Decomposing logophoric pronouns: a presuppositional account of logophoric dependencies

Natural Language Semantics https://doi.org/10.1007/s11050-025-09242-x Decomposing logophoric pronouns: a presuppositional account of logophoric dependencies Itai Bassi1·Imke Driemel2·Abigail Anne Bimpeh3·Silvia Silleresi4· Johnson Fo .lo .runs .o .Ilo .ri5·Anastasia Nuworsu6·Gerald Okey Nweya7 Received: 27 February 2023 / Revised: 26 September 2025 / Accepted: 29 September 2025 © The Author(s) 2025 Abstract Logophoric pronouns in West African languages occur in attitude environments and are anaphorically linked to an attitude holder in a superordinate clause. This has motivated theorists to treat logophoric pronouns semantically as an obligatorily bound variable (bound from the edge of an embedded complement clause). Culy (Linguistics 32:1055–1094, 1994) and Bimpeh and Sode (in Proceedings of TripleA, vol. 6 pp. 1–16, 2021), however, point out that logophoric pronouns in Ewe do not behave like obligatorily bound variables, allowing both sloppy (bound) and strict (nonbound) readings in focus contexts involving ‘only’ and ellipsis. We strengthen this line of criticism by providing novel cross-linguistic data that indicate that logophoric pronouns in Ewe, Igbo and Yoruba support strict readings. We offer an alternative formal account to existing approaches that builds on Bimpeh et al. (in Proceedings of the 40th WCCFL, pp. 1–10, 2024) and can capture both strict and sloppy interpretations, while preserving the requirement that a logophoric pronoun be anaphoric to an attitude holder. The main novelty involves decomposition of logophoric pronouns into two syntactic components at LF—a variable that can in principle be free and refer strictly, and a semantic presuppositional feature LOG that can be ignored in ellipsis and focus sites, following similar ideas in the literature on pronominal features (Sauerland in Proceedings of SALT 23, pp. 156–173, 2013). Our analysis implies that in terms of their syntactic and semantic make up, logophors are essentially no different from other pronouns, consisting of a referential index plus semantic features. Keywords Logophors ·Strict/sloppy readings ·Binding ·Presupposition ·Ewe · Yoruba ·Igbo ·Focus ·de se ·Attitude ascriptions 1Introduction Logophoric pronouns in some West African languages are special anaphoric elements that typically occur in attitude contexts and must refer back to the attitude holder I. Bassi, I. Driemel and A.A. Bimpeh contributed equally to this work. Extended author information available on the last page of the article I. Bassi et al. (Clements 1975). In Ewe, for example, the logophoric pronoun (henceforth LOGP) yè normally appears in attitude ascriptions like in (1). For convenience, the relationship between LOGP and its antecedent is represented with indexation.1 (1) EweKòfí1 Kofi bé say yè1 LOGP dzó. left ‘Kofi said that he left.’ (Clements 1975) Ewe’s LOGP is restricted to this kind of environment; in particular, it cannot be used in simple unembedded sentences to refer to an antecedent introduced in the preceding discourse (or made contextually salient otherwise). For instance, if Afi is the topic of conversation and we mention her having been at the party, then (2a) cannot be used as a follow-up; in such cases, the ordinary third person pronoun (henceforth ORDP) émust be used (2b). (2) a. Ewe*yè1 LOGP dzó leave. ‘She left.’ b. é1 ORDP dzó leave. ‘She left.’ (Pearson 2015, p. 78) Clements (1975), one of the first to systematically study the phenomenon, thus characterizes Ewe’s LOGPyè as an item “used exclusively to designate the person whose speech, thoughts, feelings or general state of consciousness are reported.” In the literature since the 70s, the term “logophoricity” has been applied to describe elements with a similar function of encoding sensitivity to the attitude of some perspective-holder (see Hagège 1974; Charnavel 2020 for French; Koopman and Sportiche 1989 for Abe; Nikitina 2012 for Wan; Kaiser 2018 for Finnish; Park 2018 for Korean; Sundaresan 2018 for Tamil; Kiemtoré 2022 for Jula; Newkirk 2019 for Ibibio; Schlenker 1999 for Amharic; and Sells 1987;Culy1994; Stirling 1994; Güldemann 2003; Deal 2020 for cross-linguistic studies). How is the dependency between LOGP and its antecedent encoded in the grammar? This paper is an attempt to advance toward an answer to this question, from the angle of the well-known strict-sloppy ambiguity of pronominal reference: our discussion and analysis will gear toward explaining why LOGPs allow strict readings in ellipsis and focus contexts (cf. Culy 1994; Bimpeh and Sode 2021). As an overview, we provide new data from three languages—Ewe, Yoruba and Igbo—that confirm that LOGPs in these languages are ambiguous between sloppy and strict interpretations (our work also provides the first cross-linguistic study on de se readings of LOGPs that includes mistaken identity scenarios across several attitude 1Glossing abbreviations follow the Leipzig glossing rules, with the addition of LOGP=logophoric pronoun, ORDP=ordinary pronoun, RP =resumptive pronoun, PREP =preposition, and RED =reduplication. Decomposing logophoric pronouns predicates). We will discuss why the existence of the strict reading is problematic for current approaches (von Stechow 2003; Pearson 2015, a.o.), and will account for the problematic generalizations with a novel theory of the syntax-semantics interface of LOGPs. The main novelty is a decomposition of logophoric pronouns into two syntactic components at Logical Form (LF)—a variable and a semantic feature LOG that induces a presupposition, like other pronominal features. The paper is organized as follows. Section 2provides general background on the basic distribution and interpretation of logophors in Ewe, Yoruba and Igbo. Section 3 presents our findings on strict and sloppy interpretations, and shows why they require modification of the existing accounts. Section 4presents our new proposal for a syntax-semantics of LOGPs that can capture strict readings, including hitherto undescribed flavours of strict readings. Section 5expands the proposal and accounts for long-distance LOGP dependencies and for the relationship between LOGP and the first person pronoun in the languages of interest. 2 Logophors and their distribution and interpretation We focus on three West African languages that have been observed to display logophoric pronouns: the Kwa language Ewe, and two Benue-Congo languages: Yoruba and Igbo. Detailed language profiles can be found in the Supplementary Material accompanying this paper. In Sect. 2.1, we present the main distributional pattern of LOGPs in the languages under investigation, while Sect. 2.2 is devoted to a discussion of the de se-de re distinction in connection to logophors. Unless indicated otherwise, all data in this paper come from original fieldwork. We elicited data from three Ewe speakers (two Anlo and one Ewedome), two Yoruba speakers and four Igbo speakers. All data was elicited via multiple Zoom sessions with each speaker, transcribed live by the experimenters and double-checked by the speakers. The elicitation language was English. Speakers’ spontaneous comments on the reasoning behind their responses were also noted.2 2.1 Obligatory co-reference with the attitude Holder In (3), we present data for Ewe’s LOGPyè embedded under the attitude predicates ‘think’, ‘say’, ‘want’, and ‘hope’.3As indicated by the indexation, yè must co-refer 2Given that we tested the distribution of LOGPandORDP across several predicates, the actual number of test items per data point was never more than 2. As for de se/de re readings (Sect. 2.2), strict/sloppy identity (Sect. 3) and multiple embeddings (Sect. 5.1), we tested the distribution of logophors and pronouns embedded under the verbs ‘say’, ‘think’ and ‘hope’ consistently with all our consultants. The verbs for ‘want’ were tested with all speakers for de se readings. We also tested with the verbs for ‘promise’ in all three languages (see Supplementary Material). For strict unknown identity, we only elicited judgments with the verbs for ‘think’. 3Although tense in Ewe (as well as in most Kwa languages) is not overtly marked, Ewe displays an irrealis marker athat expresses the possibility of an event occurring in the future (Essegbey 2008). The irrealis marker acan be optionally added in (3a)and(3b) to express such a meaning. In (3c)and(3d), however, the marker occurs obligatorily. The environments where the irrealis marker occurs obligatorily roughly match the ones known in English as complements of control verbs; see also the Supplementary Material and the discussion in Grano and Lotven (2019) for Gengbe, an Ewe dialect spoken in Togo. I. Bassi et al. with the attitude holder in the matrix clause; it cannot refer to some other contextually salient individual. (3) Logophors in Ewe (fieldwork) a. Kòfí1 Kofi súsú think bé that yè1/∗2 LOGP á IRR ãè marry Àfí. Afi ‘Kofi thinks that he will marry Afi.’ b. Kòfí1 Kofi gbl`O say bé that yè1/∗2 LOGP á IRR ãè marry Àfí. Afi ‘Kofi says that he will marry Afi.’ c. Kòfí1 Kofi dZí want bé that yè1/∗2 LOGP á IRR ãè marry Àfí. Afi ‘Kofi wants to marry Afi.’ d. Kòkú1 Koku lè COP m´O-kp´O-m path-see-PROG bé that yè1/∗2 LOGP á IRR ãè marry Àfí. Afi ‘Koku hopes that he will marry Afi.’ Yoruba and Igbo have LOGPs that exhibit comparable distributional properties to those of Ewe (Manfredi 1987; Hyman and Comrie 1981; Adésolá 2005;Lawal2006). Examples (4) and (5) illustrate.4 (4) Logophors in Yoruba (fieldwork) a. Adé1 Ade rò think wípé that òun1/∗2 LOGP fé . marry O .lá. Ola ‘Ade thinks that he married Ola.’ b. Adé1 Ade so . say wípé that òun1/∗2 LOGP fé . marry O .lá. Ola ‘Ade said that he married Ola.’ c. Adé1 Ade fé . want wípé that òun1/∗2 LOGP fé . marry O .lá Ola ‘Ade wants to marry Ola.’ d. Adé1 Ade ´n PROG rètí hope wípé that òun1/∗2 LOGP máa FUT fé . marry O .lá. Ola ‘Ade hopes that he will marry Ola.’ 4In contrast to Ewe, the forms for LOGP in Yoruba and Igbo are taken from the general paradigm of strong pronouns, and may therefore also occur in unembedded environments. The strong form òun in Yoruba is used in emphatic environments, while the strong form yá in Igbo is used additionally in object position (Pulleybank 1986; Manfredi 1987; Amaechi 2020). Decomposing logophoric pronouns (5) Logophors in Igbo (fieldwork) a. Ézè1 Eze chèrè think nà that yá1/∗2 LOGP lú .rú . marry Àdá. Ada ‘Eze thought that he married Ada.’ b. Ézè1 Eze sì .rì say nà that yá1/∗2 LOGP lú .rú . marry Àdá. Ada ‘Eze said that he married Ada.’ c. Ézè1 Eze chò .rò . want nà that yá1/∗2 LOGP gà FUT à-lú . PTCP-marry Àdá. Ada ‘Eze wants to marry Ada.’ d. Ézè1 Eze nwè-rè be-RED òlìlèányá hope nà that yá1/∗2 LOGP gà FUT à-lú . PTCP-marry Àdá. Ada ‘Eze is hopeful that he will marry Ada.’ All three languages show identical co-reference patterns. The results of our elicitation sessions confirm previous reports in the literature. In the next section, we will report on another trademark property of logophors: obligatory de se readings. 2.2 The de se-de re distinction For the elicitation of the data in this section, we used a binary acceptability judgment task designed with joint presentation for two target sentences (one with LOGP and one ORDP) to be judged against de re contexts: speakers were asked to express their acceptability judgments on both target sentences, but they were free to accept as felicitous both sentences, one sentence or none. Additionally, consultants were asked to judge the same target sentences against equivalent de se scenarios (see Sect. 2 in Supplementary Material) to confirm our methodology and the data in Sect. 2.1. 2.2.1 de se-only interpretation of logophors It falls beyond the scope of this paper to provide a complete overview of the literature on LOGPs.5But one property of LOGPs that will be integrated into our analysis deserves elaboraton, and that is the de se reading of LOGPs. The so-called de se-de re distinction of pronominal reference in attitude contexts has to do with whether the attitude holder recognizes themselves as the real referent of the pronoun (Lewis 1979, Chierchia 1989, Kaplan 1989, Schlenker 1999, Pearson 2015,Park2018, PatelGrosz 2020, a.o.). Recently, Bimpeh et al. (2024) found evidence that LOGPs in Ewe, 5Nor do we intend to cover in any depth the connection between LOGPs in West African and other phenomena sometimes discussed under the general rubric of “logophoricity”: perspectival anaphora (Sundaresan 2018), exempt anaphora (Charnavel 2020), and perhaps Indexical Shift (Deal 2020). But see Sect. 4 in the Supplementary Material for a comparison to the phenomenon of exempt anaphors. I. Bassi et al. Yoruba and Igbo must meet this “self-awareness” requirement. Specifically, Bimpeh et al. (2024) showed that LOGPs (but not ORDPs) are infelicitous in “mistaken identity” scenarios as in (6), in which Donald Duck is referring to someone who, unbeknownst to Donald, is actually him. (6) Mistaken Identity Context: (Bimpeh et al. 2024) Donald Duck (DD) went to the grocery store to buy flour. He mistakenly put sugar in his cart. DD then saw a trail of sugar going up and down the aisles and thought that someone’s bag had a hole in it and looked around for the guy. DD says: “I wonder who is losing sugar; certainly, the guy who is losing sugar is stupid, as he does not check his bag.” Later he says: “Is it me the stupid guy who is losing sugar? No, because I did not buy sugar but flour.” a. EweDonald Donald Duck Duck súsú think bé that #yè LOGP / / é ORDP dz`Omòví. stupid.person ‘Donald Duck thinks that he is stupid.’ b. IgboDonald Donald Duck Duck chèrè think nà that #yá LOGP / / ó . ORDP bù . COP ónyé´nzúzù. stupid.person ‘Donald Duck thinks that he is stupid.’ c. YorubaDonald Donald Duck Duck rò think pé that #òun LOGP / / ó ORDP jé . COP òmùgò .. stupid.person ‘Donald Duck thinks that he is stupid.’ As with ‘think’ (6), logophors embedded under ‘say’ and ‘hope’ can only be read de se, highly suggested by the infelicity in a mistaken identity scenario in (7) and (8). (7) Mistaken Identity Context: (Bimpeh et al. 2024) Elmo goes to visit Big Bird. While there, Big Bird shows him old paintings he found from back when Elmo was living there with him. After looking at several pictures, Elmo does not recognize one of the paintings, which is particularly pretty. Elmo says: “I wonder who painted this. Certainly, the person who painted is a good painter.” Later he says: “Is it me the good painter who painted this? No, because I am not very talented at painting.” a. EweElmo Elmo bé say #yè LOGP / / é ORDP nyé be nùtálá painter nyùìé good aãé. INDF ‘Elmo said that he is a good painter.’ b. IgboElmo Elmo sì . say nà that #yá LOGP / / ó . ORDP nà-ésè IPFV-paint íhé thing ` nké NKE ó .má. good ‘Elmo said that he is a good painter.’ c. YorubaElmo Elmo so say pé that #òún LOGP / / ó ORDP jé . COP akunlé painter tí REL ó RP dára. good ‘Elmo said that he is a good painter.’ Decomposing logophoric pronouns (8) Mistaken Identity Context: (fieldwork) Goofy is so drunk that he has forgotten he is a candidate in the election. He watches someone on TV and finds that that person is a terrific candidate who should definitely be elected. Unbeknownst to Goofy, the candidate he is watching on TV is Goofy himself. a. Goofy Goofy lè COP m´O-kp´O-m path-see-PROG bé that #yè LOGP / / é ORDP à IRR ãù win àk`O-dádá vote-cast lá DEF dzí. post ‘Goofy hopes that he will win the election.’ Ewe b. Goofy Goofy nà have èlé look ánya eye nà that #yá LOGP / / ó . ORDP gà FUT è-mérí PTCP-win n’ in àtú ´ mvóòtù election áhù. DEM ‘Goofy hopes that he will win the election.’ Igbo c. Goofy Goofy ´n PROG rètí hope pé that #òun LOGP / / ó ORDP yoo FUT borí win nínú in ìbò election náà. DET ‘Goofy hopes that he will win the election.’ Yoruba We take it then that LOGPs in these languages have a requirement for de se readings: LOGP can only refer to the attitude holder’s “recognized self”. 2.2.2 A note on methodology and conflicting generalizations Our claim that LOGPs only have de se readings is controversial. While it converges with the results in Bimpeh (2019) for Ewe, as well as with Adésolá (2005) and Anand (2006) for Yoruba (we are not aware of a previous discussion of de se-de re with respect to Igbo), Pearson (2015) by contrast reports that LOGPinEwealso supports a de re reading, as most of the speakers in her study judged LOGPasfelicitous and true in mistaken identity scenarios (see also O’Neill 2015; Satık 2021). These findings were partly confirmed by a quantitative questionnaire study on de se-de re readings by Bimpeh (2023) with 20 Ewe speakers. While 62.5% responses indicated that LOGP is infelicitous in mistaken identity contexts, 37.5% responses revealed that LOGP was accepted. This indeed suggests that LOGP at least sometimes seems to receive a de re reading. The questionnaire study in Bimpeh (2023), however, revealed several surprising results: most importantly, ORDP was rejected by participants over half of the times (ORDP was judged infelicitous in mistaken identity scenarios 65.7% of the time). This outcome is unexpected under every account mentioned so far. Bimpeh (2023, p. 127) concludes that “a possible explanation [for the discrepancy in results between Bimpeh (2023) and Bimpeh (2019)—our addition] is that participants were mostly confused since they are not used to such scenarios”, which is possibly why LOGP and the baseline ORDP version were both rejected over 50% of the time. The difference between one-on-one elicitation (the I. Bassi et al. current study and Bimpeh 2019) and questionnaires (Bimpeh 2023) is that with the former, consultants can more comfortably be familiarized with the rather unusual mistaken identity contexts, and ask for clarification if necessary. Consequently, the one-on-one elicitation method used in Bimpeh (2019) achieved consistent results indicating exclusively de se interpretation of LOGPs, which align with our findings. It is important to try to clarify the nature of the disagreement, and to this end we now comment about the methodology we used to obtain our results in (6)-(8) and explain how it differs from studies that arrived at conflicting conclusions to ours. Our technique to elicit de re readings is different from previous work on two points. First, as is clear from (6)-(8), and following Bimpeh’s (2023) idea, we let our speakers judge not just sentences with LOGP but also a minimally different variant with ORDP. This gave speakers the opportunity to create a baseline, and it also allowed us to verify that speakers understood the context (they were asked to express their acceptability judgments on both versions, and were