factive contexts. The aim was to determine whether there are prosodic correlates to factivity in spoken language, that is, whether sentences containing factive predicates have different acoustic realizations depending on whether they are used in factive or in non-factive contexts. We have chosen to focus on two languages, Italian and English.
The project is experimental and rely on data gathered by carrying out two experiments. Experiment 1 was geared toward production and consisted of a sentence elicitation task administered to native speakers of Italian and English; Experiment 2 was geared toward comprehension and presented the replies that were recorded in Experiment 1 to new groups of speakers of Italian and English.
Implementation and results
Experiment 1 was carried out in the period between September 2023 and November 2024. 30 native speakers of Italian and 27 native speakers of English participated in the experiment.
The data were collected both in Sweden and in Italy. In our project description, we applied for funding to travel to Italy and to UK to collect data. However, thanks to the international environment that distinguishes Lund University, we discovered that it was easy to find participants in Lund. This allowed us to be eco-friendlier and more sustainable, to run the experiment in the best physical setting – as we could use the facilities of the Humanities Lab at Lund University – and finally to find participants with different geographical origins, thus representing different linguistic varieties.
The participants engaged in dialogues with the experimenter. They were first shown a context that the experimenter did not have access to. After reading this context, the participants saw a reply that they had to read out loud in response to a question posed by the experimenter. Their task was to read this reply in such a way as to convey to the experimenter a message that fit the information in the context that they had seen. Each dialogue occurred once in a factive context and once more in a non-factive context. Moreover, in half of the dialogues that we analyzed, the factive verb was used to fulfill an evidential strategy of hearsay, referring to another source of information than the speaker, as illustrated in example (B) below.
Here are two dialogue examples from the English version of our experiment, (A) and (B):
(A) Factive context: Mary is a vegetarian. Her friend Billy is having a dinner party tonight preparing only meat. So he didn’t invite her.
Non-factive context: Mary was a vegetarian but decided recently to start eating meat again. None of her friends are aware that she is eating meat again. Her friend Billy is having a dinner party tonight preparing only meat, and he didn’t invite her.
Question: Why wasn’t Mary invited to the dinner party tonight?
Participant reply: Because Billy knows that Mary is a vegetarian.
(B) Factive context: Your friend Susan is a fashion expert and she told you that short skirts are in this year.
Non-factive context: Your friend Susan has never been one to follow fashion trends. The other day she told you that short skirts are in this year, but you don’t know whether you can trust what she says.
Question: What skirts are in fashion this year?
Participant reply: Susan knows that short skirts are in.
In the analysis we focused on the factive verb and the subject of the matrix clause – in particular their duration and pitch range, two variables that are correlated with acoustic prominence – to study the effects of factivity and evidentiality on prosody.
The data analysis in Experiment 1 shows that English 'know' is prosodically marked in all non-factive contexts, both in terms of duration and pitch range, while Italian 'sapere' is not prosodically marked in the corresponding Italian utterances, where duration and pitch range are reduced in non-factive contexts rather than expanded. This result seems to confirm the idea that while Italian 'sapere' has undergone and completed a process of semantic change, for its English counterpart 'know', this process is in an earlier stage. Speakers of English, when using 'know' non-factively, feel the need to convey the message that they are not using the verb in its canonical factive sense, and to accomplish this they mark the verb prosodically. Speakers of Italian, on the other hand, do not mark the verb prosodically, but they make use of other prosodic cues to differentiate their utterance depending on whether the context is factive or non-factive: in non-evidential answers, they mitigate the focal stress on the verb in non-factive contexts and at the same time put stress on the main clause subject; in evidential answers, they emphasize the prominence of the main clause subject in non-factive contexts compared to factive contexts. In conclusion, our findings show that factivity and evidentiality are factors that affect the prosody of the utterance, both in English and in Italian.
In December 2024, we started working on the design of Experiment 2. A selection of the responses recorded in Experiment 1 were used as stimulus material in Experiment 2. Pairs of responses given in factive and non-factive contexts were combined with either the factive or the non-factive context. Participants were asked to choose which of the two utterances sounded most natural to them given the context they were presented in.
At the time of writing, 68 Italian and 26 English listeners participated in Experiment 2. The results showed that these participants were able to discriminate the utterances produced in factive contexts from those produced in non-factive contexts, and to match them with the correct context. Within non-factive contexts they predominantly preferred responses that were originally given in a non-factive context, and within factive contexts they preferred responses given in factive contexts. This preference was significant in the statistical analysis for both languages, and the difference between the languages was not. However, while the results appear to be stable, we decided to extend the data collection for the English participants in 2026 in order to obtain a better balance between the two sample sizes. This collection is currently still ongoing.
