In the rise and fall of the active-inactive alignment across Romance (La Fauci 1988; Ledgeway 2012), object-past participle agreement started off as an alignment-driven pattern and subsequently developed into a partially discourse-driven structure, where only a subset of objects triggered agreement, namely topical objects (Bentley 2006). The paper investigates Friulian’s agreement pattern between the past participle and the in-situ postverbal direct object and its relation to information structure. As shown in (1) below, in Friulian, past participle agreement in gender and number takes place with the postverbal direct object in the absence of apparent right or left dislocation: (1) Mario al a mangjad-e le polente Mario 3SG.M.SCL has.3SG eat.PTCP-F.SG the polenta.F.SG “Mario ate the polenta” Friulian object-past participle agreement, however, does not consistently take place. Its occurrence is conditioned by the discourse-pragmatic status of the postverbal direct object: if the direct object is in narrow focus (Lambrecht 1994) object-past participle agreement cannot take place, as shown in (2): (2) CONTEXT: What did you get? a. O ai cjapât-Ø DOS MULTIS 1SG.SCL have.1SG get.PTCP two fine.F.PL “I have got two fines” b. *O ai cjapad-is DOS MULTIS 1SG.SCL have.1SG get.PTCP-F.PL two fine.F.PL Note that the focal portion of the sentence is in capitals for ease of representation. On the other hand, if the object has topical status as background (G-topic) or stage-setting (A-topic) information (Erteschirk-Shir 2007; Bianchi & Frascarelli 2010), object-past participle agreement obligatorily takes place (cf.1). The central claim of the paper is hence that Friulian object-past participle agreement is not unpredictable or, as previously proposed, governed by a competing-grammar model (see Loporcaro 1998; Paoli 2006), but it is determined by the covert topicality of the in-situ direct object. Italian has instead lost object-past participle agreement in all contexts, except when the direct object is an established left-peripheral topic (Loporcaro 2003, 2016). In such case, the object is left dislocated and obligatorily resumed by an object clitic (Cinque 1990), as shown in (3): (3) Le due multe, le ho pres-e ieri the two fine.F.PL 3PL.F.OCL have.1SG get.PTCP-F.PL yesterday “As for the two fines, I got them yesterday” In (3), the past participle and the direct object agree in number and gender. Traditionally, Italian object-past participle agreement is analysed as the result of clitic-movement (Kayne 1989; Belletti 2001): the clitic moves as a phrase and passes through the specifier position of the functional head hosting the past participle, triggering agreement (see D’Alessandro and Roberts 2008 for an Agree analysis for the phenomenon). Such analysis is however unsuitable for the pattern of object-past participle agreement in Friulian. A clitic-movement analysis may in fact account for the agreement pattern in (4), but fails to account for (5): (4) Lis dos multis, les ai cjapad-is iar the two fine.F.PL 3PL.F.OCL have.1SG get.PTCP-F.PL yesterday “As for the two fines, I got them yesterday” (5) CONTEXT: Did you really get two fines?! Ceppo! O ai cjapad-is dos multis Of course 1SG.SCL have.1SG get.PTCP-PL.F two fine.PL.F “Yes, that’s correct. I have got two fines” Unlike Italian, Friulian can exhibit past participle agreement also with the postverbal direct object (Benincà & Vanelli 1984; Haiman & Benincà 1992; Paoli 1997; 2006). As shown in (5), the past participle and the object agree in number and gender: the direct object appears in canonical position and is not resumed by an agreeing object clitic. In light on the recent literature on the relation between information structure and verb- object agreement in Bantu and Italo-Romance (Bax & Diercks 2012; Mursell 2018; D’Alessandro 2017, 2019), I propose that the difference in object-past participle agreement pattern between Italian and Friulian (see Loporcaro 1998, 2016) is due to the presence of an extra set of phi-features (uφ) on v° in Friulian, which can only be valued by an internal argument DP that enters the derivation carrying a [Topic] feature as well as a set of interpretable phi-features (iφ), namely [Number] and [Gender]. If such goal is not available the default masculine singular ending surfaces on the past participle, as the Agree operation (Chomsky 2001) fails (i.e. with focal objects). The correct discourse-pragmatic interpretation of the topical in-situ object is then ensured post-syntactically through covert topicalisation at LF (Polinsky and Potsdam 2001). As for the nature of this extra set of phi-features (uφ) on v°, I want to put forward the hypothesis that is a direct continuation of a previous stage of the language, common to most Romance languages (La Fauci 1988), where uφ indiscriminately established an Agree relation with the internal argument DP as an alignment strategy, before specialising into a discourse-driven structure in Friulian.