free to accept both sentences, one sentence or none). We think this is highly relevant, as mistaken identity contexts are rather difficult to take in (especially for non-linguists/semanticists). With the exception of Bimpeh (2023), previous work (Pearson 2015; O’Neill 2015; Satık 2020,2021) has not tested the ORDP version in mistaken-identity scenarios to establish a baseline. The fact that our speakers all accepted the ORDP version but rejected the LOGP version suggests that only the former is compatible with a de re reference. Second, most of our mistaken-identity scenarios made sure that the target sentences are false on a de se reading. In the mistaken identity context in (6), for example, it is explicitely mentioned that Donald Duck does not self-ascribe stupidity. (“Is it me the stupid guy who is losing sugar? No, because I did not buy sugar but flour.”) No previous work we are aware of (Pearson 2015; O’Neill 2015; Satık 2020,2021; Bimpeh 2023) made it unambiguously clear in the description of the context that a de se interpretation is false. We believe it is important do to so, in order to make sure that upon judging the LOGP sentence, speakers do not apply some charity principle and mentally modify the mistaken-identity context ever so slightly so as to make it possible for the sentence to be true on a de se reading. Whenever our speakers were asked if the sentence is true or false against a mistaken-identity context that made it sufficiently clear that a de se interpretation is false, our speakers indicated to us that the sentence is false.6For more discussion on the methodology, see the Supplementary Material. 6GiventhatPearson(2015) and other works reach the opposite conclusion to ours for Ewe (using different methodology), a reviewer suggests that one cannot rule out the possibility that there is a dialect split or cross-speaker variation in Ewe. On this point we side with Pearson (2022), who contends that it is unclear what sort of evidence in the input could trigger the acquisition of different grammars by different speakers on the de re property of LOGPs. Moreover, we hope that our methods for eliminating possible confounds will prove useful in clarifying the empirical picture (for Ewe). But if it turns out that a genuine variation is what underlies the disagreement in the literature, then our analysis from hereon should be taken to apply only to those speakers (like ours) who only accept de se readings of LOGPs; if there are speakers whose grammar truly allows de re construals of LOGP, our account would have to be enriched with further mechanisms to accommodate them. Decomposing logophoric pronouns 3 Strict readings of logophors 3.1 The simplex binding account of logophors and a problematic prediction In the formal semantic literature on LOGPs, it has become standard to capture their basic distributional facts—namely, the requirement for (de se) coreference with the attitude holder—by assuming that LOGPs are simplex bound variables, bound from the left periphery of complement clauses. This is the view taken for example by Schlenker (2003), von Stechow (2004), Heim (2005), and Pearson (2015). We will call this the SIMPLEX BINDING approach to LOGPs. Let us briefly go over how this approach works, as it will later be criticized based on the availability of strict readings of LOGPs. We use the account in Pearson (2015) to illustrate the approach.7 Pearson (2015), following von Stechow (2003), assumes that LOGP is like a standard pronoun in being interpreted as a bare variable (via an assignment function as usual), but it comes with a syntactic feature LOG whose purpose is to make sure that that variable ends up being bound by a λ-operator at the edge of an embedded clause (technically by feature “checking” between [LOG] and the matching λ). To illustrate, the LF representation of Kofi says that LOGP will marry Afi is in (9a), where [LOG] enforces index matching between the variable and the binder at the edge of the CP (the boldfaced λoperator). This syntax is coupled with a semantics that assigns the embedded clause a property meaning (type ⟨e,st⟩), and an appropriate meaning for attitude predicates like say that involves quantification over centered worlds (Lewis 1979); see (9b). (9) Simplex Binding account of LOGP (based on Pearson 2015) a. Syntax: Kofi says that [︁λx1 λx1 λx1λw [︁x1[LOG]/∗x2[LOG] x1[LOG]/∗x2[LOG] x1[LOG]/∗x2[LOG] ⏞⏟⏟ ⏞ LogP will marry Afi]︁]︁ b. Semantics: ⟦say⟧w=λP λx. ∀⟨w′,x′⟩∈SAYx,w,P(x ′)(w′)=1. SAYx,w :={⟨w′,x′⟩: what xsays in wis true in w′and xidentifies themselves as x′in w′} In (9a), the fact that [LOG] requires x1to be bound by λx1makes sure that LOGP ends up referring to the attitude holder’s recognized self (the “Logophoric Center”), and this yields obligatory de se coreference with the attitude holder. The paraphrase of the resulting meaning is in (10). (10) ⟦(9a)⟧≈In all worlds in which what Kofi says is true, the person Kofi identifies as himself in those worlds marries Afi.(de se reading) The Simplex Binding approach to LOGPs makes a prediction about the possible readings of LOGP with respect to the strict-sloppy ambiguity: it predicts that 7There are differences in implementation between Pearson’s proposal and the other works cited above, but those differences are immaterial for us insofar as we are merely interested in the Simplex Binding assumption that these accounts are all committed to. I. Bassi et al. we take individual-denoting phrases (proper names, pronouns) to be interpreted as simple type-e, actual-world individuals: ⟦Mary⟧=the actual individual Mary. (25) a. ⟦spy⟧=λwx.λz. z is a spy in w.(type ⟨s,et⟩) b. ⟦think⟧=λwxλp⟨s,t⟩λy. ∀w′ x′∈BELy,p(w ′ x′)=1. (type ⟨s,⟨st,et⟩⟩) The domain of quantification for think,BEL, is defined in (26). It is a set of centered worlds where the world coordinate wis as it usually is for belief predicates, and the individual coordinate xis a “recognized self” of the attitude holder. (26) BELh:= {wx:wis compatible with h’s beliefs and xis an individual that h identifies as themselves in w} Put differently, the x-coordinate in (26) is an individual in wwho, if hwas put in x’s place, hwould not experience any difference to what they take to actually be the case. We call xhere a SELF-COUNTERPART of hin w. There are two (related) problems with the LF in (24) having to do with the occurrence of Mary in the embedded clause. First, since on our Lewis-adopted ontology individuals only occupy one possible world (no trans-world individuals), an interpretation like ⟦(24)⟧=[∀wx∈BELJohn,Maryisaspyinw]doesn’t make much sense; it entails that the predicate spy, evaluated in John’s belief worlds, applies to the actual Mary, who doesn’t inhabit those worlds, and this cannot be satisfied. We need our formalism to rather deliver that the embedded subject denotes not the actual Mary, but different counterparts of hers that live in John’s belief worlds. Second, this analysis is not equipped to account for the intuition that attitudes about an individual can be both true and false in so-called “double-vision scenarios”. These scenarios give rise to certain de re puzzles discussed since Quine (1956) and Kaplan (1968). The observation is that sometimes more than one way to specify counterparts of individuals is relevant for the interpretation of such sentences. Imagine that John has been acquainted with a certain Mary in two different ways: he sees her at the beach on a sunny day, and he also sees her trying to infiltrate into a military base at night. Crucially, John doesn’t know that the person at the beach and the infiltrator are in fact one and the same person. John suspects that the infiltrator is a spy, but has no reason to believe that the beach person is a spy. In this double-vision scenario, the sentence John thinks Mary is a spy is intuitively both true (by virtue of Mary being the night infiltrator) and false (by virtue of her being the beach person). Double-vision scenarios then point to the need for having a way to semantically encode how exactly attitude holders are acquainted with individuals. We amend both of these problems, following Heim (2001) and Sauerland (2018), by enriching our representations to include COUNTERPART FUNCTIONS. Individuals indeed occupy only one possible world, but can be systematically linked to their counterparts in other worlds. The exact way they do so depends on some suitable ACQUAINTANCE FUNCTION. The Counterpart technology amounts to a device that maps the evaluation-world individual Mary in (24) into a certain counterpart of Mary that lives in w. Concretely, we postulate silent counterpart pronouns, 𝑪𝒇𝒊 𝒙, which can attach to individual-denoting phrases like Mary. An amended LF is in (27). Decomposing logophoric pronouns (27) λw∗ sJohn thinksw∗ sλwx[[ Cf1 x Cf1 x Cf1 xMary]is a spywx] Informally, here is how counterpart pronouns work. They consist of two parameters: an indexed acquaintance function fi, whose content we assume is supplied contextually (by an assignment function), and an individual “pivot” x, which is bound by an attitude holder and is therefore the individual’s self-counterpart in the relevant worlds. fiencodes a specific description, say ‘the woman seen at the beach’, by which actual-John is acquainted with actual-Mary. ‘𝑪𝒇𝒊 𝒙(Mary)’ is to be read as the fi-counterpart of Mary in the world of x. So, ‘𝑪𝒇2 𝒙(Mary)’, for example, could be a counterpart of Mary in the world of xas she is described in the mind of xas “the woman I saw at the beach”; (27) would thus be true iff John thinks that that woman is a spy. The Counterpart technology—or its parallels in other systems, notably Concept Generators (Percus and Sauerland 2003; Charlow and Sharvit 2014, a.o.)