Dissemination
We disseminated our research by holding seminars at Lund University and by taking part in international conferences. We presented our research – notably, the results of Experiment 1 – at the following conferences:
The 5th Experimental Pragmatics in Italy Conference, Venice, Italy, 26-27 September 2024;
The XXXI Congresso internazionale di linguistica e di filologia romanza, Lecce, Italy, 30 June – 5 July 2025;
The 2025 European Society for Philosophy and Psychology Conference, Warsaw, Poland, 2-5 September 2025;
The 11th Experimental Pragmatics Conference, Cambridge, UK, 17-19 September 2025.
Taking part in these conferences was an excellent opportunity to discuss our research, receive precious feedback and establish new contacts for future collaborations.
Experiment 2 was presented at the Huminfra Conference, in Stockholm on 12-13 November 2025. Moreover, the results will be presented at the Congresso degli Italianisti Scandinavi to be held in Bergen on 9-12 June 2026.
We also invited several internationally renowned scholars to Lund, which offered us the chance to discuss our research.
New research questions
In future research, we aim at investigating the interpretation of non-factive sentences in English, Swedish, and Italian – three languages that differ significantly in accepting the use of factive verbs in non-factive contexts (where Swedish is the language that accepts this use less and Italian the language that accepts it most) – by studying brain activity (“Event Related Potentials”).
Furthermore, it would be very interesting to compare the results achieved on Italian data with data collected in other languages that, like Italian, easily accept their corresponding verb ‘know’ in non-factive contexts (for instance, Modern Greek and Serbian).
References
Colonna Dahlman, R. & van de Weijer, J. 2019. Testing factivity in Italian. Experimental evidence for the hypothesis that Italian sapere is ambiguous. Language Sciences 72: 93-103.
Colonna Dahlman, R. & van de Weijer, J. 2022. Cognitive factive verbs across languages. Language Sciences 90: 1-17.
Hazlett, A. 2010. The myth of factive verbs. Philosophy and Phenomenological Research 80: 497–522.
The purpose of our project is to contribute to a broader understanding of the linguistic class of factive predicates, such as 'know', 'remember', 'discover', 'realize', etc., which typically establish a relation of implication to the truth of the proposition expressed by their complement clause (p). This relation of implication is called "factivity". To illustrate, a sentence such as (1) asserts that John holds a mental state of knowledge and implies that the object of John’s knowledge – the proposition 'Mary lives in Italy' – is true.
(1) John knows that Mary lives in Italy.
Several works by linguists and philosophers have highlighted the occurrence of factive predicates in non-factive contexts, that is, in contexts where p is false, as illustrated in (2):
(2) Everyone knew that stress caused ulcers, before two Australian doctors in the early 80s proved that ulcers are actually caused by bacterial infection. (ex. (1) in Hazlett 2010: 501)
For Italian, previous investigations have demonstrated that the verb 'sapere' ('know) is polysemous between a factive and a non-factive sense, supporting the idea that this word, when used in mental states ascriptions, has undergone a process of semantic change that has led to the generalization of its meaning into 'believe' (Colonna Dahlman & van de Weijer 2019, 2022). Based upon empirical evidence provided by several scholars, we assume a similar ongoing process of generalization for English 'know'.
The non-factive use of factive predicates is a phenomenon that is typical for spoken language, thus our research project has focused on the acoustic realization (notably the prosody) of factive predicates in factive and non-factive contexts. The aim was to determine whether there are prosodic correlates to factivity in spoken language, that is, whether sentences containing factive predicates have different acoustic realizations depending on whether they are used in factive or in non-factive contexts. We have chosen to focus on two languages, Italian and English.
The project is experimental and rely on data gathered by carrying out two experiments. Experiment 1 was geared toward production and consisted of a sentence elicitation task administered to native speakers of Italian and English; Experiment 2 was geared toward comprehension and presented the replies that were recorded in Experiment 1 to new groups of speakers of Italian and English.
Implementation and results
Experiment 1 was carried out in the period between September 2023 and November 2024. 30 native speakers of Italian and 27 native speakers of English participated in the experiment.
The data were collected both in Sweden and in Italy. In our project description, we applied for funding to travel to Italy and to UK to collect data. However, thanks to the international environment that distinguishes Lund University, we discovered that it was easy to find participants in Lund. This allowed us to be eco-friendlier and more sustainable, to run the experiment in the best physical setting – as we could use the facilities of the Humanities Lab at Lund University – and finally to find participants with different geographical origins, thus representing different linguistic varieties.
The participants engaged in dialogues with the experimenter. They were first shown a context that the experimenter did not have access to. After reading this context, the participants saw a reply that they had to read out loud in response to a question posed by the experimenter. Their task was to read this reply in such a way as to convey to the experimenter a message that fit the information in the context that they had seen. Each dialogue occurred once in a factive context and once more in a non-factive context. Moreover, in half of the dialogues that we analyzed, the factive verb was used to fulfill an evidential strategy of hearsay, referring to another source of information than the speaker, as illustrated in example (B) below.
Here are two dialogue examples from the English version of our experiment, (A) and (B):
(A) Factive context: Mary is a vegetarian. Her friend Billy is having a dinner party tonight preparing only meat. So he didn’t invite her.