—allows us to provide an account for double-vision scenarios. A sentence like John thinks Mary is a spy uttered in a context where John is acquainted with Mary through two different descriptions as illustrated above could be represented either as in (28a)or(28b)—and one of them could be true while the other false. The difference is located in the different acquaintance functions, which determine different individuals as counterparts of the same actual Mary. (28) a. λw∗ sJohn thinksw∗ sλwx[[ Cf1 x Cf1 x Cf1 xMary]is a spywx] e.g., John thinks: “the person I saw trying to infiltrate the base is a spy.” b. λw∗ sJohn thinksw∗ sλwx[[ Cf2 x Cf2 x Cf2 xMary]is a spywx] e.g., John thinks: “The person I saw at the beach is a spy.” The next section provides more content and definitions necessary for how counterpart operators are formally interpreted (but readers not interested in the details can skip to Sect. 4.3). Before that, we give in (29) the LF for a sentence with a free pronoun instead of a proper name in the embedded clause. The interpretation procedure is just like it is for (27), except that the value of the embedded subject sheiis supplied, as usual, by an assignment function. A Counterpart function applies to it to turn it into the suitable individuals that reside in the belief worlds of John. (29) a. John thinks that she is a spy. b. λw∗ sJohn thinksw∗ sλwx[[ Cf1 x Cf1 x Cf1 xshej]isaspy wx] 4.2.2 Counterparts: formal details Let us first define acquaintance functions. These, adopting the spirit of Sauerland’s (2018) suggestion, are functions that relate an individual to another by some suitable description from a first-person perspective. I. Bassi et al. (30) Acquaintance Function. A function fδof type ⟨e,e⟩mapping individuals to their world-mates is an Acquaintance Function iff there is a definite description of individuals, δ, containing a first person pronoun such that for any x∈dom(fδ),fδ(x) =the individual in the world of xwho xknows as ‘δ’. Take for instance the first-person-pronoun-containing description δ=‘my math teacher’. Many individuals are acquainted with many other (world-mate) individuals through this description. When John utters it, it refers to John’s math teacher. The acquaintance function that corresponds to it, defined in (31), maps any individual x for which the description is defined (i.e., any xwho knows someone as their math teacher), to the individual who fits this description for x. (31) fmy math teacher =λx. the person in x’s world who xknows as x’s math teacher The special case of BELIEF-BASED acquaintance functions is defined in (32): they guarantee a value for any self-counterpart of the individuals in their domain. (32) Belief-based Acquaintance Function. An Acquaintance Function fis beliefbased iff ∀h∈dom(f)[∀wx∈BELh[x∈dom(f)]] Belief-based acquaintance functions are those that are needed for attitude contexts. Informally, an acquaintance function fis belief-based if it preserves the underlying acquaintance relationship through which perspective-holders are acquainted with acquaintants, for all of their self-counterparts. Where xis a self-counterpart of some attitude holder h,f(x)is a world-mate of xand is the individual that xis acquainted with via the same description that specifies how his acquainted with f(h). Let us suppose further that self-counterparts of individuals have a unique actual SOURCE in the following sense: every possible individual is the self-counterpart of at most one actual individual. Thus, where hand jare distinct actual individuals, there is no world-individual pair <w,x>that is in both BELhand BELj(see definition of BEL in (26)).12 With this, we can explicate the semantics of Counterpart operators in (33). (33) Counterpart.⟦Cfi x⟧g(y) (‘the fi-Counterpart of yfor x’) is defined iff g(f i) is a belief-based acquaintance function and (︁g(f i))︁(h)=y, where his the Source of x(i.e., his the unique individual that satisfies wx∈BELh,forsome w). If defined, ⟦Cfi x⟧(y) =(︁g(f i))︁(x). ⟦Cfi x⟧g(y) is the individual in x’s world who xis mapped to (‘acquainted with’) via the same acquaintance function that maps x’s Source (the attitude holder) to y. Note that for ⟦Cfi x⟧g(y) to be defined, yneeds to live in the world of the Source of 12This assumption is defensible insofar as we hold that no two individuals have identical experiences and completely compatible beliefs about the world. It thus ignores metaphysical edge cases such as the two-gods scenario discussed in Lewis (1979, p. 520). Decomposing logophoric pronouns x;intheLFsin(28), for instance, yis the actual Mary and she lives in the world of the actual attitude holder John, who is the Source of his self-counterpart x.The interpretation of (34)isin(35). (34) λw∗ sJohn thinksw∗ sλwx[[ Cf1 x Cf1 x Cf1 xMary]is a spywx] (repeated from (27)) (35) ⟦34⟧=∀wx,wis compatible with John’s beliefs and xis a self-counterpart of John in w:thefi-counterpart of Mary for xis a spy in w. Counterparts are f-dependent (description-dependent), so one cannot generally talk about “the counterpart” of yin w, as there could be more than one. The same individual ycan have different counterparts in one and the same doxastically accessible world, so Cf1 x(y) ≡ Cf2 x(y) when f1and f2are two different belief-based acquaintance functions. This solves the de re puzzles described at the end of the previous section and allows us to capture the fact that the truth of attitude ascriptions is sometimes sensitive to the way individuals become known to attitude holders. 4.3 The meaning of LOGP and a derivation of basic sentences The previous section provided us with tools that are rich enough to properly handle reference in attitude contexts. Now we can go back to our innovation regarding LogPs. Consider again the example in (36). (36) EweÉlì Eli súsú think bé that yè LOGPãùdzí. win ‘Eli1thinks that hede-se 1won.’ As previewed, we propose that the logophoric pronoun (in Ewe, Yoruba and Igbo) underlyingly consists of two semantically active syntactic pieces: roughly, ignoring counterpart functions for a moment, LOGP≡[LOGx[proi]] LOGP≡[LOGx[proi]] LOGP≡[LOGx[proi]]. proiis a simple individual variable with no binding requirement: it may be free or bound, and if free, its value is supplied by an assignment function. LOGxis an indexed presuppositional feature, whose index must be bound by a centered-world abstractor at the edge of a clause, and is thus equated with (a counterpart of) the attitude holder.13 (37) A Logical Form (first pass): λw∗ sEli thinksw∗ sthat λwx[︁[LOGxproi] ⏞⏟⏟ ⏞ LOGP wonwx]︁ The meaning of LOGxis given in (38). LOGxfunctions like other pronominal features in being a pure-presupposition trigger (a partial identity function), acting as 13We are deliberately saying that the index on LOGxmust be bound at the edge of a clause, not necessarily at the edge of a complement clause as is usually assumed in accounts of logophoric pronouns (von Stechow 2004; Pearson 2015, a.o.). The reason for this more relaxed requirement will be discussed in Sect. 5.2, where we show that it can help explain a further distributional property of LOGPs. I. Bassi et al. a filter on the possible values of its argument. Specifically, it imposes the condition that its argument’s value is identified with the index on LOG.14 (38) ⟦LOGx⟧=[λz :z=x. z]15 Remember that given our ontology, if proi’s value is an actual-world individual— a possibility that we need to assume for the purpose of deriving strict readings—then its occurrence in an attitude context in (37) means that it must be “turned into” a counterpart of that individual that lives in the embedded world. Officially, then, we replace (37) with (39), which, just like we had in (29), appends a counterpart function to proi(but differently from (29), also has LOGx). Consequently, the interpretation of (29)isgivenin(40). (39) A Logical Form (amended): λw∗ sEli thinksw∗ sthat λwx[︁[LOGxCf1 x Cf1 x Cf1 x(proi)]won wx]︁ (40) ⟦LOGP⟧g=⟦LOGx[Cf x Cf x Cf xproi]⟧g(for any belief-based f) =:⟦Cf x(proi)⟧g=x ⏞⏟⏟ ⏞ LOGx’s presupposition .⟦Cf x(proi)⟧g LOGxthus forces the relevant counterpart of proito be identical to the selfcounterpart of the attitude holder. This is as desired, as it captures the de se reading of 14A reviewer asks in what sense the LOG feature can be counted as a pronominal feature. Semantically, as we lay out in the next section, our denotation for LOG adds a presuppositional restriction on the value of a variable—in that sense it matches what has been proposed about the semantics of ϕ-features. And syntactically, there is evidence that logophors trigger dedicated logophoric agreement on the verb in languages like Ibibio, shown for the logophor imo ˙in (ia). Hence, the LOG feature behaves morpho-syntactically like other ϕ-features in this respect. (i) Logophoric agreement in Ibibio (Newkirk 2019) a. Ekpei Ekpe a-bo 3SG-say ke C imo ˙i LOG ì-ma LOG-PST í-to LOG-hit Udo. Udo ‘Ekpe says that he hit Udo.’ b. Ekpei Ekpe a-bo 3SG-say ke C anyei/k 3SG a-diyo ˙¯ no ˙ 3SG-know ikwo sing ikwo song mfo ˙nmfo ˙n. well ‘Ekpe says that he sings well.’ While this property cannot be investigated in the languages we focus on in this paper since Ewe, Yoruba and Igbo do not display subject verb agreement, we note that in the languages with logophors that do display subject verb agreement, dedicated inflectional morphology is attested, as one would expect from a pronominal feature. 15We employ the notation of Heim and Kratzer (1998) for encoding partial functions, where the part between a colon and a dot defines the domain of the function (and where partiality is meant to model the presuppositional dimension). Decomposing logophoric pronouns LOGP: it ensures that the value of the whole LOGPistheself-counterpart of the attitude holder (no matter how the counterpart of proiis fixed, i.e., no matter the specific value of f). Here’s a complete semantic derivation of (36), for an arbitrary f:16 (41) (42) ⟦41⟧= Presupposition (simplified): In all wxwhere wis compatible with Eli’s beliefs and xis a self-counterpart of Eli in w:xis the f-counterpart of ⟦proi⟧gfor x. Assertion: In all such wx:thef-counterpart of ⟦proi⟧gfor xwins in w. What can ⟦proi⟧gbe? Apart from contextual recoverability, free pronouns are not normally referentially constrained. But here, proi’s value cannot be just anyone contextually salient; if it is an actual-world individual, it must be the (actual) attitude holder—Eli. This is because only in that case will the presupposition in (42) generated by LOGxbe satisfied. To wit, this presupposition equates x,aself-counterpart of Eli, with a counterpart of ⟦proi⟧g. By definition of Counterpart in (33), possible individuals have only one Source, which means that the Source of xmust be the Source of ⟦proi⟧gas well. From these two, and from the fact that Eli is the Source of x,it follows that ⟦proi⟧g=Eli. Therefore we can safely replace (42) with (43): 16Given that we now represent presuppositions, we replace the entry for think wehadin(25b) with the one in (i) that encodes how presuppositions are projected, adopted after Karttunen (1974)andHeim(1992). This entry is used in the calculation in (41). (i) ⟦thinkws∗⟧g=λp⟨s,t⟩λy :∀wx∈BELy,p(w x)is defined. ∀wx∈BELy,p(w x)=1 Similar lexical entries need to be written for predicates of speech, desire, hopes etc. All that matters for our purposes is that for each such predicate the domain of quantification is a set of centered worlds, and that the presupposition of LOG projects to BEL. I. Bassi et al. (43) ⟦41⟧= Presupposition: In all wxsuch that wis compatible with Eli’s beliefs and xis aself-counterpart of Eli in w:xis the f-counterpart of Eli for x. Assertion: In all such wx:thef-counterpart of Eli for xwins in w. Now what can the value of the acquaintance function-variable fbe? It too, in principle, merely needs to be recoverable from context (by some salient-enough description); but here too, fis semantically restricted: only those values are possible that result in a presuppositional statement that can be safely accommodated. In our case, because ⟦proi⟧gmust be Eli, fmust in turn be an acquaintance function that maps Eli to himself both in the actual world and in his alternative worlds. The option in (44a), which is just the identity function (underlined by the se description, “me”), will always be possible here; also the one in (44b) is ok in natural contexts in which Eli knows himself as the person called ‘Eli’. Some possibilities are out in most natural contexts since they would incur a presupposition failure, e.g., (44c)(ifEli= Ann); and others are heavily context-dependent, for instance (44d), which would satisfy the presupposition only in contexts where Eli wears a red costume and identifies himself as such.17 This nicety in the options for fwill become relevant soon. (44) Options for the value of fin (42) a. ✓fI/me =[λx. x] b. ✓fThe person I know as ‘Eli’ =[λx. the person who xknows as ‘Eli’] c. ✗fThe person I know as ‘Ann’ =[λx. the person who xknows as ‘Ann’] d. ? fThe person I know as wearing a red costume =[λx. the person who xknows as wearing a red costume] 4.4 Strict readings by ignoring LOG from alternatives We have just shown how to derive obligatory de se readings by splitting up LOGPs and by assuming that the LOG feature introduces a presupposition (without any formal binding dependency between its proipart and the matrix subject). By this, we merely replicated a basic result already obtained by Simplex Binding, just in a more complicated way. But the complication allows us to derive strict readings, to which we now turn. Consider again an example that brings strict-sloppy ambiguity to light. We repeat example (11) in an abbreviated form in (45), with which we will exemplify our analysis. 17Acquaintance functions that don’t map Eli to himself in the actual world (i.e., acquaintance functions that are not reliable in the sense of Pearson 2015) would not result in a satisfied presupposition in (41), even if Eli mistakenly associates himself with that description. Imagine for instance that Eli believes he’s the person wearing the red costume, but is mistaken about it (he actually wears a different costume). In that context, (44d) cannot be the value for fgiven LF (41). Decomposing logophoric pronouns (45) EweÉlì Eli kò only yé FOC súsú think bé that yè LOGPãùdzí. win ‘Only Eli thinks that he won.’ a. ⇝sloppy No onejbut Eli thinks theyjwon. b. ⇝strict No one but Eliithinks hei(=Eli) won. We continue to analyze these constructions as involving computation of focus alternatives triggered by a focus feature on the subject, as represented in (46a). For concreteness we take alternatives to be syntactic objects, LFs (Fox and Katzir 2011). only says that the prejacent (its sister) is true and all the alternatives are false.18 Now if LOG contributes its meaning across the focus alternatives, as represented in (46b), we are still short of deriving the intended strict reading. (46) a. LF:Only [︁Eli[FOC FOC FOC]thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx]]︁ b. Alt’s:{︁Kofi thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx], Koku thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx],...}︁ In particular, (46) fails at the level of focus alternatives. To appreciate why, notice first that since proirefers to Eli, which as we have shown above is forced on us by the presence of LOG at the prejacent, then proi’s reference to Eli will remain constant across the alternatives. But then, by the same reasoning, LOG’s presence in the alternatives results in the false information that the relevant alternatives to Eli (Kofi, Koku) are in actuality equated with Eli, which cannot be.19 However, we assume as discussed in Sect. 4.1 that LOG—being a pronominal semantic feature—is subject to the hypothesis in (23) and thus can be ignored when computing focus alternatives, like fake ϕ-features on bound pronouns. We implement the idea by letting LOG be deleted from the tier of alternatives (though not from the prejacent). The relevant derivation is in (47). (47) Analysis of (45) with LOG deleted from alternatives20 a. LF:Only [︁Eli[FOC FOC FOC]thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx]]︁ b. Alt’s:{︁Kofi thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx], Koku thinks λwx[[LOGP[LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]] [LOGx[Cf x Cf x Cf xproi]]]won wx],...}︁ 18We evidently assume in (46a)thatonly scopes at LF over the whole rest of the clause. This is done for simplicity’s sake. We could instead adopt the bipartite structure [[only DPFoc] VP] (Wagner 2006, a.o.), where only forms a constituent with its focused associate (the subject). Our main proposal isn’t affected by this choice, as long as the subject DP formally triggers focus alternatives. 19(46b) also results in the information that the alternatives Kofi, Koku identify themselves in their mind with Eli (this part by itself is not contradictory). 20The deletion operation is adopted to simplify the presentation. Instead of syntactically deleting LOG, we could neutralize its contribution by postulating (as in Sauerland 2013) that the meaning of a semantic feature can be reset in the alternatives to the total-identity function [λf.f ]. Both implementations capture the idea the LOG’s contribution is ignored in alternatives. I. Bassi et al. As above, the referent of proiremains Eli across the alternatives. But since LOG is active only in the prejacent now, the offending presupposition is absent in the alternatives. The interpretation of this configuration is given in (48), and only negates the alternatives. (48) The Interpretation of the prejacent and alternatives in (47) a. Prejacent: :∀wx∈BELEli,⟦Cf xEli⟧g=x ⏞⏟⏟ ⏞ presupposition .∀wx∈BELEli,⟦Cf xEli⟧gwinwx b. Alternatives:{︁∀wx∈BELKoku,⟦Cf xEli⟧gwinwx, ∀wx∈BELKofi,⟦Cf xEli⟧gwinwx,...