Non-factive context: Mary was a vegetarian but decided recently to start eating meat again. None of her friends are aware that she is eating meat again. Her friend Billy is having a dinner party tonight preparing only meat, and he didn’t invite her.
Question: Why wasn’t Mary invited to the dinner party tonight?
Participant reply: Because Billy knows that Mary is a vegetarian.
(B) Factive context: Your friend Susan is a fashion expert and she told you that short skirts are in this year.
Non-factive context: Your friend Susan has never been one to follow fashion trends. The other day she told you that short skirts are in this year, but you don’t know whether you can trust what she says.
Question: What skirts are in fashion this year?
Participant reply: Susan knows that short skirts are in.
In the analysis we focused on the factive verb and the subject of the matrix clause – in particular their duration and pitch range, two variables that are correlated with acoustic prominence – to study the effects of factivity and evidentiality on prosody.
The data analysis in Experiment 1 shows that English 'know' is prosodically marked in all non-factive contexts, both in terms of duration and pitch range, while Italian 'sapere' is not prosodically marked in the corresponding Italian utterances, where duration and pitch range are reduced in non-factive contexts rather than expanded. This result seems to confirm the idea that while Italian 'sapere' has undergone and completed a process of semantic change, for its English counterpart 'know', this process is in an earlier stage. Speakers of English, when using 'know' non-factively, feel the need to convey the message that they are not using the verb in its canonical factive sense, and to accomplish this they mark the verb prosodically. Speakers of Italian, on the other hand, do not mark the verb prosodically, but they make use of other prosodic cues to differentiate their utterance depending on whether the context is factive or non-factive: in non-evidential answers, they mitigate the focal stress on the verb in non-factive contexts and at the same time put stress on the main clause subject; in evidential answers, they emphasize the prominence of the main clause subject in non-factive contexts compared to factive contexts. In conclusion, our findings show that factivity and evidentiality are factors that affect the prosody of the utterance, both in English and in Italian.
In December 2024, we started working on the design of Experiment 2. A selection of the responses recorded in Experiment 1 were used as stimulus material in Experiment 2. Pairs of responses given in factive and non-factive contexts were combined with either the factive or the non-factive context. Participants were asked to choose which of the two utterances sounded most natural to them given the context they were presented in.
At the time of writing, 68 Italian and 26 English listeners participated in Experiment 2. The results showed that these participants were able to discriminate the utterances produced in factive contexts from those produced in non-factive contexts, and to match them with the correct context. Within non-factive contexts they predominantly preferred responses that were originally given in a non-factive context, and within factive contexts they preferred responses given in factive contexts. This preference was significant in the statistical analysis for both languages, and the difference between the languages was not. However, while the results appear to be stable, we decided to extend the data collection for the English participants in 2026 in order to obtain a better balance between the two sample sizes. This collection is currently still ongoing.
Dissemination
We disseminated our research by holding seminars at Lund University and by taking part in international conferences. We presented our research – notably, the results of Experiment 1 – at the following conferences:
The 5th Experimental Pragmatics in Italy Conference, Venice, Italy, 26-27 September 2024;
The XXXI Congresso internazionale di linguistica e di filologia romanza, Lecce, Italy, 30 June – 5 July 2025;
The 2025 European Society for Philosophy and Psychology Conference, Warsaw, Poland, 2-5 September 2025;
The 11th Experimental Pragmatics Conference, Cambridge, UK, 17-19 September 2025.
Taking part in these conferences was an excellent opportunity to discuss our research, receive precious feedback and establish new contacts for future collaborations.
Experiment 2 was presented at the Huminfra Conference, in Stockholm on 12-13 November 2025. Moreover, the results will be presented at the Congresso degli Italianisti Scandinavi to be held in Bergen on 9-12 June 2026.
We also invited several internationally renowned scholars to Lund, which offered us the chance to discuss our research.
New research questions
In future research, we aim at investigating the interpretation of non-factive sentences in English, Swedish, and Italian – three languages that differ significantly in accepting the use of factive verbs in non-factive contexts (where Swedish is the language that accepts this use less and Italian the language that accepts it most) – by studying brain activity (“Event Related Potentials”).
Furthermore, it would be very interesting to compare the results achieved on Italian data with data collected in other languages that, like Italian, easily accept their corresponding verb ‘know’ in non-factive contexts (for instance, Modern Greek and Serbian).
References
Colonna Dahlman, R. & van de Weijer, J. 2019. Testing factivity in Italian. Experimental evidence for the hypothesis that Italian sapere is ambiguous. Language Sciences 72: 93-103.
Colonna Dahlman, R. & van de Weijer, J. 2022. Cognitive factive verbs across languages. Language Sciences 90: 1-17.
Hazlett, A. 2010. The myth of factive verbs. Philosophy and Phenomenological Research 80: 497–522.