}︁ The specific interpretation of this depends, of course, on the specific acquaintance function chosen for f. All that is required is for fto represent an acquaintance function that links Eli and all of his alternatives (Koku, Kofi, ...) to Eli, and that furthermore they all mentally associate with Eli. If, for example, it is common ground that everyone knows Eli by the name ‘Eli’, then plugging in for fthe value in (44b), will result in a suitable strict reading. This proposal makes the correct prediction that the strict reading can be achieved via many different ways of specifying f. Koku and Kofi do not have to be acquainted with the referent of LOGP through the description ‘the person called Eli’; they only need to know him through some shared description, if one can be accommodated. That the prediction is borne out is exemplified in (49). In this scenario, Koku and Kofi are acquainted with a certain man in a red costume, but do not know it is Eli. We callitaS TRICT-UNKNOWN scenario. (49) Context for strict unknown identity (Ewe): (fieldwork) There is a costume contest. Eli, a participant who was wearing a red costume, overhears the judges of the contest debating, and concludes from what he hears that he is going to be declared the winner. Koku and Kofi, who watched the costume show, are wrong about the identity of the man with the red costume; they don’t know it was Eli. (They might even disagree among themselves who it was.) But they don’t think that he, whoever he is, will win. Élì Eli kò only yé FOC súsú think bé that yè LOGPãùdzí win lè in àwù-dódó dress-wear êé POSS hòVíVlí contest mè. inside ‘Only Eli thinks that he won the costume contest.’ Our Ewe consultants judged the sentence felicitous and true in the scenario. The sentence entails that Koku and Kofi don’t think that the man in the red costume won— despite their lack of awareness that it is Eli. The facts hold in Igbo and Yoruba as well. In Strict-Unknown scenarios, LOGPyá in Igbo and LOGPòun in Yoruba are felicitous; see (50) and (51). Decomposing logophoric pronouns (50) Context for strict unknown identity (Igbo): (fieldwork) There is a costume contest. Eze, a participant who was wearing a red costume, overhears the judges of the contest debating, and concludes from what he hears that he is going to be declared the winner. Aki and Ada, the other contestants, are wrong about the identity of the man with the red costume; they don’t know it was Eze. (They might even disagree among themselves who it was.) But they don’t think that he, whoever he is, will win. Nᯠaní . only Ézè Eze chèrè think nà that yá LOGP mérìrì win nà PREP ásòmpì contest igósì show ákwá. clothes ‘Only Eze thinks that he won the costume contest.’ (51) Context for strict unknown identity (Yoruba): (fieldwork) There is a costume contest. Ade, a participant who was wearing a red costume, overhears the judges of the contest debating, and concludes from what he hears that he is going to be declared the winner. Niyi and O .la, who are also contestants in the costume show, are wrong about the identity of the man with the red costume; they don’t know it was Ade. (They might even disagree among themselves who it was.) But they don’t think that he, whoever he is, will win. Adé Adé nìkan only ni FOC ó RP rò think wípé that òun LOGP máa FUT tayo . win nínú inside ìdíje contest aso . clothes náà. DET ‘Only Ade thinks that he will win the costume contest.’ The reason LOGPs are predicted to be licensed in strict-unknown cases is that the value for fin our LF configuration in (47) can be resolved to the description in (44d). This description is salient in the context, and it refers to Eli in Eli’s mind but not to who Koku and Kofi associate with the description ‘Eli’.21 As one of our consultants emphasized, if Eli does not know that he is the man in the red costume, the target sentence in (49) is not felicitous in such strict-unknown scenarios. Thus, the attitude holder in the prejacent (Eli) still has to be familiar with himself as the man in the red costume. This is expected for the same reason that de re readings of LOGPs are out in basic sentences (cf. Sect. 2.2): LOG’s presupposition, which in the prejacent cannot be ignored, imposes the de se condition on the use of LOGP. This concludes our basic proposal of the strict reading of LOGPs. 21What if a context furnishes no suitable definite description that could be the value of fin this derivation? Our predictions might change. Imagine we cook up a context where it is impossible to find any (salient) description that links Eli and all his relevant alternatives to Eli and is vivid enough in their minds (i.e., across the doxastically accessible worlds) as such. In such a context, we predict that sentence (45) would not support the strict reading (we thank Amir Anvari, p.c., for raising a similar point to us). We think, however, that finding convincing cases of such contexts is not trivial, as there is arguably always some shared description that can be accommodated whose reference in the minds of the relevant individuals would be the matrix subject. In the absence of independently justifiable constraints on what concepts can be accommodated in a context, it is not entirely clear to us how to create the relevant scenarios. We leave this as an open question for now. I. Bassi et al. (68) a. Ewe28 *mè I súsú think bé that yè LOGP dzó leave kábá. early ‘I think that I left early.’ b. mè I súsú think bé that mè I dzó leave kábá. early ‘I think that I left early.’ 5.2.1 What LOGPs denote in matrix environments Recall from Sect. 4.3 the denotation we assign to a LOGP, repeated in (69).29 (69) a. ⟦LOGx⟧=[λz :z=x. z] b. ⟦LOGP⟧g=⟦LOGxproi⟧g=⟦proi⟧g, defined only if ⟦proi⟧g=x We assumed there (see fn. 13) that LOGPs—more accurately, the index on LOG—must be bound by an abstractor over centered worlds from the edge of a clause.30 If so, nothing prohibits a matrix occurrence of a LOGP where it is bound by the top-most centered-world abstractor introduced at the matrix level. The result would be: (70) a. ‘LOGPleft’ ≡λws[︁[LOGPLOGsproi]left w]︁ b. ⟦(70a)⟧=[λws:⟦proi⟧g=s. ⟦proi⟧gleft in w] The world coordinate of a matrix centered world corresponds, as standard, to a world compatible with the actual context of utterance, and it is natural to take the individual coordinate to be the speaker of the context. This is meant to capture the intuition that the “logophoric center” of a matrix sentence is the speaker of the sentence. We technically implement this with the discourse rule in (71) relating the proposition expressed by a sentence to contexts in which the sentence can be appropriately uttered by speakers (the notions of Common Ground and Context Set in (71a) are adapted versions of their familiar predecessors from the works of Stalnaker 1978, et seq.) 28Pearson (2015, p. 97, fn. 26) and Bimpeh (2023, p. 32, fn. 7) both mention dialects of Ewe that allow LOGP to be anteceded by first person pronouns, although the robustness of that data is not clear due to insufficient accessability to speakers of those dialects. We weren’t able to verify that claim independently, but if correct, our account does not apply to such dialects. 29Counterpart operators are not needed in matrix environments, so they are left out here. 30If we didn’t place any binding restrictions on LOG, we would wildly overgenerate readings, because then it wouldn’t be clear what rules out a LOGP with a free-indexed LOG from referring to some contextually salient individual. Decomposing logophoric pronouns (71) a. A context is appropriate for a sentence ϕif it determines a Context Set, a set of worlds csuch that c={w∗:w∗is compatible with the shared knowledge among the participants in the context}. b. A sentence ϕuttered in context cby a speaker in c(notated ‘sp(c)’) is true in a world w∗∈c,iff: ⟦ϕ⟧g,c(w∗ sp(c))=1. Given all this, (70a) is predicted to be grammatical and appropriately used to express the proposition that the speaker left, wrongly so (cf. (66)). 5.2.2 Competition via Maximize Presupposition! In principle, we could avoid the problem with a syntactic stipulation to the effect that LOGPs are licensed only when bound by attitude predicates, like previous approches assumed (see Sect. 3.1). But, as previewed, we want to offer an alternative that doesn’t require this restriction, and to derive the infelicity of (66) as the result of competition with (67). To do so, we need a concrete analysis of first person pronouns. We take the first person pronoun to contain a variable proiand a semantic feature 1ST whose job is to restrict the variable’s value to be the speaker, as represented in (72). This much is fairly standard (Heim 2008, Charnavel 2019, a.o.), but we also assume that 1ST, like LOG, comes with an index that must be bound by a logophoric abstractor. The index’s value in this case is identified with proiand, unlike LOG, also with the actual speaker. The lexical entry of 1ST, a purely presuppositional function, is thus in (73a).31 (‘c’ stands for the Context Set as was defined in (71a).) (72) LF of first person pronoun:[ 1STP1STsproi] (73) a. ⟦1STs⟧g,c =[λz :z=s∧z=sp(c).z] b. ⟦1STP⟧g,c =⟦FIRSTs⟧g,c(⟦proi⟧g,c)= ⟦proi⟧g,c, defined only if ⟦proi⟧g,c =s=sp(c). Below is the LF and meaning of (67)‘Ileft’. (74) ‘I left.’ a. λws[︁[1STP1STsproi]left w]︁ b. ⟦(74a)⟧g,c =[λws:⟦proi⟧g=s=sp(c).⟦proi⟧gleft in w] The resulting partial proposition in (74b) has a stronger semantic presupposition— a strictly smaller domain—than the LOGP version of that sentence in (70b). While 31This analysis of course is not applicable to languages with Indexical Shifting, as the first person pronoun can also pick out attitude holders in those languages (see Deal 2020). In fact, if our analysis is to be extended to indexical shifting languages, then the first person pronoun in those languages can be analyzed the way we analyze LOGPs in West African (and moreover, indexical shifting languages would have to be assumed to lack a pronoun with the syntax and semantics in (72)-(73)). I. Bassi et al. (70b) is in principle defined for any individual s(=proi), (74b) is defined only for the actual speaker(=proi). But when defined, both (70b) and (74b) have the exact same assertive content and therefore convey the same information: the speaker left. A situation where two alternative propositions have the same assertive content but one has a stronger presupposition is generally taken to feed the competition principle Maximize Presupposition! (Heim 1991; Percus 2006; Sauerland 2008; Schlenker 2012), with the consequence that the presuppositionally weaker alternative is blocked. (75) Maximize Presupposition!. Don’t use a sentence ϕin context S(cf. (71a)) if there is an alternative sentence ψof ϕsuch that the following are all met: (i) ψhas stronger presuppositions than ϕ(⟦ψ⟧’s domain is smaller than ⟦ϕ⟧’s); (ii) ψ’s presuppositions are met in S(i.e., w∗ s∗∈dom(⟦ψ⟧S)for all w∗ s∗∈S); (iii) ψis contextually equivalent to ϕ(ϕ≡Sψ). We assume that 1STP is formally an alternative to LOGP (as it is syntactically no more complex than it; Katzir 2007) and therefore Maximize Presupposition! dictates that a sentence with LOGP cannot be used to refer to the speaker.32 This explains the restriction in matrix environments. 5.2.3 1ST-LOG competition in embedded positions First person pronouns are, of course, licensed not only in matrix positions but also in embedded ones. For example, John thinks that I left, a suitable analysis within our formalism is given in (76a), with its assigned interpretation in (76b). The presupposition of 1ST, once again, guarantees that the reference is to the (relevant counterpart of) the actual speaker. Note that the semantics of 1SC forces the index on 1SC to be bound by the matrix abstractor over centered worlds, and that a counterpart operator must consequently be added. (76) ‘John thinks 1STPleft.’ a. λwsJohn thinks λw′ x[︁[1STPCf x(1STsproi)] leftw′]︁ b. λws:⟦proi⟧g,c =s=sp(c).∀w′ x∈BELJohn[Cf x(⟦proi⟧g,c)left in w′] We can now proceed to suggest an explanation for the restriction on first-person antecedents, repeated from (68) (the facts hold in Yoruba and Igbo as well): 32Maximize Presupposition! has been invoked to explain certain inferences arising from the use of the presuppositionally weaker alternative (“anti-presuppositions”; Percus 2006, a.o.). In our case, however, no anti-presupposition can be detected from a use of unembedded LOGP because the relevant presupposition of the stronger alternative 1STP is always met; more precisely, a hypothetical context in which that presupposition isn’t met would be one where the sentence as parsed in (74a) could not be uttered by the speaker in the context to begin with, given the definition of appropriate context in (71). Decomposing logophoric pronouns (77) LOGPs cannot be anteceded by first-person attitude holders: a. Ewe*mè I súsú think bé that yè LOGP dzó. leave ‘I think that I left.’ b. mè I súsú think bé that mè I dzó. leave ‘I think that I left.’ Why is (77a) not acceptable? The answer, we propose, is that (77b) blocks (77a) by Maximize Presupposition!, just like (74b) blocked (70b) in matrix environments. Here is the LF-analysis of the two sentences: (78) ‘I think LOGPleft.’ a. λwsI think λw′ x[︁[LOGPCf x(LOGsproi)] leftw′]︁ b. λws:⟦proi⟧g,c =s. ∀w′ x∈BELsp(c)[Cf x(⟦proi⟧g,c)left in w′] (79) ‘I think 1STPleft.’ a. λwsI think λw′ x[︁[1STPCf x(1STsproi)] leftw′]︁ b. λws:⟦proi⟧g,c =s=sp(c).∀w′ x∈BELsp(c)[Cf x(⟦proi⟧g,c)left in w′] The underlined parts reveal again that (79) has a strictly stronger presupposition than (78). But they have the exact same assertive content. Therefore the latter is blocked by Maximize Presupposition!.33 While this competition-based account explains (77) as a direct extension of the account of matrix environments, it is difficult to see how previous accounts of the dependency between a LOGP and its antecedent can capture it without further stipulations. If LOGPs are merely syntactically required to be bound by an attitude verb (as is assumed by Simplex Binding, Sect. 3.1), then additional mechanisms are needed to ensure that a first-person subject is not a possible antecedent.34 33In (78a) the index on LOG is bound by the matrix abstractor (λws). Things change if we choose the index xbound by the embedded abstractor (λw′ x). That parse, in (i), generates the presupposition underlined in (ib). (i) a. λwsI think λw′ x[︁[LOGPLOGx[Cf x Cf x Cf xproi]] leftw′]︁ b. λws:∀wx∈BELsp(c),⟦Cf xproi⟧g=x. ∀w′ x∈BELsp(c)[Cf x(⟦proi⟧g,c)left in w′] This amounts to the presupposition that across the speaker’s belief worlds, her self-counterpart is also the counterpart of pro. We are not sure at the moment what blocks this representation. 34The logic of the account of (77) leads us to expect that LOGPs are fine with second-person antecedents. While this is correct for Ewe (Clements 1975;Bimpeh2023) and other languages (see Hyman and Com- I. Bassi et al. 6Conclusion This paper provided evidence that logophoric pronouns (LOGPs) in Ewe, Yoruba and Igbo support both strict and sloppy readings in ellipsis configurations and in sentences with only (following observations in Culy 1994; Bimpeh and Sode 2021), and offered a formal analysis that could capture this behavior. The account supplants existing accounts of LOGPs with the idea that LOGPs are pronouns that contain a semantic feature LOG in charge of encoding the de se reference to the attitude holder (following Bimpeh et al. 2024), but whose contribution can be ignored at the level of focus alternatives, like other pronominal features. We also showed how long-distance dependencies of LOGPs are handled, and provided a novel solution for the restriction on the occurrence of LOGP in matrix environments and with first-person antecedents. Supplementary information The online version contains supplementary material available at https://doi. org/10.1007/s11050-025-09242-x. Acknowledgements We are very grateful to Noble Ahiaklo-Kuz, Mary Amaechi, Daniel Aremu, Veronica Ebere Ugwu, and Chinedu Fedrick for sharing their languages with us. We also thank the audiences at GLOW 45, WCCFL 40 and SuB 27, as well as Frank Sode, for their input. Fundinginformation Open Access funding enabled and organized by Projekt DEAL. This project received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (LeibnizDream, grant agreement No 856421). Declarations Competing interests The authors have no competing interests to declare. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/ 4.0/. References Adésolá, O. 2005. Pronouns and null operators – A-bar dependencies and relations in Yorùbá. PhD diss. Rutgers University. Amaechi, M. C. 2020. A-bar-movement dependencies and their reflexes in Igbo. PhD diss. Universität Potsdam. Anand, P. 2006. De de se. PhD diss. MIT. Bassi, I. 2021. Fake feature and valuation from context. PhD diss. MIT. Bassi, I., I. Driemel, A. A. Bimpeh, and S. Silleresi. 2023. Strict logophors in Ewe, Yoruba, and Igbo. In Proceedings of Sinn und Bedeutung 27, eds. M. Onoeva, A. Staˇ nková, and R. Šimík, 52–63. Universität Tübingen. rie 1981; Stirling 1993; Nikitina 2012), it appears to not be true for all logophoric languages (see, e.g., Kiemtoré 2022). We leave this issue unresolved for now and hope to revisit it in a future occasion. Decomposing logophoric pronouns Bimpeh, A. A. 2023. Logophoricity in Ewe: An empirical-semantic assessment of yè. PhD diss. Goethe University. Bimpeh, A. A., and F. Sode. 2021. Evidence against de se binding: Strict readings of the logophoric pronoun in Ewe. In Proceedings of TripleA, eds. N. Banerjee and V. Hehl. Vol. 6. 1–16. Cambridge: MIT Working Papers in Linguistics. Bimpeh, A. A. 2019. Default de se: The interpretation of the Ewe logophor. In Proceedings of TripleA 5, eds. M. R. Bochnak, et al.. Tübingen: Universitätsbibliothek Tübingen, 1–15. Bimpeh, A. A., I. Driemel, I. Bassi, and S. Silleresi. 2024. Obligatory de se logophors in Ewe, Yoruba and Igbo: Variation and competition. In Proceedings of the 40th WCCFL, eds. J. Lu, et al., 1–10. Somerville: Cascadilla Press. Bruening, B. 2021. Generalizing the presuppositional approach to the binding conditions. Syntax 24(4):417–461. Charlow, S., and Y. Sharvit. 2014. Bound ‘de re’ pronouns and the LFs of attitude reports. Semantics and Pragmatics 7(3):1–43. Charnavel, I. 2019. Presupposition failure and intended pronominal reference: Person is not so different from gender after all. Linguistic Inquiry 50(2):372–387. Charnavel, I. 2020. Locality and logophoricity: A theory of exempt anaphora. Oxford: Oxford University Press. Chierchia, G. 1989. Anaphora and attitudes de se. In Semantics and contextual expressions,eds.R.Bartsch, J. van Benthem, and P. van Emde Boas, 1–31. Foris: Dordrecht. Clements, G. 1975. The logophoric pronoun in Ewe: Its role in discourse. Journal of West African Languages 10:141–177. Culy, C. 1994. Aspects of logophoric marking. Linguistics 32:1055–1094. Deal, A. R. 2020. A theory of indexical shift: Meaning, grammar, and crosslinguistic variation. Cambridge: MIT Press. Essegbey, J. 2008. The potential morpheme in Ewe. In Aspect and modality in Kwa languages.eds.F.K. Ameka, and M. E. Kropp Dakubu 195–214. Amsterdam: John Benjamins Publishing. Fiengo, R., and R. May. 1994. Indices and identity. Cambridge: MIT press. Fox, D., and R. Katzir. 2011. On the characterization of alternatives. Natural Language Semantics 19(1):87–107. Grano, T., and S. Lotven. 2019. Control of logophoric pronouns in Gengbe. In African linguistics across the disciplines: Selected papers from the 48th annual conference on African linguistics,ed.S.Lotven, 325–337. Berlin: Language Science Press. Güldemann, T. 2003. Logophoricity in Africa: An attempt to explain and evaluate the significance of its modern distribution. STUF – Language Typology and Universals 56(4):366–387. Hagège, C. 1974. Les pronoms logophoriques. Bulletin de la Société de Linguistique de Paris 69(1):287–310. Heim, I. 1991. Artikel und definitheit. In Semantik: ein internationales Handbuch der zeitgenössischen Forschung, Vol. 6, 487–535. Heim, I. 1992. Presupposition projection and the semantics of attitude verbs. Journal of Semantics 9:183–221. Heim, I. 2001. Semantics and morphology of person and logophoricity. Handout from a talk at the University of Tübingen, December 2001. Heim, I. 2005. Pronouns and their features in attitude reports. Manuscript, MIT. Heim, I. 2008. Features on bound pronouns. In Phi theory: Phi-features across modules and interfaces, eds. D. Harbour, D. Adger, and S. Bejar, 35–56. Heim, I., and A. Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. Hestvik, A. 1995. Reflexives and ellipsis. Natural Language Semantics 3(2):211–237. Hyman, L. M., and B. Comrie. 1981. Logophoric reference in Gokana. Journal of African Languages and Linguistics 3(1):19–37. Jacobson, P. 2012. Direct compositionality and ‘uninterpretability’: The case of (sometimes) ‘uninterpretable’ features on pronouns. Journal of Semantics 29:305–343. Kaiser, E. 2018. Pronoun use in Finnish reported speech and free indirect discourse: Effects of logophoricity. In Pronouns in embedded contexts at the syntax-semantics interface.eds.P.Patel-Grosz,P.G. Grosz, and S. Zobel 75–104. New York: Springer. Kaplan, D. 1968. Quantifying in. Synthese 19:178–214. Kaplan, D. 1989. Demonstratives. In Themes from Kaplan, eds. J. Almog, J. Perry, and H. Wettstein, 481–563. I. Bassi et al. Karttunen, L. 1974. Presupposition and linguistic context. Theoretical Linguistics 1(1–3):181–194. Katzir, R. 2007. Structurally-defined alternatives. Linguistics and Philosophy 30(6):669–690. Keenan, E. L. 1971. Names, quantifiers, and the sloppy identity problem. Research on Language and Social Interaction 4(2):211–232. Kiemtoré, A. 2022. Issues in Jula complementation: Structures, relations and matters of interpretation. PhD diss. University of Stuttgart. Koopman, H., and D. Sportiche. 1989. Pronouns, logical variables, and logophoricity in Abe. Linguistic Inquiry 20:555–588. Kratzer, A. 2009. Making a pronoun: Fake indexicals as windows into the properties of pronouns. Linguistic Inquiry 40(2):187–237. Lawal, N. S. 2006. Yoruba pronominal anaphor Òun and the binding theory. In Selected proceedings of the 35th annual conference on African linguistics, eds. J. Mugane, J. P. Hutchison, and D. A. Worman, 245–257. Somerville: Cascadilla Proceedings Project. Lewis, D. K. 1968. Counterpart theory and quantified modal logic. The Journal of Philosophy 65(5):113–126. Lewis, D. 1979. Attitudes de dicto and de se. Philosophical Review 88(4):513–543. Manfredi, V. 1987. Antilogophoricity as domain extension in Igbo and Yoruba. Niger-Congo Syntax and Semantics 1:97–117. McKillen, A. 2016. On the interpretation of reflexive pronouns. PhD diss. McGill University. Merchant, J. 2019. Ellipsis: A survey of analytical approaches. In The Oxford handbook of ellipsis, 18–46. Newkirk, L. 2019. Logophoricity in Ibibio. In Selected proceedings of the 48th annual conference on African linguistics, eds. S. Lotven, et al., 309–323. Berlin: Language Science Press. Nikitina, T. 2012. Logophoric discourse and first person reporting in Wan (West Africa). Anthropological Linguistics 54(3):280–301. O’Neill, T. 2015. The distribution of the Danyi Ewe logophor yi. Slides. LSA Annual Meeting. Park, Y. 2018. Attitudes de se and logophoricity. PhD diss. University of Massachusetts, Amherst. Partee, B. 1973. Some transformational extensions of Montague grammar. Journal of Philosophical Logic 2(4):509–534. Patel-Grosz, P. 2020. Pronominal typology and the de se/de re distinction. Linguistics and Philosophy 43(5):537–587. Pearson, H. A. 2015. The interpretation of the logophoric pronoun in Ewe. Natural Language Semantics 23(2):77–118. Pearson, H. 2022. On testing for de se and de re construals across languages. Semantic Fieldwork Methods 4:1–22. Percus, O. 2000. Constraints on some other variables in syntax. Natural Language Semantics 8(3):173–229. Percus, O. 2006. Antipresuppositions. Theoretical and empirical studies of reference and anaphora: Toward the establishment of generative grammar as an empirical science. Report of the Grant-in-Aid for Scientific Research (B), Project No. 15320052, Japan Society for the Promotion of Science, ed. A. Ueyama, 52–73. Percus, O., and U. Sauerland. 2003. Pronoun movement in dream reports. In Proceedings of NELS 33,eds. M. Kadowaki and S. Kawahara, 347–366. Amherst: GLSA. Pulleybank, D. 1986. Clitics in Yoruba. In Syntax and semantics 9: The syntax of pronominal clitics.ed. H. Borer 43–64. New York: Academic Press. Quine, W. V. 1956. Quantifiers and propositional attitudes. The Journal of Philosophy 53(5):177–187. Reinhart, T. 1983. Coreference and bound anaphora: A restatement of the anaphora questions. Linguistics and Philosophy 6:47–88. Rooth, M. 1992a. Ellipsis redundancy and reduction redundancy. In Proceedings of the Stuttgart ellipsis workshop, Vol. 29, 1–26. Rooth, M. 1992b. A theory of focus interpretation. Natural Language Semantics 1(1):75–116. Ross, J. 1967. Constraints on variables in syntax. PhD diss. MIT. Sag, I. A. 1976. Deletion and logical form. PhD diss. MIT. Satık, D. 2020. Logophoric pronouns are not (always) inherently logophoric. In Proceedings of the fiftyfifth annual meeting of the Chicago Linguistic Society, ed. O. Eren, 381–396. Chicago: Chicago Linguistic Society. Satık, D. 2021. Control is not movement: Evidence from overt PRO in Ewe. Manuscript, lingbuzz/004685. Sauerland, U. 2008. Implicated presuppositions. In The discourse potential of underspecified structures, ed. A. Stuebe, 581–600. Decomposing logophoric pronouns Sauerland, U. 2013. Presuppositions and the alternative tier. In Proceedings of SALT 23,ed.T.Snider, 156–173. Sauerland, U. 2018. Counterparts block some ‘de re’ readings. English Linguistics: Journal of the English Linguistic Society of Japan 35(1):38–64. Schlenker, P. 1999. Propositional attitudes and indexicality: A cross-categorial approach. PhD diss. MIT. Schlenker, P. 2003. A plea for monsters. Linguistics and Philosophy 26(1):29–120. Schlenker, P. 2012. Maximize presupposition and Gricean reasoning. Natural Language Semantics 20:391–429. Sells, P. 1987. Aspects of logophoricity. Linguistic Inquiry 18(3):445–479. Stalnaker, R. C. 1978. Assertion. In Syntax and semantics, Vol. 9, 315–332. New York: Academic Press. Stirling, L. 1993. Switch-reference and discourse representation. Cambridge: Cambridge University Press. Stirling, L. 1994. Logophoricity and long-distance reflexives. In The encyclopedia of language and linguistics, ed. R. E. Asher. Vol. 9, 2302–2306. Sudo, Y., and G. Spathas. 2020. Gender and interpretation in Greek: Comments on Merchant (2014). Glossa: A Journal of General Linguistics 5:1. Sundaresan, S. 2018. Perspective is syntactic: Evidence from anaphora. Glossa: A Journal of General Linguistics 3:1. Takahashi, S., and D. Fox. 2005. MaxElide and the re-binding problem. In Proceedings of SALT 15,eds. E. Georgala and J. Howell, 223–240. Tancredi, C. 1992. Deletion, deaccenting and presupposition. PhD diss. MIT. von Fintel, K., and I. Heim. 2011. Intensional semantics. Manuscript, MIT. von Stechow, A. 2003. Feature deletion under semantic binding: Tense, person, and mood under verbal quantifiers. In Proceedings of NELS 33, eds. M. Kadowaki and S. Kawahara, 133–157. Amherst: GLSA. von Stechow, A. 2004. Binding by verbs: Tense, person and mood under attitudes. In The syntax and semantics of the left periphery, eds. S. Lohnstein and H. Trissler, 431–488. Berlin: Mouton de Gruyter. Wagner, M. 2006. Association by movement: Evidence from NPI-licensing. Natural Language Semantics 14(4):297–324. Williams, E. S. 1977. Discourse and logical form. Linguistic Inquiry 8(1):101–139. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Authors and Affiliations Itai Bassi1·Imke Driemel2·Abigail Anne Bimpeh3·Silvia Silleresi4· Johnson Fo .lo .runs .o .Ilo .ri5·Anastasia Nuworsu6·Gerald Okey Nweya7 ✉I. Bassi [email protected] I. Driemel [email protected] A.A. Bimpeh [email protected] S. Silleresi [email protected] J.F. Ilo .ri [email protected] A. Nuworsu [email protected] G.O. Nweya [email protected] I. Bassi et al. 1Leibniz-ZAS, Pariser Straße 1, Berlin, 10719, Germany 2University of York, Heslington, York, YO105DD, UK 3Humboldt-Universität zu Berlin, Unter den Linden 6, Berlin, 10099, Germany 4University of Milano-Bicocca, Piazza dell’Ateneo Nuovo 1, Milan, 20126, Italy 5University of Lagos, University Road Lagos Mainland Akoka, Lagos, Nigeria 6Ho Technical University, HFRC+M4R Poly Rd, Ho, Ghana 7University of Ibadan, Ibadan 200001, Oyo, Nigeria